Small test fixes

4239: Remove the actix-web dependency from milli r=dureuill a=Kerollmops

Just remove actix-web from milli.

Co-authored-by: Clément Renault <clement@meilisearch.com>

.github/workflows/benchmarks-pr.yml

@@ -90,7 +90,8 @@ jobs:
           set -x
           export base_ref=$(git merge-base origin/main ${{ steps.comment-branch.outputs.head_ref }} | head -c8)
           export base_filename=$(echo ${{ steps.command.outputs.command-arguments }}_main_${base_ref}.json)
-          echo 'Here are your benchmarks diff 👊' >> body.txt
+          export bench_name=$(echo ${{ steps.command.outputs.command-arguments }})
+          echo "Here are your $bench_name benchmarks diff 👊" >> body.txt
           echo '```' >> body.txt
           ./benchmarks/scripts/compare.sh $base_filename ${{ steps.file.outputs.basename }}.json >> body.txt
           echo '```' >> body.txt

.github/workflows/publish-apt-brew-pkg.yml

@@ -50,7 +50,7 @@ jobs:
     needs: check-version
     steps:
       - name: Create PR to Homebrew
-        uses: mislav/bump-homebrew-formula-action@v2
+        uses: mislav/bump-homebrew-formula-action@v3
         with:
           formula-name: meilisearch
           formula-path: Formula/m/meilisearch.rb

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

@@ -63,7 +63,7 @@ jobs:
         uses: docker/setup-buildx-action@v3
 
       - name: Login to Docker Hub
-        uses: docker/login-action@v2
+        uses: docker/login-action@v3
         with:
           username: ${{ secrets.DOCKERHUB_USERNAME }}
           password: ${{ secrets.DOCKERHUB_TOKEN }}

.github/workflows/sdks-tests.yml

@@ -160,7 +160,7 @@ jobs:
         with:
           repository: meilisearch/meilisearch-js
       - name: Setup node
-        uses: actions/setup-node@v3
+        uses: actions/setup-node@v4
         with:
           cache: 'yarn'
       - name: Install dependencies
@@ -318,7 +318,7 @@ jobs:
         with:
           repository: meilisearch/meilisearch-js-plugins
       - name: Setup node
-        uses: actions/setup-node@v3
+        uses: actions/setup-node@v4
         with:
           cache: yarn
       - name: Install dependencies

.github/workflows/test-suite.yml

@@ -43,7 +43,7 @@ jobs:
           toolchain: nightly
           override: true
       - name: Cache dependencies
-        uses: Swatinem/rust-cache@v2.6.2
+        uses: Swatinem/rust-cache@v2.7.1
       - name: Run cargo check without any default features
         uses: actions-rs/cargo@v1
         with:
@@ -65,7 +65,7 @@ jobs:
     steps:
       - uses: actions/checkout@v3
       - name: Cache dependencies
-        uses: Swatinem/rust-cache@v2.6.2
+        uses: Swatinem/rust-cache@v2.7.1
       - name: Run cargo check without any default features
         uses: actions-rs/cargo@v1
         with:
@@ -149,7 +149,7 @@ jobs:
           toolchain: stable
           override: true
       - name: Cache dependencies
-        uses: Swatinem/rust-cache@v2.6.2
+        uses: Swatinem/rust-cache@v2.7.1
       - name: Run tests in debug
         uses: actions-rs/cargo@v1
         with:
@@ -168,7 +168,7 @@ jobs:
           override: true
           components: clippy
       - name: Cache dependencies
-        uses: Swatinem/rust-cache@v2.6.2
+        uses: Swatinem/rust-cache@v2.7.1
       - name: Run cargo clippy
         uses: actions-rs/cargo@v1
         with:
@@ -187,7 +187,7 @@ jobs:
           override: true
           components: rustfmt
       - name: Cache dependencies
-        uses: Swatinem/rust-cache@v2.6.2
+        uses: Swatinem/rust-cache@v2.7.1
       - name: Run cargo fmt
         # Since we never ran the `build.rs` script in the benchmark directory we are missing one auto-generated import file.
         # Since we want to trigger (and fail) this action as fast as possible, instead of building the benchmark crate

Cargo.toml

@@ -2,6 +2,7 @@
 resolver = "2"
 members = [
     "meilisearch",
+    "meilitool",
     "meilisearch-types",
     "meilisearch-auth",
     "meili-snap",
@@ -18,7 +19,7 @@ members = [
 ]
 
 [workspace.package]
-version = "1.4.1"
+version = "1.5.0"
 authors = ["Quentin de Quelen <quentin@dequelen.me>", "Clément Renault <clement@meilisearch.com>"]
 description = "Meilisearch HTTP server"
 homepage = "https://meilisearch.com"

Dockerfile

@@ -3,7 +3,7 @@ FROM    rust:alpine3.16 AS compiler
 
 RUN     apk add -q --update-cache --no-cache build-base openssl-dev
 
-WORKDIR /meilisearch
+WORKDIR /
 
 ARG     COMMIT_SHA
 ARG     COMMIT_DATE
@@ -17,7 +17,7 @@ RUN     set -eux; \
         if [ "$apkArch" = "aarch64" ]; then \
             export JEMALLOC_SYS_WITH_LG_PAGE=16; \
         fi && \
-        cargo build --release
+        cargo build --release -p meilisearch -p meilitool
 
 # Run
 FROM    alpine:3.16
@@ -28,9 +28,10 @@ ENV     MEILI_SERVER_PROVIDER docker
 RUN     apk update --quiet \
         && apk add -q --no-cache libgcc tini curl
 
-# add meilisearch to the `/bin` so you can run it from anywhere and it's easy
-# to find.
-COPY    --from=compiler /meilisearch/target/release/meilisearch /bin/meilisearch
+# add meilisearch and meilitool to the `/bin` so you can run it from anywhere
+# and it's easy to find.
+COPY    --from=compiler /target/release/meilisearch /bin/meilisearch
+COPY    --from=compiler /target/release/meilitool /bin/meilitool
 # To stay compatible with the older version of the container (pre v0.27.0) we're
 # going to symlink the meilisearch binary in the path to `/meilisearch`
 RUN     ln -s /bin/meilisearch /meilisearch

README.md

@@ -25,12 +25,6 @@
 
 <p align="center">⚡ A lightning-fast search engine that fits effortlessly into your apps, websites, and workflow 🔍</p>
 
----
-
-### 🔥 On November 2nd, we are hosting our first-ever live demo and product updates for [Meilisearch Cloud](https://www.meilisearch.com/cloud?utm_campaign=oss&utm_source=github&utm_medium=meilisearch). Make sure to [register here](https://us06web.zoom.us/meeting/register/tZMlc-mqrjIsH912-HTRe-AaT-pp41bDe81a#/registration) and bring your questions for live Q&A!
-
----
-
 Meilisearch helps you shape a delightful search experience in a snap, offering features that work out-of-the-box to speed up your workflow.
 
 <p align="center" name="demo">

benchmarks/benches/indexing.rs

@@ -36,7 +36,7 @@ fn setup_index() -> Index {
 }
 
 fn setup_settings<'t>(
-    wtxn: &mut RwTxn<'t, '_>,
+    wtxn: &mut RwTxn<'t>,
     index: &'t Index,
     primary_key: &str,
     searchable_fields: &[&str],
@@ -864,22 +864,12 @@ fn delete_documents_from_ids(index: Index, document_ids_to_delete: Vec<RoaringBi
 
     let indexer_config = IndexerConfig::default();
     for ids in document_ids_to_delete {
-        let external_documents_ids = index.external_documents_ids();
-        // FIXME: for filters matching a lot of documents, this will allocate a huge vec of external docids (strings).
-        // Since what we have is an iterator, it would be better to delete in chunks
-        let external_to_internal: std::result::Result<Vec<_>, RoaringBitmap> =
-            external_documents_ids
-                .find_external_id_of(&wtxn, ids)
-                .unwrap()
-                .only_external_ids()
-                .collect();
-        let ids = external_to_internal.unwrap();
         let config = IndexDocumentsConfig::default();
 
         let mut builder =
             IndexDocuments::new(&mut wtxn, &index, &indexer_config, config, |_| (), || false)
                 .unwrap();
-        (builder, _) = builder.remove_documents(ids).unwrap();
+        (builder, _) = builder.remove_documents_from_db_no_batch(&ids).unwrap();
         builder.execute().unwrap();
     }
 

dump/src/lib.rs

@@ -267,6 +267,7 @@ pub(crate) mod test {
             dictionary: Setting::NotSet,
             synonyms: Setting::NotSet,
             distinct_attribute: Setting::NotSet,
+            proximity_precision: Setting::NotSet,
             typo_tolerance: Setting::NotSet,
             faceting: Setting::Set(FacetingSettings {
                 max_values_per_facet: Setting::Set(111),

dump/src/reader/compat/v5_to_v6.rs

@@ -345,6 +345,7 @@ impl<T> From<v5::Settings<T>> for v6::Settings<v6::Unchecked> {
             dictionary: v6::Setting::NotSet,
             synonyms: settings.synonyms.into(),
             distinct_attribute: settings.distinct_attribute.into(),
+            proximity_precision: v6::Setting::NotSet,
             typo_tolerance: match settings.typo_tolerance {
                 v5::Setting::Set(typo) => v6::Setting::Set(v6::TypoTolerance {
                     enabled: typo.enabled.into(),

dump/src/reader/mod.rs

@@ -13,12 +13,12 @@ use crate::{Result, Version};
 
 mod compat;
 
-pub(self) mod v1;
-pub(self) mod v2;
-pub(self) mod v3;
-pub(self) mod v4;
-pub(self) mod v5;
-pub(self) mod v6;
+mod v1;
+mod v2;
+mod v3;
+mod v4;
+mod v5;
+mod v6;
 
 pub type Document = serde_json::Map<String, serde_json::Value>;
 pub type UpdateFile = dyn Iterator<Item = Result<Document>>;

dump/src/reader/v1/settings.rs

@@ -56,8 +56,7 @@ pub enum RankingRule {
     Desc(String),
 }
 
-static ASC_DESC_REGEX: Lazy<Regex> =
-    Lazy::new(|| Regex::new(r#"(asc|desc)\(([\w_-]+)\)"#).unwrap());
+static ASC_DESC_REGEX: Lazy<Regex> = Lazy::new(|| Regex::new(r"(asc|desc)\(([\w_-]+)\)").unwrap());
 
 impl FromStr for RankingRule {
     type Err = ();

filter-parser/src/lib.rs

@@ -564,10 +564,10 @@ pub mod tests {
 
     #[test]
     fn parse_escaped() {
-        insta::assert_display_snapshot!(p(r#"title = 'foo\\'"#), @r#"{title} = {foo\}"#);
-        insta::assert_display_snapshot!(p(r#"title = 'foo\\\\'"#), @r#"{title} = {foo\\}"#);
-        insta::assert_display_snapshot!(p(r#"title = 'foo\\\\\\'"#), @r#"{title} = {foo\\\}"#);
-        insta::assert_display_snapshot!(p(r#"title = 'foo\\\\\\\\'"#), @r#"{title} = {foo\\\\}"#);
+        insta::assert_display_snapshot!(p(r"title = 'foo\\'"), @r#"{title} = {foo\}"#);
+        insta::assert_display_snapshot!(p(r"title = 'foo\\\\'"), @r#"{title} = {foo\\}"#);
+        insta::assert_display_snapshot!(p(r"title = 'foo\\\\\\'"), @r#"{title} = {foo\\\}"#);
+        insta::assert_display_snapshot!(p(r"title = 'foo\\\\\\\\'"), @r#"{title} = {foo\\\\}"#);
         // but it also works with other sequencies
         insta::assert_display_snapshot!(p(r#"title = 'foo\x20\n\t\"\'"'"#), @"{title} = {foo \n\t\"\'\"}");
     }

filter-parser/src/value.rs

@@ -270,8 +270,8 @@ pub mod test {
             ("aaaa", "", rtok("", "aaaa"), "aaaa"),
             (r#"aa"aa"#, r#""aa"#, rtok("", "aa"), "aa"),
             (r#"aa\"aa"#, r#""#, rtok("", r#"aa\"aa"#), r#"aa"aa"#),
-            (r#"aa\\\aa"#, r#""#, rtok("", r#"aa\\\aa"#), r#"aa\\\aa"#),
-            (r#"aa\\"\aa"#, r#""\aa"#, rtok("", r#"aa\\"#), r#"aa\\"#),
+            (r"aa\\\aa", r#""#, rtok("", r"aa\\\aa"), r"aa\\\aa"),
+            (r#"aa\\"\aa"#, r#""\aa"#, rtok("", r"aa\\"), r"aa\\"),
             (r#"aa\\\"\aa"#, r#""#, rtok("", r#"aa\\\"\aa"#), r#"aa\\"\aa"#),
             (r#"\"\""#, r#""#, rtok("", r#"\"\""#), r#""""#),
         ];
@@ -301,12 +301,12 @@ pub mod test {
         );
         // simple quote
         assert_eq!(
-            unescape(Span::new_extra(r#"Hello \'World\'"#, ""), '\''),
+            unescape(Span::new_extra(r"Hello \'World\'", ""), '\''),
             r#"Hello 'World'"#.to_string()
         );
         assert_eq!(
-            unescape(Span::new_extra(r#"Hello \\\'World\\\'"#, ""), '\''),
-            r#"Hello \\'World\\'"#.to_string()
+            unescape(Span::new_extra(r"Hello \\\'World\\\'", ""), '\''),
+            r"Hello \\'World\\'".to_string()
         );
     }
 
@@ -335,19 +335,19 @@ pub mod test {
             ("\"cha'nnel\"", "cha'nnel", false),
             ("I'm tamo", "I", false),
             // escaped thing but not quote
-            (r#""\\""#, r#"\"#, true),
-            (r#""\\\\\\""#, r#"\\\"#, true),
-            (r#""aa\\aa""#, r#"aa\aa"#, true),
+            (r#""\\""#, r"\", true),
+            (r#""\\\\\\""#, r"\\\", true),
+            (r#""aa\\aa""#, r"aa\aa", true),
             // with double quote
             (r#""Hello \"world\"""#, r#"Hello "world""#, true),
             (r#""Hello \\\"world\\\"""#, r#"Hello \"world\""#, true),
             (r#""I'm \"super\" tamo""#, r#"I'm "super" tamo"#, true),
             (r#""\"\"""#, r#""""#, true),
             // with simple quote
-            (r#"'Hello \'world\''"#, r#"Hello 'world'"#, true),
-            (r#"'Hello \\\'world\\\''"#, r#"Hello \'world\'"#, true),
+            (r"'Hello \'world\''", r#"Hello 'world'"#, true),
+            (r"'Hello \\\'world\\\''", r"Hello \'world\'", true),
             (r#"'I\'m "super" tamo'"#, r#"I'm "super" tamo"#, true),
-            (r#"'\'\''"#, r#"''"#, true),
+            (r"'\'\''", r#"''"#, true),
         ];
 
         for (input, expected, escaped) in test_case {

fuzzers/src/bin/fuzz-indexing.rs

@@ -113,7 +113,7 @@ fn main() {
                             index.documents(&wtxn, res.documents_ids).unwrap();
                             progression.fetch_add(1, Ordering::Relaxed);
                         }
-                        wtxn.abort().unwrap();
+                        wtxn.abort();
                     });
                     if let err @ Err(_) = handle.join() {
                         stop.store(true, Ordering::Relaxed);

index-scheduler/Cargo.toml

@@ -22,7 +22,7 @@ log = "0.4.17"
 meilisearch-auth = { path = "../meilisearch-auth" }
 meilisearch-types = { path = "../meilisearch-types" }
 page_size = "0.5.0"
-puffin = "0.16.0"
+puffin = { version = "0.16.0", features = ["serialization"] }
 roaring = { version = "0.10.1", features = ["serde"] }
 serde = { version = "1.0.160", features = ["derive"] }
 serde_json = { version = "1.0.95", features = ["preserve_order"] }

index-scheduler/src/batch.rs

@@ -32,7 +32,7 @@ use meilisearch_types::milli::heed::CompactionOption;
 use meilisearch_types::milli::update::{
     IndexDocumentsConfig, IndexDocumentsMethod, IndexerConfig, Settings as MilliSettings,
 };
-use meilisearch_types::milli::{self, Filter, BEU32};
+use meilisearch_types::milli::{self, Filter};
 use meilisearch_types::settings::{apply_settings_to_builder, Settings, Unchecked};
 use meilisearch_types::tasks::{Details, IndexSwap, Kind, KindWithContent, Status, Task};
 use meilisearch_types::{compression, Index, VERSION_FILE_NAME};
@@ -104,12 +104,6 @@ pub(crate) enum IndexOperation {
         operations: Vec<DocumentOperation>,
         tasks: Vec<Task>,
     },
-    DocumentDeletion {
-        index_uid: String,
-        // The vec associated with each document deletion tasks.
-        documents: Vec<Vec<String>>,
-        tasks: Vec<Task>,
-    },
     IndexDocumentDeletionByFilter {
         index_uid: String,
         task: Task,
@@ -161,7 +155,6 @@ impl Batch {
             }
             Batch::IndexOperation { op, .. } => match op {
                 IndexOperation::DocumentOperation { tasks, .. }
-                | IndexOperation::DocumentDeletion { tasks, .. }
                 | IndexOperation::Settings { tasks, .. }
                 | IndexOperation::DocumentClear { tasks, .. } => {
                     tasks.iter().map(|task| task.uid).collect()
@@ -226,7 +219,6 @@ impl IndexOperation {
     pub fn index_uid(&self) -> &str {
         match self {
             IndexOperation::DocumentOperation { index_uid, .. }
-            | IndexOperation::DocumentDeletion { index_uid, .. }
             | IndexOperation::IndexDocumentDeletionByFilter { index_uid, .. }
             | IndexOperation::DocumentClear { index_uid, .. }
             | IndexOperation::Settings { index_uid, .. }
@@ -242,9 +234,6 @@ impl fmt::Display for IndexOperation {
             IndexOperation::DocumentOperation { .. } => {
                 f.write_str("IndexOperation::DocumentOperation")
             }
-            IndexOperation::DocumentDeletion { .. } => {
-                f.write_str("IndexOperation::DocumentDeletion")
-            }
             IndexOperation::IndexDocumentDeletionByFilter { .. } => {
                 f.write_str("IndexOperation::IndexDocumentDeletionByFilter")
             }
@@ -347,18 +336,27 @@ impl IndexScheduler {
             BatchKind::DocumentDeletion { deletion_ids } => {
                 let tasks = self.get_existing_tasks(rtxn, deletion_ids)?;
 
-                let mut documents = Vec::new();
+                let mut operations = Vec::with_capacity(tasks.len());
+                let mut documents_counts = Vec::with_capacity(tasks.len());
                 for task in &tasks {
                     match task.kind {
                         KindWithContent::DocumentDeletion { ref documents_ids, .. } => {
-                            documents.push(documents_ids.clone())
+                            operations.push(DocumentOperation::Delete(documents_ids.clone()));
+                            documents_counts.push(documents_ids.len() as u64);
                         }
                         _ => unreachable!(),
                     }
                 }
 
                 Ok(Some(Batch::IndexOperation {
-                    op: IndexOperation::DocumentDeletion { index_uid, documents, tasks },
+                    op: IndexOperation::DocumentOperation {
+                        index_uid,
+                        primary_key: None,
+                        method: IndexDocumentsMethod::ReplaceDocuments,
+                        documents_counts,
+                        operations,
+                        tasks,
+                    },
                     must_create_index,
                 }))
             }
@@ -717,7 +715,7 @@ impl IndexScheduler {
 
                 // 2. Snapshot the index-scheduler LMDB env
                 //
-                // When we call copy_to_path, LMDB opens a read transaction by itself,
+                // When we call copy_to_file, LMDB opens a read transaction by itself,
                 // we can't provide our own. It is an issue as we would like to know
                 // the update files to copy but new ones can be enqueued between the copy
                 // of the env and the new transaction we open to retrieve the enqueued tasks.
@@ -730,7 +728,7 @@ impl IndexScheduler {
                 // 2.1 First copy the LMDB env of the index-scheduler
                 let dst = temp_snapshot_dir.path().join("tasks");
                 fs::create_dir_all(&dst)?;
-                self.env.copy_to_path(dst.join("data.mdb"), CompactionOption::Enabled)?;
+                self.env.copy_to_file(dst.join("data.mdb"), CompactionOption::Enabled)?;
 
                 // 2.2 Create a read transaction on the index-scheduler
                 let rtxn = self.env.read_txn()?;
@@ -755,7 +753,7 @@ impl IndexScheduler {
                     let index = self.index_mapper.index(&rtxn, name)?;
                     let dst = temp_snapshot_dir.path().join("indexes").join(uuid.to_string());
                     fs::create_dir_all(&dst)?;
-                    index.copy_to_path(dst.join("data.mdb"), CompactionOption::Enabled)?;
+                    index.copy_to_file(dst.join("data.mdb"), CompactionOption::Enabled)?;
                 }
 
                 drop(rtxn);
@@ -768,7 +766,7 @@ impl IndexScheduler {
                     .map_size(1024 * 1024 * 1024) // 1 GiB
                     .max_dbs(2)
                     .open(&self.auth_path)?;
-                auth.copy_to_path(dst.join("data.mdb"), CompactionOption::Enabled)?;
+                auth.copy_to_file(dst.join("data.mdb"), CompactionOption::Enabled)?;
 
                 // 5. Copy and tarball the flat snapshot
                 // 5.1 Find the original name of the database
@@ -824,6 +822,10 @@ impl IndexScheduler {
                 // 2. dump the tasks
                 let mut dump_tasks = dump.create_tasks_queue()?;
                 for ret in self.all_tasks.iter(&rtxn)? {
+                    if self.must_stop_processing.get() {
+                        return Err(Error::AbortedTask);
+                    }
+
                     let (_, mut t) = ret?;
                     let status = t.status;
                     let content_file = t.content_uuid();
@@ -844,6 +846,9 @@ impl IndexScheduler {
 
                     // 2.1. Dump the `content_file` associated with the task if there is one and the task is not finished yet.
                     if let Some(content_file) = content_file {
+                        if self.must_stop_processing.get() {
+                            return Err(Error::AbortedTask);
+                        }
                         if status == Status::Enqueued {
                             let content_file = self.file_store.get_update(content_file)?;
 
@@ -883,6 +888,9 @@ impl IndexScheduler {
 
                     // 3.1. Dump the documents
                     for ret in index.all_documents(&rtxn)? {
+                        if self.must_stop_processing.get() {
+                            return Err(Error::AbortedTask);
+                        }
                         let (_id, doc) = ret?;
                         let document = milli::obkv_to_json(&all_fields, &fields_ids_map, doc)?;
                         index_dumper.push_document(&document)?;
@@ -902,6 +910,9 @@ impl IndexScheduler {
                     "[year repr:full][month repr:numerical][day padding:zero]-[hour padding:zero][minute padding:zero][second padding:zero][subsecond digits:3]"
                 )).unwrap();
 
+                if self.must_stop_processing.get() {
+                    return Err(Error::AbortedTask);
+                }
                 let path = self.dumps_path.join(format!("{}.dump", dump_uid));
                 let file = File::create(path)?;
                 dump.persist_to(BufWriter::new(file))?;
@@ -922,6 +933,10 @@ impl IndexScheduler {
                     self.index_mapper.index(&rtxn, &index_uid)?
                 };
 
+                // the index operation can take a long time, so save this handle to make it available to the search for the duration of the tick
+                *self.currently_updating_index.write().unwrap() =
+                    Some((index_uid.clone(), index.clone()));
+
                 let mut index_wtxn = index.write_txn()?;
                 let tasks = self.apply_index_operation(&mut index_wtxn, &index, op)?;
                 index_wtxn.commit()?;
@@ -1091,7 +1106,7 @@ impl IndexScheduler {
         for task_id in &index_lhs_task_ids | &index_rhs_task_ids {
             let mut task = self.get_task(wtxn, task_id)?.ok_or(Error::CorruptedTaskQueue)?;
             swap_index_uid_in_task(&mut task, (lhs, rhs));
-            self.all_tasks.put(wtxn, &BEU32::new(task_id), &task)?;
+            self.all_tasks.put(wtxn, &task_id, &task)?;
         }
 
         // 4. remove the task from indexuid = before_name
@@ -1117,7 +1132,7 @@ impl IndexScheduler {
     /// The list of processed tasks.
     fn apply_index_operation<'i>(
         &self,
-        index_wtxn: &mut RwTxn<'i, '_>,
+        index_wtxn: &mut RwTxn<'i>,
         index: &'i Index,
         operation: IndexOperation,
     ) -> Result<Vec<Task>> {
@@ -1275,45 +1290,6 @@ impl IndexScheduler {
 
                 Ok(tasks)
             }
-            IndexOperation::DocumentDeletion { index_uid: _, documents, mut tasks } => {
-                let indexer_config = self.index_mapper.indexer_config();
-                let config = IndexDocumentsConfig {
-                    update_method: IndexDocumentsMethod::ReplaceDocuments,
-                    ..Default::default()
-                };
-                let must_stop_processing = self.must_stop_processing.clone();
-
-                let mut builder = milli::update::IndexDocuments::new(
-                    index_wtxn,
-                    index,
-                    indexer_config,
-                    config,
-                    |indexing_step| trace!("update: {:?}", indexing_step),
-                    || must_stop_processing.get(),
-                )?;
-
-                let document_ids = documents.iter().flatten().cloned().collect();
-
-                let (new_builder, user_result) = builder.remove_documents(document_ids)?;
-                builder = new_builder;
-                // Uses Invariant: remove documents actually always returns Ok for the inner result
-                let count = user_result.unwrap();
-
-                for (task, documents) in tasks.iter_mut().zip(documents) {
-                    task.status = Status::Succeeded;
-                    task.details = Some(Details::DocumentDeletion {
-                        provided_ids: documents.len(),
-                        deleted_documents: Some(count.min(documents.len() as u64)),
-                    });
-                }
-
-                if !tasks.iter().all(|res| res.error.is_some()) {
-                    let addition = builder.execute()?;
-                    info!("document deletion done: {:?}", addition);
-                }
-
-                Ok(tasks)
-            }
             IndexOperation::IndexDocumentDeletionByFilter { mut task, index_uid: _ } => {
                 let filter =
                     if let KindWithContent::DocumentDeletionByFilter { filter_expr, .. } =
@@ -1367,6 +1343,9 @@ impl IndexScheduler {
 
                 for (task, (_, settings)) in tasks.iter_mut().zip(settings) {
                     let checked_settings = settings.clone().check();
+                    if checked_settings.proximity_precision.set().is_some() {
+                        self.features.features().check_proximity_precision()?;
+                    }
                     task.details = Some(Details::SettingsUpdate { settings: Box::new(settings) });
                     apply_settings_to_builder(&checked_settings, &mut builder);
 
@@ -1503,10 +1482,9 @@ impl IndexScheduler {
         }
 
         for task in to_delete_tasks.iter() {
-            self.all_tasks.delete(wtxn, &BEU32::new(task))?;
+            self.all_tasks.delete(wtxn, &task)?;
         }
         for canceled_by in affected_canceled_by {
-            let canceled_by = BEU32::new(canceled_by);
             if let Some(mut tasks) = self.canceled_by.get(wtxn, &canceled_by)? {
                 tasks -= &to_delete_tasks;
                 if tasks.is_empty() {
@@ -1554,14 +1532,14 @@ impl IndexScheduler {
             task.details = task.details.map(|d| d.to_failed());
             self.update_task(wtxn, &task)?;
         }
-        self.canceled_by.put(wtxn, &BEU32::new(cancel_task_id), &tasks_to_cancel)?;
+        self.canceled_by.put(wtxn, &cancel_task_id, &tasks_to_cancel)?;
 
         Ok(content_files_to_delete)
     }
 }
 
 fn delete_document_by_filter<'a>(
-    wtxn: &mut RwTxn<'a, '_>,
+    wtxn: &mut RwTxn<'a>,
     filter: &serde_json::Value,
     indexer_config: &IndexerConfig,
     must_stop_processing: MustStopProcessing,
@@ -1575,18 +1553,6 @@ fn delete_document_by_filter<'a>(
             }
             e => e.into(),
         })?;
-        let external_documents_ids = index.external_documents_ids();
-        // FIXME: for filters matching a lot of documents, this will allocate a huge vec of external docids (strings).
-        // Since what we have is an iterator, it would be better to delete in chunks
-        let external_to_internal: std::result::Result<Vec<_>, RoaringBitmap> =
-            external_documents_ids
-                .find_external_id_of(wtxn, candidates)?
-                .only_external_ids()
-                .collect();
-        let document_ids = match external_to_internal {
-            Ok(external_ids) => external_ids,
-            Err(remaining_ids) => panic!("Couldn't find some external ids {:?}", remaining_ids),
-        };
 
         let config = IndexDocumentsConfig {
             update_method: IndexDocumentsMethod::ReplaceDocuments,
@@ -1602,13 +1568,10 @@ fn delete_document_by_filter<'a>(
             || must_stop_processing.get(),
         )?;
 
-        let (new_builder, user_result) = builder.remove_documents(document_ids)?;
+        let (new_builder, count) = builder.remove_documents_from_db_no_batch(&candidates)?;
         builder = new_builder;
-        // Uses Invariant: remove documents actually always returns Ok for the inner result
-        let count = user_result.unwrap();
 
         let _ = builder.execute()?;
-
         count
     } else {
         0

index-scheduler/src/error.rs

@@ -108,6 +108,8 @@ pub enum Error {
     TaskDeletionWithEmptyQuery,
     #[error("Query parameters to filter the tasks to cancel are missing. Available query parameters are: `uids`, `indexUids`, `statuses`, `types`, `canceledBy`, `beforeEnqueuedAt`, `afterEnqueuedAt`, `beforeStartedAt`, `afterStartedAt`, `beforeFinishedAt`, `afterFinishedAt`.")]
     TaskCancelationWithEmptyQuery,
+    #[error("Aborted task")]
+    AbortedTask,
 
