Merge pull request #533 from meilisearch/bump-to-v0.9.0

Bump the workspace crates to 0.9.0

.dockerignore

@@ -0,0 +1,5 @@
+target
+Dockerfile
+.dockerignore
+.git
+.gitignore

.github/workflows/README.md

@@ -0,0 +1,18 @@
+# GitHub actions workflow for MeiliDB
+
+> **Note:**
+
+> - We do not use [cache](https://github.com/actions/cache) yet but we could use it to speed up CI
+
+## Workflow
+
+- On each pull request, we are triggering `cargo test`.
+- On each tag, we are building:
+    - the tagged docker image
+    - the binaries for MacOS, Ubuntu, and Windows
+    - the debian package
+- On each stable release, we are build the latest docker image.
+
+## Problems
+
+- We do not test on Windows because we are unable to make it work, there is a disk space problem.

.github/workflows/publish-binaries.yml

@@ -0,0 +1,39 @@
+name: Publish binaries to GitHub release
+
+on:
+  push:
+    tags:
+      - '*'
+
+jobs:
+  publish:
+    name: Publish for ${{ matrix.os }}
+    runs-on: ${{ matrix.os }}
+    strategy:
+      matrix:
+        os: [ubuntu-latest, macos-latest, windows-latest]
+        include:
+          - os: ubuntu-latest
+            artifact_name: meilisearch
+            asset_name: meilisearch-linux-amd64
+          - os: macos-latest
+            artifact_name: meilisearch
+            asset_name: meilisearch-macos-amd64
+          - os: windows-latest
+            artifact_name: meilisearch
+            asset_name: meilisearch-windows-amd64
+
+    steps:
+    - uses: hecrj/setup-rust-action@master
+      with:
+        rust-version: stable
+    - uses: actions/checkout@v1
+    - name: Build
+      run: cargo build --release --locked
+    - name: Upload binaries to release
+      uses: svenstaro/upload-release-action@v1-release
+      with:
+        repo_token: ${{ secrets.GITHUB_TOKEN }}
+        file: target/release/${{ matrix.artifact_name }}
+        asset_name: ${{ matrix.asset_name }}
+        tag: ${{ github.ref }}

.github/workflows/publish-deb-pkg.yml

@@ -0,0 +1,30 @@
+name: Publish deb pkg to GitHub release & apt repository
+
+on:
+  push:
+    tags:
+      - '*'
+
+jobs:
+  publish:
+    name: Publish debian packagge
+    runs-on: ubuntu-latest
+
+    steps:
+    - uses: hecrj/setup-rust-action@master
+      with:
+        rust-version: stable
+    - name: Install cargo-deb
+      run: cargo install cargo-deb
+    - uses: actions/checkout@v1
+    - name: Build deb package
+      run: cargo deb -p meilisearch-http -o target/debian/meilisearch.deb
+    - name: Upload debian pkg to release
+      uses: svenstaro/upload-release-action@v1-release
+      with:
+        repo_token: ${{ secrets.GITHUB_TOKEN }}
+        file: target/debian/meilisearch.deb
+        asset_name: meilisearch.deb
+        tag: ${{ github.ref }}
+    - name: Upload debian pkg to apt repository
+      run: curl -F package=@target/debian/meilisearch.deb https://${{ secrets.GEMFURY_PUSH_TOKEN }}@push.fury.io/meilisearch/

.github/workflows/publish-docker-latest.yml

@@ -0,0 +1,19 @@
+---
+on:
+  push:
+    tags:
+      - 'v[0-9]+.[0-9]+.[0-9]+'
+
+name: Publish latest image to Docker Hub
+
+jobs:
+  build:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v1
+      - name: Publish to Registry
+        uses: elgohr/Publish-Docker-Github-Action@master
+        with:
+          name: getmeili/meilisearch
+          username: ${{ secrets.DOCKER_USERNAME }}
+          password: ${{ secrets.DOCKER_PASSWORD }}

.github/workflows/publish-docker-tag.yml

@@ -0,0 +1,20 @@
+---
+on:
+  push:
+    tags:
+      - '*'
+
+name: Publish tagged image to Docker Hub
+
+jobs:
+  build:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v1
+      - name: Publish to Registry
+        uses: elgohr/Publish-Docker-Github-Action@master
+        with:
+          name: getmeili/meilisearch
+          username: ${{ secrets.DOCKER_USERNAME }}
+          password: ${{ secrets.DOCKER_PASSWORD }}
+          tag_names: true

.github/workflows/test.yml

@@ -0,0 +1,25 @@
+---
+on: [pull_request]
+
+name: Test binaries with cargo test
+
+jobs:
+  check:
+    name: Test on ${{ matrix.os }}
+    runs-on: ${{ matrix.os }}
+    strategy:
+      matrix:
+        os: [ubuntu-latest, macos-latest]
+
+    steps:
+    - uses: actions/checkout@v1
+    - uses: actions-rs/toolchain@v1
+      with:
+        profile: minimal
+        toolchain: stable
+        override: true
+    - name: Run cargo test
+      uses: actions-rs/cargo@v1
+      with:
+        command: test
+        args: --locked --release

.gitignore

@@ -1,7 +1,8 @@
 /target
-Cargo.lock
+meilisearch-core/target
 **/*.csv
 **/*.json_lines
 **/*.rs.bk
 /*.mdb
 /query-history.txt
+/data.ms

Cargo.toml

@@ -1,8 +1,10 @@
 [workspace]
 members = [
-    "meilidb-core",
-    "meilidb-schema",
-    "meilidb-tokenizer",
+    "meilisearch-core",
+    "meilisearch-http",
+    "meilisearch-schema",
+    "meilisearch-tokenizer",
+    "meilisearch-types",
 ]
 
 [profile.release]

Dockerfile

@@ -0,0 +1,27 @@
+# Compile
+FROM    alpine:3.10 AS compiler
+
+RUN     apk update --quiet
+RUN     apk add curl
+RUN     apk add build-base
+
+RUN     curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh -s -- -y
+
+WORKDIR /meilisearch
+
+COPY    . .
+
+ENV     RUSTFLAGS="-C target-feature=-crt-static"
+
+RUN     $HOME/.cargo/bin/cargo build --release
+
+# Run
+FROM    alpine:3.10
+
+RUN     apk update --quiet
+RUN     apk add libgcc
+
+COPY    --from=compiler /meilisearch/target/release/meilisearch .
+
+ENV     MEILI_HTTP_ADDR 0.0.0.0:7700
+CMD     ./meilisearch

LICENSE

@@ -1,13 +1,21 @@
-“Commons Clause” License Condition v1.0
+MIT License
 
-The Software is provided to you by the Licensor under the License, as defined below, subject to the following condition.
+Copyright (c) 2019-2020 Meili SAS
 
-Without limiting other conditions in the License, the grant of rights under the License will not include, and the License does not grant to you, the right to Sell the Software.
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
 
-For purposes of the foregoing, “Sell” means practicing any or all of the rights granted to you under the License to provide to third parties, for a fee or other consideration (including without limitation fees for hosting or consulting/ support services related to the Software), a product or service whose value derives, entirely or substantially, from the functionality of the Software. Any license notice or attribution required by the License must also include this Commons Clause License Condition notice.
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
 
-Software: MeiliDB
-
-License: MIT
-
-Licensor: MEILI SAS
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

Procfile

@@ -0,0 +1 @@
+web: ./target/release/meilisearch --http-addr=0.0.0.0:$PORT

README.md

@@ -1,43 +1,169 @@
-# MeiliDB
+# MeiliSearch
 
-[![Build Status](https://dev.azure.com/thomas0884/thomas/_apis/build/status/meilisearch.MeiliDB?branchName=master)](https://dev.azure.com/thomas0884/thomas/_build/latest?definitionId=1&branchName=master)
-[![dependency status](https://deps.rs/repo/github/meilisearch/MeiliDB/status.svg)](https://deps.rs/repo/github/meilisearch/MeiliDB)
-[![License](https://img.shields.io/badge/license-commons%20clause-lightgrey)](https://commonsclause.com/)
+[![Build Status](https://github.com/meilisearch/MeiliSearch/workflows/Cargo%20test/badge.svg)](https://github.com/meilisearch/MeiliSearch/actions)
+[![dependency status](https://deps.rs/repo/github/meilisearch/MeiliSearch/status.svg)](https://deps.rs/repo/github/meilisearch/MeiliSearch)
+[![License](https://img.shields.io/badge/license-MIT-informational)](https://github.com/meilisearch/MeiliSearch/blob/master/LICENSE)
 
-A _full-text search database_ based on the fast [LMDB key-value store](https://en.wikipedia.org/wiki/Lightning_Memory-Mapped_Database).
+⚡ Ultra relevant and instant full-text search API 🔍
+
+MeiliSearch is a powerful, fast, open-source, easy to use, and deploy search engine. The search and indexation are fully customizable and handles features like typo-tolerance, filters, and synonyms.
+For more [details about those features, go to our documentation](https://docs.meilisearch.com/).
+
+[![crates.io demo gif](misc/crates-io-demo.gif)](https://crates.meilisearch.com)
+> Meili helps the Rust community find crates on [crates.meilisearch.com](https://crates.meilisearch.com)
 
 ## Features
+* Search as-you-type experience (answers < 50ms)
+* Full-text search
+* Typo tolerant (understands typos and spelling mistakes)
+* Supports Kanji
+* Supports Synonym
+* Easy to install, deploy, and maintain
+* Whole documents returned
+* Highly customizable
+* RESTfull API
 
-- Provides [6 default ranking criteria](https://github.com/meilisearch/MeiliDB/blob/dc5c42821e1340e96cb90a3da472264624a26326/meilidb-core/src/criterion/mod.rs#L107-L113) used to [bucket sort](https://en.wikipedia.org/wiki/Bucket_sort) documents
-- Accepts [custom criteria](https://github.com/meilisearch/MeiliDB/blob/dc5c42821e1340e96cb90a3da472264624a26326/meilidb-core/src/criterion/mod.rs#L24-L33) and can apply them in any custom order
-- Support [ranged queries](https://github.com/meilisearch/MeiliDB/blob/dc5c42821e1340e96cb90a3da472264624a26326/meilidb-core/src/query_builder.rs#L283), useful for paginating results
-- Can [distinct](https://github.com/meilisearch/MeiliDB/blob/dc5c42821e1340e96cb90a3da472264624a26326/meilidb-core/src/query_builder.rs#L265-L270) and [filter](https://github.com/meilisearch/MeiliDB/blob/dc5c42821e1340e96cb90a3da472264624a26326/meilidb-core/src/query_builder.rs#L246-L259) returned documents based on context defined rules
-- Can store complete documents or only [user schema specified fields](https://github.com/meilisearch/MeiliDB/blob/dc5c42821e1340e96cb90a3da472264624a26326/meilidb-schema/src/lib.rs#L265-L279)
-- The [default tokenizer](https://github.com/meilisearch/MeiliDB/blob/dc5c42821e1340e96cb90a3da472264624a26326/meilidb-tokenizer/src/lib.rs) can index latin and kanji based languages
-- Returns [the matching text areas](https://github.com/meilisearch/MeiliDB/blob/dc5c42821e1340e96cb90a3da472264624a26326/meilidb-core/src/lib.rs#L66-L88), useful to highlight matched words in results
-- Accepts query time search config like the [searchable attributes](https://github.com/meilisearch/MeiliDB/blob/dc5c42821e1340e96cb90a3da472264624a26326/meilidb-core/src/query_builder.rs#L272-L275)
-- Supports [runtime incremental indexing](https://github.com/meilisearch/MeiliDB/blob/dc5c42821e1340e96cb90a3da472264624a26326/meilidb-core/src/store/mod.rs#L143-L173)
+## Quick Start
 
+### Deploy the Server
 
+#### Run it using Docker
 
-It uses [LMDB](https://en.wikipedia.org/wiki/Lightning_Memory-Mapped_Database) as the internal key-value store. The key-value store allows us to handle updates and queries with small memory and CPU overheads. The whole ranking system is [data oriented](https://github.com/meilisearch/MeiliDB/issues/82) and provides great performances.
+```bash
+docker run -it -p 7700:7700 --rm getmeili/meilisearch
+```
 
-You can [read the deep dive](deep-dive.md) if you want more information on the engine, it describes the whole process of generating updates and handling queries or you can take a look at the [typos and ranking rules](typos-ranking-rules.md) if you want to know the default rules used to sort the documents.
+#### Installation using Homebrew
 
-We will be proud if you submit issues and pull requests. You can help to grow this project and start contributing by checking [issues tagged "good-first-issue"](https://github.com/meilisearch/MeiliDB/issues?q=is%3Aissue+is%3Aopen+label%3A%22good+first+issue%22). It is a good start!
+```bash
+brew update && brew install meilisearch
+meilisearch
+```
 
-The project is only a library yet. It means that there is no binary provided yet. To get started, you can check the examples wich are made to work with the data located in the `datasets/` folder.
+#### Installation using APT
 
-MeiliDB will be a binary in a near future so you will be able to use it as a database out-of-the-box. We should be able to query it using HTTP. This is our current goal, [see the milestones](https://github.com/meilisearch/MeiliDB/milestones). In the end, the binary will be a bunch of network protocols and wrappers around the library - which will also be published on [crates.io](https://crates.io). Both the binary and the library will follow the same update cycle.
+```bash
+echo "deb [trusted=yes] https://apt.fury.io/meilisearch/ /" > /etc/apt/sources.list.d/fury.list
+apt update && apt install meilisearch-http
+meilisearch
+```
 
+#### Download the binary
 
+```bash
+curl -L https://install.meilisearch.com | sh
+./meilisearch
+```
+
+#### Run it on heroku
+
+[![Deploy](https://www.herokucdn.com/deploy/button.svg)](https://heroku.com/deploy?template=https://github.com/meilisearch/MeiliSearch)
+
+#### Compile and run it from sources
+
+If you have the Rust toolchain already installed, you can compile from the source
+
+```bash
+git clone https://github.com/meilisearch/MeiliSearch.git
+cd MeiliSearch
+cargo run --release
+```
+
+### Create an Index and Upload Some Documents
+
+We provide a movie dataset that you can use for testing purposes.
+
+```bash
+curl -L 'https://bit.ly/2PAcw9l' -o movies.json
+```
+
+MeiliSearch can serve multiple indexes, with different kinds of documents,
+therefore, it is required to create the index before sending documents to it.
+
+```bash
+curl -i -X POST 'http://127.0.0.1:7700/indexes' --data '{ "name": "Movies", "uid": "movies" }'
+```
+
+Now that the server knows about our brand new index, we can send it data.
+We provided you a small dataset that is available in the `datasets/` directory.
+
+```bash
+curl -i -X POST 'http://127.0.0.1:7700/indexes/movies/documents' \
+  --header 'content-type: application/json' \
+  --data-binary @movies.json
+```
+
+### Search for Documents
+
+#### In command line
+
+The search engine is now aware of our documents and can serve those via our HTTP server again.
+The [`jq` command-line tool](https://stedolan.github.io/jq/) can significantly help you read the server responses.
+
+```bash
+curl 'http://127.0.0.1:7700/indexes/movies/search?q=botman+robin&limit=2' | jq
+```
+
+```json
+{
+  "hits": [
+    {
+      "id": "415",
+      "title": "Batman & Robin",
+      "poster": "https://image.tmdb.org/t/p/w1280/79AYCcxw3kSKbhGpx1LiqaCAbwo.jpg",
+      "overview": "Along with crime-fighting partner Robin and new recruit Batgirl...",
+      "release_date": "1997-06-20",
+    },
+    {
+      "id": "411736",
+      "title": "Batman: Return of the Caped Crusaders",
+      "poster": "https://image.tmdb.org/t/p/w1280/GW3IyMW5Xgl0cgCN8wu96IlNpD.jpg",
+      "overview": "Adam West and Burt Ward returns to their iconic roles of Batman and Robin...",
+      "release_date": "2016-10-08",
+    }
+  ],
+  "offset": 0,
+  "limit": 2,
+  "processingTimeMs": 1,
+  "query": "botman robin"
+}
+```
+
+#### With the Web Interface
+
+MeiliSearch provides a simple web interface containing a search bar in order to quickly test the instant search experience with a given set of documents.
+
+This web interface is available in your browser at the root of the server. The default URL is [http://127.0.0.1:7700](http://127.0.0.1:7700).
+
+### Documentation
+
+Now, that you have a running MeiliSearch, you can learn more and tune your search engine using [the documentation](https://docs.meilisearch.com).
+
+## How it works
+
+MeiliSearch uses [LMDB](https://en.wikipedia.org/wiki/Lightning_Memory-Mapped_Database) as the internal key-value store. The key-value store allows us to handle updates and queries with small memory and CPU overheads. The whole ranking system is [data oriented](https://github.com/meilisearch/MeiliSearch/issues/82) and provides great performances.
+
+You can [read the deep dive](deep-dive.md) if you want more information on the engine; it describes the whole process of generating updates and handling queries. Also, you can take a look at the [typos and ranking rules](typos-ranking-rules.md) if you want to know the default rules used to sort the documents.
+
+### Technical features
+
+- Provides [6 default ranking criteria](https://github.com/meilisearch/MeiliSearch/blob/3ea5aa18a209b6973b921542d46a79e1c753c163/meilisearch-core/src/criterion/mod.rs#L106-L111) used to [bucket sort](https://en.wikipedia.org/wiki/Bucket_sort) documents
+- Accepts [custom criteria](https://github.com/meilisearch/MeiliSearch/blob/3ea5aa18a209b6973b921542d46a79e1c753c163/meilisearch-core/src/criterion/mod.rs#L20-L29) and can apply them in any custom order
+- Support [ranged queries](https://github.com/meilisearch/MeiliSearch/blob/3ea5aa18a209b6973b921542d46a79e1c753c163/meilisearch-core/src/query_builder.rs#L342), useful for paginating results
+- Can [distinct](https://github.com/meilisearch/MeiliSearch/blob/3ea5aa18a209b6973b921542d46a79e1c753c163/meilisearch-core/src/query_builder.rs#L324-L329) and [filter](https://github.com/meilisearch/MeiliSearch/blob/3ea5aa18a209b6973b921542d46a79e1c753c163/meilisearch-core/src/query_builder.rs#L313-L318) returned documents based on context defined rules
+- Searches for [concatenated](https://github.com/meilisearch/MeiliSearch/pull/164) and [splitted query words](https://github.com/meilisearch/MeiliSearch/pull/232) to improve the search quality.
+- Can store complete documents or only [user schema specified fields](https://github.com/meilisearch/MeiliSearch/blob/3ea5aa18a209b6973b921542d46a79e1c753c163/datasets/movies/schema.toml)
+- The [default tokenizer](https://github.com/meilisearch/MeiliSearch/blob/3ea5aa18a209b6973b921542d46a79e1c753c163/meilisearch-tokenizer/src/lib.rs) can index latin and kanji based languages
+- Returns [the matching text areas](https://github.com/meilisearch/MeiliSearch/blob/3ea5aa18a209b6973b921542d46a79e1c753c163/meilisearch-types/src/lib.rs#L49-L65), useful to highlight matched words in results
+- Accepts query time search config like the [searchable attributes](https://github.com/meilisearch/MeiliSearch/blob/3ea5aa18a209b6973b921542d46a79e1c753c163/meilisearch-core/src/query_builder.rs#L331-L336)
+- Supports [runtime incremental indexing](https://github.com/meilisearch/MeiliSearch/blob/3ea5aa18a209b6973b921542d46a79e1c753c163/meilisearch-core/src/store/mod.rs#L143-L212)
 
 ## Performances
 
-With a database composed of _100 353_ documents with _352_ attributes each and _3_ of them indexed.
+With a dataset composed of _100 353_ documents with _352_ attributes each and _3_ of them indexed.
 So more than _300 000_ fields indexed for _35 million_ stored we can handle more than _2.8k req/sec_ with an average response time of _9 ms_ on an Intel i7-7700 (8) @ 4.2GHz.
 
-Requests are made using [wrk](https://github.com/wg/wrk) and scripted to simulate real users queries.
+Requests are made using [wrk](https://github.com/wg/wrk) and scripted to simulate real users' queries.
 
 ```
 Running 10s test @ http://localhost:2230
@@ -50,29 +176,20 @@ Requests/sec:   2806.46
 Transfer/sec:    759.17KB
 ```
 
+We also indexed a dataset containing something like _12 millions_ cities names in _24 minutes_ on a machine with _8 cores_, _64 GB of RAM_, and a _300 GB NMVe_ SSD.<br/>
+The resulting database was _16 GB_ and search results were between _30 ms_ and _4 seconds_ for short prefix queries.
+
 ### Notes
 
 With Rust 1.32 the allocator has been [changed to use the system allocator](https://blog.rust-lang.org/2019/01/17/Rust-1.32.0.html#jemalloc-is-removed-by-default).
 We have seen much better performances when [using jemalloc as the global allocator](https://github.com/alexcrichton/jemallocator#documentation).
 
-## Usage and examples
+## Contributing
 
-Currently MeiliDB do not provide an http server but you can run the example binary.
+We will be glad if you submit issues and pull requests. You can help to grow this project and start contributing by checking [issues tagged "good-first-issue"](https://github.com/meilisearch/MeiliSearch/issues?q=is%3Aissue+is%3Aopen+label%3A%22good+first+issue%22). It is a good start!
 
-The _index_ subcommand has been made to create an index and inject documents into it. Using the command line below, the index will be named _movies_ and the _19 700_ movies of the `datasets/` will be injected in MeiliDB.
+### Analytic Events
 
-```bash
-cargo run --release --example from_file -- \
-    index example.mdb datasets/movies/data.csv \
-    --schema datasets/movies/schema.toml
-```
-
-Once the first command is done, you can query the freshly created _movies_ index using the _search_ subcomand. In this example we filtered the dataset to only show _non-adult_ movies using the non-definitive `!adult` syntax filter.
-
-```bash
-cargo run --release --example from_file -- \
-    search example.mdb
-    --number 4 \
-    --filter '!adult' \
-    id popularity adult original_title
-```
+We send events to our Amplitude instance to be aware of the number of people who use MeiliSearch.<br/>
+We only send the platform on which the server runs once by day. No other information is sent.<br/>
+If you do not want us to send events, you can disable these analytics by using the `MEILI_NO_ANALYTICS` env variable.

app.json

@@ -0,0 +1,16 @@
+{
+  "name": "MeiliSearch",
+  "description": "Ultra relevant, instant and typo-tolerant full-text search API",
+  "keywords": [
+    "search-engine",
+    "instant search",
+    "search API"
+  ],
+  "website": "https://docs.meilisearch.com/",
+  "repository": "https://github.com/meilisearch/MeiliSearch",
+  "buildpacks": [
+    {
+      "url": "https://github.com/emk/heroku-buildpack-rust"
+    }
+  ]
+}

azure-pipelines.yml

@@ -1,52 +0,0 @@
----
-trigger:
-  branches:
-    include: [ master ]
-
-pr: [ master ]
-
-jobs:
-  - job: test
-    pool:
-      vmImage: 'Ubuntu 16.04'
-    container: tpayet/chiquitita:latest
-    steps:
-      - script: |
-          curl https://sh.rustup.rs -sSf | sh -s -- -y --default-toolchain nightly
-          $HOME/.cargo/bin/rustup component add rustfmt
-        displayName: 'Install rustc and components'
-      - script: |
-          $HOME/.cargo/bin/cargo check
-        displayName: 'Check MeiliDB'
-      - script: |
-          $HOME/.cargo/bin/cargo test
-        displayName: 'Test MeiliDB'
-      - script: |
-          $HOME/.cargo/bin/cargo fmt --all -- --check
-        displayName: 'Fmt MeiliDB'
-
-  - job: build
-    dependsOn:
-      - test
-    condition: succeeded()
-    pool:
-      vmImage: 'Ubuntu 16.04'
-    container: tpayet/chiquitita:latest
-    steps:
-      - script: |
-          curl https://sh.rustup.rs -sSf | sh -s -- -y --default-toolchain nightly
-          $HOME/.cargo/bin/rustup component add rustfmt
-        displayName: 'Install rustc and components'
-      - script: |
-          $HOME/.cargo/bin/cargo build --release
-        displayName: 'Build MeiliDB'
-      - task: CopyFiles@2
-        inputs:
-          contents: '$(System.DefaultWorkingDirectory)/target/release/libmeilidb.rlib'
-          targetFolder: $(Build.ArtifactStagingDirectory)
-        displayName: 'Copy build'
-      - task: PublishBuildArtifacts@1
-        inputs:
-          artifactName: libmeilidb.rlib
-        displayName: 'Upload artifacts'
-

datasets/movies/README.md

@@ -1 +1 @@
-_datas in movies.csv are from https://www.themoviedb.org/_
+_datas in movies.csv are from https://www.themoviedb.org/_

datasets/movies/schema-movies.toml

@@ -1,21 +0,0 @@
-# This schema has been generated ...
-# The order in which the attributes are declared is important,
-# it specify the attribute xxx...
-identifier = "id"
-
-[attributes.id]
-displayed = true
-
-[attributes.title]
-displayed = true
-indexed = true
-
-[attributes.overview]
-displayed = true
-indexed = true
-
-[attributes.release_date]
-displayed = true
-
-[attributes.poster]
-displayed = true

datasets/movies/settings.json

@@ -0,0 +1,11 @@
+{
+    "primaryKey": "id",
+    "searchableAttributes": ["title", "overview"],
+    "displayedAttributes": [
+        "id",
+        "title",
+        "overview",
+        "release_date",
+        "poster"
+    ]
+}

deep-dive.md

@@ -1,8 +1,8 @@
-# A deep dive in MeiliDB
+# A deep dive in MeiliSearch
 
 On the 15 of May 2019.
 
-MeiliDB is a full text search engine based on a final state transducer named [fst](https://github.com/BurntSushi/fst) and a key-value store named [sled](https://github.com/spacejam/sled). The goal of a search engine is to store data and to respond to queries as accurate and fast as possible. To achieve this it must save the matching words in an [inverted index](https://en.wikipedia.org/wiki/Inverted_index).
+MeiliSearch is a full text search engine based on a final state transducer named [fst](https://github.com/BurntSushi/fst) and a key-value store named [sled](https://github.com/spacejam/sled). The goal of a search engine is to store data and to respond to queries as accurate and fast as possible. To achieve this it must save the matching words in an [inverted index](https://en.wikipedia.org/wiki/Inverted_index).
 
 <!-- MarkdownTOC autolink="true" -->
 
@@ -22,7 +22,7 @@ MeiliDB is a full text search engine based on a final state transducer named [fs
 
 ## Where is the data stored?
 
-MeiliDB is entirely backed by a key-value store like any good database (i.e. Postgres, MySQL). This brings a great flexibility in the way documents can be stored and updates handled along time.
+MeiliSearch is entirely backed by a key-value store like any good database (i.e. Postgres, MySQL). This brings a great flexibility in the way documents can be stored and updates handled along time.
 
 [sled will brings some](https://github.com/spacejam/sled/tree/434533332a3f485e6d2e467023be0a0b55d3a1af#plans) of the [A.C.I.D. properties](https://en.wikipedia.org/wiki/ACID_(computer_science)) to help us be sure the saved data is consistent.
 
@@ -34,7 +34,7 @@ It contain the inverted word index, the schema and the documents fields.
 
 ### The inverted word index
 
-[The inverted word index](https://github.com/meilisearch/MeiliDB/blob/3db823de002243004612e36a19b4578d800dab97/meilidb-data/src/database/words_index.rs) is a sled Tree dedicated to store and give access to all documents that contains a specific word. The information stored under the word is simply a big ordered array of where in the document the word has been found. In other word, a big list of [`DocIndex`](https://github.com/meilisearch/MeiliDB/blob/3db823de002243004612e36a19b4578d800dab97/meilidb-core/src/lib.rs#L35-L51).
+[The inverted word index](https://github.com/meilisearch/MeiliSearch/blob/3db823de002243004612e36a19b4578d800dab97/meilisearch-data/src/database/words_index.rs) is a sled Tree dedicated to store and give access to all documents that contains a specific word. The information stored under the word is simply a big ordered array of where in the document the word has been found. In other word, a big list of [`DocIndex`](https://github.com/meilisearch/MeiliSearch/blob/3db823de002243004612e36a19b4578d800dab97/meilisearch-core/src/lib.rs#L35-L51).
 
 #### A final state transducer
 
@@ -42,27 +42,27 @@ _...also abbreviated fst_
 
 This is the first entry point of the engine, you can read more about how it work with the beautiful blog post of @BurntSushi, [Index 1,600,000,000 Keys with Automata and Rust](https://blog.burntsushi.net/transducers/).
 
-To make it short it is a powerful way to store all the words that are present in the indexed documents. You construct it by giving it all the words you want to index. When you want to search in it you can provide any automaton you want, in MeiliDB [a custom levenshtein automaton](https://github.com/tantivy-search/levenshtein-automata/) is used.
+To make it short it is a powerful way to store all the words that are present in the indexed documents. You construct it by giving it all the words you want to index. When you want to search in it you can provide any automaton you want, in MeiliSearch [a custom levenshtein automaton](https://github.com/tantivy-search/levenshtein-automata/) is used.
 
 #### Document indexes
 
 The `fst` will only return the words that match with the search automaton but the goal of the search engine is to retrieve all matches in all the documents when a query is made. You want it to return some sort of position in an attribute in a document, an information about where the given word matched.
 
-To make it possible we retrieve all of the `DocIndex` corresponding to all the matching words in the fst, we use the [`WordsIndex`](https://github.com/meilisearch/MeiliDB/blob/3db823de002243004612e36a19b4578d800dab97/meilidb-data/src/database/words_index.rs#L11-L21) Tree to get the `DocIndexes` corresponding the words.
+To make it possible we retrieve all of the `DocIndex` corresponding to all the matching words in the fst, we use the [`WordsIndex`](https://github.com/meilisearch/MeiliSearch/blob/3db823de002243004612e36a19b4578d800dab97/meilisearch-data/src/database/words_index.rs#L11-L21) Tree to get the `DocIndexes` corresponding the words.
 
 ### The schema
 
-The schema is a data structure that represents which documents attributes should be stored and which should be indexed. It is stored under a the [`MainIndex`](https://github.com/meilisearch/MeiliDB/blob/3db823de002243004612e36a19b4578d800dab97/meilidb-data/src/database/main_index.rs#L12) Tree and given to MeiliDB only at the creation of an index.
+The schema is a data structure that represents which documents attributes should be stored and which should be indexed. It is stored under a the [`MainIndex`](https://github.com/meilisearch/MeiliSearch/blob/3db823de002243004612e36a19b4578d800dab97/meilisearch-data/src/database/main_index.rs#L12) Tree and given to MeiliSearch only at the creation of an index.
 
-Each document attribute is associated to a unique 16 bit number named [`SchemaAttr`](https://github.com/meilisearch/MeiliDB/blob/3db823de002243004612e36a19b4578d800dab97/meilidb-data/src/schema.rs#L186).
+Each document attribute is associated to a unique 16 bit number named [`SchemaAttr`](https://github.com/meilisearch/MeiliSearch/blob/3db823de002243004612e36a19b4578d800dab97/meilisearch-data/src/schema.rs#L186).
 
 In the future, this schema type could be given along with updates, the database could be able to handled a new schema and reindex the database according to the new one.
 
 ### Document attributes
 
-When the engine handle a query the result that the requester want is a document, not only the [`Matches`](https://github.com/meilisearch/MeiliDB/blob/3db823de002243004612e36a19b4578d800dab97/meilidb-core/src/lib.rs#L62-L88) associated to it, fields of the original document must be returned too.
+When the engine handle a query the result that the requester want is a document, not only the [`Matches`](https://github.com/meilisearch/MeiliSearch/blob/3db823de002243004612e36a19b4578d800dab97/meilisearch-core/src/lib.rs#L62-L88) associated to it, fields of the original document must be returned too.
 
-So MeiliDB again uses the power of the underlying key-value store and save the documents attributes marked as _STORE_ in the schema. The dedicated Tree for this information is the [`DocumentsIndex`](https://github.com/meilisearch/MeiliDB/blob/3db823de002243004612e36a19b4578d800dab97/meilidb-data/src/database/documents_index.rs#L11).
+So MeiliSearch again uses the power of the underlying key-value store and save the documents attributes marked as _STORE_ in the schema. The dedicated Tree for this information is the [`DocumentsIndex`](https://github.com/meilisearch/MeiliSearch/blob/3db823de002243004612e36a19b4578d800dab97/meilisearch-data/src/database/documents_index.rs#L11).
 
 When a document field is saved in the key-value store its value is binary encoded using [message pack](https://github.com/3Hren/msgpack-rust), so a document must be serializable using serde.
 
@@ -70,26 +70,26 @@ When a document field is saved in the key-value store its value is binary encode
 
 ## How is a request processed?
 
-Now that we have our inverted index we are able to return results based on a query. In the MeiliDB universe a query is a simple string containing words.
+Now that we have our inverted index we are able to return results based on a query. In the MeiliSearch universe a query is a simple string containing words.
 
 ### Query lexemes
 
-The first step to be able to call the underlying structures is to split the query in words, for that we use a [custom tokenizer](https://github.com/meilisearch/MeiliDB/blob/3db823de002243004612e36a19b4578d800dab97/meilidb-tokenizer/src/lib.rs#L82-L84). Note that a tokenizer is specialized for a human language, this is the hard part.
+The first step to be able to call the underlying structures is to split the query in words, for that we use a [custom tokenizer](https://github.com/meilisearch/MeiliSearch/blob/3db823de002243004612e36a19b4578d800dab97/meilisearch-tokenizer/src/lib.rs#L82-L84). Note that a tokenizer is specialized for a human language, this is the hard part.
 
 ### Automatons and query index
 
-So to query the fst we need an automaton, in MeiliDB we use a [levenshtein automaton](https://en.wikipedia.org/wiki/Levenshtein_automaton), this automaton is constructed using a string and a maximum distance. According to the [Algolia's blog post](https://blog.algolia.com/inside-the-algolia-engine-part-3-query-processing/#algolia%e2%80%99s-way-of-searching-for-alternatives) we [created the DFAs](https://github.com/meilisearch/MeiliDB/blob/3db823de002243004612e36a19b4578d800dab97/meilidb-core/src/automaton.rs#L59-L78) with different settings.
+So to query the fst we need an automaton, in MeiliSearch we use a [levenshtein automaton](https://en.wikipedia.org/wiki/Levenshtein_automaton), this automaton is constructed using a string and a maximum distance. According to the [Algolia's blog post](https://blog.algolia.com/inside-the-algolia-engine-part-3-query-processing/#algolia%e2%80%99s-way-of-searching-for-alternatives) we [created the DFAs](https://github.com/meilisearch/MeiliSearch/blob/3db823de002243004612e36a19b4578d800dab97/meilisearch-core/src/automaton.rs#L59-L78) with different settings.
 
 Thanks to the power of the fst library [it is possible to union multiple automatons](https://docs.rs/fst/0.3.2/fst/map/struct.OpBuilder.html#method.union) on the same fst set. The `Stream` is able to return all the matching words. We use these words to find the whole list of `DocIndexes` associated.
 
-With all these informations it is possible [to reconstruct a list of all the `DocIndexes` associated](https://github.com/meilisearch/MeiliDB/blob/3db823de002243004612e36a19b4578d800dab97/meilidb-core/src/query_builder.rs#L103-L130) with the words queried.
+With all these informations it is possible [to reconstruct a list of all the `DocIndexes` associated](https://github.com/meilisearch/MeiliSearch/blob/3db823de002243004612e36a19b4578d800dab97/meilisearch-core/src/query_builder.rs#L103-L130) with the words queried.
 
 ### Sort by criteria
 
-Now that we are able to get a big list of [DocIndexes](https://github.com/Kerollmops/MeiliDB/blob/550dc1e99224e386516877450320f694947332d4/src/lib.rs#L21-L36) it is not enough to sort them by criteria, we need more informations like the levenshtein distance or the fact that a query word match exactly the word stored in the fst. So [we stuff it a little bit](https://github.com/Kerollmops/MeiliDB/blob/550dc1e99224e386516877450320f694947332d4/src/rank/query_builder.rs#L86-L93), and aggregate all these [Matches](https://github.com/Kerollmops/MeiliDB/blob/550dc1e99224e386516877450320f694947332d4/src/lib.rs#L47-L74) for each document. This way it will be easy to sort a simple vector of document using a bunch of functions.
+Now that we are able to get a big list of [DocIndexes](https://github.com/Kerollmops/MeiliSearch/blob/550dc1e99224e386516877450320f694947332d4/src/lib.rs#L21-L36) it is not enough to sort them by criteria, we need more informations like the levenshtein distance or the fact that a query word match exactly the word stored in the fst. So [we stuff it a little bit](https://github.com/Kerollmops/MeiliSearch/blob/550dc1e99224e386516877450320f694947332d4/src/rank/query_builder.rs#L86-L93), and aggregate all these [Matches](https://github.com/Kerollmops/MeiliSearch/blob/550dc1e99224e386516877450320f694947332d4/src/lib.rs#L47-L74) for each document. This way it will be easy to sort a simple vector of document using a bunch of functions.
 
-With this big list of documents and associated matches [we are able to sort only the part of the slice that we want](https://github.com/meilisearch/MeiliDB/blob/3db823de002243004612e36a19b4578d800dab97/meilidb-core/src/query_builder.rs#L160-L188) using bucket sorting. [Each criterion](https://github.com/meilisearch/MeiliDB/blob/3db823de002243004612e36a19b4578d800dab97/meilidb-core/src/criterion/mod.rs#L95-L101) is evaluated on each subslice without copy, thanks to [GroupByMut](https://docs.rs/slice-group-by/0.2.4/slice_group_by/) which, I hope [will soon be merged](https://github.com/rust-lang/rfcs/pull/2477).
+With this big list of documents and associated matches [we are able to sort only the part of the slice that we want](https://github.com/meilisearch/MeiliSearch/blob/3db823de002243004612e36a19b4578d800dab97/meilisearch-core/src/query_builder.rs#L160-L188) using bucket sorting. [Each criterion](https://github.com/meilisearch/MeiliSearch/blob/3db823de002243004612e36a19b4578d800dab97/meilisearch-core/src/criterion/mod.rs#L95-L101) is evaluated on each subslice without copy, thanks to [GroupByMut](https://docs.rs/slice-group-by/0.2.4/slice_group_by/) which, I hope [will soon be merged](https://github.com/rust-lang/rfcs/pull/2477).
 
-Note that it is possible to customize the criteria used by using the `QueryBuilder::with_criteria` constructor, this way you can implement some custom ranking based on the document attributes using the appropriate structure and the [`document` method](https://github.com/meilisearch/MeiliDB/blob/3db823de002243004612e36a19b4578d800dab97/meilidb-data/src/database/index.rs#L86).
+Note that it is possible to customize the criteria used by using the `QueryBuilder::with_criteria` constructor, this way you can implement some custom ranking based on the document attributes using the appropriate structure and the [`document` method](https://github.com/meilisearch/MeiliSearch/blob/3db823de002243004612e36a19b4578d800dab97/meilisearch-data/src/database/index.rs#L86).
 
-At this point, MeiliDB work is over 🎉
+At this point, MeiliSearch work is over 🎉

download-latest.sh

@@ -0,0 +1,129 @@
+#!/bin/sh
+
+# COLORS
+RED='\033[31m'
+GREEN='\033[32m'
+DEFAULT='\033[0m'
+
+# GLOBALS
+GREP_SEMVER_REGEXP='\"v\([0-9]*\)[.]\([0-9]*\)[.]\([0-9]*\)\"' # i.e. "v[number].[number].[number]"
+BINARY_NAME='meilisearch'
+
+# semverParseInto and semverLT from https://github.com/cloudflare/semver_bash/blob/master/semver.sh
+
+# usage: semverParseInto version major minor patch special
+# version: the string version
+# major, minor, patch, special: will be assigned by the function
+semverParseInto() {
+    local RE='[^0-9]*\([0-9]*\)[.]\([0-9]*\)[.]\([0-9]*\)\([0-9A-Za-z-]*\)'
+    #MAJOR
+    eval $2=`echo $1 | sed -e "s#$RE#\1#"`
+    #MINOR
+    eval $3=`echo $1 | sed -e "s#$RE#\2#"`
+    #MINOR
+    eval $4=`echo $1 | sed -e "s#$RE#\3#"`
+    #SPECIAL
+    eval $5=`echo $1 | sed -e "s#$RE#\4#"`
+}
+
+# usage: semverLT version1 version2
+semverLT() {
+    local MAJOR_A=0
+    local MINOR_A=0
+    local PATCH_A=0
+    local SPECIAL_A=0
+
+    local MAJOR_B=0
+    local MINOR_B=0
+    local PATCH_B=0
+    local SPECIAL_B=0
+
+    semverParseInto $1 MAJOR_A MINOR_A PATCH_A SPECIAL_A
+    semverParseInto $2 MAJOR_B MINOR_B PATCH_B SPECIAL_B
+
+    if [ $MAJOR_A -lt $MAJOR_B ]; then
+        return 0
+    fi
+    if [ $MAJOR_A -le $MAJOR_B ] && [ $MINOR_A -lt $MINOR_B ]; then
+        return 0
+    fi
+    if [ $MAJOR_A -le $MAJOR_B ] && [ $MINOR_A -le $MINOR_B ] && [ $PATCH_A -lt $PATCH_B ]; then
+        return 0
+    fi
+    if [ "_$SPECIAL_A"  == "_" ] && [ "_$SPECIAL_B"  == "_" ] ; then
+        return 1
+    fi
+    if [ "_$SPECIAL_A"  == "_" ] && [ "_$SPECIAL_B"  != "_" ] ; then
+        return 1
+    fi
+    if [ "_$SPECIAL_A"  != "_" ] && [ "_$SPECIAL_B"  == "_" ] ; then
+        return 0
+    fi
+    if [ "_$SPECIAL_A" < "_$SPECIAL_B" ]; then
+        return 0
+    fi
+
+    return 1
+}
+
+success_usage() {
+    printf "$GREEN%s\n$DEFAULT" "MeiliSearch binary successfully downloaded as '$BINARY_NAME' file."
+    echo ''
+    echo 'Run it:'
+    echo '    $ ./meilisearch'
+    echo 'Usage:'
+    echo '    $ ./meilisearch --help'
+}
+
+failure_usage() {
+    printf "$RED%s\n$DEFAULT" 'ERROR: MeiliSearch binary is not available for your OS distribution yet.'
+    echo ''
+    echo 'However, you can easily compile the binary from the source files.'
+    echo 'Follow the steps on the docs: https://docs.meilisearch.com/advanced_guides/binary.html#how-to-compile-meilisearch'
+}
+
+# OS DETECTION
+echo 'Detecting OS distribution...'
+os_name=$(uname -s)
+if [ "$os_name" != "Darwin" ]; then
+    os_name=$(cat /etc/os-release | grep '^ID=' | tr -d '"' | cut -d '=' -f 2)
+fi
+echo "OS distribution detected: $os_name"
+case "$os_name" in
+'Darwin')
+    os='macos'
+    ;;
+'ubuntu' | 'debian')
+    os='linux'
+    ;;
+*)
+    failure_usage
+    exit 1
+esac
+
+# GET LATEST VERSION
+tags=$(curl -s 'https://api.github.com/repos/meilisearch/MeiliSearch/tags' \
+    | grep "$GREP_SEMVER_REGEXP" \
+    | grep 'name' \
+    | tr -d '"' | tr -d ',' | cut -d 'v' -f 2)
+
+latest=""
+for tag in $tags; do
+    if [ "$latest" = "" ]; then
+        latest="$tag"
+    else
+        semverLT $tag $latest
+        if [ $? -eq 1 ]; then
+            latest="$tag"
+        fi
+    fi
+done
+
+# DOWNLOAD THE LATEST
+echo "Downloading MeiliSearch binary v$latest for $os..."
+release_file="meilisearch-$os-amd64"
+link="https://github.com/meilisearch/MeiliSearch/releases/download/v$latest/$release_file"
+curl -OL "$link"
+mv "$release_file" "$BINARY_NAME"
+chmod 744 "$BINARY_NAME"
+success_usage

meilidb-core/Cargo.toml

@@ -1,45 +0,0 @@
-[package]
-name = "meilidb-core"
-version = "0.1.0"
-authors = ["Kerollmops <clement@meilisearch.com>"]
-edition = "2018"
-
-[dependencies]
-arc-swap = "0.4.3"
-bincode = "1.1.4"
-byteorder = "1.3.2"
-crossbeam-channel = "0.3.9"
-deunicode = "1.0.0"
-env_logger = "0.7.0"
-hashbrown = { version = "0.6.0", features = ["serde"] }
-heed = "0.1.0"
-log = "0.4.8"
-meilidb-schema = { path = "../meilidb-schema", version = "0.1.0" }
-meilidb-tokenizer = { path = "../meilidb-tokenizer", version = "0.1.0" }
-once_cell = "1.2.0"
-ordered-float = { version = "1.0.2", features = ["serde"] }
-sdset = "0.3.3"
-serde = { version = "1.0.101", features = ["derive"] }
-serde_json = "1.0.41"
-siphasher = "0.3.0"
-slice-group-by = "0.2.6"
-zerocopy = "0.2.8"
-
-[dependencies.levenshtein_automata]
-git = "https://github.com/Kerollmops/levenshtein-automata.git"
-branch = "arc-byte-slice"
-features = ["fst_automaton"]
-
-[dependencies.fst]
-git = "https://github.com/Kerollmops/fst.git"
-branch = "arc-byte-slice"
-
-[dev-dependencies]
-assert_matches = "1.3"
-csv = "1.0.7"
-indexmap = { version = "1.2.0", features = ["serde-1"] }
-rustyline = { version = "5.0.0", default-features = false }
-structopt = "0.3.2"
-tempfile = "3.1.0"
-termcolor = "1.0.4"
-toml = "0.5.3"

meilidb-core/src/automaton/mod.rs

@@ -1,220 +0,0 @@
-mod dfa;
-mod query_enhancer;
-
-use std::cmp::Reverse;
-use std::vec;
-
-use fst::{IntoStreamer, Streamer};
-use levenshtein_automata::DFA;
-use meilidb_tokenizer::{is_cjk, split_query_string};
-
-use crate::error::MResult;
-use crate::store;
-
-use self::dfa::{build_dfa, build_prefix_dfa};
-pub use self::query_enhancer::QueryEnhancer;
-use self::query_enhancer::QueryEnhancerBuilder;
-
-const NGRAMS: usize = 3;
-
-pub struct AutomatonProducer {
-    automatons: Vec<Vec<Automaton>>,
-}
-
-impl AutomatonProducer {
-    pub fn new(
-        reader: &heed::RoTxn,
-        query: &str,
-        main_store: store::Main,
-        synonyms_store: store::Synonyms,
-    ) -> MResult<(AutomatonProducer, QueryEnhancer)> {
-        let (automatons, query_enhancer) =
-            generate_automatons(reader, query, main_store, synonyms_store)?;
-
-        Ok((AutomatonProducer { automatons }, query_enhancer))
-    }
-
-    pub fn into_iter(self) -> vec::IntoIter<Vec<Automaton>> {
-        self.automatons.into_iter()
-    }
-}
-
-#[derive(Debug)]
-pub struct Automaton {
-    pub index: usize,
-    pub ngram: usize,
-    pub query_len: usize,
-    pub is_exact: bool,
-    pub is_prefix: bool,
-    pub query: String,
-}
-
-impl Automaton {
-    pub fn dfa(&self) -> DFA {
-        if self.is_prefix {
-            build_prefix_dfa(&self.query)
-        } else {
-            build_dfa(&self.query)
-        }
-    }
-
-    fn exact(index: usize, ngram: usize, query: &str) -> Automaton {
-        Automaton {
-            index,
-            ngram,
-            query_len: query.len(),
-            is_exact: true,
-            is_prefix: false,
-            query: query.to_string(),
-        }
-    }
-
-    fn prefix_exact(index: usize, ngram: usize, query: &str) -> Automaton {
-        Automaton {
-            index,
-            ngram,
-            query_len: query.len(),
-            is_exact: true,
-            is_prefix: true,
-            query: query.to_string(),
-        }
-    }
-
-    fn non_exact(index: usize, ngram: usize, query: &str) -> Automaton {
-        Automaton {
-            index,
-            ngram,
-            query_len: query.len(),
-            is_exact: false,
-            is_prefix: false,
-            query: query.to_string(),
-        }
-    }
-}
-
-pub fn normalize_str(string: &str) -> String {
-    let mut string = string.to_lowercase();
-
-    if !string.contains(is_cjk) {
-        string = deunicode::deunicode_with_tofu(&string, "");
-    }
-
-    string
-}
-
-fn generate_automatons(
-    reader: &heed::RoTxn,
-    query: &str,
-    main_store: store::Main,
-    synonym_store: store::Synonyms,
-) -> MResult<(Vec<Vec<Automaton>>, QueryEnhancer)> {
-    let has_end_whitespace = query.chars().last().map_or(false, char::is_whitespace);
-    let query_words: Vec<_> = split_query_string(query).map(str::to_lowercase).collect();
-    let synonyms = match main_store.synonyms_fst(reader)? {
-        Some(synonym) => synonym,
-        None => fst::Set::default(),
-    };
-
-    let mut automaton_index = 0;
-    let mut automatons = Vec::new();
-    let mut enhancer_builder = QueryEnhancerBuilder::new(&query_words);
-
-    // We must not declare the original words to the query enhancer
-    // *but* we need to push them in the automatons list first
-    let mut original_automatons = Vec::new();
-    let mut original_words = query_words.iter().peekable();
-    while let Some(word) = original_words.next() {
-        let has_following_word = original_words.peek().is_some();
-        let not_prefix_dfa = has_following_word || has_end_whitespace || word.chars().all(is_cjk);
-
-        let automaton = if not_prefix_dfa {
-            Automaton::exact(automaton_index, 1, word)
-        } else {
-            Automaton::prefix_exact(automaton_index, 1, word)
-        };
-        automaton_index += 1;
-        original_automatons.push(automaton);
-    }
-
-    automatons.push(original_automatons);
-
-    for n in 1..=NGRAMS {
-        let mut ngrams = query_words.windows(n).enumerate().peekable();
-        while let Some((query_index, ngram_slice)) = ngrams.next() {
-            let query_range = query_index..query_index + n;
-            let ngram_nb_words = ngram_slice.len();
-            let ngram = ngram_slice.join(" ");
-
-            let has_following_word = ngrams.peek().is_some();
-            let not_prefix_dfa =
-                has_following_word || has_end_whitespace || ngram.chars().all(is_cjk);
-
-            // automaton of synonyms of the ngrams
-            let normalized = normalize_str(&ngram);
-            let lev = if not_prefix_dfa {
-                build_dfa(&normalized)
-            } else {
-                build_prefix_dfa(&normalized)
-            };
-
-            let mut stream = synonyms.search(&lev).into_stream();
-            while let Some(base) = stream.next() {
-                // only trigger alternatives when the last word has been typed
-                // i.e. "new " do not but "new yo" triggers alternatives to "new york"
-                let base = std::str::from_utf8(base).unwrap();
-                let base_nb_words = split_query_string(base).count();
-                if ngram_nb_words != base_nb_words {
-                    continue;
-                }
-
-                if let Some(synonyms) = synonym_store.synonyms(reader, base.as_bytes())? {
-                    let mut stream = synonyms.into_stream();
-                    while let Some(synonyms) = stream.next() {
-                        let synonyms = std::str::from_utf8(synonyms).unwrap();
-                        let synonyms_words: Vec<_> = split_query_string(synonyms).collect();
-                        let nb_synonym_words = synonyms_words.len();
-
-                        let real_query_index = automaton_index;
-                        enhancer_builder.declare(
-                            query_range.clone(),
-                            real_query_index,
-                            &synonyms_words,
-                        );
-
-                        for synonym in synonyms_words {
-                            let automaton = if nb_synonym_words == 1 {
-                                Automaton::exact(automaton_index, n, synonym)
-                            } else {
-                                Automaton::non_exact(automaton_index, n, synonym)
-                            };
-                            automaton_index += 1;
-                            automatons.push(vec![automaton]);
-                        }
-                    }
-                }
-            }
-
-            if n != 1 {
-                // automaton of concatenation of query words
-                let concat = ngram_slice.concat();
-                let normalized = normalize_str(&concat);
-
-                let real_query_index = automaton_index;
-                enhancer_builder.declare(query_range.clone(), real_query_index, &[&normalized]);
-
-                let automaton = Automaton::exact(automaton_index, n, &normalized);
-                automaton_index += 1;
-                automatons.push(vec![automaton]);
-            }
-        }
-    }
-
-    // order automatons, the most important first,
-    // we keep the original automatons at the front.
-    automatons[1..].sort_by_key(|a| {
-        let a = a.first().unwrap();
-        (Reverse(a.is_exact), a.ngram)
-    });
-
-    Ok((automatons, enhancer_builder.build()))
-}

