Enable swedish recomposition

4798: Update version for the next release (v1.8.4) in Cargo.toml r=dureuill a=meili-bot

⚠️ This PR is automatically generated. Check the new version is the expected one and Cargo.lock has been updated before merging.

Co-authored-by: dureuill <dureuill@users.noreply.github.com>

.github/workflows/flaky-tests.yml

@@ -1,4 +1,6 @@
 name: Look for flaky tests
+env:
+  ACTIONS_ALLOW_USE_UNSECURE_NODE_VERSION: true
 on:
   workflow_dispatch:
   schedule:

.github/workflows/fuzzer-indexing.yml

@@ -1,5 +1,6 @@
 name: Run the indexing fuzzer
-
+env:
+  ACTIONS_ALLOW_USE_UNSECURE_NODE_VERSION: true
 on:
   push:
     branches:

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

@@ -15,6 +15,8 @@ jobs:
 
   debian:
     name: Publish debian packagge
+    env:
+      ACTIONS_ALLOW_USE_UNSECURE_NODE_VERSION: true
     runs-on: ubuntu-latest
     needs: check-version
     container:

.github/workflows/publish-binaries.yml

@@ -35,6 +35,8 @@ jobs:
   publish-linux:
     name: Publish binary for Linux
     runs-on: ubuntu-latest
+    env:
+      ACTIONS_ALLOW_USE_UNSECURE_NODE_VERSION: true
     needs: check-version
     container:
       # Use ubuntu-18.04 to compile with glibc 2.27
@@ -132,6 +134,8 @@ jobs:
     name: Publish binary for aarch64
     runs-on: ubuntu-latest
     needs: check-version
+    env:
+      ACTIONS_ALLOW_USE_UNSECURE_NODE_VERSION: true
     container:
       # Use ubuntu-18.04 to compile with glibc 2.27
       image: ubuntu:18.04

.github/workflows/test-suite.yml

@@ -21,6 +21,8 @@ jobs:
   test-linux:
     name: Tests on ubuntu-18.04
     runs-on: ubuntu-latest
+    env:
+      ACTIONS_ALLOW_USE_UNSECURE_NODE_VERSION: true
     container:
       # Use ubuntu-18.04 to compile with glibc 2.27, which are the production expectations
       image: ubuntu:18.04
@@ -77,6 +79,8 @@ jobs:
   test-all-features:
     name: Tests almost all features
     runs-on: ubuntu-latest
+    env:
+      ACTIONS_ALLOW_USE_UNSECURE_NODE_VERSION: true
     container:
       # Use ubuntu-18.04 to compile with glibc 2.27, which are the production expectations
       image: ubuntu:18.04
@@ -100,6 +104,8 @@ jobs:
 
   test-disabled-tokenization:
     name: Test disabled tokenization
+    env:
+      ACTIONS_ALLOW_USE_UNSECURE_NODE_VERSION: true
     runs-on: ubuntu-latest
     container:
       image: ubuntu:18.04
@@ -127,6 +133,8 @@ jobs:
   # We run tests in debug also, to make sure that the debug_assertions are hit
   test-debug:
     name: Run tests in debug
+    env:
+      ACTIONS_ALLOW_USE_UNSECURE_NODE_VERSION: true
     runs-on: ubuntu-latest
     container:
       # Use ubuntu-18.04 to compile with glibc 2.27, which are the production expectations

Cargo.toml

@@ -22,7 +22,7 @@ members = [
 ]
 
 [workspace.package]
-version = "1.8.0"
+version = "1.8.4"
 authors = [
     "Quentin de Quelen <quentin@dequelen.me>",
     "Clément Renault <clement@meilisearch.com>",

Dockerfile

@@ -17,7 +17,7 @@ RUN     set -eux; \
         if [ "$apkArch" = "aarch64" ]; then \
             export JEMALLOC_SYS_WITH_LG_PAGE=16; \
         fi && \
-        cargo build --release -p meilisearch -p meilitool
+        cargo build --release -p meilisearch -p meilitool --features "swedish-recomposition"
 
 # Run
 FROM    alpine:3.16

dump/src/reader/v3/settings.rs

@@ -152,6 +152,7 @@ impl Settings<Unchecked> {
 }
 
 #[derive(Debug, Clone, Deserialize)]
+#[allow(dead_code)] // otherwise rustc complains that the fields go unused
 #[cfg_attr(test, derive(serde::Serialize))]
 #[serde(deny_unknown_fields)]
 #[serde(rename_all = "camelCase")]

dump/src/reader/v4/settings.rs

@@ -182,6 +182,7 @@ impl Settings<Unchecked> {
     }
 }
 
+#[allow(dead_code)] // otherwise rustc complains that the fields go unused
 #[derive(Debug, Clone, Deserialize)]
 #[cfg_attr(test, derive(serde::Serialize))]
 #[serde(deny_unknown_fields)]

dump/src/reader/v5/tasks.rs

@@ -200,6 +200,7 @@ impl std::ops::Deref for IndexUid {
     }
 }
 
+#[allow(dead_code)] // otherwise rustc complains that the fields go unused
 #[derive(Debug)]
 #[cfg_attr(test, derive(serde::Serialize))]
 #[cfg_attr(test, serde(rename_all = "camelCase"))]

filter-parser/src/lib.rs

@@ -568,7 +568,7 @@ pub mod tests {
         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 sequences
+        // but it also works with other sequencies
         insta::assert_display_snapshot!(p(r#"title = 'foo\x20\n\t\"\'"'"#), @"{title} = {foo \n\t\"\'\"}");
     }
 

index-scheduler/Cargo.toml

@@ -37,7 +37,7 @@ time = { version = "0.3.31", features = [
     "macros",
 ] }
 tracing = "0.1.40"
-ureq = "2.9.7"
+ureq = "2.9.1"
 uuid = { version = "1.6.1", features = ["serde", "v4"] }
 
