Remove the unused distance

This reverts commit 7a4b6c065482f988b01298642f4c18775503f92f.

.github/scripts/check-release.sh

@@ -1,41 +1,24 @@
-#!/usr/bin/env bash
-set -eu -o pipefail
+#!/bin/bash
 
-check_tag() {
-    local expected=$1
-    local actual=$2
-    local filename=$3
-
-    if [[ $actual != $expected ]]; then
-        echo >&2 "Error: the current tag does not match the version in $filename: found $actual, expected $expected"
-        return 1
-    fi
+# check_tag $current_tag $file_tag $file_name
+function check_tag {
+  if [[ "$1" != "$2" ]]; then
+      echo "Error: the current tag does not match the version in Cargo.toml: found $2 - expected $1"
+      ret=1
+  fi
 }
 
-read_version() {
-    grep '^version = ' | cut -d \" -f 2
-}
-
-if [[ -z "${GITHUB_REF:-}" ]]; then
-    echo >&2 "Error: GITHUB_REF is not set"
-    exit 1
-fi
-
-if [[ ! "$GITHUB_REF" =~ ^refs/tags/v[0-9]+\.[0-9]+\.[0-9]+(-[a-z0-9]+)?$ ]]; then
-    echo >&2 "Error: GITHUB_REF is not a valid tag: $GITHUB_REF"
-    exit 1
-fi
-
-current_tag=${GITHUB_REF#refs/tags/v}
 ret=0
+current_tag=${GITHUB_REF#'refs/tags/v'}
 
-toml_tag="$(cat Cargo.toml | read_version)"
-check_tag "$current_tag" "$toml_tag" Cargo.toml || ret=1
+file_tag="$(grep '^version = ' Cargo.toml | cut -d '=' -f 2 | tr -d '"' | tr -d ' ')"
+check_tag $current_tag $file_tag
 
-lock_tag=$(grep -A 1 '^name = "meilisearch-auth"' Cargo.lock | read_version)
-check_tag "$current_tag" "$lock_tag" Cargo.lock || ret=1
+lock_file='Cargo.lock'
+lock_tag=$(grep -A 1 'name = "meilisearch-auth"' $lock_file | grep version | cut -d '=' -f 2 | tr -d '"' | tr -d ' ')
+check_tag $current_tag $lock_tag $lock_file
 
-if (( ret == 0 )); then
-    echo 'OK'
+if [[ "$ret" -eq 0 ]] ; then
+  echo 'OK'
 fi
 exit $ret

.github/workflows/fuzzer-indexing.yml

@@ -1,24 +0,0 @@
-name: Run the indexing fuzzer
-
-on:
-  push:
-    branches:
-      - main
-
-jobs:
-  fuzz:
-    name: Setup the action
-    runs-on: ubuntu-latest
-    timeout-minutes: 4320 # 72h
-    steps:
-      - uses: actions/checkout@v3
-      - uses: actions-rs/toolchain@v1
-        with:
-          profile: minimal
-          toolchain: stable
-          override: true
-
-      # Run benchmarks
-      - name: Run the fuzzer
-        run: |
-          cargo run --release --bin fuzz-indexing

.github/workflows/sdks-tests.yml

@@ -25,7 +25,7 @@ jobs:
       - name: Define the Docker image we need to use
         id: define-image
         run: |
-          event=${{ github.event_name }}
+          event=${{ github.event.action }}
           echo "docker-image=nightly" >> $GITHUB_OUTPUT
           if [[ $event == 'workflow_dispatch' ]]; then
             echo "docker-image=${{ github.event.inputs.docker_image }}" >> $GITHUB_OUTPUT
@@ -37,7 +37,7 @@ jobs:
     runs-on: ubuntu-latest
     services:
       meilisearch:
-        image: getmeili/meilisearch:${{ needs.define-docker-image.outputs.docker-image }}
+        image: getmeili/meilisearch:${{ github.event.inputs.docker_image }}
         env:
           MEILI_MASTER_KEY: ${{ env.MEILI_MASTER_KEY }}
           MEILI_NO_ANALYTICS: ${{ env.MEILI_NO_ANALYTICS }}
@@ -72,7 +72,7 @@ jobs:
     runs-on: ubuntu-latest
     services:
       meilisearch:
-        image: getmeili/meilisearch:${{ needs.define-docker-image.outputs.docker-image }}
+        image: getmeili/meilisearch:${{ github.event.inputs.docker_image }}
         env:
           MEILI_MASTER_KEY: ${{ env.MEILI_MASTER_KEY }}
           MEILI_NO_ANALYTICS: ${{ env.MEILI_NO_ANALYTICS }}
@@ -99,7 +99,7 @@ jobs:
     runs-on: ubuntu-latest
     services:
       meilisearch:
-        image: getmeili/meilisearch:${{ needs.define-docker-image.outputs.docker-image }}
+        image: getmeili/meilisearch:${{ github.event.inputs.docker_image }}
         env:
           MEILI_MASTER_KEY: ${{ env.MEILI_MASTER_KEY }}
           MEILI_NO_ANALYTICS: ${{ env.MEILI_NO_ANALYTICS }}
@@ -130,7 +130,7 @@ jobs:
     runs-on: ubuntu-latest
     services:
       meilisearch:
-        image: getmeili/meilisearch:${{ needs.define-docker-image.outputs.docker-image }}
+        image: getmeili/meilisearch:${{ github.event.inputs.docker_image }}
         env:
           MEILI_MASTER_KEY: ${{ env.MEILI_MASTER_KEY }}
           MEILI_NO_ANALYTICS: ${{ env.MEILI_NO_ANALYTICS }}
@@ -155,7 +155,7 @@ jobs:
     runs-on: ubuntu-latest
     services:
       meilisearch:
-        image: getmeili/meilisearch:${{ needs.define-docker-image.outputs.docker-image }}
+        image: getmeili/meilisearch:${{ github.event.inputs.docker_image }}
         env:
           MEILI_MASTER_KEY: ${{ env.MEILI_MASTER_KEY }}
           MEILI_NO_ANALYTICS: ${{ env.MEILI_NO_ANALYTICS }}
@@ -185,7 +185,7 @@ jobs:
     runs-on: ubuntu-latest
     services:
       meilisearch:
-        image: getmeili/meilisearch:${{ needs.define-docker-image.outputs.docker-image }}
+        image: getmeili/meilisearch:${{ github.event.inputs.docker_image }}
         env:
           MEILI_MASTER_KEY: ${{ env.MEILI_MASTER_KEY }}
           MEILI_NO_ANALYTICS: ${{ env.MEILI_NO_ANALYTICS }}
@@ -210,7 +210,7 @@ jobs:
     runs-on: ubuntu-latest
     services:
       meilisearch:
-        image: getmeili/meilisearch:${{ needs.define-docker-image.outputs.docker-image }}
+        image: getmeili/meilisearch:${{ github.event.inputs.docker_image }}
         env:
           MEILI_MASTER_KEY: ${{ env.MEILI_MASTER_KEY }}
           MEILI_NO_ANALYTICS: ${{ env.MEILI_NO_ANALYTICS }}

Cargo.toml

@@ -10,12 +10,10 @@ members = [
     "file-store",
     "permissive-json-pointer",
     "milli",
-    "index-stats",
     "filter-parser",
     "flatten-serde-json",
     "json-depth-checker",
-    "benchmarks",
-    "fuzzers",
+    "benchmarks"
 ]
 
 [workspace.package]

fuzzers/Cargo.toml

@@ -1,20 +0,0 @@
-[package]
-name = "fuzzers"
-publish = false
-
-version.workspace = true
-authors.workspace = true
-description.workspace = true
-homepage.workspace = true
-readme.workspace = true
-edition.workspace = true
-license.workspace = true
-
-[dependencies]
-arbitrary = { version = "1.3.0", features = ["derive"] }
-clap = { version = "4.3.0", features = ["derive"] }
-fastrand = "1.9.0"
-milli = { path = "../milli" }
-serde = { version = "1.0.160", features = ["derive"] }
-serde_json = { version = "1.0.95", features = ["preserve_order"] }
-tempfile = "3.5.0"

fuzzers/README.md

@@ -1,3 +0,0 @@
-# Fuzzers
-
-The purpose of this crate is to contains all the handmade "fuzzer" we may need.

fuzzers/src/bin/fuzz-indexing.rs

@@ -1,152 +0,0 @@
-use std::num::NonZeroUsize;
-use std::path::PathBuf;
-use std::sync::atomic::{AtomicBool, AtomicUsize, Ordering};
-use std::time::Duration;
-
-use arbitrary::{Arbitrary, Unstructured};
-use clap::Parser;
-use fuzzers::Operation;
-use milli::heed::EnvOpenOptions;
-use milli::update::{IndexDocuments, IndexDocumentsConfig, IndexerConfig};
-use milli::Index;
-use tempfile::TempDir;
-
-#[derive(Debug, Arbitrary)]
-struct Batch([Operation; 5]);
-
-#[derive(Debug, Clone, Parser)]
-struct Opt {
-    /// The number of fuzzer to run in parallel.
-    #[clap(long)]
-    par: Option<NonZeroUsize>,
-    // We need to put a lot of newlines in the following documentation or else everything gets collapsed on one line
-    /// The path in which the databases will be created.
-    /// Using a ramdisk is recommended.
-    ///
-    /// Linux:
-    ///
-    /// sudo mount -t tmpfs -o size=2g tmpfs ramdisk # to create it
-    ///
-    /// sudo umount ramdisk # to remove it
-    ///
-    /// MacOS:
-    ///
-    /// diskutil erasevolume HFS+ 'RAM Disk' `hdiutil attach -nobrowse -nomount ram://4194304 # create it
-    ///
-    /// hdiutil detach /dev/:the_disk
-    #[clap(long)]
-    path: Option<PathBuf>,
-}
-
-fn main() {
-    let opt = Opt::parse();
-    let progression: &'static AtomicUsize = Box::leak(Box::new(AtomicUsize::new(0)));
-    let stop: &'static AtomicBool = Box::leak(Box::new(AtomicBool::new(false)));
-
-    let par = opt.par.unwrap_or_else(|| std::thread::available_parallelism().unwrap()).get();
-    let mut handles = Vec::with_capacity(par);
-
-    for _ in 0..par {
-        let opt = opt.clone();
-
-        let handle = std::thread::spawn(move || {
-            let mut options = EnvOpenOptions::new();
-            options.map_size(1024 * 1024 * 1024 * 1024);
-            let tempdir = match opt.path {
-                Some(path) => TempDir::new_in(path).unwrap(),
-                None => TempDir::new().unwrap(),
-            };
-            let index = Index::new(options, tempdir.path()).unwrap();
-            let indexer_config = IndexerConfig::default();
-            let index_documents_config = IndexDocumentsConfig::default();
-
-            std::thread::scope(|s| {
-                loop {
-                    if stop.load(Ordering::Relaxed) {
-                        return;
-                    }
-                    let v: Vec<u8> =
-                        std::iter::repeat_with(|| fastrand::u8(..)).take(1000).collect();
-
-                    let mut data = Unstructured::new(&v);
-                    let batches = <[Batch; 5]>::arbitrary(&mut data).unwrap();
-                    // will be used to display the error once a thread crashes
-                    let dbg_input = format!("{:#?}", batches);
-
-                    let handle = s.spawn(|| {
-                        let mut wtxn = index.write_txn().unwrap();
-
-                        for batch in batches {
-                            let mut builder = IndexDocuments::new(
-                                &mut wtxn,
-                                &index,
-                                &indexer_config,
-                                index_documents_config.clone(),
-                                |_| (),
-                                || false,
-                            )
-                            .unwrap();
-
-                            for op in batch.0 {
-                                match op {
-                                    Operation::AddDoc(doc) => {
-                                        let documents =
-                                            milli::documents::objects_from_json_value(doc.to_d());
-                                        let documents =
-                                            milli::documents::documents_batch_reader_from_objects(
-                                                documents,
-                                            );
-                                        let (b, _added) = builder.add_documents(documents).unwrap();
-                                        builder = b;
-                                    }
-                                    Operation::DeleteDoc(id) => {
-                                        let (b, _removed) =
-                                            builder.remove_documents(vec![id.to_s()]).unwrap();
-                                        builder = b;
-                                    }
-                                }
-                            }
-                            builder.execute().unwrap();
-
-                            // after executing a batch we check if the database is corrupted
-                            let res = index.search(&wtxn).execute().unwrap();
-                            index.documents(&wtxn, res.documents_ids).unwrap();
-                            progression.fetch_add(1, Ordering::Relaxed);
-                        }
-                        wtxn.abort().unwrap();
-                    });
-                    if let err @ Err(_) = handle.join() {
-                        stop.store(true, Ordering::Relaxed);
-                        err.expect(&dbg_input);
-                    }
-                }
-            });
-        });
-        handles.push(handle);
-    }
-
-    std::thread::spawn(|| {
-        let mut last_value = 0;
-        let start = std::time::Instant::now();
-        loop {
-            let total = progression.load(Ordering::Relaxed);
-            let elapsed = start.elapsed().as_secs();
-            if elapsed > 3600 {
-                // after 1 hour, stop the fuzzer, success
-                std::process::exit(0);
-            }
-            println!(
-                "Has been running for {:?} seconds. Tested {} new values for a total of {}.",
-                elapsed,
-                total - last_value,
-                total
-            );
-            last_value = total;
-            std::thread::sleep(Duration::from_secs(1));
-        }
-    });
-
-    for handle in handles {
-        handle.join().unwrap();
-    }
-}

fuzzers/src/lib.rs

@@ -1,46 +0,0 @@
-use arbitrary::Arbitrary;
-use serde_json::{json, Value};
-
-#[derive(Debug, Arbitrary)]
-pub enum Document {
-    One,
-    Two,
-    Three,
-    Four,
-    Five,
-    Six,
-}
-
-impl Document {
-    pub fn to_d(&self) -> Value {
-        match self {
-            Document::One => json!({ "id": 0, "doggo": "bernese" }),
-            Document::Two => json!({ "id": 0, "doggo": "golden" }),
-            Document::Three => json!({ "id": 0, "catto": "jorts" }),
-            Document::Four => json!({ "id": 1, "doggo": "bernese" }),
-            Document::Five => json!({ "id": 1, "doggo": "golden" }),
-            Document::Six => json!({ "id": 1, "catto": "jorts" }),
-        }
-    }
-}
-
-#[derive(Debug, Arbitrary)]
-pub enum DocId {
-    Zero,
-    One,
-}
-
-impl DocId {
-    pub fn to_s(&self) -> String {
-        match self {
-            DocId::Zero => "0".to_string(),
-            DocId::One => "1".to_string(),
-        }
-    }
-}
-
-#[derive(Debug, Arbitrary)]
-pub enum Operation {
-    AddDoc(Document),
-    DeleteDoc(DocId),
-}

index-scheduler/src/autobatcher.rs

@@ -160,7 +160,7 @@ impl BatchKind {
 impl BatchKind {
     /// Returns a `ControlFlow::Break` if you must stop right now.
     /// The boolean tell you if an index has been created by the batched task.
-    /// To ease the writing of the code. `true` can be returned when you don't need to create an index
+    /// To ease the writting of the code. `true` can be returned when you don't need to create an index
     /// but false can't be returned if you needs to create an index.
     // TODO use an AutoBatchKind as input
     pub fn new(
@@ -214,7 +214,7 @@ impl BatchKind {
 
     /// Returns a `ControlFlow::Break` if you must stop right now.
     /// The boolean tell you if an index has been created by the batched task.
-    /// To ease the writing of the code. `true` can be returned when you don't need to create an index
+    /// To ease the writting of the code. `true` can be returned when you don't need to create an index
     /// but false can't be returned if you needs to create an index.
     #[rustfmt::skip]
     fn accumulate(self, id: TaskId, kind: AutobatchKind, index_already_exists: bool, primary_key: Option<&str>) -> ControlFlow<BatchKind, BatchKind> {
@@ -321,18 +321,9 @@ impl BatchKind {
                 })
             }
             (
-                BatchKind::DocumentOperation { method, allow_index_creation, primary_key, mut operation_ids },
+                this @ BatchKind::DocumentOperation { .. },
                 K::DocumentDeletion,
-            ) => {
-                operation_ids.push(id);
-
-                Continue(BatchKind::DocumentOperation {
-                    method,
-                    allow_index_creation,
-                    primary_key,
-                    operation_ids,
-                })
-            }
+            ) => Break(this),
             // but we can't autobatch documents if it's not the same kind
             // this match branch MUST be AFTER the previous one
             (
@@ -355,35 +346,7 @@ impl BatchKind {
                 deletion_ids.push(id);
                 Continue(BatchKind::DocumentClear { ids: deletion_ids })
             }
-            // we can autobatch the deletion and import if the index already exists
-            (
-                BatchKind::DocumentDeletion { mut deletion_ids },
-                K::DocumentImport { method, allow_index_creation, primary_key }
-            ) if index_already_exists => {
-                deletion_ids.push(id);
-
-                Continue(BatchKind::DocumentOperation {
-                    method,
-                    allow_index_creation,
-                    primary_key,
-                    operation_ids: deletion_ids,
-                })
-            }
-            // we can autobatch the deletion and import if both can't create an index
-            (
-                BatchKind::DocumentDeletion { mut deletion_ids },
-                K::DocumentImport { method, allow_index_creation, primary_key }
-            ) if !allow_index_creation => {
-                deletion_ids.push(id);
-
-                Continue(BatchKind::DocumentOperation {
-                    method,
-                    allow_index_creation,
-                    primary_key,
-                    operation_ids: deletion_ids,
-                })
-            }
-            // we can't autobatch a deletion and an import if the index does not exists but would be created by an addition
+            // we can't autobatch a deletion and an import
             (
                 this @ BatchKind::DocumentDeletion { .. },
                 K::DocumentImport { .. }
@@ -685,36 +648,36 @@ mod tests {
         debug_snapshot!(autobatch_from(false,None,  [settings(false)]), @"Some((Settings { allow_index_creation: false, settings_ids: [0] }, false))");
         debug_snapshot!(autobatch_from(false,None,  [settings(false), settings(false), settings(false)]), @"Some((Settings { allow_index_creation: false, settings_ids: [0, 1, 2] }, false))");
 
