Change implementation of MergedDocuments::iter_top_level_fields

5131: Ignore documents whose selected fields didn't change r=dureuill a=dureuill

Attempts to improve the new indexer performance by ignoring documents whose selected fields didn't change:

- Add `Update::has_changed_for_fields` function
- Ignore documents whose searchable attributes didn't change for word docids and word pair proximity extraction
- Ignore documents whose faceted attributes didn't change for facet extraction

Co-authored-by: Louis Dureuil <louis@meilisearch.com>

Cargo.lock

@@ -489,6 +489,11 @@ version = "0.22.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "72b3254f16251a8381aa12e40e3c4d2f0199f8c6508fbecb9d91f575e0fbb8c6"
 
+[[package]]
+name = "bbqueue"
+version = "0.5.1"
+source = "git+https://github.com/meilisearch/bbqueue#cbb87cc707b5af415ef203bdaf2443e06ba0d6d4"
+
 [[package]]
 name = "benchmarks"
 version = "1.12.0"
@@ -1246,19 +1251,6 @@ dependencies = [
  "itertools 0.10.5",
 ]
 
-[[package]]
-name = "crossbeam"
-version = "0.8.4"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "1137cd7e7fc0fb5d3c5a8678be38ec56e819125d8d7907411fe24ccb943faca8"
-dependencies = [
- "crossbeam-channel",
- "crossbeam-deque",
- "crossbeam-epoch",
- "crossbeam-queue",
- "crossbeam-utils",
-]
-
 [[package]]
 name = "crossbeam-channel"
 version = "0.5.13"
@@ -1918,6 +1910,15 @@ dependencies = [
  "serde_json",
 ]
 
+[[package]]
+name = "flume"
+version = "0.11.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "da0e4dd2a88388a1f4ccc7c9ce104604dab68d9f408dc34cd45823d5a9069095"
+dependencies = [
+ "spin",
+]
+
 [[package]]
 name = "fnv"
 version = "1.0.7"
@@ -2616,7 +2617,7 @@ dependencies = [
  "big_s",
  "bincode",
  "bumpalo",
- "crossbeam",
+ "crossbeam-channel",
  "csv",
  "derive_builder 0.20.0",
  "dump",
@@ -3611,6 +3612,7 @@ version = "1.12.0"
 dependencies = [
  "allocator-api2",
  "arroy 0.5.0 (registry+https://github.com/rust-lang/crates.io-index)",
+ "bbqueue",
  "big_s",
  "bimap",
  "bincode",
@@ -3630,6 +3632,7 @@ dependencies = [
  "enum-iterator",
  "filter-parser",
  "flatten-serde-json",
+ "flume",
  "fst",
  "fxhash",
  "geoutils",
@@ -4743,8 +4746,9 @@ dependencies = [
 
 [[package]]
 name = "roaring"
-version = "0.10.6"
-source = "git+https://github.com/RoaringBitmap/roaring-rs?branch=clone-iter-slice#8ff028e484fb6192a0acf5a669eaf18c30cada6e"
+version = "0.10.7"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f81dc953b2244ddd5e7860cb0bb2a790494b898ef321d4aff8e260efab60cc88"
 dependencies = [
  "bytemuck",
  "byteorder",
@@ -5186,6 +5190,9 @@ name = "spin"
 version = "0.9.8"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "6980e8d7511241f8acf4aebddbb1ff938df5eebe98691418c4468d0b72a96a67"
+dependencies = [
+ "lock_api",
+]
 
 [[package]]
 name = "spm_precompiled"

Cargo.toml

@@ -43,6 +43,3 @@ opt-level = 3
 opt-level = 3
 [profile.dev.package.roaring]
 opt-level = 3
-
-[patch.crates-io]
-roaring = { git = "https://github.com/RoaringBitmap/roaring-rs", branch = "clone-iter-slice" }

crates/benchmarks/Cargo.toml

@@ -24,7 +24,7 @@ tempfile = "3.14.0"
 criterion = { version = "0.5.1", features = ["html_reports"] }
 rand = "0.8.5"
 rand_chacha = "0.3.1"
-roaring = "0.10.6"
+roaring = "0.10.7"
 
 [build-dependencies]
 anyhow = "1.0.86"

crates/benchmarks/benches/indexing.rs

@@ -16,6 +16,7 @@ use rand::seq::SliceRandom;
 use rand_chacha::rand_core::SeedableRng;
 use roaring::RoaringBitmap;
 
+#[cfg(not(windows))]
 #[global_allocator]
 static ALLOC: mimalloc::MiMalloc = mimalloc::MiMalloc;
 