meilidb-core/src/automaton/query_enhancer.rs

@@ -1,423 +0,0 @@
-use std::cmp::Ordering::{Equal, Greater, Less};
-use std::ops::Range;
-
-/// Return `true` if the specified range can accept the given replacements words.
-/// Returns `false` if the replacements words are already present in the original query
-/// or if there is fewer replacement words than the range to replace.
-//
-//
-// ## Ignored because already present in original
-//
-//     new york city subway
-//     -------- ^^^^
-//   /          \
-//  [new york city]
-//
-//
-// ## Ignored because smaller than the original
-//
-//   new york city subway
-//   -------------
-//   \          /
-//    [new york]
-//
-//
-// ## Accepted because bigger than the original
-//
-//        NYC subway
-//        ---
-//       /   \
-//      /     \
-//     /       \
-//    /         \
-//   /           \
-//  [new york city]
-//
-fn rewrite_range_with<S, T>(query: &[S], range: Range<usize>, words: &[T]) -> bool
-where
-    S: AsRef<str>,
-    T: AsRef<str>,
-{
-    if words.len() <= range.len() {
-        // there is fewer or equal replacement words
-        // than there is already in the replaced range
-        return false;
-    }
-
-    // retrieve the part to rewrite but with the length
-    // of the replacement part
-    let original = query.iter().skip(range.start).take(words.len());
-
-    // check if the original query doesn't already contain
-    // the replacement words
-    !original
-        .map(AsRef::as_ref)
-        .eq(words.iter().map(AsRef::as_ref))
-}
-
-type Origin = usize;
-type RealLength = usize;
-
-struct FakeIntervalTree {
-    intervals: Vec<(Range<usize>, (Origin, RealLength))>,
-}
-
-impl FakeIntervalTree {
-    fn new(mut intervals: Vec<(Range<usize>, (Origin, RealLength))>) -> FakeIntervalTree {
-        intervals.sort_unstable_by_key(|(r, _)| (r.start, r.end));
-        FakeIntervalTree { intervals }
-    }
-
-    fn query(&self, point: usize) -> Option<(Range<usize>, (Origin, RealLength))> {
-        let element = self.intervals.binary_search_by(|(r, _)| {
-            if point >= r.start {
-                if point < r.end {
-                    Equal
-                } else {
-                    Less
-                }
-            } else {
-                Greater
-            }
-        });
-
-        let n = match element {
-            Ok(n) => n,
-            Err(n) => n,
-        };
-
-        match self.intervals.get(n) {
-            Some((range, value)) if range.contains(&point) => Some((range.clone(), *value)),
-            _otherwise => None,
-        }
-    }
-}
-
-pub struct QueryEnhancerBuilder<'a, S> {
-    query: &'a [S],
-    origins: Vec<usize>,
-    real_to_origin: Vec<(Range<usize>, (Origin, RealLength))>,
-}
-
-impl<S: AsRef<str>> QueryEnhancerBuilder<'_, S> {
-    pub fn new(query: &[S]) -> QueryEnhancerBuilder<S> {
-        // we initialize origins query indices based on their positions
-        let origins: Vec<_> = (0..=query.len()).collect();
-        let real_to_origin = origins.iter().map(|&o| (o..o + 1, (o, 1))).collect();
-
-        QueryEnhancerBuilder {
-            query,
-            origins,
-            real_to_origin,
-        }
-    }
-
-    /// Update the final real to origin query indices mapping.
-    ///
-    /// `range` is the original words range that this `replacement` words replace
-    /// and `real` is the first real query index of these replacement words.
-    pub fn declare<T>(&mut self, range: Range<usize>, real: usize, replacement: &[T])
-    where
-        T: AsRef<str>,
-    {
-        // check if the range of original words
-        // can be rewritten with the replacement words
-        if rewrite_range_with(self.query, range.clone(), replacement) {
-            // this range can be replaced so we need to
-            // modify the origins accordingly
-            let offset = replacement.len() - range.len();
-
-            let previous_padding = self.origins[range.end - 1];
-            let current_offset = (self.origins[range.end] - 1) - previous_padding;
-            let diff = offset.saturating_sub(current_offset);
-            self.origins[range.end] += diff;
-
-            for r in &mut self.origins[range.end + 1..] {
-                *r += diff;
-            }
-        }
-
-        // we need to store the real number and origins relations
-        // this way it will be possible to know by how many
-        // we need to pad real query indices
-        let real_range = real..real + replacement.len().max(range.len());
-        let real_length = replacement.len();
-        self.real_to_origin
-            .push((real_range, (range.start, real_length)));
-    }
-
-    pub fn build(self) -> QueryEnhancer {
-        QueryEnhancer {
-            origins: self.origins,
-            real_to_origin: FakeIntervalTree::new(self.real_to_origin),
-        }
-    }
-}
-
-pub struct QueryEnhancer {
-    origins: Vec<usize>,
-    real_to_origin: FakeIntervalTree,
-}
-
-impl QueryEnhancer {
-    /// Returns the query indices to use to replace this real query index.
-    pub fn replacement(&self, real: u32) -> Range<u32> {
-        let real = real as usize;
-
-        // query the fake interval tree with the real query index
-        let (range, (origin, real_length)) = self
-            .real_to_origin
-            .query(real)
-            .expect("real has never been declared");
-
-        // if `real` is the end bound of the range
-        if (range.start + real_length - 1) == real {
-            let mut count = range.len();
-            let mut new_origin = origin;
-            for (i, slice) in self.origins[new_origin..].windows(2).enumerate() {
-                let len = slice[1] - slice[0];
-                count = count.saturating_sub(len);
-                if count == 0 {
-                    new_origin = origin + i;
-                    break;
-                }
-            }
-
-            let n = real - range.start;
-            let start = self.origins[origin];
-            let end = self.origins[new_origin + 1];
-            let remaining = (end - start) - n;
-
-            Range {
-                start: (start + n) as u32,
-                end: (start + n + remaining) as u32,
-            }
-        } else {
-            // just return the origin along with
-            // the real position of the word
-            let n = real as usize - range.start;
-            let origin = self.origins[origin];
-
-            Range {
-                start: (origin + n) as u32,
-                end: (origin + n + 1) as u32,
-            }
-        }
-    }
-}
-
-#[cfg(test)]
-mod tests {
-    use super::*;
-
-    #[test]
-    fn original_unmodified() {
-        let query = ["new", "york", "city", "subway"];
-        //             0       1       2        3
-        let mut builder = QueryEnhancerBuilder::new(&query);
-
-        // new york = new york city
-        builder.declare(0..2, 4, &["new", "york", "city"]);
-        //                    ^      4       5       6
-
-        let enhancer = builder.build();
-
-        assert_eq!(enhancer.replacement(0), 0..1); // new
-        assert_eq!(enhancer.replacement(1), 1..2); // york
-        assert_eq!(enhancer.replacement(2), 2..3); // city
-        assert_eq!(enhancer.replacement(3), 3..4); // subway
-        assert_eq!(enhancer.replacement(4), 0..1); // new
-        assert_eq!(enhancer.replacement(5), 1..2); // york
-        assert_eq!(enhancer.replacement(6), 2..3); // city
-    }
-
-    #[test]
-    fn simple_growing() {
-        let query = ["new", "york", "subway"];
-        //             0       1        2
-        let mut builder = QueryEnhancerBuilder::new(&query);
-
-        // new york = new york city
-        builder.declare(0..2, 3, &["new", "york", "city"]);
-        //                    ^      3       4       5
-
-        let enhancer = builder.build();
-
-        assert_eq!(enhancer.replacement(0), 0..1); // new
-        assert_eq!(enhancer.replacement(1), 1..3); // york
-        assert_eq!(enhancer.replacement(2), 3..4); // subway
-        assert_eq!(enhancer.replacement(3), 0..1); // new
-        assert_eq!(enhancer.replacement(4), 1..2); // york
-        assert_eq!(enhancer.replacement(5), 2..3); // city
-    }
-
-    #[test]
-    fn same_place_growings() {
-        let query = ["NY", "subway"];
-        //             0       1
-        let mut builder = QueryEnhancerBuilder::new(&query);
-
-        // NY = new york
-        builder.declare(0..1, 2, &["new", "york"]);
-        //                    ^      2       3
-
-        // NY = new york city
-        builder.declare(0..1, 4, &["new", "york", "city"]);
-        //                    ^      4       5       6
-
-        // NY = NYC
-        builder.declare(0..1, 7, &["NYC"]);
-        //                    ^      7
-
-        // NY = new york city
-        builder.declare(0..1, 8, &["new", "york", "city"]);
-        //                    ^      8       9      10
-
-        // subway = underground train
-        builder.declare(1..2, 11, &["underground", "train"]);
-        //                    ^          11          12
-
-        let enhancer = builder.build();
-
-        assert_eq!(enhancer.replacement(0), 0..3); // NY
-        assert_eq!(enhancer.replacement(1), 3..5); // subway
-        assert_eq!(enhancer.replacement(2), 0..1); // new
-        assert_eq!(enhancer.replacement(3), 1..3); // york
-        assert_eq!(enhancer.replacement(4), 0..1); // new
-        assert_eq!(enhancer.replacement(5), 1..2); // york
-        assert_eq!(enhancer.replacement(6), 2..3); // city
-        assert_eq!(enhancer.replacement(7), 0..3); // NYC
-        assert_eq!(enhancer.replacement(8), 0..1); // new
-        assert_eq!(enhancer.replacement(9), 1..2); // york
-        assert_eq!(enhancer.replacement(10), 2..3); // city
-        assert_eq!(enhancer.replacement(11), 3..4); // underground
-        assert_eq!(enhancer.replacement(12), 4..5); // train
-    }
-
-    #[test]
-    fn bigger_growing() {
-        let query = ["NYC", "subway"];
-        //             0        1
-        let mut builder = QueryEnhancerBuilder::new(&query);
-
-        // NYC = new york city
-        builder.declare(0..1, 2, &["new", "york", "city"]);
-        //                    ^      2       3       4
-
-        let enhancer = builder.build();
-
-        assert_eq!(enhancer.replacement(0), 0..3); // NYC
-        assert_eq!(enhancer.replacement(1), 3..4); // subway
-        assert_eq!(enhancer.replacement(2), 0..1); // new
-        assert_eq!(enhancer.replacement(3), 1..2); // york
-        assert_eq!(enhancer.replacement(4), 2..3); // city
-    }
-
-    #[test]
-    fn middle_query_growing() {
-        let query = ["great", "awesome", "NYC", "subway"];
-        //              0         1        2        3
-        let mut builder = QueryEnhancerBuilder::new(&query);
-
-        // NYC = new york city
-        builder.declare(2..3, 4, &["new", "york", "city"]);
-        //                    ^      4       5       6
-
-        let enhancer = builder.build();
-
-        assert_eq!(enhancer.replacement(0), 0..1); // great
-        assert_eq!(enhancer.replacement(1), 1..2); // awesome
-        assert_eq!(enhancer.replacement(2), 2..5); // NYC
-        assert_eq!(enhancer.replacement(3), 5..6); // subway
-        assert_eq!(enhancer.replacement(4), 2..3); // new
-        assert_eq!(enhancer.replacement(5), 3..4); // york
-        assert_eq!(enhancer.replacement(6), 4..5); // city
-    }
-
-    #[test]
-    fn end_query_growing() {
-        let query = ["NYC", "subway"];
-        //             0        1
-        let mut builder = QueryEnhancerBuilder::new(&query);
-
-        // NYC = new york city
-        builder.declare(1..2, 2, &["underground", "train"]);
-        //                    ^         2            3
-
-        let enhancer = builder.build();
-
-        assert_eq!(enhancer.replacement(0), 0..1); // NYC
-        assert_eq!(enhancer.replacement(1), 1..3); // subway
-        assert_eq!(enhancer.replacement(2), 1..2); // underground
-        assert_eq!(enhancer.replacement(3), 2..3); // train
-    }
-
-    #[test]
-    fn multiple_growings() {
-        let query = ["great", "awesome", "NYC", "subway"];
-        //              0         1        2        3
-        let mut builder = QueryEnhancerBuilder::new(&query);
-
-        // NYC = new york city
-        builder.declare(2..3, 4, &["new", "york", "city"]);
-        //                    ^      4       5       6
-
-        // subway = underground train
-        builder.declare(3..4, 7, &["underground", "train"]);
-        //                    ^          7           8
-
-        let enhancer = builder.build();
-
-        assert_eq!(enhancer.replacement(0), 0..1); // great
-        assert_eq!(enhancer.replacement(1), 1..2); // awesome
-        assert_eq!(enhancer.replacement(2), 2..5); // NYC
-        assert_eq!(enhancer.replacement(3), 5..7); // subway
-        assert_eq!(enhancer.replacement(4), 2..3); // new
-        assert_eq!(enhancer.replacement(5), 3..4); // york
-        assert_eq!(enhancer.replacement(6), 4..5); // city
-        assert_eq!(enhancer.replacement(7), 5..6); // underground
-        assert_eq!(enhancer.replacement(8), 6..7); // train
-    }
-
-    #[test]
-    fn multiple_probable_growings() {
-        let query = ["great", "awesome", "NYC", "subway"];
-        //              0         1        2        3
-        let mut builder = QueryEnhancerBuilder::new(&query);
-
-        // NYC = new york city
-        builder.declare(2..3, 4, &["new", "york", "city"]);
-        //                    ^      4       5       6
-
-        // subway = underground train
-        builder.declare(3..4, 7, &["underground", "train"]);
-        //                    ^          7           8
-
-        // great awesome = good
-        builder.declare(0..2, 9, &["good"]);
-        //                    ^       9
-
-        // awesome NYC = NY
-        builder.declare(1..3, 10, &["NY"]);
-        //                    ^^     10
-
-        // NYC subway = metro
-        builder.declare(2..4, 11, &["metro"]);
-        //                    ^^      11
-
-        let enhancer = builder.build();
-
-        assert_eq!(enhancer.replacement(0), 0..1); // great
-        assert_eq!(enhancer.replacement(1), 1..2); // awesome
-        assert_eq!(enhancer.replacement(2), 2..5); // NYC
-        assert_eq!(enhancer.replacement(3), 5..7); // subway
-        assert_eq!(enhancer.replacement(4), 2..3); // new
-        assert_eq!(enhancer.replacement(5), 3..4); // york
-        assert_eq!(enhancer.replacement(6), 4..5); // city
-        assert_eq!(enhancer.replacement(7), 5..6); // underground
-        assert_eq!(enhancer.replacement(8), 6..7); // train
-        assert_eq!(enhancer.replacement(9), 0..2); // good
-        assert_eq!(enhancer.replacement(10), 1..5); // NY
-        assert_eq!(enhancer.replacement(11), 2..5); // metro
-    }
-}

meilidb-core/src/criterion/document_id.rs

@@ -1,16 +0,0 @@
-use crate::criterion::Criterion;
-use crate::RawDocument;
-use std::cmp::Ordering;
-
-#[derive(Debug, Clone, Copy)]
-pub struct DocumentId;
-
-impl Criterion for DocumentId {
-    fn evaluate(&self, lhs: &RawDocument, rhs: &RawDocument) -> Ordering {
-        lhs.id.cmp(&rhs.id)
-    }
-
-    fn name(&self) -> &str {
-        "DocumentId"
-    }
-}

meilidb-core/src/criterion/exact.rs

@@ -1,133 +0,0 @@
-use std::cmp::Ordering;
-
-use meilidb_schema::SchemaAttr;
-use sdset::Set;
-use slice_group_by::GroupBy;
-
-use crate::criterion::Criterion;
-use crate::RawDocument;
-
-#[inline]
-fn number_exact_matches(
-    query_index: &[u32],
-    attribute: &[u16],
-    is_exact: &[bool],
-    fields_counts: &Set<(SchemaAttr, u64)>,
-) -> usize {
-    let mut count = 0;
-    let mut index = 0;
-
-    for group in query_index.linear_group() {
-        let len = group.len();
-
-        let mut found_exact = false;
-        for (pos, _) in is_exact[index..index + len]
-            .iter()
-            .filter(|x| **x)
-            .enumerate()
-        {
-            found_exact = true;
-            if let Ok(pos) = fields_counts.binary_search_by_key(&attribute[pos], |(a, _)| a.0) {
-                let (_, count) = fields_counts[pos];
-                if count == 1 {
-                    return usize::max_value();
-                }
-            }
-        }
-
-        count += found_exact as usize;
-        index += len;
-    }
-
-    count
-}
-
-#[derive(Debug, Clone, Copy)]
-pub struct Exact;
-
-impl Criterion for Exact {
-    fn evaluate(&self, lhs: &RawDocument, rhs: &RawDocument) -> Ordering {
-        let lhs = {
-            let query_index = lhs.query_index();
-            let is_exact = lhs.is_exact();
-            let attribute = lhs.attribute();
-            let fields_counts = &lhs.fields_counts;
-
-            number_exact_matches(query_index, attribute, is_exact, fields_counts)
-        };
-
-        let rhs = {
-            let query_index = rhs.query_index();
-            let is_exact = rhs.is_exact();
-            let attribute = rhs.attribute();
-            let fields_counts = &rhs.fields_counts;
-
-            number_exact_matches(query_index, attribute, is_exact, fields_counts)
-        };
-
-        lhs.cmp(&rhs).reverse()
-    }
-
-    fn name(&self) -> &str {
-        "Exact"
-    }
-}
-
-#[cfg(test)]
-mod tests {
-    use super::*;
-
-    // typing: "soulier"
-    //
-    // doc0: "Soulier bleu"
-    // doc1: "souliereres rouge"
-    #[test]
-    fn easy_case() {
-        let doc0 = {
-            let query_index = &[0];
-            let attribute = &[0];
-            let is_exact = &[true];
-            let fields_counts = Set::new(&[(SchemaAttr(0), 2)]).unwrap();
-
-            number_exact_matches(query_index, attribute, is_exact, fields_counts)
-        };
-
-        let doc1 = {
-            let query_index = &[0];
-            let attribute = &[0];
-            let is_exact = &[false];
-            let fields_counts = Set::new(&[(SchemaAttr(0), 2)]).unwrap();
-
-            number_exact_matches(query_index, attribute, is_exact, fields_counts)
-        };
-
-        assert_eq!(doc0.cmp(&doc1).reverse(), Ordering::Less);
-    }
-
-    // typing: "soulier"
-    //
-    // doc0: { 0. "soulier" }
-    // doc1: { 0. "soulier bleu et blanc" }
-    #[test]
-    fn basic() {
-        let doc0 = {
-            let query_index = &[0];
-            let attribute = &[0];
-            let is_exact = &[true];
-            let fields_counts = Set::new(&[(SchemaAttr(0), 1)]).unwrap();
-
-            number_exact_matches(query_index, attribute, is_exact, fields_counts)
-        };
-
-        let doc1 = {
-            let query_index = &[0];
-            let attribute = &[0];
-            let is_exact = &[true];
-            let fields_counts = Set::new(&[(SchemaAttr(0), 4)]).unwrap();
-
-            number_exact_matches(query_index, attribute, is_exact, fields_counts)
-        };
-
-        assert_eq!(doc0.cmp(&doc1).reverse(), Ordering::Less);
-    }
-}

meilidb-core/src/criterion/mod.rs

@@ -1,121 +0,0 @@
-mod document_id;
-mod exact;
-mod number_of_words;
-mod sort_by_attr;
-mod sum_of_typos;
-mod sum_of_words_attribute;
-mod sum_of_words_position;
-mod words_proximity;
-
-use crate::RawDocument;
-use std::cmp::Ordering;
-
-pub use self::{
-    document_id::DocumentId, exact::Exact, number_of_words::NumberOfWords,
-    sort_by_attr::SortByAttr, sum_of_typos::SumOfTypos,
-    sum_of_words_attribute::SumOfWordsAttribute, sum_of_words_position::SumOfWordsPosition,
-    words_proximity::WordsProximity,
-};
-
-pub trait Criterion: Send + Sync {
-    fn evaluate(&self, lhs: &RawDocument, rhs: &RawDocument) -> Ordering;
-
-    fn name(&self) -> &str;
-
-    #[inline]
-    fn eq(&self, lhs: &RawDocument, rhs: &RawDocument) -> bool {
-        self.evaluate(lhs, rhs) == Ordering::Equal
-    }
-}
-
-impl<'a, T: Criterion + ?Sized + Send + Sync> Criterion for &'a T {
-    fn evaluate(&self, lhs: &RawDocument, rhs: &RawDocument) -> Ordering {
-        (**self).evaluate(lhs, rhs)
-    }
-
-    fn name(&self) -> &str {
-        (**self).name()
-    }
-
-    fn eq(&self, lhs: &RawDocument, rhs: &RawDocument) -> bool {
-        (**self).eq(lhs, rhs)
-    }
-}
-
-impl<T: Criterion + ?Sized> Criterion for Box<T> {
-    fn evaluate(&self, lhs: &RawDocument, rhs: &RawDocument) -> Ordering {
-        (**self).evaluate(lhs, rhs)
-    }
-
-    fn name(&self) -> &str {
-        (**self).name()
-    }
-
-    fn eq(&self, lhs: &RawDocument, rhs: &RawDocument) -> bool {
-        (**self).eq(lhs, rhs)
-    }
-}
-
-#[derive(Default)]
-pub struct CriteriaBuilder<'a> {
-    inner: Vec<Box<dyn Criterion + 'a>>,
-}
-
-impl<'a> CriteriaBuilder<'a> {
-    pub fn new() -> CriteriaBuilder<'a> {
-        CriteriaBuilder { inner: Vec::new() }
-    }
-
-    pub fn with_capacity(capacity: usize) -> CriteriaBuilder<'a> {
-        CriteriaBuilder {
-            inner: Vec::with_capacity(capacity),
-        }
-    }
-
-    pub fn reserve(&mut self, additional: usize) {
-        self.inner.reserve(additional)
-    }
-
-    pub fn add<C: 'a>(mut self, criterion: C) -> CriteriaBuilder<'a>
-    where
-        C: Criterion,
-    {
-        self.push(criterion);
-        self
-    }
-
-    pub fn push<C: 'a>(&mut self, criterion: C)
-    where
-        C: Criterion,
-    {
-        self.inner.push(Box::new(criterion));
-    }
-
-    pub fn build(self) -> Criteria<'a> {
-        Criteria { inner: self.inner }
-    }
-}
-
-pub struct Criteria<'a> {
-    inner: Vec<Box<dyn Criterion + 'a>>,
-}
-
-impl<'a> Default for Criteria<'a> {
-    fn default() -> Self {
-        CriteriaBuilder::with_capacity(7)
-            .add(SumOfTypos)
-            .add(NumberOfWords)
-            .add(WordsProximity)
-            .add(SumOfWordsAttribute)
-            .add(SumOfWordsPosition)
-            .add(Exact)
-            .add(DocumentId)
-            .build()
-    }
-}
-
-impl<'a> AsRef<[Box<dyn Criterion + 'a>]> for Criteria<'a> {
-    fn as_ref(&self) -> &[Box<dyn Criterion + 'a>] {
-        &self.inner
-    }
-}

meilidb-core/src/criterion/number_of_words.rs

@@ -1,31 +0,0 @@
-use crate::criterion::Criterion;
-use crate::RawDocument;
-use slice_group_by::GroupBy;
-use std::cmp::Ordering;
-
-#[inline]
-fn number_of_query_words(query_index: &[u32]) -> usize {
-    query_index.linear_group().count()
-}
-
-#[derive(Debug, Clone, Copy)]
-pub struct NumberOfWords;
-
-impl Criterion for NumberOfWords {
-    fn evaluate(&self, lhs: &RawDocument, rhs: &RawDocument) -> Ordering {
-        let lhs = {
-            let query_index = lhs.query_index();
-            number_of_query_words(query_index)
-        };
-        let rhs = {
-            let query_index = rhs.query_index();
-            number_of_query_words(query_index)
-        };
-
-        lhs.cmp(&rhs).reverse()
-    }
-
-    fn name(&self) -> &str {
-        "NumberOfWords"
-    }
-}

meilidb-core/src/criterion/sum_of_typos.rs

@@ -1,116 +0,0 @@
-use std::cmp::Ordering;
-
-use slice_group_by::GroupBy;
-
-use crate::criterion::Criterion;
-use crate::RawDocument;
-
-// This function is a wrong logarithmic 10 function.
-// It is safe to panic on input number higher than 3,
-// the number of typos is never bigger than that.
-#[inline]
-fn custom_log10(n: u8) -> f32 {
-    match n {
-        0 => 0.0,     // log(1)
-        1 => 0.30102, // log(2)
-        2 => 0.47712, // log(3)
-        3 => 0.60205, // log(4)
-        _ => panic!("invalid number"),
-    }
-}
-
-#[inline]
-fn sum_matches_typos(query_index: &[u32], distance: &[u8]) -> usize {
-    let mut number_words: usize = 0;
-    let mut sum_typos = 0.0;
-    let mut index = 0;
-
-    for group in query_index.linear_group() {
-        sum_typos += custom_log10(distance[index]);
-        number_words += 1;
-        index += group.len();
-    }
-
-    (number_words as f32 / (sum_typos + 1.0) * 1000.0) as usize
-}
-
-#[derive(Debug, Clone, Copy)]
-pub struct SumOfTypos;
-
-impl Criterion for SumOfTypos {
-    fn evaluate(&self, lhs: &RawDocument, rhs: &RawDocument) -> Ordering {
-        let lhs = {
-            let query_index = lhs.query_index();
-            let distance = lhs.distance();
-            sum_matches_typos(query_index, distance)
-        };
-
-        let rhs = {
-            let query_index = rhs.query_index();
-            let distance = rhs.distance();
-            sum_matches_typos(query_index, distance)
-        };
-
-        lhs.cmp(&rhs).reverse()
-    }
-
-    fn name(&self) -> &str {
-        "SumOfTypos"
-    }
-}
-
-#[cfg(test)]
-mod tests {
-    use super::*;
-
-    // typing: "Geox CEO"
-    //
-    // doc0: "Geox SpA: CEO and Executive"
-    // doc1: "Mt. Gox CEO Resigns From Bitcoin Foundation"
-    #[test]
-    fn one_typo_reference() {
-        let query_index0 = &[0, 1];
-        let distance0 = &[0, 0];
-
-        let query_index1 = &[0, 1];
-        let distance1 = &[1, 0];
-
-        let doc0 = sum_matches_typos(query_index0, distance0);
-        let doc1 = sum_matches_typos(query_index1, distance1);
-        assert_eq!(doc0.cmp(&doc1).reverse(), Ordering::Less);
-    }
-
-    // typing: "bouton manchette"
-    //
-    // doc0: "bouton manchette"
-    // doc1: "bouton"
-    #[test]
-    fn no_typo() {
-        let query_index0 = &[0, 1];
-        let distance0 = &[0, 0];
-
-        let query_index1 = &[0];
-        let distance1 = &[0];
-
-        let doc0 = sum_matches_typos(query_index0, distance0);
-        let doc1 = sum_matches_typos(query_index1, distance1);
-        assert_eq!(doc0.cmp(&doc1).reverse(), Ordering::Less);
-    }
-
-    // typing: "bouton manchztte"
-    //
-    // doc0: "bouton manchette"
-    // doc1: "bouton"
-    #[test]
-    fn one_typo() {
-        let query_index0 = &[0, 1];
-        let distance0 = &[0, 1];
-
-        let query_index1 = &[0];
-        let distance1 = &[0];
-
-        let doc0 = sum_matches_typos(query_index0, distance0);
-        let doc1 = sum_matches_typos(query_index1, distance1);
-        assert_eq!(doc0.cmp(&doc1).reverse(), Ordering::Less);
-    }
-}

meilidb-core/src/criterion/sum_of_words_attribute.rs

@@ -1,64 +0,0 @@
-use crate::criterion::Criterion;
-use crate::RawDocument;
-use slice_group_by::GroupBy;
-use std::cmp::Ordering;
-
-#[inline]
-fn sum_matches_attributes(query_index: &[u32], attribute: &[u16]) -> usize {
-    let mut sum_attributes = 0;
-    let mut index = 0;
-
-    for group in query_index.linear_group() {
-        sum_attributes += attribute[index] as usize;
-        index += group.len();
-    }
-
-    sum_attributes
-}
-
-#[derive(Debug, Clone, Copy)]
-pub struct SumOfWordsAttribute;
-
-impl Criterion for SumOfWordsAttribute {
-    fn evaluate(&self, lhs: &RawDocument, rhs: &RawDocument) -> Ordering {
-        let lhs = {
-            let query_index = lhs.query_index();
-            let attribute = lhs.attribute();
-            sum_matches_attributes(query_index, attribute)
-        };
-
-        let rhs = {
-            let query_index = rhs.query_index();
-            let attribute = rhs.attribute();
-            sum_matches_attributes(query_index, attribute)
-        };
-
-        lhs.cmp(&rhs)
-    }
-
-    fn name(&self) -> &str {
-        "SumOfWordsAttribute"
-    }
-}
-
-#[cfg(test)]
-mod tests {
-    use super::*;
-
-    // typing: "soulier"
-    //
-    // doc0: { 0. "Soulier bleu", 1. "bla bla bla" }
-    // doc1: { 0. "Botte rouge", 1. "Soulier en cuir" }
-    #[test]
-    fn title_vs_description() {
-        let query_index0 = &[0];
-        let attribute0 = &[0];
-
-        let query_index1 = &[0];
-        let attribute1 = &[1];
-
-        let doc0 = sum_matches_attributes(query_index0, attribute0);
-        let doc1 = sum_matches_attributes(query_index1, attribute1);
-        assert_eq!(doc0.cmp(&doc1), Ordering::Less);
-    }
-}

meilidb-core/src/criterion/sum_of_words_position.rs

@@ -1,64 +0,0 @@
-use crate::criterion::Criterion;
-use crate::RawDocument;
-use slice_group_by::GroupBy;
-use std::cmp::Ordering;
-
-#[inline]
-fn sum_matches_attribute_index(query_index: &[u32], word_index: &[u16]) -> usize {
-    let mut sum_word_index = 0;
-    let mut index = 0;
-
-    for group in query_index.linear_group() {
-        sum_word_index += word_index[index] as usize;
-        index += group.len();
-    }
-
-    sum_word_index
-}
-
-#[derive(Debug, Clone, Copy)]
-pub struct SumOfWordsPosition;
-
-impl Criterion for SumOfWordsPosition {
-    fn evaluate(&self, lhs: &RawDocument, rhs: &RawDocument) -> Ordering {
-        let lhs = {
-            let query_index = lhs.query_index();
-            let word_index = lhs.word_index();
-            sum_matches_attribute_index(query_index, word_index)
-        };
-
-        let rhs = {
-            let query_index = rhs.query_index();
-            let word_index = rhs.word_index();
-            sum_matches_attribute_index(query_index, word_index)
-        };
-
-        lhs.cmp(&rhs)
-    }
-
-    fn name(&self) -> &str {
-        "SumOfWordsPosition"
-    }
-}
-
-#[cfg(test)]
-mod tests {
-    use super::*;
-
-    // typing: "soulier"
-    //
-    // doc0: "Soulier bleu"
-    // doc1: "Botte rouge et soulier noir"
-    #[test]
-    fn easy_case() {
-        let query_index0 = &[0];
-        let word_index0 = &[0];
-
-        let query_index1 = &[0];
-        let word_index1 = &[3];
-
-        let doc0 = sum_matches_attribute_index(query_index0, word_index0);
-        let doc1 = sum_matches_attribute_index(query_index1, word_index1);
-        assert_eq!(doc0.cmp(&doc1), Ordering::Less);
-    }
-}

meilidb-core/src/criterion/words_proximity.rs

@@ -1,164 +0,0 @@
-use crate::criterion::Criterion;
-use crate::RawDocument;
-use slice_group_by::GroupBy;
-use std::cmp::{self, Ordering};
-
-const MAX_DISTANCE: u16 = 8;
-
-#[inline]
-fn clone_tuple<T: Clone, U: Clone>((a, b): (&T, &U)) -> (T, U) {
-    (a.clone(), b.clone())
-}
-
-fn index_proximity(lhs: u16, rhs: u16) -> u16 {
-    if lhs < rhs {
-        cmp::min(rhs - lhs, MAX_DISTANCE)
-    } else {
-        cmp::min(lhs - rhs, MAX_DISTANCE) + 1
-    }
-}
-
-fn attribute_proximity((lattr, lwi): (u16, u16), (rattr, rwi): (u16, u16)) -> u16 {
-    if lattr != rattr {
-        return MAX_DISTANCE;
-    }
-    index_proximity(lwi, rwi)
-}
-
-fn min_proximity((lattr, lwi): (&[u16], &[u16]), (rattr, rwi): (&[u16], &[u16])) -> u16 {
-    let mut min_prox = u16::max_value();
-
-    for a in lattr.iter().zip(lwi) {
-        for b in rattr.iter().zip(rwi) {
-            let a = clone_tuple(a);
-            let b = clone_tuple(b);
-            min_prox = cmp::min(min_prox, attribute_proximity(a, b));
-        }
-    }
-
-    min_prox
-}
-
-fn matches_proximity(
-    query_index: &[u32],
-    distance: &[u8],
-    attribute: &[u16],
-    word_index: &[u16],
-) -> u16 {
-    let mut query_index_groups = query_index.linear_group();
-    let mut proximity = 0;
-    let mut index = 0;
-
-    let get_attr_wi = |index: usize, group_len: usize| {
-        // retrieve the first distance group (with the lowest values)
-        let len = distance[index..index + group_len]
-            .linear_group()
-            .next()
-            .unwrap()
-            .len();
-
-        let rattr = &attribute[index..index + len];
-        let rwi = &word_index[index..index + len];
-
-        (rattr, rwi)
-    };
-
-    let mut last = query_index_groups.next().map(|group| {
-        let attr_wi = get_attr_wi(index, group.len());
-        index += group.len();
-        attr_wi
-    });
-
-    // iter by windows of size 2
-    while let (Some(lhs), Some(rhs)) = (last, query_index_groups.next()) {
-        let attr_wi = get_attr_wi(index, rhs.len());
-        proximity += min_proximity(lhs, attr_wi);
-        last = Some(attr_wi);
-        index += rhs.len();
-    }
-
-    proximity
-}
-
-#[derive(Debug, Clone, Copy)]
-pub struct WordsProximity;
-
-impl Criterion for WordsProximity {
-    fn evaluate(&self, lhs: &RawDocument, rhs: &RawDocument) -> Ordering {
-        let lhs = {
-            let query_index = lhs.query_index();
-            let distance = lhs.distance();
-            let attribute = lhs.attribute();
-            let word_index = lhs.word_index();
-            matches_proximity(query_index, distance, attribute, word_index)
-        };
-
-        let rhs = {
-            let query_index = rhs.query_index();
-            let distance = rhs.distance();
-            let attribute = rhs.attribute();
-            let word_index = rhs.word_index();
-            matches_proximity(query_index, distance, attribute, word_index)
-        };
-
-        lhs.cmp(&rhs)
-    }
-
-    fn name(&self) -> &str {
-        "WordsProximity"
-    }
-}
-
-#[cfg(test)]
-mod tests {
-    use super::*;
-
-    #[test]
-    fn three_different_attributes() {
-        // "soup" "of the" "the day"
-        //
-        // { id: 0, attr: 0, attr_index: 0 }
-        // { id: 1, attr: 1, attr_index: 0 }
-        // { id: 2, attr: 1, attr_index: 1 }
-        // { id: 2, attr: 2, attr_index: 0 }
-        // { id: 3, attr: 3, attr_index: 1 }
-
-        let query_index = &[0, 1, 2, 2, 3];
-        let distance = &[0, 0, 0, 0, 0];
-        let attribute = &[0, 1, 1, 2, 3];
-        let word_index = &[0, 0, 1, 0, 1];
-
-        //   soup -> of = 8
-        // + of -> the  = 1
-        // + the -> day = 8 (not 1)
-        assert_eq!(
-            matches_proximity(query_index, distance, attribute, word_index),
-            17
-        );
-    }
-
-    #[test]
-    fn two_different_attributes() {
-        // "soup day" "soup of the day"
-        //
-        // { id: 0, attr: 0, attr_index: 0 }
-        // { id: 0, attr: 1, attr_index: 0 }
-        // { id: 1, attr: 1, attr_index: 1 }
-        // { id: 2, attr: 1, attr_index: 2 }
-        // { id: 3, attr: 0, attr_index: 1 }
-        // { id: 3, attr: 1, attr_index: 3 }
-
-        let query_index = &[0, 0, 1, 2, 3, 3];
-        let distance = &[0, 0, 0, 0, 0, 0];
-        let attribute = &[0, 1, 1, 1, 0, 1];
-        let word_index = &[0, 0, 1, 2, 1, 3];
-
-        //   soup -> of = 1
-        // + of -> the  = 1
-        // + the -> day = 1
-        assert_eq!(
-            matches_proximity(query_index, distance, attribute, word_index),
-            3
-        );
-    }
-}

meilidb-core/src/database.rs

@@ -1,200 +0,0 @@
-use std::collections::hash_map::{Entry, HashMap};
-use std::fs::File;
-use std::path::Path;
-use std::sync::{Arc, RwLock};
-use std::{fs, thread};
-
-use crossbeam_channel::Receiver;
-use heed::types::{Str, Unit};
-use heed::{CompactionOption, Result as ZResult};
-use log::{debug, error};
-
-use crate::{store, update, Index, MResult};
-
-pub type BoxUpdateFn = Box<dyn Fn(update::UpdateResult) + Send + Sync + 'static>;
-type ArcSwapFn = arc_swap::ArcSwapOption<BoxUpdateFn>;
-
-pub struct Database {
-    pub env: heed::Env,
-    common_store: heed::PolyDatabase,
-    indexes_store: heed::Database<Str, Unit>,
-    indexes: RwLock<HashMap<String, (Index, Arc<ArcSwapFn>, thread::JoinHandle<()>)>>,
-}
-
-fn update_awaiter(receiver: Receiver<()>, env: heed::Env, update_fn: Arc<ArcSwapFn>, index: Index) {
-    for () in receiver {
-        // consume all updates in order (oldest first)
-        loop {
-            let mut writer = match env.write_txn() {
-                Ok(writer) => writer,
-                Err(e) => {
-                    error!("LMDB writer transaction begin failed: {}", e);
-                    break;
-                }
-            };
-
-            match update::update_task(&mut writer, index.clone()) {
-                Ok(Some(status)) => {
-                    if let Err(e) = writer.commit() {
-                        error!("update transaction failed: {}", e)
-                    }
-
-                    if let Some(ref callback) = *update_fn.load() {
-                        (callback)(status);
-                    }
-                }
-                // no more updates to handle for now
-                Ok(None) => {
-                    debug!("no more updates");
-                    writer.abort();
-                    break;
-                }
-                Err(e) => {
-                    error!("update task failed: {}", e);
-                    writer.abort()
-                }
-            }
-        }
-    }
-}
-
-impl Database {
-    pub fn open_or_create(path: impl AsRef<Path>) -> MResult<Database> {
-        fs::create_dir_all(path.as_ref())?;
-
-        let env = heed::EnvOpenOptions::new()
-            .map_size(10 * 1024 * 1024 * 1024) // 10GB
-            .max_dbs(3000)
-            .open(path)?;
-
-        let common_store = env.create_dyn_database(Some("common"))?;
-        let indexes_store = env.create_database::<Str, Unit>(Some("indexes"))?;
-
-        // list all indexes that needs to be opened
-        let mut must_open = Vec::new();
-        let reader = env.read_txn()?;
-        for result in indexes_store.iter(&reader)? {
-            let (index_name, _) = result?;
-            must_open.push(index_name.to_owned());
-        }
-
-        reader.abort();
-
-        // open the previously aggregated indexes
-        let mut indexes = HashMap::new();
-        for index_name in must_open {
-            let (sender, receiver) = crossbeam_channel::bounded(100);
-            let index = match store::open(&env, &index_name, sender.clone())? {
-                Some(index) => index,
-                None => {
-                    log::warn!(
-                        "the index {} doesn't exist or has not all the databases",
-                        index_name
-                    );
-                    continue;
-                }
-            };
-            let update_fn = Arc::new(ArcSwapFn::empty());
-
-            let env_clone = env.clone();
-            let index_clone = index.clone();
-            let update_fn_clone = update_fn.clone();
-
-            let handle = thread::spawn(move || {
-                update_awaiter(receiver, env_clone, update_fn_clone, index_clone)
-            });
-
-            // send an update notification to make sure that
-            // possible pre-boot updates are consumed
-            sender.send(()).unwrap();
-
-            let result = indexes.insert(index_name, (index, update_fn, handle));
-            assert!(
-                result.is_none(),
-                "The index should not have been already open"
-            );
-        }
-
-        Ok(Database {
-            env,
-            common_store,
-            indexes_store,
-            indexes: RwLock::new(indexes),
-        })
-    }
-
-    pub fn open_index(&self, name: impl AsRef<str>) -> Option<Index> {
-        let indexes_lock = self.indexes.read().unwrap();
-        match indexes_lock.get(name.as_ref()) {
-            Some((index, ..)) => Some(index.clone()),
-            None => None,
-        }
-    }
-
-    pub fn create_index(&self, name: impl AsRef<str>) -> MResult<Index> {
-        let name = name.as_ref();
-        let mut indexes_lock = self.indexes.write().unwrap();
-
-        match indexes_lock.entry(name.to_owned()) {
-            Entry::Occupied(_) => Err(crate::Error::IndexAlreadyExists),
-            Entry::Vacant(entry) => {
-                let (sender, receiver) = crossbeam_channel::bounded(100);
-                let index = store::create(&self.env, name, sender)?;
-
-                let mut writer = self.env.write_txn()?;
-                self.indexes_store.put(&mut writer, name, &())?;
-
-                let env_clone = self.env.clone();
-                let index_clone = index.clone();
-
-                let no_update_fn = Arc::new(ArcSwapFn::empty());
-                let no_update_fn_clone = no_update_fn.clone();
-
-                let handle = thread::spawn(move || {
-                    update_awaiter(receiver, env_clone, no_update_fn_clone, index_clone)
-                });
-
-                writer.commit()?;
-                entry.insert((index.clone(), no_update_fn, handle));
-
-                Ok(index)
-            }
-        }
-    }
-
-    pub fn set_update_callback(&self, name: impl AsRef<str>, update_fn: BoxUpdateFn) -> bool {
-        let indexes_lock = self.indexes.read().unwrap();
-        match indexes_lock.get(name.as_ref()) {
-            Some((_, current_update_fn, _)) => {
-                let update_fn = Some(Arc::new(update_fn));
-                current_update_fn.swap(update_fn);
-                true
-            }
-            None => false,
-        }
-    }
-
-    pub fn unset_update_callback(&self, name: impl AsRef<str>) -> bool {
-        let indexes_lock = self.indexes.read().unwrap();
-        match indexes_lock.get(name.as_ref()) {
-            Some((_, current_update_fn, _)) => {
-                current_update_fn.swap(None);
-                true
-            }
-            None => false,
-        }
-    }
-
-    pub fn copy_and_compact_to_path<P: AsRef<Path>>(&self, path: P) -> ZResult<File> {
-        self.env.copy_to_path(path, CompactionOption::Enabled)
-    }
-
-    pub fn indexes_names(&self) -> MResult<Vec<String>> {
-        let indexes = self.indexes.read().unwrap();
-        Ok(indexes.keys().cloned().collect())
-    }
-
-    pub fn common_store(&self) -> heed::PolyDatabase {
-        self.common_store
-    }
-}

meilidb-core/src/lib.rs

@@ -1,168 +0,0 @@
-#[cfg(test)]
-#[macro_use]
-extern crate assert_matches;
-
-mod automaton;
-pub mod criterion;
-mod database;
-mod distinct_map;
-mod error;
-mod number;
-mod query_builder;
-mod ranked_map;
-mod raw_document;
-pub mod raw_indexer;
-mod reordered_attrs;
-pub mod serde;
-pub mod store;
-mod update;
-
-pub use self::database::{BoxUpdateFn, Database};
-pub use self::error::{Error, MResult};
-pub use self::number::{Number, ParseNumberError};
-pub use self::ranked_map::RankedMap;
-pub use self::raw_document::RawDocument;
-pub use self::store::Index;
-pub use self::update::{UpdateResult, UpdateStatus, UpdateType};
-
-use ::serde::{Deserialize, Serialize};
-use zerocopy::{AsBytes, FromBytes};
-
-/// Represent an internally generated document unique identifier.
-///
-/// It is used to inform the database the document you want to deserialize.
-/// Helpful for custom ranking.
-#[derive(
-    Debug,
-    Copy,
-    Clone,
-    Eq,
-    PartialEq,
-    PartialOrd,
-    Ord,
-    Hash,
-    Serialize,
-    Deserialize,
-    AsBytes,
-    FromBytes,
-)]
-#[repr(C)]
-pub struct DocumentId(pub u64);
-
-/// This structure represent the position of a word
-/// in a document and its attributes.
-///
-/// This is stored in the map, generated at index time,
-/// extracted and interpreted at search time.
-#[derive(Debug, Copy, Clone, PartialEq, Eq, PartialOrd, Ord, Hash, AsBytes, FromBytes)]
-#[repr(C)]
-pub struct DocIndex {
-    /// The document identifier where the word was found.
-    pub document_id: DocumentId,
-
-    /// The attribute in the document where the word was found
-    /// along with the index in it.
-    pub attribute: u16,
-    pub word_index: u16,
-
-    /// The position in bytes where the word was found
-    /// along with the length of it.
-    ///
-    /// It informs on the original word area in the text indexed
-    /// without needing to run the tokenizer again.
-    pub char_index: u16,
-    pub char_length: u16,
-}
-
-/// This structure represent a matching word with informations
-/// on the location of the word in the document.
-///
-/// The order of the field is important because it defines
-/// the way these structures are ordered between themselves.
-#[derive(Debug, Copy, Clone, PartialEq, Eq, PartialOrd, Ord, Hash)]
-pub struct Highlight {
-    /// The attribute in the document where the word was found
-    /// along with the index in it.
-    pub attribute: u16,
-
-    /// The position in bytes where the word was found.
-    ///
-    /// It informs on the original word area in the text indexed
-    /// without needing to run the tokenizer again.
-    pub char_index: u16,
-
-    /// The length in bytes of the found word.
-    ///
-    /// It informs on the original word area in the text indexed
-    /// without needing to run the tokenizer again.
-    pub char_length: u16,
-}
-
-#[doc(hidden)]
-#[derive(Debug, Copy, Clone, PartialEq, Eq, PartialOrd, Ord, Hash)]
-pub struct TmpMatch {
-    pub query_index: u32,
-    pub distance: u8,
-    pub attribute: u16,
-    pub word_index: u16,
-    pub is_exact: bool,
-}
-
-#[derive(Debug, Clone, PartialEq, Eq, PartialOrd, Ord, Hash)]
-pub struct Document {
-    pub id: DocumentId,
-    pub highlights: Vec<Highlight>,
-
-    #[cfg(test)]
-    pub matches: Vec<TmpMatch>,
-}
-
-impl Document {
-    #[cfg(not(test))]
-    fn from_raw(raw: RawDocument) -> Document {
-        Document {
-            id: raw.id,
-            highlights: raw.highlights,
-        }
-    }
-
-    #[cfg(test)]
-    fn from_raw(raw: RawDocument) -> Document {
-        let len = raw.query_index().len();
-        let mut matches = Vec::with_capacity(len);
-
-        let query_index = raw.query_index();
-        let distance = raw.distance();
-        let attribute = raw.attribute();
-        let word_index = raw.word_index();
-        let is_exact = raw.is_exact();
-
-        for i in 0..len {
-            let match_ = TmpMatch {
-                query_index: query_index[i],
-                distance: distance[i],
-                attribute: attribute[i],
-                word_index: word_index[i],
-                is_exact: is_exact[i],
-            };
-            matches.push(match_);
-        }
-
-        Document {
-            id: raw.id,
-            matches,
-            highlights: raw.highlights,
-        }
-    }
-}
-
-#[cfg(test)]
-mod tests {
-    use super::*;
-    use std::mem;
-
-    #[test]
-    fn docindex_mem_size() {
-        assert_eq!(mem::size_of::<DocIndex>(), 16);
-    }
-}