-        // We can autobatch document addition with document deletion
-        debug_snapshot!(autobatch_from(true, None, [doc_imp(ReplaceDocuments, true, None), doc_del()]), @"Some((DocumentOperation { method: ReplaceDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0, 1] }, true))");
-        debug_snapshot!(autobatch_from(true, None, [doc_imp(UpdateDocuments, true, None), doc_del()]), @"Some((DocumentOperation { method: UpdateDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0, 1] }, true))");
-        debug_snapshot!(autobatch_from(true, None, [doc_imp(ReplaceDocuments, false, None), doc_del()]), @"Some((DocumentOperation { method: ReplaceDocuments, allow_index_creation: false, primary_key: None, operation_ids: [0, 1] }, false))");
-        debug_snapshot!(autobatch_from(true, None, [doc_imp(UpdateDocuments, false, None), doc_del()]), @"Some((DocumentOperation { method: UpdateDocuments, allow_index_creation: false, primary_key: None, operation_ids: [0, 1] }, false))");
-        debug_snapshot!(autobatch_from(true, None, [doc_imp(ReplaceDocuments, true, Some("catto")), doc_del()]), @r###"Some((DocumentOperation { method: ReplaceDocuments, allow_index_creation: true, primary_key: Some("catto"), operation_ids: [0, 1] }, true))"###);
-        debug_snapshot!(autobatch_from(true, None, [doc_imp(UpdateDocuments, true, Some("catto")), doc_del()]), @r###"Some((DocumentOperation { method: UpdateDocuments, allow_index_creation: true, primary_key: Some("catto"), operation_ids: [0, 1] }, true))"###);
-        debug_snapshot!(autobatch_from(true, None, [doc_imp(ReplaceDocuments, false, Some("catto")), doc_del()]), @r###"Some((DocumentOperation { method: ReplaceDocuments, allow_index_creation: false, primary_key: Some("catto"), operation_ids: [0, 1] }, false))"###);
-        debug_snapshot!(autobatch_from(true, None, [doc_imp(UpdateDocuments, false, Some("catto")), doc_del()]), @r###"Some((DocumentOperation { method: UpdateDocuments, allow_index_creation: false, primary_key: Some("catto"), operation_ids: [0, 1] }, false))"###);
-        debug_snapshot!(autobatch_from(false, None, [doc_imp(ReplaceDocuments, true, None), doc_del()]), @"Some((DocumentOperation { method: ReplaceDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0, 1] }, true))");
-        debug_snapshot!(autobatch_from(false, None, [doc_imp(UpdateDocuments, true, None), doc_del()]), @"Some((DocumentOperation { method: UpdateDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0, 1] }, true))");
-        debug_snapshot!(autobatch_from(false, None, [doc_imp(ReplaceDocuments, false, None), doc_del()]), @"Some((DocumentOperation { method: ReplaceDocuments, allow_index_creation: false, primary_key: None, operation_ids: [0, 1] }, false))");
-        debug_snapshot!(autobatch_from(false, None, [doc_imp(UpdateDocuments, false, None), doc_del()]), @"Some((DocumentOperation { method: UpdateDocuments, allow_index_creation: false, primary_key: None, operation_ids: [0, 1] }, false))");
-        debug_snapshot!(autobatch_from(false, None, [doc_imp(ReplaceDocuments, true, Some("catto")), doc_del()]), @r###"Some((DocumentOperation { method: ReplaceDocuments, allow_index_creation: true, primary_key: Some("catto"), operation_ids: [0, 1] }, true))"###);
-        debug_snapshot!(autobatch_from(false, None, [doc_imp(UpdateDocuments, true, Some("catto")), doc_del()]), @r###"Some((DocumentOperation { method: UpdateDocuments, allow_index_creation: true, primary_key: Some("catto"), operation_ids: [0, 1] }, true))"###);
-        debug_snapshot!(autobatch_from(false, None, [doc_imp(ReplaceDocuments, false, Some("catto")), doc_del()]), @r###"Some((DocumentOperation { method: ReplaceDocuments, allow_index_creation: false, primary_key: Some("catto"), operation_ids: [0, 1] }, false))"###);
-        debug_snapshot!(autobatch_from(false, None, [doc_imp(UpdateDocuments, false, Some("catto")), doc_del()]), @r###"Some((DocumentOperation { method: UpdateDocuments, allow_index_creation: false, primary_key: Some("catto"), operation_ids: [0, 1] }, false))"###);
-        // And the other way around
-        debug_snapshot!(autobatch_from(true, None, [doc_del(), doc_imp(ReplaceDocuments, true, None)]), @"Some((DocumentOperation { method: ReplaceDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0, 1] }, false))");
-        debug_snapshot!(autobatch_from(true, None, [doc_del(), doc_imp(UpdateDocuments, true, None)]), @"Some((DocumentOperation { method: UpdateDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0, 1] }, false))");
-        debug_snapshot!(autobatch_from(true, None, [doc_del(), doc_imp(ReplaceDocuments, false, None)]), @"Some((DocumentOperation { method: ReplaceDocuments, allow_index_creation: false, primary_key: None, operation_ids: [0, 1] }, false))");
-        debug_snapshot!(autobatch_from(true, None, [doc_del(), doc_imp(UpdateDocuments, false, None)]), @"Some((DocumentOperation { method: UpdateDocuments, allow_index_creation: false, primary_key: None, operation_ids: [0, 1] }, false))");
-        debug_snapshot!(autobatch_from(true, None, [doc_del(), doc_imp(ReplaceDocuments, true, Some("catto"))]), @r###"Some((DocumentOperation { method: ReplaceDocuments, allow_index_creation: true, primary_key: Some("catto"), operation_ids: [0, 1] }, false))"###);
-        debug_snapshot!(autobatch_from(true, None, [doc_del(), doc_imp(UpdateDocuments, true, Some("catto"))]), @r###"Some((DocumentOperation { method: UpdateDocuments, allow_index_creation: true, primary_key: Some("catto"), operation_ids: [0, 1] }, false))"###);
-        debug_snapshot!(autobatch_from(true, None, [doc_del(), doc_imp(ReplaceDocuments, false, Some("catto"))]), @r###"Some((DocumentOperation { method: ReplaceDocuments, allow_index_creation: false, primary_key: Some("catto"), operation_ids: [0, 1] }, false))"###);
-        debug_snapshot!(autobatch_from(true, None, [doc_del(), doc_imp(UpdateDocuments, false, Some("catto"))]), @r###"Some((DocumentOperation { method: UpdateDocuments, allow_index_creation: false, primary_key: Some("catto"), operation_ids: [0, 1] }, false))"###);
-        debug_snapshot!(autobatch_from(false, None, [doc_del(), doc_imp(ReplaceDocuments, false, None)]), @"Some((DocumentOperation { method: ReplaceDocuments, allow_index_creation: false, primary_key: None, operation_ids: [0, 1] }, false))");
-        debug_snapshot!(autobatch_from(false, None, [doc_del(), doc_imp(UpdateDocuments, false, None)]), @"Some((DocumentOperation { method: UpdateDocuments, allow_index_creation: false, primary_key: None, operation_ids: [0, 1] }, false))");
-        debug_snapshot!(autobatch_from(false, None, [doc_del(), doc_imp(ReplaceDocuments, false, Some("catto"))]), @r###"Some((DocumentOperation { method: ReplaceDocuments, allow_index_creation: false, primary_key: Some("catto"), operation_ids: [0, 1] }, false))"###);
-        debug_snapshot!(autobatch_from(false, None, [doc_del(), doc_imp(UpdateDocuments, false, Some("catto"))]), @r###"Some((DocumentOperation { method: UpdateDocuments, allow_index_creation: false, primary_key: Some("catto"), operation_ids: [0, 1] }, false))"###);
+        // We can't autobatch document addition with document deletion
+        debug_snapshot!(autobatch_from(true, None, [doc_imp(ReplaceDocuments, true, None), doc_del()]), @"Some((DocumentOperation { method: ReplaceDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0] }, true))");
+        debug_snapshot!(autobatch_from(true, None, [doc_imp(UpdateDocuments, true, None), doc_del()]), @"Some((DocumentOperation { method: UpdateDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0] }, true))");
+        debug_snapshot!(autobatch_from(true, None, [doc_imp(ReplaceDocuments, false, None), doc_del()]), @"Some((DocumentOperation { method: ReplaceDocuments, allow_index_creation: false, primary_key: None, operation_ids: [0] }, false))");
+        debug_snapshot!(autobatch_from(true, None, [doc_imp(UpdateDocuments, false, None), doc_del()]), @"Some((DocumentOperation { method: UpdateDocuments, allow_index_creation: false, primary_key: None, operation_ids: [0] }, false))");
+        debug_snapshot!(autobatch_from(true, None, [doc_imp(ReplaceDocuments, true, Some("catto")), doc_del()]), @r###"Some((DocumentOperation { method: ReplaceDocuments, allow_index_creation: true, primary_key: Some("catto"), operation_ids: [0] }, true))"###);
+        debug_snapshot!(autobatch_from(true, None, [doc_imp(UpdateDocuments, true, Some("catto")), doc_del()]), @r###"Some((DocumentOperation { method: UpdateDocuments, allow_index_creation: true, primary_key: Some("catto"), operation_ids: [0] }, true))"###);
+        debug_snapshot!(autobatch_from(true, None, [doc_imp(ReplaceDocuments, false, Some("catto")), doc_del()]), @r###"Some((DocumentOperation { method: ReplaceDocuments, allow_index_creation: false, primary_key: Some("catto"), operation_ids: [0] }, false))"###);
+        debug_snapshot!(autobatch_from(true, None, [doc_imp(UpdateDocuments, false, Some("catto")), doc_del()]), @r###"Some((DocumentOperation { method: UpdateDocuments, allow_index_creation: false, primary_key: Some("catto"), operation_ids: [0] }, false))"###);
+        debug_snapshot!(autobatch_from(false, None, [doc_imp(ReplaceDocuments, true, None), doc_del()]), @"Some((DocumentOperation { method: ReplaceDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0] }, true))");
+        debug_snapshot!(autobatch_from(false, None, [doc_imp(UpdateDocuments, true, None), doc_del()]), @"Some((DocumentOperation { method: UpdateDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0] }, true))");
+        debug_snapshot!(autobatch_from(false, None, [doc_imp(ReplaceDocuments, false, None), doc_del()]), @"Some((DocumentOperation { method: ReplaceDocuments, allow_index_creation: false, primary_key: None, operation_ids: [0] }, false))");
+        debug_snapshot!(autobatch_from(false, None, [doc_imp(UpdateDocuments, false, None), doc_del()]), @"Some((DocumentOperation { method: UpdateDocuments, allow_index_creation: false, primary_key: None, operation_ids: [0] }, false))");
+        debug_snapshot!(autobatch_from(false, None, [doc_imp(ReplaceDocuments, true, Some("catto")), doc_del()]), @r###"Some((DocumentOperation { method: ReplaceDocuments, allow_index_creation: true, primary_key: Some("catto"), operation_ids: [0] }, true))"###);
+        debug_snapshot!(autobatch_from(false, None, [doc_imp(UpdateDocuments, true, Some("catto")), doc_del()]), @r###"Some((DocumentOperation { method: UpdateDocuments, allow_index_creation: true, primary_key: Some("catto"), operation_ids: [0] }, true))"###);
+        debug_snapshot!(autobatch_from(false, None, [doc_imp(ReplaceDocuments, false, Some("catto")), doc_del()]), @r###"Some((DocumentOperation { method: ReplaceDocuments, allow_index_creation: false, primary_key: Some("catto"), operation_ids: [0] }, false))"###);
+        debug_snapshot!(autobatch_from(false, None, [doc_imp(UpdateDocuments, false, Some("catto")), doc_del()]), @r###"Some((DocumentOperation { method: UpdateDocuments, allow_index_creation: false, primary_key: Some("catto"), operation_ids: [0] }, false))"###);
+        // we also can't do the only way around
+        debug_snapshot!(autobatch_from(true, None, [doc_del(), doc_imp(ReplaceDocuments, true, None)]), @"Some((DocumentDeletion { deletion_ids: [0] }, false))");
+        debug_snapshot!(autobatch_from(true, None, [doc_del(), doc_imp(UpdateDocuments, true, None)]), @"Some((DocumentDeletion { deletion_ids: [0] }, false))");
+        debug_snapshot!(autobatch_from(true, None, [doc_del(), doc_imp(ReplaceDocuments, false, None)]), @"Some((DocumentDeletion { deletion_ids: [0] }, false))");
+        debug_snapshot!(autobatch_from(true, None, [doc_del(), doc_imp(UpdateDocuments, false, None)]), @"Some((DocumentDeletion { deletion_ids: [0] }, false))");
+        debug_snapshot!(autobatch_from(true, None, [doc_del(), doc_imp(ReplaceDocuments, true, Some("catto"))]), @"Some((DocumentDeletion { deletion_ids: [0] }, false))");
+        debug_snapshot!(autobatch_from(true, None, [doc_del(), doc_imp(UpdateDocuments, true, Some("catto"))]), @"Some((DocumentDeletion { deletion_ids: [0] }, false))");
+        debug_snapshot!(autobatch_from(true, None, [doc_del(), doc_imp(ReplaceDocuments, false, Some("catto"))]), @"Some((DocumentDeletion { deletion_ids: [0] }, false))");
+        debug_snapshot!(autobatch_from(true, None, [doc_del(), doc_imp(UpdateDocuments, false, Some("catto"))]), @"Some((DocumentDeletion { deletion_ids: [0] }, false))");
+        debug_snapshot!(autobatch_from(false, None, [doc_del(), doc_imp(ReplaceDocuments, false, None)]), @"Some((DocumentDeletion { deletion_ids: [0] }, false))");
+        debug_snapshot!(autobatch_from(false, None, [doc_del(), doc_imp(UpdateDocuments, false, None)]), @"Some((DocumentDeletion { deletion_ids: [0] }, false))");
+        debug_snapshot!(autobatch_from(false, None, [doc_del(), doc_imp(ReplaceDocuments, false, Some("catto"))]), @"Some((DocumentDeletion { deletion_ids: [0] }, false))");
+        debug_snapshot!(autobatch_from(false, None, [doc_del(), doc_imp(UpdateDocuments, false, Some("catto"))]), @"Some((DocumentDeletion { deletion_ids: [0] }, false))");
     }
 
     #[test]

index-scheduler/src/batch.rs

@@ -998,7 +998,7 @@ impl IndexScheduler {
                 }()
                 .unwrap_or_default();
 
-                // The write transaction is directly owned and committed inside.
+                // The write transaction is directly owned and commited inside.
                 match self.index_mapper.delete_index(wtxn, &index_uid) {
                     Ok(()) => (),
                     Err(Error::IndexNotFound(_)) if index_has_been_created => (),

index-scheduler/src/lib.rs

@@ -1785,7 +1785,7 @@ mod tests {
             assert_eq!(task.kind.as_kind(), k);
         }
 
-        snapshot!(snapshot_index_scheduler(&index_scheduler), name: "everything_is_successfully_registered");
+        snapshot!(snapshot_index_scheduler(&index_scheduler), name: "everything_is_succesfully_registered");
     }
 
     #[test]
@@ -2075,105 +2075,6 @@ mod tests {
         snapshot!(snapshot_index_scheduler(&index_scheduler), name: "both_task_succeeded");
     }
 