     #[error(transparent)]
     Dump(#[from] dump::Error),
@@ -175,6 +177,7 @@ impl Error {
             | Error::TaskNotFound(_)
             | Error::TaskDeletionWithEmptyQuery
             | Error::TaskCancelationWithEmptyQuery
+            | Error::AbortedTask
             | Error::Dump(_)
             | Error::Heed(_)
             | Error::Milli(_)
@@ -236,6 +239,9 @@ impl ErrorCode for Error {
             Error::TaskDatabaseUpdate(_) => Code::Internal,
             Error::CreateBatch(_) => Code::Internal,
 
+            // This one should never be seen by the end user
+            Error::AbortedTask => Code::Internal,
+
             #[cfg(test)]
             Error::PlannedFailure => Code::Internal,
         }

index-scheduler/src/features.rs

@@ -81,6 +81,19 @@ impl RoFeatures {
             .into())
         }
     }
+
+    pub fn check_proximity_precision(&self) -> Result<()> {
+        if self.runtime.proximity_precision {
+            Ok(())
+        } else {
+            Err(FeatureNotEnabledError {
+                disabled_action: "Using `proximityPrecision` index setting",
+                feature: "proximity precision",
+                issue_link: "https://github.com/orgs/meilisearch/discussions/710",
+            }
+            .into())
+        }
+    }
 }
 
 impl FeatureData {

index-scheduler/src/index_mapper/index_map.rs

@@ -1,12 +1,8 @@
-/// the map size to use when we don't succeed in reading it in indexes.
-const DEFAULT_MAP_SIZE: usize = 10 * 1024 * 1024 * 1024; // 10 GiB
-
 use std::collections::BTreeMap;
 use std::path::Path;
 use std::time::Duration;
 
-use meilisearch_types::heed::flags::Flags;
-use meilisearch_types::heed::{EnvClosingEvent, EnvOpenOptions};
+use meilisearch_types::heed::{EnvClosingEvent, EnvFlags, EnvOpenOptions};
 use meilisearch_types::milli::Index;
 use time::OffsetDateTime;
 use uuid::Uuid;
@@ -236,7 +232,7 @@ impl IndexMap {
         enable_mdb_writemap: bool,
         map_size_growth: usize,
     ) {
-        let map_size = index.map_size().unwrap_or(DEFAULT_MAP_SIZE) + map_size_growth;
+        let map_size = index.map_size() + map_size_growth;
         let closing_event = index.prepare_for_closing();
         let generation = self.next_generation();
         self.unavailable.insert(
@@ -309,7 +305,7 @@ fn create_or_open_index(
     options.map_size(clamp_to_page_size(map_size));
     options.max_readers(1024);
     if enable_mdb_writemap {
-        unsafe { options.flag(Flags::MdbWriteMap) };
+        unsafe { options.flags(EnvFlags::WRITE_MAP) };
     }
 
     if let Some((created, updated)) = date {
@@ -388,7 +384,7 @@ mod tests {
 
     fn assert_index_size(index: Index, expected: usize) {
         let expected = clamp_to_page_size(expected);
-        let index_map_size = index.map_size().unwrap();
+        let index_map_size = index.map_size();
         assert_eq!(index_map_size, expected);
     }
 }

index-scheduler/src/insta_snapshot.rs

@@ -1,7 +1,7 @@
 use std::collections::BTreeSet;
 use std::fmt::Write;
 
-use meilisearch_types::heed::types::{OwnedType, SerdeBincode, SerdeJson, Str};
+use meilisearch_types::heed::types::{SerdeBincode, SerdeJson, Str};
 use meilisearch_types::heed::{Database, RoTxn};
 use meilisearch_types::milli::{CboRoaringBitmapCodec, RoaringBitmapCodec, BEU32};
 use meilisearch_types::tasks::{Details, Task};
@@ -39,6 +39,7 @@ pub fn snapshot_index_scheduler(scheduler: &IndexScheduler) -> String {
         test_breakpoint_sdr: _,
         planned_failures: _,
         run_loop_iteration: _,
+        currently_updating_index: _,
     } = scheduler;
 
     let rtxn = env.read_txn().unwrap();
@@ -114,7 +115,7 @@ pub fn snapshot_bitmap(r: &RoaringBitmap) -> String {
     snap
 }
 
-pub fn snapshot_all_tasks(rtxn: &RoTxn, db: Database<OwnedType<BEU32>, SerdeJson<Task>>) -> String {
+pub fn snapshot_all_tasks(rtxn: &RoTxn, db: Database<BEU32, SerdeJson<Task>>) -> String {
     let mut snap = String::new();
     let iter = db.iter(rtxn).unwrap();
     for next in iter {
@@ -124,10 +125,7 @@ pub fn snapshot_all_tasks(rtxn: &RoTxn, db: Database<OwnedType<BEU32>, SerdeJson
     snap
 }
 
-pub fn snapshot_date_db(
-    rtxn: &RoTxn,
-    db: Database<OwnedType<BEI128>, CboRoaringBitmapCodec>,
-) -> String {
+pub fn snapshot_date_db(rtxn: &RoTxn, db: Database<BEI128, CboRoaringBitmapCodec>) -> String {
     let mut snap = String::new();
     let iter = db.iter(rtxn).unwrap();
     for next in iter {
@@ -247,10 +245,7 @@ pub fn snapshot_index_tasks(rtxn: &RoTxn, db: Database<Str, RoaringBitmapCodec>)
     }
     snap
 }
-pub fn snapshot_canceled_by(
-    rtxn: &RoTxn,
-    db: Database<OwnedType<BEU32>, RoaringBitmapCodec>,
-) -> String {
+pub fn snapshot_canceled_by(rtxn: &RoTxn, db: Database<BEU32, RoaringBitmapCodec>) -> String {
     let mut snap = String::new();
     let iter = db.iter(rtxn).unwrap();
     for next in iter {

index-scheduler/src/lib.rs

@@ -27,7 +27,7 @@ mod index_mapper;
 mod insta_snapshot;
 mod lru;
 mod utils;
-mod uuid_codec;
+pub mod uuid_codec;
 
 pub type Result<T> = std::result::Result<T, Error>;
 pub type TaskId = u32;
@@ -47,8 +47,9 @@ pub use features::RoFeatures;
 use file_store::FileStore;
 use meilisearch_types::error::ResponseError;
 use meilisearch_types::features::{InstanceTogglableFeatures, RuntimeTogglableFeatures};
-use meilisearch_types::heed::types::{OwnedType, SerdeBincode, SerdeJson, Str};
-use meilisearch_types::heed::{self, Database, Env, RoTxn, RwTxn};
+use meilisearch_types::heed::byteorder::BE;
+use meilisearch_types::heed::types::{SerdeBincode, SerdeJson, Str, I128};
+use meilisearch_types::heed::{self, Database, Env, PutFlags, RoTxn, RwTxn};
 use meilisearch_types::milli::documents::DocumentsBatchBuilder;
 use meilisearch_types::milli::update::IndexerConfig;
 use meilisearch_types::milli::{self, CboRoaringBitmapCodec, Index, RoaringBitmapCodec, BEU32};
@@ -64,8 +65,7 @@ use uuid::Uuid;
 use crate::index_mapper::IndexMapper;
 use crate::utils::{check_index_swap_validity, clamp_to_page_size};
 
-pub(crate) type BEI128 =
-    meilisearch_types::heed::zerocopy::I128<meilisearch_types::heed::byteorder::BE>;
+pub(crate) type BEI128 = I128<BE>;
 
 /// Defines a subset of tasks to be retrieved from the [`IndexScheduler`].
 ///
@@ -278,7 +278,7 @@ pub struct IndexScheduler {
     pub(crate) file_store: FileStore,
 
     // The main database, it contains all the tasks accessible by their Id.
-    pub(crate) all_tasks: Database<OwnedType<BEU32>, SerdeJson<Task>>,
+    pub(crate) all_tasks: Database<BEU32, SerdeJson<Task>>,
 
     /// All the tasks ids grouped by their status.
     // TODO we should not be able to serialize a `Status::Processing` in this database.
@@ -289,16 +289,16 @@ pub struct IndexScheduler {
     pub(crate) index_tasks: Database<Str, RoaringBitmapCodec>,
 
     /// Store the tasks that were canceled by a task uid
-    pub(crate) canceled_by: Database<OwnedType<BEU32>, RoaringBitmapCodec>,
+    pub(crate) canceled_by: Database<BEU32, RoaringBitmapCodec>,
 
     /// Store the task ids of tasks which were enqueued at a specific date
-    pub(crate) enqueued_at: Database<OwnedType<BEI128>, CboRoaringBitmapCodec>,
+    pub(crate) enqueued_at: Database<BEI128, CboRoaringBitmapCodec>,
 
     /// Store the task ids of finished tasks which started being processed at a specific date
-    pub(crate) started_at: Database<OwnedType<BEI128>, CboRoaringBitmapCodec>,
+    pub(crate) started_at: Database<BEI128, CboRoaringBitmapCodec>,
 
     /// Store the task ids of tasks which finished at a specific date
-    pub(crate) finished_at: Database<OwnedType<BEI128>, CboRoaringBitmapCodec>,
+    pub(crate) finished_at: Database<BEI128, CboRoaringBitmapCodec>,
 
     /// In charge of creating, opening, storing and returning indexes.
     pub(crate) index_mapper: IndexMapper,
@@ -331,6 +331,10 @@ pub struct IndexScheduler {
     /// The path to the version file of Meilisearch.
     pub(crate) version_file_path: PathBuf,
 
+    /// A few types of long running batches of tasks that act on a single index set this field
+    /// so that a handle to the index is available from other threads (search) in an optimized manner.
+    currently_updating_index: Arc<RwLock<Option<(String, Index)>>>,
+
     // ================= test
     // The next entry is dedicated to the tests.
     /// Provide a way to set a breakpoint in multiple part of the scheduler.
@@ -374,6 +378,7 @@ impl IndexScheduler {
             dumps_path: self.dumps_path.clone(),
             auth_path: self.auth_path.clone(),
             version_file_path: self.version_file_path.clone(),
+            currently_updating_index: self.currently_updating_index.clone(),
             #[cfg(test)]
             test_breakpoint_sdr: self.test_breakpoint_sdr.clone(),
             #[cfg(test)]
@@ -470,6 +475,7 @@ impl IndexScheduler {
             snapshots_path: options.snapshots_path,
             auth_path: options.auth_path,
             version_file_path: options.version_file_path,
+            currently_updating_index: Arc::new(RwLock::new(None)),
 
             #[cfg(test)]
             test_breakpoint_sdr,
@@ -652,6 +658,13 @@ impl IndexScheduler {
     /// If you need to fetch information from or perform an action on all indexes,
     /// see the `try_for_each_index` function.
     pub fn index(&self, name: &str) -> Result<Index> {
+        if let Some((current_name, current_index)) =
+            self.currently_updating_index.read().unwrap().as_ref()
+        {
+            if current_name == name {
+                return Ok(current_index.clone());
+            }
+        }
         let rtxn = self.env.read_txn()?;
         self.index_mapper.index(&rtxn, name)
     }
@@ -717,9 +730,7 @@ impl IndexScheduler {
         if let Some(canceled_by) = &query.canceled_by {
             let mut all_canceled_tasks = RoaringBitmap::new();
             for cancel_task_uid in canceled_by {
-                if let Some(canceled_by_uid) =
-                    self.canceled_by.get(rtxn, &BEU32::new(*cancel_task_uid))?
-                {
+                if let Some(canceled_by_uid) = self.canceled_by.get(rtxn, cancel_task_uid)? {
                     all_canceled_tasks |= canceled_by_uid;
                 }
             }
@@ -970,7 +981,7 @@ impl IndexScheduler {
 
         // if the task doesn't delete anything and 50% of the task queue is full, we must refuse to enqueue the incomming task
         if !matches!(&kind, KindWithContent::TaskDeletion { tasks, .. } if !tasks.is_empty())
-            && (self.env.non_free_pages_size()? * 100) / self.env.map_size()? as u64 > 50
+            && (self.env.non_free_pages_size()? * 100) / self.env.info().map_size as u64 > 50
         {
             return Err(Error::NoSpaceLeftInTaskQueue);
         }
@@ -996,7 +1007,7 @@ impl IndexScheduler {
         // Get rid of the mutability.
         let task = task;
 
-        self.all_tasks.append(&mut wtxn, &BEU32::new(task.uid), &task)?;
+        self.all_tasks.put_with_flags(&mut wtxn, PutFlags::APPEND, &task.uid, &task)?;
 
         for index in task.indexes() {
             self.update_index(&mut wtxn, index, |bitmap| {
@@ -1133,6 +1144,9 @@ impl IndexScheduler {
             handle.join().unwrap_or(Err(Error::ProcessBatchPanicked))
         };
 
+        // Reset the currently updating index to relinquish the index handle
+        *self.currently_updating_index.write().unwrap() = None;
+
         #[cfg(test)]
         self.maybe_fail(tests::FailureLocation::AcquiringWtxn)?;
 
@@ -1167,10 +1181,11 @@ impl IndexScheduler {
             // If we have an abortion error we must stop the tick here and re-schedule tasks.
             Err(Error::Milli(milli::Error::InternalError(
                 milli::InternalError::AbortedIndexation,
-            ))) => {
+            )))
+            | Err(Error::AbortedTask) => {
                 #[cfg(test)]
                 self.breakpoint(Breakpoint::AbortedIndexation);
-                wtxn.abort().map_err(Error::HeedTransaction)?;
+                wtxn.abort();
 
                 // We make sure that we don't call `stop_processing` on the `processing_tasks`,
                 // this is because we want to let the next tick call `create_next_batch` and keep
@@ -1191,7 +1206,7 @@ impl IndexScheduler {
                 let index_uid = index_uid.unwrap();
                 // fixme: handle error more gracefully? not sure when this could happen
                 self.index_mapper.resize_index(&wtxn, &index_uid)?;
-                wtxn.abort().map_err(Error::HeedTransaction)?;
+                wtxn.abort();
 
                 return Ok(TickOutcome::TickAgain(0));
             }
@@ -1337,7 +1352,7 @@ impl IndexScheduler {
 
 pub struct Dump<'a> {
     index_scheduler: &'a IndexScheduler,
-    wtxn: RwTxn<'a, 'a>,
+    wtxn: RwTxn<'a>,
 
     indexes: HashMap<String, RoaringBitmap>,
     statuses: HashMap<Status, RoaringBitmap>,
@@ -1452,7 +1467,7 @@ impl<'a> Dump<'a> {
             },
         };
 
-        self.index_scheduler.all_tasks.put(&mut self.wtxn, &BEU32::new(task.uid), &task)?;
+        self.index_scheduler.all_tasks.put(&mut self.wtxn, &task.uid, &task)?;
 
         for index in task.indexes() {
             match self.indexes.get_mut(index) {
@@ -1494,8 +1509,8 @@ impl<'a> Dump<'a> {
             }
         }
 
-        self.statuses.entry(task.status).or_insert(RoaringBitmap::new()).insert(task.uid);
-        self.kinds.entry(task.kind.as_kind()).or_insert(RoaringBitmap::new()).insert(task.uid);
+        self.statuses.entry(task.status).or_default().insert(task.uid);
+        self.kinds.entry(task.kind.as_kind()).or_default().insert(task.uid);
 
         Ok(task)
     }
@@ -4323,4 +4338,26 @@ mod tests {
         }
         "###);
     }
+
+    #[test]
+    fn cancel_processing_dump() {
+        let (index_scheduler, mut handle) = IndexScheduler::test(true, vec![]);
+
+        let dump_creation = KindWithContent::DumpCreation { keys: Vec::new(), instance_uid: None };
+        let dump_cancellation = KindWithContent::TaskCancelation {
+            query: "cancel dump".to_owned(),
+            tasks: RoaringBitmap::from_iter([0]),
+        };
+        let _ = index_scheduler.register(dump_creation).unwrap();
+        snapshot!(snapshot_index_scheduler(&index_scheduler), name: "after_dump_register");
+        handle.advance_till([Start, BatchCreated, InsideProcessBatch]);
+
+        let _ = index_scheduler.register(dump_cancellation).unwrap();
+        snapshot!(snapshot_index_scheduler(&index_scheduler), name: "cancel_registered");
+
+        snapshot!(format!("{:?}", handle.advance()), @"AbortedIndexation");
+
+        handle.advance_one_successful_batch();
+        snapshot!(snapshot_index_scheduler(&index_scheduler), name: "cancel_processed");
+    }
 }

index-scheduler/src/snapshots/lib.rs/cancel_processing_dump/after_dump_register.snap

@@ -0,0 +1,35 @@
+---
+source: index-scheduler/src/lib.rs
+---
+### Autobatching Enabled = true
+### Processing Tasks:
+[]
+----------------------------------------------------------------------
+### All Tasks:
+0 {uid: 0, status: enqueued, details: { dump_uid: None }, kind: DumpCreation { keys: [], instance_uid: None }}
+----------------------------------------------------------------------
+### Status:
+enqueued [0,]
+----------------------------------------------------------------------
+### Kind:
+"dumpCreation" [0,]
+----------------------------------------------------------------------
+### Index Tasks:
+----------------------------------------------------------------------
+### Index Mapper:
+
+----------------------------------------------------------------------
+### Canceled By:
+
+----------------------------------------------------------------------
+### Enqueued At:
+[timestamp] [0,]
+----------------------------------------------------------------------
+### Started At:
+----------------------------------------------------------------------
+### Finished At:
+----------------------------------------------------------------------
+### File Store:
+
+----------------------------------------------------------------------
+

index-scheduler/src/snapshots/lib.rs/cancel_processing_dump/cancel_processed.snap

@@ -0,0 +1,45 @@
+---
+source: index-scheduler/src/lib.rs
+---
+### Autobatching Enabled = true
+### Processing Tasks:
+[]
+----------------------------------------------------------------------
+### All Tasks:
+0 {uid: 0, status: canceled, canceled_by: 1, details: { dump_uid: None }, kind: DumpCreation { keys: [], instance_uid: None }}
+1 {uid: 1, status: succeeded, details: { matched_tasks: 1, canceled_tasks: Some(0), original_filter: "cancel dump" }, kind: TaskCancelation { query: "cancel dump", tasks: RoaringBitmap<[0]> }}
+----------------------------------------------------------------------
+### Status:
+enqueued []
+succeeded [1,]
+canceled [0,]
+----------------------------------------------------------------------
+### Kind:
+"taskCancelation" [1,]
+"dumpCreation" [0,]
+----------------------------------------------------------------------
+### Index Tasks:
+----------------------------------------------------------------------
+### Index Mapper:
+
+----------------------------------------------------------------------
+### Canceled By:
+1 [0,]
+
+----------------------------------------------------------------------
+### 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/cancel_processing_dump/cancel_registered.snap

@@ -0,0 +1,38 @@
+---
+source: index-scheduler/src/lib.rs
+---
+### Autobatching Enabled = true
+### Processing Tasks:
+[0,]
+----------------------------------------------------------------------
+### All Tasks:
+0 {uid: 0, status: enqueued, details: { dump_uid: None }, kind: DumpCreation { keys: [], instance_uid: None }}
+1 {uid: 1, status: enqueued, details: { matched_tasks: 1, canceled_tasks: None, original_filter: "cancel dump" }, kind: TaskCancelation { query: "cancel dump", tasks: RoaringBitmap<[0]> }}
+----------------------------------------------------------------------
+### Status:
+enqueued [0,1,]
+----------------------------------------------------------------------
+### Kind:
+"taskCancelation" [1,]
+"dumpCreation" [0,]
+----------------------------------------------------------------------
+### Index Tasks:
+----------------------------------------------------------------------
+### Index Mapper:
+
+----------------------------------------------------------------------
+### Canceled By:
+
+----------------------------------------------------------------------
+### Enqueued At:
+[timestamp] [0,]
+[timestamp] [1,]
+----------------------------------------------------------------------
+### Started At:
+----------------------------------------------------------------------
+### Finished At:
+----------------------------------------------------------------------
+### File Store:
+
+----------------------------------------------------------------------
+

index-scheduler/src/utils.rs

@@ -3,9 +3,9 @@
 use std::collections::{BTreeSet, HashSet};
 use std::ops::Bound;
 
-use meilisearch_types::heed::types::{DecodeIgnore, OwnedType};
+use meilisearch_types::heed::types::DecodeIgnore;
 use meilisearch_types::heed::{Database, RoTxn, RwTxn};
-use meilisearch_types::milli::{CboRoaringBitmapCodec, BEU32};
+use meilisearch_types::milli::CboRoaringBitmapCodec;
 use meilisearch_types::tasks::{Details, IndexSwap, Kind, KindWithContent, Status};
 use roaring::{MultiOps, RoaringBitmap};
 use time::OffsetDateTime;
@@ -18,7 +18,7 @@ impl IndexScheduler {
     }
 
     pub(crate) fn last_task_id(&self, rtxn: &RoTxn) -> Result<Option<TaskId>> {
-        Ok(self.all_tasks.remap_data_type::<DecodeIgnore>().last(rtxn)?.map(|(k, _)| k.get() + 1))
+        Ok(self.all_tasks.remap_data_type::<DecodeIgnore>().last(rtxn)?.map(|(k, _)| k + 1))
     }
 
     pub(crate) fn next_task_id(&self, rtxn: &RoTxn) -> Result<TaskId> {
@@ -26,7 +26,7 @@ impl IndexScheduler {
     }
 
     pub(crate) fn get_task(&self, rtxn: &RoTxn, task_id: TaskId) -> Result<Option<Task>> {
-        Ok(self.all_tasks.get(rtxn, &BEU32::new(task_id))?)
+        Ok(self.all_tasks.get(rtxn, &task_id)?)
     }
 
     /// Convert an iterator to a `Vec` of tasks. The tasks MUST exist or a
@@ -88,7 +88,7 @@ impl IndexScheduler {
             }
         }
 
-        self.all_tasks.put(wtxn, &BEU32::new(task.uid), task)?;
+        self.all_tasks.put(wtxn, &task.uid, task)?;
         Ok(())
     }
 
@@ -169,11 +169,11 @@ impl IndexScheduler {
 
 pub(crate) fn insert_task_datetime(
     wtxn: &mut RwTxn,
-    database: Database<OwnedType<BEI128>, CboRoaringBitmapCodec>,
+    database: Database<BEI128, CboRoaringBitmapCodec>,
     time: OffsetDateTime,
     task_id: TaskId,
 ) -> Result<()> {
-    let timestamp = BEI128::new(time.unix_timestamp_nanos());
+    let timestamp = time.unix_timestamp_nanos();
     let mut task_ids = database.get(wtxn, &timestamp)?.unwrap_or_default();
     task_ids.insert(task_id);
     database.put(wtxn, &timestamp, &RoaringBitmap::from_iter(task_ids))?;
@@ -182,11 +182,11 @@ pub(crate) fn insert_task_datetime(
 
 pub(crate) fn remove_task_datetime(
     wtxn: &mut RwTxn,
-    database: Database<OwnedType<BEI128>, CboRoaringBitmapCodec>,
+    database: Database<BEI128, CboRoaringBitmapCodec>,
     time: OffsetDateTime,
     task_id: TaskId,
 ) -> Result<()> {
-    let timestamp = BEI128::new(time.unix_timestamp_nanos());
+    let timestamp = time.unix_timestamp_nanos();
     if let Some(mut existing) = database.get(wtxn, &timestamp)? {
         existing.remove(task_id);
         if existing.is_empty() {
@@ -202,7 +202,7 @@ pub(crate) fn remove_task_datetime(
 pub(crate) fn keep_tasks_within_datetimes(
     rtxn: &RoTxn,
     tasks: &mut RoaringBitmap,
-    database: Database<OwnedType<BEI128>, CboRoaringBitmapCodec>,
+    database: Database<BEI128, CboRoaringBitmapCodec>,
     after: Option<OffsetDateTime>,
     before: Option<OffsetDateTime>,
 ) -> Result<()> {
@@ -213,8 +213,8 @@ pub(crate) fn keep_tasks_within_datetimes(
         (Some(after), Some(before)) => (Bound::Excluded(*after), Bound::Excluded(*before)),
     };
     let mut collected_task_ids = RoaringBitmap::new();
-    let start = map_bound(start, |b| BEI128::new(b.unix_timestamp_nanos()));
-    let end = map_bound(end, |b| BEI128::new(b.unix_timestamp_nanos()));
+    let start = map_bound(start, |b| b.unix_timestamp_nanos());
+    let end = map_bound(end, |b| b.unix_timestamp_nanos());
     let iter = database.range(rtxn, &(start, end))?;
     for r in iter {
         let (_timestamp, task_ids) = r?;
@@ -337,8 +337,6 @@ impl IndexScheduler {
         let rtxn = self.env.read_txn().unwrap();
         for task in self.all_tasks.iter(&rtxn).unwrap() {
             let (task_id, task) = task.unwrap();
-            let task_id = task_id.get();
-
             let task_index_uid = task.index_uid().map(ToOwned::to_owned);
 
             let Task {
@@ -361,16 +359,13 @@ impl IndexScheduler {
                     .unwrap()
                     .contains(task.uid));
             }
-            let db_enqueued_at = self
-                .enqueued_at
-                .get(&rtxn, &BEI128::new(enqueued_at.unix_timestamp_nanos()))
-                .unwrap()
-                .unwrap();
+            let db_enqueued_at =
+                self.enqueued_at.get(&rtxn, &enqueued_at.unix_timestamp_nanos()).unwrap().unwrap();
             assert!(db_enqueued_at.contains(task_id));
             if let Some(started_at) = started_at {
                 let db_started_at = self
                     .started_at
-                    .get(&rtxn, &BEI128::new(started_at.unix_timestamp_nanos()))
+                    .get(&rtxn, &started_at.unix_timestamp_nanos())
                     .unwrap()
                     .unwrap();
                 assert!(db_started_at.contains(task_id));
@@ -378,7 +373,7 @@ impl IndexScheduler {
             if let Some(finished_at) = finished_at {
                 let db_finished_at = self
                     .finished_at
-                    .get(&rtxn, &BEI128::new(finished_at.unix_timestamp_nanos()))
+                    .get(&rtxn, &finished_at.unix_timestamp_nanos())
                     .unwrap()
                     .unwrap();
                 assert!(db_finished_at.contains(task_id));

index-scheduler/src/uuid_codec.rs

@@ -1,7 +1,7 @@
 use std::borrow::Cow;
 use std::convert::TryInto;
 
-use meilisearch_types::heed::{BytesDecode, BytesEncode};
+use meilisearch_types::heed::{BoxedError, BytesDecode, BytesEncode};
 use uuid::Uuid;
 
 /// A heed codec for value of struct Uuid.
@@ -10,15 +10,15 @@ pub struct UuidCodec;
 impl<'a> BytesDecode<'a> for UuidCodec {
     type DItem = Uuid;
 
-    fn bytes_decode(bytes: &'a [u8]) -> Option<Self::DItem> {
-        bytes.try_into().ok().map(Uuid::from_bytes)
+    fn bytes_decode(bytes: &'a [u8]) -> Result<Self::DItem, BoxedError> {
+        bytes.try_into().map(Uuid::from_bytes).map_err(Into::into)
     }
 }
 
 impl BytesEncode<'_> for UuidCodec {
     type EItem = Uuid;
 
-    fn bytes_encode(item: &Self::EItem) -> Option<Cow<[u8]>> {
-        Some(Cow::Borrowed(item.as_bytes()))
+    fn bytes_encode(item: &Self::EItem) -> Result<Cow<[u8]>, BoxedError> {
+        Ok(Cow::Borrowed(item.as_bytes()))
     }
 }

meilisearch-auth/src/store.rs

@@ -4,17 +4,20 @@ use std::collections::HashSet;
 use std::convert::{TryFrom, TryInto};
 use std::fs::create_dir_all;
 use std::path::Path;
+use std::result::Result as StdResult;
 use std::str;
 use std::str::FromStr;
 use std::sync::Arc;
 
 use hmac::{Hmac, Mac};
+use meilisearch_types::heed::BoxedError;
 use meilisearch_types::index_uid_pattern::IndexUidPattern;
 use meilisearch_types::keys::KeyId;
 use meilisearch_types::milli;
-use meilisearch_types::milli::heed::types::{ByteSlice, DecodeIgnore, SerdeJson};
+use meilisearch_types::milli::heed::types::{Bytes, DecodeIgnore, SerdeJson};
 use meilisearch_types::milli::heed::{Database, Env, EnvOpenOptions, RwTxn};
 use sha2::Sha256;
+use thiserror::Error;
 use time::OffsetDateTime;
 use uuid::fmt::Hyphenated;
 use uuid::Uuid;
@@ -30,7 +33,7 @@ const KEY_ID_ACTION_INDEX_EXPIRATION_DB_NAME: &str = "keyid-action-index-expirat
 #[derive(Clone)]
 pub struct HeedAuthStore {
     env: Arc<Env>,
-    keys: Database<ByteSlice, SerdeJson<Key>>,
+    keys: Database<Bytes, SerdeJson<Key>>,
     action_keyid_index_expiration: Database<KeyIdActionCodec, SerdeJson<Option<OffsetDateTime>>>,
     should_close_on_drop: bool,
 }
@@ -276,7 +279,7 @@ impl HeedAuthStore {
     fn delete_key_from_inverted_db(&self, wtxn: &mut RwTxn, key: &KeyId) -> Result<()> {
         let mut iter = self
             .action_keyid_index_expiration
-            .remap_types::<ByteSlice, DecodeIgnore>()
+            .remap_types::<Bytes, DecodeIgnore>()
             .prefix_iter_mut(wtxn, key.as_bytes())?;
         while iter.next().transpose()?.is_some() {
             // safety: we don't keep references from inside the LMDB database.
@@ -294,23 +297,24 @@ pub struct KeyIdActionCodec;
 impl<'a> milli::heed::BytesDecode<'a> for KeyIdActionCodec {
     type DItem = (KeyId, Action, Option<&'a [u8]>);
 