meilidb-core/src/query_enhancer.rs

@@ -1,398 +0,0 @@
-use std::ops::Range;
-use std::cmp::Ordering::{Less, Greater, Equal};
-
-/// Return `true` if the specified range can accept the given replacements words.
-/// Returns `false` if the replacements words are already present in the original query
-/// or if there is fewer replacement words than the range to replace.
-//
-//
-// ## Ignored because already present in original
-//
-//     new york city subway
-//     -------- ^^^^
-//   /          \
-//  [new york city]
-//
-//
-// ## Ignored because smaller than the original
-//
-//   new york city subway
-//   -------------
-//   \          /
-//    [new york]
-//
-//
-// ## Accepted because bigger than the original
-//
-//        NYC subway
-//        ---
-//       /   \
-//      /     \
-//     /       \
-//    /         \
-//   /           \
-//  [new york city]
-//
-fn rewrite_range_with<S, T>(query: &[S], range: Range<usize>, words: &[T]) -> bool
-where S: AsRef<str>,
-      T: AsRef<str>,
-{
-    if words.len() <= range.len() {
-        // there is fewer or equal replacement words
-        // than there is already in the replaced range
-        return false
-    }
-
-    // retrieve the part to rewrite but with the length
-    // of the replacement part
-    let original = query.iter().skip(range.start).take(words.len());
-
-    // check if the original query doesn't already contain
-    // the replacement words
-    !original.map(AsRef::as_ref).eq(words.iter().map(AsRef::as_ref))
-}
-
-type Origin = usize;
-type RealLength = usize;
-
-struct FakeIntervalTree {
-    intervals: Vec<(Range<usize>, (Origin, RealLength))>,
-}
-
-impl FakeIntervalTree {
-    fn new(mut intervals: Vec<(Range<usize>, (Origin, RealLength))>) -> FakeIntervalTree {
-        intervals.sort_unstable_by_key(|(r, _)| (r.start, r.end));
-        FakeIntervalTree { intervals }
-    }
-
-    fn query(&self, point: usize) -> Option<(Range<usize>, (Origin, RealLength))> {
-        let element = self.intervals.binary_search_by(|(r, _)| {
-            if point >= r.start {
-                if point < r.end { Equal } else { Less }
-            } else { Greater }
-        });
-
-        let n = match element { Ok(n) => n, Err(n) => n };
-
-        match self.intervals.get(n) {
-            Some((range, value)) if range.contains(&point) => Some((range.clone(), *value)),
-            _otherwise => None,
-        }
-    }
-}
-
-pub struct QueryEnhancerBuilder<'a, S> {
-    query: &'a [S],
-    origins: Vec<usize>,
-    real_to_origin: Vec<(Range<usize>, (Origin, RealLength))>,
-}
-
-impl<S: AsRef<str>> QueryEnhancerBuilder<'_, S> {
-    pub fn new(query: &[S]) -> QueryEnhancerBuilder<S> {
-        // we initialize origins query indices based on their positions
-        let origins: Vec<_> = (0..query.len() + 1).collect();
-        let real_to_origin = origins.iter().map(|&o| (o..o+1, (o, 1))).collect();
-
-        QueryEnhancerBuilder { query, origins, real_to_origin }
-    }
-
-    /// Update the final real to origin query indices mapping.
-    ///
-    /// `range` is the original words range that this `replacement` words replace
-    /// and `real` is the first real query index of these replacement words.
-    pub fn declare<T>(&mut self, range: Range<usize>, real: usize, replacement: &[T])
-    where T: AsRef<str>,
-    {
-        // check if the range of original words
-        // can be rewritten with the replacement words
-        if rewrite_range_with(self.query, range.clone(), replacement) {
-
-            // this range can be replaced so we need to
-            // modify the origins accordingly
-            let offset = replacement.len() - range.len();
-
-            let previous_padding = self.origins[range.end - 1];
-            let current_offset = (self.origins[range.end] - 1) - previous_padding;
-            let diff = offset.saturating_sub(current_offset);
-            self.origins[range.end] += diff;
-
-            for r in &mut self.origins[range.end + 1..] {
-                *r += diff;
-            }
-        }
-
-        // we need to store the real number and origins relations
-        // this way it will be possible to know by how many
-        // we need to pad real query indices
-        let real_range = real..real + replacement.len().max(range.len());
-        let real_length = replacement.len();
-        self.real_to_origin.push((real_range, (range.start, real_length)));
-    }
-
-    pub fn build(self) -> QueryEnhancer {
-        QueryEnhancer {
-            origins: self.origins,
-            real_to_origin: FakeIntervalTree::new(self.real_to_origin),
-        }
-    }
-}
-
-pub struct QueryEnhancer {
-    origins: Vec<usize>,
-    real_to_origin: FakeIntervalTree,
-}
-
-impl QueryEnhancer {
-    /// Returns the query indices to use to replace this real query index.
-    pub fn replacement(&self, real: u32) -> Range<u32> {
-        let real = real as usize;
-
-        // query the fake interval tree with the real query index
-        let (range, (origin, real_length)) =
-            self.real_to_origin
-                .query(real)
-                .expect("real has never been declared");
-
-        // if `real` is the end bound of the range
-        if (range.start + real_length - 1) == real {
-            let mut count = range.len();
-            let mut new_origin = origin;
-            for (i, slice) in self.origins[new_origin..].windows(2).enumerate() {
-                let len = slice[1] - slice[0];
-                count = count.saturating_sub(len);
-                if count == 0 { new_origin = origin + i; break }
-            }
-
-            let n = real - range.start;
-            let start = self.origins[origin];
-            let end = self.origins[new_origin + 1];
-            let remaining = (end - start) - n;
-
-            Range { start: (start + n) as u32, end: (start + n + remaining) as u32 }
-
-        } else {
-            // just return the origin along with
-            // the real position of the word
-            let n = real as usize - range.start;
-            let origin = self.origins[origin];
-
-            Range { start: (origin + n) as u32, end: (origin + n + 1) as u32 }
-        }
-    }
-}
-
-#[cfg(test)]
-mod tests {
-    use super::*;
-
-    #[test]
-    fn original_unmodified() {
-        let query = ["new", "york", "city", "subway"];
-        //             0       1       2        3
-        let mut builder = QueryEnhancerBuilder::new(&query);
-
-        // new york = new york city
-        builder.declare(0..2, 4, &["new", "york", "city"]);
-        //                    ^      4       5       6
-
-        let enhancer = builder.build();
-
-        assert_eq!(enhancer.replacement(0), 0..1); // new
-        assert_eq!(enhancer.replacement(1), 1..2); // york
-        assert_eq!(enhancer.replacement(2), 2..3); // city
-        assert_eq!(enhancer.replacement(3), 3..4); // subway
-        assert_eq!(enhancer.replacement(4), 0..1); // new
-        assert_eq!(enhancer.replacement(5), 1..2); // york
-        assert_eq!(enhancer.replacement(6), 2..3); // city
-    }
-
-    #[test]
-    fn simple_growing() {
-        let query = ["new", "york", "subway"];
-        //             0       1        2
-        let mut builder = QueryEnhancerBuilder::new(&query);
-
-        // new york = new york city
-        builder.declare(0..2, 3, &["new", "york", "city"]);
-        //                    ^      3       4       5
-
-        let enhancer = builder.build();
-
-        assert_eq!(enhancer.replacement(0), 0..1); // new
-        assert_eq!(enhancer.replacement(1), 1..3); // york
-        assert_eq!(enhancer.replacement(2), 3..4); // subway
-        assert_eq!(enhancer.replacement(3), 0..1); // new
-        assert_eq!(enhancer.replacement(4), 1..2); // york
-        assert_eq!(enhancer.replacement(5), 2..3); // city
-    }
-
-    #[test]
-    fn same_place_growings() {
-        let query = ["NY", "subway"];
-        //             0       1
-        let mut builder = QueryEnhancerBuilder::new(&query);
-
-        // NY = new york
-        builder.declare(0..1, 2, &["new", "york"]);
-        //                    ^      2       3
-
-        // NY = new york city
-        builder.declare(0..1, 4, &["new", "york", "city"]);
-        //                    ^      4       5       6
-
-        // NY = NYC
-        builder.declare(0..1, 7, &["NYC"]);
-        //                    ^      7
-
-        // NY = new york city
-        builder.declare(0..1, 8, &["new", "york", "city"]);
-        //                    ^      8       9      10
-
-        // subway = underground train
-        builder.declare(1..2, 11, &["underground", "train"]);
-        //                    ^          11          12
-
-        let enhancer = builder.build();
-
-        assert_eq!(enhancer.replacement(0), 0..3); // NY
-        assert_eq!(enhancer.replacement(1), 3..5); // subway
-        assert_eq!(enhancer.replacement(2), 0..1); // new
-        assert_eq!(enhancer.replacement(3), 1..3); // york
-        assert_eq!(enhancer.replacement(4), 0..1); // new
-        assert_eq!(enhancer.replacement(5), 1..2); // york
-        assert_eq!(enhancer.replacement(6), 2..3); // city
-        assert_eq!(enhancer.replacement(7), 0..3); // NYC
-        assert_eq!(enhancer.replacement(8), 0..1); // new
-        assert_eq!(enhancer.replacement(9), 1..2); // york
-        assert_eq!(enhancer.replacement(10), 2..3); // city
-        assert_eq!(enhancer.replacement(11), 3..4); // underground
-        assert_eq!(enhancer.replacement(12), 4..5); // train
-    }
-
-    #[test]
-    fn bigger_growing() {
-        let query = ["NYC", "subway"];
-        //             0        1
-        let mut builder = QueryEnhancerBuilder::new(&query);
-
-        // NYC = new york city
-        builder.declare(0..1, 2, &["new", "york", "city"]);
-        //                    ^      2       3       4
-
-        let enhancer = builder.build();
-
-        assert_eq!(enhancer.replacement(0), 0..3); // NYC
-        assert_eq!(enhancer.replacement(1), 3..4); // subway
-        assert_eq!(enhancer.replacement(2), 0..1); // new
-        assert_eq!(enhancer.replacement(3), 1..2); // york
-        assert_eq!(enhancer.replacement(4), 2..3); // city
-    }
-
-    #[test]
-    fn middle_query_growing() {
-        let query = ["great", "awesome", "NYC", "subway"];
-        //              0         1        2        3
-        let mut builder = QueryEnhancerBuilder::new(&query);
-
-        // NYC = new york city
-        builder.declare(2..3, 4, &["new", "york", "city"]);
-        //                    ^      4       5       6
-
-        let enhancer = builder.build();
-
-        assert_eq!(enhancer.replacement(0), 0..1); // great
-        assert_eq!(enhancer.replacement(1), 1..2); // awesome
-        assert_eq!(enhancer.replacement(2), 2..5); // NYC
-        assert_eq!(enhancer.replacement(3), 5..6); // subway
-        assert_eq!(enhancer.replacement(4), 2..3); // new
-        assert_eq!(enhancer.replacement(5), 3..4); // york
-        assert_eq!(enhancer.replacement(6), 4..5); // city
-    }
-
-    #[test]
-    fn end_query_growing() {
-        let query = ["NYC", "subway"];
-        //             0        1
-        let mut builder = QueryEnhancerBuilder::new(&query);
-
-        // NYC = new york city
-        builder.declare(1..2, 2, &["underground", "train"]);
-        //                    ^         2            3
-
-        let enhancer = builder.build();
-
-        assert_eq!(enhancer.replacement(0), 0..1); // NYC
-        assert_eq!(enhancer.replacement(1), 1..3); // subway
-        assert_eq!(enhancer.replacement(2), 1..2); // underground
-        assert_eq!(enhancer.replacement(3), 2..3); // train
-    }
-
-    #[test]
-    fn multiple_growings() {
-        let query = ["great", "awesome", "NYC", "subway"];
-        //              0         1        2        3
-        let mut builder = QueryEnhancerBuilder::new(&query);
-
-        // NYC = new york city
-        builder.declare(2..3, 4, &["new", "york", "city"]);
-        //                    ^      4       5       6
-
-        // subway = underground train
-        builder.declare(3..4, 7, &["underground", "train"]);
-        //                    ^          7           8
-
-        let enhancer = builder.build();
-
-        assert_eq!(enhancer.replacement(0), 0..1); // great
-        assert_eq!(enhancer.replacement(1), 1..2); // awesome
-        assert_eq!(enhancer.replacement(2), 2..5); // NYC
-        assert_eq!(enhancer.replacement(3), 5..7); // subway
-        assert_eq!(enhancer.replacement(4), 2..3); // new
-        assert_eq!(enhancer.replacement(5), 3..4); // york
-        assert_eq!(enhancer.replacement(6), 4..5); // city
-        assert_eq!(enhancer.replacement(7), 5..6); // underground
-        assert_eq!(enhancer.replacement(8), 6..7); // train
-    }
-
-    #[test]
-    fn multiple_probable_growings() {
-        let query = ["great", "awesome", "NYC", "subway"];
-        //              0         1        2        3
-        let mut builder = QueryEnhancerBuilder::new(&query);
-
-        // NYC = new york city
-        builder.declare(2..3, 4, &["new", "york", "city"]);
-        //                    ^      4       5       6
-
-        // subway = underground train
-        builder.declare(3..4, 7, &["underground", "train"]);
-        //                    ^          7           8
-
-        // great awesome = good
-        builder.declare(0..2, 9, &["good"]);
-        //                    ^       9
-
-        // awesome NYC = NY
-        builder.declare(1..3, 10, &["NY"]);
-        //                    ^^     10
-
-        // NYC subway = metro
-        builder.declare(2..4, 11, &["metro"]);
-        //                    ^^      11
-
-        let enhancer = builder.build();
-
-        assert_eq!(enhancer.replacement(0),  0..1); // great
-        assert_eq!(enhancer.replacement(1),  1..2); // awesome
-        assert_eq!(enhancer.replacement(2),  2..5); // NYC
-        assert_eq!(enhancer.replacement(3),  5..7); // subway
-        assert_eq!(enhancer.replacement(4),  2..3); // new
-        assert_eq!(enhancer.replacement(5),  3..4); // york
-        assert_eq!(enhancer.replacement(6),  4..5); // city
-        assert_eq!(enhancer.replacement(7),  5..6); // underground
-        assert_eq!(enhancer.replacement(8),  6..7); // train
-        assert_eq!(enhancer.replacement(9),  0..2); // good
-        assert_eq!(enhancer.replacement(10), 1..5); // NY
-        assert_eq!(enhancer.replacement(11), 2..5); // metro
-    }
-}

meilidb-core/src/raw_document.rs

@@ -1,186 +0,0 @@
-use std::fmt;
-use std::sync::Arc;
-
-use meilidb_schema::SchemaAttr;
-use sdset::SetBuf;
-use slice_group_by::GroupBy;
-
-use crate::{DocumentId, Highlight, TmpMatch};
-
-#[derive(Clone)]
-pub struct RawDocument {
-    pub id: DocumentId,
-    pub matches: SharedMatches,
-    pub highlights: Vec<Highlight>,
-    pub fields_counts: SetBuf<(SchemaAttr, u64)>,
-}
-
-impl RawDocument {
-    pub fn query_index(&self) -> &[u32] {
-        let r = self.matches.range;
-        // it is safe because construction/modifications
-        // can only be done in this module
-        unsafe {
-            &self
-                .matches
-                .matches
-                .query_index
-                .get_unchecked(r.start..r.end)
-        }
-    }
-
-    pub fn distance(&self) -> &[u8] {
-        let r = self.matches.range;
-        // it is safe because construction/modifications
-        // can only be done in this module
-        unsafe { &self.matches.matches.distance.get_unchecked(r.start..r.end) }
-    }
-
-    pub fn attribute(&self) -> &[u16] {
-        let r = self.matches.range;
-        // it is safe because construction/modifications
-        // can only be done in this module
-        unsafe { &self.matches.matches.attribute.get_unchecked(r.start..r.end) }
-    }
-
-    pub fn word_index(&self) -> &[u16] {
-        let r = self.matches.range;
-        // it is safe because construction/modifications
-        // can only be done in this module
-        unsafe {
-            &self
-                .matches
-                .matches
-                .word_index
-                .get_unchecked(r.start..r.end)
-        }
-    }
-
-    pub fn is_exact(&self) -> &[bool] {
-        let r = self.matches.range;
-        // it is safe because construction/modifications
-        // can only be done in this module
-        unsafe { &self.matches.matches.is_exact.get_unchecked(r.start..r.end) }
-    }
-}
-
-impl fmt::Debug for RawDocument {
-    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
-        f.write_str("RawDocument {\r\n")?;
-        f.write_fmt(format_args!("{:>15}: {:?},\r\n", "id", self.id))?;
-        f.write_fmt(format_args!(
-            "{:>15}: {:^5?},\r\n",
-            "query_index",
-            self.query_index()
-        ))?;
-        f.write_fmt(format_args!(
-            "{:>15}: {:^5?},\r\n",
-            "distance",
-            self.distance()
-        ))?;
-        f.write_fmt(format_args!(
-            "{:>15}: {:^5?},\r\n",
-            "attribute",
-            self.attribute()
-        ))?;
-        f.write_fmt(format_args!(
-            "{:>15}: {:^5?},\r\n",
-            "word_index",
-            self.word_index()
-        ))?;
-        f.write_fmt(format_args!(
-            "{:>15}: {:^5?},\r\n",
-            "is_exact",
-            self.is_exact()
-        ))?;
-        f.write_str("}")?;
-        Ok(())
-    }
-}
-
-pub fn raw_documents_from(
-    matches: SetBuf<(DocumentId, TmpMatch)>,
-    highlights: SetBuf<(DocumentId, Highlight)>,
-    fields_counts: SetBuf<(DocumentId, SchemaAttr, u64)>,
-) -> Vec<RawDocument> {
-    let mut docs_ranges: Vec<(_, Range, _, _)> = Vec::new();
-    let mut matches2 = Matches::with_capacity(matches.len());
-
-    let matches = matches.linear_group_by_key(|(id, _)| *id);
-    let highlights = highlights.linear_group_by_key(|(id, _)| *id);
-    let fields_counts = fields_counts.linear_group_by_key(|(id, _, _)| *id);
-
-    for ((mgroup, hgroup), fgroup) in matches.zip(highlights).zip(fields_counts) {
-        debug_assert_eq!(mgroup[0].0, hgroup[0].0);
-        debug_assert_eq!(mgroup[0].0, fgroup[0].0);
-
-        let document_id = mgroup[0].0;
-        let start = docs_ranges.last().map(|(_, r, _, _)| r.end).unwrap_or(0);
-        let end = start + mgroup.len();
-        let highlights = hgroup.iter().map(|(_, h)| *h).collect();
-        let fields_counts = SetBuf::new(fgroup.iter().map(|(_, a, c)| (*a, *c)).collect()).unwrap();
-
-        docs_ranges.push((document_id, Range { start, end }, highlights, fields_counts));
-        matches2.extend_from_slice(mgroup);
-    }
-
-    let matches = Arc::new(matches2);
-    docs_ranges
-        .into_iter()
-        .map(|(id, range, highlights, fields_counts)| {
-            let matches = SharedMatches {
-                range,
-                matches: matches.clone(),
-            };
-            RawDocument {
-                id,
-                matches,
-                highlights,
-                fields_counts,
-            }
-        })
-        .collect()
-}
-
-#[derive(Debug, Copy, Clone)]
-struct Range {
-    start: usize,
-    end: usize,
-}
-
-#[derive(Clone)]
-pub struct SharedMatches {
-    range: Range,
-    matches: Arc<Matches>,
-}
-
-#[derive(Clone)]
-struct Matches {
-    query_index: Vec<u32>,
-    distance: Vec<u8>,
-    attribute: Vec<u16>,
-    word_index: Vec<u16>,
-    is_exact: Vec<bool>,
-}
-
-impl Matches {
-    fn with_capacity(cap: usize) -> Matches {
-        Matches {
-            query_index: Vec::with_capacity(cap),
-            distance: Vec::with_capacity(cap),
-            attribute: Vec::with_capacity(cap),
-            word_index: Vec::with_capacity(cap),
-            is_exact: Vec::with_capacity(cap),
-        }
-    }
-
-    fn extend_from_slice(&mut self, matches: &[(DocumentId, TmpMatch)]) {
-        for (_, match_) in matches {
-            self.query_index.push(match_.query_index);
-            self.distance.push(match_.distance);
-            self.attribute.push(match_.attribute);
-            self.word_index.push(match_.word_index);
-            self.is_exact.push(match_.is_exact);
-        }
-    }
-}

meilidb-core/src/raw_indexer.rs

@@ -1,255 +0,0 @@
-use std::collections::{BTreeMap, HashMap};
-use std::convert::TryFrom;
-
-use crate::{DocIndex, DocumentId};
-use deunicode::deunicode_with_tofu;
-use meilidb_schema::SchemaAttr;
-use meilidb_tokenizer::{is_cjk, SeqTokenizer, Token, Tokenizer};
-use sdset::SetBuf;
-
-type Word = Vec<u8>; // TODO make it be a SmallVec
-
-pub struct RawIndexer {
-    word_limit: usize, // the maximum number of indexed words
-    words_doc_indexes: BTreeMap<Word, Vec<DocIndex>>,
-    docs_words: HashMap<DocumentId, Vec<Word>>,
-}
-
-pub struct Indexed {
-    pub words_doc_indexes: BTreeMap<Word, SetBuf<DocIndex>>,
-    pub docs_words: HashMap<DocumentId, fst::Set>,
-}
-
-impl RawIndexer {
-    pub fn new() -> RawIndexer {
-        RawIndexer::with_word_limit(1000)
-    }
-
-    pub fn with_word_limit(limit: usize) -> RawIndexer {
-        RawIndexer {
-            word_limit: limit,
-            words_doc_indexes: BTreeMap::new(),
-            docs_words: HashMap::new(),
-        }
-    }
-
-    pub fn index_text(&mut self, id: DocumentId, attr: SchemaAttr, text: &str) -> usize {
-        let mut number_of_words = 0;
-        let lowercase_text = text.to_lowercase();
-        let deunicoded = deunicode_with_tofu(&lowercase_text, "");
-
-        // TODO compute the deunicoded version after the cjk check
-        let next = if !lowercase_text.contains(is_cjk) && lowercase_text != deunicoded {
-            Some(deunicoded)
-        } else {
-            None
-        };
-        let iter = Some(lowercase_text).into_iter().chain(next);
-
-        for text in iter {
-            // we must not count 2 times the same words
-            number_of_words = 0;
-
-            for token in Tokenizer::new(&text) {
-                let must_continue = index_token(
-                    token,
-                    id,
-                    attr,
-                    self.word_limit,
-                    &mut self.words_doc_indexes,
-                    &mut self.docs_words,
-                );
-
-                if !must_continue {
-                    break;
-                }
-
-                number_of_words += 1;
-            }
-        }
-
-        number_of_words
-    }
-
-    pub fn index_text_seq<'a, I, IT>(&mut self, id: DocumentId, attr: SchemaAttr, iter: I)
-    where
-        I: IntoIterator<Item = &'a str, IntoIter = IT>,
-        IT: Iterator<Item = &'a str> + Clone,
-    {
-        // TODO serialize this to one call to the SeqTokenizer loop
-
-        let lowercased: Vec<_> = iter.into_iter().map(str::to_lowercase).collect();
-        let iter = lowercased.iter().map(|t| t.as_str());
-
-        for token in SeqTokenizer::new(iter) {
-            let must_continue = index_token(
-                token,
-                id,
-                attr,
-                self.word_limit,
-                &mut self.words_doc_indexes,
-                &mut self.docs_words,
-            );
-
-            if !must_continue {
-                break;
-            }
-        }
-
-        let deunicoded: Vec<_> = lowercased
-            .into_iter()
-            .map(|lowercase_text| {
-                if lowercase_text.contains(is_cjk) {
-                    return lowercase_text;
-                }
-                let deunicoded = deunicode_with_tofu(&lowercase_text, "");
-                if lowercase_text != deunicoded {
-                    deunicoded
-                } else {
-                    lowercase_text
-                }
-            })
-            .collect();
-        let iter = deunicoded.iter().map(|t| t.as_str());
-
-        for token in SeqTokenizer::new(iter) {
-            let must_continue = index_token(
-                token,
-                id,
-                attr,
-                self.word_limit,
-                &mut self.words_doc_indexes,
-                &mut self.docs_words,
-            );
-
-            if !must_continue {
-                break;
-            }
-        }
-    }
-
-    pub fn build(self) -> Indexed {
-        let words_doc_indexes = self
-            .words_doc_indexes
-            .into_iter()
-            .map(|(word, indexes)| (word, SetBuf::from_dirty(indexes)))
-            .collect();
-
-        let docs_words = self
-            .docs_words
-            .into_iter()
-            .map(|(id, mut words)| {
-                words.sort_unstable();
-                words.dedup();
-                (id, fst::Set::from_iter(words).unwrap())
-            })
-            .collect();
-
-        Indexed {
-            words_doc_indexes,
-            docs_words,
-        }
-    }
-}
-
-impl Default for RawIndexer {
-    fn default() -> Self {
-        Self::new()
-    }
-}
-
-fn index_token(
-    token: Token,
-    id: DocumentId,
-    attr: SchemaAttr,
-    word_limit: usize,
-    words_doc_indexes: &mut BTreeMap<Word, Vec<DocIndex>>,
-    docs_words: &mut HashMap<DocumentId, Vec<Word>>,
-) -> bool {
-    if token.word_index >= word_limit {
-        return false;
-    }
-
-    match token_to_docindex(id, attr, token) {
-        Some(docindex) => {
-            let word = Vec::from(token.word);
-            words_doc_indexes
-                .entry(word.clone())
-                .or_insert_with(Vec::new)
-                .push(docindex);
-            docs_words.entry(id).or_insert_with(Vec::new).push(word);
-        }
-        None => return false,
-    }
-
-    true
-}
-
-fn token_to_docindex(id: DocumentId, attr: SchemaAttr, token: Token) -> Option<DocIndex> {
-    let word_index = u16::try_from(token.word_index).ok()?;
-    let char_index = u16::try_from(token.char_index).ok()?;
-    let char_length = u16::try_from(token.word.chars().count()).ok()?;
-
-    let docindex = DocIndex {
-        document_id: id,
-        attribute: attr.0,
-        word_index,
-        char_index,
-        char_length,
-    };
-
-    Some(docindex)
-}
-
-#[cfg(test)]
-mod tests {
-    use super::*;
-
-    #[test]
-    fn strange_apostrophe() {
-        let mut indexer = RawIndexer::new();
-
-        let docid = DocumentId(0);
-        let attr = SchemaAttr(0);
-        let text = "Zut, l’aspirateur, j’ai oublié de l’éteindre !";
-        indexer.index_text(docid, attr, text);
-
-        let Indexed {
-            words_doc_indexes, ..
-        } = indexer.build();
-
-        assert!(words_doc_indexes.get(&b"l"[..]).is_some());
-        assert!(words_doc_indexes.get(&b"aspirateur"[..]).is_some());
-        assert!(words_doc_indexes.get(&b"ai"[..]).is_some());
-        assert!(words_doc_indexes.get(&b"eteindre"[..]).is_some());
-
-        // with the ugly apostrophe...
-        assert!(words_doc_indexes
-            .get(&"l’éteindre".to_owned().into_bytes())
-            .is_some());
-    }
-
-    #[test]
-    fn strange_apostrophe_in_sequence() {
-        let mut indexer = RawIndexer::new();
-
-        let docid = DocumentId(0);
-        let attr = SchemaAttr(0);
-        let text = vec!["Zut, l’aspirateur, j’ai oublié de l’éteindre !"];
-        indexer.index_text_seq(docid, attr, text);
-
-        let Indexed {
-            words_doc_indexes, ..
-        } = indexer.build();
-
-        assert!(words_doc_indexes.get(&b"l"[..]).is_some());
-        assert!(words_doc_indexes.get(&b"aspirateur"[..]).is_some());
-        assert!(words_doc_indexes.get(&b"ai"[..]).is_some());
-        assert!(words_doc_indexes.get(&b"eteindre"[..]).is_some());
-
-        // with the ugly apostrophe...
-        assert!(words_doc_indexes
-            .get(&"l’éteindre".to_owned().into_bytes())
-            .is_some());
-    }
-}

meilidb-core/src/reordered_attrs.rs

@@ -1,27 +0,0 @@
-#[derive(Default, Clone)]
-pub struct ReorderedAttrs {
-    count: usize,
-    reorders: Vec<Option<u16>>,
-}
-
-impl ReorderedAttrs {
-    pub fn new() -> ReorderedAttrs {
-        ReorderedAttrs {
-            count: 0,
-            reorders: Vec::new(),
-        }
-    }
-
-    pub fn insert_attribute(&mut self, attribute: u16) {
-        self.reorders.resize(attribute as usize + 1, None);
-        self.reorders[attribute as usize] = Some(self.count as u16);
-        self.count += 1;
-    }
-
-    pub fn get(&self, attribute: u16) -> Option<u16> {
-        match self.reorders.get(attribute as usize) {
-            Some(Some(attribute)) => Some(*attribute),
-            _ => None,
-        }
-    }
-}

meilidb-core/src/store/documents_fields.rs

@@ -1,74 +0,0 @@
-use heed::types::{ByteSlice, OwnedType};
-use heed::Result as ZResult;
-use meilidb_schema::SchemaAttr;
-
-use super::DocumentAttrKey;
-use crate::DocumentId;
-
-#[derive(Copy, Clone)]
-pub struct DocumentsFields {
-    pub(crate) documents_fields: heed::Database<OwnedType<DocumentAttrKey>, ByteSlice>,
-}
-
-impl DocumentsFields {
-    pub fn put_document_field(
-        self,
-        writer: &mut heed::RwTxn,
-        document_id: DocumentId,
-        attribute: SchemaAttr,
-        value: &[u8],
-    ) -> ZResult<()> {
-        let key = DocumentAttrKey::new(document_id, attribute);
-        self.documents_fields.put(writer, &key, value)
-    }
-
-    pub fn del_all_document_fields(
-        self,
-        writer: &mut heed::RwTxn,
-        document_id: DocumentId,
-    ) -> ZResult<usize> {
-        let start = DocumentAttrKey::new(document_id, SchemaAttr::min());
-        let end = DocumentAttrKey::new(document_id, SchemaAttr::max());
-        self.documents_fields.delete_range(writer, start..=end)
-    }
-
-    pub fn document_attribute<'txn>(
-        self,
-        reader: &'txn heed::RoTxn,
-        document_id: DocumentId,
-        attribute: SchemaAttr,
-    ) -> ZResult<Option<&'txn [u8]>> {
-        let key = DocumentAttrKey::new(document_id, attribute);
-        self.documents_fields.get(reader, &key)
-    }
-
-    pub fn document_fields<'txn>(
-        self,
-        reader: &'txn heed::RoTxn,
-        document_id: DocumentId,
-    ) -> ZResult<DocumentFieldsIter<'txn>> {
-        let start = DocumentAttrKey::new(document_id, SchemaAttr::min());
-        let end = DocumentAttrKey::new(document_id, SchemaAttr::max());
-        let iter = self.documents_fields.range(reader, start..=end)?;
-        Ok(DocumentFieldsIter { iter })
-    }
-}
-
-pub struct DocumentFieldsIter<'txn> {
-    iter: heed::RoRange<'txn, OwnedType<DocumentAttrKey>, ByteSlice>,
-}
-
-impl<'txn> Iterator for DocumentFieldsIter<'txn> {
-    type Item = ZResult<(SchemaAttr, &'txn [u8])>;
-
-    fn next(&mut self) -> Option<Self::Item> {
-        match self.iter.next() {
-            Some(Ok((key, bytes))) => {
-                let attr = SchemaAttr(key.attr.get());
-                Some(Ok((attr, bytes)))
-            }
-            Some(Err(e)) => Some(Err(e)),
-            None => None,
-        }
-    }
-}

meilidb-core/src/store/main.rs

@@ -1,102 +0,0 @@
-use crate::RankedMap;
-use heed::types::{ByteSlice, OwnedType, SerdeBincode, Str};
-use heed::Result as ZResult;
-use meilidb_schema::Schema;
-use std::sync::Arc;
-
-const CUSTOMS_KEY: &str = "customs-key";
-const NUMBER_OF_DOCUMENTS_KEY: &str = "number-of-documents";
-const RANKED_MAP_KEY: &str = "ranked-map";
-const SCHEMA_KEY: &str = "schema";
-const SYNONYMS_KEY: &str = "synonyms";
-const WORDS_KEY: &str = "words";
-
-#[derive(Copy, Clone)]
-pub struct Main {
-    pub(crate) main: heed::PolyDatabase,
-}
-
-impl Main {
-    pub fn put_words_fst(self, writer: &mut heed::RwTxn, fst: &fst::Set) -> ZResult<()> {
-        let bytes = fst.as_fst().as_bytes();
-        self.main.put::<Str, ByteSlice>(writer, WORDS_KEY, bytes)
-    }
-
-    pub fn words_fst(self, reader: &heed::RoTxn) -> ZResult<Option<fst::Set>> {
-        match self.main.get::<Str, ByteSlice>(reader, WORDS_KEY)? {
-            Some(bytes) => {
-                let len = bytes.len();
-                let bytes = Arc::from(bytes);
-                let fst = fst::raw::Fst::from_shared_bytes(bytes, 0, len).unwrap();
-                Ok(Some(fst::Set::from(fst)))
-            }
-            None => Ok(None),
-        }
-    }
-
-    pub fn put_schema(self, writer: &mut heed::RwTxn, schema: &Schema) -> ZResult<()> {
-        self.main
-            .put::<Str, SerdeBincode<Schema>>(writer, SCHEMA_KEY, schema)
-    }
-
-    pub fn schema(self, reader: &heed::RoTxn) -> ZResult<Option<Schema>> {
-        self.main
-            .get::<Str, SerdeBincode<Schema>>(reader, SCHEMA_KEY)
-    }
-
-    pub fn put_ranked_map(self, writer: &mut heed::RwTxn, ranked_map: &RankedMap) -> ZResult<()> {
-        self.main
-            .put::<Str, SerdeBincode<RankedMap>>(writer, RANKED_MAP_KEY, &ranked_map)
-    }
-
-    pub fn ranked_map(self, reader: &heed::RoTxn) -> ZResult<Option<RankedMap>> {
-        self.main
-            .get::<Str, SerdeBincode<RankedMap>>(reader, RANKED_MAP_KEY)
-    }
-
-    pub fn put_synonyms_fst(self, writer: &mut heed::RwTxn, fst: &fst::Set) -> ZResult<()> {
-        let bytes = fst.as_fst().as_bytes();
-        self.main.put::<Str, ByteSlice>(writer, SYNONYMS_KEY, bytes)
-    }
-
-    pub fn synonyms_fst(self, reader: &heed::RoTxn) -> ZResult<Option<fst::Set>> {
-        match self.main.get::<Str, ByteSlice>(reader, SYNONYMS_KEY)? {
-            Some(bytes) => {
-                let len = bytes.len();
-                let bytes = Arc::from(bytes);
-                let fst = fst::raw::Fst::from_shared_bytes(bytes, 0, len).unwrap();
-                Ok(Some(fst::Set::from(fst)))
-            }
-            None => Ok(None),
-        }
-    }
-
-    pub fn put_number_of_documents<F>(self, writer: &mut heed::RwTxn, f: F) -> ZResult<u64>
-    where
-        F: Fn(u64) -> u64,
-    {
-        let new = self.number_of_documents(writer).map(f)?;
-        self.main
-            .put::<Str, OwnedType<u64>>(writer, NUMBER_OF_DOCUMENTS_KEY, &new)?;
-        Ok(new)
-    }
-
-    pub fn number_of_documents(self, reader: &heed::RoTxn) -> ZResult<u64> {
-        match self
-            .main
-            .get::<Str, OwnedType<u64>>(reader, NUMBER_OF_DOCUMENTS_KEY)?
-        {
-            Some(value) => Ok(value),
-            None => Ok(0),
-        }
-    }
-
-    pub fn put_customs(self, writer: &mut heed::RwTxn, customs: &[u8]) -> ZResult<()> {
-        self.main
-            .put::<Str, ByteSlice>(writer, CUSTOMS_KEY, customs)
-    }
-
-    pub fn customs<'txn>(self, reader: &'txn heed::RoTxn) -> ZResult<Option<&'txn [u8]>> {
-        self.main.get::<Str, ByteSlice>(reader, CUSTOMS_KEY)
-    }
-}

meilidb-core/src/store/mod.rs

@@ -1,325 +0,0 @@
-mod docs_words;
-mod documents_fields;
-mod documents_fields_counts;
-mod main;
-mod postings_lists;
-mod synonyms;
-mod updates;
-mod updates_results;
-
-pub use self::docs_words::DocsWords;
-pub use self::documents_fields::{DocumentFieldsIter, DocumentsFields};
-pub use self::documents_fields_counts::{
-    DocumentFieldsCountsIter, DocumentsFieldsCounts, DocumentsIdsIter,
-};
-pub use self::main::Main;
-pub use self::postings_lists::PostingsLists;
-pub use self::synonyms::Synonyms;
-pub use self::updates::Updates;
-pub use self::updates_results::UpdatesResults;
-
-use std::collections::HashSet;
-
-use heed::Result as ZResult;
-use meilidb_schema::{Schema, SchemaAttr};
-use serde::de;
-use zerocopy::{AsBytes, FromBytes};
-
-use crate::criterion::Criteria;
-use crate::serde::Deserializer;
-use crate::{query_builder::QueryBuilder, update, DocumentId, Error, MResult};
-
-type BEU64 = zerocopy::U64<byteorder::BigEndian>;
-type BEU16 = zerocopy::U16<byteorder::BigEndian>;
-
-#[derive(Debug, Copy, Clone, AsBytes, FromBytes)]
-#[repr(C)]
-pub struct DocumentAttrKey {
-    docid: BEU64,
-    attr: BEU16,
-}
-
-impl DocumentAttrKey {
-    fn new(docid: DocumentId, attr: SchemaAttr) -> DocumentAttrKey {
-        DocumentAttrKey {
-            docid: BEU64::new(docid.0),
-            attr: BEU16::new(attr.0),
-        }
-    }
-}
-
-fn main_name(name: &str) -> String {
-    format!("store-{}", name)
-}
-
-fn postings_lists_name(name: &str) -> String {
-    format!("store-{}-postings-lists", name)
-}
-
-fn documents_fields_name(name: &str) -> String {
-    format!("store-{}-documents-fields", name)
-}
-
-fn documents_fields_counts_name(name: &str) -> String {
-    format!("store-{}-documents-fields-counts", name)
-}
-
-fn synonyms_name(name: &str) -> String {
-    format!("store-{}-synonyms", name)
-}
-
-fn docs_words_name(name: &str) -> String {
-    format!("store-{}-docs-words", name)
-}
-
-fn updates_name(name: &str) -> String {
-    format!("store-{}-updates", name)
-}
-
-fn updates_results_name(name: &str) -> String {
-    format!("store-{}-updates-results", name)
-}
-
-#[derive(Clone)]
-pub struct Index {
-    pub main: Main,
-    pub postings_lists: PostingsLists,
-    pub documents_fields: DocumentsFields,
-    pub documents_fields_counts: DocumentsFieldsCounts,
-    pub synonyms: Synonyms,
-    pub docs_words: DocsWords,
-
-    pub updates: Updates,
-    pub updates_results: UpdatesResults,
-    updates_notifier: crossbeam_channel::Sender<()>,
-}
-
-impl Index {
-    pub fn document<T: de::DeserializeOwned>(
-        &self,
-        reader: &heed::RoTxn,
-        attributes: Option<&HashSet<&str>>,
-        document_id: DocumentId,
-    ) -> MResult<Option<T>> {
-        let schema = self.main.schema(reader)?;
-        let schema = schema.ok_or(Error::SchemaMissing)?;
-
-        let attributes = match attributes {
-            Some(attributes) => attributes
-                .iter()
-                .map(|name| schema.attribute(name))
-                .collect(),
-            None => None,
-        };
-
-        let mut deserializer = Deserializer {
-            document_id,
-            reader,
-            documents_fields: self.documents_fields,
-            schema: &schema,
-            attributes: attributes.as_ref(),
-        };
-
-        // TODO: currently we return an error if all document fields are missing,
-        //       returning None would have been better
-        Ok(T::deserialize(&mut deserializer).map(Some)?)
-    }
-
-    pub fn document_attribute<T: de::DeserializeOwned>(
-        &self,
-        reader: &heed::RoTxn,
-        document_id: DocumentId,
-        attribute: SchemaAttr,
-    ) -> MResult<Option<T>> {
-        let bytes = self
-            .documents_fields
-            .document_attribute(reader, document_id, attribute)?;
-        match bytes {
-            Some(bytes) => Ok(Some(serde_json::from_slice(bytes)?)),
-            None => Ok(None),
-        }
-    }
-
-    pub fn schema_update(&self, writer: &mut heed::RwTxn, schema: Schema) -> MResult<u64> {
-        let _ = self.updates_notifier.send(());
-        update::push_schema_update(writer, self.updates, self.updates_results, schema)
-    }
-
-    pub fn customs_update(&self, writer: &mut heed::RwTxn, customs: Vec<u8>) -> ZResult<u64> {
-        let _ = self.updates_notifier.send(());
-        update::push_customs_update(writer, self.updates, self.updates_results, customs)
-    }
-
-    pub fn documents_addition<D>(&self) -> update::DocumentsAddition<D> {
-        update::DocumentsAddition::new(
-            self.updates,
-            self.updates_results,
-            self.updates_notifier.clone(),
-        )
-    }
-
-    pub fn documents_deletion(&self) -> update::DocumentsDeletion {
-        update::DocumentsDeletion::new(
-            self.updates,
-            self.updates_results,
-            self.updates_notifier.clone(),
-        )
-    }
-
-    pub fn synonyms_addition(&self) -> update::SynonymsAddition {
-        update::SynonymsAddition::new(
-            self.updates,
-            self.updates_results,
-            self.updates_notifier.clone(),
-        )
-    }
-
-    pub fn synonyms_deletion(&self) -> update::SynonymsDeletion {
-        update::SynonymsDeletion::new(
-            self.updates,
-            self.updates_results,
-            self.updates_notifier.clone(),
-        )
-    }
-
-    pub fn current_update_id(&self, reader: &heed::RoTxn) -> MResult<Option<u64>> {
-        match self.updates.last_update_id(reader)? {
-            Some((id, _)) => Ok(Some(id)),
-            None => Ok(None),
-        }
-    }
-
-    pub fn update_status(
-        &self,
-        reader: &heed::RoTxn,
-        update_id: u64,
-    ) -> MResult<update::UpdateStatus> {
-        update::update_status(reader, self.updates, self.updates_results, update_id)
-    }
-
-    pub fn query_builder(&self) -> QueryBuilder {
-        QueryBuilder::new(
-            self.main,
-            self.postings_lists,
-            self.documents_fields_counts,
-            self.synonyms,
-        )
-    }
-
-    pub fn query_builder_with_criteria<'c, 'f, 'd>(
-        &self,
-        criteria: Criteria<'c>,
-    ) -> QueryBuilder<'c, 'f, 'd> {
-        QueryBuilder::with_criteria(
-            self.main,
-            self.postings_lists,
-            self.documents_fields_counts,
-            self.synonyms,
-            criteria,
-        )
-    }
-}
-
-pub fn create(
-    env: &heed::Env,
-    name: &str,
-    updates_notifier: crossbeam_channel::Sender<()>,
-) -> MResult<Index> {
-    // create all the store names
-    let main_name = main_name(name);
-    let postings_lists_name = postings_lists_name(name);
-    let documents_fields_name = documents_fields_name(name);
-    let documents_fields_counts_name = documents_fields_counts_name(name);
-    let synonyms_name = synonyms_name(name);
-    let docs_words_name = docs_words_name(name);
-    let updates_name = updates_name(name);
-    let updates_results_name = updates_results_name(name);
-
-    // open all the stores
-    let main = env.create_dyn_database(Some(&main_name))?;
-    let postings_lists = env.create_database(Some(&postings_lists_name))?;
-    let documents_fields = env.create_database(Some(&documents_fields_name))?;
-    let documents_fields_counts = env.create_database(Some(&documents_fields_counts_name))?;
-    let synonyms = env.create_database(Some(&synonyms_name))?;
-    let docs_words = env.create_database(Some(&docs_words_name))?;
-    let updates = env.create_database(Some(&updates_name))?;
-    let updates_results = env.create_database(Some(&updates_results_name))?;
-
-    Ok(Index {
-        main: Main { main },
-        postings_lists: PostingsLists { postings_lists },
-        documents_fields: DocumentsFields { documents_fields },
-        documents_fields_counts: DocumentsFieldsCounts {
-            documents_fields_counts,
-        },
-        synonyms: Synonyms { synonyms },
-        docs_words: DocsWords { docs_words },
-        updates: Updates { updates },
-        updates_results: UpdatesResults { updates_results },
-        updates_notifier,
-    })
-}
-
-pub fn open(
-    env: &heed::Env,
-    name: &str,
-    updates_notifier: crossbeam_channel::Sender<()>,
-) -> MResult<Option<Index>> {
-    // create all the store names
-    let main_name = main_name(name);
-    let postings_lists_name = postings_lists_name(name);
-    let documents_fields_name = documents_fields_name(name);
-    let documents_fields_counts_name = documents_fields_counts_name(name);
-    let synonyms_name = synonyms_name(name);
-    let docs_words_name = docs_words_name(name);
-    let updates_name = updates_name(name);
-    let updates_results_name = updates_results_name(name);
-
-    // open all the stores
-    let main = match env.open_dyn_database(Some(&main_name))? {
-        Some(main) => main,
-        None => return Ok(None),
-    };
-    let postings_lists = match env.open_database(Some(&postings_lists_name))? {
-        Some(postings_lists) => postings_lists,
-        None => return Ok(None),
-    };
-    let documents_fields = match env.open_database(Some(&documents_fields_name))? {
-        Some(documents_fields) => documents_fields,
-        None => return Ok(None),
-    };
-    let documents_fields_counts = match env.open_database(Some(&documents_fields_counts_name))? {
-        Some(documents_fields_counts) => documents_fields_counts,
-        None => return Ok(None),
-    };
-    let synonyms = match env.open_database(Some(&synonyms_name))? {
-        Some(synonyms) => synonyms,
-        None => return Ok(None),
-    };
-    let docs_words = match env.open_database(Some(&docs_words_name))? {
-        Some(docs_words) => docs_words,
-        None => return Ok(None),
-    };
-    let updates = match env.open_database(Some(&updates_name))? {
-        Some(updates) => updates,
-        None => return Ok(None),
-    };
-    let updates_results = match env.open_database(Some(&updates_results_name))? {
-        Some(updates_results) => updates_results,
-        None => return Ok(None),
-    };
-
-    Ok(Some(Index {
-        main: Main { main },
-        postings_lists: PostingsLists { postings_lists },
-        documents_fields: DocumentsFields { documents_fields },
-        documents_fields_counts: DocumentsFieldsCounts {
-            documents_fields_counts,
-        },
-        synonyms: Synonyms { synonyms },
-        docs_words: DocsWords { docs_words },
-        updates: Updates { updates },
-        updates_results: UpdatesResults { updates_results },
-        updates_notifier,
-    }))
-}

meilidb-core/src/store/postings_lists.rs

@@ -1,37 +0,0 @@
-use crate::DocIndex;
-use heed::types::{ByteSlice, CowSlice};
-use heed::Result as ZResult;
-use sdset::{Set, SetBuf};
-use std::borrow::Cow;
-
-#[derive(Copy, Clone)]
-pub struct PostingsLists {
-    pub(crate) postings_lists: heed::Database<ByteSlice, CowSlice<DocIndex>>,
-}
-
-impl PostingsLists {
-    pub fn put_postings_list(
-        self,
-        writer: &mut heed::RwTxn,
-        word: &[u8],
-        words_indexes: &Set<DocIndex>,
-    ) -> ZResult<()> {
-        self.postings_lists.put(writer, word, words_indexes)
-    }
-
-    pub fn del_postings_list(self, writer: &mut heed::RwTxn, word: &[u8]) -> ZResult<bool> {
-        self.postings_lists.delete(writer, word)
-    }
-
-    pub fn postings_list<'txn>(
-        self,
-        reader: &'txn heed::RoTxn,
-        word: &[u8],
-    ) -> ZResult<Option<Cow<'txn, Set<DocIndex>>>> {
-        match self.postings_lists.get(reader, word)? {
-            Some(Cow::Borrowed(slice)) => Ok(Some(Cow::Borrowed(Set::new_unchecked(slice)))),
-            Some(Cow::Owned(vec)) => Ok(Some(Cow::Owned(SetBuf::new_unchecked(vec)))),
-            None => Ok(None),
-        }
-    }
-}

meilidb-core/src/update/documents_addition.rs

@@ -1,194 +0,0 @@
-use std::collections::{HashMap, HashSet};
-
-use fst::{set::OpBuilder, SetBuilder};
-use sdset::{duo::Union, SetOperation};
-use serde::Serialize;
-
-use crate::raw_indexer::RawIndexer;
-use crate::serde::{extract_document_id, RamDocumentStore, Serializer};
-use crate::store;
-use crate::update::{apply_documents_deletion, next_update_id, Update};
-use crate::{Error, MResult, RankedMap};
-
-pub struct DocumentsAddition<D> {
-    updates_store: store::Updates,
-    updates_results_store: store::UpdatesResults,
-    updates_notifier: crossbeam_channel::Sender<()>,
-    documents: Vec<D>,
-}
-
-impl<D> DocumentsAddition<D> {
-    pub fn new(
-        updates_store: store::Updates,
-        updates_results_store: store::UpdatesResults,
-        updates_notifier: crossbeam_channel::Sender<()>,
-    ) -> DocumentsAddition<D> {
-        DocumentsAddition {
-            updates_store,
-            updates_results_store,
-            updates_notifier,
-            documents: Vec::new(),
-        }
-    }
-
-    pub fn update_document(&mut self, document: D) {
-        self.documents.push(document);
-    }
-
-    pub fn finalize(self, writer: &mut heed::RwTxn) -> MResult<u64>
-    where
-        D: serde::Serialize,
-    {
-        let _ = self.updates_notifier.send(());
-        let update_id = push_documents_addition(
-            writer,
-            self.updates_store,
-            self.updates_results_store,
-            self.documents,
-        )?;
-        Ok(update_id)
-    }
-}
-
-impl<D> Extend<D> for DocumentsAddition<D> {
-    fn extend<T: IntoIterator<Item = D>>(&mut self, iter: T) {
-        self.documents.extend(iter)
-    }
-}
-
-pub fn push_documents_addition<D: serde::Serialize>(
-    writer: &mut heed::RwTxn,
-    updates_store: store::Updates,
-    updates_results_store: store::UpdatesResults,
-    addition: Vec<D>,
-) -> MResult<u64> {
-    let mut values = Vec::with_capacity(addition.len());
-    for add in addition {
-        let vec = serde_json::to_vec(&add)?;
-        let add = serde_json::from_slice(&vec)?;
-        values.push(add);
-    }
-
-    let last_update_id = next_update_id(writer, updates_store, updates_results_store)?;
-
-    let update = Update::DocumentsAddition(values);
-    updates_store.put_update(writer, last_update_id, &update)?;
-
-    Ok(last_update_id)
-}
-
-pub fn apply_documents_addition(
-    writer: &mut heed::RwTxn,
-    main_store: store::Main,
-    documents_fields_store: store::DocumentsFields,
-    documents_fields_counts_store: store::DocumentsFieldsCounts,
-    postings_lists_store: store::PostingsLists,
-    docs_words_store: store::DocsWords,
-    mut ranked_map: RankedMap,
-    addition: Vec<serde_json::Value>,
-) -> MResult<()> {
-    let mut document_ids = HashSet::new();
-    let mut document_store = RamDocumentStore::new();
-    let mut document_fields_counts = HashMap::new();
-    let mut indexer = RawIndexer::new();
-
-    let schema = match main_store.schema(writer)? {
-        Some(schema) => schema,
-        None => return Err(Error::SchemaMissing),
-    };
-
-    let identifier = schema.identifier_name();
-
-    for document in addition {
-        let document_id = match extract_document_id(identifier, &document)? {
-            Some(id) => id,
-            None => return Err(Error::MissingDocumentId),
-        };
-
-        // 1. store the document id for future deletion
-        document_ids.insert(document_id);
-
-        // 2. index the document fields in ram stores
-        let serializer = Serializer {
-            schema: &schema,
-            document_store: &mut document_store,
-            document_fields_counts: &mut document_fields_counts,
-            indexer: &mut indexer,
-            ranked_map: &mut ranked_map,
-            document_id,
-        };
-
-        document.serialize(serializer)?;
-    }
-
-    // 1. remove the previous documents match indexes
-    let documents_to_insert = document_ids.iter().cloned().collect();
-    apply_documents_deletion(
-        writer,
-        main_store,
-        documents_fields_store,
-        documents_fields_counts_store,
-        postings_lists_store,
-        docs_words_store,
-        ranked_map.clone(),
-        documents_to_insert,
-    )?;
-
-    // 2. insert new document attributes in the database
-    for ((id, attr), value) in document_store.into_inner() {
-        documents_fields_store.put_document_field(writer, id, attr, &value)?;
-    }
-
-    // 3. insert new document attributes counts
-    for ((id, attr), count) in document_fields_counts {
-        documents_fields_counts_store.put_document_field_count(writer, id, attr, count)?;
-    }
-
-    let indexed = indexer.build();
-    let mut delta_words_builder = SetBuilder::memory();
-
-    for (word, delta_set) in indexed.words_doc_indexes {
-        delta_words_builder.insert(&word).unwrap();
-
-        let set = match postings_lists_store.postings_list(writer, &word)? {
-            Some(set) => Union::new(&set, &delta_set).into_set_buf(),
-            None => delta_set,
-        };
-
-        postings_lists_store.put_postings_list(writer, &word, &set)?;
-    }
-
-    for (id, words) in indexed.docs_words {
-        docs_words_store.put_doc_words(writer, id, &words)?;
-    }
-
-    let delta_words = delta_words_builder
-        .into_inner()
-        .and_then(fst::Set::from_bytes)
-        .unwrap();
-
-    let words = match main_store.words_fst(writer)? {
-        Some(words) => {
-            let op = OpBuilder::new()
-                .add(words.stream())
-                .add(delta_words.stream())
-                .r#union();
-
-            let mut words_builder = SetBuilder::memory();
-            words_builder.extend_stream(op).unwrap();
-            words_builder
-                .into_inner()
-                .and_then(fst::Set::from_bytes)
-                .unwrap()
-        }
-        None => delta_words,
-    };
-
-    main_store.put_words_fst(writer, &words)?;
-    main_store.put_ranked_map(writer, &ranked_map)?;
-
-    let inserted_documents_len = document_ids.len() as u64;
-    main_store.put_number_of_documents(writer, |old| old + inserted_documents_len)?;
-
-    Ok(())
-}