 [dev-dependencies]

index-scheduler/src/batch.rs

@@ -13,7 +13,7 @@ We can combine the two tasks in a single batch:
 1. import documents X and Y
 
 Processing this batch is functionally equivalent to processing the two
-tasks individually, but should be much faster since we are only performing
+tasks individally, but should be much faster since we are only performing
 one indexing operation.
 */
 
@@ -914,8 +914,34 @@ impl IndexScheduler {
                         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)?;
+                        let (id, doc) = ret?;
+                        let mut document = milli::obkv_to_json(&all_fields, &fields_ids_map, doc)?;
+
+                        'inject_vectors: {
+                            let embeddings = index.embeddings(&rtxn, id)?;
+
+                            if embeddings.is_empty() {
+                                break 'inject_vectors;
+                            }
+
+                            let vectors = document
+                                .entry("_vectors".to_owned())
+                                .or_insert(serde_json::Value::Object(Default::default()));
+
+                            let serde_json::Value::Object(vectors) = vectors else {
+                                break 'inject_vectors;
+                            };
+
+                            for (embedder_name, embeddings) in embeddings {
+                                vectors.entry(embedder_name).or_insert_with(|| {
+                                    serde_json::json!({
+                                        "embeddings": embeddings,
+                                        "regenerate": true
+                                    })
+                                });
+                            }
+                        }
+
                         index_dumper.push_document(&document)?;
                     }
 

meilisearch-types/src/deserr/mod.rs

@@ -26,7 +26,7 @@ pub type DeserrQueryParamError<C = BadRequest> = DeserrError<DeserrQueryParam, C
 
 /// A request deserialization error.
 ///
-/// The first generic parameter is a marker type describing the format of the request: either json (e.g. [`DeserrJson`] or [`DeserrQueryParam`]).
+/// The first generic paramater is a marker type describing the format of the request: either json (e.g. [`DeserrJson`] or [`DeserrQueryParam`]).
 /// The second generic parameter is the default error code for the deserialization error, in case it is not given.
 pub struct DeserrError<Format, C: Default + ErrorCode> {
     pub msg: String,

meilisearch/Cargo.toml

@@ -71,13 +71,13 @@ puffin = { version = "0.16.0", features = ["serialization"] }
 rand = "0.8.5"
 rayon = "1.8.0"
 regex = "1.10.2"
-reqwest = { version = "0.12.4", features = [
+reqwest = { version = "0.11.23", features = [
     "rustls-tls",
     "json",
 ], default-features = false }
-rustls = "0.21.12"
+rustls = "0.21.6"
 rustls-pemfile = "1.0.2"
-segment = { version = "0.2.4", optional = true }
+segment = { version = "0.2.3", optional = true }
 serde = { version = "1.0.195", features = ["derive"] }
 serde_json = { version = "1.0.111", features = ["preserve_order"] }
 sha2 = "0.10.8"

meilisearch/src/lib.rs

@@ -419,7 +419,41 @@ fn import_dump(
         let file = tempfile::tempfile()?;
         let mut builder = DocumentsBatchBuilder::new(BufWriter::new(file));
         for document in index_reader.documents()? {
-            builder.append_json_object(&document?)?;
+            let mut document = document?;
+
+            'remove_injected_vectors: {
+                let Some(vectors) = document.get_mut("_vectors") else {
+                    break 'remove_injected_vectors;
+                };
+
+                let Some(vectors) = vectors.as_object_mut() else { break 'remove_injected_vectors };
+
+                vectors.retain(|_embedder, embedding_object| {
+                    // don't touch values that aren't objects
+                    let Some(embedding_object) = embedding_object.as_object() else {
+                        return true;
+                    };
+
+                    let mut has_regenerate_true = false;
+                    for (field, value) in embedding_object {
+                        match (field.as_str(), value) {
+                            // detected regenerate : true
+                            // if we don't have any superfluous field, we'll remove the entire entry
+                            ("regenerate", serde_json::Value::Bool(true)) => {
+                                has_regenerate_true = true;
+                            }
+                            // ignore embeddings
+                            ("embeddings", _) => continue,
+                            // any other field: immediately retain the entry
+                            _ => return true,
+                        }
+                    }
+                    // retain the entry unless it has regenerate: true
+                    !has_regenerate_true
+                })
+            }
+
+            builder.append_json_object(&document)?;
         }
 
         // This flush the content of the batch builder.