-    fn bytes_decode(bytes: &'a [u8]) -> Option<Self::DItem> {
-        let (key_id_bytes, action_bytes) = try_split_array_at(bytes)?;
-        let (action_bytes, index) = match try_split_array_at(action_bytes)? {
-            (action, []) => (action, None),
-            (action, index) => (action, Some(index)),
-        };
+    fn bytes_decode(bytes: &'a [u8]) -> StdResult<Self::DItem, BoxedError> {
+        let (key_id_bytes, action_bytes) = try_split_array_at(bytes).ok_or(SliceTooShortError)?;
+        let (&action_byte, index) =
+            match try_split_array_at(action_bytes).ok_or(SliceTooShortError)? {
+                ([action], []) => (action, None),
+                ([action], index) => (action, Some(index)),
+            };
         let key_id = Uuid::from_bytes(*key_id_bytes);
-        let action = Action::from_repr(u8::from_be_bytes(*action_bytes))?;
+        let action = Action::from_repr(action_byte).ok_or(InvalidActionError { action_byte })?;
 
-        Some((key_id, action, index))
+        Ok((key_id, action, index))
     }
 }
 
 impl<'a> milli::heed::BytesEncode<'a> for KeyIdActionCodec {
     type EItem = (&'a KeyId, &'a Action, Option<&'a [u8]>);
 
-    fn bytes_encode((key_id, action, index): &Self::EItem) -> Option<Cow<[u8]>> {
+    fn bytes_encode((key_id, action, index): &Self::EItem) -> StdResult<Cow<[u8]>, BoxedError> {
         let mut bytes = Vec::new();
 
         bytes.extend_from_slice(key_id.as_bytes());
@@ -320,10 +324,20 @@ impl<'a> milli::heed::BytesEncode<'a> for KeyIdActionCodec {
             bytes.extend_from_slice(index);
         }
 
-        Some(Cow::Owned(bytes))
+        Ok(Cow::Owned(bytes))
     }
 }
 
+#[derive(Error, Debug)]
+#[error("the slice is too short")]
+pub struct SliceTooShortError;
+
+#[derive(Error, Debug)]
+#[error("cannot construct a valid Action from {action_byte}")]
+pub struct InvalidActionError {
+    pub action_byte: u8,
+}
+
 pub fn generate_key_as_hexa(uid: Uuid, master_key: &[u8]) -> String {
     // format uid as hyphenated allowing user to generate their own keys.
     let mut uid_buffer = [0; Hyphenated::LENGTH];

meilisearch-types/Cargo.toml

@@ -15,7 +15,7 @@ actix-web = { version = "4.3.1", default-features = false }
 anyhow = "1.0.70"
 convert_case = "0.6.0"
 csv = "1.2.1"
-deserr = { version = "0.6.0", features = ["actix-web"]}
+deserr = { version = "0.6.0", features = ["actix-web"] }
 either = { version = "1.8.1", features = ["serde"] }
 enum-iterator = "1.4.0"
 file-store = { path = "../file-store" }
@@ -50,6 +50,7 @@ hebrew = ["milli/hebrew"]
 japanese = ["milli/japanese"]
 # thai specialized tokenization
 thai = ["milli/thai"]
-
 # allow greek specialized tokenization
 greek = ["milli/greek"]
+# allow khmer specialized tokenization
+khmer = ["milli/khmer"]

meilisearch-types/src/error.rs

@@ -252,6 +252,7 @@ InvalidSearchShowRankingScoreDetails  , InvalidRequest       , BAD_REQUEST ;
 InvalidSearchSort                     , InvalidRequest       , BAD_REQUEST ;
 InvalidSettingsDisplayedAttributes    , InvalidRequest       , BAD_REQUEST ;
 InvalidSettingsDistinctAttribute      , InvalidRequest       , BAD_REQUEST ;
+InvalidSettingsProximityPrecision     , InvalidRequest       , BAD_REQUEST ;
 InvalidSettingsFaceting               , InvalidRequest       , BAD_REQUEST ;
 InvalidSettingsFilterableAttributes   , InvalidRequest       , BAD_REQUEST ;
 InvalidSettingsPagination             , InvalidRequest       , BAD_REQUEST ;
@@ -386,11 +387,11 @@ impl ErrorCode for HeedError {
             HeedError::Mdb(MdbError::Invalid) => Code::InvalidStoreFile,
             HeedError::Io(e) => e.error_code(),
             HeedError::Mdb(_)
-            | HeedError::Encoding
-            | HeedError::Decoding
+            | HeedError::Encoding(_)
+            | HeedError::Decoding(_)
             | HeedError::InvalidDatabaseTyping
             | HeedError::DatabaseClosing
-            | HeedError::BadOpenOptions => Code::Internal,
+            | HeedError::BadOpenOptions { .. } => Code::Internal,
         }
     }
 }

meilisearch-types/src/features.rs

@@ -7,6 +7,7 @@ pub struct RuntimeTogglableFeatures {
     pub vector_store: bool,
     pub metrics: bool,
     pub export_puffin_reports: bool,
+    pub proximity_precision: bool,
 }
 
 #[derive(Default, Debug, Clone, Copy)]

meilisearch-types/src/settings.rs

@@ -8,6 +8,7 @@ use std::str::FromStr;
 
 use deserr::{DeserializeError, Deserr, ErrorKind, MergeWithError, ValuePointerRef};
 use fst::IntoStreamer;
+use milli::proximity::ProximityPrecision;
 use milli::update::Setting;
 use milli::{Criterion, CriterionError, Index, DEFAULT_VALUES_PER_FACET};
 use serde::{Deserialize, Serialize, Serializer};
@@ -186,6 +187,9 @@ pub struct Settings<T> {
     #[deserr(default, error = DeserrJsonError<InvalidSettingsDistinctAttribute>)]
     pub distinct_attribute: Setting<String>,
     #[serde(default, skip_serializing_if = "Setting::is_not_set")]
+    #[deserr(default, error = DeserrJsonError<InvalidSettingsProximityPrecision>)]
+    pub proximity_precision: Setting<ProximityPrecisionView>,
+    #[serde(default, skip_serializing_if = "Setting::is_not_set")]
     #[deserr(default, error = DeserrJsonError<InvalidSettingsTypoTolerance>)]
     pub typo_tolerance: Setting<TypoSettings>,
     #[serde(default, skip_serializing_if = "Setting::is_not_set")]
@@ -214,6 +218,7 @@ impl Settings<Checked> {
             separator_tokens: Setting::Reset,
             dictionary: Setting::Reset,
             distinct_attribute: Setting::Reset,
+            proximity_precision: Setting::Reset,
             typo_tolerance: Setting::Reset,
             faceting: Setting::Reset,
             pagination: Setting::Reset,
@@ -234,6 +239,7 @@ impl Settings<Checked> {
             dictionary,
             synonyms,
             distinct_attribute,
+            proximity_precision,
             typo_tolerance,
             faceting,
             pagination,
@@ -252,6 +258,7 @@ impl Settings<Checked> {
             dictionary,
             synonyms,
             distinct_attribute,
+            proximity_precision,
             typo_tolerance,
             faceting,
             pagination,
@@ -296,6 +303,7 @@ impl Settings<Unchecked> {
             separator_tokens: self.separator_tokens,
             dictionary: self.dictionary,
             distinct_attribute: self.distinct_attribute,
+            proximity_precision: self.proximity_precision,
             typo_tolerance: self.typo_tolerance,
             faceting: self.faceting,
             pagination: self.pagination,
@@ -390,6 +398,12 @@ pub fn apply_settings_to_builder(
         Setting::NotSet => (),
     }
 
+    match settings.proximity_precision {
+        Setting::Set(ref precision) => builder.set_proximity_precision((*precision).into()),
+        Setting::Reset => builder.reset_proximity_precision(),
+        Setting::NotSet => (),
+    }
+
     match settings.typo_tolerance {
         Setting::Set(ref value) => {
             match value.enabled {
@@ -509,6 +523,8 @@ pub fn settings(
 
     let distinct_field = index.distinct_field(rtxn)?.map(String::from);
 
+    let proximity_precision = index.proximity_precision(rtxn)?.map(ProximityPrecisionView::from);
+
     let synonyms = index.user_defined_synonyms(rtxn)?;
 
     let min_typo_word_len = MinWordSizeTyposSetting {
@@ -532,7 +548,10 @@ pub fn settings(
 
     let faceting = FacetingSettings {
         max_values_per_facet: Setting::Set(
-            index.max_values_per_facet(rtxn)?.unwrap_or(DEFAULT_VALUES_PER_FACET),
+            index
+                .max_values_per_facet(rtxn)?
+                .map(|x| x as usize)
+                .unwrap_or(DEFAULT_VALUES_PER_FACET),
         ),
         sort_facet_values_by: Setting::Set(
             index
@@ -545,7 +564,10 @@ pub fn settings(
 
     let pagination = PaginationSettings {
         max_total_hits: Setting::Set(
-            index.pagination_max_total_hits(rtxn)?.unwrap_or(DEFAULT_PAGINATION_MAX_TOTAL_HITS),
+            index
+                .pagination_max_total_hits(rtxn)?
+                .map(|x| x as usize)
+                .unwrap_or(DEFAULT_PAGINATION_MAX_TOTAL_HITS),
         ),
     };
 
@@ -569,6 +591,10 @@ pub fn settings(
             Some(field) => Setting::Set(field),
             None => Setting::Reset,
         },
+        proximity_precision: match proximity_precision {
+            Some(precision) => Setting::Set(precision),
+            None => Setting::Reset,
+        },
         synonyms: Setting::Set(synonyms),
         typo_tolerance: Setting::Set(typo_tolerance),
         faceting: Setting::Set(faceting),
@@ -673,6 +699,31 @@ impl From<RankingRuleView> for Criterion {
     }
 }
 
+#[derive(Debug, Clone, Copy, PartialEq, Eq, Deserr, Serialize, Deserialize)]
+#[serde(deny_unknown_fields, rename_all = "camelCase")]
+#[deserr(error = DeserrJsonError<InvalidSettingsProximityPrecision>, rename_all = camelCase, deny_unknown_fields)]
+pub enum ProximityPrecisionView {
+    WordScale,
+    AttributeScale,
+}
+
+impl From<ProximityPrecision> for ProximityPrecisionView {
+    fn from(value: ProximityPrecision) -> Self {
+        match value {
+            ProximityPrecision::WordScale => ProximityPrecisionView::WordScale,
+            ProximityPrecision::AttributeScale => ProximityPrecisionView::AttributeScale,
+        }
+    }
+}
+impl From<ProximityPrecisionView> for ProximityPrecision {
+    fn from(value: ProximityPrecisionView) -> Self {
+        match value {
+            ProximityPrecisionView::WordScale => ProximityPrecision::WordScale,
+            ProximityPrecisionView::AttributeScale => ProximityPrecision::AttributeScale,
+        }
+    }
+}
+
 #[cfg(test)]
 pub(crate) mod test {
     use super::*;
@@ -692,6 +743,7 @@ pub(crate) mod test {
             dictionary: Setting::NotSet,
             synonyms: Setting::NotSet,
             distinct_attribute: Setting::NotSet,
+            proximity_precision: Setting::NotSet,
             typo_tolerance: Setting::NotSet,
             faceting: Setting::NotSet,
             pagination: Setting::NotSet,
@@ -716,6 +768,7 @@ pub(crate) mod test {
             dictionary: Setting::NotSet,
             synonyms: Setting::NotSet,
             distinct_attribute: Setting::NotSet,
+            proximity_precision: Setting::NotSet,
             typo_tolerance: Setting::NotSet,
             faceting: Setting::NotSet,
             pagination: Setting::NotSet,

meilisearch/Cargo.toml

@@ -39,7 +39,7 @@ byte-unit = { version = "4.0.19", default-features = false, features = [
 bytes = "1.4.0"
 clap = { version = "4.2.1", features = ["derive", "env"] }
 crossbeam-channel = "0.5.8"
-deserr = { version = "0.6.0", features = ["actix-web"]}
+deserr = { version = "0.6.0", features = ["actix-web"] }
 dump = { path = "../dump" }
 either = "1.8.1"
 env_logger = "0.10.0"
@@ -150,6 +150,7 @@ hebrew = ["meilisearch-types/hebrew"]
 japanese = ["meilisearch-types/japanese"]
 thai = ["meilisearch-types/thai"]
 greek = ["meilisearch-types/greek"]
+khmer = ["meilisearch-types/khmer"]
 
 [package.metadata.mini-dashboard]
 assets-url = "https://github.com/meilisearch/mini-dashboard/releases/download/v0.2.11/build.zip"

meilisearch/src/lib.rs

@@ -397,6 +397,7 @@ pub fn configure_data(
         .app_data(web::Data::from(analytics))
         .app_data(
             web::JsonConfig::default()
+                .limit(http_payload_size_limit)
                 .content_type(|mime| mime == mime::APPLICATION_JSON)
                 .error_handler(|err, req: &HttpRequest| match err {
                     JsonPayloadError::ContentType => match req.headers().get(CONTENT_TYPE) {

meilisearch/src/routes/features.rs

@@ -48,6 +48,8 @@ pub struct RuntimeTogglableFeatures {
     pub metrics: Option<bool>,
     #[deserr(default)]
     pub export_puffin_reports: Option<bool>,
+    #[deserr(default)]
+    pub proximity_precision: Option<bool>,
 }
 
 async fn patch_features(
@@ -70,6 +72,10 @@ async fn patch_features(
             .0
             .export_puffin_reports
             .unwrap_or(old_features.export_puffin_reports),
+        proximity_precision: new_features
+            .0
+            .proximity_precision
+            .unwrap_or(old_features.proximity_precision),
     };
 
     // explicitly destructure for analytics rather than using the `Serialize` implementation, because
@@ -80,6 +86,7 @@ async fn patch_features(
         vector_store,
         metrics,
         export_puffin_reports,
+        proximity_precision,
     } = new_features;
 
     analytics.publish(
@@ -89,6 +96,7 @@ async fn patch_features(
             "vector_store": vector_store,
             "metrics": metrics,
             "export_puffin_reports": export_puffin_reports,
+            "proximity_precision": proximity_precision,
         }),
         Some(&req),
     );

meilisearch/src/routes/indexes/documents.rs

@@ -3,7 +3,7 @@ use std::io::ErrorKind;
 use actix_web::http::header::CONTENT_TYPE;
 use actix_web::web::Data;
 use actix_web::{web, HttpMessage, HttpRequest, HttpResponse};
-use bstr::ByteSlice;
+use bstr::ByteSlice as _;
 use deserr::actix_web::{AwebJson, AwebQueryParameter};
 use deserr::Deserr;
 use futures::StreamExt;

meilisearch/src/routes/indexes/settings.rs

@@ -78,6 +78,7 @@ macro_rules! make_setting_route {
 
                 let body = body.into_inner();
 
+                #[allow(clippy::redundant_closure_call)]
                 $analytics(&body, &req);
 
                 let new_settings = Settings {
@@ -434,6 +435,30 @@ make_setting_route!(
     }
 );
 
+make_setting_route!(
+    "/proximity-precision",
+    put,
+    meilisearch_types::settings::ProximityPrecisionView,
+    meilisearch_types::deserr::DeserrJsonError<
+        meilisearch_types::error::deserr_codes::InvalidSettingsProximityPrecision,
+    >,
+    proximity_precision,
+    "proximityPrecision",
+    analytics,
+    |precision: &Option<meilisearch_types::settings::ProximityPrecisionView>, req: &HttpRequest| {
+        use serde_json::json;
+        analytics.publish(
+            "ProximityPrecision Updated".to_string(),
+            json!({
+                "proximity_precision": {
+                    "set": precision.is_some(),
+                }
+            }),
+            Some(req),
+        );
+    }
+);
+
 make_setting_route!(
     "/ranking-rules",
     put,
@@ -540,6 +565,7 @@ generate_configure!(
     displayed_attributes,
     searchable_attributes,
     distinct_attribute,
+    proximity_precision,
     stop_words,
     separator_tokens,
     non_separator_tokens,
@@ -593,6 +619,9 @@ pub async fn update_all(
             "distinct_attribute": {
                 "set": new_settings.distinct_attribute.as_ref().set().is_some()
             },
+            "proximity_precision": {
+                "set": new_settings.proximity_precision.as_ref().set().is_some()
+            },
             "typo_tolerance": {
                 "enabled": new_settings.typo_tolerance
                     .as_ref()

meilisearch/src/routes/multi_search.rs

@@ -46,49 +46,46 @@ pub async fn multi_search_with_post(
     // Explicitly expect a `(ResponseError, usize)` for the error type rather than `ResponseError` only,
     // so that `?` doesn't work if it doesn't use `with_index`, ensuring that it is not forgotten in case of code
     // changes.
-    let search_results: Result<_, (ResponseError, usize)> = (|| {
-        async {
-            let mut search_results = Vec::with_capacity(queries.len());
-            for (query_index, (index_uid, mut query)) in
-                queries.into_iter().map(SearchQueryWithIndex::into_index_query).enumerate()
+    let search_results: Result<_, (ResponseError, usize)> = async {
+        let mut search_results = Vec::with_capacity(queries.len());
+        for (query_index, (index_uid, mut query)) in
+            queries.into_iter().map(SearchQueryWithIndex::into_index_query).enumerate()
+        {
+            debug!("multi-search #{query_index}: called with params: {:?}", query);
+
+            // Check index from API key
+            if !index_scheduler.filters().is_index_authorized(&index_uid) {
+                return Err(AuthenticationError::InvalidToken).with_index(query_index);
+            }
+            // Apply search rules from tenant token
+            if let Some(search_rules) = index_scheduler.filters().get_index_search_rules(&index_uid)
             {
-                debug!("multi-search #{query_index}: called with params: {:?}", query);
+                add_search_rules(&mut query, search_rules);
+            }
 
-                // Check index from API key
-                if !index_scheduler.filters().is_index_authorized(&index_uid) {
-                    return Err(AuthenticationError::InvalidToken).with_index(query_index);
-                }
-                // Apply search rules from tenant token
-                if let Some(search_rules) =
-                    index_scheduler.filters().get_index_search_rules(&index_uid)
-                {
-                    add_search_rules(&mut query, search_rules);
-                }
+            let index = index_scheduler
+                .index(&index_uid)
+                .map_err(|err| {
+                    let mut err = ResponseError::from(err);
+                    // Patch the HTTP status code to 400 as it defaults to 404 for `index_not_found`, but
+                    // here the resource not found is not part of the URL.
+                    err.code = StatusCode::BAD_REQUEST;
+                    err
+                })
+                .with_index(query_index)?;
 
-                let index = index_scheduler
-                    .index(&index_uid)
-                    .map_err(|err| {
-                        let mut err = ResponseError::from(err);
-                        // Patch the HTTP status code to 400 as it defaults to 404 for `index_not_found`, but
-                        // here the resource not found is not part of the URL.
-                        err.code = StatusCode::BAD_REQUEST;
-                        err
-                    })
+            let search_result =
+                tokio::task::spawn_blocking(move || perform_search(&index, query, features))
+                    .await
                     .with_index(query_index)?;
 
-                let search_result =
-                    tokio::task::spawn_blocking(move || perform_search(&index, query, features))
-                        .await
-                        .with_index(query_index)?;
-
-                search_results.push(SearchResultWithIndex {
-                    index_uid: index_uid.into_inner(),
-                    result: search_result.with_index(query_index)?,
-                });
-            }
-            Ok(search_results)
+            search_results.push(SearchResultWithIndex {
+                index_uid: index_uid.into_inner(),
+                result: search_result.with_index(query_index)?,
+            });
         }
-    })()
+        Ok(search_results)
+    }
     .await;
 
     if search_results.is_ok() {

meilisearch/src/search.rs

@@ -360,6 +360,7 @@ fn prepare_search<'t>(
     let max_total_hits = index
         .pagination_max_total_hits(rtxn)
         .map_err(milli::Error::from)?
+        .map(|x| x as usize)
         .unwrap_or(DEFAULT_PAGINATION_MAX_TOTAL_HITS);
 
     search.exhaustive_number_hits(is_finite_pagination);
@@ -586,6 +587,7 @@ pub fn perform_search(
             let max_values_by_facet = index
                 .max_values_per_facet(&rtxn)
                 .map_err(milli::Error::from)?
+                .map(|x| x as usize)
                 .unwrap_or(DEFAULT_VALUES_PER_FACET);
             facet_distribution.max_values_per_facet(max_values_by_facet);
 

meilisearch/tests/common/mod.rs

@@ -5,9 +5,11 @@ pub mod service;
 
 use std::fmt::{self, Display};
 
+#[allow(unused)]
 pub use index::{GetAllDocumentsOptions, GetDocumentOptions};
 use meili_snap::json_string;
 use serde::{Deserialize, Serialize};
+#[allow(unused)]
 pub use server::{default_settings, Server};
 
 #[derive(Debug, Clone, Default, Serialize, Deserialize, PartialEq, Eq)]

meilisearch/tests/dumps/data.rs

@@ -20,6 +20,8 @@ pub enum GetDump {
     RubyGemsWithSettingsV4,
 
     TestV5,
+
+    TestV6WithExperimental,
 }
 
 impl GetDump {
@@ -68,6 +70,10 @@ impl GetDump {
             GetDump::TestV5 => {
                 exist_relative_path!("tests/assets/v5_v0.28.0_test_dump.dump").into()
             }
+            GetDump::TestV6WithExperimental => exist_relative_path!(
+                "tests/assets/v6_v1.6.0_use_deactivated_experimental_setting.dump"
+            )
+            .into(),
         }
     }
 }

meilisearch/tests/dumps/mod.rs

@@ -59,6 +59,7 @@ async fn import_dump_v1_movie_raw() {
       "dictionary": [],
       "synonyms": {},
       "distinctAttribute": null,
+      "proximityPrecision": null,
       "typoTolerance": {
         "enabled": true,
         "minWordSizeForTypos": {
@@ -219,6 +220,7 @@ async fn import_dump_v1_movie_with_settings() {
       "dictionary": [],
       "synonyms": {},
       "distinctAttribute": null,
+      "proximityPrecision": null,
       "typoTolerance": {
         "enabled": true,
         "minWordSizeForTypos": {
@@ -365,6 +367,7 @@ async fn import_dump_v1_rubygems_with_settings() {
       "dictionary": [],
       "synonyms": {},
       "distinctAttribute": null,
+      "proximityPrecision": null,
       "typoTolerance": {
         "enabled": true,
         "minWordSizeForTypos": {
@@ -497,6 +500,7 @@ async fn import_dump_v2_movie_raw() {
       "dictionary": [],
       "synonyms": {},
       "distinctAttribute": null,
+      "proximityPrecision": null,
       "typoTolerance": {
         "enabled": true,
         "minWordSizeForTypos": {
@@ -641,6 +645,7 @@ async fn import_dump_v2_movie_with_settings() {
       "dictionary": [],
       "synonyms": {},
       "distinctAttribute": null,
+      "proximityPrecision": null,
       "typoTolerance": {
         "enabled": true,
         "minWordSizeForTypos": {
@@ -784,6 +789,7 @@ async fn import_dump_v2_rubygems_with_settings() {
       "dictionary": [],
       "synonyms": {},
       "distinctAttribute": null,
+      "proximityPrecision": null,
       "typoTolerance": {
         "enabled": true,
         "minWordSizeForTypos": {
@@ -916,6 +922,7 @@ async fn import_dump_v3_movie_raw() {
       "dictionary": [],
       "synonyms": {},
       "distinctAttribute": null,
+      "proximityPrecision": null,
       "typoTolerance": {
         "enabled": true,
         "minWordSizeForTypos": {
@@ -1060,6 +1067,7 @@ async fn import_dump_v3_movie_with_settings() {
       "dictionary": [],
       "synonyms": {},
       "distinctAttribute": null,
+      "proximityPrecision": null,
       "typoTolerance": {
         "enabled": true,
         "minWordSizeForTypos": {
@@ -1203,6 +1211,7 @@ async fn import_dump_v3_rubygems_with_settings() {
       "dictionary": [],
       "synonyms": {},
       "distinctAttribute": null,
+      "proximityPrecision": null,
       "typoTolerance": {
         "enabled": true,
         "minWordSizeForTypos": {
@@ -1335,6 +1344,7 @@ async fn import_dump_v4_movie_raw() {
       "dictionary": [],
       "synonyms": {},
       "distinctAttribute": null,
+      "proximityPrecision": null,
       "typoTolerance": {
         "enabled": true,
         "minWordSizeForTypos": {
@@ -1479,6 +1489,7 @@ async fn import_dump_v4_movie_with_settings() {
       "dictionary": [],
       "synonyms": {},
       "distinctAttribute": null,
+      "proximityPrecision": null,
       "typoTolerance": {
         "enabled": true,
         "minWordSizeForTypos": {
@@ -1622,6 +1633,7 @@ async fn import_dump_v4_rubygems_with_settings() {
       "dictionary": [],
       "synonyms": {},
       "distinctAttribute": null,
+      "proximityPrecision": null,
       "typoTolerance": {
         "enabled": true,
         "minWordSizeForTypos": {
@@ -1810,3 +1822,108 @@ async fn import_dump_v5() {
         json_string!(tasks, { ".results[].details.dumpUid" => "[uid]",  ".results[].duration" => "[duration]" ,  ".results[].startedAt" => "[date]" ,  ".results[].finishedAt" => "[date]"  })
     );
 }
+
+#[actix_rt::test]
+async fn import_dump_v6_containing_experimental_features() {
+    let temp = tempfile::tempdir().unwrap();
+
+    let options = Opt {
+        import_dump: Some(GetDump::TestV6WithExperimental.path()),
+        ..default_settings(temp.path())
+    };
+    let mut server = Server::new_auth_with_options(options, temp).await;
+    server.use_api_key("MASTER_KEY");
+
+    let (indexes, code) = server.list_indexes(None, None).await;
+    assert_eq!(code, 200, "{indexes}");
+
+    assert_eq!(indexes["results"].as_array().unwrap().len(), 1);
+    assert_eq!(indexes["results"][0]["uid"], json!("movies"));
+    assert_eq!(indexes["results"][0]["primaryKey"], json!("id"));
+
+    let (response, code) = server.get_features().await;
+    meili_snap::snapshot!(code, @"200 OK");
+    meili_snap::snapshot!(meili_snap::json_string!(response), @r###"
+    {
+      "scoreDetails": false,
+      "vectorStore": false,
+      "metrics": false,
+      "exportPuffinReports": false,
+      "proximityPrecision": false
+    }
+    "###);
+
+    let index = server.index("movies");
+
+    let (response, code) = index.settings().await;
+    meili_snap::snapshot!(code, @"200 OK");
+    meili_snap::snapshot!(meili_snap::json_string!(response), @r###"
+    {
+      "displayedAttributes": [
+        "*"
+      ],
+      "searchableAttributes": [
+        "*"
+      ],
+      "filterableAttributes": [],
+      "sortableAttributes": [],
+      "rankingRules": [
+        "words",
+        "typo",
+        "proximity"
+      ],
+      "stopWords": [],
+      "nonSeparatorTokens": [],
+      "separatorTokens": [],
+      "dictionary": [],
+      "synonyms": {},
+      "distinctAttribute": null,
+      "proximityPrecision": "attributeScale",
+      "typoTolerance": {
+        "enabled": true,
+        "minWordSizeForTypos": {
+          "oneTypo": 5,
+          "twoTypos": 9
+        },
+        "disableOnWords": [],
+        "disableOnAttributes": []
+      },
+      "faceting": {
+        "maxValuesPerFacet": 100,
+        "sortFacetValuesBy": {
+          "*": "alpha"
+        }
+      },
+      "pagination": {
+        "maxTotalHits": 1000
+      }
+    }
+    "###);
+
+    // the expected order is [1, 3, 2] instead of [3, 1, 2]
+    // because the attribute scale doesn't make the difference between 1 and 3.
+    index
+        .search(json!({"q": "the soup of day"}), |response, code| {
+            snapshot!(code, @"200 OK");
+            snapshot!(json_string!(response["hits"]), @r###"
+            [
+              {
+                "id": 1,
+                "a": "Soup of the day",
+                "b": "many the fish"
+              },
+              {
+                "id": 3,
+                "a": "the Soup of day",
+                "b": "many the fish"
+              },
+              {
+                "id": 2,
+                "a": "Soup of day",
+                "b": "many the lazy fish"
+              }
+            ]
+            "###);
+        })
+        .await;
+}

meilisearch/tests/features/mod.rs

@@ -21,7 +21,8 @@ async fn experimental_features() {
       "scoreDetails": false,
       "vectorStore": false,
       "metrics": false,
-      "exportPuffinReports": false
+      "exportPuffinReports": false,
+      "proximityPrecision": false
     }
     "###);
 
@@ -33,7 +34,8 @@ async fn experimental_features() {
       "scoreDetails": false,
       "vectorStore": true,
       "metrics": false,
-      "exportPuffinReports": false
+      "exportPuffinReports": false,
+      "proximityPrecision": false
     }
     "###);
 
@@ -45,7 +47,8 @@ async fn experimental_features() {
       "scoreDetails": false,
       "vectorStore": true,
       "metrics": false,
-      "exportPuffinReports": false
+      "exportPuffinReports": false,
+      "proximityPrecision": false
     }
     "###);
 
@@ -58,7 +61,8 @@ async fn experimental_features() {
       "scoreDetails": false,
       "vectorStore": true,
       "metrics": false,
-      "exportPuffinReports": false
+      "exportPuffinReports": false,
+      "proximityPrecision": false
     }
     "###);
 
@@ -71,7 +75,8 @@ async fn experimental_features() {
       "scoreDetails": false,
       "vectorStore": true,
       "metrics": false,
-      "exportPuffinReports": false
+      "exportPuffinReports": false,
+      "proximityPrecision": false
     }
     "###);
 }
@@ -91,7 +96,8 @@ async fn experimental_feature_metrics() {
       "scoreDetails": false,
       "vectorStore": false,
       "metrics": true,
-      "exportPuffinReports": false
+      "exportPuffinReports": false,
+      "proximityPrecision": false
     }
     "###);
 
@@ -146,7 +152,7 @@ async fn errors() {
     meili_snap::snapshot!(code, @"400 Bad Request");
     meili_snap::snapshot!(meili_snap::json_string!(response), @r###"
     {
-      "message": "Unknown field `NotAFeature`: expected one of `scoreDetails`, `vectorStore`, `metrics`, `exportPuffinReports`",
+      "message": "Unknown field `NotAFeature`: expected one of `scoreDetails`, `vectorStore`, `metrics`, `exportPuffinReports`, `proximityPrecision`",
       "code": "bad_request",
       "type": "invalid_request",
       "link": "https://docs.meilisearch.com/errors#bad_request"

meilisearch/tests/search/distinct.rs

@@ -4,23 +4,111 @@ use once_cell::sync::Lazy;
 use crate::common::{Server, Value};
 use crate::json;
 