@@ -157,6 +158,7 @@ fn indexing_songs_default(c: &mut Criterion) {
                 indexer::index(
                     &mut wtxn,
                     &index,
+                    &milli::ThreadPoolNoAbortBuilder::new().build().unwrap(),
                     config.grenad_parameters(),
                     &db_fields_ids_map,
                     new_fields_ids_map,
@@ -223,6 +225,7 @@ fn reindexing_songs_default(c: &mut Criterion) {
                 indexer::index(
                     &mut wtxn,
                     &index,
+                    &milli::ThreadPoolNoAbortBuilder::new().build().unwrap(),
                     config.grenad_parameters(),
                     &db_fields_ids_map,
                     new_fields_ids_map,
@@ -267,6 +270,7 @@ fn reindexing_songs_default(c: &mut Criterion) {
                 indexer::index(
                     &mut wtxn,
                     &index,
+                    &milli::ThreadPoolNoAbortBuilder::new().build().unwrap(),
                     config.grenad_parameters(),
                     &db_fields_ids_map,
                     new_fields_ids_map,
@@ -335,6 +339,7 @@ fn deleting_songs_in_batches_default(c: &mut Criterion) {
                 indexer::index(
                     &mut wtxn,
                     &index,
+                    &milli::ThreadPoolNoAbortBuilder::new().build().unwrap(),
                     config.grenad_parameters(),
                     &db_fields_ids_map,
                     new_fields_ids_map,
@@ -411,6 +416,7 @@ fn indexing_songs_in_three_batches_default(c: &mut Criterion) {
                 indexer::index(
                     &mut wtxn,
                     &index,
+                    &milli::ThreadPoolNoAbortBuilder::new().build().unwrap(),
                     config.grenad_parameters(),
                     &db_fields_ids_map,
                     new_fields_ids_map,
@@ -455,6 +461,7 @@ fn indexing_songs_in_three_batches_default(c: &mut Criterion) {
                 indexer::index(
                     &mut wtxn,
                     &index,
+                    &milli::ThreadPoolNoAbortBuilder::new().build().unwrap(),
                     config.grenad_parameters(),
                     &db_fields_ids_map,
                     new_fields_ids_map,
@@ -495,6 +502,7 @@ fn indexing_songs_in_three_batches_default(c: &mut Criterion) {
                 indexer::index(
                     &mut wtxn,
                     &index,
+                    &milli::ThreadPoolNoAbortBuilder::new().build().unwrap(),
                     config.grenad_parameters(),
                     &db_fields_ids_map,
                     new_fields_ids_map,
@@ -562,6 +570,7 @@ fn indexing_songs_without_faceted_numbers(c: &mut Criterion) {
                 indexer::index(
                     &mut wtxn,
                     &index,
+                    &milli::ThreadPoolNoAbortBuilder::new().build().unwrap(),
                     config.grenad_parameters(),
                     &db_fields_ids_map,
                     new_fields_ids_map,
@@ -628,6 +637,7 @@ fn indexing_songs_without_faceted_fields(c: &mut Criterion) {
                 indexer::index(
                     &mut wtxn,
                     &index,
+                    &milli::ThreadPoolNoAbortBuilder::new().build().unwrap(),
                     config.grenad_parameters(),
                     &db_fields_ids_map,
                     new_fields_ids_map,
@@ -694,6 +704,7 @@ fn indexing_wiki(c: &mut Criterion) {
                 indexer::index(
                     &mut wtxn,
                     &index,
+                    &milli::ThreadPoolNoAbortBuilder::new().build().unwrap(),
                     config.grenad_parameters(),
                     &db_fields_ids_map,
                     new_fields_ids_map,
@@ -759,6 +770,7 @@ fn reindexing_wiki(c: &mut Criterion) {
                 indexer::index(
                     &mut wtxn,
                     &index,
+                    &milli::ThreadPoolNoAbortBuilder::new().build().unwrap(),
                     config.grenad_parameters(),
                     &db_fields_ids_map,
                     new_fields_ids_map,
@@ -803,6 +815,7 @@ fn reindexing_wiki(c: &mut Criterion) {
                 indexer::index(
                     &mut wtxn,
                     &index,
+                    &milli::ThreadPoolNoAbortBuilder::new().build().unwrap(),
                     config.grenad_parameters(),
                     &db_fields_ids_map,
                     new_fields_ids_map,
@@ -870,6 +883,7 @@ fn deleting_wiki_in_batches_default(c: &mut Criterion) {
                 indexer::index(
                     &mut wtxn,
                     &index,
+                    &milli::ThreadPoolNoAbortBuilder::new().build().unwrap(),
                     config.grenad_parameters(),
                     &db_fields_ids_map,
                     new_fields_ids_map,
@@ -946,6 +960,7 @@ fn indexing_wiki_in_three_batches(c: &mut Criterion) {
                 indexer::index(
                     &mut wtxn,
                     &index,
+                    &milli::ThreadPoolNoAbortBuilder::new().build().unwrap(),
                     config.grenad_parameters(),
                     &db_fields_ids_map,
                     new_fields_ids_map,
@@ -991,6 +1006,7 @@ fn indexing_wiki_in_three_batches(c: &mut Criterion) {
                 indexer::index(
                     &mut wtxn,
                     &index,
+                    &milli::ThreadPoolNoAbortBuilder::new().build().unwrap(),
                     config.grenad_parameters(),
                     &db_fields_ids_map,
                     new_fields_ids_map,
@@ -1032,6 +1048,7 @@ fn indexing_wiki_in_three_batches(c: &mut Criterion) {
                 indexer::index(
                     &mut wtxn,
                     &index,