meilidb-core/src/update/mod.rs

@@ -1,220 +0,0 @@
-mod customs_update;
-mod documents_addition;
-mod documents_deletion;
-mod schema_update;
-mod synonyms_addition;
-mod synonyms_deletion;
-
-pub use self::customs_update::{apply_customs_update, push_customs_update};
-pub use self::documents_addition::{apply_documents_addition, DocumentsAddition};
-pub use self::documents_deletion::{apply_documents_deletion, DocumentsDeletion};
-pub use self::schema_update::{apply_schema_update, push_schema_update};
-pub use self::synonyms_addition::{apply_synonyms_addition, SynonymsAddition};
-pub use self::synonyms_deletion::{apply_synonyms_deletion, SynonymsDeletion};
-
-use std::cmp;
-use std::collections::BTreeMap;
-use std::time::{Duration, Instant};
-
-use heed::Result as ZResult;
-use log::debug;
-use serde::{Deserialize, Serialize};
-
-use crate::{store, DocumentId, MResult, RankedMap};
-use meilidb_schema::Schema;
-
-#[derive(Debug, Clone, Serialize, Deserialize)]
-pub enum Update {
-    Schema(Schema),
-    Customs(Vec<u8>),
-    DocumentsAddition(Vec<serde_json::Value>),
-    DocumentsDeletion(Vec<DocumentId>),
-    SynonymsAddition(BTreeMap<String, Vec<String>>),
-    SynonymsDeletion(BTreeMap<String, Option<Vec<String>>>),
-}
-
-#[derive(Debug, Clone, Serialize, Deserialize)]
-pub enum UpdateType {
-    Schema { schema: Schema },
-    Customs,
-    DocumentsAddition { number: usize },
-    DocumentsDeletion { number: usize },
-    SynonymsAddition { number: usize },
-    SynonymsDeletion { number: usize },
-}
-
-#[derive(Clone, Serialize, Deserialize)]
-pub struct DetailedDuration {
-    pub main: Duration,
-}
-
-#[derive(Clone, Serialize, Deserialize)]
-pub struct UpdateResult {
-    pub update_id: u64,
-    pub update_type: UpdateType,
-    pub result: Result<(), String>,
-    pub detailed_duration: DetailedDuration,
-}
-
-#[derive(Clone, Serialize, Deserialize)]
-pub enum UpdateStatus {
-    Enqueued,
-    Processed(UpdateResult),
-    Unknown,
-}
-
-pub fn update_status(
-    reader: &heed::RoTxn,
-    updates_store: store::Updates,
-    updates_results_store: store::UpdatesResults,
-    update_id: u64,
-) -> MResult<UpdateStatus> {
-    match updates_results_store.update_result(reader, update_id)? {
-        Some(result) => Ok(UpdateStatus::Processed(result)),
-        None => {
-            if updates_store.contains(reader, update_id)? {
-                Ok(UpdateStatus::Enqueued)
-            } else {
-                Ok(UpdateStatus::Unknown)
-            }
-        }
-    }
-}
-
-pub fn next_update_id(
-    writer: &mut heed::RwTxn,
-    updates_store: store::Updates,
-    updates_results_store: store::UpdatesResults,
-) -> ZResult<u64> {
-    let last_update_id = updates_store.last_update_id(writer)?;
-    let last_update_id = last_update_id.map(|(n, _)| n);
-
-    let last_update_results_id = updates_results_store.last_update_id(writer)?;
-    let last_update_results_id = last_update_results_id.map(|(n, _)| n);
-
-    let max_update_id = cmp::max(last_update_id, last_update_results_id);
-    let new_update_id = max_update_id.map_or(0, |n| n + 1);
-
-    Ok(new_update_id)
-}
-
-pub fn update_task(writer: &mut heed::RwTxn, index: store::Index) -> MResult<Option<UpdateResult>> {
-    let (update_id, update) = match index.updates.pop_front(writer)? {
-        Some(value) => value,
-        None => return Ok(None),
-    };
-
-    debug!("Processing update number {}", update_id);
-
-    let (update_type, result, duration) = match update {
-        Update::Schema(schema) => {
-            let start = Instant::now();
-
-            let update_type = UpdateType::Schema {
-                schema: schema.clone(),
-            };
-            let result = apply_schema_update(writer, index.main, &schema);
-
-            (update_type, result, start.elapsed())
-        }
-        Update::Customs(customs) => {
-            let start = Instant::now();
-
-            let update_type = UpdateType::Customs;
-            let result = apply_customs_update(writer, index.main, &customs).map_err(Into::into);
-
-            (update_type, result, start.elapsed())
-        }
-        Update::DocumentsAddition(documents) => {
-            let start = Instant::now();
-
-            let ranked_map = match index.main.ranked_map(writer)? {
-                Some(ranked_map) => ranked_map,
-                None => RankedMap::default(),
-            };
-
-            let update_type = UpdateType::DocumentsAddition {
-                number: documents.len(),
-            };
-
-            let result = apply_documents_addition(
-                writer,
-                index.main,
-                index.documents_fields,
-                index.documents_fields_counts,
-                index.postings_lists,
-                index.docs_words,
-                ranked_map,
-                documents,
-            );
-
-            (update_type, result, start.elapsed())
-        }
-        Update::DocumentsDeletion(documents) => {
-            let start = Instant::now();
-
-            let ranked_map = match index.main.ranked_map(writer)? {
-                Some(ranked_map) => ranked_map,
-                None => RankedMap::default(),
-            };
-
-            let update_type = UpdateType::DocumentsDeletion {
-                number: documents.len(),
-            };
-
-            let result = apply_documents_deletion(
-                writer,
-                index.main,
-                index.documents_fields,
-                index.documents_fields_counts,
-                index.postings_lists,
-                index.docs_words,
-                ranked_map,
-                documents,
-            );
-
-            (update_type, result, start.elapsed())
-        }
-        Update::SynonymsAddition(synonyms) => {
-            let start = Instant::now();
-
-            let update_type = UpdateType::SynonymsAddition {
-                number: synonyms.len(),
-            };
-
-            let result = apply_synonyms_addition(writer, index.main, index.synonyms, synonyms);
-
-            (update_type, result, start.elapsed())
-        }
-        Update::SynonymsDeletion(synonyms) => {
-            let start = Instant::now();
-
-            let update_type = UpdateType::SynonymsDeletion {
-                number: synonyms.len(),
-            };
-
-            let result = apply_synonyms_deletion(writer, index.main, index.synonyms, synonyms);
-
-            (update_type, result, start.elapsed())
-        }
-    };
-
-    debug!(
-        "Processed update number {} {:?} {:?}",
-        update_id, update_type, result
-    );
-
-    let detailed_duration = DetailedDuration { main: duration };
-    let status = UpdateResult {
-        update_id,
-        update_type,
-        result: result.map_err(|e| e.to_string()),
-        detailed_duration,
-    };
-
-    index
-        .updates_results
-        .put_update_result(writer, update_id, &status)?;
-
-    Ok(Some(status))
-}

meilidb-core/src/update/schema_update.rs

@@ -1,31 +0,0 @@
-use crate::update::{next_update_id, Update};
-use crate::{error::UnsupportedOperation, store, MResult};
-use meilidb_schema::Schema;
-
-pub fn apply_schema_update(
-    writer: &mut heed::RwTxn,
-    main_store: store::Main,
-    new_schema: &Schema,
-) -> MResult<()> {
-    if main_store.schema(writer)?.is_some() {
-        return Err(UnsupportedOperation::SchemaAlreadyExists.into());
-    }
-
-    main_store
-        .put_schema(writer, new_schema)
-        .map_err(Into::into)
-}
-
-pub fn push_schema_update(
-    writer: &mut heed::RwTxn,
-    updates_store: store::Updates,
-    updates_results_store: store::UpdatesResults,
-    schema: Schema,
-) -> MResult<u64> {
-    let last_update_id = next_update_id(writer, updates_store, updates_results_store)?;
-
-    let update = Update::Schema(schema);
-    updates_store.put_update(writer, last_update_id, &update)?;
-
-    Ok(last_update_id)
-}

meilidb-core/src/update/synonyms_addition.rs

@@ -1,118 +0,0 @@
-use std::collections::BTreeMap;
-
-use fst::{set::OpBuilder, SetBuilder};
-use sdset::SetBuf;
-
-use crate::automaton::normalize_str;
-use crate::update::{next_update_id, Update};
-use crate::{store, MResult};
-
-pub struct SynonymsAddition {
-    updates_store: store::Updates,
-    updates_results_store: store::UpdatesResults,
-    updates_notifier: crossbeam_channel::Sender<()>,
-    synonyms: BTreeMap<String, Vec<String>>,
-}
-
-impl SynonymsAddition {
-    pub fn new(
-        updates_store: store::Updates,
-        updates_results_store: store::UpdatesResults,
-        updates_notifier: crossbeam_channel::Sender<()>,
-    ) -> SynonymsAddition {
-        SynonymsAddition {
-            updates_store,
-            updates_results_store,
-            updates_notifier,
-            synonyms: BTreeMap::new(),
-        }
-    }
-
-    pub fn add_synonym<S, T, I>(&mut self, synonym: S, alternatives: I)
-    where
-        S: AsRef<str>,
-        T: AsRef<str>,
-        I: IntoIterator<Item = T>,
-    {
-        let synonym = normalize_str(synonym.as_ref());
-        let alternatives = alternatives.into_iter().map(|s| s.as_ref().to_lowercase());
-        self.synonyms
-            .entry(synonym)
-            .or_insert_with(Vec::new)
-            .extend(alternatives);
-    }
-
-    pub fn finalize(self, writer: &mut heed::RwTxn) -> MResult<u64> {
-        let _ = self.updates_notifier.send(());
-        let update_id = push_synonyms_addition(
-            writer,
-            self.updates_store,
-            self.updates_results_store,
-            self.synonyms,
-        )?;
-        Ok(update_id)
-    }
-}
-
-pub fn push_synonyms_addition(
-    writer: &mut heed::RwTxn,
-    updates_store: store::Updates,
-    updates_results_store: store::UpdatesResults,
-    addition: BTreeMap<String, Vec<String>>,
-) -> MResult<u64> {
-    let last_update_id = next_update_id(writer, updates_store, updates_results_store)?;
-
-    let update = Update::SynonymsAddition(addition);
-    updates_store.put_update(writer, last_update_id, &update)?;
-
-    Ok(last_update_id)
-}
-
-pub fn apply_synonyms_addition(
-    writer: &mut heed::RwTxn,
-    main_store: store::Main,
-    synonyms_store: store::Synonyms,
-    addition: BTreeMap<String, Vec<String>>,
-) -> MResult<()> {
-    let mut synonyms_builder = SetBuilder::memory();
-
-    for (word, alternatives) in addition {
-        synonyms_builder.insert(&word).unwrap();
-
-        let alternatives = {
-            let alternatives = SetBuf::from_dirty(alternatives);
-            let mut alternatives_builder = SetBuilder::memory();
-            alternatives_builder.extend_iter(alternatives).unwrap();
-            let bytes = alternatives_builder.into_inner().unwrap();
-            fst::Set::from_bytes(bytes).unwrap()
-        };
-
-        synonyms_store.put_synonyms(writer, word.as_bytes(), &alternatives)?;
-    }
-
-    let delta_synonyms = synonyms_builder
-        .into_inner()
-        .and_then(fst::Set::from_bytes)
-        .unwrap();
-
-    let synonyms = match main_store.synonyms_fst(writer)? {
-        Some(synonyms) => {
-            let op = OpBuilder::new()
-                .add(synonyms.stream())
-                .add(delta_synonyms.stream())
-                .r#union();
-
-            let mut synonyms_builder = SetBuilder::memory();
-            synonyms_builder.extend_stream(op).unwrap();
-            synonyms_builder
-                .into_inner()
-                .and_then(fst::Set::from_bytes)
-                .unwrap()
-        }
-        None => delta_synonyms,
-    };
-
-    main_store.put_synonyms_fst(writer, &synonyms)?;
-
-    Ok(())
-}

meilidb-core/src/update/synonyms_deletion.rs

@@ -1,156 +0,0 @@
-use std::collections::BTreeMap;
-use std::iter::FromIterator;
-
-use fst::{set::OpBuilder, SetBuilder};
-use sdset::SetBuf;
-
-use crate::automaton::normalize_str;
-use crate::update::{next_update_id, Update};
-use crate::{store, MResult};
-
-pub struct SynonymsDeletion {
-    updates_store: store::Updates,
-    updates_results_store: store::UpdatesResults,
-    updates_notifier: crossbeam_channel::Sender<()>,
-    synonyms: BTreeMap<String, Option<Vec<String>>>,
-}
-
-impl SynonymsDeletion {
-    pub fn new(
-        updates_store: store::Updates,
-        updates_results_store: store::UpdatesResults,
-        updates_notifier: crossbeam_channel::Sender<()>,
-    ) -> SynonymsDeletion {
-        SynonymsDeletion {
-            updates_store,
-            updates_results_store,
-            updates_notifier,
-            synonyms: BTreeMap::new(),
-        }
-    }
-
-    pub fn delete_all_alternatives_of<S: AsRef<str>>(&mut self, synonym: S) {
-        let synonym = normalize_str(synonym.as_ref());
-        self.synonyms.insert(synonym, None);
-    }
-
-    pub fn delete_specific_alternatives_of<S, T, I>(&mut self, synonym: S, alternatives: I)
-    where
-        S: AsRef<str>,
-        T: AsRef<str>,
-        I: Iterator<Item = T>,
-    {
-        let synonym = normalize_str(synonym.as_ref());
-        let value = self.synonyms.entry(synonym).or_insert(None);
-        let alternatives = alternatives.map(|s| s.as_ref().to_lowercase());
-        match value {
-            Some(v) => v.extend(alternatives),
-            None => *value = Some(Vec::from_iter(alternatives)),
-        }
-    }
-
-    pub fn finalize(self, writer: &mut heed::RwTxn) -> MResult<u64> {
-        let _ = self.updates_notifier.send(());
-        let update_id = push_synonyms_deletion(
-            writer,
-            self.updates_store,
-            self.updates_results_store,
-            self.synonyms,
-        )?;
-        Ok(update_id)
-    }
-}
-
-pub fn push_synonyms_deletion(
-    writer: &mut heed::RwTxn,
-    updates_store: store::Updates,
-    updates_results_store: store::UpdatesResults,
-    deletion: BTreeMap<String, Option<Vec<String>>>,
-) -> MResult<u64> {
-    let last_update_id = next_update_id(writer, updates_store, updates_results_store)?;
-
-    let update = Update::SynonymsDeletion(deletion);
-    updates_store.put_update(writer, last_update_id, &update)?;
-
-    Ok(last_update_id)
-}
-
-pub fn apply_synonyms_deletion(
-    writer: &mut heed::RwTxn,
-    main_store: store::Main,
-    synonyms_store: store::Synonyms,
-    deletion: BTreeMap<String, Option<Vec<String>>>,
-) -> MResult<()> {
-    let mut delete_whole_synonym_builder = SetBuilder::memory();
-
-    for (synonym, alternatives) in deletion {
-        match alternatives {
-            Some(alternatives) => {
-                let prev_alternatives = synonyms_store.synonyms(writer, synonym.as_bytes())?;
-                let prev_alternatives = match prev_alternatives {
-                    Some(alternatives) => alternatives,
-                    None => continue,
-                };
-
-                let delta_alternatives = {
-                    let alternatives = SetBuf::from_dirty(alternatives);
-                    let mut builder = SetBuilder::memory();
-                    builder.extend_iter(alternatives).unwrap();
-                    builder.into_inner().and_then(fst::Set::from_bytes).unwrap()
-                };
-
-                let op = OpBuilder::new()
-                    .add(prev_alternatives.stream())
-                    .add(delta_alternatives.stream())
-                    .difference();
-
-                let (alternatives, empty_alternatives) = {
-                    let mut builder = SetBuilder::memory();
-                    let len = builder.get_ref().len();
-                    builder.extend_stream(op).unwrap();
-                    let is_empty = len == builder.get_ref().len();
-                    let bytes = builder.into_inner().unwrap();
-                    let alternatives = fst::Set::from_bytes(bytes).unwrap();
-
-                    (alternatives, is_empty)
-                };
-
-                if empty_alternatives {
-                    delete_whole_synonym_builder.insert(synonym.as_bytes())?;
-                } else {
-                    synonyms_store.put_synonyms(writer, synonym.as_bytes(), &alternatives)?;
-                }
-            }
-            None => {
-                delete_whole_synonym_builder.insert(&synonym).unwrap();
-                synonyms_store.del_synonyms(writer, synonym.as_bytes())?;
-            }
-        }
-    }
-
-    let delta_synonyms = delete_whole_synonym_builder
-        .into_inner()
-        .and_then(fst::Set::from_bytes)
-        .unwrap();
-
-    let synonyms = match main_store.synonyms_fst(writer)? {
-        Some(synonyms) => {
-            let op = OpBuilder::new()
-                .add(synonyms.stream())
-                .add(delta_synonyms.stream())
-                .difference();
-
-            let mut synonyms_builder = SetBuilder::memory();
-            synonyms_builder.extend_stream(op).unwrap();
-            synonyms_builder
-                .into_inner()
-                .and_then(fst::Set::from_bytes)
-                .unwrap()
-        }
-        None => fst::Set::default(),
-    };
-
-    main_store.put_synonyms_fst(writer, &synonyms)?;
-
-    Ok(())
-}

meilidb-schema/src/lib.rs

@@ -1,306 +0,0 @@
-use std::collections::{BTreeMap, HashMap};
-use std::ops::BitOr;
-use std::sync::Arc;
-use std::{fmt, u16};
-
-use indexmap::IndexMap;
-use serde::{Deserialize, Serialize};
-
-pub const DISPLAYED: SchemaProps = SchemaProps {
-    displayed: true,
-    indexed: false,
-    ranked: false,
-};
-pub const INDEXED: SchemaProps = SchemaProps {
-    displayed: false,
-    indexed: true,
-    ranked: false,
-};
-pub const RANKED: SchemaProps = SchemaProps {
-    displayed: false,
-    indexed: false,
-    ranked: true,
-};
-
-#[derive(Debug, Copy, Clone, PartialEq, Eq, Serialize, Deserialize)]
-pub struct SchemaProps {
-    #[serde(default)]
-    pub displayed: bool,
-
-    #[serde(default)]
-    pub indexed: bool,
-
-    #[serde(default)]
-    pub ranked: bool,
-}
-
-impl SchemaProps {
-    pub fn is_displayed(self) -> bool {
-        self.displayed
-    }
-
-    pub fn is_indexed(self) -> bool {
-        self.indexed
-    }
-
-    pub fn is_ranked(self) -> bool {
-        self.ranked
-    }
-}
-
-impl BitOr for SchemaProps {
-    type Output = Self;
-
-    fn bitor(self, other: Self) -> Self::Output {
-        SchemaProps {
-            displayed: self.displayed | other.displayed,
-            indexed: self.indexed | other.indexed,
-            ranked: self.ranked | other.ranked,
-        }
-    }
-}
-
-#[derive(Serialize, Deserialize)]
-pub struct SchemaBuilder {
-    identifier: String,
-    attributes: IndexMap<String, SchemaProps>,
-}
-
-impl SchemaBuilder {
-    pub fn with_identifier<S: Into<String>>(name: S) -> SchemaBuilder {
-        SchemaBuilder {
-            identifier: name.into(),
-            attributes: IndexMap::new(),
-        }
-    }
-
-    pub fn new_attribute<S: Into<String>>(&mut self, name: S, props: SchemaProps) -> SchemaAttr {
-        let len = self.attributes.len();
-        if self.attributes.insert(name.into(), props).is_some() {
-            panic!("Field already inserted.")
-        }
-        SchemaAttr(len as u16)
-    }
-
-    pub fn build(self) -> Schema {
-        let mut attrs = HashMap::new();
-        let mut props = Vec::new();
-
-        for (i, (name, prop)) in self.attributes.into_iter().enumerate() {
-            attrs.insert(name.clone(), SchemaAttr(i as u16));
-            props.push((name, prop));
-        }
-
-        let identifier = self.identifier;
-        Schema {
-            inner: Arc::new(InnerSchema {
-                identifier,
-                attrs,
-                props,
-            }),
-        }
-    }
-}
-
-#[derive(Debug, Clone, PartialEq, Eq)]
-pub struct Schema {
-    inner: Arc<InnerSchema>,
-}
-
-#[derive(Debug, Clone, PartialEq, Eq)]
-struct InnerSchema {
-    identifier: String,
-    attrs: HashMap<String, SchemaAttr>,
-    props: Vec<(String, SchemaProps)>,
-}
-
-impl Schema {
-    fn to_builder(&self) -> SchemaBuilder {
-        let identifier = self.inner.identifier.clone();
-        let attributes = self.attributes_ordered();
-        SchemaBuilder {
-            identifier,
-            attributes,
-        }
-    }
-
-    fn attributes_ordered(&self) -> IndexMap<String, SchemaProps> {
-        let mut ordered = BTreeMap::new();
-        for (name, attr) in &self.inner.attrs {
-            let (_, props) = self.inner.props[attr.0 as usize];
-            ordered.insert(attr.0, (name, props));
-        }
-
-        let mut attributes = IndexMap::with_capacity(ordered.len());
-        for (_, (name, props)) in ordered {
-            attributes.insert(name.clone(), props);
-        }
-
-        attributes
-    }
-
-    pub fn props(&self, attr: SchemaAttr) -> SchemaProps {
-        let (_, props) = self.inner.props[attr.0 as usize];
-        props
-    }
-
-    pub fn identifier_name(&self) -> &str {
-        &self.inner.identifier
-    }
-
-    pub fn attribute<S: AsRef<str>>(&self, name: S) -> Option<SchemaAttr> {
-        self.inner.attrs.get(name.as_ref()).cloned()
-    }
-
-    pub fn attribute_name(&self, attr: SchemaAttr) -> &str {
-        let (name, _) = &self.inner.props[attr.0 as usize];
-        name
-    }
-
-    pub fn iter<'a>(&'a self) -> impl Iterator<Item = (&str, SchemaAttr, SchemaProps)> + 'a {
-        self.inner.props.iter().map(move |(name, prop)| {
-            let attr = self.inner.attrs.get(name).unwrap();
-            (name.as_str(), *attr, *prop)
-        })
-    }
-}
-
-impl Serialize for Schema {
-    fn serialize<S>(&self, serializer: S) -> Result<S::Ok, S::Error>
-    where
-        S: serde::ser::Serializer,
-    {
-        self.to_builder().serialize(serializer)
-    }
-}
-
-impl<'de> Deserialize<'de> for Schema {
-    fn deserialize<D>(deserializer: D) -> Result<Self, D::Error>
-    where
-        D: serde::de::Deserializer<'de>,
-    {
-        let builder = SchemaBuilder::deserialize(deserializer)?;
-        Ok(builder.build())
-    }
-}
-
-#[derive(Serialize, Deserialize, Debug, Copy, Clone, PartialOrd, Ord, PartialEq, Eq, Hash)]
-pub struct SchemaAttr(pub u16);
-
-impl SchemaAttr {
-    pub const fn new(value: u16) -> SchemaAttr {
-        SchemaAttr(value)
-    }
-
-    pub const fn min() -> SchemaAttr {
-        SchemaAttr(u16::min_value())
-    }
-
-    pub const fn max() -> SchemaAttr {
-        SchemaAttr(u16::max_value())
-    }
-
-    pub fn next(self) -> Option<SchemaAttr> {
-        self.0.checked_add(1).map(SchemaAttr)
-    }
-
-    pub fn prev(self) -> Option<SchemaAttr> {
-        self.0.checked_sub(1).map(SchemaAttr)
-    }
-}
-
-impl fmt::Display for SchemaAttr {
-    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
-        self.0.fmt(f)
-    }
-}
-
-#[cfg(test)]
-mod tests {
-    use super::*;
-    use std::error::Error;
-
-    #[test]
-    fn serialize_deserialize() -> bincode::Result<()> {
-        let mut builder = SchemaBuilder::with_identifier("id");
-        builder.new_attribute("alpha", DISPLAYED);
-        builder.new_attribute("beta", DISPLAYED | INDEXED);
-        builder.new_attribute("gamma", INDEXED);
-        let schema = builder.build();
-
-        let mut buffer = Vec::new();
-        bincode::serialize_into(&mut buffer, &schema)?;
-        let schema2 = bincode::deserialize_from(buffer.as_slice())?;
-
-        assert_eq!(schema, schema2);
-
-        Ok(())
-    }
-
-    #[test]
-    fn serialize_deserialize_toml() -> Result<(), Box<dyn Error>> {
-        let mut builder = SchemaBuilder::with_identifier("id");
-        builder.new_attribute("alpha", DISPLAYED);
-        builder.new_attribute("beta", DISPLAYED | INDEXED);
-        builder.new_attribute("gamma", INDEXED);
-        let schema = builder.build();
-
-        let buffer = toml::to_vec(&schema)?;
-        let schema2 = toml::from_slice(buffer.as_slice())?;
-
-        assert_eq!(schema, schema2);
-
-        let data = r#"
-            identifier = "id"
-
-            [attributes."alpha"]
-            displayed = true
-
-            [attributes."beta"]
-            displayed = true
-            indexed = true
-
-            [attributes."gamma"]
-            indexed = true
-        "#;
-        let schema2 = toml::from_str(data)?;
-        assert_eq!(schema, schema2);
-
-        Ok(())
-    }
-
-    #[test]
-    fn serialize_deserialize_json() -> Result<(), Box<dyn Error>> {
-        let mut builder = SchemaBuilder::with_identifier("id");
-        builder.new_attribute("alpha", DISPLAYED);
-        builder.new_attribute("beta", DISPLAYED | INDEXED);
-        builder.new_attribute("gamma", INDEXED);
-        let schema = builder.build();
-
-        let buffer = serde_json::to_vec(&schema)?;
-        let schema2 = serde_json::from_slice(buffer.as_slice())?;
-
-        assert_eq!(schema, schema2);
-
-        let data = r#"
-            {
-                "identifier": "id",
-                "attributes": {
-                    "alpha": {
-                        "displayed": true
-                    },
-                    "beta": {
-                        "displayed": true,
-                        "indexed": true
-                    },
-                    "gamma": {
-                        "indexed": true
-                    }
-                }
-            }"#;
-        let schema2 = serde_json::from_str(data)?;
-        assert_eq!(schema, schema2);
-
-        Ok(())
-    }
-}

meilisearch-core/Cargo.toml

@@ -0,0 +1,50 @@
+[package]
+name = "meilisearch-core"
+version = "0.9.0"
+license = "MIT"
+authors = ["Kerollmops <clement@meilisearch.com>"]
+edition = "2018"
+
+[dependencies]
+arc-swap = "0.4.3"
+bincode = "1.1.4"
+byteorder = "1.3.2"
+chrono = { version = "0.4.9", features = ["serde"] }
+compact_arena = "0.4.0"
+crossbeam-channel = "0.4.0"
+deunicode = "1.0.0"
+env_logger = "0.7.0"
+fst = { version = "0.3.5", default-features = false }
+hashbrown = { version = "0.6.0", features = ["serde"] }
+heed = "0.6.1"
+indexmap = { version = "1.2.0", features = ["serde-1"] }
+intervaltree = "0.2.5"
+itertools = "0.8.2"
+levenshtein_automata = { version = "0.1.1", features = ["fst_automaton"] }
+log = "0.4.8"
+meilisearch-schema = { path = "../meilisearch-schema", version = "0.9.0" }
+meilisearch-tokenizer = { path = "../meilisearch-tokenizer", version = "0.9.0" }
+meilisearch-types = { path = "../meilisearch-types", version = "0.9.0" }
+once_cell = "1.2.0"
+ordered-float = { version = "1.0.2", features = ["serde"] }
+regex = "1.3.1"
+sdset = "0.3.6"
+serde = { version = "1.0.101", features = ["derive"] }
+serde_json = "1.0.41"
+siphasher = "0.3.1"
+slice-group-by = "0.2.6"
+zerocopy = "0.2.8"
+
+[dev-dependencies]
+assert_matches = "1.3"
+criterion = "0.3"
+csv = "1.0.7"
+jemallocator = "0.3.2"
+rustyline = { version = "5.0.0", default-features = false }
+structopt = "0.3.2"
+tempfile = "3.1.0"
+termcolor = "1.0.4"
+
+[[bench]]
+name = "search_benchmark"
+harness = false

meilisearch-core/benches/search_benchmark.rs

@@ -0,0 +1,95 @@
+#[cfg(test)]
+#[macro_use]
+extern crate assert_matches;
+
+use std::sync::mpsc;
+use std::path::Path;
+use std::fs;
+use std::iter;
+
+use meilisearch_core::Database;
+use meilisearch_core::{ProcessedUpdateResult, UpdateStatus};
+use serde_json::Value;
+
+use criterion::{criterion_group, criterion_main, Criterion, BenchmarkId};
+
+fn prepare_database(path: &Path) -> Database {
+    let database = Database::open_or_create(path).unwrap();
+    let db = &database;
+
+    let (sender, receiver) = mpsc::sync_channel(100);
+    let update_fn = move |_name: &str, update: ProcessedUpdateResult| {
+        sender.send(update.update_id).unwrap()
+    };
+    let index = database.create_index("bench").unwrap();
+
+    database.set_update_callback(Box::new(update_fn));
+
+    let schema = {
+        let path = concat!(env!("CARGO_MANIFEST_DIR"), "/../datasets/movies/schema.toml");
+        let string = fs::read_to_string(path).expect("find schema");
+        toml::from_str(&string).unwrap()
+    };
+
+    let mut update_writer = db.update_write_txn().unwrap();
+    let _update_id = index.schema_update(&mut update_writer, schema).unwrap();
+    update_writer.commit().unwrap();
+
+    let mut additions = index.documents_addition();
+
+    let json: Value = {
+        let path = concat!(env!("CARGO_MANIFEST_DIR"), "/../datasets/movies/movies.json");
+        let movies_file = fs::File::open(path).expect("find movies");
+        serde_json::from_reader(movies_file).unwrap()
+    };
+
+    let documents = json.as_array().unwrap();
+
+    for document in documents {
+        additions.update_document(document);
+    }
+
+    let mut update_writer = db.update_write_txn().unwrap();
+    let update_id = additions.finalize(&mut update_writer).unwrap();
+    update_writer.commit().unwrap();
+
+    // block until the transaction is processed
+    let _ = receiver.into_iter().find(|id| *id == update_id);
+
+    let update_reader = db.update_read_txn().unwrap();
+    let result = index.update_status(&update_reader, update_id).unwrap();
+    assert_matches!(result, Some(UpdateStatus::Processed { content }) if content.error.is_none());
+
+    database
+}
+
+pub fn criterion_benchmark(c: &mut Criterion) {
+    let dir = tempfile::tempdir().unwrap();
+    let database = prepare_database(dir.path());
+
+    let reader = database.main_read_txn().unwrap();
+    let index = database.open_index("bench").unwrap();
+
+    let mut count = 0;
+    let query = "I love paris ";
+
+    let iter = iter::from_fn(|| {
+        count += 1;
+        query.get(0..count)
+    });
+
+    let mut group = c.benchmark_group("searching in movies (19654 docs)");
+    group.sample_size(10);
+
+    for query in iter {
+        let bench_name = BenchmarkId::from_parameter(format!("{:?}", query));
+        group.bench_with_input(bench_name, &query, |b, query| b.iter(|| {
+            let builder = index.query_builder();
+            builder.query(&reader, query, 0..20).unwrap();
+        }));
+    }
+    group.finish();
+}
+
+criterion_group!(benches, criterion_benchmark);
+criterion_main!(benches);

meilisearch-core/examples/from_file.rs

@@ -1,7 +1,7 @@
-use std::collections::btree_map::{BTreeMap, Entry};
 use std::collections::HashSet;
+use std::collections::btree_map::{BTreeMap, Entry};
 use std::error::Error;
-use std::io::Write;
+use std::io::{Read, Write};
 use std::iter::FromIterator;
 use std::path::{Path, PathBuf};
 use std::time::{Duration, Instant};
@@ -12,10 +12,13 @@ use serde::{Deserialize, Serialize};
 use structopt::StructOpt;
 use termcolor::{Color, ColorChoice, ColorSpec, StandardStream, WriteColor};
 
-use meilidb_core::{Database, Highlight, UpdateResult};
-use meilidb_schema::SchemaAttr;
+use meilisearch_core::{Database, Highlight, ProcessedUpdateResult};
+use meilisearch_core::settings::Settings;
+use meilisearch_schema::FieldId;
 
-const INDEX_NAME: &str = "default";
+// #[cfg(target_os = "linux")]
+#[global_allocator]
+static ALLOC: jemallocator::Jemalloc = jemallocator::Jemalloc;
 
 #[derive(Debug, StructOpt)]
 struct IndexCommand {
@@ -23,13 +26,16 @@ struct IndexCommand {
     #[structopt(parse(from_os_str))]
     database_path: PathBuf,
 
-    /// The csv file to index.
+    #[structopt(long, default_value = "default")]
+    index_uid: String,
+
+    /// The csv file path to index, you can also use `-` to specify the standard input.
     #[structopt(parse(from_os_str))]
     csv_data_path: PathBuf,
 
-    /// The path to the schema.
+    /// The path to the settings.
     #[structopt(long, parse(from_os_str))]
-    schema: PathBuf,
+    settings: PathBuf,
 
     #[structopt(long)]
     update_group_size: Option<usize>,
@@ -40,10 +46,13 @@ struct IndexCommand {
 
 #[derive(Debug, StructOpt)]
 struct SearchCommand {
-    /// The destination where the database must be created.
+    /// The path of the database to work with.
     #[structopt(parse(from_os_str))]
     database_path: PathBuf,
 
+    #[structopt(long, default_value = "default")]
+    index_uid: String,
+
     /// Timeout after which the search will return results.
     #[structopt(long)]
     fetch_timeout_ms: Option<u64>,
@@ -65,10 +74,21 @@ struct SearchCommand {
     displayed_fields: Vec<String>,
 }
 
+#[derive(Debug, StructOpt)]
+struct ShowUpdatesCommand {
+    /// The path of the database to work with.
+    #[structopt(parse(from_os_str))]
+    database_path: PathBuf,
+
+    #[structopt(long, default_value = "default")]
+    index_uid: String,
+}
+
 #[derive(Debug, StructOpt)]
 enum Command {
     Index(IndexCommand),
     Search(SearchCommand),
+    ShowUpdates(ShowUpdatesCommand),
 }
 
 impl Command {
@@ -76,6 +96,7 @@ impl Command {
         match self {
             Command::Index(command) => &command.database_path,
             Command::Search(command) => &command.database_path,
+            Command::ShowUpdates(command) => &command.database_path,
         }
     }
 }
@@ -88,37 +109,34 @@ fn index_command(command: IndexCommand, database: Database) -> Result<(), Box<dy
     let start = Instant::now();
 
     let (sender, receiver) = mpsc::sync_channel(100);
-    let update_fn = move |update: UpdateResult| sender.send(update.update_id).unwrap();
-    let index = match database.open_index(INDEX_NAME) {
+    let update_fn =
+        move |_name: &str, update: ProcessedUpdateResult| sender.send(update.update_id).unwrap();
+    let index = match database.open_index(&command.index_uid) {
         Some(index) => index,
-        None => database.create_index(INDEX_NAME).unwrap(),
+        None => database.create_index(&command.index_uid).unwrap(),
     };
 
-    let done = database.set_update_callback(INDEX_NAME, Box::new(update_fn));
-    assert!(done, "could not set the index update function");
+    database.set_update_callback(Box::new(update_fn));
 
-    let env = &database.env;
+    let db = &database;
 
-    let schema = {
-        let string = fs::read_to_string(&command.schema)?;
-        toml::from_str(&string).unwrap()
+    let settings = {
+        let string = fs::read_to_string(&command.settings)?;
+        let settings: Settings = serde_json::from_str(&string).unwrap();
+        settings.into_update().unwrap()
     };
 
-    let mut writer = env.write_txn().unwrap();
-    match index.main.schema(&writer)? {
-        Some(current_schema) => {
-            if current_schema != schema {
-                return Err(meilidb_core::Error::SchemaDiffer.into());
-            }
-            writer.abort();
-        }
-        None => {
-            index.schema_update(&mut writer, schema)?;
-            writer.commit().unwrap();
-        }
-    }
+    let mut update_writer = db.update_write_txn().unwrap();
+    index.settings_update(&mut update_writer, settings)?;
+    update_writer.commit().unwrap();
+
+    let mut rdr = if command.csv_data_path.as_os_str() == "-" {
+        csv::Reader::from_reader(Box::new(io::stdin()) as Box<dyn Read>)
+    } else {
+        let file = std::fs::File::open(command.csv_data_path)?;
+        csv::Reader::from_reader(Box::new(file) as Box<dyn Read>)
+    };
 
-    let mut rdr = csv::Reader::from_path(command.csv_data_path)?;
     let mut raw_record = csv::StringRecord::new();
     let headers = rdr.headers()?.clone();
 
@@ -157,10 +175,10 @@ fn index_command(command: IndexCommand, database: Database) -> Result<(), Box<dy
 
         println!();
 
-        let mut writer = env.write_txn().unwrap();
+        let mut update_writer = db.update_write_txn().unwrap();
         println!("committing update...");
-        let update_id = additions.finalize(&mut writer)?;
-        writer.commit().unwrap();
+        let update_id = additions.finalize(&mut update_writer)?;
+        update_writer.commit().unwrap();
         max_update_id = max_update_id.max(update_id);
         println!("committed update {}", update_id);
     }
@@ -179,8 +197,9 @@ fn index_command(command: IndexCommand, database: Database) -> Result<(), Box<dy
     );
 
     if let Some(path) = command.compact_to_path {
+        fs::create_dir_all(&path)?;
         let start = Instant::now();
-        let _file = database.copy_and_compact_to_path(&path)?;
+        let _file = database.copy_and_compact_to_path(path.join("data.mdb"))?;
         println!(
             "database compacted in {:.2?} at: {:?}",
             start.elapsed(),
@@ -201,7 +220,11 @@ fn display_highlights(text: &str, ranges: &[usize]) -> io::Result<()> {
             _ => unreachable!(),
         };
         if highlighted {
-            stdout.set_color(ColorSpec::new().set_fg(Some(Color::Yellow)))?;
+            stdout.set_color(
+                ColorSpec::new()
+                    .set_fg(Some(Color::Yellow))
+                    .set_underline(true),
+            )?;
         }
         write!(&mut stdout, "{}", &text[start..end])?;
         stdout.reset()?;
@@ -295,15 +318,16 @@ fn crop_text(
 }
 
 fn search_command(command: SearchCommand, database: Database) -> Result<(), Box<dyn Error>> {
-    let env = &database.env;
+    let db = &database;
     let index = database
-        .open_index(INDEX_NAME)
+        .open_index(&command.index_uid)
         .expect("Could not find index");
 
-    let reader = env.read_txn().unwrap();
+    let reader = db.main_read_txn().unwrap();
     let schema = index.main.schema(&reader)?;
     reader.abort();
-    let schema = schema.ok_or(meilidb_core::Error::SchemaMissing)?;
+
+    let schema = schema.ok_or(meilisearch_core::Error::SchemaMissing)?;
 
     let fields = command.displayed_fields.iter().map(String::as_str);
     let fields = HashSet::from_iter(fields);
@@ -317,7 +341,7 @@ fn search_command(command: SearchCommand, database: Database) -> Result<(), Box<
             Ok(query) => {
                 let start_total = Instant::now();
 
-                let reader = env.read_txn().unwrap();
+                let reader = db.main_read_txn().unwrap();
                 let ref_index = &index;
                 let ref_reader = &reader;
 
@@ -335,7 +359,7 @@ fn search_command(command: SearchCommand, database: Database) -> Result<(), Box<
                     };
 
                     let attr = schema
-                        .attribute(&filter)
+                        .id(filter)
                         .expect("Could not find filtered attribute");
 
                     builder.with_filter(move |document_id| {
@@ -366,11 +390,11 @@ fn search_command(command: SearchCommand, database: Database) -> Result<(), Box<
                             for (name, text) in document.0 {
                                 print!("{}: ", name);
 
-                                let attr = schema.attribute(&name).unwrap();
+                                let attr = schema.id(&name).unwrap();
                                 let highlights = doc
                                     .highlights
                                     .iter()
-                                    .filter(|m| SchemaAttr::new(m.attribute) == attr)
+                                    .filter(|m| FieldId::new(m.attribute) == attr)
                                     .cloned();
                                 let (text, highlights) =
                                     crop_text(&text, highlights, command.char_context);
@@ -385,8 +409,8 @@ fn search_command(command: SearchCommand, database: Database) -> Result<(), Box<
 
                     let mut matching_attributes = HashSet::new();
                     for highlight in doc.highlights {
-                        let attr = SchemaAttr::new(highlight.attribute);
-                        let name = schema.attribute_name(attr);
+                        let attr = FieldId::new(highlight.attribute);
+                        let name = schema.name(attr);
                         matching_attributes.insert(name);
                     }
 
@@ -418,6 +442,23 @@ fn search_command(command: SearchCommand, database: Database) -> Result<(), Box<
     Ok(())
 }
 
+fn show_updates_command(
+    command: ShowUpdatesCommand,
+    database: Database,
+) -> Result<(), Box<dyn Error>> {
+    let db = &database;
+    let index = database
+        .open_index(&command.index_uid)
+        .expect("Could not find index");
+
+    let reader = db.update_read_txn().unwrap();
+    let updates = index.all_updates_status(&reader)?;
+    println!("{:#?}", updates);
+    reader.abort();
+
+    Ok(())
+}
+
 fn main() -> Result<(), Box<dyn Error>> {
     env_logger::init();
 
@@ -427,5 +468,6 @@ fn main() -> Result<(), Box<dyn Error>> {
     match opt {
         Command::Index(command) => index_command(command, database),
         Command::Search(command) => search_command(command, database),
+        Command::ShowUpdates(command) => show_updates_command(command, database),
     }
 }

meilisearch-core/src/automaton/dfa.rs

@@ -46,3 +46,8 @@ pub fn build_prefix_dfa(query: &str) -> DFA {
 pub fn build_dfa(query: &str) -> DFA {
     build_dfa_with_setting(query, PrefixSetting::NoPrefix)
 }
+
+pub fn build_exact_dfa(query: &str) -> DFA {
+    let builder = LEVDIST0.get_or_init(|| LevBuilder::new(0, true));
+    builder.build_dfa(query)
+}

meilisearch-core/src/automaton/mod.rs

@@ -0,0 +1,15 @@
+mod dfa;
+
+use meilisearch_tokenizer::is_cjk;
+
+pub use self::dfa::{build_dfa, build_prefix_dfa, build_exact_dfa};
+
+pub fn normalize_str(string: &str) -> String {
+    let mut string = string.to_lowercase();
+
+    if !string.contains(is_cjk) {
+        string = deunicode::deunicode_with_tofu(&string, "");
+    }
+
+    string
+}

meilisearch-core/src/bucket_sort.rs

@@ -0,0 +1,560 @@
+use std::borrow::Cow;
+use std::collections::HashMap;
+use std::mem;
+use std::ops::Deref;
+use std::ops::Range;
+use std::rc::Rc;
+use std::sync::atomic::{AtomicUsize, Ordering};
+use std::time::Instant;
+use std::fmt;
+
+use compact_arena::{SmallArena, Idx32, mk_arena};
+use log::debug;
+use meilisearch_types::DocIndex;
+use sdset::{Set, SetBuf, exponential_search};
+use slice_group_by::{GroupBy, GroupByMut};
+
+use crate::error::Error;
+use crate::criterion::{Criteria, Context, ContextMut};
+use crate::distinct_map::{BufferedDistinctMap, DistinctMap};
+use crate::raw_document::RawDocument;
+use crate::{database::MainT, reordered_attrs::ReorderedAttrs};
+use crate::{store, Document, DocumentId, MResult};
+use crate::query_tree::{create_query_tree, traverse_query_tree};
+use crate::query_tree::{Operation, QueryResult, QueryKind, QueryId, PostingsKey};
+use crate::query_tree::Context as QTContext;
+
+pub fn bucket_sort<'c, FI>(
+    reader: &heed::RoTxn<MainT>,
+    query: &str,
+    range: Range<usize>,
+    filter: Option<FI>,
+    criteria: Criteria<'c>,
+    searchable_attrs: Option<ReorderedAttrs>,
+    main_store: store::Main,
+    postings_lists_store: store::PostingsLists,
+    documents_fields_counts_store: store::DocumentsFieldsCounts,
+    synonyms_store: store::Synonyms,
+    prefix_documents_cache_store: store::PrefixDocumentsCache,
+    prefix_postings_lists_cache_store: store::PrefixPostingsListsCache,
+) -> MResult<Vec<Document>>
+where
+    FI: Fn(DocumentId) -> bool,
+{
+    // We delegate the filter work to the distinct query builder,
+    // specifying a distinct rule that has no effect.
+    if filter.is_some() {
+        let distinct = |_| None;
+        let distinct_size = 1;
+        return bucket_sort_with_distinct(
+            reader,
+            query,
+            range,
+            filter,
+            distinct,
+            distinct_size,
+            criteria,
+            searchable_attrs,
+            main_store,
+            postings_lists_store,
+            documents_fields_counts_store,
+            synonyms_store,
+            prefix_documents_cache_store,
+            prefix_postings_lists_cache_store,
+        );
+    }
+
+    let words_set = match unsafe { main_store.static_words_fst(reader)? } {
+        Some(words) => words,
+        None => return Ok(Vec::new()),
+    };
+
+    let stop_words = main_store.stop_words_fst(reader)?.unwrap_or_default();
+
+    let context = QTContext {
+        words_set,
+        stop_words,
+        synonyms: synonyms_store,
+        postings_lists: postings_lists_store,
+        prefix_postings_lists: prefix_postings_lists_cache_store,
+    };
+
+    let (operation, mapping) = create_query_tree(reader, &context, query)?;
+    debug!("operation:\n{:?}", operation);
+    debug!("mapping:\n{:?}", mapping);
+
+    fn recurs_operation<'o>(map: &mut HashMap<QueryId, &'o QueryKind>, operation: &'o Operation) {
+        match operation {
+            Operation::And(ops) => ops.iter().for_each(|op| recurs_operation(map, op)),
+            Operation::Or(ops) => ops.iter().for_each(|op| recurs_operation(map, op)),
+            Operation::Query(query) => { map.insert(query.id, &query.kind); },
+        }
+    }
+
+    let mut queries_kinds = HashMap::new();
+    recurs_operation(&mut queries_kinds, &operation);
+
+    let QueryResult { docids, queries } = traverse_query_tree(reader, &context, &operation)?;
+    debug!("found {} documents", docids.len());
+    debug!("number of postings {:?}", queries.len());
+
+    let before = Instant::now();
+    mk_arena!(arena);
+    let mut bare_matches = cleanup_bare_matches(&mut arena, &docids, queries);
+    debug!("matches cleaned in {:.02?}", before.elapsed());
+
+    let before_bucket_sort = Instant::now();
+
+    let before_raw_documents_building = Instant::now();
+    let mut raw_documents = Vec::new();
+    for bare_matches in bare_matches.linear_group_by_key_mut(|sm| sm.document_id) {
+        let raw_document = RawDocument::new(bare_matches, &mut arena, searchable_attrs.as_ref());
+        raw_documents.push(raw_document);
+    }
+    debug!("creating {} candidates documents took {:.02?}",
+        raw_documents.len(),
+        before_raw_documents_building.elapsed(),
+    );
+
+    let before_criterion_loop = Instant::now();
+    let proximity_count = AtomicUsize::new(0);
+
+    let mut groups = vec![raw_documents.as_mut_slice()];
+
+    'criteria: for criterion in criteria.as_ref() {
+        let tmp_groups = mem::replace(&mut groups, Vec::new());
+        let mut documents_seen = 0;
+
+        for mut group in tmp_groups {
+            let before_criterion_preparation = Instant::now();
+
+            let ctx = ContextMut {
+                reader,
+                postings_lists: &mut arena,
+                query_mapping: &mapping,
+                documents_fields_counts_store,
+            };
+
+            criterion.prepare(ctx, &mut group)?;
+            debug!("{:?} preparation took {:.02?}", criterion.name(), before_criterion_preparation.elapsed());
+
+            let ctx = Context {
+                postings_lists: &arena,
+                query_mapping: &mapping,
+            };
+
+            let before_criterion_sort = Instant::now();
+            group.sort_unstable_by(|a, b| criterion.evaluate(&ctx, a, b));
+            debug!("{:?} evaluation took {:.02?}", criterion.name(), before_criterion_sort.elapsed());
+
+            for group in group.binary_group_by_mut(|a, b| criterion.eq(&ctx, a, b)) {
+                debug!("{:?} produced a group of size {}", criterion.name(), group.len());
+
+                documents_seen += group.len();
+                groups.push(group);
+
+                // we have sort enough documents if the last document sorted is after
+                // the end of the requested range, we can continue to the next criterion
+                if documents_seen >= range.end {
+                    continue 'criteria;
+                }
+            }
+        }
+    }
+
+    debug!("criterion loop took {:.02?}", before_criterion_loop.elapsed());
+    debug!("proximity evaluation called {} times", proximity_count.load(Ordering::Relaxed));
+
+    let schema = main_store.schema(reader)?.ok_or(Error::SchemaMissing)?;
+    let iter = raw_documents.into_iter().skip(range.start).take(range.len());
+    let iter = iter.map(|rd| Document::from_raw(rd, &queries_kinds, &arena, searchable_attrs.as_ref(), &schema));
+    let documents = iter.collect();
+
+    debug!("bucket sort took {:.02?}", before_bucket_sort.elapsed());
+
+    Ok(documents)
+}
+
+pub fn bucket_sort_with_distinct<'c, FI, FD>(
+    reader: &heed::RoTxn<MainT>,
+    query: &str,
+    range: Range<usize>,
+    filter: Option<FI>,
+    distinct: FD,
+    distinct_size: usize,
+    criteria: Criteria<'c>,
+    searchable_attrs: Option<ReorderedAttrs>,
+    main_store: store::Main,
+    postings_lists_store: store::PostingsLists,
+    documents_fields_counts_store: store::DocumentsFieldsCounts,
+    synonyms_store: store::Synonyms,
+    _prefix_documents_cache_store: store::PrefixDocumentsCache,
+    prefix_postings_lists_cache_store: store::PrefixPostingsListsCache,
+) -> MResult<Vec<Document>>
+where
+    FI: Fn(DocumentId) -> bool,
+    FD: Fn(DocumentId) -> Option<u64>,
+{
+    let words_set = match unsafe { main_store.static_words_fst(reader)? } {
+        Some(words) => words,
+        None => return Ok(Vec::new()),
+    };
+
+    let stop_words = main_store.stop_words_fst(reader)?.unwrap_or_default();
+
+    let context = QTContext {
+        words_set,
+        stop_words,
+        synonyms: synonyms_store,
+        postings_lists: postings_lists_store,
+        prefix_postings_lists: prefix_postings_lists_cache_store,
+    };
+
+    let (operation, mapping) = create_query_tree(reader, &context, query)?;
+    debug!("operation:\n{:?}", operation);
+    debug!("mapping:\n{:?}", mapping);
+
+    fn recurs_operation<'o>(map: &mut HashMap<QueryId, &'o QueryKind>, operation: &'o Operation) {
+        match operation {
+            Operation::And(ops) => ops.iter().for_each(|op| recurs_operation(map, op)),
+            Operation::Or(ops) => ops.iter().for_each(|op| recurs_operation(map, op)),
+            Operation::Query(query) => { map.insert(query.id, &query.kind); },
+        }
+    }
+
+    let mut queries_kinds = HashMap::new();
+    recurs_operation(&mut queries_kinds, &operation);
+
+    let QueryResult { docids, queries } = traverse_query_tree(reader, &context, &operation)?;
+    debug!("found {} documents", docids.len());
+    debug!("number of postings {:?}", queries.len());
+
+    let before = Instant::now();
+    mk_arena!(arena);
+    let mut bare_matches = cleanup_bare_matches(&mut arena, &docids, queries);
+    debug!("matches cleaned in {:.02?}", before.elapsed());
+
+    let before_raw_documents_building = Instant::now();
+    let mut raw_documents = Vec::new();
+    for bare_matches in bare_matches.linear_group_by_key_mut(|sm| sm.document_id) {
+        let raw_document = RawDocument::new(bare_matches, &mut arena, searchable_attrs.as_ref());
+        raw_documents.push(raw_document);
+    }
+    debug!("creating {} candidates documents took {:.02?}",
+        raw_documents.len(),
+        before_raw_documents_building.elapsed(),
+    );
+
+    let mut groups = vec![raw_documents.as_mut_slice()];
+    let mut key_cache = HashMap::new();
+
+    let mut filter_map = HashMap::new();
+    // these two variables informs on the current distinct map and
+    // on the raw offset of the start of the group where the
+    // range.start bound is located according to the distinct function
+    let mut distinct_map = DistinctMap::new(distinct_size);
+    let mut distinct_raw_offset = 0;
+
+    'criteria: for criterion in criteria.as_ref() {
+        let tmp_groups = mem::replace(&mut groups, Vec::new());
+        let mut buf_distinct = BufferedDistinctMap::new(&mut distinct_map);
+        let mut documents_seen = 0;
+
+        for mut group in tmp_groups {
+            // if this group does not overlap with the requested range,
+            // push it without sorting and splitting it
+            if documents_seen + group.len() < distinct_raw_offset {
+                documents_seen += group.len();
+                groups.push(group);
+                continue;
+            }
+
+            let ctx = ContextMut {
+                reader,
+                postings_lists: &mut arena,
+                query_mapping: &mapping,
+                documents_fields_counts_store,
+            };
+
+            let before_criterion_preparation = Instant::now();
+            criterion.prepare(ctx, &mut group)?;
+            debug!("{:?} preparation took {:.02?}", criterion.name(), before_criterion_preparation.elapsed());
+
+            let ctx = Context {
+                postings_lists: &arena,
+                query_mapping: &mapping,
+            };
+
+            let before_criterion_sort = Instant::now();
+            group.sort_unstable_by(|a, b| criterion.evaluate(&ctx, a, b));
+            debug!("{:?} evaluation took {:.02?}", criterion.name(), before_criterion_sort.elapsed());
+
+            for group in group.binary_group_by_mut(|a, b| criterion.eq(&ctx, a, b)) {
+                // we must compute the real distinguished len of this sub-group
+                for document in group.iter() {
+                    let filter_accepted = match &filter {
+                        Some(filter) => {
+                            let entry = filter_map.entry(document.id);
+                            *entry.or_insert_with(|| (filter)(document.id))
+                        }
+                        None => true,
+                    };
+
+                    if filter_accepted {
+                        let entry = key_cache.entry(document.id);
+                        let key = entry.or_insert_with(|| (distinct)(document.id).map(Rc::new));
+
+                        match key.clone() {
+                            Some(key) => buf_distinct.register(key),
+                            None => buf_distinct.register_without_key(),
+                        };
+                    }
+
+                    // the requested range end is reached: stop computing distinct
+                    if buf_distinct.len() >= range.end {
+                        break;
+                    }
+                }
+
+                documents_seen += group.len();
+                groups.push(group);
+
+                // if this sub-group does not overlap with the requested range
+                // we must update the distinct map and its start index
+                if buf_distinct.len() < range.start {
+                    buf_distinct.transfert_to_internal();
+                    distinct_raw_offset = documents_seen;
+                }
+
+                // we have sort enough documents if the last document sorted is after
+                // the end of the requested range, we can continue to the next criterion
+                if buf_distinct.len() >= range.end {
+                    continue 'criteria;
+                }
+            }
+        }
+    }
+
+    // once we classified the documents related to the current
+    // automatons we save that as the next valid result
+    let mut seen = BufferedDistinctMap::new(&mut distinct_map);
+    let schema = main_store.schema(reader)?.ok_or(Error::SchemaMissing)?;
+
+    let mut documents = Vec::with_capacity(range.len());
+    for raw_document in raw_documents.into_iter().skip(distinct_raw_offset) {
+        let filter_accepted = match &filter {
+            Some(_) => filter_map.remove(&raw_document.id).unwrap(),
+            None => true,
+        };
+
+        if filter_accepted {
+            let key = key_cache.remove(&raw_document.id).unwrap();
+            let distinct_accepted = match key {
+                Some(key) => seen.register(key),
+                None => seen.register_without_key(),
+            };
+
+            if distinct_accepted && seen.len() > range.start {
+                documents.push(Document::from_raw(raw_document, &queries_kinds, &arena, searchable_attrs.as_ref(), &schema));
+                if documents.len() == range.len() {
+                    break;
+                }
+            }
+        }
+    }
+
+    Ok(documents)
+}
+
+fn cleanup_bare_matches<'tag, 'txn>(
+    arena: &mut SmallArena<'tag, PostingsListView<'txn>>,
+    docids: &Set<DocumentId>,
+    queries: HashMap<PostingsKey, Cow<'txn, Set<DocIndex>>>,
+) -> Vec<BareMatch<'tag>>
+{
+    let docidslen = docids.len() as f32;
+    let mut bare_matches = Vec::new();
+
+    for (PostingsKey { query, input, distance, is_exact }, matches) in queries {
+        let postings_list_view = PostingsListView::original(Rc::from(input), Rc::new(matches));
+        let pllen = postings_list_view.len() as f32;
+
+        if docidslen / pllen >= 0.8 {
+            let mut offset = 0;
+            for matches in postings_list_view.linear_group_by_key(|m| m.document_id) {
+                let document_id = matches[0].document_id;
+                if docids.contains(&document_id) {
+                    let range = postings_list_view.range(offset, matches.len());
+                    let posting_list_index = arena.add(range);
+
+                    let bare_match = BareMatch {
+                        document_id,
+                        query_index: query.id,
+                        distance,
+                        is_exact,
+                        postings_list: posting_list_index,
+                    };
+
+                    bare_matches.push(bare_match);
+                }
+
+                offset += matches.len();
+            }
+
+        } else {
+            let mut offset = 0;
+            for id in docids.as_slice() {
+                let di = DocIndex { document_id: *id, ..DocIndex::default() };
+                let pos = exponential_search(&postings_list_view[offset..], &di).unwrap_or_else(|x| x);
+
+                offset += pos;
+
+                let group = postings_list_view[offset..]
+                    .linear_group_by_key(|m| m.document_id)
+                    .next()
+                    .filter(|matches| matches[0].document_id == *id);
+
+                if let Some(matches) = group {
+                    let range = postings_list_view.range(offset, matches.len());
+                    let posting_list_index = arena.add(range);
+
+                    let bare_match = BareMatch {
+                        document_id: *id,
+                        query_index: query.id,
+                        distance,
+                        is_exact,
+                        postings_list: posting_list_index,
+                    };
+
+                    bare_matches.push(bare_match);
+                }
+            }
+        }
+    }
+
+    let before_raw_documents_presort = Instant::now();
+    bare_matches.sort_unstable_by_key(|sm| sm.document_id);
+    debug!("sort by documents ids took {:.02?}", before_raw_documents_presort.elapsed());
+
+    bare_matches
+}
+
+pub struct BareMatch<'tag> {
+    pub document_id: DocumentId,
+    pub query_index: usize,
+    pub distance: u8,
+    pub is_exact: bool,
+    pub postings_list: Idx32<'tag>,
+}
+
+impl fmt::Debug for BareMatch<'_> {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        f.debug_struct("BareMatch")
+            .field("document_id", &self.document_id)
+            .field("query_index", &self.query_index)
+            .field("distance", &self.distance)
+            .field("is_exact", &self.is_exact)
+            .finish()
+    }
+}
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq, PartialOrd, Ord)]
+pub struct SimpleMatch {
+    pub query_index: usize,
+    pub distance: u8,
+    pub attribute: u16,
+    pub word_index: u16,
+    pub is_exact: bool,
+}
+
+#[derive(Clone)]
+pub enum PostingsListView<'txn> {
+    Original {
+        input: Rc<[u8]>,
+        postings_list: Rc<Cow<'txn, Set<DocIndex>>>,
+        offset: usize,
+        len: usize,
+    },
+    Rewritten {
+        input: Rc<[u8]>,
+        postings_list: SetBuf<DocIndex>,
+    },
+}
+
+impl fmt::Debug for PostingsListView<'_> {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        f.debug_struct("PostingsListView")
+            .field("input", &std::str::from_utf8(&self.input()).unwrap())
+            .field("postings_list", &self.as_ref())
+            .finish()
+    }
+}
+
+impl<'txn> PostingsListView<'txn> {
+    pub fn original(input: Rc<[u8]>, postings_list: Rc<Cow<'txn, Set<DocIndex>>>) -> PostingsListView<'txn> {
+        let len = postings_list.len();
+        PostingsListView::Original { input, postings_list, offset: 0, len }
+    }
+
+    pub fn rewritten(input: Rc<[u8]>, postings_list: SetBuf<DocIndex>) -> PostingsListView<'txn> {
+        PostingsListView::Rewritten { input, postings_list }
+    }
+
+    pub fn rewrite_with(&mut self, postings_list: SetBuf<DocIndex>) {
+        let input = match self {
+            PostingsListView::Original { input, .. } => input.clone(),
+            PostingsListView::Rewritten { input, .. } => input.clone(),
+        };
+        *self = PostingsListView::rewritten(input, postings_list);
+    }
+
+    pub fn len(&self) -> usize {
+        match self {
+            PostingsListView::Original { len, .. } => *len,
+            PostingsListView::Rewritten { postings_list, .. } => postings_list.len(),
+        }
+    }
+
+    pub fn input(&self) -> &[u8] {
+        match self {
+            PostingsListView::Original { ref input, .. } => input,
+            PostingsListView::Rewritten { ref input, .. } => input,
+        }
+    }
+
+    pub fn range(&self, range_offset: usize, range_len: usize) -> PostingsListView<'txn> {
+        match self {
+            PostingsListView::Original { input, postings_list, offset, len } => {
+                assert!(range_offset + range_len <= *len);
+                PostingsListView::Original {
+                    input: input.clone(),
+                    postings_list: postings_list.clone(),
+                    offset: offset + range_offset,
+                    len: range_len,
+                }
+            },
+            PostingsListView::Rewritten { .. } => {
+                panic!("Cannot create a range on a rewritten postings list view");
+            }
+        }
+    }
+}
+
+impl AsRef<Set<DocIndex>> for PostingsListView<'_> {
+    fn as_ref(&self) -> &Set<DocIndex> {
+        self
+    }
+}
+
+impl Deref for PostingsListView<'_> {
+    type Target = Set<DocIndex>;
+
+    fn deref(&self) -> &Set<DocIndex> {
+        match *self {
+            PostingsListView::Original { ref postings_list, offset, len, .. } => {
+                Set::new_unchecked(&postings_list[offset..offset + len])
+            },
+            PostingsListView::Rewritten { ref postings_list, .. } => postings_list,
+        }
+    }
+}