-    #[test]
-    fn document_addition_and_document_deletion() {
-        let (index_scheduler, mut handle) = IndexScheduler::test(true, vec![]);
-
-        let content = r#"[
-            { "id": 1, "doggo": "jean bob" },
-            { "id": 2, "catto": "jorts" },
-            { "id": 3, "doggo": "bork" }
-        ]"#;
-
-        let (uuid, mut file) = index_scheduler.create_update_file_with_uuid(0).unwrap();
-        let documents_count = read_json(content.as_bytes(), file.as_file_mut()).unwrap();
-        file.persist().unwrap();
-        index_scheduler
-            .register(KindWithContent::DocumentAdditionOrUpdate {
-                index_uid: S("doggos"),
-                primary_key: Some(S("id")),
-                method: ReplaceDocuments,
-                content_file: uuid,
-                documents_count,
-                allow_index_creation: true,
-            })
-            .unwrap();
-        snapshot!(snapshot_index_scheduler(&index_scheduler), name: "registered_the_first_task");
-        index_scheduler
-            .register(KindWithContent::DocumentDeletion {
-                index_uid: S("doggos"),
-                documents_ids: vec![S("1"), S("2")],
-            })
-            .unwrap();
-        snapshot!(snapshot_index_scheduler(&index_scheduler), name: "registered_the_second_task");
-
-        handle.advance_one_successful_batch(); // The addition AND deletion should've been batched together
-        snapshot!(snapshot_index_scheduler(&index_scheduler), name: "after_processing_the_batch");
-
-        let index = index_scheduler.index("doggos").unwrap();
-        let rtxn = index.read_txn().unwrap();
-        let field_ids_map = index.fields_ids_map(&rtxn).unwrap();
-        let field_ids = field_ids_map.ids().collect::<Vec<_>>();
-        let documents = index
-            .all_documents(&rtxn)
-            .unwrap()
-            .map(|ret| obkv_to_json(&field_ids, &field_ids_map, ret.unwrap().1).unwrap())
-            .collect::<Vec<_>>();
-        snapshot!(serde_json::to_string_pretty(&documents).unwrap(), name: "documents");
-    }
-
-    #[test]
-    fn document_deletion_and_document_addition() {
-        let (index_scheduler, mut handle) = IndexScheduler::test(true, vec![]);
-        index_scheduler
-            .register(KindWithContent::DocumentDeletion {
-                index_uid: S("doggos"),
-                documents_ids: vec![S("1"), S("2")],
-            })
-            .unwrap();
-        snapshot!(snapshot_index_scheduler(&index_scheduler), name: "registered_the_first_task");
-
-        let content = r#"[
-            { "id": 1, "doggo": "jean bob" },
-            { "id": 2, "catto": "jorts" },
-            { "id": 3, "doggo": "bork" }
-        ]"#;
-
-        let (uuid, mut file) = index_scheduler.create_update_file_with_uuid(0).unwrap();
-        let documents_count = read_json(content.as_bytes(), file.as_file_mut()).unwrap();
-        file.persist().unwrap();
-        index_scheduler
-            .register(KindWithContent::DocumentAdditionOrUpdate {
-                index_uid: S("doggos"),
-                primary_key: Some(S("id")),
-                method: ReplaceDocuments,
-                content_file: uuid,
-                documents_count,
-                allow_index_creation: true,
-            })
-            .unwrap();
-        snapshot!(snapshot_index_scheduler(&index_scheduler), name: "registered_the_second_task");
-
-        // The deletion should have failed because it can't create an index
-        handle.advance_one_failed_batch();
-        snapshot!(snapshot_index_scheduler(&index_scheduler), name: "after_failing_the_deletion");
-
-        // The addition should works
-        handle.advance_one_successful_batch();
-        snapshot!(snapshot_index_scheduler(&index_scheduler), name: "after_last_successful_addition");
-
-        let index = index_scheduler.index("doggos").unwrap();
-        let rtxn = index.read_txn().unwrap();
-        let field_ids_map = index.fields_ids_map(&rtxn).unwrap();
-        let field_ids = field_ids_map.ids().collect::<Vec<_>>();
-        let documents = index
-            .all_documents(&rtxn)
-            .unwrap()
-            .map(|ret| obkv_to_json(&field_ids, &field_ids_map, ret.unwrap().1).unwrap())
-            .collect::<Vec<_>>();
-        snapshot!(serde_json::to_string_pretty(&documents).unwrap(), name: "documents");
-    }
-
     #[test]
     fn do_not_batch_task_of_different_indexes() {
         let (index_scheduler, mut handle) = IndexScheduler::test(true, vec![]);

index-scheduler/src/snapshots/lib.rs/document_addition_and_document_deletion/after_processing_the_batch.snap

@@ -1,43 +0,0 @@
----
-source: index-scheduler/src/lib.rs
----
-### Autobatching Enabled = true
-### Processing Tasks:
-[]
-----------------------------------------------------------------------
-### All Tasks:
-0 {uid: 0, status: succeeded, details: { received_documents: 3, indexed_documents: Some(3) }, kind: DocumentAdditionOrUpdate { index_uid: "doggos", primary_key: Some("id"), method: ReplaceDocuments, content_file: 00000000-0000-0000-0000-000000000000, documents_count: 3, allow_index_creation: true }}
-1 {uid: 1, status: succeeded, details: { received_document_ids: 2, deleted_documents: Some(2) }, kind: DocumentDeletion { index_uid: "doggos", documents_ids: ["1", "2"] }}
-----------------------------------------------------------------------
-### Status:
-enqueued []
-succeeded [0,1,]
-----------------------------------------------------------------------
-### Kind:
-"documentAdditionOrUpdate" [0,]
-"documentDeletion" [1,]
-----------------------------------------------------------------------
-### Index Tasks:
-doggos [0,1,]
-----------------------------------------------------------------------
-### Index Mapper:
-doggos: { number_of_documents: 1, field_distribution: {"doggo": 1, "id": 1} }
-
-----------------------------------------------------------------------
-### Canceled By:
-
-----------------------------------------------------------------------
-### Enqueued At:
-[timestamp] [0,]
-[timestamp] [1,]
-----------------------------------------------------------------------
-### Started At:
-[timestamp] [0,1,]
-----------------------------------------------------------------------
-### Finished At:
-[timestamp] [0,1,]
-----------------------------------------------------------------------
-### File Store:
-
-----------------------------------------------------------------------
-

index-scheduler/src/snapshots/lib.rs/document_addition_and_document_deletion/documents.snap

@@ -1,9 +0,0 @@
----
-source: index-scheduler/src/lib.rs
----
-[
-  {
-    "id": 3,
-    "doggo": "bork"
-  }
-]

index-scheduler/src/snapshots/lib.rs/document_addition_and_document_deletion/registered_the_first_task.snap

@@ -1,37 +0,0 @@
----
-source: index-scheduler/src/lib.rs
----
-### Autobatching Enabled = true
-### Processing Tasks:
-[]
-----------------------------------------------------------------------
-### All Tasks:
-0 {uid: 0, status: enqueued, details: { received_documents: 3, indexed_documents: None }, kind: DocumentAdditionOrUpdate { index_uid: "doggos", primary_key: Some("id"), method: ReplaceDocuments, content_file: 00000000-0000-0000-0000-000000000000, documents_count: 3, allow_index_creation: true }}
-----------------------------------------------------------------------
-### Status:
-enqueued [0,]
-----------------------------------------------------------------------
-### Kind:
-"documentAdditionOrUpdate" [0,]
-----------------------------------------------------------------------
-### Index Tasks:
-doggos [0,]
-----------------------------------------------------------------------
-### Index Mapper:
-
-----------------------------------------------------------------------
-### Canceled By:
-
-----------------------------------------------------------------------
-### Enqueued At:
-[timestamp] [0,]
-----------------------------------------------------------------------
-### Started At:
-----------------------------------------------------------------------
-### Finished At:
-----------------------------------------------------------------------
-### File Store:
-00000000-0000-0000-0000-000000000000
-
-----------------------------------------------------------------------
-

index-scheduler/src/snapshots/lib.rs/document_addition_and_document_deletion/registered_the_second_task.snap

@@ -1,40 +0,0 @@
----
-source: index-scheduler/src/lib.rs
----
-### Autobatching Enabled = true
-### Processing Tasks:
-[]
-----------------------------------------------------------------------
-### All Tasks:
-0 {uid: 0, status: enqueued, details: { received_documents: 3, indexed_documents: None }, kind: DocumentAdditionOrUpdate { index_uid: "doggos", primary_key: Some("id"), method: ReplaceDocuments, content_file: 00000000-0000-0000-0000-000000000000, documents_count: 3, allow_index_creation: true }}
-1 {uid: 1, status: enqueued, details: { received_document_ids: 2, deleted_documents: None }, kind: DocumentDeletion { index_uid: "doggos", documents_ids: ["1", "2"] }}
-----------------------------------------------------------------------
-### Status:
-enqueued [0,1,]
-----------------------------------------------------------------------
-### Kind:
-"documentAdditionOrUpdate" [0,]
-"documentDeletion" [1,]
-----------------------------------------------------------------------
-### Index Tasks:
-doggos [0,1,]
-----------------------------------------------------------------------
-### Index Mapper:
-
-----------------------------------------------------------------------
-### Canceled By:
-
-----------------------------------------------------------------------
-### Enqueued At:
-[timestamp] [0,]
-[timestamp] [1,]
-----------------------------------------------------------------------
-### Started At:
-----------------------------------------------------------------------
-### Finished At:
-----------------------------------------------------------------------
-### File Store:
-00000000-0000-0000-0000-000000000000
-
-----------------------------------------------------------------------
-

index-scheduler/src/snapshots/lib.rs/document_deletion_and_document_addition/after_failing_the_deletion.snap

@@ -1,43 +0,0 @@
----
-source: index-scheduler/src/lib.rs
----
-### Autobatching Enabled = true
-### Processing Tasks:
-[]
-----------------------------------------------------------------------
-### All Tasks:
-0 {uid: 0, status: failed, error: ResponseError { code: 200, message: "Index `doggos` not found.", error_code: "index_not_found", error_type: "invalid_request", error_link: "https://docs.meilisearch.com/errors#index_not_found" }, details: { received_document_ids: 2, deleted_documents: Some(0) }, kind: DocumentDeletion { index_uid: "doggos", documents_ids: ["1", "2"] }}
-1 {uid: 1, status: enqueued, details: { received_documents: 3, indexed_documents: None }, kind: DocumentAdditionOrUpdate { index_uid: "doggos", primary_key: Some("id"), method: ReplaceDocuments, content_file: 00000000-0000-0000-0000-000000000000, documents_count: 3, allow_index_creation: true }}
-----------------------------------------------------------------------
-### Status:
-enqueued [1,]
-failed [0,]
-----------------------------------------------------------------------
-### Kind:
-"documentAdditionOrUpdate" [1,]
-"documentDeletion" [0,]
-----------------------------------------------------------------------
-### Index Tasks:
-doggos [0,1,]
-----------------------------------------------------------------------
-### Index Mapper:
-
-----------------------------------------------------------------------
-### Canceled By:
-
-----------------------------------------------------------------------
-### Enqueued At:
-[timestamp] [0,]
-[timestamp] [1,]
-----------------------------------------------------------------------
-### Started At:
-[timestamp] [0,]
-----------------------------------------------------------------------
-### Finished At:
-[timestamp] [0,]
-----------------------------------------------------------------------
-### File Store:
-00000000-0000-0000-0000-000000000000
-
-----------------------------------------------------------------------
-

index-scheduler/src/snapshots/lib.rs/document_deletion_and_document_addition/after_last_successful_addition.snap

@@ -1,46 +0,0 @@
----
-source: index-scheduler/src/lib.rs
----
-### Autobatching Enabled = true
-### Processing Tasks:
-[]
-----------------------------------------------------------------------
-### All Tasks:
-0 {uid: 0, status: failed, error: ResponseError { code: 200, message: "Index `doggos` not found.", error_code: "index_not_found", error_type: "invalid_request", error_link: "https://docs.meilisearch.com/errors#index_not_found" }, details: { received_document_ids: 2, deleted_documents: Some(0) }, kind: DocumentDeletion { index_uid: "doggos", documents_ids: ["1", "2"] }}
-1 {uid: 1, status: succeeded, details: { received_documents: 3, indexed_documents: Some(3) }, kind: DocumentAdditionOrUpdate { index_uid: "doggos", primary_key: Some("id"), method: ReplaceDocuments, content_file: 00000000-0000-0000-0000-000000000000, documents_count: 3, allow_index_creation: true }}
-----------------------------------------------------------------------
-### Status:
-enqueued []
-succeeded [1,]
-failed [0,]
-----------------------------------------------------------------------
-### Kind:
-"documentAdditionOrUpdate" [1,]
-"documentDeletion" [0,]
-----------------------------------------------------------------------
-### Index Tasks:
-doggos [0,1,]
-----------------------------------------------------------------------
-### Index Mapper:
-doggos: { number_of_documents: 3, field_distribution: {"catto": 1, "doggo": 2, "id": 3} }
-
-----------------------------------------------------------------------
-### Canceled By:
-
-----------------------------------------------------------------------
-### Enqueued At:
-[timestamp] [0,]
-[timestamp] [1,]
-----------------------------------------------------------------------
-### Started At:
-[timestamp] [0,]
-[timestamp] [1,]
-----------------------------------------------------------------------
-### Finished At:
-[timestamp] [0,]
-[timestamp] [1,]
-----------------------------------------------------------------------
-### File Store:
-
-----------------------------------------------------------------------
-

index-scheduler/src/snapshots/lib.rs/document_deletion_and_document_addition/documents.snap

@@ -1,17 +0,0 @@
----
-source: index-scheduler/src/lib.rs
----
-[
-  {
-    "id": 1,
-    "doggo": "jean bob"
-  },
-  {
-    "id": 2,
-    "catto": "jorts"
-  },
-  {
-    "id": 3,
-    "doggo": "bork"
-  }
-]

index-scheduler/src/snapshots/lib.rs/document_deletion_and_document_addition/registered_the_first_task.snap

@@ -1,36 +0,0 @@
----
-source: index-scheduler/src/lib.rs
----
-### Autobatching Enabled = true
-### Processing Tasks:
-[]
-----------------------------------------------------------------------
-### All Tasks:
-0 {uid: 0, status: enqueued, details: { received_document_ids: 2, deleted_documents: None }, kind: DocumentDeletion { index_uid: "doggos", documents_ids: ["1", "2"] }}
-----------------------------------------------------------------------
-### Status:
-enqueued [0,]
-----------------------------------------------------------------------
-### Kind:
-"documentDeletion" [0,]
-----------------------------------------------------------------------
-### Index Tasks:
-doggos [0,]
-----------------------------------------------------------------------
-### Index Mapper:
-
-----------------------------------------------------------------------
-### Canceled By:
-
-----------------------------------------------------------------------
-### Enqueued At:
-[timestamp] [0,]
-----------------------------------------------------------------------
-### Started At:
-----------------------------------------------------------------------
-### Finished At:
-----------------------------------------------------------------------
-### File Store:
-
-----------------------------------------------------------------------
-

index-scheduler/src/snapshots/lib.rs/document_deletion_and_document_addition/registered_the_second_task.snap

@@ -1,40 +0,0 @@
----
-source: index-scheduler/src/lib.rs
----
-### Autobatching Enabled = true
-### Processing Tasks:
-[]
-----------------------------------------------------------------------
-### All Tasks:
-0 {uid: 0, status: enqueued, details: { received_document_ids: 2, deleted_documents: None }, kind: DocumentDeletion { index_uid: "doggos", documents_ids: ["1", "2"] }}
-1 {uid: 1, status: enqueued, details: { received_documents: 3, indexed_documents: None }, kind: DocumentAdditionOrUpdate { index_uid: "doggos", primary_key: Some("id"), method: ReplaceDocuments, content_file: 00000000-0000-0000-0000-000000000000, documents_count: 3, allow_index_creation: true }}
-----------------------------------------------------------------------
-### Status:
-enqueued [0,1,]
-----------------------------------------------------------------------
-### Kind:
-"documentAdditionOrUpdate" [1,]
-"documentDeletion" [0,]
-----------------------------------------------------------------------
-### Index Tasks:
-doggos [0,1,]
-----------------------------------------------------------------------
-### Index Mapper:
-
-----------------------------------------------------------------------
-### Canceled By:
-
-----------------------------------------------------------------------
-### Enqueued At:
-[timestamp] [0,]
-[timestamp] [1,]
-----------------------------------------------------------------------
-### Started At:
-----------------------------------------------------------------------
-### Finished At:
-----------------------------------------------------------------------
-### File Store:
-00000000-0000-0000-0000-000000000000
-
-----------------------------------------------------------------------
-

index-stats/Cargo.toml

@@ -1,12 +0,0 @@
-[package]
-name = "index-stats"
-description = "A small program that computes internal stats of a Meilisearch index"
-version = "0.1.0"
-edition = "2021"
-publish = false
-
-[dependencies]
-anyhow = "1.0.71"
-clap = { version = "4.3.5", features = ["derive"] }
-milli = { path = "../milli" }
-piechart = "1.0.0"