+                    &milli::ThreadPoolNoAbortBuilder::new().build().unwrap(),
                     config.grenad_parameters(),
                     &db_fields_ids_map,
                     new_fields_ids_map,
@@ -1098,6 +1115,7 @@ fn indexing_movies_default(c: &mut Criterion) {
                 indexer::index(
                     &mut wtxn,
                     &index,
+                    &milli::ThreadPoolNoAbortBuilder::new().build().unwrap(),
                     config.grenad_parameters(),
                     &db_fields_ids_map,
                     new_fields_ids_map,
@@ -1163,6 +1181,7 @@ fn reindexing_movies_default(c: &mut Criterion) {
                 indexer::index(
                     &mut wtxn,
                     &index,
+                    &milli::ThreadPoolNoAbortBuilder::new().build().unwrap(),
                     config.grenad_parameters(),
                     &db_fields_ids_map,
                     new_fields_ids_map,
@@ -1207,6 +1226,7 @@ fn reindexing_movies_default(c: &mut Criterion) {
                 indexer::index(
                     &mut wtxn,
                     &index,
+                    &milli::ThreadPoolNoAbortBuilder::new().build().unwrap(),
                     config.grenad_parameters(),
                     &db_fields_ids_map,
                     new_fields_ids_map,
@@ -1274,6 +1294,7 @@ fn deleting_movies_in_batches_default(c: &mut Criterion) {
                 indexer::index(
                     &mut wtxn,
                     &index,
+                    &milli::ThreadPoolNoAbortBuilder::new().build().unwrap(),
                     config.grenad_parameters(),
                     &db_fields_ids_map,
                     new_fields_ids_map,
@@ -1321,6 +1342,7 @@ fn delete_documents_from_ids(index: Index, document_ids_to_delete: Vec<RoaringBi
         indexer::index(
             &mut wtxn,
             &index,
+            &milli::ThreadPoolNoAbortBuilder::new().build().unwrap(),
             config.grenad_parameters(),
             &db_fields_ids_map,
             new_fields_ids_map,
@@ -1385,6 +1407,7 @@ fn indexing_movies_in_three_batches(c: &mut Criterion) {
                 indexer::index(
                     &mut wtxn,
                     &index,
+                    &milli::ThreadPoolNoAbortBuilder::new().build().unwrap(),
                     config.grenad_parameters(),
                     &db_fields_ids_map,
                     new_fields_ids_map,
@@ -1429,6 +1452,7 @@ fn indexing_movies_in_three_batches(c: &mut Criterion) {
                 indexer::index(
                     &mut wtxn,
                     &index,
+                    &milli::ThreadPoolNoAbortBuilder::new().build().unwrap(),
                     config.grenad_parameters(),
                     &db_fields_ids_map,
                     new_fields_ids_map,
@@ -1469,6 +1493,7 @@ fn indexing_movies_in_three_batches(c: &mut Criterion) {
                 indexer::index(
                     &mut wtxn,
                     &index,
+                    &milli::ThreadPoolNoAbortBuilder::new().build().unwrap(),
                     config.grenad_parameters(),
                     &db_fields_ids_map,
                     new_fields_ids_map,
@@ -1558,6 +1583,7 @@ fn indexing_nested_movies_default(c: &mut Criterion) {
                 indexer::index(
                     &mut wtxn,
                     &index,
+                    &milli::ThreadPoolNoAbortBuilder::new().build().unwrap(),
                     config.grenad_parameters(),
                     &db_fields_ids_map,
                     new_fields_ids_map,
@@ -1648,6 +1674,7 @@ fn deleting_nested_movies_in_batches_default(c: &mut Criterion) {
                 indexer::index(
                     &mut wtxn,
                     &index,
+                    &milli::ThreadPoolNoAbortBuilder::new().build().unwrap(),
                     config.grenad_parameters(),
                     &db_fields_ids_map,
                     new_fields_ids_map,
@@ -1730,6 +1757,7 @@ fn indexing_nested_movies_without_faceted_fields(c: &mut Criterion) {
                 indexer::index(
                     &mut wtxn,
                     &index,
+                    &milli::ThreadPoolNoAbortBuilder::new().build().unwrap(),
                     config.grenad_parameters(),
                     &db_fields_ids_map,
                     new_fields_ids_map,
@@ -1796,6 +1824,7 @@ fn indexing_geo(c: &mut Criterion) {
                 indexer::index(
                     &mut wtxn,
                     &index,
+                    &milli::ThreadPoolNoAbortBuilder::new().build().unwrap(),
                     config.grenad_parameters(),
                     &db_fields_ids_map,
                     new_fields_ids_map,
@@ -1861,6 +1890,7 @@ fn reindexing_geo(c: &mut Criterion) {
                 indexer::index(
                     &mut wtxn,
                     &index,
+                    &milli::ThreadPoolNoAbortBuilder::new().build().unwrap(),
                     config.grenad_parameters(),
                     &db_fields_ids_map,
                     new_fields_ids_map,
@@ -1905,6 +1935,7 @@ fn reindexing_geo(c: &mut Criterion) {
                 indexer::index(
                     &mut wtxn,
                     &index,
+                    &milli::ThreadPoolNoAbortBuilder::new().build().unwrap(),
                     config.grenad_parameters(),
                     &db_fields_ids_map,
                     new_fields_ids_map,
@@ -1972,6 +2003,7 @@ fn deleting_geo_in_batches_default(c: &mut Criterion) {
                 indexer::index(
                     &mut wtxn,
                     &index,
+                    &milli::ThreadPoolNoAbortBuilder::new().build().unwrap(),
                     config.grenad_parameters(),
                     &db_fields_ids_map,
                     new_fields_ids_map,