meilisearch-core/src/criterion/attribute.rs

@@ -0,0 +1,37 @@
+use std::cmp::Ordering;
+use slice_group_by::GroupBy;
+use crate::{RawDocument, MResult};
+use crate::bucket_sort::SimpleMatch;
+use super::{Criterion, Context, ContextMut, prepare_bare_matches};
+
+pub struct Attribute;
+
+impl Criterion for Attribute {
+    fn name(&self) -> &str { "attribute" }
+
+    fn prepare<'h, 'p, 'tag, 'txn, 'q, 'r>(
+        &self,
+        ctx: ContextMut<'h, 'p, 'tag, 'txn, 'q>,
+        documents: &mut [RawDocument<'r, 'tag>],
+    ) -> MResult<()>
+    {
+        prepare_bare_matches(documents, ctx.postings_lists, ctx.query_mapping);
+        Ok(())
+    }
+
+    fn evaluate(&self, _ctx: &Context, lhs: &RawDocument, rhs: &RawDocument) -> Ordering {
+        #[inline]
+        fn sum_of_attribute(matches: &[SimpleMatch]) -> usize {
+            let mut sum_of_attribute = 0;
+            for group in matches.linear_group_by_key(|bm| bm.query_index) {
+                sum_of_attribute += group[0].attribute as usize;
+            }
+            sum_of_attribute
+        }
+
+        let lhs = sum_of_attribute(&lhs.processed_matches);
+        let rhs = sum_of_attribute(&rhs.processed_matches);
+
+        lhs.cmp(&rhs)
+    }
+}

meilisearch-core/src/criterion/document_id.rs

@@ -0,0 +1,16 @@
+use std::cmp::Ordering;
+use crate::RawDocument;
+use super::{Criterion, Context};
+
+pub struct DocumentId;
+
+impl Criterion for DocumentId {
+    fn name(&self) -> &str { "stable document id" }
+
+    fn evaluate(&self, _ctx: &Context, lhs: &RawDocument, rhs: &RawDocument) -> Ordering {
+        let lhs = &lhs.id;
+        let rhs = &rhs.id;
+
+        lhs.cmp(rhs)
+    }
+}

meilisearch-core/src/criterion/exactness.rs

@@ -0,0 +1,78 @@
+use std::cmp::{Ordering, Reverse};
+use std::collections::hash_map::{HashMap, Entry};
+use meilisearch_schema::IndexedPos;
+use slice_group_by::GroupBy;
+use crate::{RawDocument, MResult};
+use crate::bucket_sort::BareMatch;
+use super::{Criterion, Context, ContextMut};
+
+pub struct Exactness;
+
+impl Criterion for Exactness {
+    fn name(&self) -> &str { "exactness" }
+
+    fn prepare<'h, 'p, 'tag, 'txn, 'q, 'r>(
+        &self,
+        ctx: ContextMut<'h, 'p, 'tag, 'txn, 'q>,
+        documents: &mut [RawDocument<'r, 'tag>],
+    ) -> MResult<()>
+    {
+        let store = ctx.documents_fields_counts_store;
+        let reader = ctx.reader;
+
+        'documents: for doc in documents {
+            doc.bare_matches.sort_unstable_by_key(|bm| (bm.query_index, Reverse(bm.is_exact)));
+
+            // mark the document if we find a "one word field" that matches
+            let mut fields_counts = HashMap::new();
+            for group in doc.bare_matches.linear_group_by_key(|bm| bm.query_index) {
+                for group in group.linear_group_by_key(|bm| bm.is_exact) {
+                    if !group[0].is_exact { break }
+
+                    for bm in group {
+                        for di in ctx.postings_lists[bm.postings_list].as_ref() {
+
+                            let attr = IndexedPos(di.attribute);
+                            let count = match fields_counts.entry(attr) {
+                                Entry::Occupied(entry) => *entry.get(),
+                                Entry::Vacant(entry) => {
+                                    let count = store.document_field_count(reader, doc.id, attr)?;
+                                    *entry.insert(count)
+                                },
+                            };
+
+                            if count == Some(1) {
+                                doc.contains_one_word_field = true;
+                                continue 'documents
+                            }
+                        }
+                    }
+                }
+            }
+        }
+
+        Ok(())
+    }
+
+    fn evaluate(&self, _ctx: &Context, lhs: &RawDocument, rhs: &RawDocument) -> Ordering {
+        #[inline]
+        fn sum_exact_query_words(matches: &[BareMatch]) -> usize {
+            let mut sum_exact_query_words = 0;
+
+            for group in matches.linear_group_by_key(|bm| bm.query_index) {
+                sum_exact_query_words += group[0].is_exact as usize;
+            }
+
+            sum_exact_query_words
+        }
+
+        // does it contains a "one word field"
+        lhs.contains_one_word_field.cmp(&rhs.contains_one_word_field).reverse()
+        // if not, with document contains the more exact words
+        .then_with(|| {
+            let lhs = sum_exact_query_words(&lhs.bare_matches);
+            let rhs = sum_exact_query_words(&rhs.bare_matches);
+            lhs.cmp(&rhs).reverse()
+        })
+    }
+}

meilisearch-core/src/criterion/mod.rs

@@ -0,0 +1,291 @@
+use std::cmp::{self, Ordering};
+use std::collections::HashMap;
+use std::ops::Range;
+
+use compact_arena::SmallArena;
+use sdset::SetBuf;
+use slice_group_by::GroupBy;
+
+use crate::bucket_sort::{SimpleMatch, PostingsListView};
+use crate::database::MainT;
+use crate::query_tree::QueryId;
+use crate::{store, RawDocument, MResult};
+
+mod typo;
+mod words;
+mod proximity;
+mod attribute;
+mod words_position;
+mod exactness;
+mod document_id;
+mod sort_by_attr;
+
+pub use self::typo::Typo;
+pub use self::words::Words;
+pub use self::proximity::Proximity;
+pub use self::attribute::Attribute;
+pub use self::words_position::WordsPosition;
+pub use self::exactness::Exactness;
+pub use self::document_id::DocumentId;
+pub use self::sort_by_attr::SortByAttr;
+
+pub trait Criterion {
+    fn name(&self) -> &str;
+
+    fn prepare<'h, 'p, 'tag, 'txn, 'q, 'r>(
+        &self,
+        _ctx: ContextMut<'h, 'p, 'tag, 'txn, 'q>,
+        _documents: &mut [RawDocument<'r, 'tag>],
+    ) -> MResult<()>
+    {
+        Ok(())
+    }
+
+    fn evaluate<'p, 'tag, 'txn, 'q, 'r>(
+        &self,
+        ctx: &Context<'p, 'tag, 'txn, 'q>,
+        lhs: &RawDocument<'r, 'tag>,
+        rhs: &RawDocument<'r, 'tag>,
+    ) -> Ordering;
+
+    #[inline]
+    fn eq<'p, 'tag, 'txn, 'q, 'r>(
+        &self,
+        ctx: &Context<'p, 'tag, 'txn, 'q>,
+        lhs: &RawDocument<'r, 'tag>,
+        rhs: &RawDocument<'r, 'tag>,
+    ) -> bool
+    {
+        self.evaluate(ctx, lhs, rhs) == Ordering::Equal
+    }
+}
+
+pub struct ContextMut<'h, 'p, 'tag, 'txn, 'q> {
+    pub reader: &'h heed::RoTxn<MainT>,
+    pub postings_lists: &'p mut SmallArena<'tag, PostingsListView<'txn>>,
+    pub query_mapping: &'q HashMap<QueryId, Range<usize>>,
+    pub documents_fields_counts_store: store::DocumentsFieldsCounts,
+}
+
+pub struct Context<'p, 'tag, 'txn, 'q> {
+    pub postings_lists: &'p SmallArena<'tag, PostingsListView<'txn>>,
+    pub query_mapping: &'q HashMap<QueryId, Range<usize>>,
+}
+
+#[derive(Default)]
+pub struct CriteriaBuilder<'a> {
+    inner: Vec<Box<dyn Criterion + 'a>>,
+}
+
+impl<'a> CriteriaBuilder<'a> {
+    pub fn new() -> CriteriaBuilder<'a> {
+        CriteriaBuilder { inner: Vec::new() }
+    }
+
+    pub fn with_capacity(capacity: usize) -> CriteriaBuilder<'a> {
+        CriteriaBuilder {
+            inner: Vec::with_capacity(capacity),
+        }
+    }
+
+    pub fn reserve(&mut self, additional: usize) {
+        self.inner.reserve(additional)
+    }
+
+    pub fn add<C: 'a>(mut self, criterion: C) -> CriteriaBuilder<'a>
+    where
+        C: Criterion,
+    {
+        self.push(criterion);
+        self
+    }
+
+    pub fn push<C: 'a>(&mut self, criterion: C)
+    where
+        C: Criterion,
+    {
+        self.inner.push(Box::new(criterion));
+    }
+
+    pub fn build(self) -> Criteria<'a> {
+        Criteria { inner: self.inner }
+    }
+}
+
+pub struct Criteria<'a> {
+    inner: Vec<Box<dyn Criterion + 'a>>,
+}
+
+impl<'a> Default for Criteria<'a> {
+    fn default() -> Self {
+        CriteriaBuilder::with_capacity(7)
+            .add(Typo)
+            .add(Words)
+            .add(Proximity)
+            .add(Attribute)
+            .add(WordsPosition)
+            .add(Exactness)
+            .add(DocumentId)
+            .build()
+    }
+}
+
+impl<'a> AsRef<[Box<dyn Criterion + 'a>]> for Criteria<'a> {
+    fn as_ref(&self) -> &[Box<dyn Criterion + 'a>] {
+        &self.inner
+    }
+}
+
+fn prepare_query_distances<'a, 'tag, 'txn>(
+    documents: &mut [RawDocument<'a, 'tag>],
+    query_mapping: &HashMap<QueryId, Range<usize>>,
+    postings_lists: &SmallArena<'tag, PostingsListView<'txn>>,
+) {
+    for document in documents {
+        if !document.processed_distances.is_empty() { continue }
+
+        let mut processed = Vec::new();
+        for m in document.bare_matches.iter() {
+            if postings_lists[m.postings_list].is_empty() { continue }
+
+            let range = query_mapping[&(m.query_index as usize)].clone();
+            let new_len = cmp::max(range.end as usize, processed.len());
+            processed.resize(new_len, None);
+
+            for index in range {
+                let index = index as usize;
+                processed[index] = match processed[index] {
+                    Some(distance) if distance > m.distance => Some(m.distance),
+                    Some(distance) => Some(distance),
+                    None => Some(m.distance),
+                };
+            }
+        }
+
+        document.processed_distances = processed;
+    }
+}
+
+fn prepare_bare_matches<'a, 'tag, 'txn>(
+    documents: &mut [RawDocument<'a, 'tag>],
+    postings_lists: &mut SmallArena<'tag, PostingsListView<'txn>>,
+    query_mapping: &HashMap<QueryId, Range<usize>>,
+) {
+    for document in documents {
+        if !document.processed_matches.is_empty() { continue }
+
+        let mut processed = Vec::new();
+        for m in document.bare_matches.iter() {
+            let postings_list = &postings_lists[m.postings_list];
+            processed.reserve(postings_list.len());
+            for di in postings_list.as_ref() {
+                let simple_match = SimpleMatch {
+                    query_index: m.query_index,
+                    distance: m.distance,
+                    attribute: di.attribute,
+                    word_index: di.word_index,
+                    is_exact: m.is_exact,
+                };
+                processed.push(simple_match);
+            }
+        }
+
+        let processed = multiword_rewrite_matches(&mut processed, query_mapping);
+        document.processed_matches = processed.into_vec();
+    }
+}
+
+fn multiword_rewrite_matches(
+    matches: &mut [SimpleMatch],
+    query_mapping: &HashMap<QueryId, Range<usize>>,
+) -> SetBuf<SimpleMatch>
+{
+    matches.sort_unstable_by_key(|m| (m.attribute, m.word_index));
+
+    let mut padded_matches = Vec::with_capacity(matches.len());
+
+    // let before_padding = Instant::now();
+    // for each attribute of each document
+    for same_document_attribute in matches.linear_group_by_key(|m| m.attribute) {
+        // padding will only be applied
+        // to word indices in the same attribute
+        let mut padding = 0;
+        let mut iter = same_document_attribute.linear_group_by_key(|m| m.word_index);
+
+        // for each match at the same position
+        // in this document attribute
+        while let Some(same_word_index) = iter.next() {
+            // find the biggest padding
+            let mut biggest = 0;
+            for match_ in same_word_index {
+                let mut replacement = query_mapping[&(match_.query_index as usize)].clone();
+                let replacement_len = replacement.len();
+                let nexts = iter.remainder().linear_group_by_key(|m| m.word_index);
+
+                if let Some(query_index) = replacement.next() {
+                    let word_index = match_.word_index + padding as u16;
+                    let match_ = SimpleMatch { query_index, word_index, ..*match_ };
+                    padded_matches.push(match_);
+                }
+
+                let mut found = false;
+
+                // look ahead and if there already is a match
+                // corresponding to this padding word, abort the padding
+                'padding: for (x, next_group) in nexts.enumerate() {
+                    for (i, query_index) in replacement.clone().enumerate().skip(x) {
+                        let word_index = match_.word_index + padding as u16 + (i + 1) as u16;
+                        let padmatch = SimpleMatch { query_index, word_index, ..*match_ };
+
+                        for nmatch_ in next_group {
+                            let mut rep = query_mapping[&(nmatch_.query_index as usize)].clone();
+                            let query_index = rep.next().unwrap();
+                            if query_index == padmatch.query_index {
+                                if !found {
+                                    // if we find a corresponding padding for the
+                                    // first time we must push preceding paddings
+                                    for (i, query_index) in replacement.clone().enumerate().take(i) {
+                                        let word_index = match_.word_index + padding as u16 + (i + 1) as u16;
+                                        let match_ = SimpleMatch { query_index, word_index, ..*match_ };
+                                        padded_matches.push(match_);
+                                        biggest = biggest.max(i + 1);
+                                    }
+                                }
+
+                                padded_matches.push(padmatch);
+                                found = true;
+                                continue 'padding;
+                            }
+                        }
+                    }
+
+                    // if we do not find a corresponding padding in the
+                    // next groups so stop here and pad what was found
+                    break;
+                }
+
+                if !found {
+                    // if no padding was found in the following matches
+                    // we must insert the entire padding
+                    for (i, query_index) in replacement.enumerate() {
+                        let word_index = match_.word_index + padding as u16 + (i + 1) as u16;
+                        let match_ = SimpleMatch { query_index, word_index, ..*match_ };
+                        padded_matches.push(match_);
+                    }
+
+                    biggest = biggest.max(replacement_len - 1);
+                }
+            }
+
+            padding += biggest;
+        }
+    }
+
+    // debug!("padding matches took {:.02?}", before_padding.elapsed());
+
+    // With this check we can see that the loop above takes something
+    // like 43% of the search time even when no rewrite is needed.
+    // assert_eq!(before_matches, padded_matches);
+
+    SetBuf::from_dirty(padded_matches)
+}

meilisearch-core/src/criterion/proximity.rs

@@ -0,0 +1,68 @@
+use std::cmp::{self, Ordering};
+use slice_group_by::GroupBy;
+use crate::bucket_sort::{SimpleMatch};
+use crate::{RawDocument, MResult};
+use super::{Criterion, Context, ContextMut, prepare_bare_matches};
+
+const MAX_DISTANCE: u16 = 8;
+
+pub struct Proximity;
+
+impl Criterion for Proximity {
+    fn name(&self) -> &str { "proximity" }
+
+    fn prepare<'h, 'p, 'tag, 'txn, 'q, 'r>(
+        &self,
+        ctx: ContextMut<'h, 'p, 'tag, 'txn, 'q>,
+        documents: &mut [RawDocument<'r, 'tag>],
+    ) -> MResult<()>
+    {
+        prepare_bare_matches(documents, ctx.postings_lists, ctx.query_mapping);
+        Ok(())
+    }
+
+    fn evaluate(&self, _ctx: &Context, lhs: &RawDocument, rhs: &RawDocument) -> Ordering {
+        fn index_proximity(lhs: u16, rhs: u16) -> u16 {
+            if lhs < rhs {
+                cmp::min(rhs - lhs, MAX_DISTANCE)
+            } else {
+                cmp::min(lhs - rhs, MAX_DISTANCE) + 1
+            }
+        }
+
+        fn attribute_proximity(lhs: SimpleMatch, rhs: SimpleMatch) -> u16 {
+            if lhs.attribute != rhs.attribute { MAX_DISTANCE }
+            else { index_proximity(lhs.word_index, rhs.word_index) }
+        }
+
+        fn min_proximity(lhs: &[SimpleMatch], rhs: &[SimpleMatch]) -> u16 {
+            let mut min_prox = u16::max_value();
+            for a in lhs {
+                for b in rhs {
+                    let prox = attribute_proximity(*a, *b);
+                    min_prox = cmp::min(min_prox, prox);
+                }
+            }
+            min_prox
+        }
+
+        fn matches_proximity(matches: &[SimpleMatch],) -> u16 {
+            let mut proximity = 0;
+            let mut iter = matches.linear_group_by_key(|m| m.query_index);
+
+            // iterate over groups by windows of size 2
+            let mut last = iter.next();
+            while let (Some(lhs), Some(rhs)) = (last, iter.next()) {
+                proximity += min_proximity(lhs, rhs);
+                last = Some(rhs);
+            }
+
+            proximity
+        }
+
+        let lhs = matches_proximity(&lhs.processed_matches);
+        let rhs = matches_proximity(&rhs.processed_matches);
+
+        lhs.cmp(&rhs)
+    }
+}

meilisearch-core/src/criterion/sort_by_attr.rs

@@ -1,10 +1,9 @@
 use std::cmp::Ordering;
 use std::error::Error;
 use std::fmt;
-
-use crate::criterion::Criterion;
+use meilisearch_schema::{Schema, FieldId};
 use crate::{RankedMap, RawDocument};
-use meilidb_schema::{Schema, SchemaAttr};
+use super::{Criterion, Context};
 
 /// An helper struct that permit to sort documents by
 /// some of their stored attributes.
@@ -23,17 +22,17 @@ use meilidb_schema::{Schema, SchemaAttr};
 ///
 /// ```ignore
 /// use serde_derive::Deserialize;
-/// use meilidb::rank::criterion::*;
+/// use meilisearch::rank::criterion::*;
 ///
 /// let custom_ranking = SortByAttr::lower_is_better(&ranked_map, &schema, "published_at")?;
 ///
 /// let builder = CriteriaBuilder::with_capacity(8)
-///        .add(SumOfTypos)
-///        .add(NumberOfWords)
-///        .add(WordsProximity)
-///        .add(SumOfWordsAttribute)
-///        .add(SumOfWordsPosition)
-///        .add(Exact)
+///        .add(Typo)
+///        .add(Words)
+///        .add(Proximity)
+///        .add(Attribute)
+///        .add(WordsPosition)
+///        .add(Exactness)
 ///        .add(custom_ranking)
 ///        .add(DocumentId);
 ///
@@ -42,7 +41,7 @@ use meilidb_schema::{Schema, SchemaAttr};
 /// ```
 pub struct SortByAttr<'a> {
     ranked_map: &'a RankedMap,
-    attr: SchemaAttr,
+    field_id: FieldId,
     reversed: bool,
 }
 
@@ -69,27 +68,31 @@ impl<'a> SortByAttr<'a> {
         attr_name: &str,
         reversed: bool,
     ) -> Result<SortByAttr<'a>, SortByAttrError> {
-        let attr = match schema.attribute(attr_name) {
-            Some(attr) => attr,
+        let field_id = match schema.id(attr_name) {
+            Some(field_id) => field_id,
             None => return Err(SortByAttrError::AttributeNotFound),
         };
 
-        if !schema.props(attr).is_ranked() {
+        if !schema.is_ranked(field_id) {
             return Err(SortByAttrError::AttributeNotRegisteredForRanking);
         }
 
         Ok(SortByAttr {
             ranked_map,
-            attr,
+            field_id,
             reversed,
         })
     }
 }
 
-impl<'a> Criterion for SortByAttr<'a> {
-    fn evaluate(&self, lhs: &RawDocument, rhs: &RawDocument) -> Ordering {
-        let lhs = self.ranked_map.get(lhs.id, self.attr);
-        let rhs = self.ranked_map.get(rhs.id, self.attr);
+impl Criterion for SortByAttr<'_> {
+    fn name(&self) -> &str {
+        "sort by attribute"
+    }
+
+    fn evaluate(&self, _ctx: &Context, lhs: &RawDocument, rhs: &RawDocument) -> Ordering {
+        let lhs = self.ranked_map.get(lhs.id, self.field_id);
+        let rhs = self.ranked_map.get(rhs.id, self.field_id);
 
         match (lhs, rhs) {
             (Some(lhs), Some(rhs)) => {
@@ -105,10 +108,6 @@ impl<'a> Criterion for SortByAttr<'a> {
             (None, None) => Ordering::Equal,
         }
     }
-
-    fn name(&self) -> &str {
-        "SortByAttr"
-    }
 }
 
 #[derive(Debug, Clone, Copy, PartialEq, Eq, Hash)]

meilisearch-core/src/criterion/typo.rs

@@ -0,0 +1,55 @@
+use std::cmp::Ordering;
+use crate::{RawDocument, MResult};
+use super::{Criterion, Context, ContextMut, prepare_query_distances};
+
+pub struct Typo;
+
+impl Criterion for Typo {
+    fn name(&self) -> &str { "typo" }
+
+    fn prepare<'h, 'p, 'tag, 'txn, 'q, 'r>(
+        &self,
+        ctx: ContextMut<'h, 'p, 'tag, 'txn, 'q>,
+        documents: &mut [RawDocument<'r, 'tag>],
+    ) -> MResult<()>
+    {
+        prepare_query_distances(documents, ctx.query_mapping, ctx.postings_lists);
+        Ok(())
+    }
+
+    fn evaluate(&self, _ctx: &Context, lhs: &RawDocument, rhs: &RawDocument) -> Ordering {
+        // This function is a wrong logarithmic 10 function.
+        // It is safe to panic on input number higher than 3,
+        // the number of typos is never bigger than that.
+        #[inline]
+        fn custom_log10(n: u8) -> f32 {
+            match n {
+                0 => 0.0,     // log(1)
+                1 => 0.30102, // log(2)
+                2 => 0.47712, // log(3)
+                3 => 0.60205, // log(4)
+                _ => panic!("invalid number"),
+            }
+        }
+
+        #[inline]
+        fn compute_typos(distances: &[Option<u8>]) -> usize {
+            let mut number_words: usize = 0;
+            let mut sum_typos = 0.0;
+
+            for distance in distances {
+                if let Some(distance) = distance {
+                    sum_typos += custom_log10(*distance);
+                    number_words += 1;
+                }
+            }
+
+            (number_words as f32 / (sum_typos + 1.0) * 1000.0) as usize
+        }
+
+        let lhs = compute_typos(&lhs.processed_distances);
+        let rhs = compute_typos(&rhs.processed_distances);
+
+        lhs.cmp(&rhs).reverse()
+    }
+}

meilisearch-core/src/criterion/words.rs

@@ -0,0 +1,31 @@
+use std::cmp::Ordering;
+use crate::{RawDocument, MResult};
+use super::{Criterion, Context, ContextMut, prepare_query_distances};
+
+pub struct Words;
+
+impl Criterion for Words {
+    fn name(&self) -> &str { "words" }
+
+    fn prepare<'h, 'p, 'tag, 'txn, 'q, 'r>(
+        &self,
+        ctx: ContextMut<'h, 'p, 'tag, 'txn, 'q>,
+        documents: &mut [RawDocument<'r, 'tag>],
+    ) -> MResult<()>
+    {
+        prepare_query_distances(documents, ctx.query_mapping, ctx.postings_lists);
+        Ok(())
+    }
+
+    fn evaluate(&self, _ctx: &Context, lhs: &RawDocument, rhs: &RawDocument) -> Ordering {
+        #[inline]
+        fn number_of_query_words(distances: &[Option<u8>]) -> usize {
+            distances.iter().cloned().filter(Option::is_some).count()
+        }
+
+        let lhs = number_of_query_words(&lhs.processed_distances);
+        let rhs = number_of_query_words(&rhs.processed_distances);
+
+        lhs.cmp(&rhs).reverse()
+    }
+}

meilisearch-core/src/criterion/words_position.rs

@@ -0,0 +1,37 @@
+use std::cmp::Ordering;
+use slice_group_by::GroupBy;
+use crate::bucket_sort::SimpleMatch;
+use crate::{RawDocument, MResult};
+use super::{Criterion, Context, ContextMut, prepare_bare_matches};
+
+pub struct WordsPosition;
+
+impl Criterion for WordsPosition {
+    fn name(&self) -> &str { "words position" }
+
+    fn prepare<'h, 'p, 'tag, 'txn, 'q, 'r>(
+        &self,
+        ctx: ContextMut<'h, 'p, 'tag, 'txn, 'q>,
+        documents: &mut [RawDocument<'r, 'tag>],
+    ) -> MResult<()>
+    {
+        prepare_bare_matches(documents, ctx.postings_lists, ctx.query_mapping);
+        Ok(())
+    }
+
+    fn evaluate(&self, _ctx: &Context, lhs: &RawDocument, rhs: &RawDocument) -> Ordering {
+        #[inline]
+        fn sum_words_position(matches: &[SimpleMatch]) -> usize {
+            let mut sum_words_position = 0;
+            for group in matches.linear_group_by_key(|bm| bm.query_index) {
+                sum_words_position += group[0].word_index as usize;
+            }
+            sum_words_position
+        }
+
+        let lhs = sum_words_position(&lhs.processed_matches);
+        let rhs = sum_words_position(&rhs.processed_matches);
+
+        lhs.cmp(&rhs)
+    }
+}

meilisearch-core/src/error.rs

@@ -2,16 +2,22 @@ use crate::serde::{DeserializerError, SerializerError};
 use serde_json::Error as SerdeJsonError;
 use std::{error, fmt, io};
 
+pub use heed::Error as HeedError;
+pub use fst::Error as FstError;
+pub use bincode::Error as BincodeError;
+
 pub type MResult<T> = Result<T, Error>;
 
 #[derive(Debug)]
 pub enum Error {
     Io(io::Error),
     IndexAlreadyExists,
-    SchemaDiffer,
+    MissingPrimaryKey,
     SchemaMissing,
     WordIndexMissing,
     MissingDocumentId,
+    MaxFieldsLimitExceeded,
+    Schema(meilisearch_schema::Error),
     Zlmdb(heed::Error),
     Fst(fst::Error),
     SerdeJson(SerdeJsonError),
@@ -27,14 +33,20 @@ impl From<io::Error> for Error {
     }
 }
 
-impl From<heed::Error> for Error {
-    fn from(error: heed::Error) -> Error {
+impl From<meilisearch_schema::Error> for Error {
+    fn from(error: meilisearch_schema::Error) -> Error {
+        Error::Schema(error)
+    }
+}
+
+impl From<HeedError> for Error {
+    fn from(error: HeedError) -> Error {
         Error::Zlmdb(error)
     }
 }
 
-impl From<fst::Error> for Error {
-    fn from(error: fst::Error) -> Error {
+impl From<FstError> for Error {
+    fn from(error: FstError) -> Error {
         Error::Fst(error)
     }
 }
@@ -45,8 +57,8 @@ impl From<SerdeJsonError> for Error {
     }
 }
 
-impl From<bincode::Error> for Error {
-    fn from(error: bincode::Error) -> Error {
+impl From<BincodeError> for Error {
+    fn from(error: BincodeError) -> Error {
         Error::Bincode(error)
     }
 }
@@ -75,10 +87,12 @@ impl fmt::Display for Error {
         match self {
             Io(e) => write!(f, "{}", e),
             IndexAlreadyExists => write!(f, "index already exists"),
-            SchemaDiffer => write!(f, "schemas differ"),
+            MissingPrimaryKey => write!(f, "schema cannot be built without a primary key"),
             SchemaMissing => write!(f, "this index does not have a schema"),
             WordIndexMissing => write!(f, "this index does not have a word index"),
             MissingDocumentId => write!(f, "document id is missing"),
+            MaxFieldsLimitExceeded => write!(f, "maximum number of fields in a document exceeded"),
+            Schema(e) => write!(f, "schema error; {}", e),
             Zlmdb(e) => write!(f, "heed error; {}", e),
             Fst(e) => write!(f, "fst error; {}", e),
             SerdeJson(e) => write!(f, "serde json error; {}", e),
@@ -95,6 +109,10 @@ impl error::Error for Error {}
 #[derive(Debug)]
 pub enum UnsupportedOperation {
     SchemaAlreadyExists,
+    CannotUpdateSchemaPrimaryKey,
+    CannotReorderSchemaAttribute,
+    CanOnlyIntroduceNewSchemaAttributesAtEnd,
+    CannotRemoveSchemaAttribute,
 }
 
 impl fmt::Display for UnsupportedOperation {
@@ -102,6 +120,12 @@ impl fmt::Display for UnsupportedOperation {
         use self::UnsupportedOperation::*;
         match self {
             SchemaAlreadyExists => write!(f, "Cannot update index which already have a schema"),
+            CannotUpdateSchemaPrimaryKey => write!(f, "Cannot update the primary key of a schema"),
+            CannotReorderSchemaAttribute => write!(f, "Cannot reorder the attributes of a schema"),
+            CanOnlyIntroduceNewSchemaAttributesAtEnd => {
+                write!(f, "Can only introduce new attributes at end of a schema")
+            }
+            CannotRemoveSchemaAttribute => write!(f, "Cannot remove attributes from a schema"),
         }
     }
 }

meilisearch-core/src/levenshtein.rs

@@ -0,0 +1,134 @@
+use std::cmp::min;
+use std::collections::BTreeMap;
+use std::ops::{Index, IndexMut};
+
+// A simple wrapper around vec so we can get contiguous but index it like it's 2D array.
+struct N2Array<T> {
+    y_size: usize,
+    buf: Vec<T>,
+}
+
+impl<T: Clone> N2Array<T> {
+    fn new(x: usize, y: usize, value: T) -> N2Array<T> {
+        N2Array {
+            y_size: y,
+            buf: vec![value; x * y],
+        }
+    }
+}
+
+impl<T> Index<(usize, usize)> for N2Array<T> {
+    type Output = T;
+
+    #[inline]
+    fn index(&self, (x, y): (usize, usize)) -> &T {
+        &self.buf[(x * self.y_size) + y]
+    }
+}
+
+impl<T> IndexMut<(usize, usize)> for N2Array<T> {
+    #[inline]
+    fn index_mut(&mut self, (x, y): (usize, usize)) -> &mut T {
+        &mut self.buf[(x * self.y_size) + y]
+    }
+}
+
+pub fn prefix_damerau_levenshtein(source: &[u8], target: &[u8]) -> (u32, usize) {
+    let (n, m) = (source.len(), target.len());
+
+    assert!(
+        n <= m,
+        "the source string must be shorter than the target one"
+    );
+
+    if n == 0 {
+        return (m as u32, 0);
+    }
+    if m == 0 {
+        return (n as u32, 0);
+    }
+
+    if n == m && source == target {
+        return (0, m);
+    }
+
+    let inf = n + m;
+    let mut matrix = N2Array::new(n + 2, m + 2, 0);
+
+    matrix[(0, 0)] = inf;
+    for i in 0..n + 1 {
+        matrix[(i + 1, 0)] = inf;
+        matrix[(i + 1, 1)] = i;
+    }
+    for j in 0..m + 1 {
+        matrix[(0, j + 1)] = inf;
+        matrix[(1, j + 1)] = j;
+    }
+
+    let mut last_row = BTreeMap::new();
+
+    for (row, char_s) in source.iter().enumerate() {
+        let mut last_match_col = 0;
+        let row = row + 1;
+
+        for (col, char_t) in target.iter().enumerate() {
+            let col = col + 1;
+            let last_match_row = *last_row.get(&char_t).unwrap_or(&0);
+            let cost = if char_s == char_t { 0 } else { 1 };
+
+            let dist_add = matrix[(row, col + 1)] + 1;
+            let dist_del = matrix[(row + 1, col)] + 1;
+            let dist_sub = matrix[(row, col)] + cost;
+            let dist_trans = matrix[(last_match_row, last_match_col)]
+                + (row - last_match_row - 1)
+                + 1
+                + (col - last_match_col - 1);
+
+            let dist = min(min(dist_add, dist_del), min(dist_sub, dist_trans));
+
+            matrix[(row + 1, col + 1)] = dist;
+
+            if cost == 0 {
+                last_match_col = col;
+            }
+        }
+
+        last_row.insert(char_s, row);
+    }
+
+    let mut minimum = (u32::max_value(), 0);
+
+    for x in n..=m {
+        let dist = matrix[(n + 1, x + 1)] as u32;
+        if dist < minimum.0 {
+            minimum = (dist, x)
+        }
+    }
+
+    minimum
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn matched_length() {
+        let query = "Levenste";
+        let text = "Levenshtein";
+
+        let (dist, length) = prefix_damerau_levenshtein(query.as_bytes(), text.as_bytes());
+        assert_eq!(dist, 1);
+        assert_eq!(&text[..length], "Levenshte");
+    }
+
+    #[test]
+    #[should_panic]
+    fn matched_length_panic() {
+        let query = "Levenshtein";
+        let text = "Levenste";
+
+        // this function will panic if source if longer than target
+        prefix_damerau_levenshtein(query.as_bytes(), text.as_bytes());
+    }
+}

meilisearch-core/src/lib.rs

@@ -0,0 +1,191 @@
+#[cfg(test)]
+#[macro_use]
+extern crate assert_matches;
+
+mod automaton;
+mod bucket_sort;
+mod database;
+mod distinct_map;
+mod error;
+mod levenshtein;
+mod number;
+mod query_builder;
+mod query_tree;
+mod query_words_mapper;
+mod ranked_map;
+mod raw_document;
+mod reordered_attrs;
+mod update;
+pub mod settings;
+pub mod criterion;
+pub mod raw_indexer;
+pub mod serde;
+pub mod store;
+
+pub use self::database::{BoxUpdateFn, Database, MainT, UpdateT};
+pub use self::error::{Error, HeedError, FstError, MResult};
+pub use self::number::{Number, ParseNumberError};
+pub use self::ranked_map::RankedMap;
+pub use self::raw_document::RawDocument;
+pub use self::store::Index;
+pub use self::update::{EnqueuedUpdateResult, ProcessedUpdateResult, UpdateStatus, UpdateType};
+pub use meilisearch_types::{DocIndex, DocumentId, Highlight};
+pub use meilisearch_schema::Schema;
+pub use query_words_mapper::QueryWordsMapper;
+
+use std::convert::TryFrom;
+use std::collections::HashMap;
+use compact_arena::SmallArena;
+use log::{error, trace};
+
+use crate::bucket_sort::PostingsListView;
+use crate::levenshtein::prefix_damerau_levenshtein;
+use crate::query_tree::{QueryId, QueryKind};
+use crate::reordered_attrs::ReorderedAttrs;
+
+#[derive(Debug, Clone, PartialEq, Eq, PartialOrd, Ord)]
+pub struct Document {
+    pub id: DocumentId,
+    pub highlights: Vec<Highlight>,
+
+    #[cfg(test)]
+    pub matches: Vec<crate::bucket_sort::SimpleMatch>,
+}
+
+fn highlights_from_raw_document<'a, 'tag, 'txn>(
+    raw_document: &RawDocument<'a, 'tag>,
+    queries_kinds: &HashMap<QueryId, &QueryKind>,
+    arena: &SmallArena<'tag, PostingsListView<'txn>>,
+    searchable_attrs: Option<&ReorderedAttrs>,
+    schema: &Schema,
+) -> Vec<Highlight>
+{
+    let mut highlights = Vec::new();
+
+    for bm in raw_document.bare_matches.iter() {
+        let postings_list = &arena[bm.postings_list];
+        let input = postings_list.input();
+        let kind = &queries_kinds.get(&bm.query_index);
+
+        for di in postings_list.iter() {
+            let covered_area = match kind {
+                Some(QueryKind::NonTolerant(query)) | Some(QueryKind::Tolerant(query)) => {
+                    let len = if query.len() > input.len() {
+                        input.len()
+                    } else {
+                        prefix_damerau_levenshtein(query.as_bytes(), input).1
+                    };
+                    u16::try_from(len).unwrap_or(u16::max_value())
+                },
+                _ => di.char_length,
+            };
+
+            let attribute = searchable_attrs
+                .and_then(|sa| sa.reverse(di.attribute))
+                .unwrap_or(di.attribute);
+
+            let attribute = match schema.indexed_pos_to_field_id(attribute) {
+                Some(field_id) => field_id.0,
+                None => {
+                    error!("Cannot convert indexed_pos {} to field_id", attribute);
+                    trace!("Schema is compromized; {:?}", schema);
+                    continue
+                }
+            };
+
+            let highlight = Highlight {
+                attribute,
+                char_index: di.char_index,
+                char_length: covered_area,
+            };
+
+            highlights.push(highlight);
+        }
+    }
+
+    highlights
+}
+
+impl Document {
+    #[cfg(not(test))]
+    pub fn from_highlights(id: DocumentId, highlights: &[Highlight]) -> Document {
+        Document { id, highlights: highlights.to_owned() }
+    }
+
+    #[cfg(test)]
+    pub fn from_highlights(id: DocumentId, highlights: &[Highlight]) -> Document {
+        Document { id, highlights: highlights.to_owned(), matches: Vec::new() }
+    }
+
+    #[cfg(not(test))]
+    pub fn from_raw<'a, 'tag, 'txn>(
+        raw_document: RawDocument<'a, 'tag>,
+        queries_kinds: &HashMap<QueryId, &QueryKind>,
+        arena: &SmallArena<'tag, PostingsListView<'txn>>,
+        searchable_attrs: Option<&ReorderedAttrs>,
+        schema: &Schema,
+    ) -> Document
+    {
+        let highlights = highlights_from_raw_document(
+            &raw_document,
+            queries_kinds,
+            arena,
+            searchable_attrs,
+            schema,
+        );
+
+        Document { id: raw_document.id, highlights }
+    }
+
+    #[cfg(test)]
+    pub fn from_raw<'a, 'tag, 'txn>(
+        raw_document: RawDocument<'a, 'tag>,
+        queries_kinds: &HashMap<QueryId, &QueryKind>,
+        arena: &SmallArena<'tag, PostingsListView<'txn>>,
+        searchable_attrs: Option<&ReorderedAttrs>,
+        schema: &Schema,
+    ) -> Document
+    {
+        use crate::bucket_sort::SimpleMatch;
+
+        let highlights = highlights_from_raw_document(
+            &raw_document,
+            queries_kinds,
+            arena,
+            searchable_attrs,
+            schema,
+        );
+
+        let mut matches = Vec::new();
+        for sm in raw_document.processed_matches {
+            let attribute = searchable_attrs
+                .and_then(|sa| sa.reverse(sm.attribute))
+                .unwrap_or(sm.attribute);
+
+            let attribute = match schema.indexed_pos_to_field_id(attribute) {
+                Some(field_id) => field_id.0,
+                None => {
+                    error!("Cannot convert indexed_pos {} to field_id", attribute);
+                    trace!("Schema is compromized; {:?}", schema);
+                    continue
+                }
+            };
+
+            matches.push(SimpleMatch { attribute, ..sm });
+        }
+        matches.sort_unstable();
+
+        Document { id: raw_document.id, highlights, matches }
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use std::mem;
+
+    #[test]
+    fn docindex_mem_size() {
+        assert_eq!(mem::size_of::<DocIndex>(), 16);
+    }
+}

meilisearch-core/src/number.rs

@@ -1,3 +1,4 @@
+use std::cmp::Ordering;
 use std::fmt;
 use std::num::{ParseFloatError, ParseIntError};
 use std::str::FromStr;
@@ -5,11 +6,18 @@ use std::str::FromStr;
 use ordered_float::OrderedFloat;
 use serde::{Deserialize, Serialize};
 