index-stats/src/main.rs

@@ -1,224 +0,0 @@
-use std::cmp::Reverse;
-use std::path::PathBuf;
-
-use clap::Parser;
-use milli::heed::{types::ByteSlice, EnvOpenOptions, PolyDatabase, RoTxn};
-use milli::index::db_name::*;
-use milli::index::Index;
-use piechart::{Chart, Color, Data};
-
-/// Simple program to greet a person
-#[derive(Parser, Debug)]
-#[command(author, version, about, long_about = None)]
-struct Args {
-    /// The path to the LMDB Meilisearch index database.
-    path: PathBuf,
-
-    /// The radius of the graphs
-    #[clap(long, default_value_t = 10)]
-    graph_radius: u16,
-
-    /// The radius of the graphs
-    #[clap(long, default_value_t = 6)]
-    graph_aspect_ratio: u16,
-}
-
-fn main() -> anyhow::Result<()> {
-    let Args { path, graph_radius, graph_aspect_ratio } = Args::parse();
-    let env = EnvOpenOptions::new().max_dbs(24).open(path)?;
-
-    // TODO not sure to keep that...
-    //      if removed put the pub(crate) back in the Index struct
-    matches!(
-        Option::<Index>::None,
-        Some(Index {
-            env: _,
-            main: _,
-            word_docids: _,
-            exact_word_docids: _,
-            word_prefix_docids: _,
-            exact_word_prefix_docids: _,
-            word_pair_proximity_docids: _,
-            word_prefix_pair_proximity_docids: _,
-            prefix_word_pair_proximity_docids: _,
-            word_position_docids: _,
-            word_fid_docids: _,
-            field_id_word_count_docids: _,
-            word_prefix_position_docids: _,
-            word_prefix_fid_docids: _,
-            script_language_docids: _,
-            facet_id_exists_docids: _,
-            facet_id_is_null_docids: _,
-            facet_id_is_empty_docids: _,
-            facet_id_f64_docids: _,
-            facet_id_string_docids: _,
-            field_id_docid_facet_f64s: _,
-            field_id_docid_facet_strings: _,
-            documents: _,
-        })
-    );
-
-    let mut wtxn = env.write_txn()?;
-    let main = env.create_poly_database(&mut wtxn, Some(MAIN))?;
-    let word_docids = env.create_poly_database(&mut wtxn, Some(WORD_DOCIDS))?;
-    let exact_word_docids = env.create_poly_database(&mut wtxn, Some(EXACT_WORD_DOCIDS))?;
-    let word_prefix_docids = env.create_poly_database(&mut wtxn, Some(WORD_PREFIX_DOCIDS))?;
-    let exact_word_prefix_docids =
-        env.create_poly_database(&mut wtxn, Some(EXACT_WORD_PREFIX_DOCIDS))?;
-    let word_pair_proximity_docids =
-        env.create_poly_database(&mut wtxn, Some(WORD_PAIR_PROXIMITY_DOCIDS))?;
-    let script_language_docids =
-        env.create_poly_database(&mut wtxn, Some(SCRIPT_LANGUAGE_DOCIDS))?;
-    let word_prefix_pair_proximity_docids =
-        env.create_poly_database(&mut wtxn, Some(WORD_PREFIX_PAIR_PROXIMITY_DOCIDS))?;
-    let prefix_word_pair_proximity_docids =
-        env.create_poly_database(&mut wtxn, Some(PREFIX_WORD_PAIR_PROXIMITY_DOCIDS))?;
-    let word_position_docids = env.create_poly_database(&mut wtxn, Some(WORD_POSITION_DOCIDS))?;
-    let word_fid_docids = env.create_poly_database(&mut wtxn, Some(WORD_FIELD_ID_DOCIDS))?;
-    let field_id_word_count_docids =
-        env.create_poly_database(&mut wtxn, Some(FIELD_ID_WORD_COUNT_DOCIDS))?;
-    let word_prefix_position_docids =
-        env.create_poly_database(&mut wtxn, Some(WORD_PREFIX_POSITION_DOCIDS))?;
-    let word_prefix_fid_docids =
-        env.create_poly_database(&mut wtxn, Some(WORD_PREFIX_FIELD_ID_DOCIDS))?;
-    let facet_id_f64_docids = env.create_poly_database(&mut wtxn, Some(FACET_ID_F64_DOCIDS))?;
-    let facet_id_string_docids =
-        env.create_poly_database(&mut wtxn, Some(FACET_ID_STRING_DOCIDS))?;
-    let facet_id_exists_docids =
-        env.create_poly_database(&mut wtxn, Some(FACET_ID_EXISTS_DOCIDS))?;
-    let facet_id_is_null_docids =
-        env.create_poly_database(&mut wtxn, Some(FACET_ID_IS_NULL_DOCIDS))?;
-    let facet_id_is_empty_docids =
-        env.create_poly_database(&mut wtxn, Some(FACET_ID_IS_EMPTY_DOCIDS))?;
-    let field_id_docid_facet_f64s =
-        env.create_poly_database(&mut wtxn, Some(FIELD_ID_DOCID_FACET_F64S))?;
-    let field_id_docid_facet_strings =
-        env.create_poly_database(&mut wtxn, Some(FIELD_ID_DOCID_FACET_STRINGS))?;
-    let documents = env.create_poly_database(&mut wtxn, Some(DOCUMENTS))?;
-    wtxn.commit()?;
-
-    let list = [
-        (main, MAIN),
-        (word_docids, WORD_DOCIDS),
-        (exact_word_docids, EXACT_WORD_DOCIDS),
-        (word_prefix_docids, WORD_PREFIX_DOCIDS),
-        (exact_word_prefix_docids, EXACT_WORD_PREFIX_DOCIDS),
-        (word_pair_proximity_docids, WORD_PAIR_PROXIMITY_DOCIDS),
-        (script_language_docids, SCRIPT_LANGUAGE_DOCIDS),
-        (word_prefix_pair_proximity_docids, WORD_PREFIX_PAIR_PROXIMITY_DOCIDS),
-        (prefix_word_pair_proximity_docids, PREFIX_WORD_PAIR_PROXIMITY_DOCIDS),
-        (word_position_docids, WORD_POSITION_DOCIDS),
-        (word_fid_docids, WORD_FIELD_ID_DOCIDS),
-        (field_id_word_count_docids, FIELD_ID_WORD_COUNT_DOCIDS),
-        (word_prefix_position_docids, WORD_PREFIX_POSITION_DOCIDS),
-        (word_prefix_fid_docids, WORD_PREFIX_FIELD_ID_DOCIDS),
-        (facet_id_f64_docids, FACET_ID_F64_DOCIDS),
-        (facet_id_string_docids, FACET_ID_STRING_DOCIDS),
-        (facet_id_exists_docids, FACET_ID_EXISTS_DOCIDS),
-        (facet_id_is_null_docids, FACET_ID_IS_NULL_DOCIDS),
-        (facet_id_is_empty_docids, FACET_ID_IS_EMPTY_DOCIDS),
-        (field_id_docid_facet_f64s, FIELD_ID_DOCID_FACET_F64S),
-        (field_id_docid_facet_strings, FIELD_ID_DOCID_FACET_STRINGS),
-        (documents, DOCUMENTS),
-    ];
-
-    let rtxn = env.read_txn()?;
-    let result: Result<Vec<_>, _> =
-        list.into_iter().map(|(db, name)| compute_stats(&rtxn, db).map(|s| (s, name))).collect();
-    let mut stats = result?;
-
-    println!("{:1$} Number of Entries", "", graph_radius as usize * 2);
-    stats.sort_by_key(|(s, _)| Reverse(s.number_of_entries));
-    let data = compute_graph_data(stats.iter().map(|(s, n)| (s.number_of_entries as f32, *n)));
-    Chart::new().radius(graph_radius).aspect_ratio(graph_aspect_ratio).draw(&data);
-    display_legend(&data);
-    print!("\r\n");
-
-    println!("{:1$} Size of Entries", "", graph_radius as usize * 2);
-    stats.sort_by_key(|(s, _)| Reverse(s.size_of_entries));
-    let data = compute_graph_data(stats.iter().map(|(s, n)| (s.size_of_entries as f32, *n)));
-    Chart::new().radius(graph_radius).aspect_ratio(graph_aspect_ratio).draw(&data);
-    display_legend(&data);
-    print!("\r\n");
-
-    println!("{:1$} Size of Data", "", graph_radius as usize * 2);
-    stats.sort_by_key(|(s, _)| Reverse(s.size_of_data));
-    let data = compute_graph_data(stats.iter().map(|(s, n)| (s.size_of_data as f32, *n)));
-    Chart::new().radius(graph_radius).aspect_ratio(graph_aspect_ratio).draw(&data);
-    display_legend(&data);
-    print!("\r\n");
-
-    println!("{:1$} Size of Keys", "", graph_radius as usize * 2);
-    stats.sort_by_key(|(s, _)| Reverse(s.size_of_keys));
-    let data = compute_graph_data(stats.iter().map(|(s, n)| (s.size_of_keys as f32, *n)));
-    Chart::new().radius(graph_radius).aspect_ratio(graph_aspect_ratio).draw(&data);
-    display_legend(&data);
-
-    Ok(())
-}
-
-fn display_legend(data: &[Data]) {
-    let total: f32 = data.iter().map(|d| d.value).sum();
-    for Data { label, value, color, fill } in data {
-        println!(
-            "{} {} {:.02}%",
-            color.unwrap().paint(fill.to_string()),
-            label,
-            value / total * 100.0
-        );
-    }
-}
-
-fn compute_graph_data<'a>(stats: impl IntoIterator<Item = (f32, &'a str)>) -> Vec<Data> {
-    let mut colors = [
-        Color::Red,
-        Color::Green,
-        Color::Yellow,
-        Color::Blue,
-        Color::Purple,
-        Color::Cyan,
-        Color::White,
-    ]
-    .into_iter()
-    .cycle();
-
-    let mut characters = ['▴', '▵', '▾', '▿', '▪', '▫', '•', '◦'].into_iter().cycle();
-
-    stats
-        .into_iter()
-        .map(|(value, name)| Data {
-            label: (*name).into(),
-            value,
-            color: Some(colors.next().unwrap().into()),
-            fill: characters.next().unwrap(),
-        })
-        .collect()
-}
-
-#[derive(Debug)]
-pub struct Stats {
-    pub number_of_entries: u64,
-    pub size_of_keys: u64,
-    pub size_of_data: u64,
-    pub size_of_entries: u64,
-}
-
-fn compute_stats(rtxn: &RoTxn, db: PolyDatabase) -> anyhow::Result<Stats> {
-    let mut number_of_entries = 0;
-    let mut size_of_keys = 0;
-    let mut size_of_data = 0;
-
-    for result in db.iter::<_, ByteSlice, ByteSlice>(rtxn)? {
-        let (key, data) = result?;
-        number_of_entries += 1;
-        size_of_keys += key.len() as u64;
-        size_of_data += data.len() as u64;
-    }
-
-    Ok(Stats {
-        number_of_entries,
-        size_of_keys,
-        size_of_data,
-        size_of_entries: size_of_keys + size_of_data,
-    })
-}

meilisearch-types/src/error.rs

@@ -217,6 +217,7 @@ InvalidDocumentFields                 , InvalidRequest       , BAD_REQUEST ;
 MissingDocumentFilter                 , InvalidRequest       , BAD_REQUEST ;
 InvalidDocumentFilter                 , InvalidRequest       , BAD_REQUEST ;
 InvalidDocumentGeoField               , InvalidRequest       , BAD_REQUEST ;
+InvalidVectorDimensions               , InvalidRequest       , BAD_REQUEST ;
 InvalidDocumentId                     , InvalidRequest       , BAD_REQUEST ;
 InvalidDocumentLimit                  , InvalidRequest       , BAD_REQUEST ;
 InvalidDocumentOffset                 , InvalidRequest       , BAD_REQUEST ;
@@ -332,6 +333,7 @@ impl ErrorCode for milli::Error {
                     UserError::InvalidSortableAttribute { .. } => Code::InvalidSearchSort,
                     UserError::CriterionError(_) => Code::InvalidSettingsRankingRules,
                     UserError::InvalidGeoField { .. } => Code::InvalidDocumentGeoField,
+                    UserError::InvalidVectorDimensions { .. } => Code::InvalidVectorDimensions,
                     UserError::SortError(_) => Code::InvalidSearchSort,
                     UserError::InvalidMinTypoWordLenSetting(_, _) => {
                         Code::InvalidSettingsTypoTolerance

meilisearch/src/routes/indexes/search.rs

@@ -34,6 +34,8 @@ pub fn configure(cfg: &mut web::ServiceConfig) {
 pub struct SearchQueryGet {
     #[deserr(default, error = DeserrQueryParamError<InvalidSearchQ>)]
     q: Option<String>,
+    #[deserr(default, error = DeserrQueryParamError<InvalidSearchQ>)]
+    vector: Option<Vec<f32>>,
     #[deserr(default = Param(DEFAULT_SEARCH_OFFSET()), error = DeserrQueryParamError<InvalidSearchOffset>)]
     offset: Param<usize>,
     #[deserr(default = Param(DEFAULT_SEARCH_LIMIT()), error = DeserrQueryParamError<InvalidSearchLimit>)]
@@ -80,6 +82,7 @@ impl From<SearchQueryGet> for SearchQuery {
 
         Self {
             q: other.q,
+            vector: other.vector,
             offset: other.offset.0,
             limit: other.limit.0,
             page: other.page.as_deref().copied(),

meilisearch/src/search.rs

@@ -31,11 +31,13 @@ pub const DEFAULT_CROP_MARKER: fn() -> String = || "…".to_string();
 pub const DEFAULT_HIGHLIGHT_PRE_TAG: fn() -> String = || "<em>".to_string();
 pub const DEFAULT_HIGHLIGHT_POST_TAG: fn() -> String = || "</em>".to_string();
 
-#[derive(Debug, Clone, Default, PartialEq, Eq, Deserr)]
+#[derive(Debug, Clone, Default, PartialEq, Deserr)]
 #[deserr(error = DeserrJsonError, rename_all = camelCase, deny_unknown_fields)]
 pub struct SearchQuery {
     #[deserr(default, error = DeserrJsonError<InvalidSearchQ>)]
     pub q: Option<String>,
+    #[deserr(default, error = DeserrJsonError<InvalidSearchQ>)]
+    pub vector: Option<Vec<f32>>,
     #[deserr(default = DEFAULT_SEARCH_OFFSET(), error = DeserrJsonError<InvalidSearchOffset>)]
     pub offset: usize,
     #[deserr(default = DEFAULT_SEARCH_LIMIT(), error = DeserrJsonError<InvalidSearchLimit>)]
@@ -80,13 +82,15 @@ impl SearchQuery {
 // This struct contains the fields of `SearchQuery` inline.
 // This is because neither deserr nor serde support `flatten` when using `deny_unknown_fields.
 // The `From<SearchQueryWithIndex>` implementation ensures both structs remain up to date.
-#[derive(Debug, Clone, PartialEq, Eq, Deserr)]
+#[derive(Debug, Clone, PartialEq, Deserr)]
 #[deserr(error = DeserrJsonError, rename_all = camelCase, deny_unknown_fields)]
 pub struct SearchQueryWithIndex {
     #[deserr(error = DeserrJsonError<InvalidIndexUid>, missing_field_error = DeserrJsonError::missing_index_uid)]
     pub index_uid: IndexUid,
     #[deserr(default, error = DeserrJsonError<InvalidSearchQ>)]
     pub q: Option<String>,
+    #[deserr(default, error = DeserrJsonError<InvalidSearchQ>)]
+    pub vector: Option<Vec<f32>>,
     #[deserr(default = DEFAULT_SEARCH_OFFSET(), error = DeserrJsonError<InvalidSearchOffset>)]
     pub offset: usize,
     #[deserr(default = DEFAULT_SEARCH_LIMIT(), error = DeserrJsonError<InvalidSearchLimit>)]
@@ -126,6 +130,7 @@ impl SearchQueryWithIndex {
         let SearchQueryWithIndex {
             index_uid,
             q,
+            vector,
             offset,
             limit,
             page,
@@ -147,6 +152,7 @@ impl SearchQueryWithIndex {
             index_uid,
             SearchQuery {
                 q,
+                vector,
                 offset,
                 limit,
                 page,
@@ -270,6 +276,10 @@ pub fn perform_search(
 
     let mut search = index.search(&rtxn);
 
+    if let Some(ref vector) = query.vector {
+        search.vector(vector.clone());
+    }
+
     if let Some(ref query) = query.q {
         search.query(query);
     }

milli/Cargo.toml

@@ -15,6 +15,7 @@ license.workspace = true
 bimap = { version = "0.6.3", features = ["serde"] }
 bincode = "1.3.3"
 bstr = "1.4.0"
+bytemuck = { version = "1.13.1", features = ["extern_crate_alloc"] }
 byteorder = "1.4.3"
 charabia = { version = "0.7.2", default-features = false }
 concat-arrays = "0.1.2"
@@ -32,18 +33,21 @@ heed = { git = "https://github.com/meilisearch/heed", tag = "v0.12.6", default-f
     "lmdb",
     "sync-read-txn",
 ] }
+hnsw = { version = "0.11.0", features = ["serde1"] }
 json-depth-checker = { path = "../json-depth-checker" }
 levenshtein_automata = { version = "0.2.1", features = ["fst_automaton"] }
 memmap2 = "0.5.10"
 obkv = "0.2.0"
 once_cell = "1.17.1"
 ordered-float = "3.6.0"
+rand_pcg = { version = "0.3.1", features = ["serde1"] }
 rayon = "1.7.0"
 roaring = "0.10.1"
 rstar = { version = "0.10.0", features = ["serde"] }
 serde = { version = "1.0.160", features = ["derive"] }
 serde_json = { version = "1.0.95", features = ["preserve_order"] }
 slice-group-by = "0.3.0"
+space = "0.17.0"
 smallstr = { version = "0.3.0", features = ["serde"] }
 smallvec = "1.10.0"
 smartstring = "1.0.1"
@@ -75,6 +79,9 @@ maplit = "1.0.2"
 md5 = "0.7.0"
 rand = { version = "0.8.5", features = ["small_rng"] }
 