crates/benchmarks/benches/search_geo.rs

@@ -5,6 +5,7 @@ use criterion::{criterion_group, criterion_main};
 use milli::update::Settings;
 use utils::Conf;
 
+#[cfg(not(windows))]
 #[global_allocator]
 static ALLOC: mimalloc::MiMalloc = mimalloc::MiMalloc;
 

crates/benchmarks/benches/search_songs.rs

@@ -5,6 +5,7 @@ use criterion::{criterion_group, criterion_main};
 use milli::update::Settings;
 use utils::Conf;
 
+#[cfg(not(windows))]
 #[global_allocator]
 static ALLOC: mimalloc::MiMalloc = mimalloc::MiMalloc;
 

crates/benchmarks/benches/search_wiki.rs

@@ -5,6 +5,7 @@ use criterion::{criterion_group, criterion_main};
 use milli::update::Settings;
 use utils::Conf;
 
+#[cfg(not(windows))]
 #[global_allocator]
 static ALLOC: mimalloc::MiMalloc = mimalloc::MiMalloc;
 

crates/benchmarks/benches/utils.rs

@@ -117,6 +117,7 @@ pub fn base_setup(conf: &Conf) -> Index {
     indexer::index(
         &mut wtxn,
         &index,
+        &milli::ThreadPoolNoAbortBuilder::new().build().unwrap(),
         config.grenad_parameters(),
         &db_fields_ids_map,
         new_fields_ids_map,

crates/dump/Cargo.toml

@@ -17,7 +17,7 @@ http = "1.1.0"
 meilisearch-types = { path = "../meilisearch-types" }
 once_cell = "1.19.0"
 regex = "1.10.5"
-roaring = { version = "0.10.6", features = ["serde"] }
+roaring = { version = "0.10.7", features = ["serde"] }
 serde = { version = "1.0.204", features = ["derive"] }
 serde_json = { version = "1.0.120", features = ["preserve_order"] }
 tar = "0.4.41"

crates/fuzzers/src/bin/fuzz-indexing.rs

@@ -135,6 +135,7 @@ fn main() {
                             indexer::index(
                                 &mut wtxn,
                                 &index,
+                                &milli::ThreadPoolNoAbortBuilder::new().build().unwrap(),
                                 indexer_config.grenad_parameters(),
                                 &db_fields_ids_map,
                                 new_fields_ids_map,

crates/index-scheduler/Cargo.toml

@@ -24,7 +24,7 @@ meilisearch-types = { path = "../meilisearch-types" }
 page_size = "0.6.0"
 raw-collections = { git = "https://github.com/meilisearch/raw-collections.git", version = "0.1.0" }
 rayon = "1.10.0"
-roaring = { version = "0.10.6", features = ["serde"] }
+roaring = { version = "0.10.7", features = ["serde"] }
 serde = { version = "1.0.204", features = ["derive"] }
 serde_json = { version = "1.0.120", features = ["preserve_order"] }
 synchronoise = "1.0.1"
@@ -45,7 +45,7 @@ bumpalo = "3.16.0"
 [dev-dependencies]
 arroy = "0.5.0"
 big_s = "1.0.2"
-crossbeam = "0.8.4"
+crossbeam-channel = "0.5.13"
 insta = { version = "1.39.0", features = ["json", "redactions"] }
 maplit = "1.0.2"
 meili-snap = { path = "../meili-snap" }

crates/index-scheduler/src/autobatcher.rs

@@ -115,13 +115,6 @@ pub enum BatchKind {
         allow_index_creation: bool,
         settings_ids: Vec<TaskId>,
     },
-    SettingsAndDocumentOperation {
-        settings_ids: Vec<TaskId>,
-        method: IndexDocumentsMethod,
-        allow_index_creation: bool,
-        primary_key: Option<String>,
-        operation_ids: Vec<TaskId>,
-    },
     Settings {
         allow_index_creation: bool,
         settings_ids: Vec<TaskId>,
@@ -146,7 +139,6 @@ impl BatchKind {
         match self {
             BatchKind::DocumentOperation { allow_index_creation, .. }
             | BatchKind::ClearAndSettings { allow_index_creation, .. }
-            | BatchKind::SettingsAndDocumentOperation { allow_index_creation, .. }
             | BatchKind::Settings { allow_index_creation, .. } => Some(*allow_index_creation),
             _ => None,
         }
@@ -154,10 +146,7 @@ impl BatchKind {
 
     fn primary_key(&self) -> Option<Option<&str>> {
         match self {
-            BatchKind::DocumentOperation { primary_key, .. }
-            | BatchKind::SettingsAndDocumentOperation { primary_key, .. } => {
-                Some(primary_key.as_deref())
-            }
+            BatchKind::DocumentOperation { primary_key, .. } => Some(primary_key.as_deref()),
             _ => None,
         }
     }
@@ -275,8 +264,7 @@ impl BatchKind {
                 Break(BatchKind::IndexDeletion { ids })
             }
             (
-                BatchKind::ClearAndSettings { settings_ids: mut ids, allow_index_creation: _, mut other }
-                | BatchKind::SettingsAndDocumentOperation { operation_ids: mut ids, method: _, allow_index_creation: _, primary_key: _, settings_ids: mut other },
+                BatchKind::ClearAndSettings { settings_ids: mut ids, allow_index_creation: _, mut other },
                 K::IndexDeletion,
             ) => {
                 ids.push(id);
@@ -356,15 +344,9 @@ impl BatchKind {
             ) => Break(this),
 
             (
-                BatchKind::DocumentOperation { method, allow_index_creation, primary_key, operation_ids },
+                this @ BatchKind::DocumentOperation { .. },
                 K::Settings { .. },
-            ) => Continue(BatchKind::SettingsAndDocumentOperation {
-                settings_ids: vec![id],
-                method,
-                allow_index_creation,
-                primary_key,
-                operation_ids,
-            }),
+            ) => Break(this),
 
             (BatchKind::DocumentDeletion { mut deletion_ids, includes_by_filter: _ }, K::DocumentClear) => {
                 deletion_ids.push(id);
@@ -477,63 +459,7 @@ impl BatchKind {
                     allow_index_creation,
                 })
             }
-            (
-                BatchKind::SettingsAndDocumentOperation { settings_ids, method: _, mut operation_ids, allow_index_creation, primary_key: _ },
-                K::DocumentClear,
-            ) => {
-                operation_ids.push(id);
-                Continue(BatchKind::ClearAndSettings {
-                    settings_ids,
-                    other: operation_ids,
-                    allow_index_creation,
-                })
-            }
 