meilisearch/src/middleware.rs

@@ -59,12 +59,10 @@ where
             let request_path = req.path();
             let is_registered_resource = req.resource_map().has_resource(request_path);
             if is_registered_resource {
-                let request_pattern = req.match_pattern();
-                let metric_path = request_pattern.as_ref().map_or(request_path, String::as_str);
                 let request_method = req.method().to_string();
                 histogram_timer = Some(
                     crate::metrics::MEILISEARCH_HTTP_RESPONSE_TIME_SECONDS
-                        .with_label_values(&[&request_method, metric_path])
+                        .with_label_values(&[&request_method, request_path])
                         .start_timer(),
                 );
             }

meilisearch/src/search_queue.rs

@@ -40,8 +40,9 @@ pub struct Permit {
 
 impl Drop for Permit {
     fn drop(&mut self) {
+        let sender = self.sender.clone();
         // if the channel is closed then the whole instance is down
-        let _ = futures::executor::block_on(self.sender.send(()));
+        std::mem::drop(tokio::spawn(async move { sender.send(()).await }));
     }
 }
 

meilisearch/tests/search/geo.rs

@@ -117,3 +117,69 @@ async fn geo_bounding_box_with_string_and_number() {
         )
         .await;
 }
+
+#[actix_rt::test]
+async fn bug_4640() {
+    // https://github.com/meilisearch/meilisearch/issues/4640
+    let server = Server::new().await;
+    let index = server.index("test");
+
+    let documents = DOCUMENTS.clone();
+    index.add_documents(documents, None).await;
+    index.update_settings_filterable_attributes(json!(["_geo"])).await;
+    let (ret, _code) = index.update_settings_sortable_attributes(json!(["_geo"])).await;
+    index.wait_task(ret.uid()).await;
+
+    // Sort the document with the second one first
+    index
+        .search(
+            json!({
+                "sort": ["_geoPoint(45.4777599, 9.1967508):asc"],
+            }),
+            |response, code| {
+                assert_eq!(code, 200, "{}", response);
+                snapshot!(json_string!(response, { ".processingTimeMs" => "[time]" }), @r###"
+                {
+                  "hits": [
+                    {
+                      "id": 2,
+                      "name": "La Bella Italia",
+                      "address": "456 Elm Street, Townsville",
+                      "type": "Italian",
+                      "rating": 9,
+                      "_geo": {
+                        "lat": "45.4777599",
+                        "lng": "9.1967508"
+                      }
+                    },
+                    {
+                      "id": 1,
+                      "name": "Taco Truck",
+                      "address": "444 Salsa Street, Burritoville",
+                      "type": "Mexican",
+                      "rating": 9,
+                      "_geo": {
+                        "lat": 34.0522,
+                        "lng": -118.2437
+                      },
+                      "_geoDistance": 9714063
+                    },
+                    {
+                      "id": 3,
+                      "name": "Crêpe Truck",
+                      "address": "2 Billig Avenue, Rouenville",
+                      "type": "French",
+                      "rating": 10
+                    }
+                  ],
+                  "query": "",
+                  "processingTimeMs": "[time]",
+                  "limit": 20,
+                  "offset": 0,
+                  "estimatedTotalHits": 3
+                }
+                "###);
+            },
+        )
+        .await;
+}

meilitool/src/main.rs

@@ -129,7 +129,7 @@ fn clear_task_queue(db_path: PathBuf) -> anyhow::Result<()> {
         }
     }
 
-    eprintln!("Successfully deleted {count} content files from disk!");
+    eprintln!("Sucessfully deleted {count} content files from disk!");
 
     Ok(())
 }

milli/Cargo.toml

@@ -74,10 +74,10 @@ csv = "1.3.0"
 candle-core = { version = "0.4.1" }
 candle-transformers = { version = "0.4.1" }
 candle-nn = { version = "0.4.1" }
-tokenizers = { git = "https://github.com/huggingface/tokenizers.git", tag = "v0.15.2", version = "0.15.2", default_features = false, features = [
+tokenizers = { git = "https://github.com/huggingface/tokenizers.git", tag = "v0.15.2", version = "0.15.2", default-features = false, features = [
     "onig",
 ] }
-hf-hub = { git = "https://github.com/dureuill/hf-hub.git", branch = "rust_tls", default_features = false, features = [
+hf-hub = { git = "https://github.com/dureuill/hf-hub.git", branch = "rust_tls", default-features = false, features = [
     "online",
 ] }
 tiktoken-rs = "0.5.8"
@@ -85,7 +85,7 @@ liquid = "0.26.4"
 arroy = "0.2.0"
 rand = "0.8.5"
 tracing = "0.1.40"
-ureq = { version = "2.9.7", features = ["json"] }
+ureq = { version = "2.9.6", features = ["json"] }
 url = "2.5.0"
 
 [dev-dependencies]

milli/src/documents/builder.rs

@@ -203,7 +203,7 @@ fn parse_csv_header(header: &str) -> (&str, AllowedType) {
             "string" => (field_name, AllowedType::String),
             "boolean" => (field_name, AllowedType::Boolean),
             "number" => (field_name, AllowedType::Number),
-            // if the pattern isn't recognized, we keep the whole field.
+            // if the pattern isn't reconized, we keep the whole field.
             _otherwise => (header, AllowedType::String),
         },
         None => (header, AllowedType::String),

milli/src/index.rs

@@ -22,7 +22,7 @@ use crate::heed_codec::{
 };
 use crate::order_by_map::OrderByMap;
 use crate::proximity::ProximityPrecision;
-use crate::vector::EmbeddingConfig;
+use crate::vector::{Embedding, EmbeddingConfig};
 use crate::{
     default_criteria, CboRoaringBitmapCodec, Criterion, DocumentId, ExternalDocumentsIds,
     FacetDistribution, FieldDistribution, FieldId, FieldIdWordCountCodec, GeoPoint, ObkvCodec,
@@ -1516,6 +1516,42 @@ impl Index {
             .unwrap_or_default())
     }
 