-#[derive(Serialize, Deserialize, Debug, Copy, Clone, PartialEq, Eq, PartialOrd, Ord, Hash)]
+#[derive(Serialize, Deserialize, Debug, Copy, Clone, Hash)]
 pub enum Number {
     Unsigned(u64),
     Signed(i64),
     Float(OrderedFloat<f64>),
+    Null,
+}
+
+impl Default for Number {
+    fn default() -> Self {
+        Self::Null
+    }
 }
 
 impl FromStr for Number {
@@ -39,6 +47,53 @@ impl FromStr for Number {
     }
 }
 
+impl PartialEq for Number {
+    fn eq(&self, other: &Number) -> bool {
+        self.cmp(other) == Ordering::Equal
+    }
+}
+
+impl Eq for Number {}
+
+impl PartialOrd for Number {
+    fn partial_cmp(&self, other: &Number) -> Option<Ordering> {
+        Some(self.cmp(other))
+    }
+}
+
+impl Ord for Number {
+    fn cmp(&self, other: &Self) -> Ordering {
+        use Number::{Float, Signed, Unsigned, Null};
+
+        match (*self, *other) {
+            (Unsigned(a), Unsigned(b)) => a.cmp(&b),
+            (Unsigned(a), Signed(b)) => {
+                if b < 0 {
+                    Ordering::Greater
+                } else {
+                    a.cmp(&(b as u64))
+                }
+            }
+            (Unsigned(a), Float(b)) => (OrderedFloat(a as f64)).cmp(&b),
+            (Signed(a), Unsigned(b)) => {
+                if a < 0 {
+                    Ordering::Less
+                } else {
+                    (a as u64).cmp(&b)
+                }
+            }
+            (Signed(a), Signed(b)) => a.cmp(&b),
+            (Signed(a), Float(b)) => OrderedFloat(a as f64).cmp(&b),
+            (Float(a), Unsigned(b)) => a.cmp(&OrderedFloat(b as f64)),
+            (Float(a), Signed(b)) => a.cmp(&OrderedFloat(b as f64)),
+            (Float(a), Float(b)) => a.cmp(&b),
+            (Null, Null) => Ordering::Equal,
+            (_, Null) => Ordering::Less,
+            (Null, _) => Ordering::Greater,
+        }
+    }
+}
+
 #[derive(Debug, Clone, PartialEq, Eq)]
 pub struct ParseNumberError {
     uint_error: ParseIntError,

meilisearch-core/src/query_tree.rs

@@ -0,0 +1,560 @@
+use std::borrow::Cow;
+use std::collections::HashMap;
+use std::hash::{Hash, Hasher};
+use std::ops::Range;
+use std::time::Instant;
+use std::{cmp, fmt, iter::once};
+
+use fst::{IntoStreamer, Streamer};
+use itertools::{EitherOrBoth, merge_join_by};
+use meilisearch_tokenizer::split_query_string;
+use sdset::{Set, SetBuf, SetOperation};
+use log::debug;
+
+use crate::database::MainT;
+use crate::{store, DocumentId, DocIndex, MResult};
+use crate::automaton::{normalize_str, build_dfa, build_prefix_dfa, build_exact_dfa};
+use crate::QueryWordsMapper;
+
+#[derive(Clone, PartialEq, Eq, Hash)]
+pub enum Operation {
+    And(Vec<Operation>),
+    Or(Vec<Operation>),
+    Query(Query),
+}
+
+impl fmt::Debug for Operation {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        fn pprint_tree(f: &mut fmt::Formatter<'_>, op: &Operation, depth: usize) -> fmt::Result {
+            match op {
+                Operation::And(children) => {
+                    writeln!(f, "{:1$}AND", "", depth * 2)?;
+                    children.iter().try_for_each(|c| pprint_tree(f, c, depth + 1))
+                },
+                Operation::Or(children) => {
+                    writeln!(f, "{:1$}OR", "", depth * 2)?;
+                    children.iter().try_for_each(|c| pprint_tree(f, c, depth + 1))
+                },
+                Operation::Query(query) => writeln!(f, "{:2$}{:?}", "", query, depth * 2),
+            }
+        }
+
+        pprint_tree(f, self, 0)
+    }
+}
+
+impl Operation {
+    fn tolerant(id: QueryId, prefix: bool, s: &str) -> Operation {
+        Operation::Query(Query { id, prefix, exact: true, kind: QueryKind::Tolerant(s.to_string()) })
+    }
+
+    fn non_tolerant(id: QueryId, prefix: bool, s: &str) -> Operation {
+        Operation::Query(Query { id, prefix, exact: true, kind: QueryKind::NonTolerant(s.to_string()) })
+    }
+
+    fn phrase2(id: QueryId, prefix: bool, (left, right): (&str, &str)) -> Operation {
+        let kind = QueryKind::Phrase(vec![left.to_owned(), right.to_owned()]);
+        Operation::Query(Query { id, prefix, exact: true, kind })
+    }
+}
+
+pub type QueryId = usize;
+
+#[derive(Clone, Eq)]
+pub struct Query {
+    pub id: QueryId,
+    pub prefix: bool,
+    pub exact: bool,
+    pub kind: QueryKind,
+}
+
+impl PartialEq for Query {
+    fn eq(&self, other: &Self) -> bool {
+        self.prefix == other.prefix && self.kind == other.kind
+    }
+}
+
+impl Hash for Query {
+    fn hash<H: Hasher>(&self, state: &mut H) {
+        self.prefix.hash(state);
+        self.kind.hash(state);
+    }
+}
+
+#[derive(Clone, PartialEq, Eq, Hash)]
+pub enum QueryKind {
+    Tolerant(String),
+    NonTolerant(String),
+    Phrase(Vec<String>),
+}
+
+impl fmt::Debug for Query {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        let Query { id, prefix, kind, .. } = self;
+        let prefix = if *prefix { String::from("Prefix") } else { String::default() };
+        match kind {
+            QueryKind::NonTolerant(word) => {
+                f.debug_struct(&(prefix + "NonTolerant")).field("id", &id).field("word", &word).finish()
+            },
+            QueryKind::Tolerant(word) => {
+                f.debug_struct(&(prefix + "Tolerant")).field("id", &id).field("word", &word).finish()
+            },
+            QueryKind::Phrase(words) => {
+                f.debug_struct(&(prefix + "Phrase")).field("id", &id).field("words", &words).finish()
+            },
+        }
+    }
+}
+
+#[derive(Debug, Default)]
+pub struct PostingsList {
+    docids: SetBuf<DocumentId>,
+    matches: SetBuf<DocIndex>,
+}
+
+pub struct Context {
+    pub words_set: fst::Set,
+    pub stop_words: fst::Set,
+    pub synonyms: store::Synonyms,
+    pub postings_lists: store::PostingsLists,
+    pub prefix_postings_lists: store::PrefixPostingsListsCache,
+}
+
+fn split_best_frequency<'a>(reader: &heed::RoTxn<MainT>, ctx: &Context, word: &'a str) -> MResult<Option<(&'a str, &'a str)>> {
+    let chars = word.char_indices().skip(1);
+    let mut best = None;
+
+    for (i, _) in chars {
+        let (left, right) = word.split_at(i);
+
+        let left_freq = ctx.postings_lists
+            .postings_list(reader, left.as_bytes())?
+            .map(|p| p.docids.len())
+            .unwrap_or(0);
+        let right_freq = ctx.postings_lists
+            .postings_list(reader, right.as_bytes())?
+            .map(|p| p.docids.len())
+            .unwrap_or(0);
+
+        let min_freq = cmp::min(left_freq, right_freq);
+        if min_freq != 0 && best.map_or(true, |(old, _, _)| min_freq > old) {
+            best = Some((min_freq, left, right));
+        }
+    }
+
+    Ok(best.map(|(_, l, r)| (l, r)))
+}
+
+fn fetch_synonyms(reader: &heed::RoTxn<MainT>, ctx: &Context, words: &[&str]) -> MResult<Vec<Vec<String>>> {
+    let words = normalize_str(&words.join(" "));
+    let set = ctx.synonyms.synonyms(reader, words.as_bytes())?.unwrap_or_default();
+
+    let mut strings = Vec::new();
+    let mut stream = set.stream();
+    while let Some(input) = stream.next() {
+        if let Ok(input) = std::str::from_utf8(input) {
+            let alts = input.split_ascii_whitespace().map(ToOwned::to_owned).collect();
+            strings.push(alts);
+        }
+    }
+
+    Ok(strings)
+}
+
+fn create_operation<I, F>(iter: I, f: F) -> Operation
+where I: IntoIterator<Item=Operation>,
+      F: Fn(Vec<Operation>) -> Operation,
+{
+    let mut iter = iter.into_iter();
+    match (iter.next(), iter.next()) {
+        (Some(first), None) => first,
+        (first, second) => f(first.into_iter().chain(second).chain(iter).collect()),
+    }
+}
+
+const MAX_NGRAM: usize = 3;
+
+pub fn create_query_tree(
+    reader: &heed::RoTxn<MainT>,
+    ctx: &Context,
+    query: &str,
+) -> MResult<(Operation, HashMap<QueryId, Range<usize>>)>
+{
+    let words = split_query_string(query).map(str::to_lowercase);
+    let words = words.filter(|w| !ctx.stop_words.contains(w));
+    let words: Vec<_> = words.enumerate().collect();
+
+    let mut mapper = QueryWordsMapper::new(words.iter().map(|(_, w)| w));
+
+    fn create_inner(
+        reader: &heed::RoTxn<MainT>,
+        ctx: &Context,
+        mapper: &mut QueryWordsMapper,
+        words: &[(usize, String)],
+    ) -> MResult<Vec<Operation>>
+    {
+        let mut alts = Vec::new();
+
+        for ngram in 1..=MAX_NGRAM {
+            if let Some(group) = words.get(..ngram) {
+                let mut group_ops = Vec::new();
+
+                let tail = &words[ngram..];
+                let is_last = tail.is_empty();
+
+                let mut group_alts = Vec::new();
+                match group {
+                    [(id, word)] => {
+                        let mut idgen = ((id + 1) * 100)..;
+                        let range = (*id)..id+1;
+
+                        let phrase = split_best_frequency(reader, ctx, word)?
+                            .map(|ws| {
+                                let id = idgen.next().unwrap();
+                                idgen.next().unwrap();
+                                mapper.declare(range.clone(), id, &[ws.0, ws.1]);
+                                Operation::phrase2(id, is_last, ws)
+                            });
+
+                        let synonyms = fetch_synonyms(reader, ctx, &[word])?
+                            .into_iter()
+                            .map(|alts| {
+                                let exact = alts.len() == 1;
+                                let id = idgen.next().unwrap();
+                                mapper.declare(range.clone(), id, &alts);
+
+                                let mut idgen = once(id).chain(&mut idgen);
+                                let iter = alts.into_iter().map(|w| {
+                                    let id = idgen.next().unwrap();
+                                    let kind = QueryKind::NonTolerant(w);
+                                    Operation::Query(Query { id, prefix: false, exact, kind })
+                                });
+
+                                create_operation(iter, Operation::And)
+                            });
+
+                        let original = Operation::tolerant(*id, is_last, word);
+
+                        group_alts.push(original);
+                        group_alts.extend(synonyms.chain(phrase));
+                    },
+                    words => {
+                        let id = words[0].0;
+                        let mut idgen = ((id + 1) * 100_usize.pow(ngram as u32))..;
+                        let range = id..id+ngram;
+
+                        let words: Vec<_> = words.iter().map(|(_, s)| s.as_str()).collect();
+
+                        for synonym in fetch_synonyms(reader, ctx, &words)? {
+                            let exact = synonym.len() == 1;
+                            let id = idgen.next().unwrap();
+                            mapper.declare(range.clone(), id, &synonym);
+
+                            let mut idgen = once(id).chain(&mut idgen);
+                            let synonym = synonym.into_iter().map(|s| {
+                                let id = idgen.next().unwrap();
+                                let kind = QueryKind::NonTolerant(s);
+                                Operation::Query(Query { id, prefix: false, exact, kind })
+                            });
+                            group_alts.push(create_operation(synonym, Operation::And));
+                        }
+
+                        let id = idgen.next().unwrap();
+                        let concat = words.concat();
+                        mapper.declare(range.clone(), id, &[&concat]);
+                        group_alts.push(Operation::non_tolerant(id, is_last, &concat));
+                    }
+                }
+
+                group_ops.push(create_operation(group_alts, Operation::Or));
+
+                if !tail.is_empty() {
+                    let tail_ops = create_inner(reader, ctx, mapper, tail)?;
+                    group_ops.push(create_operation(tail_ops, Operation::Or));
+                }
+
+                alts.push(create_operation(group_ops, Operation::And));
+            }
+        }
+
+        Ok(alts)
+    }
+
+    let alternatives = create_inner(reader, ctx, &mut mapper, &words)?;
+    let operation = Operation::Or(alternatives);
+    let mapping = mapper.mapping();
+
+    Ok((operation, mapping))
+}
+
+#[derive(Debug, Clone, PartialEq, Eq, Hash)]
+pub struct PostingsKey<'o> {
+    pub query: &'o Query,
+    pub input: Vec<u8>,
+    pub distance: u8,
+    pub is_exact: bool,
+}
+
+pub type Postings<'o, 'txn> = HashMap<PostingsKey<'o>, Cow<'txn, Set<DocIndex>>>;
+pub type Cache<'o, 'txn> = HashMap<&'o Operation, Cow<'txn, Set<DocumentId>>>;
+
+pub struct QueryResult<'o, 'txn> {
+    pub docids: Cow<'txn, Set<DocumentId>>,
+    pub queries: Postings<'o, 'txn>,
+}
+
+pub fn traverse_query_tree<'o, 'txn>(
+    reader: &'txn heed::RoTxn<MainT>,
+    ctx: &Context,
+    tree: &'o Operation,
+) -> MResult<QueryResult<'o, 'txn>>
+{
+    fn execute_and<'o, 'txn>(
+        reader: &'txn heed::RoTxn<MainT>,
+        ctx: &Context,
+        cache: &mut Cache<'o, 'txn>,
+        postings: &mut Postings<'o, 'txn>,
+        depth: usize,
+        operations: &'o [Operation],
+    ) -> MResult<Cow<'txn, Set<DocumentId>>>
+    {
+        debug!("{:1$}AND", "", depth * 2);
+
+        let before = Instant::now();
+        let mut results = Vec::new();
+
+        for op in operations {
+            if cache.get(op).is_none() {
+                let docids = match op {
+                    Operation::And(ops) => execute_and(reader, ctx, cache, postings, depth + 1, &ops)?,
+                    Operation::Or(ops) => execute_or(reader, ctx, cache, postings, depth + 1, &ops)?,
+                    Operation::Query(query) => execute_query(reader, ctx, postings, depth + 1, &query)?,
+                };
+                cache.insert(op, docids);
+            }
+        }
+
+        for op in operations {
+            if let Some(docids) = cache.get(op) {
+                results.push(docids.as_ref());
+            }
+        }
+
+        let op = sdset::multi::Intersection::new(results);
+        let docids = op.into_set_buf();
+
+        debug!("{:3$}--- AND fetched {} documents in {:.02?}", "", docids.len(), before.elapsed(), depth * 2);
+
+        Ok(Cow::Owned(docids))
+    }
+
+    fn execute_or<'o, 'txn>(
+        reader: &'txn heed::RoTxn<MainT>,
+        ctx: &Context,
+        cache: &mut Cache<'o, 'txn>,
+        postings: &mut Postings<'o, 'txn>,
+        depth: usize,
+        operations: &'o [Operation],
+    ) -> MResult<Cow<'txn, Set<DocumentId>>>
+    {
+        debug!("{:1$}OR", "", depth * 2);
+
+        let before = Instant::now();
+        let mut results = Vec::new();
+
+        for op in operations {
+            if cache.get(op).is_none() {
+                let docids = match op {
+                    Operation::And(ops) => execute_and(reader, ctx, cache, postings, depth + 1, &ops)?,
+                    Operation::Or(ops) => execute_or(reader, ctx, cache, postings, depth + 1, &ops)?,
+                    Operation::Query(query) => execute_query(reader, ctx, postings, depth + 1, &query)?,
+                };
+                cache.insert(op, docids);
+            }
+        }
+
+        for op in operations {
+            if let Some(docids) = cache.get(op) {
+                results.push(docids.as_ref());
+            }
+        }
+
+        let op = sdset::multi::Union::new(results);
+        let docids = op.into_set_buf();
+
+        debug!("{:3$}--- OR fetched {} documents in {:.02?}", "", docids.len(), before.elapsed(), depth * 2);
+
+        Ok(Cow::Owned(docids))
+    }
+
+    fn execute_query<'o, 'txn>(
+        reader: &'txn heed::RoTxn<MainT>,
+        ctx: &Context,
+        postings: &mut Postings<'o, 'txn>,
+        depth: usize,
+        query: &'o Query,
+    ) -> MResult<Cow<'txn, Set<DocumentId>>>
+    {
+        let before = Instant::now();
+
+        let Query { prefix, kind, exact, .. } = query;
+        let docids: Cow<Set<_>> = match kind {
+            QueryKind::Tolerant(word) => {
+                if *prefix && word.len() <= 2 {
+                    let prefix = {
+                        let mut array = [0; 4];
+                        let bytes = word.as_bytes();
+                        array[..bytes.len()].copy_from_slice(bytes);
+                        array
+                    };
+
+                    // We retrieve the cached postings lists for all
+                    // the words that starts with this short prefix.
+                    let result = ctx.prefix_postings_lists.prefix_postings_list(reader, prefix)?.unwrap_or_default();
+                    let key = PostingsKey { query, input: word.clone().into_bytes(), distance: 0, is_exact: false };
+                    postings.insert(key, result.matches);
+                    let prefix_docids = &result.docids;
+
+                    // We retrieve the exact postings list for the prefix,
+                    // because we must consider these matches as exact.
+                    let result = ctx.postings_lists.postings_list(reader, word.as_bytes())?.unwrap_or_default();
+                    let key = PostingsKey { query, input: word.clone().into_bytes(), distance: 0, is_exact: true };
+                    postings.insert(key, result.matches);
+                    let exact_docids = &result.docids;
+
+                    let before = Instant::now();
+                    let docids = sdset::duo::Union::new(prefix_docids, exact_docids).into_set_buf();
+                    debug!("{:4$}prefix docids ({} and {}) construction took {:.02?}",
+                        "", prefix_docids.len(), exact_docids.len(), before.elapsed(), depth * 2);
+
+                    Cow::Owned(docids)
+
+                } else {
+                    let dfa = if *prefix { build_prefix_dfa(word) } else { build_dfa(word) };
+
+                    let byte = word.as_bytes()[0];
+                    let mut stream = if byte == u8::max_value() {
+                        ctx.words_set.search(&dfa).ge(&[byte]).into_stream()
+                    } else {
+                        ctx.words_set.search(&dfa).ge(&[byte]).lt(&[byte + 1]).into_stream()
+                    };
+
+                    let before = Instant::now();
+                    let mut results = Vec::new();
+                    while let Some(input) = stream.next() {
+                        if let Some(result) = ctx.postings_lists.postings_list(reader, input)? {
+                            let distance = dfa.eval(input).to_u8();
+                            let is_exact = *exact && distance == 0 && input.len() == word.len();
+                            results.push(result.docids);
+                            let key = PostingsKey { query, input: input.to_owned(), distance, is_exact };
+                            postings.insert(key, result.matches);
+                        }
+                    }
+                    debug!("{:3$}docids retrieval ({:?}) took {:.02?}", "", results.len(), before.elapsed(), depth * 2);
+
+                    let before = Instant::now();
+                    let docids = if results.len() > 10 {
+                        let cap = results.iter().map(|dis| dis.len()).sum();
+                        let mut docids = Vec::with_capacity(cap);
+                        for dis in results {
+                            docids.extend_from_slice(&dis);
+                        }
+                        SetBuf::from_dirty(docids)
+                    } else {
+                        let sets = results.iter().map(AsRef::as_ref).collect();
+                        sdset::multi::Union::new(sets).into_set_buf()
+                    };
+                    debug!("{:2$}docids construction took {:.02?}", "", before.elapsed(), depth * 2);
+
+                    Cow::Owned(docids)
+                }
+            },
+            QueryKind::NonTolerant(word) => {
+                // TODO support prefix and non-prefix exact DFA
+                let dfa = build_exact_dfa(word);
+
+                let byte = word.as_bytes()[0];
+                let mut stream = if byte == u8::max_value() {
+                    ctx.words_set.search(&dfa).ge(&[byte]).into_stream()
+                } else {
+                    ctx.words_set.search(&dfa).ge(&[byte]).lt(&[byte + 1]).into_stream()
+                };
+
+                let before = Instant::now();
+                let mut results = Vec::new();
+                while let Some(input) = stream.next() {
+                    if let Some(result) = ctx.postings_lists.postings_list(reader, input)? {
+                        let distance = dfa.eval(input).to_u8();
+                        results.push(result.docids);
+                        let key = PostingsKey { query, input: input.to_owned(), distance, is_exact: *exact };
+                        postings.insert(key, result.matches);
+                    }
+                }
+                debug!("{:3$}docids retrieval ({:?}) took {:.02?}", "", results.len(), before.elapsed(), depth * 2);
+
+                let before = Instant::now();
+                let docids = if results.len() > 10 {
+                    let cap = results.iter().map(|dis| dis.len()).sum();
+                    let mut docids = Vec::with_capacity(cap);
+                    for dis in results {
+                        docids.extend_from_slice(&dis);
+                    }
+                    SetBuf::from_dirty(docids)
+                } else {
+                    let sets = results.iter().map(AsRef::as_ref).collect();
+                    sdset::multi::Union::new(sets).into_set_buf()
+                };
+                debug!("{:2$}docids construction took {:.02?}", "", before.elapsed(), depth * 2);
+
+                Cow::Owned(docids)
+            },
+            QueryKind::Phrase(words) => {
+                // TODO support prefix and non-prefix exact DFA
+                if let [first, second] = words.as_slice() {
+                    let first = ctx.postings_lists.postings_list(reader, first.as_bytes())?.unwrap_or_default();
+                    let second = ctx.postings_lists.postings_list(reader, second.as_bytes())?.unwrap_or_default();
+
+                    let iter = merge_join_by(first.matches.as_slice(), second.matches.as_slice(), |a, b| {
+                        let x = (a.document_id, a.attribute, (a.word_index as u32) + 1);
+                        let y = (b.document_id, b.attribute, b.word_index as u32);
+                        x.cmp(&y)
+                    });
+
+                    let matches: Vec<_> = iter
+                        .filter_map(EitherOrBoth::both)
+                        .flat_map(|(a, b)| once(*a).chain(Some(*b)))
+                        .collect();
+
+                    let before = Instant::now();
+                    let mut docids: Vec<_> = matches.iter().map(|m| m.document_id).collect();
+                    docids.dedup();
+                    let docids = SetBuf::new(docids).unwrap();
+                    debug!("{:2$}docids construction took {:.02?}", "", before.elapsed(), depth * 2);
+
+                    let matches = Cow::Owned(SetBuf::new(matches).unwrap());
+                    let key = PostingsKey { query, input: vec![], distance: 0, is_exact: true };
+                    postings.insert(key, matches);
+
+                    Cow::Owned(docids)
+                } else {
+                    debug!("{:2$}{:?} skipped", "", words, depth * 2);
+                    Cow::default()
+                }
+            },
+        };
+
+        debug!("{:4$}{:?} fetched {:?} documents in {:.02?}", "", query, docids.len(), before.elapsed(), depth * 2);
+        Ok(docids)
+    }
+
+    let mut cache = Cache::new();
+    let mut postings = Postings::new();
+
+    let docids = match tree {
+        Operation::And(ops) => execute_and(reader, ctx, &mut cache, &mut postings, 0, &ops)?,
+        Operation::Or(ops) => execute_or(reader, ctx, &mut cache, &mut postings, 0, &ops)?,
+        Operation::Query(query) => execute_query(reader, ctx, &mut postings, 0, &query)?,
+    };
+
+    Ok(QueryResult { docids, queries: postings })
+}

meilisearch-core/src/query_words_mapper.rs

@@ -0,0 +1,415 @@
+use std::collections::HashMap;
+use std::iter::FromIterator;
+use std::ops::Range;
+use intervaltree::{Element, IntervalTree};
+
+pub type QueryId = usize;
+
+pub struct QueryWordsMapper {
+    originals: Vec<String>,
+    mappings: HashMap<QueryId, (Range<usize>, Vec<String>)>,
+}
+
+impl QueryWordsMapper {
+    pub fn new<I, A>(originals: I) -> QueryWordsMapper
+    where I: IntoIterator<Item = A>,
+          A: ToString,
+    {
+        let originals = originals.into_iter().map(|s| s.to_string()).collect();
+        QueryWordsMapper { originals, mappings: HashMap::new() }
+    }
+
+    pub fn declare<I, A>(&mut self, range: Range<usize>, id: QueryId, replacement: I)
+    where I: IntoIterator<Item = A>,
+          A: ToString,
+    {
+        assert!(range.len() != 0);
+        assert!(self.originals.get(range.clone()).is_some());
+        assert!(id >= self.originals.len());
+
+        let replacement: Vec<_> = replacement.into_iter().map(|s| s.to_string()).collect();
+
+        assert!(!replacement.is_empty());
+
+        // We detect words at the end and at the front of the
+        // replacement that are common with the originals:
+        //
+        //     x a b c d e f g
+        //       ^^^/   \^^^
+        //     a b x c d k j e f
+        //     ^^^           ^^^
+        //
+
+        let left = &self.originals[..range.start];
+        let right = &self.originals[range.end..];
+
+        let common_left = longest_common_prefix(left, &replacement);
+        let common_right = longest_common_prefix(&replacement, right);
+
+        for i in 0..common_left {
+            let range = range.start - common_left + i..range.start - common_left + i + 1;
+            let replacement = vec![replacement[i].clone()];
+            self.mappings.insert(id + i, (range, replacement));
+        }
+
+        {
+            let replacement = replacement[common_left..replacement.len() - common_right].iter().cloned().collect();
+            self.mappings.insert(id + common_left, (range.clone(), replacement));
+        }
+
+        for i in 0..common_right {
+            let id = id + replacement.len() - common_right + i;
+            let range = range.end + i..range.end + i + 1;
+            let replacement = vec![replacement[replacement.len() - common_right + i].clone()];
+            self.mappings.insert(id, (range, replacement));
+        }
+    }
+
+    pub fn mapping(self) -> HashMap<QueryId, Range<usize>> {
+        let mappings = self.mappings.into_iter().map(|(i, (r, v))| (r, (i, v)));
+        let intervals = IntervalTree::from_iter(mappings);
+
+        let mut output = HashMap::new();
+        let mut offset = 0;
+
+        // We map each original word to the biggest number of
+        // associated words.
+        for i in 0..self.originals.len() {
+            let max = intervals.query_point(i)
+                .filter_map(|e| {
+                    if e.range.end - 1 == i {
+                        let len = e.value.1.iter().skip(i - e.range.start).count();
+                        if len != 0 { Some(len) } else { None }
+                    } else { None }
+                })
+                .max()
+                .unwrap_or(1);
+
+            let range = i + offset..i + offset + max;
+            output.insert(i, range);
+            offset += max - 1;
+        }
+
+        // We retrieve the range that each original word
+        // is mapped to and apply it to each of the words.
+        for i in 0..self.originals.len() {
+
+            let iter = intervals.query_point(i).filter(|e| e.range.end - 1 == i);
+            for Element { range, value: (id, words) } in iter {
+
+                // We ask for the complete range mapped to the area we map.
+                let start = output.get(&range.start).map(|r| r.start).unwrap_or(range.start);
+                let end = output.get(&(range.end - 1)).map(|r| r.end).unwrap_or(range.end);
+                let range = start..end;
+
+                // We map each query id to one word until the last,
+                // we map it to the remainings words.
+                let add = range.len() - words.len();
+                for (j, x) in range.take(words.len()).enumerate() {
+                    let add = if j == words.len() - 1 { add } else { 0 }; // is last?
+                    let range = x..x + 1 + add;
+                    output.insert(id + j, range);
+                }
+            }
+        }
+
+        output
+    }
+}
+
+fn longest_common_prefix<T: Eq + std::fmt::Debug>(a: &[T], b: &[T]) -> usize {
+    let mut best = None;
+    for i in (0..a.len()).rev() {
+        let count = a[i..].iter().zip(b).take_while(|(a, b)| a == b).count();
+        best = match best {
+            Some(old) if count > old => Some(count),
+            Some(_) => break,
+            None => Some(count),
+        };
+    }
+    best.unwrap_or(0)
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn original_unmodified() {
+        let query = ["new", "york", "city", "subway"];
+        //             0       1       2        3
+        let mut builder = QueryWordsMapper::new(&query);
+
+        // new york = new york city
+        builder.declare(0..2, 4, &["new", "york", "city"]);
+        //                    ^      4       5       6
+
+        // new = new york city
+        builder.declare(0..1, 7, &["new", "york", "city"]);
+        //                    ^      7       8       9
+
+        let mapping = builder.mapping();
+
+        assert_eq!(mapping[&0], 0..1); // new
+        assert_eq!(mapping[&1], 1..2); // york
+        assert_eq!(mapping[&2], 2..3); // city
+        assert_eq!(mapping[&3], 3..4); // subway
+
+        assert_eq!(mapping[&4], 0..1); // new
+        assert_eq!(mapping[&5], 1..2); // york
+        assert_eq!(mapping[&6], 2..3); // city
+
+        assert_eq!(mapping[&7], 0..1); // new
+        assert_eq!(mapping[&8], 1..2); // york
+        assert_eq!(mapping[&9], 2..3); // city
+    }
+
+    #[test]
+    fn original_unmodified2() {
+        let query = ["new", "york", "city", "subway"];
+        //             0       1       2        3
+        let mut builder = QueryWordsMapper::new(&query);
+
+        // city subway = new york city underground train
+        builder.declare(2..4, 4, &["new", "york", "city", "underground", "train"]);
+        //                    ^      4      5       6           7           8
+
+        let mapping = builder.mapping();
+
+        assert_eq!(mapping[&0], 0..1); // new
+        assert_eq!(mapping[&1], 1..2); // york
+        assert_eq!(mapping[&2], 2..3); // city
+        assert_eq!(mapping[&3], 3..5); // subway
+
+        assert_eq!(mapping[&4], 0..1); // new
+        assert_eq!(mapping[&5], 1..2); // york
+        assert_eq!(mapping[&6], 2..3); // city
+        assert_eq!(mapping[&7], 3..4); // underground
+        assert_eq!(mapping[&8], 4..5); // train
+    }
+
+    #[test]
+    fn original_unmodified3() {
+        let query = ["a", "b", "x", "x", "a", "b", "c", "d", "e", "f", "g"];
+        //            0    1    2    3    4    5    6    7    8    9    10
+        let mut builder = QueryWordsMapper::new(&query);
+
+        // c d = a b x c d k j e f
+        builder.declare(6..8, 11, &["a", "b", "x", "c", "d", "k", "j", "e", "f"]);
+        //                    ^^    11   12   13   14   15   16   17   18   19
+
+        let mapping = builder.mapping();
+
+        assert_eq!(mapping[&0],  0..1); // a
+        assert_eq!(mapping[&1],  1..2); // b
+        assert_eq!(mapping[&2],  2..3); // x
+        assert_eq!(mapping[&3],  3..4); // x
+        assert_eq!(mapping[&4],  4..5); // a
+        assert_eq!(mapping[&5],  5..6); // b
+        assert_eq!(mapping[&6],  6..7); // c
+        assert_eq!(mapping[&7],  7..11); // d
+        assert_eq!(mapping[&8],  11..12); // e
+        assert_eq!(mapping[&9],  12..13); // f
+        assert_eq!(mapping[&10], 13..14); // g
+
+        assert_eq!(mapping[&11], 4..5); // a
+        assert_eq!(mapping[&12], 5..6); // b
+        assert_eq!(mapping[&13], 6..7); // x
+        assert_eq!(mapping[&14], 7..8); // c
+        assert_eq!(mapping[&15], 8..9); // d
+        assert_eq!(mapping[&16], 9..10); // k
+        assert_eq!(mapping[&17], 10..11); // j
+        assert_eq!(mapping[&18], 11..12); // e
+        assert_eq!(mapping[&19], 12..13); // f
+    }
+
+    #[test]
+    fn simple_growing() {
+        let query = ["new", "york", "subway"];
+        //             0       1        2
+        let mut builder = QueryWordsMapper::new(&query);
+
+        // new york = new york city
+        builder.declare(0..2, 3, &["new", "york", "city"]);
+        //                    ^      3       4       5
+
+        let mapping = builder.mapping();
+
+        assert_eq!(mapping[&0], 0..1); // new
+        assert_eq!(mapping[&1], 1..3); // york
+        assert_eq!(mapping[&2], 3..4); // subway
+        assert_eq!(mapping[&3], 0..1); // new
+        assert_eq!(mapping[&4], 1..2); // york
+        assert_eq!(mapping[&5], 2..3); // city
+    }
+
+    #[test]
+    fn same_place_growings() {
+        let query = ["NY", "subway"];
+        //             0       1
+        let mut builder = QueryWordsMapper::new(&query);
+
+        // NY = new york
+        builder.declare(0..1, 2, &["new", "york"]);
+        //                    ^      2       3
+
+        // NY = new york city
+        builder.declare(0..1, 4, &["new", "york", "city"]);
+        //                    ^      4       5       6
+
+        // NY = NYC
+        builder.declare(0..1, 7, &["NYC"]);
+        //                    ^      7
+
+        // NY = new york city
+        builder.declare(0..1, 8, &["new", "york", "city"]);
+        //                    ^      8       9      10
+
+        // subway = underground train
+        builder.declare(1..2, 11, &["underground", "train"]);
+        //                    ^          11          12
+
+        let mapping = builder.mapping();
+
+        assert_eq!(mapping[&0], 0..3); // NY
+        assert_eq!(mapping[&1], 3..5); // subway
+        assert_eq!(mapping[&2], 0..1); // new
+        assert_eq!(mapping[&3], 1..3); // york
+        assert_eq!(mapping[&4], 0..1); // new
+        assert_eq!(mapping[&5], 1..2); // york
+        assert_eq!(mapping[&6], 2..3); // city
+        assert_eq!(mapping[&7], 0..3); // NYC
+        assert_eq!(mapping[&8], 0..1); // new
+        assert_eq!(mapping[&9], 1..2); // york
+        assert_eq!(mapping[&10], 2..3); // city
+        assert_eq!(mapping[&11], 3..4); // underground
+        assert_eq!(mapping[&12], 4..5); // train
+    }
+
+    #[test]
+    fn bigger_growing() {
+        let query = ["NYC", "subway"];
+        //             0        1
+        let mut builder = QueryWordsMapper::new(&query);
+
+        // NYC = new york city
+        builder.declare(0..1, 2, &["new", "york", "city"]);
+        //                    ^      2       3       4
+
+        let mapping = builder.mapping();
+
+        assert_eq!(mapping[&0], 0..3); // NYC
+        assert_eq!(mapping[&1], 3..4); // subway
+        assert_eq!(mapping[&2], 0..1); // new
+        assert_eq!(mapping[&3], 1..2); // york
+        assert_eq!(mapping[&4], 2..3); // city
+    }
+
+    #[test]
+    fn middle_query_growing() {
+        let query = ["great", "awesome", "NYC", "subway"];
+        //              0         1        2        3
+        let mut builder = QueryWordsMapper::new(&query);
+
+        // NYC = new york city
+        builder.declare(2..3, 4, &["new", "york", "city"]);
+        //                    ^      4       5       6
+
+        let mapping = builder.mapping();
+
+        assert_eq!(mapping[&0], 0..1); // great
+        assert_eq!(mapping[&1], 1..2); // awesome
+        assert_eq!(mapping[&2], 2..5); // NYC
+        assert_eq!(mapping[&3], 5..6); // subway
+        assert_eq!(mapping[&4], 2..3); // new
+        assert_eq!(mapping[&5], 3..4); // york
+        assert_eq!(mapping[&6], 4..5); // city
+    }
+
+    #[test]
+    fn end_query_growing() {
+        let query = ["NYC", "subway"];
+        //             0        1
+        let mut builder = QueryWordsMapper::new(&query);
+
+        // NYC = new york city
+        builder.declare(1..2, 2, &["underground", "train"]);
+        //                    ^         2            3
+
+        let mapping = builder.mapping();
+
+        assert_eq!(mapping[&0], 0..1); // NYC
+        assert_eq!(mapping[&1], 1..3); // subway
+        assert_eq!(mapping[&2], 1..2); // underground
+        assert_eq!(mapping[&3], 2..3); // train
+    }
+
+    #[test]
+    fn multiple_growings() {
+        let query = ["great", "awesome", "NYC", "subway"];
+        //              0         1        2        3
+        let mut builder = QueryWordsMapper::new(&query);
+
+        // NYC = new york city
+        builder.declare(2..3, 4, &["new", "york", "city"]);
+        //                    ^      4       5       6
+
+        // subway = underground train
+        builder.declare(3..4, 7, &["underground", "train"]);
+        //                    ^          7           8
+
+        let mapping = builder.mapping();
+
+        assert_eq!(mapping[&0], 0..1); // great
+        assert_eq!(mapping[&1], 1..2); // awesome
+        assert_eq!(mapping[&2], 2..5); // NYC
+        assert_eq!(mapping[&3], 5..7); // subway
+        assert_eq!(mapping[&4], 2..3); // new
+        assert_eq!(mapping[&5], 3..4); // york
+        assert_eq!(mapping[&6], 4..5); // city
+        assert_eq!(mapping[&7], 5..6); // underground
+        assert_eq!(mapping[&8], 6..7); // train
+    }
+
+    #[test]
+    fn multiple_probable_growings() {
+        let query = ["great", "awesome", "NYC", "subway"];
+        //              0         1        2        3
+        let mut builder = QueryWordsMapper::new(&query);
+
+        // NYC = new york city
+        builder.declare(2..3, 4, &["new", "york", "city"]);
+        //                    ^      4       5       6
+
+        // subway = underground train
+        builder.declare(3..4, 7, &["underground", "train"]);
+        //                    ^          7           8
+
+        // great awesome = good
+        builder.declare(0..2, 9, &["good"]);
+        //                    ^       9
+
+        // awesome NYC = NY
+        builder.declare(1..3, 10, &["NY"]);
+        //                    ^^     10
+
+        // NYC subway = metro
+        builder.declare(2..4, 11, &["metro"]);
+        //                    ^^      11
+
+        let mapping = builder.mapping();
+
+        assert_eq!(mapping[&0], 0..1); // great
+        assert_eq!(mapping[&1], 1..2); // awesome
+        assert_eq!(mapping[&2], 2..5); // NYC
+        assert_eq!(mapping[&3], 5..7); // subway
+        assert_eq!(mapping[&4], 2..3); // new
+        assert_eq!(mapping[&5], 3..4); // york
+        assert_eq!(mapping[&6], 4..5); // city
+        assert_eq!(mapping[&7], 5..6); // underground
+        assert_eq!(mapping[&8], 6..7); // train
+        assert_eq!(mapping[&9], 0..2); // good
+        assert_eq!(mapping[&10], 1..5); // NY
+        assert_eq!(mapping[&11], 2..7); // metro
+    }
+}

meilisearch-core/src/ranked_map.rs

@@ -1,14 +1,14 @@
 use std::io::{Read, Write};
 
 use hashbrown::HashMap;
-use meilidb_schema::SchemaAttr;
+use meilisearch_schema::FieldId;
 use serde::{Deserialize, Serialize};
 
 use crate::{DocumentId, Number};
 
 #[derive(Debug, Default, Clone, PartialEq, Eq, Serialize, Deserialize)]
 #[serde(transparent)]
-pub struct RankedMap(HashMap<(DocumentId, SchemaAttr), Number>);
+pub struct RankedMap(HashMap<(DocumentId, FieldId), Number>);
 
 impl RankedMap {
     pub fn len(&self) -> usize {
@@ -19,16 +19,16 @@ impl RankedMap {
         self.0.is_empty()
     }
 
-    pub fn insert(&mut self, document: DocumentId, attribute: SchemaAttr, number: Number) {
-        self.0.insert((document, attribute), number);
+    pub fn insert(&mut self, document: DocumentId, field: FieldId, number: Number) {
+        self.0.insert((document, field), number);
     }
 
-    pub fn remove(&mut self, document: DocumentId, attribute: SchemaAttr) {
-        self.0.remove(&(document, attribute));
+    pub fn remove(&mut self, document: DocumentId, field: FieldId) {
+        self.0.remove(&(document, field));
     }
 
-    pub fn get(&self, document: DocumentId, attribute: SchemaAttr) -> Option<Number> {
-        self.0.get(&(document, attribute)).cloned()
+    pub fn get(&self, document: DocumentId, field: FieldId) -> Option<Number> {
+        self.0.get(&(document, field)).cloned()
     }
 
     pub fn read_from_bin<R: Read>(reader: R) -> bincode::Result<RankedMap> {

meilisearch-core/src/raw_document.rs

@@ -0,0 +1,51 @@
+use compact_arena::SmallArena;
+use sdset::SetBuf;
+use crate::DocIndex;
+use crate::bucket_sort::{SimpleMatch, BareMatch, PostingsListView};
+use crate::reordered_attrs::ReorderedAttrs;
+
+pub struct RawDocument<'a, 'tag> {
+    pub id: crate::DocumentId,
+    pub bare_matches: &'a mut [BareMatch<'tag>],
+    pub processed_matches: Vec<SimpleMatch>,
+    /// The list of minimum `distance` found
+    pub processed_distances: Vec<Option<u8>>,
+    /// Does this document contains a field
+    /// with one word that is exactly matching
+    pub contains_one_word_field: bool,
+}
+
+impl<'a, 'tag> RawDocument<'a, 'tag> {
+    pub fn new<'txn>(
+        bare_matches: &'a mut [BareMatch<'tag>],
+        postings_lists: &mut SmallArena<'tag, PostingsListView<'txn>>,
+        searchable_attrs: Option<&ReorderedAttrs>,
+    ) -> RawDocument<'a, 'tag>
+    {
+        if let Some(reordered_attrs) = searchable_attrs {
+            for bm in bare_matches.iter() {
+                let postings_list = &postings_lists[bm.postings_list];
+
+                let mut rewritten = Vec::new();
+                for di in postings_list.iter() {
+                    if let Some(attribute) = reordered_attrs.get(di.attribute) {
+                        rewritten.push(DocIndex { attribute, ..*di });
+                    }
+                }
+
+                let new_postings = SetBuf::from_dirty(rewritten);
+                postings_lists[bm.postings_list].rewrite_with(new_postings);
+            }
+        }
+
+        bare_matches.sort_unstable_by_key(|m| m.query_index);
+
+        RawDocument {
+            id: bare_matches[0].document_id,
+            bare_matches,
+            processed_matches: Vec::new(),
+            processed_distances: Vec::new(),
+            contains_one_word_field: false,
+        }
+    }
+}

meilisearch-core/src/raw_indexer.rs

@@ -0,0 +1,272 @@
+use std::collections::{BTreeMap, HashMap};
+use std::convert::TryFrom;
+
+use crate::{DocIndex, DocumentId};
+use deunicode::deunicode_with_tofu;
+use meilisearch_schema::IndexedPos;
+use meilisearch_tokenizer::{is_cjk, SeqTokenizer, Token, Tokenizer};
+use sdset::SetBuf;
+
+const WORD_LENGTH_LIMIT: usize = 80;
+
+type Word = Vec<u8>; // TODO make it be a SmallVec
+
+pub struct RawIndexer {
+    word_limit: usize, // the maximum number of indexed words
+    stop_words: fst::Set,
+    words_doc_indexes: BTreeMap<Word, Vec<DocIndex>>,
+    docs_words: HashMap<DocumentId, Vec<Word>>,
+}
+
+pub struct Indexed {
+    pub words_doc_indexes: BTreeMap<Word, SetBuf<DocIndex>>,
+    pub docs_words: HashMap<DocumentId, fst::Set>,
+}
+
+impl RawIndexer {
+    pub fn new(stop_words: fst::Set) -> RawIndexer {
+        RawIndexer::with_word_limit(stop_words, 1000)
+    }
+
+    pub fn with_word_limit(stop_words: fst::Set, limit: usize) -> RawIndexer {
+        RawIndexer {
+            word_limit: limit,
+            stop_words,
+            words_doc_indexes: BTreeMap::new(),
+            docs_words: HashMap::new(),
+        }
+    }
+
+    pub fn index_text(&mut self, id: DocumentId, indexed_pos: IndexedPos, text: &str) -> usize {
+        let mut number_of_words = 0;
+
+        for token in Tokenizer::new(text) {
+            let must_continue = index_token(
+                token,
+                id,
+                indexed_pos,
+                self.word_limit,
+                &self.stop_words,
+                &mut self.words_doc_indexes,
+                &mut self.docs_words,
+            );
+
+            number_of_words += 1;
+
+            if !must_continue {
+                break;
+            }
+        }
+
+        number_of_words
+    }
+
+    pub fn index_text_seq<'a, I>(&mut self, id: DocumentId, indexed_pos: IndexedPos, iter: I)
+    where
+        I: IntoIterator<Item = &'a str>,
+    {
+        let iter = iter.into_iter();
+        for token in SeqTokenizer::new(iter) {
+            let must_continue = index_token(
+                token,
+                id,
+                indexed_pos,
+                self.word_limit,
+                &self.stop_words,
+                &mut self.words_doc_indexes,
+                &mut self.docs_words,
+            );
+
+            if !must_continue {
+                break;
+            }
+        }
+    }
+
+    pub fn build(self) -> Indexed {
+        let words_doc_indexes = self
+            .words_doc_indexes
+            .into_iter()
+            .map(|(word, indexes)| (word, SetBuf::from_dirty(indexes)))
+            .collect();
+
+        let docs_words = self
+            .docs_words
+            .into_iter()
+            .map(|(id, mut words)| {
+                words.sort_unstable();
+                words.dedup();
+                (id, fst::Set::from_iter(words).unwrap())
+            })
+            .collect();
+
+        Indexed {
+            words_doc_indexes,
+            docs_words,
+        }
+    }
+}
+
+fn index_token(
+    token: Token,
+    id: DocumentId,
+    indexed_pos: IndexedPos,
+    word_limit: usize,
+    stop_words: &fst::Set,
+    words_doc_indexes: &mut BTreeMap<Word, Vec<DocIndex>>,
+    docs_words: &mut HashMap<DocumentId, Vec<Word>>,
+) -> bool {
+    if token.word_index >= word_limit {
+        return false;
+    }
+
+    let lower = token.word.to_lowercase();
+    let token = Token {
+        word: &lower,
+        ..token
+    };
+
+    if !stop_words.contains(&token.word) {
+        match token_to_docindex(id, indexed_pos, token) {
+            Some(docindex) => {
+                let word = Vec::from(token.word);
+
+                if word.len() <= WORD_LENGTH_LIMIT {
+                    words_doc_indexes
+                        .entry(word.clone())
+                        .or_insert_with(Vec::new)
+                        .push(docindex);
+                    docs_words.entry(id).or_insert_with(Vec::new).push(word);
+
+                    if !lower.contains(is_cjk) {
+                        let unidecoded = deunicode_with_tofu(&lower, "");
+                        if unidecoded != lower && !unidecoded.is_empty() {
+                            let word = Vec::from(unidecoded);
+                            if word.len() <= WORD_LENGTH_LIMIT {
+                                words_doc_indexes
+                                    .entry(word.clone())
+                                    .or_insert_with(Vec::new)
+                                    .push(docindex);
+                                docs_words.entry(id).or_insert_with(Vec::new).push(word);
+                            }
+                        }
+                    }
+                }
+            }
+            None => return false,
+        }
+    }
+
+    true
+}
+
+fn token_to_docindex(id: DocumentId, indexed_pos: IndexedPos, token: Token) -> Option<DocIndex> {
+    let word_index = u16::try_from(token.word_index).ok()?;
+    let char_index = u16::try_from(token.char_index).ok()?;
+    let char_length = u16::try_from(token.word.chars().count()).ok()?;
+
+    let docindex = DocIndex {
+        document_id: id,
+        attribute: indexed_pos.0,
+        word_index,
+        char_index,
+        char_length,
+    };
+
+    Some(docindex)
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use meilisearch_schema::IndexedPos;
+
+    #[test]
+    fn strange_apostrophe() {
+        let mut indexer = RawIndexer::new(fst::Set::default());
+
+        let docid = DocumentId(0);
+        let indexed_pos = IndexedPos(0);
+        let text = "Zut, l’aspirateur, j’ai oublié de l’éteindre !";
+        indexer.index_text(docid, indexed_pos, text);
+
+        let Indexed {
+            words_doc_indexes, ..
+        } = indexer.build();
+
+        assert!(words_doc_indexes.get(&b"l"[..]).is_some());
+        assert!(words_doc_indexes.get(&b"aspirateur"[..]).is_some());
+        assert!(words_doc_indexes.get(&b"ai"[..]).is_some());
+        assert!(words_doc_indexes.get(&b"eteindre"[..]).is_some());
+        assert!(words_doc_indexes
+            .get(&"éteindre".to_owned().into_bytes())
+            .is_some());
+    }
+
+    #[test]
+    fn strange_apostrophe_in_sequence() {
+        let mut indexer = RawIndexer::new(fst::Set::default());
+
+        let docid = DocumentId(0);
+        let indexed_pos = IndexedPos(0);
+        let text = vec!["Zut, l’aspirateur, j’ai oublié de l’éteindre !"];
+        indexer.index_text_seq(docid, indexed_pos, text);
+
+        let Indexed {
+            words_doc_indexes, ..
+        } = indexer.build();
+
+        assert!(words_doc_indexes.get(&b"l"[..]).is_some());
+        assert!(words_doc_indexes.get(&b"aspirateur"[..]).is_some());
+        assert!(words_doc_indexes.get(&b"ai"[..]).is_some());
+        assert!(words_doc_indexes.get(&b"eteindre"[..]).is_some());
+        assert!(words_doc_indexes
+            .get(&"éteindre".to_owned().into_bytes())
+            .is_some());
+    }
+
+    #[test]
+    fn basic_stop_words() {
+        let stop_words = sdset::SetBuf::from_dirty(vec!["l", "j", "ai", "de"]);
+        let stop_words = fst::Set::from_iter(stop_words).unwrap();
+
+        let mut indexer = RawIndexer::new(stop_words);
+
+        let docid = DocumentId(0);
+        let indexed_pos = IndexedPos(0);
+        let text = "Zut, l’aspirateur, j’ai oublié de l’éteindre !";
+        indexer.index_text(docid, indexed_pos, text);
+
+        let Indexed {
+            words_doc_indexes, ..
+        } = indexer.build();
+
+        assert!(words_doc_indexes.get(&b"l"[..]).is_none());
+        assert!(words_doc_indexes.get(&b"aspirateur"[..]).is_some());
+        assert!(words_doc_indexes.get(&b"j"[..]).is_none());
+        assert!(words_doc_indexes.get(&b"ai"[..]).is_none());
+        assert!(words_doc_indexes.get(&b"de"[..]).is_none());
+        assert!(words_doc_indexes.get(&b"eteindre"[..]).is_some());
+        assert!(words_doc_indexes
+            .get(&"éteindre".to_owned().into_bytes())
+            .is_some());
+    }
+
+    #[test]
+    fn no_empty_unidecode() {
+        let mut indexer = RawIndexer::new(fst::Set::default());
+
+        let docid = DocumentId(0);
+        let indexed_pos = IndexedPos(0);
+        let text = "🇯🇵";
+        indexer.index_text(docid, indexed_pos, text);
+
+        let Indexed {
+            words_doc_indexes, ..
+        } = indexer.build();
+
+        assert!(words_doc_indexes
+            .get(&"🇯🇵".to_owned().into_bytes())
+            .is_some());
+    }
+}

meilisearch-core/src/reordered_attrs.rs

@@ -0,0 +1,31 @@
+use std::cmp;
+
+#[derive(Default, Clone)]
+pub struct ReorderedAttrs {
+    reorders: Vec<Option<u16>>,
+    reverse: Vec<u16>,
+}
+
+impl ReorderedAttrs {
+    pub fn new() -> ReorderedAttrs {
+        ReorderedAttrs { reorders: Vec::new(), reverse: Vec::new() }
+    }
+
+    pub fn insert_attribute(&mut self, attribute: u16) {
+        let new_len = cmp::max(attribute as usize + 1, self.reorders.len());
+        self.reorders.resize(new_len, None);
+        self.reorders[attribute as usize] = Some(self.reverse.len() as u16);
+        self.reverse.push(attribute);
+    }
+
+    pub fn get(&self, attribute: u16) -> Option<u16> {
+        match self.reorders.get(attribute as usize)? {
+            Some(attribute) => Some(*attribute),
+            None => None,
+        }
+    }
+
+    pub fn reverse(&self, attribute: u16) -> Option<u16> {
+        self.reverse.get(attribute as usize).copied()
+    }
+}

meilisearch-core/src/serde/convert_to_string.rs

@@ -12,8 +12,8 @@ impl ser::Serializer for ConvertToString {
     type SerializeTuple = ser::Impossible<Self::Ok, Self::Error>;
     type SerializeTupleStruct = ser::Impossible<Self::Ok, Self::Error>;
     type SerializeTupleVariant = ser::Impossible<Self::Ok, Self::Error>;
-    type SerializeMap = ser::Impossible<Self::Ok, Self::Error>;
-    type SerializeStruct = ser::Impossible<Self::Ok, Self::Error>;
+    type SerializeMap = MapConvertToString;
+    type SerializeStruct = StructConvertToString;
     type SerializeStructVariant = ser::Impossible<Self::Ok, Self::Error>;
 
     fn serialize_bool(self, _value: bool) -> Result<Self::Ok, Self::Error> {
@@ -169,7 +169,9 @@ impl ser::Serializer for ConvertToString {
     }
 
     fn serialize_map(self, _len: Option<usize>) -> Result<Self::SerializeMap, Self::Error> {
-        Err(SerializerError::UnserializableType { type_name: "map" })
+        Ok(MapConvertToString {
+            text: String::new(),
+        })
     }
 
     fn serialize_struct(
@@ -177,8 +179,8 @@ impl ser::Serializer for ConvertToString {
         _name: &'static str,
         _len: usize,
     ) -> Result<Self::SerializeStruct, Self::Error> {
-        Err(SerializerError::UnserializableType {
-            type_name: "struct",
+        Ok(StructConvertToString {
+            text: String::new(),
         })
     }
 
@@ -194,3 +196,63 @@ impl ser::Serializer for ConvertToString {
         })
     }
 }
+
+pub struct MapConvertToString {
+    text: String,
+}
+
+impl ser::SerializeMap for MapConvertToString {
+    type Ok = String;
+    type Error = SerializerError;
+
+    fn serialize_key<T: ?Sized>(&mut self, key: &T) -> Result<(), Self::Error>
+    where
+        T: ser::Serialize,
+    {
+        let text = key.serialize(ConvertToString)?;
+        self.text.push_str(&text);
+        self.text.push_str(" ");
+        Ok(())
+    }
+
+    fn serialize_value<T: ?Sized>(&mut self, value: &T) -> Result<(), Self::Error>
+    where
+        T: ser::Serialize,
+    {
+        let text = value.serialize(ConvertToString)?;
+        self.text.push_str(&text);
+        Ok(())
+    }
+
+    fn end(self) -> Result<Self::Ok, Self::Error> {
+        Ok(self.text)
+    }
+}
+
+pub struct StructConvertToString {
+    text: String,
+}
+
+impl ser::SerializeStruct for StructConvertToString {
+    type Ok = String;
+    type Error = SerializerError;
+
+    fn serialize_field<T: ?Sized>(
+        &mut self,
+        key: &'static str,
+        value: &T,
+    ) -> Result<(), Self::Error>
+    where
+        T: ser::Serialize,
+    {
+        let value = value.serialize(ConvertToString)?;
+        self.text.push_str(key);
+        self.text.push_str(" ");
+        self.text.push_str(&value);
+        Ok(())
+    }
+
+    fn end(self) -> Result<Self::Ok, Self::Error> {
+        Ok(self.text)
+    }
+}

meilisearch-core/src/serde/deserializer.rs

@@ -2,12 +2,13 @@ use std::collections::HashSet;
 use std::io::Cursor;
 use std::{error::Error, fmt};
 