-            (
-                BatchKind::SettingsAndDocumentOperation { settings_ids, method: ReplaceDocuments, mut operation_ids, allow_index_creation, primary_key: _},
-                K::DocumentImport { method: ReplaceDocuments, primary_key: pk2, .. },
-            ) => {
-                operation_ids.push(id);
-                Continue(BatchKind::SettingsAndDocumentOperation {
-                    settings_ids,
-                    method: ReplaceDocuments,
-                    allow_index_creation,
-                        primary_key: pk2,
-                    operation_ids,
-                })
-            }
-            (
-                BatchKind::SettingsAndDocumentOperation { settings_ids, method: UpdateDocuments, allow_index_creation, primary_key: _, mut operation_ids },
-                K::DocumentImport { method: UpdateDocuments, primary_key: pk2, .. },
-            ) => {
-                operation_ids.push(id);
-                Continue(BatchKind::SettingsAndDocumentOperation {
-                    settings_ids,
-                    method: UpdateDocuments,
-                    allow_index_creation,
-                    primary_key: pk2,
-                    operation_ids,
-                })
-            }
-            // But we can't batch a settings and a doc op with another doc op
-            // this MUST be AFTER the two previous branch
-            (
-                this @ BatchKind::SettingsAndDocumentOperation { .. },
-                K::DocumentDeletion { .. } | K::DocumentImport { .. },
-            ) => Break(this),
-            (
-                BatchKind::SettingsAndDocumentOperation { mut settings_ids, method, allow_index_creation,primary_key, operation_ids },
-                K::Settings { .. },
-            ) => {
-                settings_ids.push(id);
-                Continue(BatchKind::SettingsAndDocumentOperation {
-                    settings_ids,
-                    method,
-                    allow_index_creation,
-                    primary_key,
-                    operation_ids,
-                })
-            }
             (
                 BatchKind::IndexCreation { .. }
                 | BatchKind::IndexDeletion { .. }
@@ -808,30 +734,30 @@ mod tests {
     }
 
     #[test]
-    fn document_addition_batch_with_settings() {
+    fn document_addition_doesnt_batch_with_settings() {
         // simple case
-        debug_snapshot!(autobatch_from(true, None, [doc_imp(ReplaceDocuments, true, None), settings(true)]), @"Some((SettingsAndDocumentOperation { settings_ids: [1], method: ReplaceDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0] }, true))");
-        debug_snapshot!(autobatch_from(true, None, [doc_imp(UpdateDocuments, true, None), settings(true)]), @"Some((SettingsAndDocumentOperation { settings_ids: [1], method: UpdateDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0] }, true))");
+        debug_snapshot!(autobatch_from(true, None, [doc_imp(ReplaceDocuments, true, None), settings(true)]), @"Some((DocumentOperation { method: ReplaceDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0] }, true))");
+        debug_snapshot!(autobatch_from(true, None, [doc_imp(UpdateDocuments, true, None), settings(true)]), @"Some((DocumentOperation { method: UpdateDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0] }, true))");
 
         // multiple settings and doc addition
-        debug_snapshot!(autobatch_from(true, None, [doc_imp(ReplaceDocuments, true, None), doc_imp(ReplaceDocuments, true, None), settings(true), settings(true)]), @"Some((SettingsAndDocumentOperation { settings_ids: [2, 3], method: ReplaceDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0, 1] }, true))");
-        debug_snapshot!(autobatch_from(true, None, [doc_imp(ReplaceDocuments, true, None), doc_imp(ReplaceDocuments, true, None), settings(true), settings(true)]), @"Some((SettingsAndDocumentOperation { settings_ids: [2, 3], method: ReplaceDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0, 1] }, true))");
+        debug_snapshot!(autobatch_from(true, None, [doc_imp(ReplaceDocuments, true, None), doc_imp(ReplaceDocuments, true, None), settings(true), settings(true)]), @"Some((DocumentOperation { method: ReplaceDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0, 1] }, true))");
+        debug_snapshot!(autobatch_from(true, None, [doc_imp(ReplaceDocuments, true, None), doc_imp(ReplaceDocuments, true, None), settings(true), settings(true)]), @"Some((DocumentOperation { method: ReplaceDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0, 1] }, true))");
 
         // addition and setting unordered
-        debug_snapshot!(autobatch_from(true, None, [doc_imp(ReplaceDocuments, true, None), settings(true), doc_imp(ReplaceDocuments, true, None), settings(true)]), @"Some((SettingsAndDocumentOperation { settings_ids: [1, 3], method: ReplaceDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0, 2] }, true))");
-        debug_snapshot!(autobatch_from(true, None, [doc_imp(UpdateDocuments, true, None), settings(true), doc_imp(UpdateDocuments, true, None), settings(true)]), @"Some((SettingsAndDocumentOperation { settings_ids: [1, 3], method: UpdateDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0, 2] }, true))");
+        debug_snapshot!(autobatch_from(true, None, [doc_imp(ReplaceDocuments, true, None), settings(true), doc_imp(ReplaceDocuments, true, None), settings(true)]), @"Some((DocumentOperation { method: ReplaceDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0] }, true))");
+        debug_snapshot!(autobatch_from(true, None, [doc_imp(UpdateDocuments, true, None), settings(true), doc_imp(UpdateDocuments, true, None), settings(true)]), @"Some((DocumentOperation { method: UpdateDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0] }, true))");
 