+[target.'cfg(fuzzing)'.dev-dependencies]
+fuzzcheck = "0.12.1"
+
 [features]
 all-tokenizations = ["charabia/default"]
 

milli/examples/search.rs

@@ -52,6 +52,7 @@ fn main() -> Result<(), Box<dyn Error>> {
             let docs = execute_search(
                 &mut ctx,
                 &(!query.trim().is_empty()).then(|| query.trim().to_owned()),
+                &None,
                 TermsMatchingStrategy::Last,
                 false,
                 &None,

milli/src/distance.rs

@@ -0,0 +1,34 @@
+use serde::{Deserialize, Serialize};
+use space::Metric;
+
+#[derive(Debug, Default, Clone, Copy, Serialize, Deserialize)]
+pub struct DotProduct;
+
+impl Metric<Vec<f32>> for DotProduct {
+    type Unit = u32;
+
+    // TODO explain me this function, I don't understand why f32.to_bits is ordered.
+    // I tried to do this and it wasn't OK <https://stackoverflow.com/a/43305015/1941280>
+    //
+    // Following <https://docs.rs/space/0.17.0/space/trait.Metric.html>.
+    fn distance(&self, a: &Vec<f32>, b: &Vec<f32>) -> Self::Unit {
+        let dist: f32 = a.iter().zip(b).map(|(a, b)| a * b).sum();
+        let dist = 1.0 - dist;
+        debug_assert!(!dist.is_nan());
+        dist.to_bits()
+    }
+}
+
+#[derive(Debug, Default, Clone, Copy, Serialize, Deserialize)]
+pub struct Euclidean;
+
+impl Metric<Vec<f32>> for Euclidean {
+    type Unit = u32;
+
+    fn distance(&self, a: &Vec<f32>, b: &Vec<f32>) -> Self::Unit {
+        let squared: f32 = a.iter().zip(b).map(|(a, b)| (a - b).powi(2)).sum();
+        let dist = squared.sqrt();
+        debug_assert!(!dist.is_nan());
+        dist.to_bits()
+    }
+}

milli/src/documents/mod.rs

@@ -111,6 +111,7 @@ pub enum Error {
     Io(#[from] io::Error),
 }
 
+#[cfg(test)]
 pub fn objects_from_json_value(json: serde_json::Value) -> Vec<crate::Object> {
     let documents = match json {
         object @ serde_json::Value::Object(_) => vec![object],
@@ -140,6 +141,7 @@ macro_rules! documents {
     }};
 }
 
+#[cfg(test)]
 pub fn documents_batch_reader_from_objects(
     objects: impl IntoIterator<Item = Object>,
 ) -> DocumentsBatchReader<std::io::Cursor<Vec<u8>>> {

milli/src/error.rs

@@ -110,9 +110,11 @@ only composed of alphanumeric characters (a-z A-Z 0-9), hyphens (-) and undersco
     },
     #[error(transparent)]
     InvalidGeoField(#[from] GeoError),
+    #[error("Invalid vector dimensions: expected: `{}`, found: `{}`.", .expected, .found)]
+    InvalidVectorDimensions { expected: usize, found: usize },
     #[error("{0}")]
     InvalidFilter(String),
-    #[error("Invalid type for filter subexpression: `expected {}, found: {1}`.", .0.join(", "))]
+    #[error("Invalid type for filter subexpression: expected: {}, found: {1}.", .0.join(", "))]
     InvalidFilterExpression(&'static [&'static str], Value),
     #[error("Attribute `{}` is not sortable. {}",
         .field,

milli/src/external_documents_ids.rs

@@ -106,30 +106,22 @@ impl<'a> ExternalDocumentsIds<'a> {
         map
     }
 
-    /// Return an fst of the combined hard and soft deleted ID.
-    pub fn to_fst<'b>(&'b self) -> fst::Result<Cow<'b, fst::Map<Cow<'a, [u8]>>>> {
-        if self.soft.is_empty() {
-            return Ok(Cow::Borrowed(&self.hard));
-        }
-        let union_op = self.hard.op().add(&self.soft).r#union();
-
-        let mut iter = union_op.into_stream();
-        let mut new_hard_builder = fst::MapBuilder::memory();
-        while let Some((external_id, marked_docids)) = iter.next() {
-            let value = indexed_last_value(marked_docids).unwrap();
-            if value != DELETED_ID {
-                new_hard_builder.insert(external_id, value)?;
-            }
-        }
-
-        drop(iter);
-
-        Ok(Cow::Owned(new_hard_builder.into_map().map_data(Cow::Owned)?))
-    }
-
     fn merge_soft_into_hard(&mut self) -> fst::Result<()> {
         if self.soft.len() >= self.hard.len() / 2 {
-            self.hard = self.to_fst()?.into_owned();
+            let union_op = self.hard.op().add(&self.soft).r#union();
+
+            let mut iter = union_op.into_stream();
+            let mut new_hard_builder = fst::MapBuilder::memory();
+            while let Some((external_id, marked_docids)) = iter.next() {
+                let value = indexed_last_value(marked_docids).unwrap();
+                if value != DELETED_ID {
+                    new_hard_builder.insert(external_id, value)?;
+                }
+            }
+
+            drop(iter);
+
+            self.hard = new_hard_builder.into_map().map_data(Cow::Owned)?;
             self.soft = fst::Map::default().map_data(Cow::Owned)?;
         }
 

milli/src/index.rs

@@ -8,10 +8,12 @@ use charabia::{Language, Script};
 use heed::flags::Flags;
 use heed::types::*;
 use heed::{CompactionOption, Database, PolyDatabase, RoTxn, RwTxn};
+use rand_pcg::Pcg32;
 use roaring::RoaringBitmap;
 use rstar::RTree;
 use time::OffsetDateTime;
 
+use crate::distance::DotProduct;
 use crate::error::{InternalError, UserError};
 use crate::facet::FacetType;
 use crate::fields_ids_map::FieldsIdsMap;
@@ -26,6 +28,9 @@ use crate::{
     Result, RoaringBitmapCodec, RoaringBitmapLenCodec, Search, U8StrStrCodec, BEU16, BEU32,
 };
 
+/// The HNSW data-structure that we serialize, fill and search in.
+pub type Hnsw = hnsw::Hnsw<DotProduct, Vec<f32>, Pcg32, 12, 24>;
+
 pub const DEFAULT_MIN_WORD_LEN_ONE_TYPO: u8 = 5;
 pub const DEFAULT_MIN_WORD_LEN_TWO_TYPOS: u8 = 9;
 
@@ -42,6 +47,7 @@ pub mod main_key {
     pub const FIELDS_IDS_MAP_KEY: &str = "fields-ids-map";
     pub const GEO_FACETED_DOCUMENTS_IDS_KEY: &str = "geo-faceted-documents-ids";
     pub const GEO_RTREE_KEY: &str = "geo-rtree";
+    pub const VECTOR_HNSW_KEY: &str = "vector-hnsw";
     pub const HARD_EXTERNAL_DOCUMENTS_IDS_KEY: &str = "hard-external-documents-ids";
     pub const NUMBER_FACETED_DOCUMENTS_IDS_PREFIX: &str = "number-faceted-documents-ids";
     pub const PRIMARY_KEY_KEY: &str = "primary-key";
@@ -86,6 +92,7 @@ pub mod db_name {
     pub const FACET_ID_STRING_DOCIDS: &str = "facet-id-string-docids";
     pub const FIELD_ID_DOCID_FACET_F64S: &str = "field-id-docid-facet-f64s";
     pub const FIELD_ID_DOCID_FACET_STRINGS: &str = "field-id-docid-facet-strings";
+    pub const VECTOR_ID_DOCID: &str = "vector-id-docids";
     pub const DOCUMENTS: &str = "documents";
     pub const SCRIPT_LANGUAGE_DOCIDS: &str = "script_language_docids";
 }
@@ -93,10 +100,10 @@ pub mod db_name {
 #[derive(Clone)]
 pub struct Index {
     /// The LMDB environment which this index is associated with.
-    pub env: heed::Env,
+    pub(crate) env: heed::Env,
 
     /// Contains many different types (e.g. the fields ids map).
-    pub main: PolyDatabase,
+    pub(crate) main: PolyDatabase,
 
     /// A word and all the documents ids containing the word.
     pub word_docids: Database<Str, RoaringBitmapCodec>,
@@ -149,8 +156,11 @@ pub struct Index {
     /// Maps the document id, the facet field id and the strings.
     pub field_id_docid_facet_strings: Database<FieldDocIdFacetStringCodec, Str>,
 
+    /// Maps a vector id to the document id that have it.
+    pub vector_id_docid: Database<OwnedType<BEU32>, OwnedType<BEU32>>,
+
     /// Maps the document id to the document as an obkv store.
-    pub documents: Database<OwnedType<BEU32>, ObkvCodec>,
+    pub(crate) documents: Database<OwnedType<BEU32>, ObkvCodec>,
 }
 
 impl Index {
@@ -162,7 +172,7 @@ impl Index {
     ) -> Result<Index> {
         use db_name::*;
 
-        options.max_dbs(23);
+        options.max_dbs(24);
         unsafe { options.flag(Flags::MdbAlwaysFreePages) };
 
         let env = options.open(path)?;
@@ -198,11 +208,11 @@ impl Index {
             env.create_database(&mut wtxn, Some(FACET_ID_IS_NULL_DOCIDS))?;
         let facet_id_is_empty_docids =
             env.create_database(&mut wtxn, Some(FACET_ID_IS_EMPTY_DOCIDS))?;
-
         let field_id_docid_facet_f64s =
             env.create_database(&mut wtxn, Some(FIELD_ID_DOCID_FACET_F64S))?;
         let field_id_docid_facet_strings =
             env.create_database(&mut wtxn, Some(FIELD_ID_DOCID_FACET_STRINGS))?;
+        let vector_id_docid = env.create_database(&mut wtxn, Some(VECTOR_ID_DOCID))?;
         let documents = env.create_database(&mut wtxn, Some(DOCUMENTS))?;
         wtxn.commit()?;
 
@@ -231,6 +241,7 @@ impl Index {
             facet_id_is_empty_docids,
             field_id_docid_facet_f64s,
             field_id_docid_facet_strings,
+            vector_id_docid,
             documents,
         })
     }
@@ -502,6 +513,26 @@ impl Index {
         }
     }
 
+    /* vector HNSW */
+
+    /// Writes the provided `hnsw`.
+    pub(crate) fn put_vector_hnsw(&self, wtxn: &mut RwTxn, hnsw: &Hnsw) -> heed::Result<()> {
+        self.main.put::<_, Str, SerdeBincode<Hnsw>>(wtxn, main_key::VECTOR_HNSW_KEY, hnsw)
+    }
+
+    /// Delete the `hnsw`.
+    pub(crate) fn delete_vector_hnsw(&self, wtxn: &mut RwTxn) -> heed::Result<bool> {
+        self.main.delete::<_, Str>(wtxn, main_key::VECTOR_HNSW_KEY)
+    }
+
+    /// Returns the `hnsw`.
+    pub fn vector_hnsw(&self, rtxn: &RoTxn) -> Result<Option<Hnsw>> {
+        match self.main.get::<_, Str, SerdeBincode<Hnsw>>(rtxn, main_key::VECTOR_HNSW_KEY)? {
+            Some(hnsw) => Ok(Some(hnsw)),
+            None => Ok(None),
+        }
+    }
+
     /* field distribution */
 
     /// Writes the field distribution which associates every field name with

milli/src/lib.rs

@@ -10,6 +10,7 @@ pub mod documents;
 
 mod asc_desc;
 mod criterion;
+pub mod distance;
 mod error;
 mod external_documents_ids;
 pub mod facet;

milli/src/search/mod.rs

@@ -22,6 +22,7 @@ pub mod new;
 
 pub struct Search<'a> {
     query: Option<String>,
+    vector: Option<Vec<f32>>,
     // this should be linked to the String in the query
     filter: Option<Filter<'a>>,
     offset: usize,
@@ -39,6 +40,7 @@ impl<'a> Search<'a> {
     pub fn new(rtxn: &'a heed::RoTxn, index: &'a Index) -> Search<'a> {
         Search {
             query: None,
+            vector: None,
             filter: None,
             offset: 0,
             limit: 20,
@@ -57,6 +59,11 @@ impl<'a> Search<'a> {
         self
     }
 
+    pub fn vector(&mut self, vector: impl Into<Vec<f32>>) -> &mut Search<'a> {
+        self.vector = Some(vector.into());
+        self
+    }
+
     pub fn offset(&mut self, offset: usize) -> &mut Search<'a> {
         self.offset = offset;
         self
@@ -106,6 +113,7 @@ impl<'a> Search<'a> {
             execute_search(
                 &mut ctx,
                 &self.query,
+                &self.vector,
                 self.terms_matching_strategy,
                 self.exhaustive_number_hits,
                 &self.filter,
@@ -132,6 +140,7 @@ impl fmt::Debug for Search<'_> {
     fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
         let Search {
             query,
+            vector: _,
             filter,
             offset,
             limit,
@@ -145,6 +154,7 @@ impl fmt::Debug for Search<'_> {
         } = self;
         f.debug_struct("Search")
             .field("query", query)
+            .field("vector", &"[...]")
             .field("filter", filter)
             .field("offset", offset)
             .field("limit", limit)

milli/src/search/new/distinct.rs

@@ -26,6 +26,7 @@ pub fn apply_distinct_rule(
     ctx: &mut SearchContext,
     field_id: u16,
     candidates: &RoaringBitmap,
+    // TODO: add a universe here, such that the `excluded` are a subset of the universe?
 ) -> Result<DistinctOutput> {
     let mut excluded = RoaringBitmap::new();
     let mut remaining = RoaringBitmap::new();

milli/src/search/new/exact_attribute.rs

@@ -206,7 +206,7 @@ impl State {
             )?;
             intersection &= &candidates;
             if !intersection.is_empty() {
-                // Although not really worth it in terms of performance,
+                // TODO: although not really worth it in terms of performance,
                 // if would be good to put this in cache for the sake of consistency
                 let candidates_with_exact_word_count = if count_all_positions < u8::MAX as usize {
                     ctx.index

milli/src/search/new/interner.rs

@@ -32,7 +32,7 @@ impl<T> Interned<T> {
 #[derive(Clone)]
 pub struct DedupInterner<T> {
     stable_store: Vec<T>,
-    lookup: FxHashMap<T, Interned<T>>,
+    lookup: FxHashMap<T, Interned<T>>, // TODO: Arc
 }
 impl<T> Default for DedupInterner<T> {
     fn default() -> Self {

milli/src/search/new/limits.rs

@@ -1,4 +1,5 @@
 /// Maximum number of tokens we consider in a single search.
+// TODO: Loic, find proper value here so we don't overflow the interner.
 pub const MAX_TOKEN_COUNT: usize = 1_000;
 
 /// Maximum number of prefixes that can be derived from a single word.

milli/src/search/new/matches/mod.rs

@@ -509,6 +509,7 @@ mod tests {
             let crate::search::PartialSearchResult { located_query_terms, .. } = execute_search(
                 &mut ctx,
                 &Some(query.to_string()),
+                &None,
                 crate::TermsMatchingStrategy::default(),
                 false,
                 &None,

milli/src/search/new/mod.rs

@@ -28,6 +28,7 @@ use db_cache::DatabaseCache;
 use exact_attribute::ExactAttribute;
 use graph_based_ranking_rule::{Exactness, Fid, Position, Proximity, Typo};
 use heed::RoTxn;
+use hnsw::Searcher;
 use interner::{DedupInterner, Interner};
 pub use logger::visual::VisualSearchLogger;
 pub use logger::{DefaultSearchLogger, SearchLogger};
@@ -39,13 +40,16 @@ use ranking_rules::{
 use resolve_query_graph::{compute_query_graph_docids, PhraseDocIdsCache};
 use roaring::RoaringBitmap;
 use sort::Sort;
+use space::Neighbor;
 
 use self::geo_sort::GeoSort;
 pub use self::geo_sort::Strategy as GeoSortStrategy;
 use self::graph_based_ranking_rule::Words;
 use self::interner::Interned;
 use crate::search::new::distinct::apply_distinct_rule;
-use crate::{AscDesc, DocumentId, Filter, Index, Member, Result, TermsMatchingStrategy, UserError};
+use crate::{
+    AscDesc, DocumentId, Filter, Index, Member, Result, TermsMatchingStrategy, UserError, BEU32,
+};
 
 /// A structure used throughout the execution of a search query.
 pub struct SearchContext<'ctx> {
@@ -349,6 +353,7 @@ fn resolve_sort_criteria<'ctx, Query: RankingRuleQueryTrait>(
 pub fn execute_search(
     ctx: &mut SearchContext,
     query: &Option<String>,
+    vector: &Option<Vec<f32>>,
     terms_matching_strategy: TermsMatchingStrategy,
     exhaustive_number_hits: bool,
     filters: &Option<Filter>,
@@ -428,6 +433,33 @@ pub fn execute_search(
 
     let BucketSortOutput { docids, mut all_candidates } = bucket_sort_output;
 
+    let docids = match vector {
+        Some(vector) => {
+            // return the nearest documents that are also part of the candidates.
+            let mut searcher = Searcher::new();
+            let hnsw = ctx.index.vector_hnsw(ctx.txn)?.unwrap_or_default();
+            let ef = hnsw.len().min(100);
+            let mut dest = vec![Neighbor { index: 0, distance: 0 }; ef];
+            let neighbors = hnsw.nearest(vector, ef, &mut searcher, &mut dest[..]);
+
+            let mut docids = Vec::new();
+            for Neighbor { index, distance: _ } in neighbors.iter() {
+                let index = BEU32::new(*index as u32);
+                let docid = ctx.index.vector_id_docid.get(ctx.txn, &index)?.unwrap().get();
+                if universe.contains(docid) {
+                    docids.push(docid);
+                    if docids.len() == (from + length) {
+                        break;
+                    }
+                }
+            }
+
+            docids.into_iter().skip(from).take(length).collect()
+        }
+        // return the search docids if the vector field is not specified
+        None => docids,
+    };
+
     // The candidates is the universe unless the exhaustive number of hits
     // is requested and a distinct attribute is set.
     if exhaustive_number_hits {

milli/src/search/new/query_graph.rs

@@ -92,7 +92,7 @@ impl QueryGraph {
     /// which contains ngrams.
     pub fn from_query(
         ctx: &mut SearchContext,
-        // The terms here must be consecutive
+        // NOTE: the terms here must be consecutive
         terms: &[LocatedQueryTerm],
     ) -> Result<(QueryGraph, Vec<LocatedQueryTerm>)> {
         let mut new_located_query_terms = terms.to_vec();
@@ -103,7 +103,7 @@ impl QueryGraph {
         let root_node = 0;
         let end_node = 1;
 