-use meilidb_schema::{Schema, SchemaAttr};
+use meilisearch_schema::{Schema, FieldId};
 use serde::{de, forward_to_deserialize_any};
 use serde_json::de::IoRead as SerdeJsonIoRead;
 use serde_json::Deserializer as SerdeJsonDeserializer;
 use serde_json::Error as SerdeJsonError;
 
+use crate::database::MainT;
 use crate::store::DocumentsFields;
 use crate::DocumentId;
 
@@ -50,10 +51,10 @@ impl From<heed::Error> for DeserializerError {
 
 pub struct Deserializer<'a> {
     pub document_id: DocumentId,
-    pub reader: &'a heed::RoTxn,
+    pub reader: &'a heed::RoTxn<MainT>,
     pub documents_fields: DocumentsFields,
     pub schema: &'a Schema,
-    pub attributes: Option<&'a HashSet<SchemaAttr>>,
+    pub fields: Option<&'a HashSet<FieldId>>,
 }
 
 impl<'de, 'a, 'b> de::Deserializer<'de> for &'b mut Deserializer<'a> {
@@ -63,13 +64,14 @@ impl<'de, 'a, 'b> de::Deserializer<'de> for &'b mut Deserializer<'a> {
     where
         V: de::Visitor<'de>,
     {
-        self.deserialize_map(visitor)
+        self.deserialize_option(visitor)
     }
 
-    forward_to_deserialize_any! {
-        bool i8 i16 i32 i64 i128 u8 u16 u32 u64 u128 f32 f64 char str string
-        bytes byte_buf option unit unit_struct newtype_struct seq tuple
-        tuple_struct struct enum identifier ignored_any
+    fn deserialize_option<V>(self, visitor: V) -> Result<V::Value, Self::Error>
+    where
+        V: de::Visitor<'de>,
+    {
+        self.deserialize_map(visitor)
     }
 
     fn deserialize_map<V>(self, visitor: V) -> Result<V::Value, Self::Error>
@@ -90,30 +92,45 @@ impl<'de, 'a, 'b> de::Deserializer<'de> for &'b mut Deserializer<'a> {
                     }
                 };
 
-                let is_displayed = self.schema.props(attr).is_displayed();
-                if is_displayed && self.attributes.map_or(true, |f| f.contains(&attr)) {
-                    let attribute_name = self.schema.attribute_name(attr);
+                let is_displayed = self.schema.is_displayed(attr);
+                if is_displayed && self.fields.map_or(true, |f| f.contains(&attr)) {
+                    if let Some(attribute_name) = self.schema.name(attr) {
+                        let cursor = Cursor::new(value.to_owned());
+                        let ioread = SerdeJsonIoRead::new(cursor);
+                        let value = Value(SerdeJsonDeserializer::new(ioread));
 
-                    let cursor = Cursor::new(value.to_owned());
-                    let ioread = SerdeJsonIoRead::new(cursor);
-                    let value = Value(SerdeJsonDeserializer::new(ioread));
-
-                    Some((attribute_name, value))
+                        Some((attribute_name, value))
+                    } else {
+                        None
+                    }
                 } else {
                     None
                 }
             });
 
-        let map_deserializer = de::value::MapDeserializer::new(iter);
-        let result = visitor
-            .visit_map(map_deserializer)
-            .map_err(DeserializerError::from);
+        let mut iter = iter.peekable();
+
+        let result = match iter.peek() {
+            Some(_) => {
+                let map_deserializer = de::value::MapDeserializer::new(iter);
+                visitor
+                    .visit_some(map_deserializer)
+                    .map_err(DeserializerError::from)
+            }
+            None => visitor.visit_none(),
+        };
 
         match error.take() {
             Some(error) => Err(error.into()),
             None => result,
         }
     }
+
+    forward_to_deserialize_any! {
+        bool i8 i16 i32 i64 i128 u8 u16 u32 u64 u128 f32 f64 char str string
+        bytes byte_buf unit unit_struct newtype_struct seq tuple
+        tuple_struct struct enum identifier ignored_any
+    }
 }
 
 struct Value(SerdeJsonDeserializer<SerdeJsonIoRead<Cursor<Vec<u8>>>>);

meilisearch-core/src/serde/extract_document_id.rs

@@ -2,28 +2,43 @@ use std::hash::{Hash, Hasher};
 
 use crate::DocumentId;
 use serde::{ser, Serialize};
-use serde_json::Value;
+use serde_json::{Value, Number};
 use siphasher::sip::SipHasher;
 
 use super::{ConvertToString, SerializerError};
 
 pub fn extract_document_id<D>(
-    identifier: &str,
+    primary_key: &str,
     document: &D,
 ) -> Result<Option<DocumentId>, SerializerError>
 where
     D: serde::Serialize,
 {
-    let serializer = ExtractDocumentId { identifier };
+    let serializer = ExtractDocumentId { primary_key };
     document.serialize(serializer)
 }
 
+fn validate_number(value: &Number) -> Option<String> {
+    if value.is_f64() {
+        return None
+    }
+    Some(value.to_string())
+}
+
+fn validate_string(value: &str) -> Option<String> {
+    if value.chars().all(|x| x.is_ascii_alphanumeric() || x == '-' || x == '_') {
+        Some(value.to_string())
+    } else {
+        None
+    }
+}
+
 pub fn value_to_string(value: &Value) -> Option<String> {
     match value {
         Value::Null => None,
         Value::Bool(_) => None,
-        Value::Number(value) => Some(value.to_string()),
-        Value::String(value) => Some(value.to_string()),
+        Value::Number(value) => validate_number(value),
+        Value::String(value) => validate_string(value),
         Value::Array(_) => None,
         Value::Object(_) => None,
     }
@@ -37,7 +52,7 @@ pub fn compute_document_id<H: Hash>(t: H) -> DocumentId {
 }
 
 struct ExtractDocumentId<'a> {
-    identifier: &'a str,
+    primary_key: &'a str,
 }
 
 impl<'a> ser::Serializer for ExtractDocumentId<'a> {
@@ -173,7 +188,7 @@ impl<'a> ser::Serializer for ExtractDocumentId<'a> {
 
     fn serialize_map(self, _len: Option<usize>) -> Result<Self::SerializeMap, Self::Error> {
         let serializer = ExtractDocumentIdMapSerializer {
-            identifier: self.identifier,
+            primary_key: self.primary_key,
             document_id: None,
             current_key_name: None,
         };
@@ -187,7 +202,7 @@ impl<'a> ser::Serializer for ExtractDocumentId<'a> {
         _len: usize,
     ) -> Result<Self::SerializeStruct, Self::Error> {
         let serializer = ExtractDocumentIdStructSerializer {
-            identifier: self.identifier,
+            primary_key: self.primary_key,
             document_id: None,
         };
 
@@ -208,7 +223,7 @@ impl<'a> ser::Serializer for ExtractDocumentId<'a> {
 }
 
 pub struct ExtractDocumentIdMapSerializer<'a> {
-    identifier: &'a str,
+    primary_key: &'a str,
     document_id: Option<DocumentId>,
     current_key_name: Option<String>,
 }
@@ -245,7 +260,7 @@ impl<'a> ser::SerializeMap for ExtractDocumentIdMapSerializer<'a> {
     {
         let key = key.serialize(ConvertToString)?;
 
-        if self.identifier == key {
+        if self.primary_key == key {
             let value = serde_json::to_string(value).and_then(|s| serde_json::from_str(&s))?;
             match value_to_string(&value).map(|s| compute_document_id(&s)) {
                 Some(document_id) => self.document_id = Some(document_id),
@@ -262,7 +277,7 @@ impl<'a> ser::SerializeMap for ExtractDocumentIdMapSerializer<'a> {
 }
 
 pub struct ExtractDocumentIdStructSerializer<'a> {
-    identifier: &'a str,
+    primary_key: &'a str,
     document_id: Option<DocumentId>,
 }
 
@@ -278,7 +293,7 @@ impl<'a> ser::SerializeStruct for ExtractDocumentIdStructSerializer<'a> {
     where
         T: Serialize,
     {
-        if self.identifier == key {
+        if self.primary_key == key {
             let value = serde_json::to_string(value).and_then(|s| serde_json::from_str(&s))?;
             match value_to_string(&value).map(compute_document_id) {
                 Some(document_id) => self.document_id = Some(document_id),

meilisearch-core/src/serde/indexer.rs

@@ -1,4 +1,4 @@
-use meilidb_schema::SchemaAttr;
+use meilisearch_schema::IndexedPos;
 use serde::ser;
 use serde::Serialize;
 
@@ -7,7 +7,7 @@ use crate::raw_indexer::RawIndexer;
 use crate::DocumentId;
 
 pub struct Indexer<'a> {
-    pub attribute: SchemaAttr,
+    pub pos: IndexedPos,
     pub indexer: &'a mut RawIndexer,
     pub document_id: DocumentId,
 }
@@ -20,13 +20,11 @@ impl<'a> ser::Serializer for Indexer<'a> {
     type SerializeTupleStruct = ser::Impossible<Self::Ok, Self::Error>;
     type SerializeTupleVariant = ser::Impossible<Self::Ok, Self::Error>;
     type SerializeMap = MapIndexer<'a>;
-    type SerializeStruct = StructSerializer<'a>;
+    type SerializeStruct = StructIndexer<'a>;
     type SerializeStructVariant = ser::Impossible<Self::Ok, Self::Error>;
 
     fn serialize_bool(self, _value: bool) -> Result<Self::Ok, Self::Error> {
-        Err(SerializerError::UnindexableType {
-            type_name: "boolean",
-        })
+        Ok(None)
     }
 
     fn serialize_char(self, value: char) -> Result<Self::Ok, Self::Error> {
@@ -87,7 +85,7 @@ impl<'a> ser::Serializer for Indexer<'a> {
     fn serialize_str(self, text: &str) -> Result<Self::Ok, Self::Error> {
         let number_of_words = self
             .indexer
-            .index_text(self.document_id, self.attribute, text);
+            .index_text(self.document_id, self.pos, text);
         Ok(Some(number_of_words))
     }
 
@@ -96,9 +94,7 @@ impl<'a> ser::Serializer for Indexer<'a> {
     }
 
     fn serialize_none(self) -> Result<Self::Ok, Self::Error> {
-        Err(SerializerError::UnindexableType {
-            type_name: "Option",
-        })
+        Ok(None)
     }
 
     fn serialize_some<T: ?Sized>(self, value: &T) -> Result<Self::Ok, Self::Error>
@@ -108,18 +104,16 @@ impl<'a> ser::Serializer for Indexer<'a> {
         let text = value.serialize(ConvertToString)?;
         let number_of_words = self
             .indexer
-            .index_text(self.document_id, self.attribute, &text);
+            .index_text(self.document_id, self.pos, &text);
         Ok(Some(number_of_words))
     }
 
     fn serialize_unit(self) -> Result<Self::Ok, Self::Error> {
-        Err(SerializerError::UnindexableType { type_name: "()" })
+        Ok(None)
     }
 
     fn serialize_unit_struct(self, _name: &'static str) -> Result<Self::Ok, Self::Error> {
-        Err(SerializerError::UnindexableType {
-            type_name: "unit struct",
-        })
+        Ok(None)
     }
 
     fn serialize_unit_variant(
@@ -128,9 +122,7 @@ impl<'a> ser::Serializer for Indexer<'a> {
         _variant_index: u32,
         _variant: &'static str,
     ) -> Result<Self::Ok, Self::Error> {
-        Err(SerializerError::UnindexableType {
-            type_name: "unit variant",
-        })
+        Ok(None)
     }
 
     fn serialize_newtype_struct<T: ?Sized>(
@@ -161,7 +153,7 @@ impl<'a> ser::Serializer for Indexer<'a> {
 
     fn serialize_seq(self, _len: Option<usize>) -> Result<Self::SerializeSeq, Self::Error> {
         let indexer = SeqIndexer {
-            attribute: self.attribute,
+            pos: self.pos,
             document_id: self.document_id,
             indexer: self.indexer,
             texts: Vec::new(),
@@ -172,7 +164,7 @@ impl<'a> ser::Serializer for Indexer<'a> {
 
     fn serialize_tuple(self, _len: usize) -> Result<Self::SerializeTuple, Self::Error> {
         let indexer = TupleIndexer {
-            attribute: self.attribute,
+            pos: self.pos,
             document_id: self.document_id,
             indexer: self.indexer,
             texts: Vec::new(),
@@ -205,7 +197,7 @@ impl<'a> ser::Serializer for Indexer<'a> {
 
     fn serialize_map(self, _len: Option<usize>) -> Result<Self::SerializeMap, Self::Error> {
         let indexer = MapIndexer {
-            attribute: self.attribute,
+            pos: self.pos,
             document_id: self.document_id,
             indexer: self.indexer,
             texts: Vec::new(),
@@ -219,9 +211,14 @@ impl<'a> ser::Serializer for Indexer<'a> {
         _name: &'static str,
         _len: usize,
     ) -> Result<Self::SerializeStruct, Self::Error> {
-        Err(SerializerError::UnindexableType {
-            type_name: "struct",
-        })
+        let indexer = StructIndexer {
+            pos: self.pos,
+            document_id: self.document_id,
+            indexer: self.indexer,
+            texts: Vec::new(),
+        };
+
+        Ok(indexer)
     }
 
     fn serialize_struct_variant(
@@ -238,7 +235,7 @@ impl<'a> ser::Serializer for Indexer<'a> {
 }
 
 pub struct SeqIndexer<'a> {
-    attribute: SchemaAttr,
+    pos: IndexedPos,
     document_id: DocumentId,
     indexer: &'a mut RawIndexer,
     texts: Vec<String>,
@@ -260,13 +257,13 @@ impl<'a> ser::SerializeSeq for SeqIndexer<'a> {
     fn end(self) -> Result<Self::Ok, Self::Error> {
         let texts = self.texts.iter().map(String::as_str);
         self.indexer
-            .index_text_seq(self.document_id, self.attribute, texts);
+            .index_text_seq(self.document_id, self.pos, texts);
         Ok(None)
     }
 }
 
 pub struct MapIndexer<'a> {
-    attribute: SchemaAttr,
+    pos: IndexedPos,
     document_id: DocumentId,
     indexer: &'a mut RawIndexer,
     texts: Vec<String>,
@@ -297,19 +294,19 @@ impl<'a> ser::SerializeMap for MapIndexer<'a> {
     fn end(self) -> Result<Self::Ok, Self::Error> {
         let texts = self.texts.iter().map(String::as_str);
         self.indexer
-            .index_text_seq(self.document_id, self.attribute, texts);
+            .index_text_seq(self.document_id, self.pos, texts);
         Ok(None)
     }
 }
 
-pub struct StructSerializer<'a> {
-    attribute: SchemaAttr,
+pub struct StructIndexer<'a> {
+    pos: IndexedPos,
     document_id: DocumentId,
     indexer: &'a mut RawIndexer,
     texts: Vec<String>,
 }
 
-impl<'a> ser::SerializeStruct for StructSerializer<'a> {
+impl<'a> ser::SerializeStruct for StructIndexer<'a> {
     type Ok = Option<usize>;
     type Error = SerializerError;
 
@@ -331,13 +328,13 @@ impl<'a> ser::SerializeStruct for StructSerializer<'a> {
     fn end(self) -> Result<Self::Ok, Self::Error> {
         let texts = self.texts.iter().map(String::as_str);
         self.indexer
-            .index_text_seq(self.document_id, self.attribute, texts);
+            .index_text_seq(self.document_id, self.pos, texts);
         Ok(None)
     }
 }
 
 pub struct TupleIndexer<'a> {
-    attribute: SchemaAttr,
+    pos: IndexedPos,
     document_id: DocumentId,
     indexer: &'a mut RawIndexer,
     texts: Vec<String>,
@@ -359,7 +356,7 @@ impl<'a> ser::SerializeTuple for TupleIndexer<'a> {
     fn end(self) -> Result<Self::Ok, Self::Error> {
         let texts = self.texts.iter().map(String::as_str);
         self.indexer
-            .index_text_seq(self.document_id, self.attribute, texts);
+            .index_text_seq(self.document_id, self.pos, texts);
         Ok(None)
     }
 }

meilisearch-core/src/serde/mod.rs

@@ -20,16 +20,15 @@ pub use self::convert_to_string::ConvertToString;
 pub use self::deserializer::{Deserializer, DeserializerError};
 pub use self::extract_document_id::{compute_document_id, extract_document_id, value_to_string};
 pub use self::indexer::Indexer;
-pub use self::serializer::Serializer;
+pub use self::serializer::{serialize_value, serialize_value_with_id, Serializer};
 
-use std::collections::BTreeMap;
 use std::{error::Error, fmt};
 
-use meilidb_schema::SchemaAttr;
 use serde::ser;
 use serde_json::Error as SerdeJsonError;
+use meilisearch_schema::Error as SchemaError;
 
-use crate::{DocumentId, ParseNumberError};
+use crate::ParseNumberError;
 
 #[derive(Debug)]
 pub enum SerializerError {
@@ -38,6 +37,7 @@ pub enum SerializerError {
     Zlmdb(heed::Error),
     SerdeJson(SerdeJsonError),
     ParseNumber(ParseNumberError),
+    Schema(SchemaError),
     UnserializableType { type_name: &'static str },
     UnindexableType { type_name: &'static str },
     UnrankableType { type_name: &'static str },
@@ -57,13 +57,14 @@ impl fmt::Display for SerializerError {
                 f.write_str("serialized document does not have an id according to the schema")
             }
             SerializerError::InvalidDocumentIdType => {
-                f.write_str("document identifier can only be of type string or number")
+                f.write_str("a document primary key can be of type integer or string only composed of alphanumeric characters, hyphens (-) and underscores (_).")
             }
             SerializerError::Zlmdb(e) => write!(f, "heed related error: {}", e),
             SerializerError::SerdeJson(e) => write!(f, "serde json error: {}", e),
             SerializerError::ParseNumber(e) => {
                 write!(f, "error while trying to parse a number: {}", e)
             }
+            SerializerError::Schema(e) => write!(f, "impossible to update schema: {}", e),
             SerializerError::UnserializableType { type_name } => {
                 write!(f, "{} is not a serializable type", type_name)
             }
@@ -104,24 +105,8 @@ impl From<ParseNumberError> for SerializerError {
     }
 }
 
-pub struct RamDocumentStore(BTreeMap<(DocumentId, SchemaAttr), Vec<u8>>);
-
-impl RamDocumentStore {
-    pub fn new() -> RamDocumentStore {
-        RamDocumentStore(BTreeMap::new())
-    }
-
-    pub fn set_document_field(&mut self, id: DocumentId, attr: SchemaAttr, value: Vec<u8>) {
-        self.0.insert((id, attr), value);
-    }
-
-    pub fn into_inner(self) -> BTreeMap<(DocumentId, SchemaAttr), Vec<u8>> {
-        self.0
-    }
-}
-
-impl Default for RamDocumentStore {
-    fn default() -> Self {
-        Self::new()
-    }
+impl From<SchemaError> for SerializerError {
+    fn from(error: SchemaError) -> SerializerError {
+        SerializerError::Schema(error)
+   }
 }

meilisearch-core/src/serde/serializer.rs

@@ -1,31 +1,32 @@
-use meilidb_schema::{Schema, SchemaAttr};
+use meilisearch_schema::{Schema, FieldId};
 use serde::ser;
-use std::collections::HashMap;
 
+use crate::database::MainT;
 use crate::raw_indexer::RawIndexer;
-use crate::serde::RamDocumentStore;
+use crate::store::{DocumentsFields, DocumentsFieldsCounts};
 use crate::{DocumentId, RankedMap};
 
 use super::{ConvertToNumber, ConvertToString, Indexer, SerializerError};
 
-pub struct Serializer<'a> {
-    pub schema: &'a Schema,
-    pub document_store: &'a mut RamDocumentStore,
-    pub document_fields_counts: &'a mut HashMap<(DocumentId, SchemaAttr), u64>,
+pub struct Serializer<'a, 'b> {
+    pub txn: &'a mut heed::RwTxn<'b, MainT>,
+    pub schema: &'a mut Schema,
+    pub document_store: DocumentsFields,
+    pub document_fields_counts: DocumentsFieldsCounts,
     pub indexer: &'a mut RawIndexer,
     pub ranked_map: &'a mut RankedMap,
     pub document_id: DocumentId,
 }
 
-impl<'a> ser::Serializer for Serializer<'a> {
+impl<'a, 'b> ser::Serializer for Serializer<'a, 'b> {
     type Ok = ();
     type Error = SerializerError;
     type SerializeSeq = ser::Impossible<Self::Ok, Self::Error>;
     type SerializeTuple = ser::Impossible<Self::Ok, Self::Error>;
     type SerializeTupleStruct = ser::Impossible<Self::Ok, Self::Error>;
     type SerializeTupleVariant = ser::Impossible<Self::Ok, Self::Error>;
-    type SerializeMap = MapSerializer<'a>;
-    type SerializeStruct = StructSerializer<'a>;
+    type SerializeMap = MapSerializer<'a, 'b>;
+    type SerializeStruct = StructSerializer<'a, 'b>;
     type SerializeStructVariant = ser::Impossible<Self::Ok, Self::Error>;
 
     forward_to_unserializable_type! {
@@ -150,6 +151,7 @@ impl<'a> ser::Serializer for Serializer<'a> {
 
     fn serialize_map(self, _len: Option<usize>) -> Result<Self::SerializeMap, Self::Error> {
         Ok(MapSerializer {
+            txn: self.txn,
             schema: self.schema,
             document_id: self.document_id,
             document_store: self.document_store,
@@ -166,6 +168,7 @@ impl<'a> ser::Serializer for Serializer<'a> {
         _len: usize,
     ) -> Result<Self::SerializeStruct, Self::Error> {
         Ok(StructSerializer {
+            txn: self.txn,
             schema: self.schema,
             document_id: self.document_id,
             document_store: self.document_store,
@@ -188,17 +191,18 @@ impl<'a> ser::Serializer for Serializer<'a> {
     }
 }
 
-pub struct MapSerializer<'a> {
-    schema: &'a Schema,
+pub struct MapSerializer<'a, 'b> {
+    txn: &'a mut heed::RwTxn<'b, MainT>,
+    schema: &'a mut Schema,
     document_id: DocumentId,
-    document_store: &'a mut RamDocumentStore,
-    document_fields_counts: &'a mut HashMap<(DocumentId, SchemaAttr), u64>,
+    document_store: DocumentsFields,
+    document_fields_counts: DocumentsFieldsCounts,
     indexer: &'a mut RawIndexer,
     ranked_map: &'a mut RankedMap,
     current_key_name: Option<String>,
 }
 
-impl<'a> ser::SerializeMap for MapSerializer<'a> {
+impl<'a, 'b> ser::SerializeMap for MapSerializer<'a, 'b> {
     type Ok = ();
     type Error = SerializerError;
 
@@ -229,15 +233,15 @@ impl<'a> ser::SerializeMap for MapSerializer<'a> {
         V: ser::Serialize,
     {
         let key = key.serialize(ConvertToString)?;
-
         serialize_value(
+            self.txn,
+            key.as_str(),
             self.schema,
             self.document_id,
             self.document_store,
             self.document_fields_counts,
             self.indexer,
             self.ranked_map,
-            &key,
             value,
         )
     }
@@ -247,16 +251,17 @@ impl<'a> ser::SerializeMap for MapSerializer<'a> {
     }
 }
 
-pub struct StructSerializer<'a> {
-    schema: &'a Schema,
+pub struct StructSerializer<'a, 'b> {
+    txn: &'a mut heed::RwTxn<'b, MainT>,
+    schema: &'a mut Schema,
     document_id: DocumentId,
-    document_store: &'a mut RamDocumentStore,
-    document_fields_counts: &'a mut HashMap<(DocumentId, SchemaAttr), u64>,
+    document_store: DocumentsFields,
+    document_fields_counts: DocumentsFieldsCounts,
     indexer: &'a mut RawIndexer,
     ranked_map: &'a mut RankedMap,
 }
 
-impl<'a> ser::SerializeStruct for StructSerializer<'a> {
+impl<'a, 'b> ser::SerializeStruct for StructSerializer<'a, 'b> {
     type Ok = ();
     type Error = SerializerError;
 
@@ -269,13 +274,14 @@ impl<'a> ser::SerializeStruct for StructSerializer<'a> {
         T: ser::Serialize,
     {
         serialize_value(
+            self.txn,
+            key,
             self.schema,
             self.document_id,
             self.document_store,
             self.document_fields_counts,
             self.indexer,
             self.ranked_map,
-            key,
             value,
         )
     }
@@ -285,40 +291,70 @@ impl<'a> ser::SerializeStruct for StructSerializer<'a> {
     }
 }
 
-fn serialize_value<T: ?Sized>(
-    schema: &Schema,
+pub fn serialize_value<'a, T: ?Sized>(
+    txn: &mut heed::RwTxn<MainT>,
+    attribute: &str,
+    schema: &'a mut Schema,
     document_id: DocumentId,
-    document_store: &mut RamDocumentStore,
-    documents_fields_counts: &mut HashMap<(DocumentId, SchemaAttr), u64>,
+    document_store: DocumentsFields,
+    documents_fields_counts: DocumentsFieldsCounts,
     indexer: &mut RawIndexer,
     ranked_map: &mut RankedMap,
-    key: &str,
     value: &T,
 ) -> Result<(), SerializerError>
 where
     T: ser::Serialize,
 {
-    if let Some(attribute) = schema.attribute(key) {
-        let props = schema.props(attribute);
+    let field_id = schema.insert_and_index(&attribute)?;
+    serialize_value_with_id(
+        txn,
+        field_id,
+        schema,
+        document_id,
+        document_store,
+        documents_fields_counts,
+        indexer,
+        ranked_map,
+        value,
+    )
+}
 
-        let serialized = serde_json::to_vec(value)?;
-        document_store.set_document_field(document_id, attribute, serialized);
+pub fn serialize_value_with_id<'a, T: ?Sized>(
+    txn: &mut heed::RwTxn<MainT>,
+    field_id: FieldId,
+    schema: &'a Schema,
+    document_id: DocumentId,
+    document_store: DocumentsFields,
+    documents_fields_counts: DocumentsFieldsCounts,
+    indexer: &mut RawIndexer,
+    ranked_map: &mut RankedMap,
+    value: &T,
+) -> Result<(), SerializerError>
+where
+    T: ser::Serialize,
+{
+    let serialized = serde_json::to_vec(value)?;
+    document_store.put_document_field(txn, document_id, field_id, &serialized)?;
 
-        if props.is_indexed() {
-            let indexer = Indexer {
-                attribute,
-                indexer,
+    if let Some(indexed_pos) = schema.is_indexed(field_id) {
+        let indexer = Indexer {
+            pos: *indexed_pos,
+            indexer,
+            document_id,
+        };
+        if let Some(number_of_words) = value.serialize(indexer)? {
+            documents_fields_counts.put_document_field_count(
+                txn,
                 document_id,
-            };
-            if let Some(number_of_words) = value.serialize(indexer)? {
-                documents_fields_counts.insert((document_id, attribute), number_of_words as u64);
-            }
+                *indexed_pos,
+                number_of_words as u16,
+            )?;
         }
+    }
 
-        if props.is_ranked() {
-            let number = value.serialize(ConvertToNumber)?;
-            ranked_map.insert(document_id, attribute, number);
-        }
+    if schema.is_ranked(field_id) {
+        let number = value.serialize(ConvertToNumber).unwrap_or_default();
+        ranked_map.insert(document_id, field_id, number);
     }
 
     Ok(())

meilisearch-core/src/settings.rs

@@ -0,0 +1,184 @@
+use std::collections::{BTreeMap, BTreeSet, HashSet};
+use std::str::FromStr;
+use std::iter::IntoIterator;
+
+use serde::{Deserialize, Deserializer, Serialize};
+use once_cell::sync::Lazy;
+
+use self::RankingRule::*;
+
+pub const DEFAULT_RANKING_RULES: [RankingRule; 6] = [Typo, Words, Proximity, Attribute, WordsPosition, Exactness];
+
+static RANKING_RULE_REGEX: Lazy<regex::Regex> = Lazy::new(|| {
+    let regex = regex::Regex::new(r"(asc|desc)\(([a-zA-Z0-9-_]*)\)").unwrap();
+    regex
+});
+
+#[derive(Default, Clone, Serialize, Deserialize)]
+#[serde(rename_all = "camelCase", deny_unknown_fields)]
+pub struct Settings {
+    #[serde(default, deserialize_with = "deserialize_some")]
+    pub ranking_rules: Option<Option<Vec<String>>>,
+    #[serde(default, deserialize_with = "deserialize_some")]
+    pub distinct_attribute: Option<Option<String>>,
+    #[serde(default, deserialize_with = "deserialize_some")]
+    pub searchable_attributes: Option<Option<Vec<String>>>,
+    #[serde(default, deserialize_with = "deserialize_some")]
+    pub displayed_attributes: Option<Option<HashSet<String>>>,
+    #[serde(default, deserialize_with = "deserialize_some")]
+    pub stop_words: Option<Option<BTreeSet<String>>>,
+    #[serde(default, deserialize_with = "deserialize_some")]
+    pub synonyms: Option<Option<BTreeMap<String, Vec<String>>>>,
+    #[serde(default, deserialize_with = "deserialize_some")]
+    pub accept_new_fields: Option<Option<bool>>,
+}
+
+// Any value that is present is considered Some value, including null.
+fn deserialize_some<'de, T, D>(deserializer: D) -> Result<Option<T>, D::Error>
+    where T: Deserialize<'de>,
+          D: Deserializer<'de>
+{
+    Deserialize::deserialize(deserializer).map(Some)
+}
+
+impl Settings {
+    pub fn into_update(&self) -> Result<SettingsUpdate, RankingRuleConversionError> {
+        let settings = self.clone();
+
+        let ranking_rules = match settings.ranking_rules {
+            Some(Some(rules)) => UpdateState::Update(RankingRule::from_iter(rules.iter())?),
+            Some(None) => UpdateState::Clear,
+            None => UpdateState::Nothing,
+        };
+
+        Ok(SettingsUpdate {
+            ranking_rules,
+            distinct_attribute: settings.distinct_attribute.into(),
+            primary_key: UpdateState::Nothing,
+            searchable_attributes: settings.searchable_attributes.into(),
+            displayed_attributes: settings.displayed_attributes.into(),
+            stop_words: settings.stop_words.into(),
+            synonyms: settings.synonyms.into(),
+            accept_new_fields: settings.accept_new_fields.into(),
+        })
+    }
+}
+
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub enum UpdateState<T> {
+    Update(T),
+    Clear,
+    Nothing,
+}
+
+impl <T> From<Option<Option<T>>> for UpdateState<T> {
+    fn from(opt: Option<Option<T>>) -> UpdateState<T> {
+        match opt {
+            Some(Some(t)) => UpdateState::Update(t),
+            Some(None) => UpdateState::Clear,
+            None => UpdateState::Nothing,
+        }
+    }
+}
+
+#[derive(Debug, Clone)]
+pub struct RankingRuleConversionError;
+
+impl std::fmt::Display for RankingRuleConversionError {
+    fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result {
+        write!(f, "impossible to convert into RankingRule")
+    }
+}
+
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub enum RankingRule {
+    Typo,
+    Words,
+    Proximity,
+    Attribute,
+    WordsPosition,
+    Exactness,
+    Asc(String),
+    Desc(String),
+}
+
+impl std::fmt::Display for RankingRule {
+    fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result {
+        match self {
+            RankingRule::Typo => f.write_str("typo"),
+            RankingRule::Words => f.write_str("words"),
+            RankingRule::Proximity => f.write_str("proximity"),
+            RankingRule::Attribute => f.write_str("attribute"),
+            RankingRule::WordsPosition => f.write_str("wordsPosition"),
+            RankingRule::Exactness => f.write_str("exactness"),
+            RankingRule::Asc(field) => write!(f, "asc({})", field),
+            RankingRule::Desc(field) => write!(f, "desc({})", field),
+        }
+    }
+}
+
+impl FromStr for RankingRule {
+    type Err = RankingRuleConversionError;
+
+    fn from_str(s: &str) -> Result<Self, Self::Err> {
+        let rule = match s {
+            "typo" => RankingRule::Typo,
+            "words" => RankingRule::Words,
+            "proximity" => RankingRule::Proximity,
+            "attribute" => RankingRule::Attribute,
+            "wordsPosition" => RankingRule::WordsPosition,
+            "exactness" => RankingRule::Exactness,
+            _ => {
+                let captures = RANKING_RULE_REGEX.captures(s).ok_or(RankingRuleConversionError)?;
+                match (captures.get(1).map(|m| m.as_str()), captures.get(2)) {
+                    (Some("asc"), Some(field)) => RankingRule::Asc(field.as_str().to_string()),
+                    (Some("desc"), Some(field)) => RankingRule::Desc(field.as_str().to_string()),
+                    _ => return Err(RankingRuleConversionError)
+                }
+            }
+        };
+        Ok(rule)
+    }
+}
+
+impl RankingRule {
+    pub fn field(&self) -> Option<&str> {
+        match self {
+            RankingRule::Asc(field) | RankingRule::Desc(field) => Some(field),
+            _ => None,
+        }
+    }
+
+    pub fn from_iter(rules: impl IntoIterator<Item = impl AsRef<str>>) -> Result<Vec<RankingRule>, RankingRuleConversionError> {
+        rules.into_iter()
+            .map(|s| RankingRule::from_str(s.as_ref()))
+            .collect()
+    }
+}
+
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct SettingsUpdate {
+    pub ranking_rules: UpdateState<Vec<RankingRule>>,
+    pub distinct_attribute: UpdateState<String>,
+    pub primary_key: UpdateState<String>,
+    pub searchable_attributes: UpdateState<Vec<String>>,
+    pub displayed_attributes: UpdateState<HashSet<String>>,
+    pub stop_words: UpdateState<BTreeSet<String>>,
+    pub synonyms: UpdateState<BTreeMap<String, Vec<String>>>,
+    pub accept_new_fields: UpdateState<bool>,
+}
+
+impl Default for SettingsUpdate {
+    fn default() -> Self {
+        Self {
+            ranking_rules: UpdateState::Nothing,
+            distinct_attribute: UpdateState::Nothing,
+            primary_key: UpdateState::Nothing,
+            searchable_attributes: UpdateState::Nothing,
+            displayed_attributes: UpdateState::Nothing,
+            stop_words: UpdateState::Nothing,
+            synonyms: UpdateState::Nothing,
+            accept_new_fields: UpdateState::Nothing,
+        }
+    }
+}

meilisearch-core/src/store/docs_words.rs

@@ -1,4 +1,5 @@
 use super::BEU64;
+use crate::database::MainT;
 use crate::DocumentId;
 use heed::types::{ByteSlice, OwnedType};
 use heed::Result as ZResult;
@@ -12,7 +13,7 @@ pub struct DocsWords {
 impl DocsWords {
     pub fn put_doc_words(
         self,
-        writer: &mut heed::RwTxn,
+        writer: &mut heed::RwTxn<MainT>,
         document_id: DocumentId,
         words: &fst::Set,
     ) -> ZResult<()> {
@@ -21,21 +22,25 @@ impl DocsWords {
         self.docs_words.put(writer, &document_id, bytes)
     }
 
-    pub fn del_doc_words(self, writer: &mut heed::RwTxn, document_id: DocumentId) -> ZResult<bool> {
+    pub fn del_doc_words(self, writer: &mut heed::RwTxn<MainT>, document_id: DocumentId) -> ZResult<bool> {
         let document_id = BEU64::new(document_id.0);
         self.docs_words.delete(writer, &document_id)
     }
 
+    pub fn clear(self, writer: &mut heed::RwTxn<MainT>) -> ZResult<()> {
+        self.docs_words.clear(writer)
+    }
+
     pub fn doc_words(
         self,
-        reader: &heed::RoTxn,
+        reader: &heed::RoTxn<MainT>,
         document_id: DocumentId,
     ) -> ZResult<Option<fst::Set>> {
         let document_id = BEU64::new(document_id.0);
         match self.docs_words.get(reader, &document_id)? {
             Some(bytes) => {
                 let len = bytes.len();
-                let bytes = Arc::from(bytes);
+                let bytes = Arc::new(bytes.to_owned());
                 let fst = fst::raw::Fst::from_shared_bytes(bytes, 0, len).unwrap();
                 Ok(Some(fst::Set::from(fst)))
             }

meilisearch-core/src/store/documents_fields.rs

@@ -0,0 +1,79 @@
+use heed::types::{ByteSlice, OwnedType};
+use crate::database::MainT;
+use heed::Result as ZResult;
+use meilisearch_schema::FieldId;
+
+use super::DocumentFieldStoredKey;
+use crate::DocumentId;
+
+#[derive(Copy, Clone)]
+pub struct DocumentsFields {
+    pub(crate) documents_fields: heed::Database<OwnedType<DocumentFieldStoredKey>, ByteSlice>,
+}
+
+impl DocumentsFields {
+    pub fn put_document_field(
+        self,
+        writer: &mut heed::RwTxn<MainT>,
+        document_id: DocumentId,
+        field: FieldId,
+        value: &[u8],
+    ) -> ZResult<()> {
+        let key = DocumentFieldStoredKey::new(document_id, field);
+        self.documents_fields.put(writer, &key, value)
+    }
+
+    pub fn del_all_document_fields(
+        self,
+        writer: &mut heed::RwTxn<MainT>,
+        document_id: DocumentId,
+    ) -> ZResult<usize> {
+        let start = DocumentFieldStoredKey::new(document_id, FieldId::min());
+        let end = DocumentFieldStoredKey::new(document_id, FieldId::max());
+        self.documents_fields.delete_range(writer, &(start..=end))
+    }
+
+    pub fn clear(self, writer: &mut heed::RwTxn<MainT>) -> ZResult<()> {
+        self.documents_fields.clear(writer)
+    }
+
+    pub fn document_attribute<'txn>(
+        self,
+        reader: &'txn heed::RoTxn<MainT>,
+        document_id: DocumentId,
+        field: FieldId,
+    ) -> ZResult<Option<&'txn [u8]>> {
+        let key = DocumentFieldStoredKey::new(document_id, field);
+        self.documents_fields.get(reader, &key)
+    }
+
+    pub fn document_fields<'txn>(
+        self,
+        reader: &'txn heed::RoTxn<MainT>,
+        document_id: DocumentId,
+    ) -> ZResult<DocumentFieldsIter<'txn>> {
+        let start = DocumentFieldStoredKey::new(document_id, FieldId::min());
+        let end = DocumentFieldStoredKey::new(document_id, FieldId::max());
+        let iter = self.documents_fields.range(reader, &(start..=end))?;
+        Ok(DocumentFieldsIter { iter })
+    }
+}
+
+pub struct DocumentFieldsIter<'txn> {
+    iter: heed::RoRange<'txn, OwnedType<DocumentFieldStoredKey>, ByteSlice>,
+}
+
+impl<'txn> Iterator for DocumentFieldsIter<'txn> {
+    type Item = ZResult<(FieldId, &'txn [u8])>;
+
+    fn next(&mut self) -> Option<Self::Item> {
+        match self.iter.next() {
+            Some(Ok((key, bytes))) => {
+                let field_id = FieldId(key.field_id.get());
+                Some(Ok((field_id, bytes)))
+            }
+            Some(Err(e)) => Some(Err(e)),
+            None => None,
+        }
+    }
+}

meilisearch-core/src/store/documents_fields_counts.rs

@@ -1,44 +1,48 @@
-use super::DocumentAttrKey;
+use super::DocumentFieldIndexedKey;
+use crate::database::MainT;
 use crate::DocumentId;
 use heed::types::OwnedType;
 use heed::Result as ZResult;
-use meilidb_schema::SchemaAttr;
+use meilisearch_schema::IndexedPos;
 
 #[derive(Copy, Clone)]
 pub struct DocumentsFieldsCounts {
-    pub(crate) documents_fields_counts: heed::Database<OwnedType<DocumentAttrKey>, OwnedType<u64>>,
+    pub(crate) documents_fields_counts: heed::Database<OwnedType<DocumentFieldIndexedKey>, OwnedType<u16>>,
 }
 
 impl DocumentsFieldsCounts {
     pub fn put_document_field_count(
         self,
-        writer: &mut heed::RwTxn,
+        writer: &mut heed::RwTxn<MainT>,
         document_id: DocumentId,
-        attribute: SchemaAttr,
-        value: u64,
+        attribute: IndexedPos,
+        value: u16,
     ) -> ZResult<()> {
-        let key = DocumentAttrKey::new(document_id, attribute);
+        let key = DocumentFieldIndexedKey::new(document_id, attribute);
         self.documents_fields_counts.put(writer, &key, &value)
     }
 
     pub fn del_all_document_fields_counts(
         self,
-        writer: &mut heed::RwTxn,
+        writer: &mut heed::RwTxn<MainT>,
         document_id: DocumentId,
     ) -> ZResult<usize> {
-        let start = DocumentAttrKey::new(document_id, SchemaAttr::min());
-        let end = DocumentAttrKey::new(document_id, SchemaAttr::max());
-        self.documents_fields_counts
-            .delete_range(writer, start..=end)
+        let start = DocumentFieldIndexedKey::new(document_id, IndexedPos::min());
+        let end = DocumentFieldIndexedKey::new(document_id, IndexedPos::max());
+        self.documents_fields_counts.delete_range(writer, &(start..=end))
+    }
+
+    pub fn clear(self, writer: &mut heed::RwTxn<MainT>) -> ZResult<()> {
+        self.documents_fields_counts.clear(writer)
     }
 
     pub fn document_field_count(
         self,
-        reader: &heed::RoTxn,
+        reader: &heed::RoTxn<MainT>,
         document_id: DocumentId,
-        attribute: SchemaAttr,
-    ) -> ZResult<Option<u64>> {
-        let key = DocumentAttrKey::new(document_id, attribute);
+        attribute: IndexedPos,
+    ) -> ZResult<Option<u16>> {
+        let key = DocumentFieldIndexedKey::new(document_id, attribute);
         match self.documents_fields_counts.get(reader, &key)? {
             Some(count) => Ok(Some(count)),
             None => Ok(None),
@@ -47,16 +51,16 @@ impl DocumentsFieldsCounts {
 
     pub fn document_fields_counts<'txn>(
         self,
-        reader: &'txn heed::RoTxn,
+        reader: &'txn heed::RoTxn<MainT>,
         document_id: DocumentId,
     ) -> ZResult<DocumentFieldsCountsIter<'txn>> {
-        let start = DocumentAttrKey::new(document_id, SchemaAttr::min());
-        let end = DocumentAttrKey::new(document_id, SchemaAttr::max());
-        let iter = self.documents_fields_counts.range(reader, start..=end)?;
+        let start = DocumentFieldIndexedKey::new(document_id, IndexedPos::min());
+        let end = DocumentFieldIndexedKey::new(document_id, IndexedPos::max());
+        let iter = self.documents_fields_counts.range(reader, &(start..=end))?;
         Ok(DocumentFieldsCountsIter { iter })
     }
 
-    pub fn documents_ids<'txn>(self, reader: &'txn heed::RoTxn) -> ZResult<DocumentsIdsIter<'txn>> {
+    pub fn documents_ids<'txn>(self, reader: &'txn heed::RoTxn<MainT>) -> ZResult<DocumentsIdsIter<'txn>> {
         let iter = self.documents_fields_counts.iter(reader)?;
         Ok(DocumentsIdsIter {
             last_seen_id: None,
@@ -66,7 +70,7 @@ impl DocumentsFieldsCounts {
 
     pub fn all_documents_fields_counts<'txn>(
         self,
-        reader: &'txn heed::RoTxn,
+        reader: &'txn heed::RoTxn<MainT>,
     ) -> ZResult<AllDocumentsFieldsCountsIter<'txn>> {
         let iter = self.documents_fields_counts.iter(reader)?;
         Ok(AllDocumentsFieldsCountsIter { iter })
@@ -74,17 +78,17 @@ impl DocumentsFieldsCounts {
 }
 
 pub struct DocumentFieldsCountsIter<'txn> {
-    iter: heed::RoRange<'txn, OwnedType<DocumentAttrKey>, OwnedType<u64>>,
+    iter: heed::RoRange<'txn, OwnedType<DocumentFieldIndexedKey>, OwnedType<u16>>,
 }
 
 impl Iterator for DocumentFieldsCountsIter<'_> {
-    type Item = ZResult<(SchemaAttr, u64)>;
+    type Item = ZResult<(IndexedPos, u16)>;
 
     fn next(&mut self) -> Option<Self::Item> {
         match self.iter.next() {
             Some(Ok((key, count))) => {
-                let attr = SchemaAttr(key.attr.get());
-                Some(Ok((attr, count)))
+                let indexed_pos = IndexedPos(key.indexed_pos.get());
+                Some(Ok((indexed_pos, count)))
             }
             Some(Err(e)) => Some(Err(e)),
             None => None,
@@ -94,7 +98,7 @@ impl Iterator for DocumentFieldsCountsIter<'_> {
 
 pub struct DocumentsIdsIter<'txn> {
     last_seen_id: Option<DocumentId>,
-    iter: heed::RoIter<'txn, OwnedType<DocumentAttrKey>, OwnedType<u64>>,
+    iter: heed::RoIter<'txn, OwnedType<DocumentFieldIndexedKey>, OwnedType<u16>>,
 }
 
 impl Iterator for DocumentsIdsIter<'_> {
@@ -118,18 +122,18 @@ impl Iterator for DocumentsIdsIter<'_> {
 }
 
 pub struct AllDocumentsFieldsCountsIter<'txn> {
-    iter: heed::RoIter<'txn, OwnedType<DocumentAttrKey>, OwnedType<u64>>,
+    iter: heed::RoIter<'txn, OwnedType<DocumentFieldIndexedKey>, OwnedType<u16>>,
 }
 