-        // We ensure this kind of batch doesn't batch with forbidden operations
-        debug_snapshot!(autobatch_from(true, None, [doc_imp(ReplaceDocuments, true, None), settings(true), doc_imp(UpdateDocuments, true, None)]), @"Some((SettingsAndDocumentOperation { settings_ids: [1], method: ReplaceDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0] }, true))");
-        debug_snapshot!(autobatch_from(true, None, [doc_imp(UpdateDocuments, true, None), settings(true), doc_imp(ReplaceDocuments, true, None)]), @"Some((SettingsAndDocumentOperation { settings_ids: [1], method: UpdateDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0] }, true))");
-        debug_snapshot!(autobatch_from(true, None, [doc_imp(ReplaceDocuments, true, None), settings(true), doc_del()]), @"Some((SettingsAndDocumentOperation { settings_ids: [1], method: ReplaceDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0] }, true))");
-        debug_snapshot!(autobatch_from(true, None, [doc_imp(UpdateDocuments, true, None), settings(true), doc_del()]), @"Some((SettingsAndDocumentOperation { settings_ids: [1], method: UpdateDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0] }, true))");
-        debug_snapshot!(autobatch_from(true, None, [doc_imp(ReplaceDocuments, true, None), settings(true), idx_create()]), @"Some((SettingsAndDocumentOperation { settings_ids: [1], method: ReplaceDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0] }, true))");
-        debug_snapshot!(autobatch_from(true, None, [doc_imp(UpdateDocuments, true, None), settings(true), idx_create()]), @"Some((SettingsAndDocumentOperation { settings_ids: [1], method: UpdateDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0] }, true))");
-        debug_snapshot!(autobatch_from(true, None, [doc_imp(ReplaceDocuments, true, None), settings(true), idx_update()]), @"Some((SettingsAndDocumentOperation { settings_ids: [1], method: ReplaceDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0] }, true))");
-        debug_snapshot!(autobatch_from(true, None, [doc_imp(UpdateDocuments, true, None), settings(true), idx_update()]), @"Some((SettingsAndDocumentOperation { settings_ids: [1], method: UpdateDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0] }, true))");
-        debug_snapshot!(autobatch_from(true, None, [doc_imp(ReplaceDocuments, true, None), settings(true), idx_swap()]), @"Some((SettingsAndDocumentOperation { settings_ids: [1], method: ReplaceDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0] }, true))");
-        debug_snapshot!(autobatch_from(true, None, [doc_imp(UpdateDocuments, true, None), settings(true), idx_swap()]), @"Some((SettingsAndDocumentOperation { settings_ids: [1], method: UpdateDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0] }, true))");
+        // Doesn't batch with other forbidden operations
+        debug_snapshot!(autobatch_from(true, None, [doc_imp(ReplaceDocuments, true, None), settings(true), doc_imp(UpdateDocuments, true, None)]), @"Some((DocumentOperation { method: ReplaceDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0] }, true))");
+        debug_snapshot!(autobatch_from(true, None, [doc_imp(UpdateDocuments, true, None), settings(true), doc_imp(ReplaceDocuments, true, None)]), @"Some((DocumentOperation { method: UpdateDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0] }, true))");
+        debug_snapshot!(autobatch_from(true, None, [doc_imp(ReplaceDocuments, true, None), settings(true), doc_del()]), @"Some((DocumentOperation { method: ReplaceDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0] }, true))");
+        debug_snapshot!(autobatch_from(true, None, [doc_imp(UpdateDocuments, true, None), settings(true), doc_del()]), @"Some((DocumentOperation { method: UpdateDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0] }, true))");
+        debug_snapshot!(autobatch_from(true, None, [doc_imp(ReplaceDocuments, true, None), settings(true), idx_create()]), @"Some((DocumentOperation { method: ReplaceDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0] }, true))");
+        debug_snapshot!(autobatch_from(true, None, [doc_imp(UpdateDocuments, true, None), settings(true), idx_create()]), @"Some((DocumentOperation { method: UpdateDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0] }, true))");
+        debug_snapshot!(autobatch_from(true, None, [doc_imp(ReplaceDocuments, true, None), settings(true), idx_update()]), @"Some((DocumentOperation { method: ReplaceDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0] }, true))");
+        debug_snapshot!(autobatch_from(true, None, [doc_imp(UpdateDocuments, true, None), settings(true), idx_update()]), @"Some((DocumentOperation { method: UpdateDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0] }, true))");
+        debug_snapshot!(autobatch_from(true, None, [doc_imp(ReplaceDocuments, true, None), settings(true), idx_swap()]), @"Some((DocumentOperation { method: ReplaceDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0] }, true))");
+        debug_snapshot!(autobatch_from(true, None, [doc_imp(UpdateDocuments, true, None), settings(true), idx_swap()]), @"Some((DocumentOperation { method: UpdateDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0] }, true))");
     }
 
     #[test]
@@ -859,8 +785,8 @@ mod tests {
         debug_snapshot!(autobatch_from(true, None, [doc_clr(), settings(true)]), @"Some((DocumentClear { ids: [0] }, false))");
 
         debug_snapshot!(autobatch_from(true, None, [settings(true), doc_clr(), settings(true)]), @"Some((ClearAndSettings { other: [1], allow_index_creation: true, settings_ids: [0, 2] }, true))");
-        debug_snapshot!(autobatch_from(true, None, [doc_imp(ReplaceDocuments, true, None), settings(true), doc_clr()]), @"Some((ClearAndSettings { other: [0, 2], allow_index_creation: true, settings_ids: [1] }, true))");
-        debug_snapshot!(autobatch_from(true, None, [doc_imp(UpdateDocuments, true, None), settings(true), doc_clr()]), @"Some((ClearAndSettings { other: [0, 2], allow_index_creation: true, settings_ids: [1] }, true))");
+        debug_snapshot!(autobatch_from(true, None, [doc_imp(ReplaceDocuments, true, None), settings(true), doc_clr()]), @"Some((DocumentOperation { method: ReplaceDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0] }, true))");
+        debug_snapshot!(autobatch_from(true, None, [doc_imp(UpdateDocuments, true, None), settings(true), doc_clr()]), @"Some((DocumentOperation { method: UpdateDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0] }, true))");
     }
 