-        // Ee could consider generalizing to 4,5,6,7,etc. ngrams
+        // TODO: we could consider generalizing to 4,5,6,7,etc. ngrams
         let (mut prev2, mut prev1, mut prev0): (Vec<u16>, Vec<u16>, Vec<u16>) =
             (vec![], vec![], vec![root_node]);
 

milli/src/search/new/query_term/mod.rs

@@ -132,6 +132,7 @@ impl QueryTermSubset {
         if full_query_term.ngram_words.is_some() {
             return None;
         }
+        // TODO: included in subset
         if let Some(phrase) = full_query_term.zero_typo.phrase {
             self.zero_typo_subset.contains_phrase(phrase).then_some(ExactTerm::Phrase(phrase))
         } else if let Some(word) = full_query_term.zero_typo.exact {
@@ -181,6 +182,7 @@ impl QueryTermSubset {
         let word = match &self.zero_typo_subset {
             NTypoTermSubset::All => Some(use_prefix_db),
             NTypoTermSubset::Subset { words, phrases: _ } => {
+                // TODO: use a subset of prefix words instead
                 if words.contains(&use_prefix_db) {
                     Some(use_prefix_db)
                 } else {
@@ -202,6 +204,7 @@ impl QueryTermSubset {
         ctx: &mut SearchContext,
     ) -> Result<BTreeSet<Word>> {
         let mut result = BTreeSet::default();
+        // TODO: a compute_partially funtion
         if !self.one_typo_subset.is_empty() || !self.two_typo_subset.is_empty() {
             self.original.compute_fully_if_needed(ctx)?;
         }
@@ -297,6 +300,7 @@ impl QueryTermSubset {
         let mut result = BTreeSet::default();
 
         if !self.one_typo_subset.is_empty() {
+            // TODO: compute less than fully if possible
             self.original.compute_fully_if_needed(ctx)?;
         }
         let original = ctx.term_interner.get_mut(self.original);

milli/src/search/new/query_term/parse_query.rs

@@ -139,6 +139,7 @@ pub fn number_of_typos_allowed<'ctx>(
     let min_len_one_typo = ctx.index.min_word_len_one_typo(ctx.txn)?;
     let min_len_two_typos = ctx.index.min_word_len_two_typos(ctx.txn)?;
 
+    // TODO: should `exact_words` also disable prefix search, ngrams, split words, or synonyms?
     let exact_words = ctx.index.exact_words(ctx.txn)?;
 
     Ok(Box::new(move |word: &str| {
@@ -249,6 +250,8 @@ impl PhraseBuilder {
         } else {
             // token has kind Word
             let word = ctx.word_interner.insert(token.lemma().to_string());
+            // TODO: in a phrase, check that every word exists
+            // otherwise return an empty term
             self.words.push(Some(word));
         }
     }

milli/src/search/new/ranking_rule_graph/cheapest_paths.rs

@@ -1,48 +1,5 @@
-/** Implements a "PathVisitor" which finds all paths of a certain cost
-from the START to END node of a ranking rule graph.
+#![allow(clippy::too_many_arguments)]
 
-A path is a list of conditions. A condition is the data associated with
-an edge, given by the ranking rule. Some edges don't have a condition associated
-with them, they are "unconditional". These kinds of edges are used to "skip" a node.
-
-The algorithm uses a depth-first search. It benefits from two main optimisations:
-- The list of all possible costs to go from any node to the END node is precomputed
-- The `DeadEndsCache` reduces the number of valid paths drastically, by making some edges
-untraversable depending on what other edges were selected.
-
-These two optimisations are meant to avoid traversing edges that wouldn't lead
-to a valid path. In practically all cases, we avoid the exponential complexity
-that is inherent to depth-first search in a large ranking rule graph.
-
-The DeadEndsCache is a sort of prefix tree which associates a list of forbidden
-conditions to a list of traversed conditions.
-For example, the DeadEndsCache could say the following:
-- Immediately, from the start, the conditions `[a,b]` are forbidden
-    - if we take the condition `c`, then the conditions `[e]` are also forbidden
-        - and if after that, we take `f`, then `[h,i]` are also forbidden
-            - etc.
-    - if we take `g`, then `[f]` is also forbidden
-        - etc.
-    - etc.
-As we traverse the graph, we also traverse the `DeadEndsCache` and keep a list of forbidden
-conditions in memory. Then, we know to avoid all edges which have a condition that is forbidden.
-
-When a path is found from START to END, we give it to the `visit` closure.
-This closure takes a mutable reference to the `DeadEndsCache`. This means that
-the caller can update this cache. Therefore, we must handle the case where the
-DeadEndsCache has been updated. This means potentially backtracking up to the point
-where the traversed conditions are all allowed by the new DeadEndsCache.
-
-The algorithm also implements the `TermsMatchingStrategy` logic.
-Some edges are augmented with a list of "nodes_to_skip". Skipping
-a node means "reaching this node through an unconditional edge". If we have
-already traversed (ie. not skipped) a node that is in this list, then we know that we
-can't traverse this edge. Otherwise, we traverse the edge but make sure to skip any
-future node that was present in the "nodes_to_skip" list.
-
-The caller can decide to stop the path finding algorithm
-by returning a `ControlFlow::Break` from the `visit` closure.
-*/
 use std::collections::{BTreeSet, VecDeque};
 use std::iter::FromIterator;
 use std::ops::ControlFlow;
@@ -55,41 +12,30 @@ use crate::search::new::query_graph::QueryNode;
 use crate::search::new::small_bitmap::SmallBitmap;
 use crate::Result;
 
-/// Closure which processes a path found by the `PathVisitor`
 type VisitFn<'f, G> = &'f mut dyn FnMut(
-    // the path as a list of conditions
     &[Interned<<G as RankingRuleGraphTrait>::Condition>],
     &mut RankingRuleGraph<G>,
-    // a mutable reference to the DeadEndsCache, to update it in case the given
-    // path doesn't resolve to any valid document ids
     &mut DeadEndsCache<<G as RankingRuleGraphTrait>::Condition>,
 ) -> Result<ControlFlow<()>>;
 
-/// A structure which is kept but not updated during the traversal of the graph.
-/// It can however be updated by the `visit` closure once a valid path has been found.
 struct VisitorContext<'a, G: RankingRuleGraphTrait> {
     graph: &'a mut RankingRuleGraph<G>,
     all_costs_from_node: &'a MappedInterner<QueryNode, Vec<u64>>,
     dead_ends_cache: &'a mut DeadEndsCache<G::Condition>,
 }
 
-/// The internal state of the traversal algorithm
 struct VisitorState<G: RankingRuleGraphTrait> {
-    /// Budget from the current node to the end node
     remaining_cost: u64,
-    /// Previously visited conditions, in order.
+
     path: Vec<Interned<G::Condition>>,
-    /// Previously visited conditions, as an efficient and compact set.
+
     visited_conditions: SmallBitmap<G::Condition>,
-    /// Previously visited (ie not skipped) nodes, as an efficient and compact set.
     visited_nodes: SmallBitmap<QueryNode>,
-    /// The conditions that cannot be visited anymore
+
     forbidden_conditions: SmallBitmap<G::Condition>,
-    /// The nodes that cannot be visited anymore (they must be skipped)
-    nodes_to_skip: SmallBitmap<QueryNode>,
+    forbidden_conditions_to_nodes: SmallBitmap<QueryNode>,
 }
 
-/// See module documentation
 pub struct PathVisitor<'a, G: RankingRuleGraphTrait> {
     state: VisitorState<G>,
     ctx: VisitorContext<'a, G>,
@@ -110,13 +56,14 @@ impl<'a, G: RankingRuleGraphTrait> PathVisitor<'a, G> {
                 forbidden_conditions: SmallBitmap::for_interned_values_in(
                     &graph.conditions_interner,
                 ),
-                nodes_to_skip: SmallBitmap::for_interned_values_in(&graph.query_graph.nodes),
+                forbidden_conditions_to_nodes: SmallBitmap::for_interned_values_in(
+                    &graph.query_graph.nodes,
+                ),
             },
             ctx: VisitorContext { graph, all_costs_from_node, dead_ends_cache },
         }
     }
 
-    /// See module documentation
     pub fn visit_paths(mut self, visit: VisitFn<G>) -> Result<()> {
         let _ =
             self.state.visit_node(self.ctx.graph.query_graph.root_node, visit, &mut self.ctx)?;
@@ -125,31 +72,22 @@ impl<'a, G: RankingRuleGraphTrait> PathVisitor<'a, G> {
 }
 
 impl<G: RankingRuleGraphTrait> VisitorState<G> {
-    /// Visits a node: traverse all its valid conditional and unconditional edges.
-    ///
-    /// Returns ControlFlow::Break if the path finding algorithm should stop.
-    /// Returns whether a valid path was found from this node otherwise.
     fn visit_node(
         &mut self,
         from_node: Interned<QueryNode>,
         visit: VisitFn<G>,
         ctx: &mut VisitorContext<G>,
     ) -> Result<ControlFlow<(), bool>> {
-        // any valid path will be found from this point
-        // if a valid path was found, then we know that the DeadEndsCache may have been updated,
-        // and we will need to do more work to potentially backtrack
         let mut any_valid = false;
 
         let edges = ctx.graph.edges_of_node.get(from_node).clone();
         for edge_idx in edges.iter() {
-            // could be none if the edge was deleted
             let Some(edge) = ctx.graph.edges_store.get(edge_idx).clone() else { continue };
 
             if self.remaining_cost < edge.cost as u64 {
                 continue;
             }
             self.remaining_cost -= edge.cost as u64;
-
             let cf = match edge.condition {
                 Some(condition) => self.visit_condition(
                     condition,
@@ -181,10 +119,6 @@ impl<G: RankingRuleGraphTrait> VisitorState<G> {
         Ok(ControlFlow::Continue(any_valid))
     }
 
-    /// Visits an unconditional edge.
-    ///
-    /// Returns ControlFlow::Break if the path finding algorithm should stop.
-    /// Returns whether a valid path was found from this node otherwise.
     fn visit_no_condition(
         &mut self,
         dest_node: Interned<QueryNode>,
@@ -200,29 +134,20 @@ impl<G: RankingRuleGraphTrait> VisitorState<G> {
         {
             return Ok(ControlFlow::Continue(false));
         }
-        // We've reached the END node!
         if dest_node == ctx.graph.query_graph.end_node {
             let control_flow = visit(&self.path, ctx.graph, ctx.dead_ends_cache)?;
-            // We could change the return type of the visit closure such that the caller
-            // tells us whether the dead ends cache was updated or not.
-            // Alternatively, maybe the DeadEndsCache should have a generation number
-            // to it, so that we don't need to play with these booleans at all.
             match control_flow {
                 ControlFlow::Continue(_) => Ok(ControlFlow::Continue(true)),
                 ControlFlow::Break(_) => Ok(ControlFlow::Break(())),
             }
         } else {
-            let old_fbct = self.nodes_to_skip.clone();
-            self.nodes_to_skip.union(edge_new_nodes_to_skip);
+            let old_fbct = self.forbidden_conditions_to_nodes.clone();
+            self.forbidden_conditions_to_nodes.union(edge_new_nodes_to_skip);
             let cf = self.visit_node(dest_node, visit, ctx)?;
-            self.nodes_to_skip = old_fbct;
+            self.forbidden_conditions_to_nodes = old_fbct;
             Ok(cf)
         }
     }
-    /// Visits a conditional edge.
-    ///
-    /// Returns ControlFlow::Break if the path finding algorithm should stop.
-    /// Returns whether a valid path was found from this node otherwise.
     fn visit_condition(
         &mut self,
         condition: Interned<G::Condition>,
@@ -234,7 +159,7 @@ impl<G: RankingRuleGraphTrait> VisitorState<G> {
         assert!(dest_node != ctx.graph.query_graph.end_node);
 
         if self.forbidden_conditions.contains(condition)
-            || self.nodes_to_skip.contains(dest_node)
+            || self.forbidden_conditions_to_nodes.contains(dest_node)
             || edge_new_nodes_to_skip.intersects(&self.visited_nodes)
         {
             return Ok(ControlFlow::Continue(false));
@@ -255,19 +180,19 @@ impl<G: RankingRuleGraphTrait> VisitorState<G> {
         self.visited_nodes.insert(dest_node);
         self.visited_conditions.insert(condition);
 
-        let old_forb_cond = self.forbidden_conditions.clone();
+        let old_fc = self.forbidden_conditions.clone();
         if let Some(next_forbidden) =
             ctx.dead_ends_cache.forbidden_conditions_after_prefix(self.path.iter().copied())
         {
             self.forbidden_conditions.union(&next_forbidden);
         }
-        let old_nodes_to_skip = self.nodes_to_skip.clone();
-        self.nodes_to_skip.union(edge_new_nodes_to_skip);
+        let old_fctn = self.forbidden_conditions_to_nodes.clone();
+        self.forbidden_conditions_to_nodes.union(edge_new_nodes_to_skip);
 
         let cf = self.visit_node(dest_node, visit, ctx)?;
 
-        self.nodes_to_skip = old_nodes_to_skip;
-        self.forbidden_conditions = old_forb_cond;
+        self.forbidden_conditions_to_nodes = old_fctn;
+        self.forbidden_conditions = old_fc;
 
         self.visited_conditions.remove(condition);
         self.visited_nodes.remove(dest_node);

milli/src/search/new/ranking_rule_graph/condition_docids_cache.rs

@@ -9,8 +9,12 @@ use crate::search::new::query_term::LocatedQueryTermSubset;
 use crate::search::new::SearchContext;
 use crate::Result;
 
+// TODO: give a generation to each universe, then be able to get the exact
+// delta of docids between two universes of different generations!
+
 /// A cache storing the document ids associated with each ranking rule edge
 pub struct ConditionDocIdsCache<G: RankingRuleGraphTrait> {
+    // TOOD: should be a mapped interner?
     pub cache: FxHashMap<Interned<G::Condition>, ComputedCondition>,
     _phantom: PhantomData<G>,
 }
@@ -50,7 +54,7 @@ impl<G: RankingRuleGraphTrait> ConditionDocIdsCache<G> {
         }
         let condition = graph.conditions_interner.get_mut(interned_condition);
         let computed = G::resolve_condition(ctx, condition, universe)?;
-        // Can we put an assert here for computed.universe_len == universe.len() ?
+        // TODO: if computed.universe_len != universe.len() ?
         let _ = self.cache.insert(interned_condition, computed);
         let computed = &self.cache[&interned_condition];
         Ok(computed)

milli/src/search/new/ranking_rule_graph/dead_ends_cache.rs

@@ -2,7 +2,6 @@ use crate::search::new::interner::{FixedSizeInterner, Interned};
 use crate::search::new::small_bitmap::SmallBitmap;
 
 pub struct DeadEndsCache<T> {
-    // conditions and next could/should be part of the same vector
     conditions: Vec<Interned<T>>,
     next: Vec<Self>,
     pub forbidden: SmallBitmap<T>,
@@ -28,7 +27,7 @@ impl<T> DeadEndsCache<T> {
         self.forbidden.insert(condition);
     }
 