-pub(self) static DOCUMENTS: Lazy<Value> = Lazy::new(|| {
+static DOCUMENTS: Lazy<Value> = Lazy::new(|| {
     json!([
-        {"productId": 1, "shopId": 1},
-        {"productId": 2, "shopId": 1},
-        {"productId": 3, "shopId": 2},
-        {"productId": 4, "shopId": 2},
-        {"productId": 5, "shopId": 3},
-        {"productId": 6, "shopId": 3},
-        {"productId": 7, "shopId": 4},
-        {"productId": 8, "shopId": 4},
-        {"productId": 9, "shopId": 5},
-        {"productId": 10, "shopId": 5}
+      {
+        "id": 1,
+        "description": "Leather Jacket",
+        "brand": "Lee Jeans",
+        "product_id": "123456",
+        "color": "Brown"
+      },
+      {
+        "id": 2,
+        "description": "Leather Jacket",
+        "brand": "Lee Jeans",
+        "product_id": "123456",
+        "color": "Black"
+      },
+      {
+        "id": 3,
+        "description": "Leather Jacket",
+        "brand": "Lee Jeans",
+        "product_id": "123456",
+        "color": "Blue"
+      },
+      {
+        "id": 4,
+        "description": "T-Shirt",
+        "brand": "Nike",
+        "product_id": "789012",
+        "color": "Red"
+      },
+      {
+        "id": 5,
+        "description": "T-Shirt",
+        "brand": "Nike",
+        "product_id": "789012",
+        "color": "Blue"
+      },
+      {
+        "id": 6,
+        "description": "Running Shoes",
+        "brand": "Adidas",
+        "product_id": "456789",
+        "color": "Black"
+      },
+      {
+        "id": 7,
+        "description": "Running Shoes",
+        "brand": "Adidas",
+        "product_id": "456789",
+        "color": "White"
+      },
+      {
+        "id": 8,
+        "description": "Hoodie",
+        "brand": "Puma",
+        "product_id": "987654",
+        "color": "Gray"
+      },
+      {
+        "id": 9,
+        "description": "Sweater",
+        "brand": "Gap",
+        "product_id": "234567",
+        "color": "Green"
+      },
+      {
+        "id": 10,
+        "description": "Sweater",
+        "brand": "Gap",
+        "product_id": "234567",
+        "color": "Red"
+      },
+      {
+        "id": 11,
+        "description": "Sweater",
+        "brand": "Gap",
+        "product_id": "234567",
+        "color": "Blue"
+      },
+      {
+        "id": 12,
+        "description": "Jeans",
+        "brand": "Levi's",
+        "product_id": "345678",
+        "color": "Indigo"
+      },
+      {
+        "id": 13,
+        "description": "Jeans",
+        "brand": "Levi's",
+        "product_id": "345678",
+        "color": "Black"
+      },
+      {
+        "id": 14,
+        "description": "Jeans",
+        "brand": "Levi's",
+        "product_id": "345678",
+        "color": "Stone Wash"
+      }
     ])
 });
 
-pub(self) static DOCUMENT_PRIMARY_KEY: &str = "productId";
-pub(self) static DOCUMENT_DISTINCT_KEY: &str = "shopId";
+static DOCUMENT_PRIMARY_KEY: &str = "id";
+static DOCUMENT_DISTINCT_KEY: &str = "product_id";
 
 /// testing: https://github.com/meilisearch/meilisearch/issues/4078
 #[actix_rt::test]
@@ -33,31 +121,121 @@ async fn distinct_search_with_offset_no_ranking() {
     index.update_distinct_attribute(json!(DOCUMENT_DISTINCT_KEY)).await;
     index.wait_task(1).await;
 
-    fn get_hits(Value(response): Value) -> Vec<i64> {
+    fn get_hits(response: &Value) -> Vec<&str> {
         let hits_array = response["hits"].as_array().unwrap();
-        hits_array.iter().map(|h| h[DOCUMENT_DISTINCT_KEY].as_i64().unwrap()).collect::<Vec<_>>()
+        hits_array.iter().map(|h| h[DOCUMENT_DISTINCT_KEY].as_str().unwrap()).collect::<Vec<_>>()
     }
 
-    let (response, code) = index.search_post(json!({"limit": 2, "offset": 0})).await;
-    let hits = get_hits(response);
+    let (response, code) = index.search_post(json!({"offset": 0, "limit": 2})).await;
+    let hits = get_hits(&response);
     snapshot!(code, @"200 OK");
     snapshot!(hits.len(), @"2");
-    snapshot!(format!("{:?}", hits), @"[1, 2]");
+    snapshot!(format!("{:?}", hits), @r#"["123456", "789012"]"#);
+    snapshot!(response["estimatedTotalHits"] , @"11");
 
-    let (response, code) = index.search_post(json!({"limit": 2, "offset": 2})).await;
-    let hits = get_hits(response);
+    let (response, code) = index.search_post(json!({"offset": 2, "limit": 2})).await;
+    let hits = get_hits(&response);
     snapshot!(code, @"200 OK");
     snapshot!(hits.len(), @"2");
-    snapshot!(format!("{:?}", hits), @"[3, 4]");
+    snapshot!(format!("{:?}", hits), @r#"["456789", "987654"]"#);
+    snapshot!(response["estimatedTotalHits"], @"10");
 
-    let (response, code) = index.search_post(json!({"limit": 10, "offset": 4})).await;
-    let hits = get_hits(response);
+    let (response, code) = index.search_post(json!({"offset": 4, "limit": 2})).await;
+    let hits = get_hits(&response);
+    snapshot!(code, @"200 OK");
+    snapshot!(hits.len(), @"2");
+    snapshot!(format!("{:?}", hits), @r#"["234567", "345678"]"#);
+    snapshot!(response["estimatedTotalHits"], @"6");
+
+    let (response, code) = index.search_post(json!({"offset": 5, "limit": 2})).await;
+    let hits = get_hits(&response);
     snapshot!(code, @"200 OK");
     snapshot!(hits.len(), @"1");
-    snapshot!(format!("{:?}", hits), @"[5]");
+    snapshot!(format!("{:?}", hits), @r#"["345678"]"#);
+    snapshot!(response["estimatedTotalHits"], @"6");
 
-    let (response, code) = index.search_post(json!({"limit": 10, "offset": 5})).await;
-    let hits = get_hits(response);
+    let (response, code) = index.search_post(json!({"offset": 6, "limit": 2})).await;
+    let hits = get_hits(&response);
     snapshot!(code, @"200 OK");
     snapshot!(hits.len(), @"0");
+    snapshot!(format!("{:?}", hits), @r#"[]"#);
+    snapshot!(response["estimatedTotalHits"], @"6");
+
+    let (response, code) = index.search_post(json!({"offset": 7, "limit": 2})).await;
+    let hits = get_hits(&response);
+    snapshot!(code, @"200 OK");
+    snapshot!(hits.len(), @"0");
+    snapshot!(format!("{:?}", hits), @r#"[]"#);
+    snapshot!(response["estimatedTotalHits"], @"6");
+}
+
+/// testing: https://github.com/meilisearch/meilisearch/issues/4130
+#[actix_rt::test]
+async fn distinct_search_with_pagination_no_ranking() {
+    let server = Server::new().await;
+    let index = server.index("test");
+
+    let documents = DOCUMENTS.clone();
+    index.add_documents(documents, Some(DOCUMENT_PRIMARY_KEY)).await;
+    index.update_distinct_attribute(json!(DOCUMENT_DISTINCT_KEY)).await;
+    index.wait_task(1).await;
+
+    fn get_hits(response: &Value) -> Vec<&str> {
+        let hits_array = response["hits"].as_array().unwrap();
+        hits_array.iter().map(|h| h[DOCUMENT_DISTINCT_KEY].as_str().unwrap()).collect::<Vec<_>>()
+    }
+
+    let (response, code) = index.search_post(json!({"page": 0, "hitsPerPage": 2})).await;
+    let hits = get_hits(&response);
+    snapshot!(code, @"200 OK");
+    snapshot!(hits.len(), @"0");
+    snapshot!(format!("{:?}", hits), @r#"[]"#);
+    snapshot!(response["page"], @"0");
+    snapshot!(response["totalPages"], @"3");
+    snapshot!(response["totalHits"], @"6");
+
+    let (response, code) = index.search_post(json!({"page": 1, "hitsPerPage": 2})).await;
+    let hits = get_hits(&response);
+    snapshot!(code, @"200 OK");
+    snapshot!(hits.len(), @"2");
+    snapshot!(format!("{:?}", hits), @r#"["123456", "789012"]"#);
+    snapshot!(response["page"], @"1");
+    snapshot!(response["totalPages"], @"3");
+    snapshot!(response["totalHits"], @"6");
+
+    let (response, code) = index.search_post(json!({"page": 2, "hitsPerPage": 2})).await;
+    let hits = get_hits(&response);
+    snapshot!(code, @"200 OK");
+    snapshot!(hits.len(), @"2");
+    snapshot!(format!("{:?}", hits), @r#"["456789", "987654"]"#);
+    snapshot!(response["page"], @"2");
+    snapshot!(response["totalPages"], @"3");
+    snapshot!(response["totalHits"], @"6");
+
+    let (response, code) = index.search_post(json!({"page": 3, "hitsPerPage": 2})).await;
+    let hits = get_hits(&response);
+    snapshot!(code, @"200 OK");
+    snapshot!(hits.len(), @"2");
+    snapshot!(format!("{:?}", hits), @r#"["234567", "345678"]"#);
+    snapshot!(response["page"], @"3");
+    snapshot!(response["totalPages"], @"3");
+    snapshot!(response["totalHits"], @"6");
+
+    let (response, code) = index.search_post(json!({"page": 4, "hitsPerPage": 2})).await;
+    let hits = get_hits(&response);
+    snapshot!(code, @"200 OK");
+    snapshot!(hits.len(), @"0");
+    snapshot!(format!("{:?}", hits), @r#"[]"#);
+    snapshot!(response["page"], @"4");
+    snapshot!(response["totalPages"], @"3");
+    snapshot!(response["totalHits"], @"6");
+
+    let (response, code) = index.search_post(json!({"page": 2, "hitsPerPage": 3})).await;
+    let hits = get_hits(&response);
+    snapshot!(code, @"200 OK");
+    snapshot!(hits.len(), @"3");
+    snapshot!(format!("{:?}", hits), @r#"["987654", "234567", "345678"]"#);
+    snapshot!(response["page"], @"2");
+    snapshot!(response["totalPages"], @"2");
+    snapshot!(response["totalHits"], @"6");
 }

meilisearch/tests/search/facet_search.rs

@@ -4,7 +4,7 @@ use once_cell::sync::Lazy;
 use crate::common::{Server, Value};
 use crate::json;
 
-pub(self) static DOCUMENTS: Lazy<Value> = Lazy::new(|| {
+static DOCUMENTS: Lazy<Value> = Lazy::new(|| {
     json!([
         {
             "title": "Shazam!",

meilisearch/tests/search/geo.rs

@@ -4,7 +4,7 @@ use once_cell::sync::Lazy;
 use crate::common::{Server, Value};
 use crate::json;
 
-pub(self) static DOCUMENTS: Lazy<Value> = Lazy::new(|| {
+static DOCUMENTS: Lazy<Value> = Lazy::new(|| {
     json!([
         {
             "id": 1,

meilisearch/tests/search/mod.rs

@@ -15,7 +15,7 @@ use once_cell::sync::Lazy;
 use crate::common::{Server, Value};
 use crate::json;
 
-pub(self) static DOCUMENTS: Lazy<Value> = Lazy::new(|| {
+static DOCUMENTS: Lazy<Value> = Lazy::new(|| {
     json!([
         {
             "title": "Shazam!",
@@ -40,7 +40,7 @@ pub(self) static DOCUMENTS: Lazy<Value> = Lazy::new(|| {
     ])
 });
 
-pub(self) static NESTED_DOCUMENTS: Lazy<Value> = Lazy::new(|| {
+static NESTED_DOCUMENTS: Lazy<Value> = Lazy::new(|| {
     json!([
         {
             "id": 852,

meilisearch/tests/settings/get_settings.rs

@@ -54,7 +54,7 @@ async fn get_settings() {
     let (response, code) = index.settings().await;
     assert_eq!(code, 200);
     let settings = response.as_object().unwrap();
-    assert_eq!(settings.keys().len(), 14);
+    assert_eq!(settings.keys().len(), 15);
     assert_eq!(settings["displayedAttributes"], json!(["*"]));
     assert_eq!(settings["searchableAttributes"], json!(["*"]));
     assert_eq!(settings["filterableAttributes"], json!([]));

meilisearch/tests/settings/mod.rs

@@ -1,4 +1,5 @@
 mod distinct;
 mod errors;
 mod get_settings;
+mod proximity_settings;
 mod tokenizer_customization;

meilisearch/tests/settings/proximity_settings.rs

@@ -0,0 +1,396 @@
+use meili_snap::{json_string, snapshot};
+use once_cell::sync::Lazy;
+
+use crate::common::Server;
+use crate::json;
+
+static DOCUMENTS: Lazy<crate::common::Value> = Lazy::new(|| {
+    json!([
+        {
+            "id": 1,
+            "a": "Soup of the day",
+            "b": "many the fish",
+        },
+        {
+            "id": 2,
+            "a": "Soup of day",
+            "b": "many the lazy fish",
+        },
+        {
+            "id": 3,
+            "a": "the Soup of day",
+            "b": "many the fish",
+        },
+    ])
+});
+
+#[actix_rt::test]
+async fn attribute_scale_search() {
+    let server = Server::new().await;
+    let (response, code) = server.set_features(json!({"proximityPrecision": true})).await;
+    meili_snap::snapshot!(code, @"200 OK");
+    meili_snap::snapshot!(meili_snap::json_string!(response), @r###"
+    {
+      "scoreDetails": false,
+      "vectorStore": false,
+      "metrics": false,
+      "exportPuffinReports": false,
+      "proximityPrecision": true
+    }
+    "###);
+    let index = server.index("test");
+
+    index.add_documents(DOCUMENTS.clone(), None).await;
+    index.wait_task(0).await;
+
+    let (response, code) = index
+        .update_settings(json!({
+            "proximityPrecision": "attributeScale",
+            "rankingRules": ["words", "typo", "proximity"],
+        }))
+        .await;
+    assert_eq!("202", code.as_str(), "{:?}", response);
+    index.wait_task(1).await;
+
+    // the expected order is [1, 3, 2] instead of [3, 1, 2]
+    // because the attribute scale doesn't make the difference between 1 and 3.
+    index
+        .search(json!({"q": "the soup of day"}), |response, code| {
+            snapshot!(code, @"200 OK");
+            snapshot!(json_string!(response["hits"]), @r###"
+            [
+              {
+                "id": 1,
+                "a": "Soup of the day",
+                "b": "many the fish"
+              },
+              {
+                "id": 3,
+                "a": "the Soup of day",
+                "b": "many the fish"
+              },
+              {
+                "id": 2,
+                "a": "Soup of day",
+                "b": "many the lazy fish"
+              }
+            ]
+            "###);
+        })
+        .await;
+
+    // the expected order is [1, 2, 3] instead of [1, 3, 2]
+    // because the attribute scale sees all the word in the same attribute
+    // and so doesn't make the difference between the documents.
+    index
+        .search(json!({"q": "many the fish"}), |response, code| {
+            snapshot!(code, @"200 OK");
+            snapshot!(json_string!(response["hits"]), @r###"
+            [
+              {
+                "id": 1,
+                "a": "Soup of the day",
+                "b": "many the fish"
+              },
+              {
+                "id": 2,
+                "a": "Soup of day",
+                "b": "many the lazy fish"
+              },
+              {
+                "id": 3,
+                "a": "the Soup of day",
+                "b": "many the fish"
+              }
+            ]
+            "###);
+        })
+        .await;
+}
+
+#[actix_rt::test]
+async fn attribute_scale_phrase_search() {
+    let server = Server::new().await;
+    let (response, code) = server.set_features(json!({"proximityPrecision": true})).await;
+    meili_snap::snapshot!(code, @"200 OK");
+    meili_snap::snapshot!(meili_snap::json_string!(response), @r###"
+    {
+      "scoreDetails": false,
+      "vectorStore": false,
+      "metrics": false,
+      "exportPuffinReports": false,
+      "proximityPrecision": true
+    }
+    "###);
+    let index = server.index("test");
+
+    index.add_documents(DOCUMENTS.clone(), None).await;
+    index.wait_task(0).await;
+
+    let (_response, _code) = index
+        .update_settings(json!({
+            "proximityPrecision": "attributeScale",
+            "rankingRules": ["words", "typo", "proximity"],
+        }))
+        .await;
+    index.wait_task(1).await;
+
+    // the expected order is [1, 3] instead of [3, 1]
+    // because the attribute scale doesn't make the difference between 1 and 3.
+    // But 2 shouldn't be returned because "the" is not in the same attribute.
+    index
+        .search(json!({"q": "\"the soup of day\""}), |response, code| {
+            snapshot!(code, @"200 OK");
+            snapshot!(json_string!(response["hits"]), @r###"
+            [
+              {
+                "id": 1,
+                "a": "Soup of the day",
+                "b": "many the fish"
+              },
+              {
+                "id": 3,
+                "a": "the Soup of day",
+                "b": "many the fish"
+              }
+            ]
+            "###);
+        })
+        .await;
+
+    // the expected order is [1, 2, 3] instead of [1, 3]
+    // because the attribute scale sees all the word in the same attribute
+    // and so doesn't make the difference between the documents.
+    index
+        .search(json!({"q": "\"many the fish\""}), |response, code| {
+            snapshot!(code, @"200 OK");
+            snapshot!(json_string!(response["hits"]), @r###"
+            [
+              {
+                "id": 1,
+                "a": "Soup of the day",
+                "b": "many the fish"
+              },
+              {
+                "id": 2,
+                "a": "Soup of day",
+                "b": "many the lazy fish"
+              },
+              {
+                "id": 3,
+                "a": "the Soup of day",
+                "b": "many the fish"
+              }
+            ]
+            "###);
+        })
+        .await;
+}
+
+#[actix_rt::test]
+async fn word_scale_set_and_reset() {
+    let server = Server::new().await;
+    let (response, code) = server.set_features(json!({"proximityPrecision": true})).await;
+    meili_snap::snapshot!(code, @"200 OK");
+    meili_snap::snapshot!(meili_snap::json_string!(response), @r###"
+    {
+      "scoreDetails": false,
+      "vectorStore": false,
+      "metrics": false,
+      "exportPuffinReports": false,
+      "proximityPrecision": true
+    }
+    "###);
+    let index = server.index("test");
+
+    index.add_documents(DOCUMENTS.clone(), None).await;
+    index.wait_task(0).await;
+
+    // Set and reset the setting ensuring the swap between the 2 settings is applied.
+    let (_response, _code) = index
+        .update_settings(json!({
+            "proximityPrecision": "attributeScale",
+            "rankingRules": ["words", "typo", "proximity"],
+        }))
+        .await;
+    index.wait_task(1).await;
+
+    let (_response, _code) = index
+        .update_settings(json!({
+            "proximityPrecision": "wordScale",
+            "rankingRules": ["words", "typo", "proximity"],
+        }))
+        .await;
+    index.wait_task(2).await;
+
+    // [3, 1, 2]
+    index
+        .search(json!({"q": "the soup of day"}), |response, code| {
+            snapshot!(code, @"200 OK");
+            snapshot!(json_string!(response["hits"]), @r###"
+            [
+              {
+                "id": 3,
+                "a": "the Soup of day",
+                "b": "many the fish"
+              },
+              {
+                "id": 1,
+                "a": "Soup of the day",
+                "b": "many the fish"
+              },
+              {
+                "id": 2,
+                "a": "Soup of day",
+                "b": "many the lazy fish"
+              }
+            ]
+            "###);
+        })
+        .await;
+
+    // [1, 3, 2]
+    index
+        .search(json!({"q": "many the fish"}), |response, code| {
+            snapshot!(code, @"200 OK");
+            snapshot!(json_string!(response["hits"]), @r###"
+            [
+              {
+                "id": 1,
+                "a": "Soup of the day",
+                "b": "many the fish"
+              },
+              {
+                "id": 3,
+                "a": "the Soup of day",
+                "b": "many the fish"
+              },
+              {
+                "id": 2,
+                "a": "Soup of day",
+                "b": "many the lazy fish"
+              }
+            ]
+            "###);
+        })
+        .await;
+
+    // [3]
+    index
+        .search(json!({"q": "\"the soup of day\""}), |response, code| {
+            snapshot!(code, @"200 OK");
+            snapshot!(json_string!(response["hits"]), @r###"
+            [
+              {
+                "id": 3,
+                "a": "the Soup of day",
+                "b": "many the fish"
+              }
+            ]
+            "###);
+        })
+        .await;
+
+    // [1, 3]
+    index
+        .search(json!({"q": "\"many the fish\""}), |response, code| {
+            snapshot!(code, @"200 OK");
+            snapshot!(json_string!(response["hits"]), @r###"
+            [
+              {
+                "id": 1,
+                "a": "Soup of the day",
+                "b": "many the fish"
+              },
+              {
+                "id": 3,
+                "a": "the Soup of day",
+                "b": "many the fish"
+              }
+            ]
+            "###);
+        })
+        .await;
+}
+
+#[actix_rt::test]
+async fn attribute_scale_default_ranking_rules() {
+    let server = Server::new().await;
+    let (response, code) = server.set_features(json!({"proximityPrecision": true})).await;
+    meili_snap::snapshot!(code, @"200 OK");
+    meili_snap::snapshot!(meili_snap::json_string!(response), @r###"
+    {
+      "scoreDetails": false,
+      "vectorStore": false,
+      "metrics": false,
+      "exportPuffinReports": false,
+      "proximityPrecision": true
+    }
+    "###);
+    let index = server.index("test");
+
+    index.add_documents(DOCUMENTS.clone(), None).await;
+    index.wait_task(0).await;
+
+    let (response, code) = index
+        .update_settings(json!({
+            "proximityPrecision": "attributeScale"
+        }))
+        .await;
+    assert_eq!("202", code.as_str(), "{:?}", response);
+    index.wait_task(1).await;
+
+    // the expected order is [3, 1, 2]
+    index
+        .search(json!({"q": "the soup of day"}), |response, code| {
+            snapshot!(code, @"200 OK");
+            snapshot!(json_string!(response["hits"]), @r###"
+            [
+              {
+                "id": 3,
+                "a": "the Soup of day",
+                "b": "many the fish"
+              },
+              {
+                "id": 1,
+                "a": "Soup of the day",
+                "b": "many the fish"
+              },
+              {
+                "id": 2,
+                "a": "Soup of day",
+                "b": "many the lazy fish"
+              }
+            ]
+            "###);
+        })
+        .await;
+
+    // the expected order is [1, 3, 2] instead of [1, 3]
+    // because the attribute scale sees all the word in the same attribute
+    // and so doesn't remove the document 2.
+    index
+        .search(json!({"q": "\"many the fish\""}), |response, code| {
+            snapshot!(code, @"200 OK");
+            snapshot!(json_string!(response["hits"]), @r###"
+            [
+              {
+                "id": 1,
+                "a": "Soup of the day",
+                "b": "many the fish"
+              },
+              {
+                "id": 3,
+                "a": "the Soup of day",
+                "b": "many the fish"
+              },
+              {
+                "id": 2,
+                "a": "Soup of day",
+                "b": "many the lazy fish"
+              }
+            ]
+            "###);
+        })
+        .await;
+}

meilitool/Cargo.toml

@@ -0,0 +1,19 @@
+[package]
+name = "meilitool"
+description = "A CLI to edit a Meilisearch database from the command line"
+version.workspace = true
+authors.workspace = true
+homepage.workspace = true
+readme.workspace = true
+edition.workspace = true
+license.workspace = true
+
+[dependencies]
+anyhow = "1.0.75"
+clap = { version = "4.2.1", features = ["derive"] }
+dump = { path = "../dump" }
+file-store = { path = "../file-store" }
+meilisearch-auth = { path = "../meilisearch-auth" }
+meilisearch-types = { path = "../meilisearch-types" }
+time = { version = "0.3.30", features = ["formatting"] }
+uuid = { version = "1.5.0", features = ["v4"], default-features = false }