     #[test]
@@ -907,50 +833,6 @@ mod tests {
         debug_snapshot!(autobatch_from(false,None,  [doc_clr(), idx_del()]), @"Some((IndexDeletion { ids: [0, 1] }, false))");
         debug_snapshot!(autobatch_from(false,None,  [settings(true), idx_del()]), @"Some((IndexDeletion { ids: [0, 1] }, true))");
         debug_snapshot!(autobatch_from(false,None,  [settings(false), idx_del()]), @"Some((IndexDeletion { ids: [0, 1] }, false))");
-
-        // Then the mixed cases.
-        // The index already exists, whatever is the right of the tasks it shouldn't change the result.
-        debug_snapshot!(autobatch_from(true, None, [doc_imp(ReplaceDocuments, true, None), settings(true), idx_del()]), @"Some((IndexDeletion { ids: [0, 2, 1] }, true))");
-        debug_snapshot!(autobatch_from(true, None, [doc_imp(UpdateDocuments, true, None), settings(true), idx_del()]), @"Some((IndexDeletion { ids: [0, 2, 1] }, true))");
-        debug_snapshot!(autobatch_from(true, None, [doc_imp(ReplaceDocuments, true, None), settings(true), doc_clr(), idx_del()]), @"Some((IndexDeletion { ids: [1, 3, 0, 2] }, true))");
-        debug_snapshot!(autobatch_from(true, None, [doc_imp(UpdateDocuments, true, None), settings(true), doc_clr(), idx_del()]), @"Some((IndexDeletion { ids: [1, 3, 0, 2] }, true))");
-        debug_snapshot!(autobatch_from(true, None, [doc_imp(ReplaceDocuments,false, None), settings(false), idx_del()]), @"Some((IndexDeletion { ids: [0, 2, 1] }, false))");
-        debug_snapshot!(autobatch_from(true, None, [doc_imp(UpdateDocuments, false, None), settings(false), idx_del()]), @"Some((IndexDeletion { ids: [0, 2, 1] }, false))");
-        debug_snapshot!(autobatch_from(true, None, [doc_imp(ReplaceDocuments,false, None), settings(false), doc_clr(), idx_del()]), @"Some((IndexDeletion { ids: [1, 3, 0, 2] }, false))");
-        debug_snapshot!(autobatch_from(true, None, [doc_imp(UpdateDocuments, false, None), settings(false), doc_clr(), idx_del()]), @"Some((IndexDeletion { ids: [1, 3, 0, 2] }, false))");
-        debug_snapshot!(autobatch_from(true, None, [doc_imp(ReplaceDocuments,false, None), settings(true), idx_del()]), @"Some((IndexDeletion { ids: [0, 2, 1] }, false))");
-        debug_snapshot!(autobatch_from(true, None, [doc_imp(UpdateDocuments, false, None), settings(true), idx_del()]), @"Some((IndexDeletion { ids: [0, 2, 1] }, false))");
-        debug_snapshot!(autobatch_from(true, None, [doc_imp(ReplaceDocuments,false, None), settings(true), doc_clr(), idx_del()]), @"Some((IndexDeletion { ids: [1, 3, 0, 2] }, false))");
-        debug_snapshot!(autobatch_from(true, None, [doc_imp(UpdateDocuments, false, None), settings(true), doc_clr(), idx_del()]), @"Some((IndexDeletion { ids: [1, 3, 0, 2] }, false))");
-        debug_snapshot!(autobatch_from(true, None, [doc_imp(ReplaceDocuments,true, None), settings(false), idx_del()]), @"Some((IndexDeletion { ids: [0, 2, 1] }, true))");
-        debug_snapshot!(autobatch_from(true, None, [doc_imp(UpdateDocuments, true, None), settings(false), idx_del()]), @"Some((IndexDeletion { ids: [0, 2, 1] }, true))");
-        debug_snapshot!(autobatch_from(true, None, [doc_imp(ReplaceDocuments,true, None), settings(false), doc_clr(), idx_del()]), @"Some((IndexDeletion { ids: [1, 3, 0, 2] }, true))");
-        debug_snapshot!(autobatch_from(true, None, [doc_imp(UpdateDocuments, true, None), settings(false), doc_clr(), idx_del()]), @"Some((IndexDeletion { ids: [1, 3, 0, 2] }, true))");
-
-        // When the index doesn't exists yet it's more complicated.
-        // Either the first task we encounter create it, in which case we can create a big batch with everything.
-        debug_snapshot!(autobatch_from(false,None,  [doc_imp(ReplaceDocuments, true, None), settings(true), idx_del()]), @"Some((IndexDeletion { ids: [0, 2, 1] }, true))");
-        debug_snapshot!(autobatch_from(false,None,  [doc_imp(UpdateDocuments, true, None), settings(true), idx_del()]), @"Some((IndexDeletion { ids: [0, 2, 1] }, true))");
-        debug_snapshot!(autobatch_from(false,None,  [doc_imp(ReplaceDocuments, true, None), settings(true), doc_clr(), idx_del()]), @"Some((IndexDeletion { ids: [1, 3, 0, 2] }, true))");
-        debug_snapshot!(autobatch_from(false,None,  [doc_imp(UpdateDocuments, true, None), settings(true), doc_clr(), idx_del()]), @"Some((IndexDeletion { ids: [1, 3, 0, 2] }, true))");
-        // The right of the tasks following isn't really important.
-        debug_snapshot!(autobatch_from(false,None,  [doc_imp(ReplaceDocuments,true, None), settings(false), idx_del()]), @"Some((IndexDeletion { ids: [0, 2, 1] }, true))");
-        debug_snapshot!(autobatch_from(false,None,  [doc_imp(UpdateDocuments, true, None), settings(false), idx_del()]), @"Some((IndexDeletion { ids: [0, 2, 1] }, true))");
-        debug_snapshot!(autobatch_from(false,None,  [doc_imp(ReplaceDocuments,true, None), settings(false), doc_clr(), idx_del()]), @"Some((IndexDeletion { ids: [1, 3, 0, 2] }, true))");
-        debug_snapshot!(autobatch_from(false,None,  [doc_imp(UpdateDocuments, true, None), settings(false), doc_clr(), idx_del()]), @"Some((IndexDeletion { ids: [1, 3, 0, 2] }, true))");
-        // Or, the second case; the first task doesn't create the index and thus we wants to batch it with only tasks that can't create an index.
-        // that can be a second task that don't have the right to create an index. Or anything that can't create an index like an index deletion, document deletion, document clear, etc.
-        // All theses tasks are going to throw an error `Index doesn't exist` once the batch is processed.
-        debug_snapshot!(autobatch_from(false,None,  [doc_imp(ReplaceDocuments,false, None), settings(false), idx_del()]), @"Some((IndexDeletion { ids: [0, 2, 1] }, false))");
-        debug_snapshot!(autobatch_from(false,None,  [doc_imp(UpdateDocuments, false, None), settings(false), idx_del()]), @"Some((IndexDeletion { ids: [0, 2, 1] }, false))");
-        debug_snapshot!(autobatch_from(false,None,  [doc_imp(ReplaceDocuments,false, None), settings(false), doc_clr(), idx_del()]), @"Some((IndexDeletion { ids: [1, 3, 0, 2] }, false))");
-        debug_snapshot!(autobatch_from(false,None,  [doc_imp(UpdateDocuments, false, None), settings(false), doc_clr(), idx_del()]), @"Some((IndexDeletion { ids: [1, 3, 0, 2] }, false))");
-        // The third and final case is when the first task doesn't create an index but is directly followed by a task creating an index. In this case we can't batch whit what
-        // follows because we first need to process the erronous batch.
-        debug_snapshot!(autobatch_from(false,None,  [doc_imp(ReplaceDocuments,false, None), settings(true), idx_del()]), @"Some((DocumentOperation { method: ReplaceDocuments, allow_index_creation: false, primary_key: None, operation_ids: [0] }, false))");
-        debug_snapshot!(autobatch_from(false,None,  [doc_imp(UpdateDocuments, false, None), settings(true), idx_del()]), @"Some((DocumentOperation { method: UpdateDocuments, allow_index_creation: false, primary_key: None, operation_ids: [0] }, false))");
-        debug_snapshot!(autobatch_from(false,None,  [doc_imp(ReplaceDocuments,false, None), settings(true), doc_clr(), idx_del()]), @"Some((DocumentOperation { method: ReplaceDocuments, allow_index_creation: false, primary_key: None, operation_ids: [0] }, false))");
-        debug_snapshot!(autobatch_from(false,None,  [doc_imp(UpdateDocuments, false, None), settings(true), doc_clr(), idx_del()]), @"Some((DocumentOperation { method: UpdateDocuments, allow_index_creation: false, primary_key: None, operation_ids: [0] }, false))");
     }
 