-    fn advance(&mut self, condition: Interned<T>) -> Option<&mut Self> {
+    pub fn advance(&mut self, condition: Interned<T>) -> Option<&mut Self> {
         if let Some(idx) = self.conditions.iter().position(|c| *c == condition) {
             Some(&mut self.next[idx])
         } else {

milli/src/search/new/ranking_rule_graph/fid/mod.rs

@@ -69,9 +69,14 @@ impl RankingRuleGraphTrait for FidGraph {
 
         let mut edges = vec![];
         for fid in all_fields {
+            // TODO: We can improve performances and relevancy by storing
+            //       the term subsets associated to each field ids fetched.
             edges.push((
-                fid as u32 * term.term_ids.len() as u32,
-                conditions_interner.insert(FidCondition { term: term.clone(), fid }),
+                fid as u32 * term.term_ids.len() as u32, // TODO improve the fid score i.e. fid^10.
+                conditions_interner.insert(FidCondition {
+                    term: term.clone(), // TODO remove this ugly clone
+                    fid,
+                }),
             ));
         }
 

milli/src/search/new/ranking_rule_graph/position/mod.rs

@@ -94,9 +94,14 @@ impl RankingRuleGraphTrait for PositionGraph {
         let mut edges = vec![];
 
         for (cost, positions) in positions_for_costs {
+            // TODO: We can improve performances and relevancy by storing
+            //       the term subsets associated to each position fetched
             edges.push((
                 cost,
-                conditions_interner.insert(PositionCondition { term: term.clone(), positions }),
+                conditions_interner.insert(PositionCondition {
+                    term: term.clone(), // TODO remove this ugly clone
+                    positions,
+                }),
             ));
         }
 

milli/src/search/new/ranking_rule_graph/proximity/compute_docids.rs

@@ -65,6 +65,13 @@ pub fn compute_docids(
         }
     }
 
+    // TODO: add safeguard in case the cartesian product is too large!
+    // even if we restrict the word derivations to a maximum of 100, the size of the
+    // caterisan product could reach a maximum of 10_000 derivations, which is way too much.
+    // Maybe prioritise the product of zero typo derivations, then the product of zero-typo/one-typo
+    // + one-typo/zero-typo, then one-typo/one-typo, then ... until an arbitrary limit has been
+    // reached
+
     for (left_phrase, left_word) in last_words_of_term_derivations(ctx, &left_term.term_subset)? {
         // Before computing the edges, check that the left word and left phrase
         // aren't disjoint with the universe, but only do it if there is more than
@@ -104,6 +111,8 @@ pub fn compute_docids(
     Ok(ComputedCondition {
         docids,
         universe_len: universe.len(),
+        // TODO: think about whether we want to reduce the subset,
+        // we probably should!
         start_term_subset: Some(left_term.clone()),
         end_term_subset: right_term.clone(),
     })
@@ -194,7 +203,12 @@ fn compute_non_prefix_edges(
             *docids |= new_docids;
         }
     }
-    if backward_proximity >= 1 && left_phrase.is_none() && right_phrase.is_none() {
+    if backward_proximity >= 1
+            // TODO: for now, we don't do any swapping when either term is a phrase
+            // but maybe we should. We'd need to look at the first/last word of the phrase
+            // depending on the context.
+            && left_phrase.is_none() && right_phrase.is_none()
+    {
         if let Some(new_docids) =
             ctx.get_db_word_pair_proximity_docids(word2, word1, backward_proximity)?
         {

milli/src/search/new/resolve_query_graph.rs

@@ -33,6 +33,8 @@ pub fn compute_query_term_subset_docids(
     ctx: &mut SearchContext,
     term: &QueryTermSubset,
 ) -> Result<RoaringBitmap> {
+    // TODO Use the roaring::MultiOps trait
+
     let mut docids = RoaringBitmap::new();
     for word in term.all_single_words_except_prefix_db(ctx)? {
         if let Some(word_docids) = ctx.word_docids(word)? {
@@ -57,6 +59,8 @@ pub fn compute_query_term_subset_docids_within_field_id(
     term: &QueryTermSubset,
     fid: u16,
 ) -> Result<RoaringBitmap> {
+    // TODO Use the roaring::MultiOps trait
+
     let mut docids = RoaringBitmap::new();
     for word in term.all_single_words_except_prefix_db(ctx)? {
         if let Some(word_fid_docids) = ctx.get_db_word_fid_docids(word.interned(), fid)? {
@@ -67,6 +71,7 @@ pub fn compute_query_term_subset_docids_within_field_id(
     for phrase in term.all_phrases(ctx)? {
         // There may be false positives when resolving a phrase, so we're not
         // guaranteed that all of its words are within a single fid.
+        // TODO: fix this?
         if let Some(word) = phrase.words(ctx).iter().flatten().next() {
             if let Some(word_fid_docids) = ctx.get_db_word_fid_docids(*word, fid)? {
                 docids |= ctx.get_phrase_docids(phrase)? & word_fid_docids;
@@ -90,6 +95,7 @@ pub fn compute_query_term_subset_docids_within_position(
     term: &QueryTermSubset,
     position: u16,
 ) -> Result<RoaringBitmap> {
+    // TODO Use the roaring::MultiOps trait
     let mut docids = RoaringBitmap::new();
     for word in term.all_single_words_except_prefix_db(ctx)? {
         if let Some(word_position_docids) =
@@ -102,6 +108,7 @@ pub fn compute_query_term_subset_docids_within_position(
     for phrase in term.all_phrases(ctx)? {
         // It's difficult to know the expected position of the words in the phrase,
         // so instead we just check the first one.
+        // TODO: fix this?
         if let Some(word) = phrase.words(ctx).iter().flatten().next() {
             if let Some(word_position_docids) = ctx.get_db_word_position_docids(*word, position)? {
                 docids |= ctx.get_phrase_docids(phrase)? & word_position_docids
@@ -125,6 +132,9 @@ pub fn compute_query_graph_docids(
     q: &QueryGraph,
     universe: &RoaringBitmap,
 ) -> Result<RoaringBitmap> {
+    // TODO: there must be a faster way to compute this big
+    // roaring bitmap expression
+
     let mut nodes_resolved = SmallBitmap::for_interned_values_in(&q.nodes);
     let mut path_nodes_docids = q.nodes.map(|_| RoaringBitmap::new());
 

milli/src/search/new/sort.rs

@@ -141,6 +141,10 @@ impl<'ctx, Query: RankingRuleQueryTrait> RankingRule<'ctx, Query> for Sort<'ctx,
         universe: &RoaringBitmap,
     ) -> Result<Option<RankingRuleOutput<Query>>> {
         let iter = self.iter.as_mut().unwrap();
+        // TODO: we should make use of the universe in the function below
+        // good for correctness, but ideally iter.next_bucket would take the current universe into account,
+        // as right now it could return buckets that don't intersect with the universe, meaning we will make many
+        // unneeded calls.
         if let Some(mut bucket) = iter.next_bucket()? {
             bucket.candidates &= universe;
             Ok(Some(bucket))

milli/src/search/new/tests/distinct.rs

@@ -527,7 +527,7 @@ fn test_distinct_all_candidates() {
     let SearchResult { documents_ids, candidates, .. } = s.execute().unwrap();
     let candidates = candidates.iter().collect::<Vec<_>>();
     insta::assert_snapshot!(format!("{documents_ids:?}"), @"[14, 26, 4, 7, 17, 23, 1, 19, 25, 8, 20, 24]");
-    // This is incorrect, but unfortunately impossible to do better efficiently.
+    // TODO: this is incorrect!
     insta::assert_snapshot!(format!("{candidates:?}"), @"[1, 4, 7, 8, 14, 17, 19, 20, 23, 24, 25, 26]");
 }
 

milli/src/search/new/tests/proximity.rs

@@ -122,11 +122,11 @@ fn create_edge_cases_index() -> TempIndex {
             sta stb stc ste stf stg sth sti stj stk stl stm stn sto stp stq str stst stt stu stv stw stx sty stz
             "
         },
-        // The next 5 documents lay out a trap with the split word, phrase search, or synonym `sun flower`.
-        // If the search query is "sunflower", the split word "Sun Flower" will match some documents.
+        // The next 5 documents lay out a trap with the split word, phrase search, or synonym `sun flower`. 
+        // If the search query is "sunflower", the split word "Sun Flower" will match some documents. 
         // If the query is `sunflower wilting`, then we should make sure that
-        // the proximity condition `flower wilting: sprx N` also comes with the condition
-        // `sun wilting: sprx N+1`, but this is not the exact condition we use for now.
+        // the sprximity condition `flower wilting: sprx N` also comes with the condition
+        // `sun wilting: sprx N+1`. TODO: this is not the exact condition we use for now. 
         // We only check that the phrase `sun flower` exists and `flower wilting: sprx N`, which
         // is better than nothing but not the best.
         {
@@ -139,7 +139,7 @@ fn create_edge_cases_index() -> TempIndex {
         },
         {
             "id": 3,
-            // This document matches the query `sunflower wilting`, but the sprximity condition
+            // This document matches the query `sunflower wilting`, but the sprximity condition 
             // between `sunflower` and `wilting` cannot be through the split-word `Sun Flower`
             // which would reduce to only `flower` and `wilting` being in sprximity.
             "text": "A flower wilting under the sun, unlike a sunflower"
@@ -299,7 +299,7 @@ fn test_proximity_split_word() {
     let SearchResult { documents_ids, .. } = s.execute().unwrap();
     insta::assert_snapshot!(format!("{documents_ids:?}"), @"[2, 4, 5, 1, 3]");
     let texts = collect_field_values(&index, &txn, "text", &documents_ids);
-    // "2" and "4" should be swapped ideally
+    // TODO: "2" and "4" should be swapped ideally
     insta::assert_debug_snapshot!(texts, @r###"
     [
         "\"Sun Flower sounds like the title of a painting, maybe about a flower wilting under the heat.\"",
@@ -316,7 +316,7 @@ fn test_proximity_split_word() {
     let SearchResult { documents_ids, .. } = s.execute().unwrap();
     insta::assert_snapshot!(format!("{documents_ids:?}"), @"[2, 4, 1]");
     let texts = collect_field_values(&index, &txn, "text", &documents_ids);
-    // "2" and "4" should be swapped ideally
+    // TODO: "2" and "4" should be swapped ideally
     insta::assert_debug_snapshot!(texts, @r###"
     [
         "\"Sun Flower sounds like the title of a painting, maybe about a flower wilting under the heat.\"",
@@ -341,7 +341,7 @@ fn test_proximity_split_word() {
     let SearchResult { documents_ids, .. } = s.execute().unwrap();
     insta::assert_snapshot!(format!("{documents_ids:?}"), @"[2, 4, 1]");
     let texts = collect_field_values(&index, &txn, "text", &documents_ids);
-    // "2" and "4" should be swapped ideally
+    // TODO: "2" and "4" should be swapped ideally
     insta::assert_debug_snapshot!(texts, @r###"
     [
         "\"Sun Flower sounds like the title of a painting, maybe about a flower wilting under the heat.\"",

milli/src/search/new/tests/proximity_typo.rs

@@ -2,8 +2,9 @@
 This module tests the interactions between the proximity and typo ranking rules.
 
 The proximity ranking rule should transform the query graph such that it
-only contains the word pairs that it used to compute its bucket, but this is not currently
-implemented.
+only contains the word pairs that it used to compute its bucket.
+
+TODO: This is not currently implemented.
 */
 
 use crate::index::tests::TempIndex;
@@ -63,7 +64,7 @@ fn test_trap_basic() {
     let SearchResult { documents_ids, .. } = s.execute().unwrap();
     insta::assert_snapshot!(format!("{documents_ids:?}"), @"[0, 1]");
     let texts = collect_field_values(&index, &txn, "text", &documents_ids);
-    // This is incorrect, 1 should come before 0
+    // TODO: this is incorrect, 1 should come before 0
     insta::assert_debug_snapshot!(texts, @r###"
     [
         "\"summer. holiday. sommer holidty\"",

milli/src/search/new/tests/typo.rs

@@ -571,8 +571,8 @@ fn test_typo_synonyms() {
     s.terms_matching_strategy(TermsMatchingStrategy::All);
     s.query("the fast brownish fox jumps over the lackadaisical dog");
 
-    // The interaction of ngrams + synonyms means that the multi-word synonyms end up having a typo cost.
-    // This is probably not what we want.
+    // TODO: is this correct? interaction of ngrams + synonyms means that the
+    // multi-word synonyms end up having a typo cost. This is probably not what we want.
     let SearchResult { documents_ids, .. } = s.execute().unwrap();
     insta::assert_snapshot!(format!("{documents_ids:?}"), @"[21, 0, 22]");
     let texts = collect_field_values(&index, &txn, "text", &documents_ids);

milli/src/update/clear_documents.rs

@@ -39,6 +39,7 @@ impl<'t, 'u, 'i> ClearDocuments<'t, 'u, 'i> {
             facet_id_is_empty_docids,
             field_id_docid_facet_f64s,
             field_id_docid_facet_strings,
+            vector_id_docid,
             documents,
         } = self.index;
 
@@ -57,6 +58,7 @@ impl<'t, 'u, 'i> ClearDocuments<'t, 'u, 'i> {
         self.index.put_field_distribution(self.wtxn, &FieldDistribution::default())?;
         self.index.delete_geo_rtree(self.wtxn)?;
         self.index.delete_geo_faceted_documents_ids(self.wtxn)?;
+        self.index.delete_vector_hnsw(self.wtxn)?;
 
         // We clean all the faceted documents ids.
         for field_id in faceted_fields {
@@ -95,6 +97,7 @@ impl<'t, 'u, 'i> ClearDocuments<'t, 'u, 'i> {
         facet_id_string_docids.clear(self.wtxn)?;
         field_id_docid_facet_f64s.clear(self.wtxn)?;
         field_id_docid_facet_strings.clear(self.wtxn)?;
+        vector_id_docid.clear(self.wtxn)?;
         documents.clear(self.wtxn)?;
 
         Ok(number_of_documents)

milli/src/update/delete_documents.rs

@@ -4,8 +4,10 @@ use std::collections::{BTreeSet, HashMap, HashSet};
 use fst::IntoStreamer;
 use heed::types::{ByteSlice, DecodeIgnore, Str, UnalignedSlice};
 use heed::{BytesDecode, BytesEncode, Database, RwIter};
+use hnsw::Searcher;
 use roaring::RoaringBitmap;
 use serde::{Deserialize, Serialize};
+use space::KnnPoints;
 use time::OffsetDateTime;
 
 use super::facet::delete::FacetsDelete;
@@ -14,6 +16,7 @@ use crate::error::InternalError;
 use crate::facet::FacetType;
 use crate::heed_codec::facet::FieldDocIdFacetCodec;
 use crate::heed_codec::CboRoaringBitmapCodec;
+use crate::index::Hnsw;
 use crate::{
     ExternalDocumentsIds, FieldId, FieldIdMapMissingEntry, Index, Result, RoaringBitmapCodec, BEU32,
 };
@@ -71,6 +74,7 @@ impl std::fmt::Display for DeletionStrategy {
 pub(crate) struct DetailedDocumentDeletionResult {
     pub deleted_documents: u64,
     pub remaining_documents: u64,
+    pub soft_deletion_used: bool,
 }
 
 impl<'t, 'u, 'i> DeleteDocuments<'t, 'u, 'i> {
@@ -107,8 +111,11 @@ impl<'t, 'u, 'i> DeleteDocuments<'t, 'u, 'i> {
         Some(docid)
     }
     pub fn execute(self) -> Result<DocumentDeletionResult> {
-        let DetailedDocumentDeletionResult { deleted_documents, remaining_documents } =
-            self.execute_inner()?;
+        let DetailedDocumentDeletionResult {
+            deleted_documents,
+            remaining_documents,
+            soft_deletion_used: _,
+        } = self.execute_inner()?;
 
         Ok(DocumentDeletionResult { deleted_documents, remaining_documents })
     }
@@ -129,6 +136,7 @@ impl<'t, 'u, 'i> DeleteDocuments<'t, 'u, 'i> {
             return Ok(DetailedDocumentDeletionResult {
                 deleted_documents: 0,
                 remaining_documents: 0,
+                soft_deletion_used: false,
             });
         }
 
@@ -144,6 +152,7 @@ impl<'t, 'u, 'i> DeleteDocuments<'t, 'u, 'i> {
             return Ok(DetailedDocumentDeletionResult {
                 deleted_documents: current_documents_ids_len,
                 remaining_documents,
+                soft_deletion_used: false,
             });
         }
 
@@ -212,6 +221,7 @@ impl<'t, 'u, 'i> DeleteDocuments<'t, 'u, 'i> {
             return Ok(DetailedDocumentDeletionResult {
                 deleted_documents: self.to_delete_docids.len(),
                 remaining_documents: documents_ids.len(),
+                soft_deletion_used: true,
             });
         }
 
@@ -240,6 +250,7 @@ impl<'t, 'u, 'i> DeleteDocuments<'t, 'u, 'i> {
             facet_id_exists_docids,
             facet_id_is_null_docids,
             facet_id_is_empty_docids,
+            vector_id_docid,
             documents,
         } = self.index;
         // Remove from the documents database
@@ -429,11 +440,36 @@ impl<'t, 'u, 'i> DeleteDocuments<'t, 'u, 'i> {
             &self.to_delete_docids,
         )?;
 
+        // An ugly and slow way to remove the vectors from the HNSW
+        // It basically reconstructs the HNSW from scratch without editing the current one.
+        let current_hnsw = self.index.vector_hnsw(self.wtxn)?.unwrap_or_default();
+        if !current_hnsw.is_empty() {
+            let mut new_hnsw = Hnsw::default();
+            let mut searcher = Searcher::new();
+            let mut new_vector_id_docids = Vec::new();
+
+            for result in vector_id_docid.iter(self.wtxn)? {
+                let (vector_id, docid) = result?;
+                if !self.to_delete_docids.contains(docid.get()) {
+                    let vector = current_hnsw.get_point(vector_id.get() as usize).clone();
+                    let vector_id = new_hnsw.insert(vector, &mut searcher);
+                    new_vector_id_docids.push((vector_id as u32, docid));
+                }
+            }
+
+            vector_id_docid.clear(self.wtxn)?;
+            for (vector_id, docid) in new_vector_id_docids {
+                vector_id_docid.put(self.wtxn, &BEU32::new(vector_id), &docid)?;
+            }
+            self.index.put_vector_hnsw(self.wtxn, &new_hnsw)?;
+        }
+
         self.index.put_soft_deleted_documents_ids(self.wtxn, &RoaringBitmap::new())?;
 