meilitool/src/main.rs

@@ -0,0 +1,314 @@
+use std::fs::{read_dir, read_to_string, remove_file, File};
+use std::io::BufWriter;
+use std::path::PathBuf;
+
+use anyhow::Context;
+use clap::{Parser, Subcommand};
+use dump::{DumpWriter, IndexMetadata};
+use file_store::FileStore;
+use meilisearch_auth::AuthController;
+use meilisearch_types::heed::types::{SerdeJson, Str};
+use meilisearch_types::heed::{Database, Env, EnvOpenOptions, RoTxn, RwTxn, Unspecified};
+use meilisearch_types::milli::documents::{obkv_to_object, DocumentsBatchReader};
+use meilisearch_types::milli::{obkv_to_json, BEU32};
+use meilisearch_types::tasks::{Status, Task};
+use meilisearch_types::versioning::check_version_file;
+use meilisearch_types::Index;
+use time::macros::format_description;
+use time::OffsetDateTime;
+use uuid_codec::UuidCodec;
+
+mod uuid_codec;
+
+#[derive(Parser)]
+#[command(author, version, about, long_about = None)]
+struct Cli {
+    /// The database path where the Meilisearch is running.
+    #[arg(long, default_value = "data.ms/")]
+    db_path: PathBuf,
+
+    #[command(subcommand)]
+    command: Command,
+}
+
+#[derive(Subcommand)]
+enum Command {
+    /// Clears the task queue and make it empty.
+    ///
+    /// This command can be safely executed even if Meilisearch is running and processing tasks.
+    /// Once the task queue is empty you can restart Meilisearch and no more tasks must be visible,
+    /// even the ones that were processing. However, it's highly possible that you see the processing
+    /// tasks in the queue again with an associated internal error message.
+    ClearTaskQueue,
+
+    /// Exports a dump from the Meilisearch database.
+    ///
+    /// Make sure to run this command when Meilisearch is not running or running but not processing tasks.
+    /// If tasks are being processed while a dump is being exported there are chances for the dump to be
+    /// malformed with missing tasks.
+    ///
+    /// TODO Verify this claim or make sure it cannot happen and we can export dumps
+    ///      without caring about killing Meilisearch first!
+    ExportADump {
+        /// The directory in which the dump will be created.
+        #[arg(long, default_value = "dumps/")]
+        dump_dir: PathBuf,
+
+        /// Skip dumping the enqueued or processing tasks.
+        ///
+        /// Can be useful when there are a lot of them and it is not particularly useful
+        /// to keep them. Note that only the enqueued tasks takes up space so skipping
+        /// the processed ones is not particularly interesting.
+        #[arg(long)]
+        skip_enqueued_tasks: bool,
+    },
+}
+
+fn main() -> anyhow::Result<()> {
+    let Cli { db_path, command } = Cli::parse();
+
+    check_version_file(&db_path).context("While checking the version file")?;
+
+    match command {
+        Command::ClearTaskQueue => clear_task_queue(db_path),
+        Command::ExportADump { dump_dir, skip_enqueued_tasks } => {
+            export_a_dump(db_path, dump_dir, skip_enqueued_tasks)
+        }
+    }
+}
+
+/// Clears the task queue located at `db_path`.
+fn clear_task_queue(db_path: PathBuf) -> anyhow::Result<()> {
+    let path = db_path.join("tasks");
+    let env = EnvOpenOptions::new()
+        .max_dbs(100)
+        .open(&path)
+        .with_context(|| format!("While trying to open {:?}", path.display()))?;
+
+    eprintln!("Deleting tasks from the database...");
+
+    let mut wtxn = env.write_txn()?;
+    let all_tasks = try_opening_poly_database(&env, &wtxn, "all-tasks")?;
+    let total = all_tasks.len(&wtxn)?;
+    let status = try_opening_poly_database(&env, &wtxn, "status")?;
+    let kind = try_opening_poly_database(&env, &wtxn, "kind")?;
+    let index_tasks = try_opening_poly_database(&env, &wtxn, "index-tasks")?;
+    let canceled_by = try_opening_poly_database(&env, &wtxn, "canceled_by")?;
+    let enqueued_at = try_opening_poly_database(&env, &wtxn, "enqueued-at")?;
+    let started_at = try_opening_poly_database(&env, &wtxn, "started-at")?;
+    let finished_at = try_opening_poly_database(&env, &wtxn, "finished-at")?;
+
+    try_clearing_poly_database(&mut wtxn, all_tasks, "all-tasks")?;
+    try_clearing_poly_database(&mut wtxn, status, "status")?;
+    try_clearing_poly_database(&mut wtxn, kind, "kind")?;
+    try_clearing_poly_database(&mut wtxn, index_tasks, "index-tasks")?;
+    try_clearing_poly_database(&mut wtxn, canceled_by, "canceled_by")?;
+    try_clearing_poly_database(&mut wtxn, enqueued_at, "enqueued-at")?;
+    try_clearing_poly_database(&mut wtxn, started_at, "started-at")?;
+    try_clearing_poly_database(&mut wtxn, finished_at, "finished-at")?;
+
+    wtxn.commit().context("While committing the transaction")?;
+
+    eprintln!("Successfully deleted {total} tasks from the tasks database!");
+    eprintln!("Deleting the content files from disk...");
+
+    let mut count = 0usize;
+    let update_files = db_path.join("update_files");
+    let entries = read_dir(&update_files).with_context(|| {
+        format!("While trying to read the content of {:?}", update_files.display())
+    })?;
+    for result in entries {
+        match result {
+            Ok(ent) => match remove_file(ent.path()) {
+                Ok(_) => count += 1,
+                Err(e) => eprintln!("Error while deleting {:?}: {}", ent.path().display(), e),
+            },
+            Err(e) => {
+                eprintln!("Error while reading a file in {:?}: {}", update_files.display(), e)
+            }
+        }
+    }
+
+    eprintln!("Sucessfully deleted {count} content files from disk!");
+
+    Ok(())
+}
+
+fn try_opening_database<KC: 'static, DC: 'static>(
+    env: &Env,
+    rtxn: &RoTxn,
+    db_name: &str,
+) -> anyhow::Result<Database<KC, DC>> {
+    env.open_database(rtxn, Some(db_name))
+        .with_context(|| format!("While opening the {db_name:?} database"))?
+        .with_context(|| format!("Missing the {db_name:?} database"))
+}
+
+fn try_opening_poly_database(
+    env: &Env,
+    rtxn: &RoTxn,
+    db_name: &str,
+) -> anyhow::Result<Database<Unspecified, Unspecified>> {
+    env.database_options()
+        .name(db_name)
+        .open(rtxn)
+        .with_context(|| format!("While opening the {db_name:?} poly database"))?
+        .with_context(|| format!("Missing the {db_name:?} poly database"))
+}
+
+fn try_clearing_poly_database(
+    wtxn: &mut RwTxn,
+    database: Database<Unspecified, Unspecified>,
+    db_name: &str,
+) -> anyhow::Result<()> {
+    database.clear(wtxn).with_context(|| format!("While clearing the {db_name:?} database"))
+}
+
+/// Exports a dump into the dump directory.
+fn export_a_dump(
+    db_path: PathBuf,
+    dump_dir: PathBuf,
+    skip_enqueued_tasks: bool,
+) -> Result<(), anyhow::Error> {
+    let started_at = OffsetDateTime::now_utc();
+
+    // 1. Extracts the instance UID from disk
+    let instance_uid_path = db_path.join("instance-uid");
+    let instance_uid = match read_to_string(&instance_uid_path) {
+        Ok(content) => match content.trim().parse() {
+            Ok(uuid) => Some(uuid),
+            Err(e) => {
+                eprintln!("Impossible to parse instance-uid: {e}");
+                None
+            }
+        },
+        Err(e) => {
+            eprintln!("Impossible to read {}: {}", instance_uid_path.display(), e);
+            None
+        }
+    };
+
+    let dump = DumpWriter::new(instance_uid).context("While creating a new dump")?;
+    let file_store =
+        FileStore::new(db_path.join("update_files")).context("While opening the FileStore")?;
+
+    let index_scheduler_path = db_path.join("tasks");
+    let env = EnvOpenOptions::new()
+        .max_dbs(100)
+        .open(&index_scheduler_path)
+        .with_context(|| format!("While trying to open {:?}", index_scheduler_path.display()))?;
+
+    eprintln!("Dumping the keys...");
+
+    // 2. dump the keys
+    let auth_store = AuthController::new(&db_path, &None)
+        .with_context(|| format!("While opening the auth store at {}", db_path.display()))?;
+    let mut dump_keys = dump.create_keys()?;
+    let mut count = 0;
+    for key in auth_store.list_keys()? {
+        dump_keys.push_key(&key)?;
+        count += 1;
+    }
+    dump_keys.flush()?;
+
+    eprintln!("Successfully dumped {count} keys!");
+
+    let rtxn = env.read_txn()?;
+    let all_tasks: Database<BEU32, SerdeJson<Task>> =
+        try_opening_database(&env, &rtxn, "all-tasks")?;
+    let index_mapping: Database<Str, UuidCodec> =
+        try_opening_database(&env, &rtxn, "index-mapping")?;
+
+    if skip_enqueued_tasks {
+        eprintln!("Skip dumping the enqueued tasks...");
+    } else {
+        eprintln!("Dumping the enqueued tasks...");
+
+        // 3. dump the tasks
+        let mut dump_tasks = dump.create_tasks_queue()?;
+        let mut count = 0;
+        for ret in all_tasks.iter(&rtxn)? {
+            let (_, t) = ret?;
+            let status = t.status;
+            let content_file = t.content_uuid();
+            let mut dump_content_file = dump_tasks.push_task(&t.into())?;
+
+            // 3.1. Dump the `content_file` associated with the task if there is one and the task is not finished yet.
+            if let Some(content_file_uuid) = content_file {
+                if status == Status::Enqueued {
+                    let content_file = file_store.get_update(content_file_uuid)?;
+
+                    let reader =
+                        DocumentsBatchReader::from_reader(content_file).with_context(|| {
+                            format!("While reading content file {:?}", content_file_uuid)
+                        })?;
+
+                    let (mut cursor, documents_batch_index) = reader.into_cursor_and_fields_index();
+                    while let Some(doc) = cursor.next_document().with_context(|| {
+                        format!("While iterating on content file {:?}", content_file_uuid)
+                    })? {
+                        dump_content_file
+                            .push_document(&obkv_to_object(&doc, &documents_batch_index)?)?;
+                    }
+                    dump_content_file.flush()?;
+                    count += 1;
+                }
+            }
+        }
+        dump_tasks.flush()?;
+
+        eprintln!("Successfully dumped {count} enqueued tasks!");
+    }
+
+    eprintln!("Dumping the indexes...");
+
+    // 4. Dump the indexes
+    let mut count = 0;
+    for result in index_mapping.iter(&rtxn)? {
+        let (uid, uuid) = result?;
+        let index_path = db_path.join("indexes").join(uuid.to_string());
+        let index = Index::new(EnvOpenOptions::new(), &index_path).with_context(|| {
+            format!("While trying to open the index at path {:?}", index_path.display())
+        })?;
+
+        let rtxn = index.read_txn()?;
+        let metadata = IndexMetadata {
+            uid: uid.to_owned(),
+            primary_key: index.primary_key(&rtxn)?.map(String::from),
+            created_at: index.created_at(&rtxn)?,
+            updated_at: index.updated_at(&rtxn)?,
+        };
+        let mut index_dumper = dump.create_index(uid, &metadata)?;
+
+        let fields_ids_map = index.fields_ids_map(&rtxn)?;
+        let all_fields: Vec<_> = fields_ids_map.iter().map(|(id, _)| id).collect();
+
+        // 4.1. Dump the documents
+        for ret in index.all_documents(&rtxn)? {
+            let (_id, doc) = ret?;
+            let document = obkv_to_json(&all_fields, &fields_ids_map, doc)?;
+            index_dumper.push_document(&document)?;
+        }
+
+        // 4.2. Dump the settings
+        let settings = meilisearch_types::settings::settings(&index, &rtxn)?;
+        index_dumper.settings(&settings)?;
+        count += 1;
+    }
+
+    eprintln!("Successfully dumped {count} indexes!");
+    // We will not dump experimental feature settings
+    eprintln!("The tool is not dumping experimental features, please set them by hand afterward");
+
+    let dump_uid = started_at.format(format_description!(
+        "[year repr:full][month repr:numerical][day padding:zero]-[hour padding:zero][minute padding:zero][second padding:zero][subsecond digits:3]"
+    )).unwrap();
+
+    let path = dump_dir.join(format!("{}.dump", dump_uid));
+    let file = File::create(&path)?;
+    dump.persist_to(BufWriter::new(file))?;
+
+    eprintln!("Dump exported at path {:?}", path.display());
+
+    Ok(())
+}

meilitool/src/uuid_codec.rs

@@ -0,0 +1,24 @@
+use std::borrow::Cow;
+use std::convert::TryInto;
+
+use meilisearch_types::heed::{BoxedError, BytesDecode, BytesEncode};
+use uuid::Uuid;
+
+/// A heed codec for value of struct Uuid.
+pub struct UuidCodec;
+
+impl<'a> BytesDecode<'a> for UuidCodec {
+    type DItem = Uuid;
+
+    fn bytes_decode(bytes: &'a [u8]) -> Result<Self::DItem, BoxedError> {
+        bytes.try_into().map(Uuid::from_bytes).map_err(Into::into)
+    }
+}
+
+impl BytesEncode<'_> for UuidCodec {
+    type EItem = Uuid;
+
+    fn bytes_encode(item: &Self::EItem) -> Result<Cow<[u8]>, BoxedError> {
+        Ok(Cow::Borrowed(item.as_bytes()))
+    }
+}

milli/Cargo.toml

@@ -17,10 +17,10 @@ bincode = "1.3.3"
 bstr = "1.4.0"
 bytemuck = { version = "1.13.1", features = ["extern_crate_alloc"] }
 byteorder = "1.4.3"
-charabia = { version = "0.8.3", default-features = false }
+charabia = { version = "0.8.5", default-features = false }
 concat-arrays = "0.1.2"
 crossbeam-channel = "0.5.8"
-deserr = { version = "0.6.0", features = ["actix-web"]}
+deserr = "0.6.0"
 either = { version = "1.8.1", features = ["serde"] }
 flatten-serde-json = { path = "../flatten-serde-json" }
 fst = "0.4.7"
@@ -29,8 +29,8 @@ geoutils = "0.5.1"
 grenad = { version = "0.4.5", default-features = false, features = [
     "rayon", "tempfile"
 ] }
-heed = { git = "https://github.com/meilisearch/heed", tag = "v0.12.7", default-features = false, features = [
-    "lmdb", "read-txn-no-tls"
+heed = { version = "0.20.0-alpha.9", default-features = false, features = [
+    "serde-json", "serde-bincode", "read-txn-no-tls"
 ] }
 indexmap = { version = "2.0.0", features = ["serde"] }
 instant-distance = { version = "0.6.1", features = ["with-serde"] }
@@ -83,7 +83,7 @@ meili-snap = { path = "../meili-snap" }
 rand = { version = "0.8.5", features = ["small_rng"] }
 
 [features]
-all-tokenizations = ["charabia/chinese", "charabia/hebrew", "charabia/japanese", "charabia/thai", "charabia/korean", "charabia/greek"]
+all-tokenizations = ["charabia/chinese", "charabia/hebrew", "charabia/japanese", "charabia/thai", "charabia/korean", "charabia/greek", "charabia/khmer"]
 
 # Use POSIX semaphores instead of SysV semaphores in LMDB
 # For more information on this feature, see heed's Cargo.toml
@@ -107,3 +107,6 @@ thai = ["charabia/thai"]
 
 # allow greek specialized tokenization
 greek = ["charabia/greek"]
+
+# allow khmer specialized tokenization
+khmer = ["charabia/khmer"]

milli/src/documents/mod.rs

@@ -1,5 +1,6 @@
 mod builder;
 mod enriched;
+mod primary_key;
 mod reader;
 mod serde_impl;
 
@@ -11,6 +12,7 @@ use bimap::BiHashMap;
 pub use builder::DocumentsBatchBuilder;
 pub use enriched::{EnrichedDocument, EnrichedDocumentsBatchCursor, EnrichedDocumentsBatchReader};
 use obkv::KvReader;
+pub use primary_key::{DocumentIdExtractionError, FieldIdMapper, PrimaryKey, DEFAULT_PRIMARY_KEY};
 pub use reader::{DocumentsBatchCursor, DocumentsBatchCursorError, DocumentsBatchReader};
 use serde::{Deserialize, Serialize};
 
@@ -87,6 +89,12 @@ impl DocumentsBatchIndex {
     }
 }
 
+impl FieldIdMapper for DocumentsBatchIndex {
+    fn id(&self, name: &str) -> Option<FieldId> {
+        self.id(name)
+    }
+}
+
 #[derive(Debug, thiserror::Error)]
 pub enum Error {
     #[error("Error parsing number {value:?} at line {line}: {error}")]

milli/src/documents/primary_key.rs

@@ -0,0 +1,172 @@
+use std::iter;
+use std::result::Result as StdResult;
+
+use serde_json::Value;
+
+use crate::{FieldId, InternalError, Object, Result, UserError};
+
+/// The symbol used to define levels in a nested primary key.
+const PRIMARY_KEY_SPLIT_SYMBOL: char = '.';
+
+/// The default primary that is used when not specified.
+pub const DEFAULT_PRIMARY_KEY: &str = "id";
+
+/// Trait for objects that can map the name of a field to its [`FieldId`].
+pub trait FieldIdMapper {
+    /// Attempts to map the passed name to its [`FieldId`].
+    ///
+    /// `None` if the field with this name was not found.
+    fn id(&self, name: &str) -> Option<FieldId>;
+}
+
+/// A type that represent the type of primary key that has been set
+/// for this index, a classic flat one or a nested one.
+#[derive(Debug, Clone, Copy)]
+pub enum PrimaryKey<'a> {
+    Flat { name: &'a str, field_id: FieldId },
+    Nested { name: &'a str },
+}
+
+pub enum DocumentIdExtractionError {
+    InvalidDocumentId(UserError),
+    MissingDocumentId,
+    TooManyDocumentIds(usize),
+}
+
+impl<'a> PrimaryKey<'a> {
+    pub fn new(path: &'a str, fields: &impl FieldIdMapper) -> Option<Self> {
+        Some(if path.contains(PRIMARY_KEY_SPLIT_SYMBOL) {
+            Self::Nested { name: path }
+        } else {
+            let field_id = fields.id(path)?;
+            Self::Flat { name: path, field_id }
+        })
+    }
+
+    pub fn name(&self) -> &str {
+        match self {
+            PrimaryKey::Flat { name, .. } => name,
+            PrimaryKey::Nested { name } => name,
+        }
+    }
+
+    pub fn document_id(
+        &self,
+        document: &obkv::KvReader<FieldId>,
+        fields: &impl FieldIdMapper,
+    ) -> Result<StdResult<String, DocumentIdExtractionError>> {
+        match self {
+            PrimaryKey::Flat { name: _, field_id } => match document.get(*field_id) {
+                Some(document_id_bytes) => {
+                    let document_id = serde_json::from_slice(document_id_bytes)
+                        .map_err(InternalError::SerdeJson)?;
+                    match validate_document_id_value(document_id)? {
+                        Ok(document_id) => Ok(Ok(document_id)),
+                        Err(user_error) => {
+                            Ok(Err(DocumentIdExtractionError::InvalidDocumentId(user_error)))
+                        }
+                    }
+                }
+                None => Ok(Err(DocumentIdExtractionError::MissingDocumentId)),
+            },
+            nested @ PrimaryKey::Nested { .. } => {
+                let mut matching_documents_ids = Vec::new();
+                for (first_level_name, right) in nested.possible_level_names() {
+                    if let Some(field_id) = fields.id(first_level_name) {
+                        if let Some(value_bytes) = document.get(field_id) {
+                            let object = serde_json::from_slice(value_bytes)
+                                .map_err(InternalError::SerdeJson)?;
+                            fetch_matching_values(object, right, &mut matching_documents_ids);
+
+                            if matching_documents_ids.len() >= 2 {
+                                return Ok(Err(DocumentIdExtractionError::TooManyDocumentIds(
+                                    matching_documents_ids.len(),
+                                )));
+                            }
+                        }
+                    }
+                }
+
+                match matching_documents_ids.pop() {
+                    Some(document_id) => match validate_document_id_value(document_id)? {
+                        Ok(document_id) => Ok(Ok(document_id)),
+                        Err(user_error) => {
+                            Ok(Err(DocumentIdExtractionError::InvalidDocumentId(user_error)))
+                        }
+                    },
+                    None => Ok(Err(DocumentIdExtractionError::MissingDocumentId)),
+                }
+            }
+        }
+    }
+
+    /// Returns an `Iterator` that gives all the possible fields names the primary key
+    /// can have depending of the first level name and depth of the objects.
+    pub fn possible_level_names(&self) -> impl Iterator<Item = (&str, &str)> + '_ {
+        let name = self.name();
+        name.match_indices(PRIMARY_KEY_SPLIT_SYMBOL)
+            .map(move |(i, _)| (&name[..i], &name[i + PRIMARY_KEY_SPLIT_SYMBOL.len_utf8()..]))
+            .chain(iter::once((name, "")))
+    }
+}
+
+fn fetch_matching_values(value: Value, selector: &str, output: &mut Vec<Value>) {
+    match value {
+        Value::Object(object) => fetch_matching_values_in_object(object, selector, "", output),
+        otherwise => output.push(otherwise),
+    }
+}
+
+fn fetch_matching_values_in_object(
+    object: Object,
+    selector: &str,
+    base_key: &str,
+    output: &mut Vec<Value>,
+) {
+    for (key, value) in object {
+        let base_key = if base_key.is_empty() {
+            key.to_string()
+        } else {
+            format!("{}{}{}", base_key, PRIMARY_KEY_SPLIT_SYMBOL, key)
+        };
+
+        if starts_with(selector, &base_key) {
+            match value {
+                Value::Object(object) => {
+                    fetch_matching_values_in_object(object, selector, &base_key, output)
+                }
+                value => output.push(value),
+            }
+        }
+    }
+}
+
+fn starts_with(selector: &str, key: &str) -> bool {
+    selector.strip_prefix(key).map_or(false, |tail| {
+        tail.chars().next().map(|c| c == PRIMARY_KEY_SPLIT_SYMBOL).unwrap_or(true)
+    })
+}
+
+// FIXME: move to a DocumentId struct
+
+fn validate_document_id(document_id: &str) -> Option<&str> {
+    if !document_id.is_empty()
+        && document_id.chars().all(|c| matches!(c, 'a'..='z' | 'A'..='Z' | '0'..='9' | '-' | '_'))
+    {
+        Some(document_id)
+    } else {
+        None
+    }
+}
+
+pub fn validate_document_id_value(document_id: Value) -> Result<StdResult<String, UserError>> {
+    match document_id {
+        Value::String(string) => match validate_document_id(&string) {
+            Some(s) if s.len() == string.len() => Ok(Ok(string)),
+            Some(s) => Ok(Ok(s.to_string())),
+            None => Ok(Err(UserError::InvalidDocumentId { document_id: Value::String(string) })),
+        },
+        Value::Number(number) if number.is_i64() => Ok(Ok(number.to_string())),
+        content => Ok(Err(UserError::InvalidDocumentId { document_id: content })),
+    }
+}

milli/src/error.rs

@@ -152,7 +152,7 @@ only composed of alphanumeric characters (a-z A-Z 0-9), hyphens (-) and undersco
         valid_fields: BTreeSet<String>,
         hidden_fields: bool,
     },
-    #[error("{}", HeedError::BadOpenOptions)]
+    #[error("an environment is already opened with different options")]
     InvalidLmdbOpenOptions,
     #[error("You must specify where `sort` is listed in the rankingRules setting to use the sort parameter at search time.")]
     SortRankingRuleMissing,
@@ -326,11 +326,12 @@ impl From<HeedError> for Error {
             HeedError::Mdb(MdbError::MapFull) => UserError(MaxDatabaseSizeReached),
             HeedError::Mdb(MdbError::Invalid) => UserError(InvalidStoreFile),
             HeedError::Mdb(error) => InternalError(Store(error)),
-            HeedError::Encoding => InternalError(Serialization(Encoding { db_name: None })),
-            HeedError::Decoding => InternalError(Serialization(Decoding { db_name: None })),
+            // TODO use the encoding
+            HeedError::Encoding(_) => InternalError(Serialization(Encoding { db_name: None })),
+            HeedError::Decoding(_) => InternalError(Serialization(Decoding { db_name: None })),
             HeedError::InvalidDatabaseTyping => InternalError(InvalidDatabaseTyping),
             HeedError::DatabaseClosing => InternalError(DatabaseClosing),
-            HeedError::BadOpenOptions => UserError(InvalidLmdbOpenOptions),
+            HeedError::BadOpenOptions { .. } => UserError(InvalidLmdbOpenOptions),
         }
     }
 }

milli/src/external_documents_ids.rs

@@ -1,8 +1,7 @@
 use std::collections::HashMap;
 
-use heed::types::{OwnedType, Str};
+use heed::types::Str;
 use heed::{Database, RoIter, RoTxn, RwTxn};
-use roaring::RoaringBitmap;
 
 use crate::{DocumentId, BEU32};
 
@@ -17,10 +16,10 @@ pub struct DocumentOperation {
     pub kind: DocumentOperationKind,
 }
 
-pub struct ExternalDocumentsIds(Database<Str, OwnedType<BEU32>>);
+pub struct ExternalDocumentsIds(Database<Str, BEU32>);
 
 impl ExternalDocumentsIds {
-    pub fn new(db: Database<Str, OwnedType<BEU32>>) -> ExternalDocumentsIds {
+    pub fn new(db: Database<Str, BEU32>) -> ExternalDocumentsIds {
         ExternalDocumentsIds(db)
     }
 
@@ -30,7 +29,7 @@ impl ExternalDocumentsIds {
     }
 
     pub fn get<A: AsRef<str>>(&self, rtxn: &RoTxn, external_id: A) -> heed::Result<Option<u32>> {
-        Ok(self.0.get(rtxn, external_id.as_ref())?.map(|x| x.get()))
+        self.0.get(rtxn, external_id.as_ref())
     }
 
     /// An helper function to debug this type, returns an `HashMap` of both,
@@ -39,28 +38,11 @@ impl ExternalDocumentsIds {
         let mut map = HashMap::default();
         for result in self.0.iter(rtxn)? {
             let (external, internal) = result?;
-            map.insert(external.to_owned(), internal.get());
+            map.insert(external.to_owned(), internal);
         }
         Ok(map)
     }
 
-    /// Looks for the internal ids in the passed bitmap, and returns an iterator over the mapping between
-    /// these internal ids and their external id.
-    ///
-    /// The returned iterator has `Result<(String, DocumentId), RoaringBitmap>` as `Item`,
-    /// where the returned values can be:
-    /// - `Ok((external_id, internal_id))`: if a mapping was found
-    /// - `Err(remaining_ids)`: if the external ids for some of the requested internal ids weren't found.
-    ///   In that case the returned bitmap contains the internal ids whose external ids were not found after traversing
-    ///   the entire fst.
-    pub fn find_external_id_of<'t>(
-        &self,
-        rtxn: &'t RoTxn,
-        internal_ids: RoaringBitmap,
-    ) -> heed::Result<ExternalToInternalOwnedIterator<'t>> {
-        self.0.iter(rtxn).map(|iter| ExternalToInternalOwnedIterator { iter, internal_ids })
-    }
-
     /// Applies the list of operations passed as argument, modifying the current external to internal id mapping.
     ///
     /// If the list contains multiple operations on the same external id, then the result is unspecified.
@@ -73,7 +55,7 @@ impl ExternalDocumentsIds {
         for DocumentOperation { external_id, internal_id, kind } in operations {
             match kind {
                 DocumentOperationKind::Create => {
-                    self.0.put(wtxn, &external_id, &BEU32::new(internal_id))?;
+                    self.0.put(wtxn, &external_id, &internal_id)?;
                 }
                 DocumentOperationKind::Delete => {
                     if !self.0.delete(wtxn, &external_id)? {
@@ -87,60 +69,7 @@ impl ExternalDocumentsIds {
     }
 
     /// Returns an iterator over all the external ids.
-    pub fn iter<'t>(&self, rtxn: &'t RoTxn) -> heed::Result<RoIter<'t, Str, OwnedType<BEU32>>> {
+    pub fn iter<'t>(&self, rtxn: &'t RoTxn) -> heed::Result<RoIter<'t, Str, BEU32>> {
         self.0.iter(rtxn)
     }
 }
-
-/// An iterator over mappings between requested internal ids and external ids.
-///
-/// See [`ExternalDocumentsIds::find_external_id_of`] for details.
-pub struct ExternalToInternalOwnedIterator<'t> {
-    iter: RoIter<'t, Str, OwnedType<BEU32>>,
-    internal_ids: RoaringBitmap,
-}
-
-impl<'t> Iterator for ExternalToInternalOwnedIterator<'t> {
-    /// A result indicating if a mapping was found, or if the stream was exhausted without finding all internal ids.
-    type Item = Result<(&'t str, DocumentId), RoaringBitmap>;
-
-    fn next(&mut self) -> Option<Self::Item> {
-        // if all requested ids were found, we won't find any other, so short-circuit
-        if self.internal_ids.is_empty() {
-            return None;
-        }
-        loop {
-            let (external, internal) = match self.iter.next() {
-                Some(Ok((external, internal))) => (external, internal),
-                // TODO manage this better, remove panic
-                Some(Err(e)) => panic!("{}", e),
-                _ => {
-                    // we exhausted the stream but we still have some internal ids to find
-                    let remaining_ids = std::mem::take(&mut self.internal_ids);
-                    return Some(Err(remaining_ids));
-                    // note: next calls to `next` will return `None` since we replaced the internal_ids
-                    // with the default empty bitmap
-                }
-            };
-            let internal = internal.get();
-            let was_contained = self.internal_ids.remove(internal);
-            if was_contained {
-                return Some(Ok((external, internal)));
-            }
-        }
-    }
-}
-
-impl<'t> ExternalToInternalOwnedIterator<'t> {
-    /// Returns the bitmap of internal ids whose external id are yet to be found
-    pub fn remaining_internal_ids(&self) -> &RoaringBitmap {
-        &self.internal_ids
-    }
-
-    /// Consumes this iterator and returns an iterator over only the external ids, ignoring the internal ids.
-    ///
-    /// Use this when you don't need the mapping between the external and the internal ids.
-    pub fn only_external_ids(self) -> impl Iterator<Item = Result<String, RoaringBitmap>> + 't {
-        self.map(|res| res.map(|(external, _internal)| external.to_owned()))
-    }
-}

milli/src/fields_ids_map.rs

@@ -81,6 +81,12 @@ impl Default for FieldsIdsMap {
     }
 }
 
+impl crate::documents::FieldIdMapper for FieldsIdsMap {
+    fn id(&self, name: &str) -> Option<FieldId> {
+        self.id(name)
+    }
+}
+
 #[cfg(test)]
 mod tests {
     use super::*;

milli/src/heed_codec/beu16_str_codec.rs

@@ -2,26 +2,28 @@ use std::borrow::Cow;
 use std::convert::TryInto;
 use std::str;
 
+use heed::BoxedError;
+
 pub struct BEU16StrCodec;
 
 impl<'a> heed::BytesDecode<'a> for BEU16StrCodec {
     type DItem = (u16, &'a str);
 
-    fn bytes_decode(bytes: &'a [u8]) -> Option<Self::DItem> {
+    fn bytes_decode(bytes: &'a [u8]) -> Result<Self::DItem, BoxedError> {
         let (n_bytes, str_bytes) = bytes.split_at(2);
-        let n = n_bytes.try_into().map(u16::from_be_bytes).ok()?;
-        let s = str::from_utf8(str_bytes).ok()?;
-        Some((n, s))
+        let n = n_bytes.try_into().map(u16::from_be_bytes)?;
+        let s = str::from_utf8(str_bytes)?;
+        Ok((n, s))
     }
 }
 
 impl<'a> heed::BytesEncode<'a> for BEU16StrCodec {
     type EItem = (u16, &'a str);
 
-    fn bytes_encode((n, s): &Self::EItem) -> Option<Cow<[u8]>> {
+    fn bytes_encode((n, s): &Self::EItem) -> Result<Cow<[u8]>, BoxedError> {
         let mut bytes = Vec::with_capacity(s.len() + 2);
         bytes.extend_from_slice(&n.to_be_bytes());
         bytes.extend_from_slice(s.as_bytes());
-        Some(Cow::Owned(bytes))
+        Ok(Cow::Owned(bytes))
     }
 }

milli/src/heed_codec/beu32_str_codec.rs

@@ -2,26 +2,28 @@ use std::borrow::Cow;
 use std::convert::TryInto;
 use std::str;
 
+use heed::BoxedError;
+
 pub struct BEU32StrCodec;
 
 impl<'a> heed::BytesDecode<'a> for BEU32StrCodec {
     type DItem = (u32, &'a str);
 
-    fn bytes_decode(bytes: &'a [u8]) -> Option<Self::DItem> {
+    fn bytes_decode(bytes: &'a [u8]) -> Result<Self::DItem, BoxedError> {
         let (n_bytes, str_bytes) = bytes.split_at(4);
-        let n = n_bytes.try_into().map(u32::from_be_bytes).ok()?;
-        let s = str::from_utf8(str_bytes).ok()?;
-        Some((n, s))
+        let n = n_bytes.try_into().map(u32::from_be_bytes)?;
+        let s = str::from_utf8(str_bytes)?;
+        Ok((n, s))
     }
 }
 
 impl<'a> heed::BytesEncode<'a> for BEU32StrCodec {
     type EItem = (u32, &'a str);
 
-    fn bytes_encode((n, s): &Self::EItem) -> Option<Cow<[u8]>> {
+    fn bytes_encode((n, s): &Self::EItem) -> Result<Cow<[u8]>, BoxedError> {
         let mut bytes = Vec::with_capacity(s.len() + 4);
         bytes.extend_from_slice(&n.to_be_bytes());
         bytes.extend_from_slice(s.as_bytes());
-        Some(Cow::Owned(bytes))
+        Ok(Cow::Owned(bytes))
     }
 }

milli/src/heed_codec/byte_slice_ref.rs

@@ -1,23 +1,23 @@
 use std::borrow::Cow;
 
-use heed::{BytesDecode, BytesEncode};
+use heed::{BoxedError, BytesDecode, BytesEncode};
 
-/// A codec for values of type `&[u8]`. Unlike `ByteSlice`, its `EItem` and `DItem` associated
+/// A codec for values of type `&[u8]`. Unlike `Bytes`, its `EItem` and `DItem` associated
 /// types are equivalent (= `&'a [u8]`) and these values can reside within another structure.
-pub struct ByteSliceRefCodec;
+pub struct BytesRefCodec;
 
-impl<'a> BytesEncode<'a> for ByteSliceRefCodec {
+impl<'a> BytesEncode<'a> for BytesRefCodec {
     type EItem = &'a [u8];
 
-    fn bytes_encode(item: &'a Self::EItem) -> Option<Cow<'a, [u8]>> {
-        Some(Cow::Borrowed(item))
+    fn bytes_encode(item: &'a Self::EItem) -> Result<Cow<'a, [u8]>, BoxedError> {
+        Ok(Cow::Borrowed(item))
     }
 }
 
-impl<'a> BytesDecode<'a> for ByteSliceRefCodec {
+impl<'a> BytesDecode<'a> for BytesRefCodec {
     type DItem = &'a [u8];
 