+    pub fn embeddings(
+        &self,
+        rtxn: &RoTxn<'_>,
+        docid: DocumentId,
+    ) -> Result<BTreeMap<String, Vec<Embedding>>> {
+        let mut res = BTreeMap::new();
+        for row in self.embedder_category_id.iter(rtxn)? {
+            let (embedder_name, embedder_id) = row?;
+            let embedder_id = (embedder_id as u16) << 8;
+            let mut embeddings = Vec::new();
+            'vectors: for i in 0..=u8::MAX {
+                let reader = arroy::Reader::open(rtxn, embedder_id | (i as u16), self.vector_arroy)
+                    .map(Some)
+                    .or_else(|e| match e {
+                        arroy::Error::MissingMetadata => Ok(None),
+                        e => Err(e),
+                    })
+                    .transpose();
+
+                let Some(reader) = reader else {
+                    break 'vectors;
+                };
+
+                let embedding = reader?.item_vector(rtxn, docid)?;
+                if let Some(embedding) = embedding {
+                    embeddings.push(embedding)
+                } else {
+                    break 'vectors;
+                }
+            }
+
+            res.insert(embedder_name.to_owned(), embeddings);
+        }
+        Ok(res)
+    }
+
     pub(crate) fn put_search_cutoff(&self, wtxn: &mut RwTxn<'_>, cutoff: u64) -> heed::Result<()> {
         self.main.remap_types::<Str, BEU64>().put(wtxn, main_key::SEARCH_CUTOFF, &cutoff)
     }

milli/src/search/new/geo_sort.rs

@@ -42,7 +42,7 @@ fn facet_number_values<'a>(
 }
 
 /// Define the strategy used by the geo sort.
-/// The parameter represents the cache size, and, in the case of the Dynamic strategy,
+/// The paramater represents the cache size, and, in the case of the Dynamic strategy,
 /// the point where we move from using the iterative strategy to the rtree.
 #[derive(Debug, Clone, Copy)]
 pub enum Strategy {

milli/src/search/new/logger/visual.rs

@@ -22,7 +22,7 @@ pub enum SearchEvents {
     RankingRuleStartIteration { ranking_rule_idx: usize, universe_len: u64 },
     RankingRuleNextBucket { ranking_rule_idx: usize, universe_len: u64, bucket_len: u64 },
     RankingRuleSkipBucket { ranking_rule_idx: usize, bucket_len: u64 },
-    RankingRuleEndIteration { ranking_rule_idx: usize, universe_len: u64 },
+    RankingRuleEndIteration { ranking_rule_idx: usize },
     ExtendResults { new: Vec<u32> },
     ProximityGraph { graph: RankingRuleGraph<ProximityGraph> },
     ProximityPaths { paths: Vec<Vec<Interned<ProximityCondition>>> },
@@ -123,12 +123,9 @@ impl SearchLogger<QueryGraph> for VisualSearchLogger {
         &mut self,
         ranking_rule_idx: usize,
         _ranking_rule: &dyn RankingRule<QueryGraph>,
-        universe: &RoaringBitmap,
+        _universe: &RoaringBitmap,
     ) {
-        self.events.push(SearchEvents::RankingRuleEndIteration {
-            ranking_rule_idx,
-            universe_len: universe.len(),
-        });
+        self.events.push(SearchEvents::RankingRuleEndIteration { ranking_rule_idx });
         self.location.pop();
     }
     fn add_to_results(&mut self, docids: &[u32]) {
@@ -326,7 +323,7 @@ impl<'ctx> DetailedLoggerFinish<'ctx> {
                 assert!(ranking_rule_idx == self.rr_action_counter.len() - 1);
                 self.write_skip_bucket(bucket_len)?;
             }
-            SearchEvents::RankingRuleEndIteration { ranking_rule_idx, universe_len: _ } => {
+            SearchEvents::RankingRuleEndIteration { ranking_rule_idx } => {
                 assert!(ranking_rule_idx == self.rr_action_counter.len() - 1);
                 self.write_end_iteration()?;
             }

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

@@ -134,7 +134,7 @@ impl<'t> Matcher<'t, '_> {
             for (token_position, word_position, word) in words_positions {
                 partial = match partial.match_token(word) {
                     // token matches the partial match, but the match is not full,
-                    // we temporarily save the current token then we try to match the next one.
+                    // we temporarly save the current token then we try to match the next one.
                     Some(MatchType::Partial(partial)) => {
                         potential_matches.push((token_position, word_position, partial.char_len()));
                         partial
@@ -722,7 +722,7 @@ mod tests {
             @"…void void void void void split the world void void"
         );
 
-        // Text containing matches with different density.
+        // Text containing matches with diferent density.
         let text = "split void the void void world void void void void void void void void void void split the world void void";
         let mut matcher = builder.build(text);
         // crop should return 10 last words with a marker at the start.