-impl<'r> Iterator for AllDocumentsFieldsCountsIter<'r> {
-    type Item = ZResult<(DocumentId, SchemaAttr, u64)>;
+impl Iterator for AllDocumentsFieldsCountsIter<'_> {
+    type Item = ZResult<(DocumentId, IndexedPos, u16)>;
 
     fn next(&mut self) -> Option<Self::Item> {
         match self.iter.next() {
             Some(Ok((key, count))) => {
                 let docid = DocumentId(key.docid.get());
-                let attr = SchemaAttr(key.attr.get());
-                Some(Ok((docid, attr, count)))
+                let indexed_pos = IndexedPos(key.indexed_pos.get());
+                Some(Ok((docid, indexed_pos, count)))
             }
             Some(Err(e)) => Some(Err(e)),
             None => None,

meilisearch-core/src/store/main.rs

@@ -0,0 +1,226 @@
+use std::sync::Arc;
+use std::collections::HashMap;
+
+use chrono::{DateTime, Utc};
+use heed::types::{ByteSlice, OwnedType, SerdeBincode, Str};
+use heed::Result as ZResult;
+use meilisearch_schema::Schema;
+
+use crate::database::MainT;
+use crate::RankedMap;
+use crate::settings::RankingRule;
+
+const CREATED_AT_KEY: &str = "created-at";
+const RANKING_RULES_KEY: &str = "ranking-rules";
+const DISTINCT_ATTRIBUTE_KEY: &str = "distinct-attribute";
+const STOP_WORDS_KEY: &str = "stop-words";
+const SYNONYMS_KEY: &str = "synonyms";
+const CUSTOMS_KEY: &str = "customs";
+const FIELDS_FREQUENCY_KEY: &str = "fields-frequency";
+const NAME_KEY: &str = "name";
+const NUMBER_OF_DOCUMENTS_KEY: &str = "number-of-documents";
+const RANKED_MAP_KEY: &str = "ranked-map";
+const SCHEMA_KEY: &str = "schema";
+const UPDATED_AT_KEY: &str = "updated-at";
+const WORDS_KEY: &str = "words";
+
+pub type FreqsMap = HashMap<String, usize>;
+type SerdeFreqsMap = SerdeBincode<FreqsMap>;
+type SerdeDatetime = SerdeBincode<DateTime<Utc>>;
+
+#[derive(Copy, Clone)]
+pub struct Main {
+    pub(crate) main: heed::PolyDatabase,
+}
+
+impl Main {
+    pub fn clear(self, writer: &mut heed::RwTxn<MainT>) -> ZResult<()> {
+        self.main.clear(writer)
+    }
+
+    pub fn put_name(self, writer: &mut heed::RwTxn<MainT>, name: &str) -> ZResult<()> {
+        self.main.put::<_, Str, Str>(writer, NAME_KEY, name)
+    }
+
+    pub fn name(self, reader: &heed::RoTxn<MainT>) -> ZResult<Option<String>> {
+        Ok(self
+            .main
+            .get::<_, Str, Str>(reader, NAME_KEY)?
+            .map(|name| name.to_owned()))
+    }
+
+    pub fn put_created_at(self, writer: &mut heed::RwTxn<MainT>) -> ZResult<()> {
+        self.main
+            .put::<_, Str, SerdeDatetime>(writer, CREATED_AT_KEY, &Utc::now())
+    }
+
+    pub fn created_at(self, reader: &heed::RoTxn<MainT>) -> ZResult<Option<DateTime<Utc>>> {
+        self.main.get::<_, Str, SerdeDatetime>(reader, CREATED_AT_KEY)
+    }
+
+    pub fn put_updated_at(self, writer: &mut heed::RwTxn<MainT>) -> ZResult<()> {
+        self.main
+            .put::<_, Str, SerdeDatetime>(writer, UPDATED_AT_KEY, &Utc::now())
+    }
+
+    pub fn updated_at(self, reader: &heed::RoTxn<MainT>) -> ZResult<Option<DateTime<Utc>>> {
+        self.main.get::<_, Str, SerdeDatetime>(reader, UPDATED_AT_KEY)
+    }
+
+    pub fn put_words_fst(self, writer: &mut heed::RwTxn<MainT>, fst: &fst::Set) -> ZResult<()> {
+        let bytes = fst.as_fst().as_bytes();
+        self.main.put::<_, Str, ByteSlice>(writer, WORDS_KEY, bytes)
+    }
+
+    pub unsafe fn static_words_fst(self, reader: &heed::RoTxn<MainT>) -> ZResult<Option<fst::Set>> {
+        match self.main.get::<_, Str, ByteSlice>(reader, WORDS_KEY)? {
+            Some(bytes) => {
+                let bytes: &'static [u8] = std::mem::transmute(bytes);
+                let set = fst::Set::from_static_slice(bytes).unwrap();
+                Ok(Some(set))
+            }
+            None => Ok(None),
+        }
+    }
+
+    pub fn words_fst(self, reader: &heed::RoTxn<MainT>) -> ZResult<Option<fst::Set>> {
+        match self.main.get::<_, Str, ByteSlice>(reader, WORDS_KEY)? {
+            Some(bytes) => {
+                let len = bytes.len();
+                let bytes = Arc::new(bytes.to_owned());
+                let fst = fst::raw::Fst::from_shared_bytes(bytes, 0, len).unwrap();
+                Ok(Some(fst::Set::from(fst)))
+            }
+            None => Ok(None),
+        }
+    }
+
+    pub fn put_schema(self, writer: &mut heed::RwTxn<MainT>, schema: &Schema) -> ZResult<()> {
+        self.main.put::<_, Str, SerdeBincode<Schema>>(writer, SCHEMA_KEY, schema)
+    }
+
+    pub fn schema(self, reader: &heed::RoTxn<MainT>) -> ZResult<Option<Schema>> {
+        self.main.get::<_, Str, SerdeBincode<Schema>>(reader, SCHEMA_KEY)
+    }
+
+    pub fn delete_schema(self, writer: &mut heed::RwTxn<MainT>) -> ZResult<bool> {
+        self.main.delete::<_, Str>(writer, SCHEMA_KEY)
+    }
+
+    pub fn put_ranked_map(self, writer: &mut heed::RwTxn<MainT>, ranked_map: &RankedMap) -> ZResult<()> {
+        self.main.put::<_, Str, SerdeBincode<RankedMap>>(writer, RANKED_MAP_KEY, &ranked_map)
+    }
+
+    pub fn ranked_map(self, reader: &heed::RoTxn<MainT>) -> ZResult<Option<RankedMap>> {
+        self.main.get::<_, Str, SerdeBincode<RankedMap>>(reader, RANKED_MAP_KEY)
+    }
+
+    pub fn put_synonyms_fst(self, writer: &mut heed::RwTxn<MainT>, fst: &fst::Set) -> ZResult<()> {
+        let bytes = fst.as_fst().as_bytes();
+        self.main.put::<_, Str, ByteSlice>(writer, SYNONYMS_KEY, bytes)
+    }
+
+    pub fn synonyms_fst(self, reader: &heed::RoTxn<MainT>) -> ZResult<Option<fst::Set>> {
+        match self.main.get::<_, Str, ByteSlice>(reader, SYNONYMS_KEY)? {
+            Some(bytes) => {
+                let len = bytes.len();
+                let bytes = Arc::new(bytes.to_owned());
+                let fst = fst::raw::Fst::from_shared_bytes(bytes, 0, len).unwrap();
+                Ok(Some(fst::Set::from(fst)))
+            }
+            None => Ok(None),
+        }
+    }
+
+    pub fn put_stop_words_fst(self, writer: &mut heed::RwTxn<MainT>, fst: &fst::Set) -> ZResult<()> {
+        let bytes = fst.as_fst().as_bytes();
+        self.main.put::<_, Str, ByteSlice>(writer, STOP_WORDS_KEY, bytes)
+    }
+
+    pub fn stop_words_fst(self, reader: &heed::RoTxn<MainT>) -> ZResult<Option<fst::Set>> {
+        match self.main.get::<_, Str, ByteSlice>(reader, STOP_WORDS_KEY)? {
+            Some(bytes) => {
+                let len = bytes.len();
+                let bytes = Arc::new(bytes.to_owned());
+                let fst = fst::raw::Fst::from_shared_bytes(bytes, 0, len).unwrap();
+                Ok(Some(fst::Set::from(fst)))
+            }
+            None => Ok(None),
+        }
+    }
+
+    pub fn put_number_of_documents<F>(self, writer: &mut heed::RwTxn<MainT>, f: F) -> ZResult<u64>
+    where
+        F: Fn(u64) -> u64,
+    {
+        let new = self.number_of_documents(&*writer).map(f)?;
+        self.main
+            .put::<_, Str, OwnedType<u64>>(writer, NUMBER_OF_DOCUMENTS_KEY, &new)?;
+        Ok(new)
+    }
+
+    pub fn number_of_documents(self, reader: &heed::RoTxn<MainT>) -> ZResult<u64> {
+        match self
+            .main
+            .get::<_, Str, OwnedType<u64>>(reader, NUMBER_OF_DOCUMENTS_KEY)?
+        {
+            Some(value) => Ok(value),
+            None => Ok(0),
+        }
+    }
+
+    pub fn put_fields_frequency(
+        self,
+        writer: &mut heed::RwTxn<MainT>,
+        fields_frequency: &FreqsMap,
+    ) -> ZResult<()> {
+        self.main
+            .put::<_, Str, SerdeFreqsMap>(writer, FIELDS_FREQUENCY_KEY, fields_frequency)
+    }
+
+    pub fn fields_frequency(&self, reader: &heed::RoTxn<MainT>) -> ZResult<Option<FreqsMap>> {
+        match self
+            .main
+            .get::<_, Str, SerdeFreqsMap>(reader, FIELDS_FREQUENCY_KEY)?
+        {
+            Some(freqs) => Ok(Some(freqs)),
+            None => Ok(None),
+        }
+    }
+
+    pub fn ranking_rules(&self, reader: &heed::RoTxn<MainT>) -> ZResult<Option<Vec<RankingRule>>> {
+        self.main.get::<_, Str, SerdeBincode<Vec<RankingRule>>>(reader, RANKING_RULES_KEY)
+    }
+
+    pub fn put_ranking_rules(self, writer: &mut heed::RwTxn<MainT>, value: &[RankingRule]) -> ZResult<()> {
+        self.main.put::<_, Str, SerdeBincode<Vec<RankingRule>>>(writer, RANKING_RULES_KEY, &value.to_vec())
+    }
+
+    pub fn delete_ranking_rules(self, writer: &mut heed::RwTxn<MainT>) -> ZResult<bool> {
+        self.main.delete::<_, Str>(writer, RANKING_RULES_KEY)
+    }
+
+    pub fn distinct_attribute(&self, reader: &heed::RoTxn<MainT>) -> ZResult<Option<String>> {
+        if let Some(value) = self.main.get::<_, Str, Str>(reader, DISTINCT_ATTRIBUTE_KEY)? {
+            return Ok(Some(value.to_owned()))
+        }
+        return Ok(None)
+    }
+
+    pub fn put_distinct_attribute(self, writer: &mut heed::RwTxn<MainT>, value: &str) -> ZResult<()> {
+        self.main.put::<_, Str, Str>(writer, DISTINCT_ATTRIBUTE_KEY, value)
+    }
+
+    pub fn delete_distinct_attribute(self, writer: &mut heed::RwTxn<MainT>) -> ZResult<bool> {
+        self.main.delete::<_, Str>(writer, DISTINCT_ATTRIBUTE_KEY)
+    }
+
+    pub fn put_customs(self, writer: &mut heed::RwTxn<MainT>, customs: &[u8]) -> ZResult<()> {
+        self.main
+            .put::<_, Str, ByteSlice>(writer, CUSTOMS_KEY, customs)
+    }
+
+    pub fn customs<'txn>(self, reader: &'txn heed::RoTxn<MainT>) -> ZResult<Option<&'txn [u8]>> {
+        self.main.get::<_, Str, ByteSlice>(reader, CUSTOMS_KEY)
+    }
+}

meilisearch-core/src/store/mod.rs

@@ -0,0 +1,518 @@
+mod docs_words;
+mod prefix_documents_cache;
+mod prefix_postings_lists_cache;
+mod documents_fields;
+mod documents_fields_counts;
+mod main;
+mod postings_lists;
+mod synonyms;
+mod updates;
+mod updates_results;
+
+pub use self::docs_words::DocsWords;
+pub use self::prefix_documents_cache::PrefixDocumentsCache;
+pub use self::prefix_postings_lists_cache::PrefixPostingsListsCache;
+pub use self::documents_fields::{DocumentFieldsIter, DocumentsFields};
+pub use self::documents_fields_counts::{
+    DocumentFieldsCountsIter, DocumentsFieldsCounts, DocumentsIdsIter,
+};
+pub use self::main::Main;
+pub use self::postings_lists::PostingsLists;
+pub use self::synonyms::Synonyms;
+pub use self::updates::Updates;
+pub use self::updates_results::UpdatesResults;
+
+use std::borrow::Cow;
+use std::collections::HashSet;
+use std::convert::TryInto;
+use std::{mem, ptr};
+
+use heed::Result as ZResult;
+use heed::{BytesEncode, BytesDecode};
+use meilisearch_schema::{IndexedPos, FieldId};
+use sdset::{Set, SetBuf};
+use serde::de::{self, Deserialize};
+use zerocopy::{AsBytes, FromBytes};
+
+use crate::criterion::Criteria;
+use crate::database::{MainT, UpdateT};
+use crate::database::{UpdateEvent, UpdateEventsEmitter};
+use crate::serde::Deserializer;
+use crate::settings::SettingsUpdate;
+use crate::{query_builder::QueryBuilder, update, DocIndex, DocumentId, Error, MResult};
+
+type BEU64 = zerocopy::U64<byteorder::BigEndian>;
+type BEU16 = zerocopy::U16<byteorder::BigEndian>;
+
+#[derive(Debug, Copy, Clone, AsBytes, FromBytes)]
+#[repr(C)]
+pub struct DocumentFieldIndexedKey {
+    docid: BEU64,
+    indexed_pos: BEU16,
+}
+
+impl DocumentFieldIndexedKey {
+    fn new(docid: DocumentId, indexed_pos: IndexedPos) -> DocumentFieldIndexedKey {
+        DocumentFieldIndexedKey {
+            docid: BEU64::new(docid.0),
+            indexed_pos: BEU16::new(indexed_pos.0),
+        }
+    }
+}
+
+#[derive(Debug, Copy, Clone, AsBytes, FromBytes)]
+#[repr(C)]
+pub struct DocumentFieldStoredKey {
+    docid: BEU64,
+    field_id: BEU16,
+}
+
+impl DocumentFieldStoredKey {
+    fn new(docid: DocumentId, field_id: FieldId) -> DocumentFieldStoredKey {
+        DocumentFieldStoredKey {
+            docid: BEU64::new(docid.0),
+            field_id: BEU16::new(field_id.0),
+        }
+    }
+}
+
+#[derive(Default, Debug)]
+pub struct Postings<'a> {
+    pub docids: Cow<'a, Set<DocumentId>>,
+    pub matches: Cow<'a, Set<DocIndex>>,
+}
+
+pub struct PostingsCodec;
+
+impl<'a> BytesEncode<'a> for PostingsCodec {
+    type EItem = Postings<'a>;
+
+    fn bytes_encode(item: &'a Self::EItem) -> Option<Cow<'a, [u8]>> {
+        let u64_size = mem::size_of::<u64>();
+        let docids_size = item.docids.len() * mem::size_of::<DocumentId>();
+        let matches_size = item.matches.len() * mem::size_of::<DocIndex>();
+
+        let mut buffer = Vec::with_capacity(u64_size + docids_size + matches_size);
+
+        let docids_len = item.docids.len();
+        buffer.extend_from_slice(&docids_len.to_be_bytes());
+        buffer.extend_from_slice(item.docids.as_bytes());
+        buffer.extend_from_slice(item.matches.as_bytes());
+
+        Some(Cow::Owned(buffer))
+    }
+}
+
+fn aligned_to(bytes: &[u8], align: usize) -> bool {
+    (bytes as *const _ as *const () as usize) % align == 0
+}
+
+fn from_bytes_to_set<'a, T: 'a>(bytes: &'a [u8]) -> Option<Cow<'a, Set<T>>>
+where T: Clone + FromBytes
+{
+    match zerocopy::LayoutVerified::<_, [T]>::new_slice(bytes) {
+        Some(layout) => Some(Cow::Borrowed(Set::new_unchecked(layout.into_slice()))),
+        None => {
+            let len = bytes.len();
+            let elem_size = mem::size_of::<T>();
+
+            // ensure that it is the alignment that is wrong
+            // and the length is valid
+            if len % elem_size == 0 && !aligned_to(bytes, mem::align_of::<T>()) {
+                let elems = len / elem_size;
+                let mut vec = Vec::<T>::with_capacity(elems);
+
+                unsafe {
+                    let dst = vec.as_mut_ptr() as *mut u8;
+                    ptr::copy_nonoverlapping(bytes.as_ptr(), dst, len);
+                    vec.set_len(elems);
+                }
+
+                return Some(Cow::Owned(SetBuf::new_unchecked(vec)));
+            }
+
+            None
+        }
+    }
+}
+
+impl<'a> BytesDecode<'a> for PostingsCodec {
+    type DItem = Postings<'a>;
+
+    fn bytes_decode(bytes: &'a [u8]) -> Option<Self::DItem> {
+        let u64_size = mem::size_of::<u64>();
+        let docid_size = mem::size_of::<DocumentId>();
+
+        let (len_bytes, bytes) = bytes.split_at(u64_size);
+        let docids_len = len_bytes.try_into().ok().map(u64::from_be_bytes)? as usize;
+        let docids_size = docids_len * docid_size;
+
+        let docids_bytes = &bytes[..docids_size];
+        let matches_bytes = &bytes[docids_size..];
+
+        let docids = from_bytes_to_set(docids_bytes)?;
+        let matches = from_bytes_to_set(matches_bytes)?;
+
+        Some(Postings { docids, matches })
+    }
+}
+
+fn main_name(name: &str) -> String {
+    format!("store-{}", name)
+}
+
+fn postings_lists_name(name: &str) -> String {
+    format!("store-{}-postings-lists", name)
+}
+
+fn documents_fields_name(name: &str) -> String {
+    format!("store-{}-documents-fields", name)
+}
+
+fn documents_fields_counts_name(name: &str) -> String {
+    format!("store-{}-documents-fields-counts", name)
+}
+
+fn synonyms_name(name: &str) -> String {
+    format!("store-{}-synonyms", name)
+}
+
+fn docs_words_name(name: &str) -> String {
+    format!("store-{}-docs-words", name)
+}
+
+fn prefix_documents_cache_name(name: &str) -> String {
+    format!("store-{}-prefix-documents-cache", name)
+}
+
+fn prefix_postings_lists_cache_name(name: &str) -> String {
+    format!("store-{}-prefix-postings-lists-cache", name)
+}
+
+fn updates_name(name: &str) -> String {
+    format!("store-{}-updates", name)
+}
+
+fn updates_results_name(name: &str) -> String {
+    format!("store-{}-updates-results", name)
+}
+
+#[derive(Clone)]
+pub struct Index {
+    pub main: Main,
+    pub postings_lists: PostingsLists,
+    pub documents_fields: DocumentsFields,
+    pub documents_fields_counts: DocumentsFieldsCounts,
+    pub synonyms: Synonyms,
+    pub docs_words: DocsWords,
+    pub prefix_documents_cache: PrefixDocumentsCache,
+    pub prefix_postings_lists_cache: PrefixPostingsListsCache,
+
+    pub updates: Updates,
+    pub updates_results: UpdatesResults,
+    pub(crate) updates_notifier: UpdateEventsEmitter,
+}
+
+impl Index {
+    pub fn document<T: de::DeserializeOwned>(
+        &self,
+        reader: &heed::RoTxn<MainT>,
+        attributes: Option<&HashSet<&str>>,
+        document_id: DocumentId,
+    ) -> MResult<Option<T>> {
+        let schema = self.main.schema(reader)?;
+        let schema = schema.ok_or(Error::SchemaMissing)?;
+
+        let attributes = match attributes {
+            Some(attributes) => Some(attributes.iter().filter_map(|name| schema.id(*name)).collect()),
+            None => None,
+        };
+
+        let mut deserializer = Deserializer {
+            document_id,
+            reader,
+            documents_fields: self.documents_fields,
+            schema: &schema,
+            fields: attributes.as_ref(),
+        };
+
+        Ok(Option::<T>::deserialize(&mut deserializer)?)
+    }
+
+    pub fn document_attribute<T: de::DeserializeOwned>(
+        &self,
+        reader: &heed::RoTxn<MainT>,
+        document_id: DocumentId,
+        attribute: FieldId,
+    ) -> MResult<Option<T>> {
+        let bytes = self
+            .documents_fields
+            .document_attribute(reader, document_id, attribute)?;
+        match bytes {
+            Some(bytes) => Ok(Some(serde_json::from_slice(bytes)?)),
+            None => Ok(None),
+        }
+    }
+
+    pub fn document_attribute_bytes<'txn>(
+        &self,
+        reader: &'txn heed::RoTxn<MainT>,
+        document_id: DocumentId,
+        attribute: FieldId,
+    ) -> MResult<Option<&'txn [u8]>> {
+        let bytes = self
+            .documents_fields
+            .document_attribute(reader, document_id, attribute)?;
+        match bytes {
+            Some(bytes) => Ok(Some(bytes)),
+            None => Ok(None),
+        }
+    }
+
+    pub fn customs_update(&self, writer: &mut heed::RwTxn<UpdateT>, customs: Vec<u8>) -> ZResult<u64> {
+        let _ = self.updates_notifier.send(UpdateEvent::NewUpdate);
+        update::push_customs_update(writer, self.updates, self.updates_results, customs)
+    }
+
+    pub fn settings_update(&self, writer: &mut heed::RwTxn<UpdateT>, update: SettingsUpdate) -> ZResult<u64> {
+        let _ = self.updates_notifier.send(UpdateEvent::NewUpdate);
+        update::push_settings_update(writer, self.updates, self.updates_results, update)
+    }
+
+    pub fn documents_addition<D>(&self) -> update::DocumentsAddition<D> {
+        update::DocumentsAddition::new(
+            self.updates,
+            self.updates_results,
+            self.updates_notifier.clone(),
+        )
+    }
+
+    pub fn documents_partial_addition<D>(&self) -> update::DocumentsAddition<D> {
+        update::DocumentsAddition::new_partial(
+            self.updates,
+            self.updates_results,
+            self.updates_notifier.clone(),
+        )
+    }
+
+    pub fn documents_deletion(&self) -> update::DocumentsDeletion {
+        update::DocumentsDeletion::new(
+            self.updates,
+            self.updates_results,
+            self.updates_notifier.clone(),
+        )
+    }
+
+    pub fn clear_all(&self, writer: &mut heed::RwTxn<UpdateT>) -> MResult<u64> {
+        let _ = self.updates_notifier.send(UpdateEvent::NewUpdate);
+        update::push_clear_all(writer, self.updates, self.updates_results)
+    }
+
+    pub fn current_update_id(&self, reader: &heed::RoTxn<UpdateT>) -> MResult<Option<u64>> {
+        match self.updates.last_update(reader)? {
+            Some((id, _)) => Ok(Some(id)),
+            None => Ok(None),
+        }
+    }
+
+    pub fn update_status(
+        &self,
+        reader: &heed::RoTxn<UpdateT>,
+        update_id: u64,
+    ) -> MResult<Option<update::UpdateStatus>> {
+        update::update_status(reader, self.updates, self.updates_results, update_id)
+    }
+
+    pub fn all_updates_status(&self, reader: &heed::RoTxn<UpdateT>) -> MResult<Vec<update::UpdateStatus>> {
+        let mut updates = Vec::new();
+        let mut last_update_result_id = 0;
+
+        // retrieve all updates results
+        if let Some((last_id, _)) = self.updates_results.last_update(reader)? {
+            updates.reserve(last_id as usize);
+
+            for id in 0..=last_id {
+                if let Some(update) = self.update_status(reader, id)? {
+                    updates.push(update);
+                    last_update_result_id = id;
+                }
+            }
+        }
+
+        // retrieve all enqueued updates
+        if let Some((last_id, _)) = self.updates.last_update(reader)? {
+            for id in last_update_result_id + 1..=last_id {
+                if let Some(update) = self.update_status(reader, id)? {
+                    updates.push(update);
+                }
+            }
+        }
+
+        Ok(updates)
+    }
+
+    pub fn query_builder(&self) -> QueryBuilder {
+        QueryBuilder::new(
+            self.main,
+            self.postings_lists,
+            self.documents_fields_counts,
+            self.synonyms,
+            self.prefix_documents_cache,
+            self.prefix_postings_lists_cache,
+        )
+    }
+
+    pub fn query_builder_with_criteria<'c, 'f, 'd>(
+        &self,
+        criteria: Criteria<'c>,
+    ) -> QueryBuilder<'c, 'f, 'd> {
+        QueryBuilder::with_criteria(
+            self.main,
+            self.postings_lists,
+            self.documents_fields_counts,
+            self.synonyms,
+            self.prefix_documents_cache,
+            self.prefix_postings_lists_cache,
+            criteria,
+        )
+    }
+}
+
+pub fn create(
+    env: &heed::Env,
+    update_env: &heed::Env,
+    name: &str,
+    updates_notifier: UpdateEventsEmitter,
+) -> MResult<Index> {
+    // create all the store names
+    let main_name = main_name(name);
+    let postings_lists_name = postings_lists_name(name);
+    let documents_fields_name = documents_fields_name(name);
+    let documents_fields_counts_name = documents_fields_counts_name(name);
+    let synonyms_name = synonyms_name(name);
+    let docs_words_name = docs_words_name(name);
+    let prefix_documents_cache_name = prefix_documents_cache_name(name);
+    let prefix_postings_lists_cache_name = prefix_postings_lists_cache_name(name);
+    let updates_name = updates_name(name);
+    let updates_results_name = updates_results_name(name);
+
+    // open all the stores
+    let main = env.create_poly_database(Some(&main_name))?;
+    let postings_lists = env.create_database(Some(&postings_lists_name))?;
+    let documents_fields = env.create_database(Some(&documents_fields_name))?;
+    let documents_fields_counts = env.create_database(Some(&documents_fields_counts_name))?;
+    let synonyms = env.create_database(Some(&synonyms_name))?;
+    let docs_words = env.create_database(Some(&docs_words_name))?;
+    let prefix_documents_cache = env.create_database(Some(&prefix_documents_cache_name))?;
+    let prefix_postings_lists_cache = env.create_database(Some(&prefix_postings_lists_cache_name))?;
+    let updates = update_env.create_database(Some(&updates_name))?;
+    let updates_results = update_env.create_database(Some(&updates_results_name))?;
+
+    Ok(Index {
+        main: Main { main },
+        postings_lists: PostingsLists { postings_lists },
+        documents_fields: DocumentsFields { documents_fields },
+        documents_fields_counts: DocumentsFieldsCounts { documents_fields_counts },
+        synonyms: Synonyms { synonyms },
+        docs_words: DocsWords { docs_words },
+        prefix_postings_lists_cache: PrefixPostingsListsCache { prefix_postings_lists_cache },
+        prefix_documents_cache: PrefixDocumentsCache { prefix_documents_cache },
+        updates: Updates { updates },
+        updates_results: UpdatesResults { updates_results },
+        updates_notifier,
+    })
+}
+
+pub fn open(
+    env: &heed::Env,
+    update_env: &heed::Env,
+    name: &str,
+    updates_notifier: UpdateEventsEmitter,
+) -> MResult<Option<Index>> {
+    // create all the store names
+    let main_name = main_name(name);
+    let postings_lists_name = postings_lists_name(name);
+    let documents_fields_name = documents_fields_name(name);
+    let documents_fields_counts_name = documents_fields_counts_name(name);
+    let synonyms_name = synonyms_name(name);
+    let docs_words_name = docs_words_name(name);
+    let prefix_documents_cache_name = prefix_documents_cache_name(name);
+    let prefix_postings_lists_cache_name = prefix_postings_lists_cache_name(name);
+    let updates_name = updates_name(name);
+    let updates_results_name = updates_results_name(name);
+
+    // open all the stores
+    let main = match env.open_poly_database(Some(&main_name))? {
+        Some(main) => main,
+        None => return Ok(None),
+    };
+    let postings_lists = match env.open_database(Some(&postings_lists_name))? {
+        Some(postings_lists) => postings_lists,
+        None => return Ok(None),
+    };
+    let documents_fields = match env.open_database(Some(&documents_fields_name))? {
+        Some(documents_fields) => documents_fields,
+        None => return Ok(None),
+    };
+    let documents_fields_counts = match env.open_database(Some(&documents_fields_counts_name))? {
+        Some(documents_fields_counts) => documents_fields_counts,
+        None => return Ok(None),
+    };
+    let synonyms = match env.open_database(Some(&synonyms_name))? {
+        Some(synonyms) => synonyms,
+        None => return Ok(None),
+    };
+    let docs_words = match env.open_database(Some(&docs_words_name))? {
+        Some(docs_words) => docs_words,
+        None => return Ok(None),
+    };
+    let prefix_documents_cache = match env.open_database(Some(&prefix_documents_cache_name))? {
+        Some(prefix_documents_cache) => prefix_documents_cache,
+        None => return Ok(None),
+    };
+    let prefix_postings_lists_cache = match env.open_database(Some(&prefix_postings_lists_cache_name))? {
+        Some(prefix_postings_lists_cache) => prefix_postings_lists_cache,
+        None => return Ok(None),
+    };
+    let updates = match update_env.open_database(Some(&updates_name))? {
+        Some(updates) => updates,
+        None => return Ok(None),
+    };
+    let updates_results = match update_env.open_database(Some(&updates_results_name))? {
+        Some(updates_results) => updates_results,
+        None => return Ok(None),
+    };
+
+    Ok(Some(Index {
+        main: Main { main },
+        postings_lists: PostingsLists { postings_lists },
+        documents_fields: DocumentsFields { documents_fields },
+        documents_fields_counts: DocumentsFieldsCounts { documents_fields_counts },
+        synonyms: Synonyms { synonyms },
+        docs_words: DocsWords { docs_words },
+        prefix_documents_cache: PrefixDocumentsCache { prefix_documents_cache },
+        prefix_postings_lists_cache: PrefixPostingsListsCache { prefix_postings_lists_cache },
+        updates: Updates { updates },
+        updates_results: UpdatesResults { updates_results },
+        updates_notifier,
+    }))
+}
+
+pub fn clear(
+    writer: &mut heed::RwTxn<MainT>,
+    update_writer: &mut heed::RwTxn<UpdateT>,
+    index: &Index,
+) -> MResult<()> {
+    // clear all the stores
+    index.main.clear(writer)?;
+    index.postings_lists.clear(writer)?;
+    index.documents_fields.clear(writer)?;
+    index.documents_fields_counts.clear(writer)?;
+    index.synonyms.clear(writer)?;
+    index.docs_words.clear(writer)?;
+    index.prefix_documents_cache.clear(writer)?;
+    index.prefix_postings_lists_cache.clear(writer)?;
+    index.updates.clear(update_writer)?;
+    index.updates_results.clear(update_writer)?;
+    Ok(())
+}

meilisearch-core/src/store/postings_lists.rs

@@ -0,0 +1,47 @@
+use std::borrow::Cow;
+
+use heed::Result as ZResult;
+use heed::types::ByteSlice;
+use sdset::{Set, SetBuf};
+use slice_group_by::GroupBy;
+
+use crate::database::MainT;
+use crate::DocIndex;
+use crate::store::{Postings, PostingsCodec};
+
+#[derive(Copy, Clone)]
+pub struct PostingsLists {
+    pub(crate) postings_lists: heed::Database<ByteSlice, PostingsCodec>,
+}
+
+impl PostingsLists {
+    pub fn put_postings_list(
+        self,
+        writer: &mut heed::RwTxn<MainT>,
+        word: &[u8],
+        matches: &Set<DocIndex>,
+    ) -> ZResult<()> {
+        let docids = matches.linear_group_by_key(|m| m.document_id).map(|g| g[0].document_id).collect();
+        let docids = Cow::Owned(SetBuf::new_unchecked(docids));
+        let matches = Cow::Borrowed(matches);
+        let postings = Postings { docids, matches };
+
+        self.postings_lists.put(writer, word, &postings)
+    }
+
+    pub fn del_postings_list(self, writer: &mut heed::RwTxn<MainT>, word: &[u8]) -> ZResult<bool> {
+        self.postings_lists.delete(writer, word)
+    }
+
+    pub fn clear(self, writer: &mut heed::RwTxn<MainT>) -> ZResult<()> {
+        self.postings_lists.clear(writer)
+    }
+
+    pub fn postings_list<'txn>(
+        self,
+        reader: &'txn heed::RoTxn<MainT>,
+        word: &[u8],
+    ) -> ZResult<Option<Postings<'txn>>> {
+        self.postings_lists.get(reader, word)
+    }
+}

meilisearch-core/src/store/prefix_documents_cache.rs

@@ -0,0 +1,80 @@
+use std::borrow::Cow;
+
+use heed::types::{OwnedType, CowSlice};
+use heed::Result as ZResult;
+use zerocopy::{AsBytes, FromBytes};
+
+use super::BEU64;
+use crate::{DocumentId, Highlight};
+use crate::database::MainT;
+
+#[derive(Debug, Copy, Clone, AsBytes, FromBytes)]
+#[repr(C)]
+pub struct PrefixKey {
+    prefix: [u8; 4],
+    index: BEU64,
+    docid: BEU64,
+}
+
+impl PrefixKey {
+    pub fn new(prefix: [u8; 4], index: u64, docid: u64) -> PrefixKey {
+        PrefixKey {
+            prefix,
+            index: BEU64::new(index),
+            docid: BEU64::new(docid),
+        }
+    }
+}
+
+#[derive(Copy, Clone)]
+pub struct PrefixDocumentsCache {
+    pub(crate) prefix_documents_cache: heed::Database<OwnedType<PrefixKey>, CowSlice<Highlight>>,
+}
+
+impl PrefixDocumentsCache {
+    pub fn put_prefix_document(
+        self,
+        writer: &mut heed::RwTxn<MainT>,
+        prefix: [u8; 4],
+        index: usize,
+        docid: DocumentId,
+        highlights: &[Highlight],
+    ) -> ZResult<()> {
+        let key = PrefixKey::new(prefix, index as u64, docid.0);
+        self.prefix_documents_cache.put(writer, &key, highlights)
+    }
+
+    pub fn clear(self, writer: &mut heed::RwTxn<MainT>) -> ZResult<()> {
+        self.prefix_documents_cache.clear(writer)
+    }
+
+    pub fn prefix_documents<'txn>(
+        self,
+        reader: &'txn heed::RoTxn<MainT>,
+        prefix: [u8; 4],
+    ) -> ZResult<PrefixDocumentsIter<'txn>> {
+        let start = PrefixKey::new(prefix, 0, 0);
+        let end = PrefixKey::new(prefix, u64::max_value(), u64::max_value());
+        let iter = self.prefix_documents_cache.range(reader, &(start..=end))?;
+        Ok(PrefixDocumentsIter { iter })
+    }
+}
+
+pub struct PrefixDocumentsIter<'txn> {
+    iter: heed::RoRange<'txn, OwnedType<PrefixKey>, CowSlice<Highlight>>,
+}
+
+impl<'txn> Iterator for PrefixDocumentsIter<'txn> {
+    type Item = ZResult<(DocumentId, Cow<'txn, [Highlight]>)>;
+
+    fn next(&mut self) -> Option<Self::Item> {
+        match self.iter.next() {
+            Some(Ok((key, highlights))) => {
+                let docid = DocumentId(key.docid.get());
+                Some(Ok((docid, highlights)))
+            }
+            Some(Err(e)) => Some(Err(e)),
+            None => None,
+        }
+    }
+}

meilisearch-core/src/store/prefix_postings_lists_cache.rs

@@ -0,0 +1,45 @@
+use std::borrow::Cow;
+
+use heed::Result as ZResult;
+use heed::types::OwnedType;
+use sdset::{Set, SetBuf};
+use slice_group_by::GroupBy;
+
+use crate::database::MainT;
+use crate::DocIndex;
+use crate::store::{PostingsCodec, Postings};
+
+#[derive(Copy, Clone)]
+pub struct PrefixPostingsListsCache {
+    pub(crate) prefix_postings_lists_cache: heed::Database<OwnedType<[u8; 4]>, PostingsCodec>,
+}
+
+impl PrefixPostingsListsCache {
+    pub fn put_prefix_postings_list(
+        self,
+        writer: &mut heed::RwTxn<MainT>,
+        prefix: [u8; 4],
+        matches: &Set<DocIndex>,
+    ) -> ZResult<()>
+    {
+        let docids = matches.linear_group_by_key(|m| m.document_id).map(|g| g[0].document_id).collect();
+        let docids = Cow::Owned(SetBuf::new_unchecked(docids));
+        let matches = Cow::Borrowed(matches);
+        let postings = Postings { docids, matches };
+
+        self.prefix_postings_lists_cache.put(writer, &prefix, &postings)
+    }
+
+    pub fn clear(self, writer: &mut heed::RwTxn<MainT>) -> ZResult<()> {
+        self.prefix_postings_lists_cache.clear(writer)
+    }
+
+    pub fn prefix_postings_list<'txn>(
+        self,
+        reader: &'txn heed::RoTxn<MainT>,
+        prefix: [u8; 4],
+    ) -> ZResult<Option<Postings<'txn>>>
+    {
+        self.prefix_postings_lists_cache.get(reader, &prefix)
+    }
+}

meilisearch-core/src/store/synonyms.rs

@@ -1,4 +1,5 @@
 use heed::types::ByteSlice;
+use crate::database::MainT;
 use heed::Result as ZResult;
 use std::sync::Arc;
 
@@ -10,7 +11,7 @@ pub struct Synonyms {
 impl Synonyms {
     pub fn put_synonyms(
         self,
-        writer: &mut heed::RwTxn,
+        writer: &mut heed::RwTxn<MainT>,
         word: &[u8],
         synonyms: &fst::Set,
     ) -> ZResult<()> {
@@ -18,15 +19,19 @@ impl Synonyms {
         self.synonyms.put(writer, word, bytes)
     }
 
-    pub fn del_synonyms(self, writer: &mut heed::RwTxn, word: &[u8]) -> ZResult<bool> {
+    pub fn del_synonyms(self, writer: &mut heed::RwTxn<MainT>, word: &[u8]) -> ZResult<bool> {
         self.synonyms.delete(writer, word)
     }
 
-    pub fn synonyms(self, reader: &heed::RoTxn, word: &[u8]) -> ZResult<Option<fst::Set>> {
+    pub fn clear(self, writer: &mut heed::RwTxn<MainT>) -> ZResult<()> {
+        self.synonyms.clear(writer)
+    }
+
+    pub fn synonyms(self, reader: &heed::RoTxn<MainT>, word: &[u8]) -> ZResult<Option<fst::Set>> {
         match self.synonyms.get(reader, word)? {
             Some(bytes) => {
                 let len = bytes.len();
-                let bytes = Arc::from(bytes);
+                let bytes = Arc::new(bytes.to_owned());
                 let fst = fst::raw::Fst::from_shared_bytes(bytes, 0, len).unwrap();
                 Ok(Some(fst::Set::from(fst)))
             }

meilisearch-core/src/store/updates.rs

@@ -1,4 +1,5 @@
 use super::BEU64;
+use crate::database::UpdateT;
 use crate::update::Update;
 use heed::types::{OwnedType, SerdeJson};
 use heed::Result as ZResult;
@@ -10,7 +11,7 @@ pub struct Updates {
 
 impl Updates {
     // TODO do not trigger deserialize if possible
-    pub fn last_update_id(self, reader: &heed::RoTxn) -> ZResult<Option<(u64, Update)>> {
+    pub fn last_update(self, reader: &heed::RoTxn<UpdateT>) -> ZResult<Option<(u64, Update)>> {
         match self.updates.last(reader)? {
             Some((key, data)) => Ok(Some((key.get(), data))),
             None => Ok(None),
@@ -18,7 +19,7 @@ impl Updates {
     }
 
     // TODO do not trigger deserialize if possible
-    fn first_update_id(self, reader: &heed::RoTxn) -> ZResult<Option<(u64, Update)>> {
+    pub fn first_update(self, reader: &heed::RoTxn<UpdateT>) -> ZResult<Option<(u64, Update)>> {
         match self.updates.first(reader)? {
             Some((key, data)) => Ok(Some((key.get(), data))),
             None => Ok(None),
@@ -26,14 +27,14 @@ impl Updates {
     }
 
     // TODO do not trigger deserialize if possible
-    pub fn contains(self, reader: &heed::RoTxn, update_id: u64) -> ZResult<bool> {
+    pub fn get(self, reader: &heed::RoTxn<UpdateT>, update_id: u64) -> ZResult<Option<Update>> {
         let update_id = BEU64::new(update_id);
-        self.updates.get(reader, &update_id).map(|v| v.is_some())
+        self.updates.get(reader, &update_id)
     }
 
     pub fn put_update(
         self,
-        writer: &mut heed::RwTxn,
+        writer: &mut heed::RwTxn<UpdateT>,
         update_id: u64,
         update: &Update,
     ) -> ZResult<()> {
@@ -42,8 +43,13 @@ impl Updates {
         self.updates.put(writer, &update_id, update)
     }
 
-    pub fn pop_front(self, writer: &mut heed::RwTxn) -> ZResult<Option<(u64, Update)>> {
-        match self.first_update_id(writer)? {
+    pub fn del_update(self, writer: &mut heed::RwTxn<UpdateT>, update_id: u64) -> ZResult<bool> {
+        let update_id = BEU64::new(update_id);
+        self.updates.delete(writer, &update_id)
+    }
+
+    pub fn pop_front(self, writer: &mut heed::RwTxn<UpdateT>) -> ZResult<Option<(u64, Update)>> {
+        match self.first_update(writer)? {
             Some((update_id, update)) => {
                 let key = BEU64::new(update_id);
                 self.updates.delete(writer, &key)?;
@@ -52,4 +58,8 @@ impl Updates {
             None => Ok(None),
         }
     }
+
+    pub fn clear(self, writer: &mut heed::RwTxn<UpdateT>) -> ZResult<()> {
+        self.updates.clear(writer)
+    }
 }

meilisearch-core/src/store/updates_results.rs

@@ -1,15 +1,19 @@
 use super::BEU64;
-use crate::update::UpdateResult;
-use heed::types::{OwnedType, SerdeBincode};
+use crate::database::UpdateT;
+use crate::update::ProcessedUpdateResult;
+use heed::types::{OwnedType, SerdeJson};
 use heed::Result as ZResult;
 
 #[derive(Copy, Clone)]
 pub struct UpdatesResults {
-    pub(crate) updates_results: heed::Database<OwnedType<BEU64>, SerdeBincode<UpdateResult>>,
+    pub(crate) updates_results: heed::Database<OwnedType<BEU64>, SerdeJson<ProcessedUpdateResult>>,
 }
 
 impl UpdatesResults {
-    pub fn last_update_id(self, reader: &heed::RoTxn) -> ZResult<Option<(u64, UpdateResult)>> {
+    pub fn last_update(
+        self,
+        reader: &heed::RoTxn<UpdateT>,
+    ) -> ZResult<Option<(u64, ProcessedUpdateResult)>> {
         match self.updates_results.last(reader)? {
             Some((key, data)) => Ok(Some((key.get(), data))),
             None => Ok(None),
@@ -18,9 +22,9 @@ impl UpdatesResults {
 
     pub fn put_update_result(
         self,
-        writer: &mut heed::RwTxn,
+        writer: &mut heed::RwTxn<UpdateT>,
         update_id: u64,
-        update_result: &UpdateResult,
+        update_result: &ProcessedUpdateResult,
     ) -> ZResult<()> {
         let update_id = BEU64::new(update_id);
         self.updates_results.put(writer, &update_id, update_result)
@@ -28,10 +32,14 @@ impl UpdatesResults {
 
     pub fn update_result(
         self,
-        reader: &heed::RoTxn,
+        reader: &heed::RoTxn<UpdateT>,
         update_id: u64,
-    ) -> ZResult<Option<UpdateResult>> {
+    ) -> ZResult<Option<ProcessedUpdateResult>> {
         let update_id = BEU64::new(update_id);
         self.updates_results.get(reader, &update_id)
     }
+
+    pub fn clear(self, writer: &mut heed::RwTxn<UpdateT>) -> ZResult<()> {
+        self.updates_results.clear(writer)
+    }
 }

meilisearch-core/src/update/clear_all.rs

@@ -0,0 +1,32 @@
+use crate::database::{MainT, UpdateT};
+use crate::update::{next_update_id, Update};
+use crate::{store, MResult, RankedMap};
+
+pub fn apply_clear_all(
+    writer: &mut heed::RwTxn<MainT>,
+    index: &store::Index,
+) -> MResult<()> {
+    index.main.put_words_fst(writer, &fst::Set::default())?;
+    index.main.put_ranked_map(writer, &RankedMap::default())?;
+    index.main.put_number_of_documents(writer, |_| 0)?;
+    index.documents_fields.clear(writer)?;
+    index.documents_fields_counts.clear(writer)?;
+    index.postings_lists.clear(writer)?;
+    index.docs_words.clear(writer)?;
+    index.prefix_documents_cache.clear(writer)?;
+    index.prefix_postings_lists_cache.clear(writer)?;
+
+    Ok(())
+}
+
+pub fn push_clear_all(
+    writer: &mut heed::RwTxn<UpdateT>,
+    updates_store: store::Updates,
+    updates_results_store: store::UpdatesResults,
+) -> MResult<u64> {
+    let last_update_id = next_update_id(writer, updates_store, updates_results_store)?;
+    let update = Update::clear_all();
+    updates_store.put_update(writer, last_update_id, &update)?;
+
+    Ok(last_update_id)
+}

meilisearch-core/src/update/customs_update.rs

@@ -1,9 +1,11 @@
-use crate::store;
-use crate::update::{next_update_id, Update};
 use heed::Result as ZResult;
 
+use crate::database::{MainT, UpdateT};
+use crate::store;
+use crate::update::{next_update_id, Update};
+
 pub fn apply_customs_update(
-    writer: &mut heed::RwTxn,
+    writer: &mut heed::RwTxn<MainT>,
     main_store: store::Main,
     customs: &[u8],
 ) -> ZResult<()> {
@@ -11,14 +13,14 @@ pub fn apply_customs_update(
 }
 
 pub fn push_customs_update(
-    writer: &mut heed::RwTxn,
+    writer: &mut heed::RwTxn<UpdateT>,
     updates_store: store::Updates,
     updates_results_store: store::UpdatesResults,
     customs: Vec<u8>,
 ) -> ZResult<u64> {
     let last_update_id = next_update_id(writer, updates_store, updates_results_store)?;
 
-    let update = Update::Customs(customs);
+    let update = Update::customs(customs);
     updates_store.put_update(writer, last_update_id, &update)?;
 
     Ok(last_update_id)

meilisearch-core/src/update/documents_addition.rs

@@ -0,0 +1,382 @@
+use std::collections::HashMap;
+
+use fst::{set::OpBuilder, SetBuilder};
+use indexmap::IndexMap;
+use sdset::{duo::Union, SetOperation};
+use serde::{Deserialize, Serialize};
+
+use crate::database::{MainT, UpdateT};
+use crate::database::{UpdateEvent, UpdateEventsEmitter};
+use crate::raw_indexer::RawIndexer;
+use crate::serde::{extract_document_id, serialize_value_with_id, Deserializer, Serializer};
+use crate::store;
+use crate::update::{apply_documents_deletion, compute_short_prefixes, next_update_id, Update};
+use crate::{Error, MResult, RankedMap};
+
+pub struct DocumentsAddition<D> {
+    updates_store: store::Updates,
+    updates_results_store: store::UpdatesResults,
+    updates_notifier: UpdateEventsEmitter,
+    documents: Vec<D>,
+    is_partial: bool,
+}
+
+impl<D> DocumentsAddition<D> {
+    pub fn new(
+        updates_store: store::Updates,
+        updates_results_store: store::UpdatesResults,
+        updates_notifier: UpdateEventsEmitter,
+    ) -> DocumentsAddition<D> {
+        DocumentsAddition {
+            updates_store,
+            updates_results_store,
+            updates_notifier,
+            documents: Vec::new(),
+            is_partial: false,
+        }
+    }
+
+    pub fn new_partial(
+        updates_store: store::Updates,
+        updates_results_store: store::UpdatesResults,
+        updates_notifier: UpdateEventsEmitter,
+    ) -> DocumentsAddition<D> {
+        DocumentsAddition {
+            updates_store,
+            updates_results_store,
+            updates_notifier,
+            documents: Vec::new(),
+            is_partial: true,
+        }
+    }
+
+    pub fn update_document(&mut self, document: D) {
+        self.documents.push(document);
+    }
+
+    pub fn finalize(self, writer: &mut heed::RwTxn<UpdateT>) -> MResult<u64>
+    where
+        D: serde::Serialize,
+    {
+        let _ = self.updates_notifier.send(UpdateEvent::NewUpdate);
+        let update_id = push_documents_addition(
+            writer,
+            self.updates_store,
+            self.updates_results_store,
+            self.documents,
+            self.is_partial,
+        )?;
+        Ok(update_id)
+    }
+}
+
+impl<D> Extend<D> for DocumentsAddition<D> {
+    fn extend<T: IntoIterator<Item = D>>(&mut self, iter: T) {
+        self.documents.extend(iter)
+    }
+}
+
+pub fn push_documents_addition<D: serde::Serialize>(
+    writer: &mut heed::RwTxn<UpdateT>,
+    updates_store: store::Updates,
+    updates_results_store: store::UpdatesResults,
+    addition: Vec<D>,
+    is_partial: bool,
+) -> MResult<u64> {
+    let mut values = Vec::with_capacity(addition.len());
+    for add in addition {
+        let vec = serde_json::to_vec(&add)?;
+        let add = serde_json::from_slice(&vec)?;
+        values.push(add);
+    }
+
+    let last_update_id = next_update_id(writer, updates_store, updates_results_store)?;
+
+    let update = if is_partial {
+        Update::documents_partial(values)
+    } else {
+        Update::documents_addition(values)
+    };
+
+    updates_store.put_update(writer, last_update_id, &update)?;
+
+    Ok(last_update_id)
+}
+
+pub fn apply_documents_addition<'a, 'b>(
+    writer: &'a mut heed::RwTxn<'b, MainT>,
+    index: &store::Index,
+    addition: Vec<IndexMap<String, serde_json::Value>>,
+) -> MResult<()> {
+    let mut documents_additions = HashMap::new();
+
+    let mut schema = match index.main.schema(writer)? {
+        Some(schema) => schema,
+        None => return Err(Error::SchemaMissing),
+    };
+
+    let primary_key = schema.primary_key().ok_or(Error::MissingPrimaryKey)?;
+
+    // 1. store documents ids for future deletion
+    for document in addition {
+        let document_id = match extract_document_id(&primary_key, &document)? {
+            Some(id) => id,
+            None => return Err(Error::MissingDocumentId),
+        };
+
+        documents_additions.insert(document_id, document);
+    }
+
+    // 2. remove the documents posting lists
+    let number_of_inserted_documents = documents_additions.len();
+    let documents_ids = documents_additions.iter().map(|(id, _)| *id).collect();
+    apply_documents_deletion(writer, index, documents_ids)?;
+
+    let mut ranked_map = match index.main.ranked_map(writer)? {
+        Some(ranked_map) => ranked_map,
+        None => RankedMap::default(),
+    };
+
+    let stop_words = match index.main.stop_words_fst(writer)? {
+        Some(stop_words) => stop_words,
+        None => fst::Set::default(),
+    };
+
+    // 3. index the documents fields in the stores
+    let mut indexer = RawIndexer::new(stop_words);
+
+    for (document_id, document) in documents_additions {
+        let serializer = Serializer {
+            txn: writer,
+            schema: &mut schema,
+            document_store: index.documents_fields,
+            document_fields_counts: index.documents_fields_counts,
+            indexer: &mut indexer,
+            ranked_map: &mut ranked_map,
+            document_id,
+        };
+
+        document.serialize(serializer)?;
+    }
+
+    write_documents_addition_index(
+        writer,
+        index,
+        &ranked_map,
+        number_of_inserted_documents,
+        indexer,
+    )?;
+
+    index.main.put_schema(writer, &schema)?;
+
+    Ok(())
+}
+
+pub fn apply_documents_partial_addition<'a, 'b>(
+    writer: &'a mut heed::RwTxn<'b, MainT>,
+    index: &store::Index,
+    addition: Vec<IndexMap<String, serde_json::Value>>,
+) -> MResult<()> {
+    let mut documents_additions = HashMap::new();
+
+    let mut schema = match index.main.schema(writer)? {
+        Some(schema) => schema,
+        None => return Err(Error::SchemaMissing),
+    };
+
+    let primary_key = schema.primary_key().ok_or(Error::MissingPrimaryKey)?;
+
+    // 1. store documents ids for future deletion
+    for mut document in addition {
+        let document_id = match extract_document_id(&primary_key, &document)? {
+            Some(id) => id,
+            None => return Err(Error::MissingDocumentId),
+        };
+
+        let mut deserializer = Deserializer {
+            document_id,
+            reader: writer,
+            documents_fields: index.documents_fields,
+            schema: &schema,
+            fields: None,
+        };
+
+        // retrieve the old document and
+        // update the new one with missing keys found in the old one
+        let result = Option::<HashMap<String, serde_json::Value>>::deserialize(&mut deserializer)?;
+        if let Some(old_document) = result {
+            for (key, value) in old_document {
+                document.entry(key).or_insert(value);
+            }
+        }
+
+        documents_additions.insert(document_id, document);
+    }
+
+    // 2. remove the documents posting lists
+    let number_of_inserted_documents = documents_additions.len();
+    let documents_ids = documents_additions.iter().map(|(id, _)| *id).collect();
+    apply_documents_deletion(writer, index, documents_ids)?;
+
+    let mut ranked_map = match index.main.ranked_map(writer)? {
+        Some(ranked_map) => ranked_map,
+        None => RankedMap::default(),
+    };
+
+    let stop_words = match index.main.stop_words_fst(writer)? {
+        Some(stop_words) => stop_words,
+        None => fst::Set::default(),
+    };
+
+    // 3. index the documents fields in the stores
+    let mut indexer = RawIndexer::new(stop_words);
+
+    for (document_id, document) in documents_additions {
+        let serializer = Serializer {
+            txn: writer,
+            schema: &mut schema,
+            document_store: index.documents_fields,
+            document_fields_counts: index.documents_fields_counts,
+            indexer: &mut indexer,
+            ranked_map: &mut ranked_map,
+            document_id,
+        };
+
+        document.serialize(serializer)?;
+    }
+
+    write_documents_addition_index(
+        writer,
+        index,
+        &ranked_map,
+        number_of_inserted_documents,
+        indexer,
+    )?;
+
+    index.main.put_schema(writer, &schema)?;
+
+    Ok(())
+}
+
+pub fn reindex_all_documents(writer: &mut heed::RwTxn<MainT>, index: &store::Index) -> MResult<()> {
+    let schema = match index.main.schema(writer)? {
+        Some(schema) => schema,
+        None => return Err(Error::SchemaMissing),
+    };
+
+    let mut ranked_map = RankedMap::default();
+
+    // 1. retrieve all documents ids
+    let mut documents_ids_to_reindex = Vec::new();
+    for result in index.documents_fields_counts.documents_ids(writer)? {
+        let document_id = result?;
+        documents_ids_to_reindex.push(document_id);
+    }
+
+    // 2. remove the documents posting lists
+    index.main.put_words_fst(writer, &fst::Set::default())?;
+    index.main.put_ranked_map(writer, &ranked_map)?;
+    index.main.put_number_of_documents(writer, |_| 0)?;
+    index.postings_lists.clear(writer)?;
+    index.docs_words.clear(writer)?;
+
+    let stop_words = match index.main.stop_words_fst(writer)? {
+        Some(stop_words) => stop_words,
+        None => fst::Set::default(),
+    };
+
+    let number_of_inserted_documents = documents_ids_to_reindex.len();
+    let mut indexer = RawIndexer::new(stop_words);
+    let mut ram_store = HashMap::new();
+
+    for document_id in documents_ids_to_reindex {
+        for result in index.documents_fields.document_fields(writer, document_id)? {
+            let (field_id, bytes) = result?;
+            let value: serde_json::Value = serde_json::from_slice(bytes)?;
+            ram_store.insert((document_id, field_id), value);
+        }
+
+        for ((docid, field_id), value) in ram_store.drain() {
+            serialize_value_with_id(
+                writer,
+                field_id,
+                &schema,
+                docid,
+                index.documents_fields,
+                index.documents_fields_counts,
+                &mut indexer,
+                &mut ranked_map,
+                &value
+            )?;
+        }
+    }
+
+    // 4. write the new index in the main store
+    write_documents_addition_index(
+        writer,
+        index,
+        &ranked_map,
+        number_of_inserted_documents,
+        indexer,
+    )?;
+
+    index.main.put_schema(writer, &schema)?;
+
+    Ok(())
+}
+
+pub fn write_documents_addition_index(
+    writer: &mut heed::RwTxn<MainT>,
+    index: &store::Index,
+    ranked_map: &RankedMap,
+    number_of_inserted_documents: usize,
+    indexer: RawIndexer,
+) -> MResult<()> {
+    let indexed = indexer.build();
+    let mut delta_words_builder = SetBuilder::memory();
+
+    for (word, delta_set) in indexed.words_doc_indexes {
+        delta_words_builder.insert(&word).unwrap();
+
+        let set = match index.postings_lists.postings_list(writer, &word)? {
+            Some(postings) => Union::new(&postings.matches, &delta_set).into_set_buf(),
+            None => delta_set,
+        };
+
+        index.postings_lists.put_postings_list(writer, &word, &set)?;
+    }
+
+    for (id, words) in indexed.docs_words {
+        index.docs_words.put_doc_words(writer, id, &words)?;
+    }
+
+    let delta_words = delta_words_builder
+        .into_inner()
+        .and_then(fst::Set::from_bytes)
+        .unwrap();
+
+    let words = match index.main.words_fst(writer)? {
+        Some(words) => {
+            let op = OpBuilder::new()
+                .add(words.stream())
+                .add(delta_words.stream())
+                .r#union();
+
+            let mut words_builder = SetBuilder::memory();
+            words_builder.extend_stream(op).unwrap();
+            words_builder
+                .into_inner()
+                .and_then(fst::Set::from_bytes)
+                .unwrap()
+        }
+        None => delta_words,
+    };
+
+    index.main.put_words_fst(writer, &words)?;
+    index.main.put_ranked_map(writer, ranked_map)?;
+    index.main.put_number_of_documents(writer, |old| old + number_of_inserted_documents as u64)?;
+
+    compute_short_prefixes(writer, index)?;
+
+    Ok(())
+}