-    fn bytes_decode(bytes: &'a [u8]) -> Option<Self::DItem> {
-        Some(bytes)
+    fn bytes_decode(bytes: &'a [u8]) -> Result<Self::DItem, BoxedError> {
+        Ok(bytes)
     }
 }

milli/src/heed_codec/facet/field_doc_id_facet_codec.rs

@@ -1,8 +1,9 @@
 use std::borrow::Cow;
 use std::marker::PhantomData;
 
-use heed::{BytesDecode, BytesEncode};
+use heed::{BoxedError, BytesDecode, BytesEncode};
 
+use crate::heed_codec::SliceTooShortError;
 use crate::{try_split_array_at, DocumentId, FieldId};
 
 pub struct FieldDocIdFacetCodec<C>(PhantomData<C>);
@@ -13,16 +14,16 @@ where
 {
     type DItem = (FieldId, DocumentId, C::DItem);
 
-    fn bytes_decode(bytes: &'a [u8]) -> Option<Self::DItem> {
-        let (field_id_bytes, bytes) = try_split_array_at(bytes)?;
+    fn bytes_decode(bytes: &'a [u8]) -> Result<Self::DItem, BoxedError> {
+        let (field_id_bytes, bytes) = try_split_array_at(bytes).ok_or(SliceTooShortError)?;
         let field_id = u16::from_be_bytes(field_id_bytes);
 
-        let (document_id_bytes, bytes) = try_split_array_at(bytes)?;
+        let (document_id_bytes, bytes) = try_split_array_at(bytes).ok_or(SliceTooShortError)?;
         let document_id = u32::from_be_bytes(document_id_bytes);
 
         let value = C::bytes_decode(bytes)?;
 
-        Some((field_id, document_id, value))
+        Ok((field_id, document_id, value))
     }
 }
 
@@ -32,13 +33,15 @@ where
 {
     type EItem = (FieldId, DocumentId, C::EItem);
 
-    fn bytes_encode((field_id, document_id, value): &'a Self::EItem) -> Option<Cow<[u8]>> {
+    fn bytes_encode(
+        (field_id, document_id, value): &'a Self::EItem,
+    ) -> Result<Cow<[u8]>, BoxedError> {
         let mut bytes = Vec::with_capacity(32);
         bytes.extend_from_slice(&field_id.to_be_bytes()); // 2 bytes
         bytes.extend_from_slice(&document_id.to_be_bytes()); // 4 bytes
         let value_bytes = C::bytes_encode(value)?;
         // variable length, if f64 -> 16 bytes, if string -> large, potentially
         bytes.extend_from_slice(&value_bytes);
-        Some(Cow::Owned(bytes))
+        Ok(Cow::Owned(bytes))
     }
 }

milli/src/heed_codec/facet/mod.rs

@@ -5,8 +5,8 @@ use std::borrow::Cow;
 use std::convert::TryFrom;
 use std::marker::PhantomData;
 
-use heed::types::{DecodeIgnore, OwnedType};
-use heed::{BytesDecode, BytesEncode};
+use heed::types::DecodeIgnore;
+use heed::{BoxedError, BytesDecode, BytesEncode};
 use roaring::RoaringBitmap;
 
 pub use self::field_doc_id_facet_codec::FieldDocIdFacetCodec;
@@ -18,7 +18,7 @@ pub type FieldDocIdFacetF64Codec = FieldDocIdFacetCodec<OrderedF64Codec>;
 pub type FieldDocIdFacetStringCodec = FieldDocIdFacetCodec<StrRefCodec>;
 pub type FieldDocIdFacetIgnoreCodec = FieldDocIdFacetCodec<DecodeIgnore>;
 
-pub type FieldIdCodec = OwnedType<BEU16>;
+pub type FieldIdCodec = BEU16;
 
 /// Tries to split a slice in half at the given middle point,
 /// `None` if the slice is too short.
@@ -58,7 +58,7 @@ where
 {
     type EItem = FacetGroupKey<T::EItem>;
 
-    fn bytes_encode(value: &'a Self::EItem) -> Option<Cow<'a, [u8]>> {
+    fn bytes_encode(value: &'a Self::EItem) -> Result<Cow<'a, [u8]>, BoxedError> {
         let mut v = vec![];
         v.extend_from_slice(&value.field_id.to_be_bytes());
         v.extend_from_slice(&[value.level]);
@@ -66,7 +66,7 @@ where
         let bound = T::bytes_encode(&value.left_bound)?;
         v.extend_from_slice(&bound);
 
-        Some(Cow::Owned(v))
+        Ok(Cow::Owned(v))
     }
 }
 impl<'a, T> heed::BytesDecode<'a> for FacetGroupKeyCodec<T>
@@ -75,11 +75,11 @@ where
 {
     type DItem = FacetGroupKey<T::DItem>;
 
-    fn bytes_decode(bytes: &'a [u8]) -> Option<Self::DItem> {
-        let fid = u16::from_be_bytes(<[u8; 2]>::try_from(&bytes[0..=1]).ok()?);
+    fn bytes_decode(bytes: &'a [u8]) -> Result<Self::DItem, BoxedError> {
+        let fid = u16::from_be_bytes(<[u8; 2]>::try_from(&bytes[0..=1])?);
         let level = bytes[2];
         let bound = T::bytes_decode(&bytes[3..])?;
-        Some(FacetGroupKey { field_id: fid, level, left_bound: bound })
+        Ok(FacetGroupKey { field_id: fid, level, left_bound: bound })
     }
 }
 
@@ -87,17 +87,17 @@ pub struct FacetGroupValueCodec;
 impl<'a> heed::BytesEncode<'a> for FacetGroupValueCodec {
     type EItem = FacetGroupValue;
 
-    fn bytes_encode(value: &'a Self::EItem) -> Option<Cow<'a, [u8]>> {
+    fn bytes_encode(value: &'a Self::EItem) -> Result<Cow<'a, [u8]>, BoxedError> {
         let mut v = vec![value.size];
         CboRoaringBitmapCodec::serialize_into(&value.bitmap, &mut v);
-        Some(Cow::Owned(v))
+        Ok(Cow::Owned(v))
     }
 }
 impl<'a> heed::BytesDecode<'a> for FacetGroupValueCodec {
     type DItem = FacetGroupValue;
-    fn bytes_decode(bytes: &'a [u8]) -> Option<Self::DItem> {
+    fn bytes_decode(bytes: &'a [u8]) -> Result<Self::DItem, BoxedError> {
         let size = bytes[0];
-        let bitmap = CboRoaringBitmapCodec::deserialize_from(&bytes[1..]).ok()?;
-        Some(FacetGroupValue { size, bitmap })
+        let bitmap = CboRoaringBitmapCodec::deserialize_from(&bytes[1..])?;
+        Ok(FacetGroupValue { size, bitmap })
     }
 }

milli/src/heed_codec/facet/ordered_f64_codec.rs

@@ -1,37 +1,45 @@
 use std::borrow::Cow;
 use std::convert::TryInto;
 
-use heed::BytesDecode;
+use heed::{BoxedError, BytesDecode};
+use thiserror::Error;
 
 use crate::facet::value_encoding::f64_into_bytes;
+use crate::heed_codec::SliceTooShortError;
 
 pub struct OrderedF64Codec;
 
 impl<'a> BytesDecode<'a> for OrderedF64Codec {
     type DItem = f64;
 
-    fn bytes_decode(bytes: &'a [u8]) -> Option<Self::DItem> {
+    fn bytes_decode(bytes: &'a [u8]) -> Result<Self::DItem, BoxedError> {
         if bytes.len() < 16 {
-            return None;
+            Err(SliceTooShortError.into())
+        } else {
+            bytes[8..].try_into().map(f64::from_be_bytes).map_err(Into::into)
         }
-        let f = bytes[8..].try_into().ok().map(f64::from_be_bytes)?;
-        Some(f)
     }
 }
 
 impl heed::BytesEncode<'_> for OrderedF64Codec {
     type EItem = f64;
 
-    fn bytes_encode(f: &Self::EItem) -> Option<Cow<[u8]>> {
+    fn bytes_encode(f: &Self::EItem) -> Result<Cow<[u8]>, BoxedError> {
         let mut buffer = [0u8; 16];
 
         // write the globally ordered float
-        let bytes = f64_into_bytes(*f)?;
+        let bytes = f64_into_bytes(*f).ok_or(InvalidGloballyOrderedFloatError { float: *f })?;
         buffer[..8].copy_from_slice(&bytes[..]);
         // Then the f64 value just to be able to read it back
         let bytes = f.to_be_bytes();
         buffer[8..16].copy_from_slice(&bytes[..]);
 
-        Some(Cow::Owned(buffer.to_vec()))
+        Ok(Cow::Owned(buffer.to_vec()))
     }
 }
+
+#[derive(Error, Debug)]
+#[error("the float {float} cannot be converted to a globally ordered representation")]
+pub struct InvalidGloballyOrderedFloatError {
+    float: f64,
+}

milli/src/heed_codec/field_id_word_count_codec.rs

@@ -1,5 +1,8 @@
 use std::borrow::Cow;
 
+use heed::BoxedError;
+
+use super::SliceTooShortError;
 use crate::{try_split_array_at, FieldId};
 
 pub struct FieldIdWordCountCodec;
@@ -7,21 +10,21 @@ pub struct FieldIdWordCountCodec;
 impl<'a> heed::BytesDecode<'a> for FieldIdWordCountCodec {
     type DItem = (FieldId, u8);
 
-    fn bytes_decode(bytes: &'a [u8]) -> Option<Self::DItem> {
-        let (field_id_bytes, bytes) = try_split_array_at(bytes)?;
+    fn bytes_decode(bytes: &'a [u8]) -> Result<Self::DItem, BoxedError> {
+        let (field_id_bytes, bytes) = try_split_array_at(bytes).ok_or(SliceTooShortError)?;
         let field_id = u16::from_be_bytes(field_id_bytes);
-        let ([word_count], _nothing) = try_split_array_at(bytes)?;
-        Some((field_id, word_count))
+        let ([word_count], _nothing) = try_split_array_at(bytes).ok_or(SliceTooShortError)?;
+        Ok((field_id, word_count))
     }
 }
 
 impl<'a> heed::BytesEncode<'a> for FieldIdWordCountCodec {
     type EItem = (FieldId, u8);
 
-    fn bytes_encode((field_id, word_count): &Self::EItem) -> Option<Cow<[u8]>> {
+    fn bytes_encode((field_id, word_count): &Self::EItem) -> Result<Cow<[u8]>, BoxedError> {
         let mut bytes = Vec::with_capacity(2 + 1);
         bytes.extend_from_slice(&field_id.to_be_bytes());
         bytes.push(*word_count);
-        Some(Cow::Owned(bytes))
+        Ok(Cow::Owned(bytes))
     }
 }

milli/src/heed_codec/fst_set_codec.rs

@@ -1,7 +1,7 @@
 use std::borrow::Cow;
 
 use fst::Set;
-use heed::{BytesDecode, BytesEncode};
+use heed::{BoxedError, BytesDecode, BytesEncode};
 
 /// A codec for values of type `Set<&[u8]>`.
 pub struct FstSetCodec;
@@ -9,15 +9,15 @@ pub struct FstSetCodec;
 impl<'a> BytesEncode<'a> for FstSetCodec {
     type EItem = Set<Vec<u8>>;
 
-    fn bytes_encode(item: &'a Self::EItem) -> Option<Cow<'a, [u8]>> {
-        Some(Cow::Borrowed(item.as_fst().as_bytes()))
+    fn bytes_encode(item: &'a Self::EItem) -> Result<Cow<'a, [u8]>, BoxedError> {
+        Ok(Cow::Borrowed(item.as_fst().as_bytes()))
     }
 }
 
 impl<'a> BytesDecode<'a> for FstSetCodec {
     type DItem = Set<&'a [u8]>;
 
-    fn bytes_decode(bytes: &'a [u8]) -> Option<Self::DItem> {
-        Set::new(bytes).ok()
+    fn bytes_decode(bytes: &'a [u8]) -> Result<Self::DItem, BoxedError> {
+        Set::new(bytes).map_err(Into::into)
     }
 }

milli/src/heed_codec/mod.rs

@@ -12,8 +12,10 @@ mod str_beu32_codec;
 mod str_ref;
 mod str_str_u8_codec;
 
-pub use byte_slice_ref::ByteSliceRefCodec;
+pub use byte_slice_ref::BytesRefCodec;
+use heed::BoxedError;
 pub use str_ref::StrRefCodec;
+use thiserror::Error;
 
 pub use self::beu16_str_codec::BEU16StrCodec;
 pub use self::beu32_str_codec::BEU32StrCodec;
@@ -31,5 +33,9 @@ pub use self::str_str_u8_codec::{U8StrStrCodec, UncheckedU8StrStrCodec};
 pub trait BytesDecodeOwned {
     type DItem;
 
-    fn bytes_decode_owned(bytes: &[u8]) -> Option<Self::DItem>;
+    fn bytes_decode_owned(bytes: &[u8]) -> Result<Self::DItem, BoxedError>;
 }
+
+#[derive(Error, Debug)]
+#[error("the slice is too short")]
+pub struct SliceTooShortError;

milli/src/heed_codec/obkv_codec.rs

@@ -1,5 +1,6 @@
 use std::borrow::Cow;
 
+use heed::BoxedError;
 use obkv::{KvReaderU16, KvWriterU16};
 
 pub struct ObkvCodec;
@@ -7,15 +8,15 @@ pub struct ObkvCodec;
 impl<'a> heed::BytesDecode<'a> for ObkvCodec {
     type DItem = KvReaderU16<'a>;
 
-    fn bytes_decode(bytes: &'a [u8]) -> Option<Self::DItem> {
-        Some(KvReaderU16::new(bytes))
+    fn bytes_decode(bytes: &'a [u8]) -> Result<Self::DItem, BoxedError> {
+        Ok(KvReaderU16::new(bytes))
     }
 }
 
 impl heed::BytesEncode<'_> for ObkvCodec {
     type EItem = KvWriterU16<Vec<u8>>;
 
-    fn bytes_encode(item: &Self::EItem) -> Option<Cow<[u8]>> {
-        item.clone().into_inner().map(Cow::Owned).ok()
+    fn bytes_encode(item: &Self::EItem) -> Result<Cow<[u8]>, BoxedError> {
+        item.clone().into_inner().map(Cow::Owned).map_err(Into::into)
     }
 }

milli/src/heed_codec/roaring_bitmap/bo_roaring_bitmap_codec.rs

@@ -2,7 +2,7 @@ use std::borrow::Cow;
 use std::convert::TryInto;
 use std::mem::size_of;
 
-use heed::BytesDecode;
+use heed::{BoxedError, BytesDecode};
 use roaring::RoaringBitmap;
 
 use crate::heed_codec::BytesDecodeOwned;
@@ -19,22 +19,22 @@ impl BoRoaringBitmapCodec {
 impl BytesDecode<'_> for BoRoaringBitmapCodec {
     type DItem = RoaringBitmap;
 
-    fn bytes_decode(bytes: &[u8]) -> Option<Self::DItem> {
+    fn bytes_decode(bytes: &[u8]) -> Result<Self::DItem, BoxedError> {
         let mut bitmap = RoaringBitmap::new();
 
         for chunk in bytes.chunks(size_of::<u32>()) {
-            let bytes = chunk.try_into().ok()?;
+            let bytes = chunk.try_into()?;
             bitmap.push(u32::from_ne_bytes(bytes));
         }
 
-        Some(bitmap)
+        Ok(bitmap)
     }
 }
 
 impl BytesDecodeOwned for BoRoaringBitmapCodec {
     type DItem = RoaringBitmap;
 
-    fn bytes_decode_owned(bytes: &[u8]) -> Option<Self::DItem> {
+    fn bytes_decode_owned(bytes: &[u8]) -> Result<Self::DItem, BoxedError> {
         Self::bytes_decode(bytes)
     }
 }
@@ -42,9 +42,9 @@ impl BytesDecodeOwned for BoRoaringBitmapCodec {
 impl heed::BytesEncode<'_> for BoRoaringBitmapCodec {
     type EItem = RoaringBitmap;
 
-    fn bytes_encode(item: &Self::EItem) -> Option<Cow<[u8]>> {
+    fn bytes_encode(item: &Self::EItem) -> Result<Cow<[u8]>, BoxedError> {
         let mut out = Vec::new();
         BoRoaringBitmapCodec::serialize_into(item, &mut out);
-        Some(Cow::Owned(out))
+        Ok(Cow::Owned(out))
     }
 }

milli/src/heed_codec/roaring_bitmap/cbo_roaring_bitmap_codec.rs

@@ -3,6 +3,7 @@ use std::io;
 use std::mem::size_of;
 
 use byteorder::{NativeEndian, ReadBytesExt, WriteBytesExt};
+use heed::BoxedError;
 use roaring::RoaringBitmap;
 
 use crate::heed_codec::BytesDecodeOwned;
@@ -132,26 +133,26 @@ impl CboRoaringBitmapCodec {
 impl heed::BytesDecode<'_> for CboRoaringBitmapCodec {
     type DItem = RoaringBitmap;
 
-    fn bytes_decode(bytes: &[u8]) -> Option<Self::DItem> {
-        Self::deserialize_from(bytes).ok()
+    fn bytes_decode(bytes: &[u8]) -> Result<Self::DItem, BoxedError> {
+        Self::deserialize_from(bytes).map_err(Into::into)
     }
 }
 
 impl BytesDecodeOwned for CboRoaringBitmapCodec {
     type DItem = RoaringBitmap;
 
-    fn bytes_decode_owned(bytes: &[u8]) -> Option<Self::DItem> {
-        Self::deserialize_from(bytes).ok()
+    fn bytes_decode_owned(bytes: &[u8]) -> Result<Self::DItem, BoxedError> {
+        Self::deserialize_from(bytes).map_err(Into::into)
     }
 }
 
 impl heed::BytesEncode<'_> for CboRoaringBitmapCodec {
     type EItem = RoaringBitmap;
 
-    fn bytes_encode(item: &Self::EItem) -> Option<Cow<[u8]>> {
+    fn bytes_encode(item: &Self::EItem) -> Result<Cow<[u8]>, BoxedError> {
         let mut vec = Vec::with_capacity(Self::serialized_size(item));
         Self::serialize_into(item, &mut vec);
-        Some(Cow::Owned(vec))
+        Ok(Cow::Owned(vec))
     }
 }
 

milli/src/heed_codec/roaring_bitmap/roaring_bitmap_codec.rs

@@ -1,5 +1,6 @@
 use std::borrow::Cow;
 
+use heed::BoxedError;
 use roaring::RoaringBitmap;
 
 use crate::heed_codec::BytesDecodeOwned;
@@ -9,25 +10,25 @@ pub struct RoaringBitmapCodec;
 impl heed::BytesDecode<'_> for RoaringBitmapCodec {
     type DItem = RoaringBitmap;
 
-    fn bytes_decode(bytes: &[u8]) -> Option<Self::DItem> {
-        RoaringBitmap::deserialize_unchecked_from(bytes).ok()
+    fn bytes_decode(bytes: &[u8]) -> Result<Self::DItem, BoxedError> {
+        RoaringBitmap::deserialize_unchecked_from(bytes).map_err(Into::into)
     }
 }
 
 impl BytesDecodeOwned for RoaringBitmapCodec {
     type DItem = RoaringBitmap;
 
-    fn bytes_decode_owned(bytes: &[u8]) -> Option<Self::DItem> {
-        RoaringBitmap::deserialize_from(bytes).ok()
+    fn bytes_decode_owned(bytes: &[u8]) -> Result<Self::DItem, BoxedError> {
+        RoaringBitmap::deserialize_from(bytes).map_err(Into::into)
     }
 }
 
 impl heed::BytesEncode<'_> for RoaringBitmapCodec {
     type EItem = RoaringBitmap;
 
-    fn bytes_encode(item: &Self::EItem) -> Option<Cow<[u8]>> {
+    fn bytes_encode(item: &Self::EItem) -> Result<Cow<[u8]>, BoxedError> {
         let mut bytes = Vec::with_capacity(item.serialized_size());
-        item.serialize_into(&mut bytes).ok()?;
-        Some(Cow::Owned(bytes))
+        item.serialize_into(&mut bytes)?;
+        Ok(Cow::Owned(bytes))
     }
 }

milli/src/heed_codec/roaring_bitmap_length/bo_roaring_bitmap_len_codec.rs

@@ -1,6 +1,6 @@
 use std::mem;
 
-use heed::BytesDecode;
+use heed::{BoxedError, BytesDecode};
 
 use crate::heed_codec::BytesDecodeOwned;
 
@@ -9,15 +9,15 @@ pub struct BoRoaringBitmapLenCodec;
 impl BytesDecode<'_> for BoRoaringBitmapLenCodec {
     type DItem = u64;
 
-    fn bytes_decode(bytes: &[u8]) -> Option<Self::DItem> {
-        Some((bytes.len() / mem::size_of::<u32>()) as u64)
+    fn bytes_decode(bytes: &[u8]) -> Result<Self::DItem, BoxedError> {
+        Ok((bytes.len() / mem::size_of::<u32>()) as u64)
     }
 }
 
 impl BytesDecodeOwned for BoRoaringBitmapLenCodec {
     type DItem = u64;
 
-    fn bytes_decode_owned(bytes: &[u8]) -> Option<Self::DItem> {
+    fn bytes_decode_owned(bytes: &[u8]) -> Result<Self::DItem, BoxedError> {
         Self::bytes_decode(bytes)
     }
 }

milli/src/heed_codec/roaring_bitmap_length/cbo_roaring_bitmap_len_codec.rs

@@ -1,6 +1,6 @@
 use std::mem;
 
-use heed::BytesDecode;
+use heed::{BoxedError, BytesDecode};
 
 use super::{BoRoaringBitmapLenCodec, RoaringBitmapLenCodec};
 use crate::heed_codec::roaring_bitmap::cbo_roaring_bitmap_codec::THRESHOLD;
@@ -11,7 +11,7 @@ pub struct CboRoaringBitmapLenCodec;
 impl BytesDecode<'_> for CboRoaringBitmapLenCodec {
     type DItem = u64;
 
-    fn bytes_decode(bytes: &[u8]) -> Option<Self::DItem> {
+    fn bytes_decode(bytes: &[u8]) -> Result<Self::DItem, BoxedError> {
         if bytes.len() <= THRESHOLD * mem::size_of::<u32>() {
             // If there is threshold or less than threshold integers that can fit into this array
             // of bytes it means that we used the ByteOrder codec serializer.
@@ -27,7 +27,7 @@ impl BytesDecode<'_> for CboRoaringBitmapLenCodec {
 impl BytesDecodeOwned for CboRoaringBitmapLenCodec {
     type DItem = u64;
 
-    fn bytes_decode_owned(bytes: &[u8]) -> Option<Self::DItem> {
+    fn bytes_decode_owned(bytes: &[u8]) -> Result<Self::DItem, BoxedError> {
         Self::bytes_decode(bytes)
     }
 }

milli/src/heed_codec/roaring_bitmap_length/roaring_bitmap_len_codec.rs

@@ -2,6 +2,7 @@ use std::io::{self, BufRead, Read};
 use std::mem;
 
 use byteorder::{LittleEndian, ReadBytesExt};
+use heed::BoxedError;
 
 use crate::heed_codec::BytesDecodeOwned;
 
@@ -56,16 +57,16 @@ impl RoaringBitmapLenCodec {
 impl heed::BytesDecode<'_> for RoaringBitmapLenCodec {
     type DItem = u64;
 
-    fn bytes_decode(bytes: &[u8]) -> Option<Self::DItem> {
-        RoaringBitmapLenCodec::deserialize_from_slice(bytes).ok()
+    fn bytes_decode(bytes: &[u8]) -> Result<Self::DItem, BoxedError> {
+        RoaringBitmapLenCodec::deserialize_from_slice(bytes).map_err(Into::into)
     }
 }
 
 impl BytesDecodeOwned for RoaringBitmapLenCodec {
     type DItem = u64;
 
-    fn bytes_decode_owned(bytes: &[u8]) -> Option<Self::DItem> {
-        RoaringBitmapLenCodec::deserialize_from_slice(bytes).ok()
+    fn bytes_decode_owned(bytes: &[u8]) -> Result<Self::DItem, BoxedError> {
+        RoaringBitmapLenCodec::deserialize_from_slice(bytes).map_err(Into::into)
     }
 }
 

milli/src/heed_codec/script_language_codec.rs

@@ -1,30 +1,31 @@
 use std::borrow::Cow;
+use std::ffi::CStr;
 use std::str;
 
 use charabia::{Language, Script};
+use heed::BoxedError;
 
 pub struct ScriptLanguageCodec;
 
 impl<'a> heed::BytesDecode<'a> for ScriptLanguageCodec {
     type DItem = (Script, Language);
 
-    fn bytes_decode(bytes: &'a [u8]) -> Option<Self::DItem> {
-        let sep = bytes.iter().position(|b| *b == 0)?;
-        let (s_bytes, l_bytes) = bytes.split_at(sep);
-        let script = str::from_utf8(s_bytes).ok()?;
+    fn bytes_decode(bytes: &'a [u8]) -> Result<Self::DItem, BoxedError> {
+        let cstr = CStr::from_bytes_until_nul(bytes)?;
+        let script = cstr.to_str()?;
         let script_name = Script::from_name(script);
-        let lan = str::from_utf8(l_bytes).ok()?;
         // skip '\0' byte between the two strings.
-        let lan_name = Language::from_name(&lan[1..]);
+        let lan = str::from_utf8(&bytes[script.len() + 1..])?;
+        let lan_name = Language::from_name(lan);
 
-        Some((script_name, lan_name))
+        Ok((script_name, lan_name))
     }
 }
 
 impl<'a> heed::BytesEncode<'a> for ScriptLanguageCodec {
     type EItem = (Script, Language);
 
-    fn bytes_encode((script, lan): &Self::EItem) -> Option<Cow<[u8]>> {
+    fn bytes_encode((script, lan): &Self::EItem) -> Result<Cow<[u8]>, BoxedError> {
         let script_name = script.name().as_bytes();
         let lan_name = lan.name().as_bytes();
 
@@ -33,6 +34,6 @@ impl<'a> heed::BytesEncode<'a> for ScriptLanguageCodec {
         bytes.push(0);
         bytes.extend_from_slice(lan_name);
 
-        Some(Cow::Owned(bytes))
+        Ok(Cow::Owned(bytes))
     }
 }

milli/src/heed_codec/str_beu32_codec.rs

@@ -3,37 +3,41 @@ use std::convert::TryInto;
 use std::mem::size_of;
 use std::str;
 
+use heed::BoxedError;
+
+use super::SliceTooShortError;
+
 pub struct StrBEU32Codec;
 
 impl<'a> heed::BytesDecode<'a> for StrBEU32Codec {
     type DItem = (&'a str, u32);
 
-    fn bytes_decode(bytes: &'a [u8]) -> Option<Self::DItem> {
+    fn bytes_decode(bytes: &'a [u8]) -> Result<Self::DItem, BoxedError> {
         let footer_len = size_of::<u32>();
 
         if bytes.len() < footer_len {
-            return None;
+            return Err(SliceTooShortError.into());
         }
 
         let (word, bytes) = bytes.split_at(bytes.len() - footer_len);
-        let word = str::from_utf8(word).ok()?;
-        let pos = bytes.try_into().map(u32::from_be_bytes).ok()?;
+        let word = str::from_utf8(word)?;
+        let pos = bytes.try_into().map(u32::from_be_bytes)?;
 
-        Some((word, pos))
+        Ok((word, pos))
     }
 }
 
 impl<'a> heed::BytesEncode<'a> for StrBEU32Codec {
     type EItem = (&'a str, u32);
 
-    fn bytes_encode((word, pos): &Self::EItem) -> Option<Cow<[u8]>> {
+    fn bytes_encode((word, pos): &Self::EItem) -> Result<Cow<[u8]>, BoxedError> {
         let pos = pos.to_be_bytes();
 
         let mut bytes = Vec::with_capacity(word.len() + pos.len());
         bytes.extend_from_slice(word.as_bytes());
         bytes.extend_from_slice(&pos[..]);
 
-        Some(Cow::Owned(bytes))
+        Ok(Cow::Owned(bytes))
     }
 }
 
@@ -42,26 +46,27 @@ pub struct StrBEU16Codec;
 impl<'a> heed::BytesDecode<'a> for StrBEU16Codec {
     type DItem = (&'a str, u16);
 
-    fn bytes_decode(bytes: &'a [u8]) -> Option<Self::DItem> {
+    fn bytes_decode(bytes: &'a [u8]) -> Result<Self::DItem, BoxedError> {
         let footer_len = size_of::<u16>();
 
         if bytes.len() < footer_len + 1 {
-            return None;
+            return Err(SliceTooShortError.into());
         }
 
         let (word_plus_nul_byte, bytes) = bytes.split_at(bytes.len() - footer_len);
-        let (_, word) = word_plus_nul_byte.split_last()?;
-        let word = str::from_utf8(word).ok()?;
-        let pos = bytes.try_into().map(u16::from_be_bytes).ok()?;
+        // unwrap: we just checked the footer + 1 above.
+        let (_, word) = word_plus_nul_byte.split_last().unwrap();
+        let word = str::from_utf8(word)?;
+        let pos = bytes.try_into().map(u16::from_be_bytes)?;
 
-        Some((word, pos))
+        Ok((word, pos))
     }
 }
 
 impl<'a> heed::BytesEncode<'a> for StrBEU16Codec {
     type EItem = (&'a str, u16);
 
-    fn bytes_encode((word, pos): &Self::EItem) -> Option<Cow<[u8]>> {
+    fn bytes_encode((word, pos): &Self::EItem) -> Result<Cow<[u8]>, BoxedError> {
         let pos = pos.to_be_bytes();
 
         let mut bytes = Vec::with_capacity(word.len() + 1 + pos.len());
@@ -69,6 +74,6 @@ impl<'a> heed::BytesEncode<'a> for StrBEU16Codec {
         bytes.push(0);
         bytes.extend_from_slice(&pos[..]);
 
-        Some(Cow::Owned(bytes))
+        Ok(Cow::Owned(bytes))
     }
 }

milli/src/heed_codec/str_ref.rs

@@ -1,6 +1,6 @@
 use std::borrow::Cow;
 