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

@@ -119,7 +119,7 @@ pub fn located_query_terms_from_tokens(
                             if let Some(located_query_term) = phrase.build(ctx) {
                                 // as we are evaluating a negative operator we put the phrase
                                 // in the negative one *but* we don't reset the negative operator
-                                // as we are immediately starting a new negative phrase.
+                                // as we are immediatly starting a new negative phrase.
                                 if negative_phrase {
                                     negative_phrases.push(located_query_term);
                                 } else {

milli/src/update/facet/incremental.rs

@@ -499,7 +499,7 @@ impl FacetsUpdateIncrementalInner {
                     ModificationResult::Expand | ModificationResult::Reduce { .. }
                 )
             {
-                // if any modification occurred, insert it in the database.
+                // if any modification occured, insert it in the database.
                 self.db.put(txn, &insertion_key.as_ref(), &updated_value)?;
                 Ok(insertion_key_modification)
             } else {

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

@@ -37,7 +37,7 @@ pub struct ExtractedFacetValues {
 
 /// Extracts the facet values of each faceted field of each document.
 ///
-/// Returns the generated grenad reader containing the docid the fid and the original value as key
+/// Returns the generated grenad reader containing the docid the fid and the orginal value as key
 /// and the normalized value as value extracted from the given chunk of documents.
 /// We need the fid of the geofields to correctly parse them as numbers if they were sent as strings initially.
 #[tracing::instrument(level = "trace", skip_all, target = "indexing::extract")]
@@ -45,7 +45,6 @@ pub fn extract_fid_docid_facet_values<R: io::Read + io::Seek>(
     obkv_documents: grenad::Reader<R>,
     indexer: GrenadParameters,
     settings_diff: &InnerIndexSettingsDiff,
-    geo_fields_ids: Option<(FieldId, FieldId)>,
 ) -> Result<ExtractedFacetValues> {
     puffin::profile_function!();
 
@@ -127,12 +126,18 @@ pub fn extract_fid_docid_facet_values<R: io::Read + io::Seek>(
                     add_exists.insert(document);
                 }
 
-                let geo_support =
-                    geo_fields_ids.map_or(false, |(lat, lng)| field_id == lat || field_id == lng);
+                let del_geo_support = settings_diff
+                    .old
+                    .geo_fields_ids
+                    .map_or(false, |(lat, lng)| field_id == lat || field_id == lng);
+                let add_geo_support = settings_diff
+                    .new
+                    .geo_fields_ids
+                    .map_or(false, |(lat, lng)| field_id == lat || field_id == lng);
                 let del_filterable_values =
-                    del_value.map(|value| extract_facet_values(&value, geo_support));
+                    del_value.map(|value| extract_facet_values(&value, del_geo_support));
                 let add_filterable_values =
-                    add_value.map(|value| extract_facet_values(&value, geo_support));
+                    add_value.map(|value| extract_facet_values(&value, add_geo_support));
 
                 // Those closures are just here to simplify things a bit.
                 let mut insert_numbers_diff = |del_numbers, add_numbers| {

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

@@ -8,6 +8,7 @@ use super::helpers::{create_writer, writer_into_reader, GrenadParameters};
 use crate::error::GeoError;
 use crate::update::del_add::{DelAdd, KvReaderDelAdd, KvWriterDelAdd};
 use crate::update::index_documents::extract_finite_float_from_value;
+use crate::update::settings::{InnerIndexSettings, InnerIndexSettingsDiff};
 use crate::{FieldId, InternalError, Result};
 
 /// Extracts the geographical coordinates contained in each document under the `_geo` field.
@@ -18,7 +19,7 @@ pub fn extract_geo_points<R: io::Read + io::Seek>(
     obkv_documents: grenad::Reader<R>,
     indexer: GrenadParameters,
     primary_key_id: FieldId,
-    (lat_fid, lng_fid): (FieldId, FieldId),
+    settings_diff: &InnerIndexSettingsDiff,
 ) -> Result<grenad::Reader<BufReader<File>>> {
     puffin::profile_function!();
 
@@ -40,47 +41,27 @@ pub fn extract_geo_points<R: io::Read + io::Seek>(
             serde_json::from_slice(document_id).unwrap()
         };
 
-        // first we get the two fields
-        match (obkv.get(lat_fid), obkv.get(lng_fid)) {
-            (Some(lat), Some(lng)) => {
-                let deladd_lat_obkv = KvReaderDelAdd::new(lat);
-                let deladd_lng_obkv = KvReaderDelAdd::new(lng);
+        // extract old version
+        let del_lat_lng =
+            extract_lat_lng(&obkv, &settings_diff.old, DelAdd::Deletion, document_id)?;
+        // extract new version
+        let add_lat_lng =
+            extract_lat_lng(&obkv, &settings_diff.new, DelAdd::Addition, document_id)?;
 