         Ok(DetailedDocumentDeletionResult {
             deleted_documents: self.to_delete_docids.len(),
             remaining_documents: documents_ids.len(),
+            soft_deletion_used: false,
         })
     }
 

milli/src/update/index_documents/extract/extract_vector_points.rs

@@ -0,0 +1,40 @@
+use std::fs::File;
+use std::io;
+
+use bytemuck::cast_slice;
+use serde_json::from_slice;
+
+use super::helpers::{create_writer, writer_into_reader, GrenadParameters};
+use crate::{FieldId, InternalError, Result};
+
+/// Extracts the embedding vector contained in each document under the `_vector` field.
+///
+/// Returns the generated grenad reader containing the docid as key associated to the Vec<f32>
+#[logging_timer::time]
+pub fn extract_vector_points<R: io::Read + io::Seek>(
+    obkv_documents: grenad::Reader<R>,
+    indexer: GrenadParameters,
+    vector_fid: FieldId,
+) -> Result<grenad::Reader<File>> {
+    let mut writer = create_writer(
+        indexer.chunk_compression_type,
+        indexer.chunk_compression_level,
+        tempfile::tempfile()?,
+    );
+
+    let mut cursor = obkv_documents.into_cursor()?;
+    while let Some((docid_bytes, value)) = cursor.move_on_next()? {
+        let obkv = obkv::KvReader::new(value);
+
+        // first we get the _vector field
+        if let Some(vector) = obkv.get(vector_fid) {
+            // try to extract the vector
+            let vector: Vec<f32> = from_slice(vector).map_err(InternalError::SerdeJson).unwrap();
+            let bytes = cast_slice(&vector);
+            writer.insert(docid_bytes, bytes)?;
+        }
+        // else => the _vector object was `null`, there is nothing to do
+    }
+
+    writer_into_reader(writer)
+}

milli/src/update/index_documents/extract/mod.rs

@@ -4,6 +4,7 @@ mod extract_facet_string_docids;
 mod extract_fid_docid_facet_values;
 mod extract_fid_word_count_docids;
 mod extract_geo_points;
+mod extract_vector_points;
 mod extract_word_docids;
 mod extract_word_fid_docids;
 mod extract_word_pair_proximity_docids;
@@ -22,6 +23,7 @@ use self::extract_facet_string_docids::extract_facet_string_docids;
 use self::extract_fid_docid_facet_values::{extract_fid_docid_facet_values, ExtractedFacetValues};
 use self::extract_fid_word_count_docids::extract_fid_word_count_docids;
 use self::extract_geo_points::extract_geo_points;
+use self::extract_vector_points::extract_vector_points;
 use self::extract_word_docids::extract_word_docids;
 use self::extract_word_fid_docids::extract_word_fid_docids;
 use self::extract_word_pair_proximity_docids::extract_word_pair_proximity_docids;
@@ -45,6 +47,7 @@ pub(crate) fn data_from_obkv_documents(
     faceted_fields: HashSet<FieldId>,
     primary_key_id: FieldId,
     geo_fields_ids: Option<(FieldId, FieldId)>,
+    vector_field_id: Option<FieldId>,
     stop_words: Option<fst::Set<&[u8]>>,
     max_positions_per_attributes: Option<u32>,
     exact_attributes: HashSet<FieldId>,
@@ -69,6 +72,7 @@ pub(crate) fn data_from_obkv_documents(
                     &faceted_fields,
                     primary_key_id,
                     geo_fields_ids,
+                    vector_field_id,
                     &stop_words,
                     max_positions_per_attributes,
                 )
@@ -279,6 +283,7 @@ fn send_and_extract_flattened_documents_data(
     faceted_fields: &HashSet<FieldId>,
     primary_key_id: FieldId,
     geo_fields_ids: Option<(FieldId, FieldId)>,
+    vector_field_id: Option<FieldId>,
     stop_words: &Option<fst::Set<&[u8]>>,
     max_positions_per_attributes: Option<u32>,
 ) -> Result<(
@@ -307,6 +312,20 @@ fn send_and_extract_flattened_documents_data(
         });
     }
 
+    if let Some(vector_field_id) = vector_field_id {
+        let documents_chunk_cloned = flattened_documents_chunk.clone();
+        let lmdb_writer_sx_cloned = lmdb_writer_sx.clone();
+        rayon::spawn(move || {
+            let result = extract_vector_points(documents_chunk_cloned, indexer, vector_field_id);
+            let _ = match result {
+                Ok(vector_points) => {
+                    lmdb_writer_sx_cloned.send(Ok(TypedChunk::VectorPoints(vector_points)))
+                }
+                Err(error) => lmdb_writer_sx_cloned.send(Err(error)),
+            };
+        });
+    }
+
     let (docid_word_positions_chunk, docid_fid_facet_values_chunks): (Result<_>, Result<_>) =
         rayon::join(
             || {

milli/src/update/index_documents/helpers/clonable_mmap.rs

@@ -2,7 +2,7 @@ use std::sync::Arc;
 
 use memmap2::Mmap;
 
-/// Wrapper around Mmap allowing to virtually clone grenad-chunks
+/// Wrapper around Mmap allowing to virtualy clone grenad-chunks
 /// in a parallel process like the indexing.
 #[derive(Debug, Clone)]
 pub struct ClonableMmap {

milli/src/update/index_documents/mod.rs

@@ -236,7 +236,7 @@ where
             primary_key,
             fields_ids_map,
             field_distribution,
-            new_external_documents_ids,
+            mut external_documents_ids,
             new_documents_ids,
             replaced_documents_ids,
             documents_count,
@@ -304,6 +304,8 @@ where
             }
             None => None,
         };
+        // get the fid of the `_vector` field.
+        let vector_field_id = self.index.fields_ids_map(self.wtxn)?.id("_vector");
 
         let stop_words = self.index.stop_words(self.wtxn)?;
         let exact_attributes = self.index.exact_attributes_ids(self.wtxn)?;
@@ -340,6 +342,7 @@ where
                     faceted_fields,
                     primary_key_id,
                     geo_fields_ids,
+                    vector_field_id,
                     stop_words,
                     max_positions_per_attributes,
                     exact_attributes,
@@ -363,6 +366,9 @@ where
             deletion_builder.delete_documents(&replaced_documents_ids);
             let deleted_documents_result = deletion_builder.execute_inner()?;
             debug!("{} documents actually deleted", deleted_documents_result.deleted_documents);
+            if !deleted_documents_result.soft_deletion_used {
+                external_documents_ids.delete_soft_deleted_documents_ids_from_fsts()?;
+            }
         }
 
         let index_documents_ids = self.index.documents_ids(self.wtxn)?;
@@ -442,9 +448,6 @@ where
         self.index.put_primary_key(self.wtxn, &primary_key)?;
 
         // We write the external documents ids into the main database.
-        let mut external_documents_ids = self.index.external_documents_ids(self.wtxn)?;
-        external_documents_ids.insert_ids(&new_external_documents_ids)?;
-        let external_documents_ids = external_documents_ids.into_static();
         self.index.put_external_documents_ids(self.wtxn, &external_documents_ids)?;
 
         let all_documents_ids = index_documents_ids | new_documents_ids;
@@ -2514,170 +2517,4 @@ mod tests {
         db_snap!(index, word_fid_docids, 3, @"4c2e2a1832e5802796edc1638136d933");
         db_snap!(index, word_position_docids, 3, @"74f556b91d161d997a89468b4da1cb8f");
     }
-
-    #[test]
-    fn reproduce_the_bug() {
-        /*
-            [milli/examples/fuzz.rs:69] &batches = [
-            Batch(
-                [
-                    AddDoc(
-                        { "id": 1, "doggo": "bernese" }, => internal 0
-                    ),
-                ],
-            ),
-            Batch(
-                [
-                    DeleteDoc(
-                        1, => delete internal 0
-                    ),
-                    AddDoc(
-                        { "id": 0, "catto": "jorts" }, => internal 1
-                    ),
-                ],
-            ),
-            Batch(
-                [
-                    AddDoc(
-                        { "id": 1, "catto": "jorts" }, => internal 2
-                    ),
-                ],
-            ),
-        ]
-        */
-        let mut index = TempIndex::new();
-        index.index_documents_config.deletion_strategy = DeletionStrategy::AlwaysHard;
-
-        // START OF BATCH
-
-        println!("--- ENTERING BATCH 1");
-
-        let mut wtxn = index.write_txn().unwrap();
-
-        let builder = IndexDocuments::new(
-            &mut wtxn,
-            &index,
-            &index.indexer_config,
-            index.index_documents_config.clone(),
-            |_| (),
-            || false,
-        )
-        .unwrap();
-
-        // OP
-
-        let documents = documents!([
-            { "id": 1, "doggo": "bernese" },
-        ]);
-        let (builder, added) = builder.add_documents(documents).unwrap();
-        insta::assert_display_snapshot!(added.unwrap(), @"1");
-
-        // FINISHING
-        let addition = builder.execute().unwrap();
-        insta::assert_debug_snapshot!(addition, @r###"
-        DocumentAdditionResult {
-            indexed_documents: 1,
-            number_of_documents: 1,
-        }
-        "###);
-        wtxn.commit().unwrap();
-
-        db_snap!(index, documents, @r###"
-        {"id":1,"doggo":"bernese"}
-        "###);
-        db_snap!(index, external_documents_ids, @r###"
-        soft:
-        hard:
-        1                        0
-        "###);
-
-        // A first batch of documents has been inserted
-
-        // BATCH 2
-
-        println!("--- ENTERING BATCH 2");
-
-        let mut wtxn = index.write_txn().unwrap();
-
-        let builder = IndexDocuments::new(
-            &mut wtxn,
-            &index,
-            &index.indexer_config,
-            index.index_documents_config.clone(),
-            |_| (),
-            || false,
-        )
-        .unwrap();
-
-        let (builder, removed) = builder.remove_documents(vec![S("1")]).unwrap();
-        insta::assert_display_snapshot!(removed.unwrap(), @"1");
-
-        let documents = documents!([
-            { "id": 0, "catto": "jorts" },
-        ]);
-        let (builder, added) = builder.add_documents(documents).unwrap();
-        insta::assert_display_snapshot!(added.unwrap(), @"1");
-
-        let addition = builder.execute().unwrap();
-        insta::assert_debug_snapshot!(addition, @r###"
-        DocumentAdditionResult {
-            indexed_documents: 1,
-            number_of_documents: 1,
-        }
-        "###);
-        wtxn.commit().unwrap();
-
-        db_snap!(index, documents, @r###"
-        {"id":0,"catto":"jorts"}
-        "###);
-
-        db_snap!(index, external_documents_ids, @r###"
-        soft:
-        hard:
-        0                        1
-        "###);
-
-        db_snap!(index, soft_deleted_documents_ids, @"[]");
-
-        // BATCH 3
-
-        println!("--- ENTERING BATCH 3");
-
-        let mut wtxn = index.write_txn().unwrap();
-
-        let builder = IndexDocuments::new(
-            &mut wtxn,
-            &index,
-            &index.indexer_config,
-            index.index_documents_config.clone(),
-            |_| (),
-            || false,
-        )
-        .unwrap();
-
-        let documents = documents!([
-            { "id": 1, "catto": "jorts" },
-        ]);
-        let (builder, added) = builder.add_documents(documents).unwrap();
-        insta::assert_display_snapshot!(added.unwrap(), @"1");
-
-        let addition = builder.execute().unwrap();
-        insta::assert_debug_snapshot!(addition, @r###"
-        DocumentAdditionResult {
-            indexed_documents: 1,
-            number_of_documents: 2,
-        }
-        "###);
-        wtxn.commit().unwrap();
-
-        db_snap!(index, documents, @r###"
-        {"id":1,"catto":"jorts"}
-        {"id":0,"catto":"jorts"}
-        "###);
-
-        // Ensuring all the returned IDs actually exists
-        let rtxn = index.read_txn().unwrap();
-        let res = index.search(&rtxn).execute().unwrap();
-        index.documents(&rtxn, res.documents_ids).unwrap();
-    }
 }

milli/src/update/index_documents/transform.rs

@@ -21,14 +21,15 @@ use crate::error::{Error, InternalError, UserError};
 use crate::index::{db_name, main_key};
 use crate::update::{AvailableDocumentsIds, ClearDocuments, UpdateIndexingStep};
 use crate::{
-    FieldDistribution, FieldId, FieldIdMapMissingEntry, FieldsIdsMap, Index, Result, BEU32,
+    ExternalDocumentsIds, FieldDistribution, FieldId, FieldIdMapMissingEntry, FieldsIdsMap, Index,
+    Result, BEU32,
 };
 
 pub struct TransformOutput {
     pub primary_key: String,
     pub fields_ids_map: FieldsIdsMap,
     pub field_distribution: FieldDistribution,
-    pub new_external_documents_ids: fst::Map<Cow<'static, [u8]>>,
+    pub external_documents_ids: ExternalDocumentsIds<'static>,
     pub new_documents_ids: RoaringBitmap,
     pub replaced_documents_ids: RoaringBitmap,
     pub documents_count: usize,
@@ -567,6 +568,8 @@ impl<'a, 'i> Transform<'a, 'i> {
             }))?
             .to_string();
 
+        let mut external_documents_ids = self.index.external_documents_ids(wtxn)?;
+
         // We create a final writer to write the new documents in order from the sorter.
         let mut writer = create_writer(
             self.indexer_settings.chunk_compression_type,
@@ -648,12 +651,13 @@ impl<'a, 'i> Transform<'a, 'i> {
             fst_new_external_documents_ids_builder.insert(key, value)
         })?;
         let new_external_documents_ids = fst_new_external_documents_ids_builder.into_map();
+        external_documents_ids.insert_ids(&new_external_documents_ids)?;
 
         Ok(TransformOutput {
             primary_key,
             fields_ids_map: self.fields_ids_map,
             field_distribution,
-            new_external_documents_ids: new_external_documents_ids.map_data(Cow::Owned).unwrap(),
+            external_documents_ids: external_documents_ids.into_static(),
             new_documents_ids: self.new_documents_ids,
             replaced_documents_ids: self.replaced_documents_ids,
             documents_count: self.documents_count,
@@ -687,8 +691,7 @@ impl<'a, 'i> Transform<'a, 'i> {
         let new_external_documents_ids = {
             let mut external_documents_ids = self.index.external_documents_ids(wtxn)?;
             external_documents_ids.delete_soft_deleted_documents_ids_from_fsts()?;
-            // This call should be free and can't fail since the previous method merged both fsts.
-            external_documents_ids.into_static().to_fst()?.into_owned()
+            external_documents_ids
         };
 
         let documents_ids = self.index.documents_ids(wtxn)?;
@@ -773,7 +776,7 @@ impl<'a, 'i> Transform<'a, 'i> {
             primary_key,
             fields_ids_map: new_fields_ids_map,
             field_distribution,
-            new_external_documents_ids,
+            external_documents_ids: new_external_documents_ids.into_static(),
             new_documents_ids: documents_ids,
             replaced_documents_ids: RoaringBitmap::default(),
             documents_count,

milli/src/update/index_documents/typed_chunk.rs

@@ -4,20 +4,24 @@ use std::convert::TryInto;
 use std::fs::File;
 use std::io;
 
+use bytemuck::allocation::pod_collect_to_vec;
 use charabia::{Language, Script};
 use grenad::MergerBuilder;
 use heed::types::ByteSlice;
 use heed::RwTxn;
+use hnsw::Searcher;
 use roaring::RoaringBitmap;
+use space::KnnPoints;
 
 use super::helpers::{
     self, merge_ignore_values, serialize_roaring_bitmap, valid_lmdb_key, CursorClonableMmap,
 };
 use super::{ClonableMmap, MergeFn};
+use crate::error::UserError;
 use crate::facet::FacetType;
 use crate::update::facet::FacetsUpdate;
 use crate::update::index_documents::helpers::as_cloneable_grenad;
-use crate::{lat_lng_to_xyz, CboRoaringBitmapCodec, DocumentId, GeoPoint, Index, Result};
+use crate::{lat_lng_to_xyz, CboRoaringBitmapCodec, DocumentId, GeoPoint, Index, Result, BEU32};
 
 pub(crate) enum TypedChunk {
     FieldIdDocidFacetStrings(grenad::Reader<CursorClonableMmap>),
@@ -38,6 +42,7 @@ pub(crate) enum TypedChunk {
     FieldIdFacetIsNullDocids(grenad::Reader<File>),
     FieldIdFacetIsEmptyDocids(grenad::Reader<File>),
     GeoPoints(grenad::Reader<File>),
+    VectorPoints(grenad::Reader<File>),
     ScriptLanguageDocids(HashMap<(Script, Language), RoaringBitmap>),
 }
 
@@ -221,6 +226,38 @@ pub(crate) fn write_typed_chunk_into_index(
             index.put_geo_rtree(wtxn, &rtree)?;
             index.put_geo_faceted_documents_ids(wtxn, &geo_faceted_docids)?;
         }
+        TypedChunk::VectorPoints(vector_points) => {
+            let mut hnsw = index.vector_hnsw(wtxn)?.unwrap_or_default();
+            let mut searcher = Searcher::new();
+
+            let mut expected_dimensions = match index.vector_id_docid.iter(wtxn)?.next() {
+                Some(result) => {
+                    let (vector_id, _) = result?;
+                    Some(hnsw.get_point(vector_id.get() as usize).len())
+                }
+                None => None,
+            };
+
+            let mut cursor = vector_points.into_cursor()?;
+            while let Some((key, value)) = cursor.move_on_next()? {
+                // convert the key back to a u32 (4 bytes)
+                let docid = key.try_into().map(DocumentId::from_be_bytes).unwrap();
+                // convert the vector back to a Vec<f32>
+                let vector: Vec<f32> = pod_collect_to_vec(value);
+
+                // TODO Move this error in the vector extractor
+                let found = vector.len();
+                let expected = *expected_dimensions.get_or_insert(found);
+                if expected != found {
+                    return Err(UserError::InvalidVectorDimensions { expected, found })?;
+                }
+
+                let vector_id = hnsw.insert(vector, &mut searcher) as u32;
+                index.vector_id_docid.put(wtxn, &BEU32::new(vector_id), &BEU32::new(docid))?;
+            }
+            log::debug!("There are {} entries in the HNSW so far", hnsw.len());
+            index.put_vector_hnsw(wtxn, &hnsw)?;
+        }
         TypedChunk::ScriptLanguageDocids(hash_pair) => {
             let mut buffer = Vec::new();
             for (key, value) in hash_pair {