-use heed::{BytesDecode, BytesEncode};
+use heed::{BoxedError, BytesDecode, BytesEncode};
 
 /// A codec for values of type `&str`. Unlike `Str`, its `EItem` and `DItem` associated
 /// types are equivalent (= `&'a str`) and these values can reside within another structure.
@@ -8,15 +8,14 @@ pub struct StrRefCodec;
 impl<'a> BytesEncode<'a> for StrRefCodec {
     type EItem = &'a str;
 
-    fn bytes_encode(item: &'a &'a str) -> Option<Cow<'a, [u8]>> {
-        Some(Cow::Borrowed(item.as_bytes()))
+    fn bytes_encode(item: &'a &'a str) -> Result<Cow<'a, [u8]>, BoxedError> {
+        Ok(Cow::Borrowed(item.as_bytes()))
     }
 }
 impl<'a> BytesDecode<'a> for StrRefCodec {
     type DItem = &'a str;
 
-    fn bytes_decode(bytes: &'a [u8]) -> Option<Self::DItem> {
-        let s = std::str::from_utf8(bytes).ok()?;
-        Some(s)
+    fn bytes_decode(bytes: &'a [u8]) -> Result<Self::DItem, BoxedError> {
+        std::str::from_utf8(bytes).map_err(Into::into)
     }
 }

milli/src/heed_codec/str_str_u8_codec.rs

@@ -1,32 +1,36 @@
 use std::borrow::Cow;
+use std::ffi::CStr;
 use std::str;
 
+use heed::BoxedError;
+
+use super::SliceTooShortError;
+
 pub struct U8StrStrCodec;
 
 impl<'a> heed::BytesDecode<'a> for U8StrStrCodec {
     type DItem = (u8, &'a str, &'a str);
 
-    fn bytes_decode(bytes: &'a [u8]) -> Option<Self::DItem> {
-        let (n, bytes) = bytes.split_first()?;
-        let s1_end = bytes.iter().position(|b| *b == 0)?;
-        let (s1_bytes, rest) = bytes.split_at(s1_end);
-        let s2_bytes = &rest[1..];
-        let s1 = str::from_utf8(s1_bytes).ok()?;
-        let s2 = str::from_utf8(s2_bytes).ok()?;
-        Some((*n, s1, s2))
+    fn bytes_decode(bytes: &'a [u8]) -> Result<Self::DItem, BoxedError> {
+        let (n, bytes) = bytes.split_first().ok_or(SliceTooShortError)?;
+        let cstr = CStr::from_bytes_until_nul(bytes)?;
+        let s1 = cstr.to_str()?;
+        // skip '\0' byte between the two strings.
+        let s2 = str::from_utf8(&bytes[s1.len() + 1..])?;
+        Ok((*n, s1, s2))
     }
 }
 
 impl<'a> heed::BytesEncode<'a> for U8StrStrCodec {
     type EItem = (u8, &'a str, &'a str);
 
-    fn bytes_encode((n, s1, s2): &Self::EItem) -> Option<Cow<[u8]>> {
+    fn bytes_encode((n, s1, s2): &Self::EItem) -> Result<Cow<[u8]>, BoxedError> {
         let mut bytes = Vec::with_capacity(s1.len() + s2.len() + 1);
         bytes.push(*n);
         bytes.extend_from_slice(s1.as_bytes());
         bytes.push(0);
         bytes.extend_from_slice(s2.as_bytes());
-        Some(Cow::Owned(bytes))
+        Ok(Cow::Owned(bytes))
     }
 }
 pub struct UncheckedU8StrStrCodec;
@@ -34,24 +38,25 @@ pub struct UncheckedU8StrStrCodec;
 impl<'a> heed::BytesDecode<'a> for UncheckedU8StrStrCodec {
     type DItem = (u8, &'a [u8], &'a [u8]);
 
-    fn bytes_decode(bytes: &'a [u8]) -> Option<Self::DItem> {
-        let (n, bytes) = bytes.split_first()?;
-        let s1_end = bytes.iter().position(|b| *b == 0)?;
-        let (s1_bytes, rest) = bytes.split_at(s1_end);
-        let s2_bytes = &rest[1..];
-        Some((*n, s1_bytes, s2_bytes))
+    fn bytes_decode(bytes: &'a [u8]) -> Result<Self::DItem, BoxedError> {
+        let (n, bytes) = bytes.split_first().ok_or(SliceTooShortError)?;
+        let cstr = CStr::from_bytes_until_nul(bytes)?;
+        let s1_bytes = cstr.to_bytes();
+        // skip '\0' byte between the two strings.
+        let s2_bytes = &bytes[s1_bytes.len() + 1..];
+        Ok((*n, s1_bytes, s2_bytes))
     }
 }
 
 impl<'a> heed::BytesEncode<'a> for UncheckedU8StrStrCodec {
     type EItem = (u8, &'a [u8], &'a [u8]);
 
-    fn bytes_encode((n, s1, s2): &Self::EItem) -> Option<Cow<[u8]>> {
+    fn bytes_encode((n, s1, s2): &Self::EItem) -> Result<Cow<[u8]>, BoxedError> {
         let mut bytes = Vec::with_capacity(s1.len() + s2.len() + 1);
         bytes.push(*n);
         bytes.extend_from_slice(s1);
         bytes.push(0);
         bytes.extend_from_slice(s2);
-        Some(Cow::Owned(bytes))
+        Ok(Cow::Owned(bytes))
     }
 }

milli/src/lib.rs

@@ -66,9 +66,9 @@ pub use self::search::{
 pub type Result<T> = std::result::Result<T, error::Error>;
 
 pub type Attribute = u32;
-pub type BEU16 = heed::zerocopy::U16<heed::byteorder::BE>;
-pub type BEU32 = heed::zerocopy::U32<heed::byteorder::BE>;
-pub type BEU64 = heed::zerocopy::U64<heed::byteorder::BE>;
+pub type BEU16 = heed::types::U16<heed::byteorder::BE>;
+pub type BEU32 = heed::types::U32<heed::byteorder::BE>;
+pub type BEU64 = heed::types::U64<heed::byteorder::BE>;
 pub type DocumentId = u32;
 pub type FastMap4<K, V> = HashMap<K, V, BuildHasherDefault<FxHasher32>>;
 pub type FastMap8<K, V> = HashMap<K, V, BuildHasherDefault<FxHasher64>>;

milli/src/proximity.rs

@@ -1,5 +1,7 @@
 use std::cmp;
 
+use serde::{Deserialize, Serialize};
+
 use crate::{relative_from_absolute_position, Position};
 
 pub const MAX_DISTANCE: u32 = 4;
@@ -25,3 +27,11 @@ pub fn positions_proximity(lhs: Position, rhs: Position) -> u32 {
 pub fn path_proximity(path: &[Position]) -> u32 {
     path.windows(2).map(|w| positions_proximity(w[0], w[1])).sum::<u32>()
 }
+
+#[derive(Debug, Clone, Copy, Serialize, Deserialize, PartialEq, Eq, Default)]
+#[serde(rename_all = "camelCase")]
+pub enum ProximityPrecision {
+    #[default]
+    WordScale,
+    AttributeScale,
+}

milli/src/search/facet/facet_distribution.rs

@@ -2,7 +2,7 @@ use std::collections::{BTreeMap, HashMap, HashSet};
 use std::ops::ControlFlow;
 use std::{fmt, mem};
 
-use heed::types::ByteSlice;
+use heed::types::Bytes;
 use heed::BytesDecode;
 use indexmap::IndexMap;
 use roaring::RoaringBitmap;
@@ -13,7 +13,7 @@ use crate::facet::FacetType;
 use crate::heed_codec::facet::{
     FacetGroupKeyCodec, FieldDocIdFacetF64Codec, FieldDocIdFacetStringCodec, OrderedF64Codec,
 };
-use crate::heed_codec::{ByteSliceRefCodec, StrRefCodec};
+use crate::heed_codec::{BytesRefCodec, StrRefCodec};
 use crate::search::facet::facet_distribution_iter::{
     count_iterate_over_facet_distribution, lexicographically_iterate_over_facet_distribution,
 };
@@ -105,7 +105,7 @@ impl<'a> FacetDistribution<'a> {
                     key_buffer.truncate(mem::size_of::<FieldId>());
                     key_buffer.extend_from_slice(&docid.to_be_bytes());
                     let iter = db
-                        .remap_key_type::<ByteSlice>()
+                        .remap_key_type::<Bytes>()
                         .prefix_iter(self.rtxn, &key_buffer)?
                         .remap_key_type::<FieldDocIdFacetF64Codec>();
 
@@ -129,7 +129,7 @@ impl<'a> FacetDistribution<'a> {
                     key_buffer.truncate(mem::size_of::<FieldId>());
                     key_buffer.extend_from_slice(&docid.to_be_bytes());
                     let iter = db
-                        .remap_key_type::<ByteSlice>()
+                        .remap_key_type::<Bytes>()
                         .prefix_iter(self.rtxn, &key_buffer)?
                         .remap_key_type::<FieldDocIdFacetStringCodec>();
 
@@ -172,9 +172,7 @@ impl<'a> FacetDistribution<'a> {
 
         search_function(
             self.rtxn,
-            self.index
-                .facet_id_f64_docids
-                .remap_key_type::<FacetGroupKeyCodec<ByteSliceRefCodec>>(),
+            self.index.facet_id_f64_docids.remap_key_type::<FacetGroupKeyCodec<BytesRefCodec>>(),
             field_id,
             candidates,
             |facet_key, nbr_docids, _| {
@@ -203,9 +201,7 @@ impl<'a> FacetDistribution<'a> {
 
         search_function(
             self.rtxn,
-            self.index
-                .facet_id_string_docids
-                .remap_key_type::<FacetGroupKeyCodec<ByteSliceRefCodec>>(),
+            self.index.facet_id_string_docids.remap_key_type::<FacetGroupKeyCodec<BytesRefCodec>>(),
             field_id,
             candidates,
             |facet_key, nbr_docids, any_docid| {

milli/src/search/facet/facet_distribution_iter.rs

@@ -7,7 +7,7 @@ use roaring::RoaringBitmap;
 
 use super::{get_first_facet_value, get_highest_level};
 use crate::heed_codec::facet::{FacetGroupKey, FacetGroupKeyCodec, FacetGroupValueCodec};
-use crate::heed_codec::ByteSliceRefCodec;
+use crate::heed_codec::BytesRefCodec;
 use crate::DocumentId;
 
 /// Call the given closure on the facet distribution of the candidate documents.
@@ -23,7 +23,7 @@ use crate::DocumentId;
 /// keep iterating over the different facet values or stop.
 pub fn lexicographically_iterate_over_facet_distribution<'t, CB>(
     rtxn: &'t heed::RoTxn<'t>,
-    db: heed::Database<FacetGroupKeyCodec<ByteSliceRefCodec>, FacetGroupValueCodec>,
+    db: heed::Database<FacetGroupKeyCodec<BytesRefCodec>, FacetGroupValueCodec>,
     field_id: u16,
     candidates: &RoaringBitmap,
     callback: CB,
@@ -34,11 +34,11 @@ where
     let mut fd = LexicographicFacetDistribution { rtxn, db, field_id, callback };
     let highest_level = get_highest_level(
         rtxn,
-        db.remap_key_type::<FacetGroupKeyCodec<ByteSliceRefCodec>>(),
+        db.remap_key_type::<FacetGroupKeyCodec<BytesRefCodec>>(),
         field_id,
     )?;
 
-    if let Some(first_bound) = get_first_facet_value::<ByteSliceRefCodec>(rtxn, db, field_id)? {
+    if let Some(first_bound) = get_first_facet_value::<BytesRefCodec>(rtxn, db, field_id)? {
         fd.iterate(candidates, highest_level, first_bound, usize::MAX)?;
         Ok(())
     } else {
@@ -48,7 +48,7 @@ where
 
 pub fn count_iterate_over_facet_distribution<'t, CB>(
     rtxn: &'t heed::RoTxn<'t>,
-    db: heed::Database<FacetGroupKeyCodec<ByteSliceRefCodec>, FacetGroupValueCodec>,
+    db: heed::Database<FacetGroupKeyCodec<BytesRefCodec>, FacetGroupValueCodec>,
     field_id: u16,
     candidates: &RoaringBitmap,
     mut callback: CB,
@@ -77,11 +77,11 @@ where
     let mut heap = BinaryHeap::new();
     let highest_level = get_highest_level(
         rtxn,
-        db.remap_key_type::<FacetGroupKeyCodec<ByteSliceRefCodec>>(),
+        db.remap_key_type::<FacetGroupKeyCodec<BytesRefCodec>>(),
         field_id,
     )?;
 
-    if let Some(first_bound) = get_first_facet_value::<ByteSliceRefCodec>(rtxn, db, field_id)? {
+    if let Some(first_bound) = get_first_facet_value::<BytesRefCodec>(rtxn, db, field_id)? {
         // We first fill the heap with values from the highest level
         let starting_key =
             FacetGroupKey { field_id, level: highest_level, left_bound: first_bound };
@@ -146,7 +146,7 @@ where
     CB: FnMut(&'t [u8], u64, DocumentId) -> Result<ControlFlow<()>>,
 {
     rtxn: &'t heed::RoTxn<'t>,
-    db: heed::Database<FacetGroupKeyCodec<ByteSliceRefCodec>, FacetGroupValueCodec>,
+    db: heed::Database<FacetGroupKeyCodec<BytesRefCodec>, FacetGroupValueCodec>,
     field_id: u16,
     callback: CB,
 }

milli/src/search/facet/facet_range_search.rs

@@ -5,7 +5,7 @@ use roaring::RoaringBitmap;
 
 use super::{get_first_facet_value, get_highest_level, get_last_facet_value};
 use crate::heed_codec::facet::{FacetGroupKey, FacetGroupKeyCodec, FacetGroupValueCodec};
-use crate::heed_codec::ByteSliceRefCodec;
+use crate::heed_codec::BytesRefCodec;
 use crate::Result;
 
 /// Find all the document ids for which the given field contains a value contained within
@@ -25,11 +25,11 @@ where
     let inner;
     let left = match left {
         Bound::Included(left) => {
-            inner = BoundCodec::bytes_encode(left).ok_or(heed::Error::Encoding)?;
+            inner = BoundCodec::bytes_encode(left).map_err(heed::Error::Encoding)?;
             Bound::Included(inner.as_ref())
         }
         Bound::Excluded(left) => {
-            inner = BoundCodec::bytes_encode(left).ok_or(heed::Error::Encoding)?;
+            inner = BoundCodec::bytes_encode(left).map_err(heed::Error::Encoding)?;
             Bound::Excluded(inner.as_ref())
         }
         Bound::Unbounded => Bound::Unbounded,
@@ -37,25 +37,22 @@ where
     let inner;
     let right = match right {
         Bound::Included(right) => {
-            inner = BoundCodec::bytes_encode(right).ok_or(heed::Error::Encoding)?;
+            inner = BoundCodec::bytes_encode(right).map_err(heed::Error::Encoding)?;
             Bound::Included(inner.as_ref())
         }
         Bound::Excluded(right) => {
-            inner = BoundCodec::bytes_encode(right).ok_or(heed::Error::Encoding)?;
+            inner = BoundCodec::bytes_encode(right).map_err(heed::Error::Encoding)?;
             Bound::Excluded(inner.as_ref())
         }
         Bound::Unbounded => Bound::Unbounded,
     };
-    let db = db.remap_key_type::<FacetGroupKeyCodec<ByteSliceRefCodec>>();
+    let db = db.remap_key_type::<FacetGroupKeyCodec<BytesRefCodec>>();
     let mut f = FacetRangeSearch { rtxn, db, field_id, left, right, docids };
     let highest_level = get_highest_level(rtxn, db, field_id)?;
 
-    if let Some(starting_left_bound) =
-        get_first_facet_value::<ByteSliceRefCodec>(rtxn, db, field_id)?
-    {
-        let rightmost_bound = Bound::Included(
-            get_last_facet_value::<ByteSliceRefCodec>(rtxn, db, field_id)?.unwrap(),
-        ); // will not fail because get_first_facet_value succeeded
+    if let Some(starting_left_bound) = get_first_facet_value::<BytesRefCodec>(rtxn, db, field_id)? {
+        let rightmost_bound =
+            Bound::Included(get_last_facet_value::<BytesRefCodec>(rtxn, db, field_id)?.unwrap()); // will not fail because get_first_facet_value succeeded
         let group_size = usize::MAX;
         f.run(highest_level, starting_left_bound, rightmost_bound, group_size)?;
         Ok(())
@@ -67,7 +64,7 @@ where
 /// Fetch the document ids that have a facet with a value between the two given bounds
 struct FacetRangeSearch<'t, 'b, 'bitmap> {
     rtxn: &'t heed::RoTxn<'t>,
-    db: heed::Database<FacetGroupKeyCodec<ByteSliceRefCodec>, FacetGroupValueCodec>,
+    db: heed::Database<FacetGroupKeyCodec<BytesRefCodec>, FacetGroupValueCodec>,
     field_id: u16,
     left: Bound<&'b [u8]>,
     right: Bound<&'b [u8]>,

milli/src/search/facet/facet_sort_ascending.rs

@@ -5,7 +5,7 @@ use super::{get_first_facet_value, get_highest_level};
 use crate::heed_codec::facet::{
     FacetGroupKey, FacetGroupKeyCodec, FacetGroupValue, FacetGroupValueCodec,
 };
-use crate::heed_codec::ByteSliceRefCodec;
+use crate::heed_codec::BytesRefCodec;
 
 /// Return an iterator which iterates over the given candidate documents in
 /// ascending order of their facet value for the given field id.
@@ -31,12 +31,12 @@ use crate::heed_codec::ByteSliceRefCodec;
 /// Note that once a document id is returned by the iterator, it is never returned again.
 pub fn ascending_facet_sort<'t>(
     rtxn: &'t heed::RoTxn<'t>,
-    db: heed::Database<FacetGroupKeyCodec<ByteSliceRefCodec>, FacetGroupValueCodec>,
+    db: heed::Database<FacetGroupKeyCodec<BytesRefCodec>, FacetGroupValueCodec>,
     field_id: u16,
     candidates: RoaringBitmap,
 ) -> Result<impl Iterator<Item = Result<(RoaringBitmap, &'t [u8])>> + 't> {
     let highest_level = get_highest_level(rtxn, db, field_id)?;
-    if let Some(first_bound) = get_first_facet_value::<ByteSliceRefCodec>(rtxn, db, field_id)? {
+    if let Some(first_bound) = get_first_facet_value::<BytesRefCodec>(rtxn, db, field_id)? {
         let first_key = FacetGroupKey { field_id, level: highest_level, left_bound: first_bound };
         let iter = db.range(rtxn, &(first_key..)).unwrap().take(usize::MAX);
 
@@ -53,14 +53,12 @@ pub fn ascending_facet_sort<'t>(
 
 struct AscendingFacetSort<'t, 'e> {
     rtxn: &'t heed::RoTxn<'e>,
-    db: heed::Database<FacetGroupKeyCodec<ByteSliceRefCodec>, FacetGroupValueCodec>,
+    db: heed::Database<FacetGroupKeyCodec<BytesRefCodec>, FacetGroupValueCodec>,
     field_id: u16,
     #[allow(clippy::type_complexity)]
     stack: Vec<(
         RoaringBitmap,
-        std::iter::Take<
-            heed::RoRange<'t, FacetGroupKeyCodec<ByteSliceRefCodec>, FacetGroupValueCodec>,
-        >,
+        std::iter::Take<heed::RoRange<'t, FacetGroupKeyCodec<BytesRefCodec>, FacetGroupValueCodec>>,
     )>,
 }
 

milli/src/search/facet/facet_sort_descending.rs

@@ -7,21 +7,21 @@ use super::{get_first_facet_value, get_highest_level, get_last_facet_value};
 use crate::heed_codec::facet::{
     FacetGroupKey, FacetGroupKeyCodec, FacetGroupValue, FacetGroupValueCodec,
 };
-use crate::heed_codec::ByteSliceRefCodec;
+use crate::heed_codec::BytesRefCodec;
 
 /// See documentationg for [`ascending_facet_sort`](super::ascending_facet_sort).
 ///
 /// This function does the same thing, but in the opposite order.
 pub fn descending_facet_sort<'t>(
     rtxn: &'t heed::RoTxn<'t>,
-    db: heed::Database<FacetGroupKeyCodec<ByteSliceRefCodec>, FacetGroupValueCodec>,
+    db: heed::Database<FacetGroupKeyCodec<BytesRefCodec>, FacetGroupValueCodec>,
     field_id: u16,
     candidates: RoaringBitmap,
 ) -> Result<impl Iterator<Item = Result<(RoaringBitmap, &'t [u8])>> + 't> {
     let highest_level = get_highest_level(rtxn, db, field_id)?;
-    if let Some(first_bound) = get_first_facet_value::<ByteSliceRefCodec>(rtxn, db, field_id)? {
+    if let Some(first_bound) = get_first_facet_value::<BytesRefCodec>(rtxn, db, field_id)? {
         let first_key = FacetGroupKey { field_id, level: highest_level, left_bound: first_bound };
-        let last_bound = get_last_facet_value::<ByteSliceRefCodec>(rtxn, db, field_id)?.unwrap();
+        let last_bound = get_last_facet_value::<BytesRefCodec>(rtxn, db, field_id)?.unwrap();
         let last_key = FacetGroupKey { field_id, level: highest_level, left_bound: last_bound };
         let iter = db.rev_range(rtxn, &(first_key..=last_key))?.take(usize::MAX);
         Ok(itertools::Either::Left(DescendingFacetSort {
@@ -37,13 +37,13 @@ pub fn descending_facet_sort<'t>(
 
 struct DescendingFacetSort<'t> {
     rtxn: &'t heed::RoTxn<'t>,
-    db: heed::Database<FacetGroupKeyCodec<ByteSliceRefCodec>, FacetGroupValueCodec>,
+    db: heed::Database<FacetGroupKeyCodec<BytesRefCodec>, FacetGroupValueCodec>,
     field_id: u16,
     #[allow(clippy::type_complexity)]
     stack: Vec<(
         RoaringBitmap,
         std::iter::Take<
-            heed::RoRevRange<'t, FacetGroupKeyCodec<ByteSliceRefCodec>, FacetGroupValueCodec>,
+            heed::RoRevRange<'t, FacetGroupKeyCodec<BytesRefCodec>, FacetGroupValueCodec>,
         >,
         Bound<&'t [u8]>,
     )>,
@@ -100,7 +100,7 @@ impl<'t> Iterator for DescendingFacetSort<'t> {
                     *right_bound = Bound::Excluded(left_bound);
                     let iter = match self
                         .db
-                        .remap_key_type::<FacetGroupKeyCodec<ByteSliceRefCodec>>()
+                        .remap_key_type::<FacetGroupKeyCodec<BytesRefCodec>>()
                         .rev_range(self.rtxn, &(Bound::Included(starting_key_below), end_key_kelow))
                     {
                         Ok(iter) => iter,
@@ -123,7 +123,7 @@ mod tests {
     use roaring::RoaringBitmap;
 
     use crate::heed_codec::facet::FacetGroupKeyCodec;
-    use crate::heed_codec::ByteSliceRefCodec;
+    use crate::heed_codec::BytesRefCodec;
     use crate::milli_snap;
     use crate::search::facet::facet_sort_descending::descending_facet_sort;
     use crate::search::facet::tests::{
@@ -144,7 +144,7 @@ mod tests {
             let txn = index.env.read_txn().unwrap();
             let candidates = (200..=300).collect::<RoaringBitmap>();
             let mut results = String::new();
-            let db = index.content.remap_key_type::<FacetGroupKeyCodec<ByteSliceRefCodec>>();
+            let db = index.content.remap_key_type::<FacetGroupKeyCodec<BytesRefCodec>>();
             let iter = descending_facet_sort(&txn, db, 0, candidates).unwrap();
             for el in iter {
                 let (docids, _) = el.unwrap();
@@ -167,7 +167,7 @@ mod tests {
             let txn = index.env.read_txn().unwrap();
             let candidates = (200..=300).collect::<RoaringBitmap>();
             let mut results = String::new();
-            let db = index.content.remap_key_type::<FacetGroupKeyCodec<ByteSliceRefCodec>>();
+            let db = index.content.remap_key_type::<FacetGroupKeyCodec<BytesRefCodec>>();
             let iter = descending_facet_sort(&txn, db, 0, candidates.clone()).unwrap();
             for el in iter {
                 let (docids, _) = el.unwrap();

milli/src/search/facet/filter.rs

@@ -3,7 +3,7 @@ use std::fmt::{Debug, Display};
 use std::ops::Bound::{self, Excluded, Included};
 
 use either::Either;
-pub use filter_parser::{Condition, Error as FPError, FilterCondition, Span, Token};
+pub use filter_parser::{Condition, Error as FPError, FilterCondition, Token};
 use roaring::RoaringBitmap;
 use serde_json::Value;
 

milli/src/search/facet/mod.rs

@@ -1,13 +1,13 @@
 pub use facet_sort_ascending::ascending_facet_sort;
 pub use facet_sort_descending::descending_facet_sort;
-use heed::types::{ByteSlice, DecodeIgnore};
+use heed::types::{Bytes, DecodeIgnore};
 use heed::{BytesDecode, RoTxn};
 use roaring::RoaringBitmap;
 
 pub use self::facet_distribution::{FacetDistribution, OrderBy, DEFAULT_VALUES_PER_FACET};
 pub use self::filter::{BadGeoError, Filter};
 use crate::heed_codec::facet::{FacetGroupKeyCodec, FacetGroupValueCodec, OrderedF64Codec};
-use crate::heed_codec::ByteSliceRefCodec;
+use crate::heed_codec::BytesRefCodec;
 use crate::{Index, Result};
 mod facet_distribution;
 mod facet_distribution_iter;
@@ -22,8 +22,10 @@ fn facet_extreme_value<'t>(
     let extreme_value =
         if let Some(extreme_value) = extreme_it.next() { extreme_value } else { return Ok(None) };
     let (_, extreme_value) = extreme_value?;
-
-    Ok(OrderedF64Codec::bytes_decode(extreme_value))
+    OrderedF64Codec::bytes_decode(extreme_value)
+        .map(Some)
+        .map_err(heed::Error::Decoding)
+        .map_err(Into::into)
 }
 
 pub fn facet_min_value<'t>(
@@ -32,7 +34,7 @@ pub fn facet_min_value<'t>(
     field_id: u16,
     candidates: RoaringBitmap,
 ) -> Result<Option<f64>> {
-    let db = index.facet_id_f64_docids.remap_key_type::<FacetGroupKeyCodec<ByteSliceRefCodec>>();
+    let db = index.facet_id_f64_docids.remap_key_type::<FacetGroupKeyCodec<BytesRefCodec>>();
     let it = ascending_facet_sort(rtxn, db, field_id, candidates)?;
     facet_extreme_value(it)
 }
@@ -43,7 +45,7 @@ pub fn facet_max_value<'t>(
     field_id: u16,
     candidates: RoaringBitmap,
 ) -> Result<Option<f64>> {
-    let db = index.facet_id_f64_docids.remap_key_type::<FacetGroupKeyCodec<ByteSliceRefCodec>>();
+    let db = index.facet_id_f64_docids.remap_key_type::<FacetGroupKeyCodec<BytesRefCodec>>();
     let it = descending_facet_sort(rtxn, db, field_id, candidates)?;
     facet_extreme_value(it)
 }
@@ -51,7 +53,7 @@ pub fn facet_max_value<'t>(
 /// Get the first facet value in the facet database
 pub(crate) fn get_first_facet_value<'t, BoundCodec>(
     txn: &'t RoTxn,
-    db: heed::Database<FacetGroupKeyCodec<ByteSliceRefCodec>, FacetGroupValueCodec>,
+    db: heed::Database<FacetGroupKeyCodec<BytesRefCodec>, FacetGroupValueCodec>,
     field_id: u16,
 ) -> heed::Result<Option<BoundCodec::DItem>>
 where
@@ -60,13 +62,12 @@ where
     let mut level0prefix = vec![];
     level0prefix.extend_from_slice(&field_id.to_be_bytes());
     level0prefix.push(0);
-    let mut level0_iter_forward = db
-        .as_polymorph()
-        .prefix_iter::<_, ByteSlice, DecodeIgnore>(txn, level0prefix.as_slice())?;
+    let mut level0_iter_forward =
+        db.remap_types::<Bytes, DecodeIgnore>().prefix_iter(txn, level0prefix.as_slice())?;
     if let Some(first) = level0_iter_forward.next() {
         let (first_key, _) = first?;
         let first_key = FacetGroupKeyCodec::<BoundCodec>::bytes_decode(first_key)
-            .ok_or(heed::Error::Encoding)?;
+            .map_err(heed::Error::Decoding)?;
         Ok(Some(first_key.left_bound))
     } else {
         Ok(None)
@@ -76,7 +77,7 @@ where
 /// Get the last facet value in the facet database
 pub(crate) fn get_last_facet_value<'t, BoundCodec>(
     txn: &'t RoTxn,
-    db: heed::Database<FacetGroupKeyCodec<ByteSliceRefCodec>, FacetGroupValueCodec>,
+    db: heed::Database<FacetGroupKeyCodec<BytesRefCodec>, FacetGroupValueCodec>,
     field_id: u16,
 ) -> heed::Result<Option<BoundCodec::DItem>>
 where
@@ -85,13 +86,12 @@ where
     let mut level0prefix = vec![];
     level0prefix.extend_from_slice(&field_id.to_be_bytes());
     level0prefix.push(0);
-    let mut level0_iter_backward = db
-        .as_polymorph()
-        .rev_prefix_iter::<_, ByteSlice, DecodeIgnore>(txn, level0prefix.as_slice())?;
+    let mut level0_iter_backward =
+        db.remap_types::<Bytes, DecodeIgnore>().rev_prefix_iter(txn, level0prefix.as_slice())?;
     if let Some(last) = level0_iter_backward.next() {
         let (last_key, _) = last?;
         let last_key = FacetGroupKeyCodec::<BoundCodec>::bytes_decode(last_key)
-            .ok_or(heed::Error::Encoding)?;
+            .map_err(heed::Error::Decoding)?;
         Ok(Some(last_key.left_bound))
     } else {
         Ok(None)
@@ -101,17 +101,17 @@ where
 /// Get the height of the highest level in the facet database
 pub(crate) fn get_highest_level<'t>(
     txn: &'t RoTxn<'t>,
-    db: heed::Database<FacetGroupKeyCodec<ByteSliceRefCodec>, FacetGroupValueCodec>,
+    db: heed::Database<FacetGroupKeyCodec<BytesRefCodec>, FacetGroupValueCodec>,
     field_id: u16,
 ) -> heed::Result<u8> {
     let field_id_prefix = &field_id.to_be_bytes();
     Ok(db
-        .as_polymorph()
-        .rev_prefix_iter::<_, ByteSlice, DecodeIgnore>(txn, field_id_prefix)?
+        .remap_types::<Bytes, DecodeIgnore>()
+        .rev_prefix_iter(txn, field_id_prefix)?
         .next()
         .map(|el| {
             let (key, _) = el.unwrap();
-            let key = FacetGroupKeyCodec::<ByteSliceRefCodec>::bytes_decode(key).unwrap();
+            let key = FacetGroupKeyCodec::<BytesRefCodec>::bytes_decode(key).unwrap();
             key.level
         })
         .unwrap_or(0))

milli/src/search/mod.rs

@@ -11,14 +11,13 @@ use once_cell::sync::Lazy;
 use roaring::bitmap::RoaringBitmap;
 
 pub use self::facet::{FacetDistribution, Filter, OrderBy, DEFAULT_VALUES_PER_FACET};
-pub use self::new::matches::{FormatOptions, MatchBounds, Matcher, MatcherBuilder, MatchingWords};
+pub use self::new::matches::{FormatOptions, MatchBounds, MatcherBuilder, MatchingWords};
 use self::new::PartialSearchResult;
 use crate::error::UserError;
 use crate::heed_codec::facet::{FacetGroupKey, FacetGroupValue};
 use crate::score_details::{ScoreDetails, ScoringStrategy};
 use crate::{
-    execute_search, AscDesc, DefaultSearchLogger, DocumentId, FieldId, Index, Result,
-    SearchContext, BEU16,
+    execute_search, AscDesc, DefaultSearchLogger, DocumentId, FieldId, Index, Result, SearchContext,
 };
 