-                // then we extract the values
-                let del_lat_lng = deladd_lat_obkv
-                    .get(DelAdd::Deletion)
-                    .zip(deladd_lng_obkv.get(DelAdd::Deletion))
-                    .map(|(lat, lng)| extract_lat_lng(lat, lng, document_id))
-                    .transpose()?;
-                let add_lat_lng = deladd_lat_obkv
-                    .get(DelAdd::Addition)
-                    .zip(deladd_lng_obkv.get(DelAdd::Addition))
-                    .map(|(lat, lng)| extract_lat_lng(lat, lng, document_id))
-                    .transpose()?;
-
-                if del_lat_lng != add_lat_lng {
-                    let mut obkv = KvWriterDelAdd::memory();
-                    if let Some([lat, lng]) = del_lat_lng {
-                        #[allow(clippy::drop_non_drop)]
-                        let bytes: [u8; 16] = concat_arrays![lat.to_ne_bytes(), lng.to_ne_bytes()];
-                        obkv.insert(DelAdd::Deletion, bytes)?;
-                    }
-                    if let Some([lat, lng]) = add_lat_lng {
-                        #[allow(clippy::drop_non_drop)]
-                        let bytes: [u8; 16] = concat_arrays![lat.to_ne_bytes(), lng.to_ne_bytes()];
-                        obkv.insert(DelAdd::Addition, bytes)?;
-                    }
-                    let bytes = obkv.into_inner()?;
-                    writer.insert(docid_bytes, bytes)?;
-                }
+        if del_lat_lng != add_lat_lng {
+            let mut obkv = KvWriterDelAdd::memory();
+            if let Some([lat, lng]) = del_lat_lng {
+                #[allow(clippy::drop_non_drop)]
+                let bytes: [u8; 16] = concat_arrays![lat.to_ne_bytes(), lng.to_ne_bytes()];
+                obkv.insert(DelAdd::Deletion, bytes)?;
             }
-            (None, Some(_)) => {
-                return Err(GeoError::MissingLatitude { document_id: document_id() }.into())
+            if let Some([lat, lng]) = add_lat_lng {
+                #[allow(clippy::drop_non_drop)]
+                let bytes: [u8; 16] = concat_arrays![lat.to_ne_bytes(), lng.to_ne_bytes()];
+                obkv.insert(DelAdd::Addition, bytes)?;
             }
-            (Some(_), None) => {
-                return Err(GeoError::MissingLongitude { document_id: document_id() }.into())
-            }
-            (None, None) => (),
+            let bytes = obkv.into_inner()?;
+            writer.insert(docid_bytes, bytes)?;
         }
     }
 
@@ -88,16 +69,37 @@ pub fn extract_geo_points<R: io::Read + io::Seek>(
 }
 
 /// Extract the finite floats lat and lng from two bytes slices.
-fn extract_lat_lng(lat: &[u8], lng: &[u8], document_id: impl Fn() -> Value) -> Result<[f64; 2]> {
-    let lat = extract_finite_float_from_value(
-        serde_json::from_slice(lat).map_err(InternalError::SerdeJson)?,
-    )
-    .map_err(|lat| GeoError::BadLatitude { document_id: document_id(), value: lat })?;
+fn extract_lat_lng(
+    document: &obkv::KvReader<FieldId>,
+    settings: &InnerIndexSettings,
+    deladd: DelAdd,
+    document_id: impl Fn() -> Value,
+) -> Result<Option<[f64; 2]>> {
+    match settings.geo_fields_ids {
+        Some((lat_fid, lng_fid)) => {
+            let lat = document.get(lat_fid).map(KvReaderDelAdd::new).and_then(|r| r.get(deladd));
+            let lng = document.get(lng_fid).map(KvReaderDelAdd::new).and_then(|r| r.get(deladd));
+            let (lat, lng) = match (lat, lng) {
+                (Some(lat), Some(lng)) => (lat, lng),
+                (Some(_), None) => {
+                    return Err(GeoError::MissingLatitude { document_id: document_id() }.into())
+                }
+                (None, Some(_)) => {
+                    return Err(GeoError::MissingLongitude { document_id: document_id() }.into())
+                }
+                (None, None) => return Ok(None),
+            };
+            let lat = extract_finite_float_from_value(
+                serde_json::from_slice(lat).map_err(InternalError::SerdeJson)?,
+            )
+            .map_err(|lat| GeoError::BadLatitude { document_id: document_id(), value: lat })?;
 
-    let lng = extract_finite_float_from_value(
-        serde_json::from_slice(lng).map_err(InternalError::SerdeJson)?,
-    )
-    .map_err(|lng| GeoError::BadLongitude { document_id: document_id(), value: lng })?;
-
-    Ok([lat, lng])
+            let lng = extract_finite_float_from_value(
+                serde_json::from_slice(lng).map_err(InternalError::SerdeJson)?,
+            )
+            .map_err(|lng| GeoError::BadLongitude { document_id: document_id(), value: lng })?;
+            Ok(Some([lat, lng]))
+        }
+        None => Ok(None),
+    }
 }

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

@@ -11,7 +11,7 @@ mod extract_word_position_docids;
 
 use std::fs::File;
 use std::io::BufReader;
-use std::sync::Arc;
+use std::sync::{Arc, OnceLock};
 
 use crossbeam_channel::Sender;
 use rayon::prelude::*;
@@ -31,7 +31,7 @@ use self::extract_word_position_docids::extract_word_position_docids;
 use super::helpers::{as_cloneable_grenad, CursorClonableMmap, GrenadParameters};
 use super::{helpers, TypedChunk};
 use crate::update::settings::InnerIndexSettingsDiff;
-use crate::{FieldId, Result, ThreadPoolNoAbortBuilder};
+use crate::{FieldId, Result, ThreadPoolNoAbort, ThreadPoolNoAbortBuilder};
 