 // Building these factories is not free.
@@ -299,7 +298,7 @@ impl<'a> SearchForFacetValues<'a> {
             None => return Ok(Vec::new()),
         };
 
-        let fst = match self.search_query.index.facet_id_string_fst.get(rtxn, &BEU16::new(fid))? {
+        let fst = match self.search_query.index.facet_id_string_fst.get(rtxn, &fid)? {
             Some(fst) => fst,
             None => return Ok(vec![]),
         };

milli/src/search/new/bucket_sort.rs

@@ -46,9 +46,8 @@ pub fn bucket_sort<'ctx, Q: RankingRuleQueryTrait>(
         if let Some(distinct_fid) = distinct_fid {
             let mut excluded = RoaringBitmap::new();
             let mut results = vec![];
-            let mut skip = 0;
             for docid in universe.iter() {
-                if results.len() >= length {
+                if results.len() >= from + length {
                     break;
                 }
                 if excluded.contains(docid) {
@@ -56,16 +55,19 @@ pub fn bucket_sort<'ctx, Q: RankingRuleQueryTrait>(
                 }
 
                 distinct_single_docid(ctx.index, ctx.txn, distinct_fid, docid, &mut excluded)?;
-                skip += 1;
-                if skip <= from {
-                    continue;
-                }
-
                 results.push(docid);
             }
 
             let mut all_candidates = universe - excluded;
             all_candidates.extend(results.iter().copied());
+            // drain the results of the skipped elements
+            // this **must** be done **after** writing the entire results in `all_candidates` to ensure
+            // e.g. estimatedTotalHits is correct.
+            if results.len() >= from {
+                results.drain(..from);
+            } else {
+                results.clear();
+            }
 
             return Ok(BucketSortOutput {
                 scores: vec![Default::default(); results.len()],

milli/src/search/new/db_cache.rs

@@ -3,15 +3,18 @@ use std::collections::hash_map::Entry;
 use std::hash::Hash;
 
 use fxhash::FxHashMap;
-use heed::types::ByteSlice;
+use heed::types::Bytes;
 use heed::{BytesEncode, Database, RoTxn};
 use roaring::RoaringBitmap;
 
 use super::interner::Interned;
 use super::Word;
 use crate::heed_codec::{BytesDecodeOwned, StrBEU16Codec};
+use crate::proximity::ProximityPrecision;
 use crate::update::{merge_cbo_roaring_bitmaps, MergeFn};
-use crate::{CboRoaringBitmapCodec, CboRoaringBitmapLenCodec, Result, SearchContext};
+use crate::{
+    CboRoaringBitmapCodec, CboRoaringBitmapLenCodec, Result, SearchContext, U8StrStrCodec,
+};
 
 /// A cache storing pointers to values in the LMDB databases.
 ///
@@ -23,7 +26,7 @@ pub struct DatabaseCache<'ctx> {
     pub word_pair_proximity_docids:
         FxHashMap<(u8, Interned<String>, Interned<String>), Option<Cow<'ctx, [u8]>>>,
     pub word_prefix_pair_proximity_docids:
-        FxHashMap<(u8, Interned<String>, Interned<String>), Option<Cow<'ctx, [u8]>>>,
+        FxHashMap<(u8, Interned<String>, Interned<String>), Option<RoaringBitmap>>,
     pub prefix_word_pair_proximity_docids:
         FxHashMap<(u8, Interned<String>, Interned<String>), Option<Cow<'ctx, [u8]>>>,
     pub word_docids: FxHashMap<Interned<String>, Option<Cow<'ctx, [u8]>>>,
@@ -48,7 +51,7 @@ impl<'ctx> DatabaseCache<'ctx> {
         cache_key: K1,
         db_key: &'v KC::EItem,
         cache: &mut FxHashMap<K1, Option<Cow<'ctx, [u8]>>>,
-        db: Database<KC, ByteSlice>,
+        db: Database<KC, Bytes>,
     ) -> Result<Option<DC::DItem>>
     where
         K1: Copy + Eq + Hash,
@@ -61,12 +64,14 @@ impl<'ctx> DatabaseCache<'ctx> {
         }
 
         match cache.get(&cache_key).unwrap() {
-            Some(Cow::Borrowed(bytes)) => {
-                DC::bytes_decode_owned(bytes).ok_or(heed::Error::Decoding.into()).map(Some)
-            }
-            Some(Cow::Owned(bytes)) => {
-                DC::bytes_decode_owned(bytes).ok_or(heed::Error::Decoding.into()).map(Some)
-            }
+            Some(Cow::Borrowed(bytes)) => DC::bytes_decode_owned(bytes)
+                .map(Some)
+                .map_err(heed::Error::Decoding)
+                .map_err(Into::into),
+            Some(Cow::Owned(bytes)) => DC::bytes_decode_owned(bytes)
+                .map(Some)
+                .map_err(heed::Error::Decoding)
+                .map_err(Into::into),
             None => Ok(None),
         }
     }
@@ -76,7 +81,7 @@ impl<'ctx> DatabaseCache<'ctx> {
         cache_key: K1,
         db_keys: &'v [KC::EItem],
         cache: &mut FxHashMap<K1, Option<Cow<'ctx, [u8]>>>,
-        db: Database<KC, ByteSlice>,
+        db: Database<KC, Bytes>,
         merger: MergeFn,
     ) -> Result<Option<DC::DItem>>
     where
@@ -108,12 +113,14 @@ impl<'ctx> DatabaseCache<'ctx> {
         }
 
         match cache.get(&cache_key).unwrap() {
-            Some(Cow::Borrowed(bytes)) => {
-                DC::bytes_decode_owned(bytes).ok_or(heed::Error::Decoding.into()).map(Some)
-            }
-            Some(Cow::Owned(bytes)) => {
-                DC::bytes_decode_owned(bytes).ok_or(heed::Error::Decoding.into()).map(Some)
-            }
+            Some(Cow::Borrowed(bytes)) => DC::bytes_decode_owned(bytes)
+                .map(Some)
+                .map_err(heed::Error::Decoding)
+                .map_err(Into::into),
+            Some(Cow::Owned(bytes)) => DC::bytes_decode_owned(bytes)
+                .map(Some)
+                .map_err(heed::Error::Decoding)
+                .map_err(Into::into),
             None => Ok(None),
         }
     }
@@ -162,7 +169,7 @@ impl<'ctx> SearchContext<'ctx> {
                     word,
                     &keys[..],
                     &mut self.db_cache.word_docids,
-                    self.index.word_fid_docids.remap_data_type::<ByteSlice>(),
+                    self.index.word_fid_docids.remap_data_type::<Bytes>(),
                     merge_cbo_roaring_bitmaps,
                 )
             }
@@ -171,7 +178,7 @@ impl<'ctx> SearchContext<'ctx> {
                 word,
                 self.word_interner.get(word).as_str(),
                 &mut self.db_cache.word_docids,
-                self.index.word_docids.remap_data_type::<ByteSlice>(),
+                self.index.word_docids.remap_data_type::<Bytes>(),
             ),
         }
     }
@@ -185,7 +192,7 @@ impl<'ctx> SearchContext<'ctx> {
             word,
             self.word_interner.get(word).as_str(),
             &mut self.db_cache.exact_word_docids,
-            self.index.exact_word_docids.remap_data_type::<ByteSlice>(),
+            self.index.exact_word_docids.remap_data_type::<Bytes>(),
         )
     }
 
@@ -224,7 +231,7 @@ impl<'ctx> SearchContext<'ctx> {
                     prefix,
                     &keys[..],
                     &mut self.db_cache.word_prefix_docids,
-                    self.index.word_prefix_fid_docids.remap_data_type::<ByteSlice>(),
+                    self.index.word_prefix_fid_docids.remap_data_type::<Bytes>(),
                     merge_cbo_roaring_bitmaps,
                 )
             }
@@ -233,7 +240,7 @@ impl<'ctx> SearchContext<'ctx> {
                 prefix,
                 self.word_interner.get(prefix).as_str(),
                 &mut self.db_cache.word_prefix_docids,
-                self.index.word_prefix_docids.remap_data_type::<ByteSlice>(),
+                self.index.word_prefix_docids.remap_data_type::<Bytes>(),
             ),
         }
     }
@@ -247,7 +254,7 @@ impl<'ctx> SearchContext<'ctx> {
             prefix,
             self.word_interner.get(prefix).as_str(),
             &mut self.db_cache.exact_word_prefix_docids,
-            self.index.exact_word_prefix_docids.remap_data_type::<ByteSlice>(),
+            self.index.exact_word_prefix_docids.remap_data_type::<Bytes>(),
         )
     }
 
@@ -257,17 +264,67 @@ impl<'ctx> SearchContext<'ctx> {
         word2: Interned<String>,
         proximity: u8,
     ) -> Result<Option<RoaringBitmap>> {
-        DatabaseCache::get_value::<_, _, CboRoaringBitmapCodec>(
-            self.txn,
-            (proximity, word1, word2),
-            &(
-                proximity,
-                self.word_interner.get(word1).as_str(),
-                self.word_interner.get(word2).as_str(),
-            ),
-            &mut self.db_cache.word_pair_proximity_docids,
-            self.index.word_pair_proximity_docids.remap_data_type::<ByteSlice>(),
-        )
+        match self.index.proximity_precision(self.txn)?.unwrap_or_default() {
+            ProximityPrecision::AttributeScale => {
+                // Force proximity to 0 because:
+                // in AttributeScale, there are only 2 possible distances:
+                // 1. words in same attribute: in that the DB contains (0, word1, word2)
+                // 2. words in different attributes: no DB entry for these two words.
+                let proximity = 0;
+                let docids = if let Some(docids) =
+                    self.db_cache.word_pair_proximity_docids.get(&(proximity, word1, word2))
+                {
+                    docids
+                        .as_ref()
+                        .map(|d| CboRoaringBitmapCodec::bytes_decode_owned(d))
+                        .transpose()
+                        .map_err(heed::Error::Decoding)?
+                } else {
+                    // Compute the distance at the attribute level and store it in the cache.
+                    let fids = if let Some(fids) = self.index.searchable_fields_ids(self.txn)? {
+                        fids
+                    } else {
+                        self.index.fields_ids_map(self.txn)?.ids().collect()
+                    };
+                    let mut docids = RoaringBitmap::new();
+                    for fid in fids {
+                        // for each field, intersect left word bitmap and right word bitmap,
+                        // then merge the result in a global bitmap before storing it in the cache.
+                        let word1_docids = self.get_db_word_fid_docids(word1, fid)?;
+                        let word2_docids = self.get_db_word_fid_docids(word2, fid)?;
+                        if let (Some(word1_docids), Some(word2_docids)) =
+                            (word1_docids, word2_docids)
+                        {
+                            docids |= word1_docids & word2_docids;
+                        }
+                    }
+                    let encoded = CboRoaringBitmapCodec::bytes_encode(&docids)
+                        .map(Cow::into_owned)
+                        .map(Cow::Owned)
+                        .map(Some)
+                        .map_err(heed::Error::Decoding)?;
+                    self.db_cache
+                        .word_pair_proximity_docids
+                        .insert((proximity, word1, word2), encoded);
+                    Some(docids)
+                };
+
+                Ok(docids)
+            }
+            ProximityPrecision::WordScale => {
+                DatabaseCache::get_value::<_, _, CboRoaringBitmapCodec>(
+                    self.txn,
+                    (proximity, word1, word2),
+                    &(
+                        proximity,
+                        self.word_interner.get(word1).as_str(),
+                        self.word_interner.get(word2).as_str(),
+                    ),
+                    &mut self.db_cache.word_pair_proximity_docids,
+                    self.index.word_pair_proximity_docids.remap_data_type::<Bytes>(),
+                )
+            }
+        }
     }
 
     pub fn get_db_word_pair_proximity_docids_len(
@@ -276,54 +333,107 @@ impl<'ctx> SearchContext<'ctx> {
         word2: Interned<String>,
         proximity: u8,
     ) -> Result<Option<u64>> {
-        DatabaseCache::get_value::<_, _, CboRoaringBitmapLenCodec>(
-            self.txn,
-            (proximity, word1, word2),
-            &(
-                proximity,
-                self.word_interner.get(word1).as_str(),
-                self.word_interner.get(word2).as_str(),
-            ),
-            &mut self.db_cache.word_pair_proximity_docids,
-            self.index.word_pair_proximity_docids.remap_data_type::<ByteSlice>(),
-        )
+        match self.index.proximity_precision(self.txn)?.unwrap_or_default() {
+            ProximityPrecision::AttributeScale => Ok(self
+                .get_db_word_pair_proximity_docids(word1, word2, proximity)?
+                .map(|d| d.len())),
+            ProximityPrecision::WordScale => {
+                DatabaseCache::get_value::<_, _, CboRoaringBitmapLenCodec>(
+                    self.txn,
+                    (proximity, word1, word2),
+                    &(
+                        proximity,
+                        self.word_interner.get(word1).as_str(),
+                        self.word_interner.get(word2).as_str(),
+                    ),
+                    &mut self.db_cache.word_pair_proximity_docids,
+                    self.index.word_pair_proximity_docids.remap_data_type::<Bytes>(),
+                )
+            }
+        }
     }
 
     pub fn get_db_word_prefix_pair_proximity_docids(
         &mut self,
         word1: Interned<String>,
         prefix2: Interned<String>,
-        proximity: u8,
+        mut proximity: u8,
     ) -> Result<Option<RoaringBitmap>> {
-        DatabaseCache::get_value::<_, _, CboRoaringBitmapCodec>(
-            self.txn,
-            (proximity, word1, prefix2),
-            &(
-                proximity,
-                self.word_interner.get(word1).as_str(),
-                self.word_interner.get(prefix2).as_str(),
-            ),
-            &mut self.db_cache.word_prefix_pair_proximity_docids,
-            self.index.word_prefix_pair_proximity_docids.remap_data_type::<ByteSlice>(),
-        )
+        let proximity_precision = self.index.proximity_precision(self.txn)?.unwrap_or_default();
+        if proximity_precision == ProximityPrecision::AttributeScale {
+            // Force proximity to 0 because:
+            // in AttributeScale, there are only 2 possible distances:
+            // 1. words in same attribute: in that the DB contains (0, word1, word2)
+            // 2. words in different attributes: no DB entry for these two words.
+            proximity = 0;
+        }
+
+        let docids = if let Some(docids) =
+            self.db_cache.word_prefix_pair_proximity_docids.get(&(proximity, word1, prefix2))
+        {
+            docids.clone()
+        } else {
+            let prefix_docids = match proximity_precision {
+                ProximityPrecision::AttributeScale => {
+                    // Compute the distance at the attribute level and store it in the cache.
+                    let fids = if let Some(fids) = self.index.searchable_fields_ids(self.txn)? {
+                        fids
+                    } else {
+                        self.index.fields_ids_map(self.txn)?.ids().collect()
+                    };
+                    let mut prefix_docids = RoaringBitmap::new();
+                    // for each field, intersect left word bitmap and right word bitmap,
+                    // then merge the result in a global bitmap before storing it in the cache.
+                    for fid in fids {
+                        let word1_docids = self.get_db_word_fid_docids(word1, fid)?;
+                        let prefix2_docids = self.get_db_word_prefix_fid_docids(prefix2, fid)?;
+                        if let (Some(word1_docids), Some(prefix2_docids)) =
+                            (word1_docids, prefix2_docids)
+                        {
+                            prefix_docids |= word1_docids & prefix2_docids;
+                        }
+                    }
+                    prefix_docids
+                }
+                ProximityPrecision::WordScale => {
+                    // compute docids using prefix iter and store the result in the cache.
+                    let key = U8StrStrCodec::bytes_encode(&(
+                        proximity,
+                        self.word_interner.get(word1).as_str(),
+                        self.word_interner.get(prefix2).as_str(),
+                    ))
+                    .unwrap()
+                    .into_owned();
+                    let mut prefix_docids = RoaringBitmap::new();
+                    let remap_key_type = self
+                        .index
+                        .word_pair_proximity_docids
+                        .remap_key_type::<Bytes>()
+                        .prefix_iter(self.txn, &key)?;
+                    for result in remap_key_type {
+                        let (_, docids) = result?;
+
+                        prefix_docids |= docids;
+                    }
+                    prefix_docids
+                }
+            };
+            self.db_cache
+                .word_prefix_pair_proximity_docids
+                .insert((proximity, word1, prefix2), Some(prefix_docids.clone()));
+            Some(prefix_docids)
+        };
+        Ok(docids)
     }
+
     pub fn get_db_prefix_word_pair_proximity_docids(
         &mut self,
         left_prefix: Interned<String>,
         right: Interned<String>,
         proximity: u8,
     ) -> Result<Option<RoaringBitmap>> {
-        DatabaseCache::get_value::<_, _, CboRoaringBitmapCodec>(
-            self.txn,
-            (proximity, left_prefix, right),
-            &(
-                proximity,
-                self.word_interner.get(left_prefix).as_str(),
-                self.word_interner.get(right).as_str(),
-            ),
-            &mut self.db_cache.prefix_word_pair_proximity_docids,
-            self.index.prefix_word_pair_proximity_docids.remap_data_type::<ByteSlice>(),
-        )
+        // only accept exact matches on reverted positions
+        self.get_db_word_pair_proximity_docids(left_prefix, right, proximity)
     }
 
     pub fn get_db_word_fid_docids(
@@ -341,7 +451,7 @@ impl<'ctx> SearchContext<'ctx> {
             (word, fid),
             &(self.word_interner.get(word).as_str(), fid),
             &mut self.db_cache.word_fid_docids,
-            self.index.word_fid_docids.remap_data_type::<ByteSlice>(),
+            self.index.word_fid_docids.remap_data_type::<Bytes>(),
         )
     }
 
@@ -360,7 +470,7 @@ impl<'ctx> SearchContext<'ctx> {
             (word_prefix, fid),
             &(self.word_interner.get(word_prefix).as_str(), fid),
             &mut self.db_cache.word_prefix_fid_docids,
-            self.index.word_prefix_fid_docids.remap_data_type::<ByteSlice>(),
+            self.index.word_prefix_fid_docids.remap_data_type::<Bytes>(),
         )
     }
 
@@ -374,7 +484,7 @@ impl<'ctx> SearchContext<'ctx> {
                 let remap_key_type = self
                     .index
                     .word_fid_docids
-                    .remap_types::<ByteSlice, ByteSlice>()
+                    .remap_types::<Bytes, Bytes>()
                     .prefix_iter(self.txn, &key)?
                     .remap_key_type::<StrBEU16Codec>();
                 for result in remap_key_type {
@@ -400,7 +510,7 @@ impl<'ctx> SearchContext<'ctx> {
                 let remap_key_type = self
                     .index
                     .word_prefix_fid_docids
-                    .remap_types::<ByteSlice, ByteSlice>()
+                    .remap_types::<Bytes, Bytes>()
                     .prefix_iter(self.txn, &key)?
                     .remap_key_type::<StrBEU16Codec>();
                 for result in remap_key_type {
@@ -428,7 +538,7 @@ impl<'ctx> SearchContext<'ctx> {
             (word, position),
             &(self.word_interner.get(word).as_str(), position),
             &mut self.db_cache.word_position_docids,
-            self.index.word_position_docids.remap_data_type::<ByteSlice>(),
+            self.index.word_position_docids.remap_data_type::<Bytes>(),
         )
     }
 
@@ -442,7 +552,7 @@ impl<'ctx> SearchContext<'ctx> {
             (word_prefix, position),
             &(self.word_interner.get(word_prefix).as_str(), position),
             &mut self.db_cache.word_prefix_position_docids,
-            self.index.word_prefix_position_docids.remap_data_type::<ByteSlice>(),
+            self.index.word_prefix_position_docids.remap_data_type::<Bytes>(),
         )
     }
 
@@ -456,7 +566,7 @@ impl<'ctx> SearchContext<'ctx> {
                 let remap_key_type = self
                     .index
                     .word_position_docids
-                    .remap_types::<ByteSlice, ByteSlice>()
+                    .remap_types::<Bytes, Bytes>()
                     .prefix_iter(self.txn, &key)?
                     .remap_key_type::<StrBEU16Codec>();
                 for result in remap_key_type {
@@ -487,7 +597,7 @@ impl<'ctx> SearchContext<'ctx> {
                 let remap_key_type = self
                     .index
                     .word_prefix_position_docids
-                    .remap_types::<ByteSlice, ByteSlice>()
+                    .remap_types::<Bytes, Bytes>()
                     .prefix_iter(self.txn, &key)?
                     .remap_key_type::<StrBEU16Codec>();
                 for result in remap_key_type {

milli/src/search/new/distinct.rs

@@ -1,4 +1,4 @@
-use heed::types::{ByteSlice, Str, Unit};
+use heed::types::{Bytes, Str, Unit};
 use heed::{Database, RoPrefix, RoTxn};
 use roaring::RoaringBitmap;
 
@@ -8,7 +8,7 @@ const DOCID_SIZE: usize = 4;
 use crate::heed_codec::facet::{
     FacetGroupKey, FacetGroupKeyCodec, FacetGroupValueCodec, FieldDocIdFacetCodec,
 };
-use crate::heed_codec::ByteSliceRefCodec;
+use crate::heed_codec::BytesRefCodec;
 use crate::{Index, Result, SearchContext};
 
 pub struct DistinctOutput {
@@ -71,7 +71,7 @@ pub fn distinct_single_docid(
 
 /// Return all the docids containing the given value in the given field
 fn facet_value_docids(
-    database: Database<FacetGroupKeyCodec<ByteSliceRefCodec>, FacetGroupValueCodec>,
+    database: Database<FacetGroupKeyCodec<BytesRefCodec>, FacetGroupValueCodec>,
     txn: &RoTxn,
     field_id: u16,
     facet_value: &[u8],
@@ -87,12 +87,12 @@ fn facet_number_values<'a>(
     field_id: u16,
     index: &Index,
     txn: &'a RoTxn,
-) -> Result<RoPrefix<'a, FieldDocIdFacetCodec<ByteSliceRefCodec>, Unit>> {
+) -> Result<RoPrefix<'a, FieldDocIdFacetCodec<BytesRefCodec>, Unit>> {
     let key = facet_values_prefix_key(field_id, docid);
 
     let iter = index
         .field_id_docid_facet_f64s
-        .remap_key_type::<ByteSlice>()
+        .remap_key_type::<Bytes>()
         .prefix_iter(txn, &key)?
         .remap_key_type();
 
@@ -105,12 +105,12 @@ pub fn facet_string_values<'a>(
     field_id: u16,
     index: &Index,
     txn: &'a RoTxn,
-) -> Result<RoPrefix<'a, FieldDocIdFacetCodec<ByteSliceRefCodec>, Str>> {
+) -> Result<RoPrefix<'a, FieldDocIdFacetCodec<BytesRefCodec>, Str>> {
     let key = facet_values_prefix_key(field_id, docid);
 
     let iter = index
         .field_id_docid_facet_strings
-        .remap_key_type::<ByteSlice>()
+        .remap_key_type::<Bytes>()
         .prefix_iter(txn, &key)?
         .remap_types();
 

milli/src/search/new/geo_sort.rs

@@ -1,7 +1,7 @@
 use std::collections::VecDeque;
 use std::iter::FromIterator;
 
-use heed::types::{ByteSlice, Unit};
+use heed::types::{Bytes, Unit};
 use heed::{RoPrefix, RoTxn};
 use roaring::RoaringBitmap;
 use rstar::RTree;
@@ -34,7 +34,7 @@ fn facet_number_values<'a>(
 
     let iter = index
         .field_id_docid_facet_f64s
-        .remap_key_type::<ByteSlice>()
+        .remap_key_type::<Bytes>()
         .prefix_iter(txn, &key)?
         .remap_key_type();
 
@@ -163,7 +163,7 @@ impl<Q: RankingRuleQueryTrait> GeoSort<Q> {
             // computing the distance between two points is expensive thus we cache the result
             documents
                 .sort_by_cached_key(|(_, p)| distance_between_two_points(&self.point, p) as usize);
-            self.cached_sorted_docids.extend(documents.into_iter());
+            self.cached_sorted_docids.extend(documents);
         };
 
         Ok(())

milli/src/search/new/interner.rs

@@ -228,7 +228,7 @@ impl<T> Ord for Interned<T> {
 
 impl<T> PartialOrd for Interned<T> {
     fn partial_cmp(&self, other: &Self) -> Option<std::cmp::Ordering> {
-        self.idx.partial_cmp(&other.idx)
+        Some(self.cmp(other))
     }
 }
 
@@ -241,7 +241,7 @@ impl<T> PartialEq for Interned<T> {
 }
 impl<T> Clone for Interned<T> {
     fn clone(&self) -> Self {
-        Self { idx: self.idx, _phantom: PhantomData }
+        *self
     }
 }
 

milli/src/search/new/mod.rs

@@ -50,9 +50,7 @@ use crate::distance::NDotProductPoint;
 use crate::error::FieldIdMapMissingEntry;
 use crate::score_details::{ScoreDetails, ScoringStrategy};
 use crate::search::new::distinct::apply_distinct_rule;
-use crate::{
-    AscDesc, DocumentId, Filter, Index, Member, Result, TermsMatchingStrategy, UserError, BEU32,
-};
+use crate::{AscDesc, DocumentId, Filter, Index, Member, Result, TermsMatchingStrategy, UserError};
 
 /// A structure used throughout the execution of a search query.
 pub struct SearchContext<'ctx> {
@@ -434,14 +432,25 @@ pub fn execute_search(
         let mut search = Search::default();
         let docids = match ctx.index.vector_hnsw(ctx.txn)? {
             Some(hnsw) => {
+                if let Some(expected_size) = hnsw.iter().map(|(_, point)| point.len()).next() {
+                    if vector.len() != expected_size {
+                        return Err(UserError::InvalidVectorDimensions {
+                            expected: expected_size,
+                            found: vector.len(),
+                        }
+                        .into());
+                    }
+                }
+
                 let vector = NDotProductPoint::new(vector.clone());
+
                 let neighbors = hnsw.search(&vector, &mut search);
 
                 let mut docids = Vec::new();
                 let mut uniq_docids = RoaringBitmap::new();
                 for instant_distance::Item { distance: _, pid, point: _ } in neighbors {
-                    let index = BEU32::new(pid.into_inner());
-                    let docid = ctx.index.vector_id_docid.get(ctx.txn, &index)?.unwrap().get();
+                    let index = pid.into_inner();
+                    let docid = ctx.index.vector_id_docid.get(ctx.txn, &index)?.unwrap();
                     if universe.contains(docid) && uniq_docids.insert(docid) {
                         docids.push(docid);
                         if docids.len() == (from + length) {
@@ -598,7 +607,8 @@ fn check_sort_criteria(ctx: &SearchContext, sort_criteria: Option<&Vec<AscDesc>>
                     field: field.to_string(),
                     valid_fields,
                     hidden_fields,
-                })?;
+                }
+                .into());
             }
             Member::Geo(_) if !sortable_fields.contains("_geo") => {
                 let (valid_fields, hidden_fields) =
@@ -608,7 +618,8 @@ fn check_sort_criteria(ctx: &SearchContext, sort_criteria: Option<&Vec<AscDesc>>
                     field: "_geo".to_string(),
                     valid_fields,
                     hidden_fields,
-                })?;
+                }
+                .into());
             }
             _ => (),
         }

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

@@ -175,7 +175,7 @@ impl QueryTermSubset {
 
     pub fn use_prefix_db(&self, ctx: &SearchContext) -> Option<Word> {
         let original = ctx.term_interner.get(self.original);
-        let Some(use_prefix_db) = original.zero_typo.use_prefix_db else { return None };
+        let use_prefix_db = original.zero_typo.use_prefix_db?;
         let word = match &self.zero_typo_subset {
             NTypoTermSubset::All => Some(use_prefix_db),
             NTypoTermSubset::Subset { words, phrases: _ } => {