 /// Extract data for each databases from obkv documents in parallel.
 /// Send data in grenad file over provided Sender.
@@ -43,7 +43,6 @@ pub(crate) fn data_from_obkv_documents(
     indexer: GrenadParameters,
     lmdb_writer_sx: Sender<Result<TypedChunk>>,
     primary_key_id: FieldId,
-    geo_fields_ids: Option<(FieldId, FieldId)>,
     settings_diff: Arc<InnerIndexSettingsDiff>,
     max_positions_per_attributes: Option<u32>,
 ) -> Result<()> {
@@ -72,7 +71,6 @@ pub(crate) fn data_from_obkv_documents(
                         indexer,
                         lmdb_writer_sx.clone(),
                         primary_key_id,
-                        geo_fields_ids,
                         settings_diff.clone(),
                         max_positions_per_attributes,
                     )
@@ -215,6 +213,18 @@ fn run_extraction_task<FE, FS, M>(
     })
 }
 
+fn request_threads() -> &'static ThreadPoolNoAbort {
+    static REQUEST_THREADS: OnceLock<ThreadPoolNoAbort> = OnceLock::new();
+
+    REQUEST_THREADS.get_or_init(|| {
+        ThreadPoolNoAbortBuilder::new()
+            .num_threads(crate::vector::REQUEST_PARALLELISM)
+            .thread_name(|index| format!("embedding-request-{index}"))
+            .build()
+            .unwrap()
+    })
+}
+
 /// Extract chunked data and send it into lmdb_writer_sx sender:
 /// - documents
 fn send_original_documents_data(
@@ -229,11 +239,6 @@ fn send_original_documents_data(
     let documents_chunk_cloned = original_documents_chunk.clone();
     let lmdb_writer_sx_cloned = lmdb_writer_sx.clone();
 
-    let request_threads = ThreadPoolNoAbortBuilder::new()
-        .num_threads(crate::vector::REQUEST_PARALLELISM)
-        .thread_name(|index| format!("embedding-request-{index}"))
-        .build()?;
-
     if settings_diff.reindex_vectors() || !settings_diff.settings_update_only() {
         let settings_diff = settings_diff.clone();
         rayon::spawn(move || {
@@ -251,7 +256,7 @@ fn send_original_documents_data(
                             prompts,
                             indexer,
                             embedder.clone(),
-                            &request_threads,
+                            request_threads(),
                         ) {
                             Ok(results) => Some(results),
                             Err(error) => {
@@ -300,7 +305,6 @@ fn send_and_extract_flattened_documents_data(
     indexer: GrenadParameters,
     lmdb_writer_sx: Sender<Result<TypedChunk>>,
     primary_key_id: FieldId,
-    geo_fields_ids: Option<(FieldId, FieldId)>,
     settings_diff: Arc<InnerIndexSettingsDiff>,
     max_positions_per_attributes: Option<u32>,
 ) -> Result<(
@@ -310,12 +314,13 @@ fn send_and_extract_flattened_documents_data(
     let flattened_documents_chunk =
         flattened_documents_chunk.and_then(|c| unsafe { as_cloneable_grenad(&c) })?;
 
-    if let Some(geo_fields_ids) = geo_fields_ids {
+    if settings_diff.run_geo_indexing() {
         let documents_chunk_cloned = flattened_documents_chunk.clone();
         let lmdb_writer_sx_cloned = lmdb_writer_sx.clone();
+        let settings_diff = settings_diff.clone();
         rayon::spawn(move || {
             let result =
-                extract_geo_points(documents_chunk_cloned, indexer, primary_key_id, geo_fields_ids);
+                extract_geo_points(documents_chunk_cloned, indexer, primary_key_id, &settings_diff);
             let _ = match result {
                 Ok(geo_points) => lmdb_writer_sx_cloned.send(Ok(TypedChunk::GeoPoints(geo_points))),
                 Err(error) => lmdb_writer_sx_cloned.send(Err(error)),
@@ -354,7 +359,6 @@ fn send_and_extract_flattened_documents_data(
                     flattened_documents_chunk.clone(),
                     indexer,
                     &settings_diff,
-                    geo_fields_ids,
                 )?;
 
                 // send fid_docid_facet_numbers_chunk to DB writer

milli/src/update/index_documents/mod.rs

@@ -324,28 +324,6 @@ where
         // get the primary key field id
         let primary_key_id = settings_diff.new.fields_ids_map.id(&primary_key).unwrap();
 
-        // get the fid of the `_geo.lat` and `_geo.lng` fields.
-        let mut field_id_map = self.index.fields_ids_map(self.wtxn)?;
-
-        // self.index.fields_ids_map($a)? ==>> field_id_map
-        let geo_fields_ids = match field_id_map.id("_geo") {
-            Some(gfid) => {
-                let is_sortable = self.index.sortable_fields_ids(self.wtxn)?.contains(&gfid);
-                let is_filterable = self.index.filterable_fields_ids(self.wtxn)?.contains(&gfid);
-                // if `_geo` is faceted then we get the `lat` and `lng`
-                if is_sortable || is_filterable {
-                    let field_ids = field_id_map
-                        .insert("_geo.lat")
-                        .zip(field_id_map.insert("_geo.lng"))
-                        .ok_or(UserError::AttributeLimitReached)?;
-                    Some(field_ids)
-                } else {
-                    None
-                }
-            }
-            None => None,
-        };
-
         let pool_params = GrenadParameters {
             chunk_compression_type: self.indexer_config.chunk_compression_type,
             chunk_compression_level: self.indexer_config.chunk_compression_level,
@@ -412,7 +390,6 @@ where
                         pool_params,
                         lmdb_writer_sx.clone(),
                         primary_key_id,
-                        geo_fields_ids,
                         settings_diff.clone(),
                         max_positions_per_attributes,
                     )

milli/src/update/index_documents/transform.rs

@@ -48,7 +48,6 @@ pub struct Transform<'a, 'i> {
     fields_ids_map: FieldsIdsMap,
 
     indexer_settings: &'a IndexerConfig,
-    pub autogenerate_docids: bool,
     pub index_documents_method: IndexDocumentsMethod,
     available_documents_ids: AvailableDocumentsIds,
 
@@ -102,7 +101,7 @@ impl<'a, 'i> Transform<'a, 'i> {
         index: &'i Index,
         indexer_settings: &'a IndexerConfig,
         index_documents_method: IndexDocumentsMethod,
-        autogenerate_docids: bool,
+        _autogenerate_docids: bool,
     ) -> Result<Self> {
         // We must choose the appropriate merge function for when two or more documents
         // with the same user id must be merged or fully replaced in the same batch.
@@ -136,7 +135,6 @@ impl<'a, 'i> Transform<'a, 'i> {
             index,
             fields_ids_map: index.fields_ids_map(wtxn)?,
             indexer_settings,
-            autogenerate_docids,
             available_documents_ids: AvailableDocumentsIds::from_documents_ids(&documents_ids),
             original_sorter,
             flattened_sorter,

milli/src/update/settings.rs

@@ -1161,6 +1161,11 @@ impl InnerIndexSettingsDiff {
     pub fn settings_update_only(&self) -> bool {
         self.settings_update_only
     }
+
+    pub fn run_geo_indexing(&self) -> bool {
+        self.old.geo_fields_ids != self.new.geo_fields_ids
+            || (!self.settings_update_only && self.new.geo_fields_ids.is_some())
+    }
 }
 
 #[derive(Clone)]
@@ -1177,6 +1182,7 @@ pub(crate) struct InnerIndexSettings {
     pub proximity_precision: ProximityPrecision,
     pub embedding_configs: EmbeddingConfigs,
     pub existing_fields: HashSet<String>,
+    pub geo_fields_ids: Option<(FieldId, FieldId)>,
 }
 
 impl InnerIndexSettings {
@@ -1185,7 +1191,7 @@ impl InnerIndexSettings {
         let stop_words = stop_words.map(|sw| sw.map_data(Vec::from).unwrap());
         let allowed_separators = index.allowed_separators(rtxn)?;
         let dictionary = index.dictionary(rtxn)?;
-        let fields_ids_map = index.fields_ids_map(rtxn)?;
+        let mut fields_ids_map = index.fields_ids_map(rtxn)?;
         let user_defined_searchable_fields = index.user_defined_searchable_fields(rtxn)?;
         let user_defined_searchable_fields =
             user_defined_searchable_fields.map(|sf| sf.into_iter().map(String::from).collect());
@@ -1200,6 +1206,24 @@ impl InnerIndexSettings {
             .into_iter()
             .filter_map(|(field, count)| (count != 0).then_some(field))
             .collect();
+        // index.fields_ids_map($a)? ==>> fields_ids_map
+        let geo_fields_ids = match fields_ids_map.id("_geo") {
+            Some(gfid) => {
+                let is_sortable = index.sortable_fields_ids(rtxn)?.contains(&gfid);
+                let is_filterable = index.filterable_fields_ids(rtxn)?.contains(&gfid);
+                // if `_geo` is faceted then we get the `lat` and `lng`
+                if is_sortable || is_filterable {
+                    let field_ids = fields_ids_map
+                        .insert("_geo.lat")
+                        .zip(fields_ids_map.insert("_geo.lng"))
+                        .ok_or(UserError::AttributeLimitReached)?;
+                    Some(field_ids)
+                } else {
+                    None
+                }
+            }
+            None => None,
+        };
 
         Ok(Self {
             stop_words,
@@ -1214,6 +1238,7 @@ impl InnerIndexSettings {
             proximity_precision,
             embedding_configs,
             existing_fields,
+            geo_fields_ids,
         })
     }
 

xtask/Cargo.toml

@@ -21,7 +21,7 @@ reqwest = { version = "0.11.23", features = [
     "stream",
     "json",
     "rustls-tls",
-], default_features = false }
+], default-features = false }
 serde = { version = "1.0.195", features = ["derive"] }
 serde_json = "1.0.111"
 sha2 = "0.10.8"