     #[test]
@@ -959,13 +841,13 @@ mod tests {
         debug_snapshot!(autobatch_from(true, None, [doc_imp(ReplaceDocuments, false, None), doc_imp(ReplaceDocuments, true, None)]), @"Some((DocumentOperation { method: ReplaceDocuments, allow_index_creation: false, primary_key: None, operation_ids: [0, 1] }, false))");
         debug_snapshot!(autobatch_from(true, None, [doc_imp(ReplaceDocuments, true, None), doc_imp(ReplaceDocuments, true, None)]), @"Some((DocumentOperation { method: ReplaceDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0, 1] }, true))");
         debug_snapshot!(autobatch_from(true, None, [doc_imp(ReplaceDocuments, false, None), doc_imp(ReplaceDocuments, false, None)]), @"Some((DocumentOperation { method: ReplaceDocuments, allow_index_creation: false, primary_key: None, operation_ids: [0, 1] }, false))");
-        debug_snapshot!(autobatch_from(true, None, [doc_imp(ReplaceDocuments, true, None), settings(true)]), @"Some((SettingsAndDocumentOperation { settings_ids: [1], method: ReplaceDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0] }, true))");
-        debug_snapshot!(autobatch_from(true, None, [doc_imp(ReplaceDocuments, false, None), settings(true)]), @"Some((SettingsAndDocumentOperation { settings_ids: [1], method: ReplaceDocuments, allow_index_creation: false, primary_key: None, operation_ids: [0] }, false))");
+        debug_snapshot!(autobatch_from(true, None, [doc_imp(ReplaceDocuments, true, None), settings(true)]), @"Some((DocumentOperation { method: ReplaceDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0] }, true))");
+        debug_snapshot!(autobatch_from(true, None, [doc_imp(ReplaceDocuments, false, None), settings(true)]), @"Some((DocumentOperation { method: ReplaceDocuments, allow_index_creation: false, primary_key: None, operation_ids: [0] }, false))");
 
         debug_snapshot!(autobatch_from(false,None,  [doc_imp(ReplaceDocuments, false, None), doc_imp(ReplaceDocuments, true, None)]), @"Some((DocumentOperation { method: ReplaceDocuments, allow_index_creation: false, primary_key: None, operation_ids: [0] }, false))");
         debug_snapshot!(autobatch_from(false,None,  [doc_imp(ReplaceDocuments, true, None), doc_imp(ReplaceDocuments, true, None)]), @"Some((DocumentOperation { method: ReplaceDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0, 1] }, true))");
         debug_snapshot!(autobatch_from(false,None,  [doc_imp(ReplaceDocuments, false, None), doc_imp(ReplaceDocuments, false, None)]), @"Some((DocumentOperation { method: ReplaceDocuments, allow_index_creation: false, primary_key: None, operation_ids: [0, 1] }, false))");
-        debug_snapshot!(autobatch_from(false,None,  [doc_imp(ReplaceDocuments, true, None), settings(true)]), @"Some((SettingsAndDocumentOperation { settings_ids: [1], method: ReplaceDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0] }, true))");
+        debug_snapshot!(autobatch_from(false,None,  [doc_imp(ReplaceDocuments, true, None), settings(true)]), @"Some((DocumentOperation { method: ReplaceDocuments, allow_index_creation: true, primary_key: None, operation_ids: [0] }, true))");
         debug_snapshot!(autobatch_from(false,None,  [doc_imp(ReplaceDocuments, false, None), settings(true)]), @"Some((DocumentOperation { method: ReplaceDocuments, allow_index_creation: false, primary_key: None, operation_ids: [0] }, false))");
 
         // batch deletion and addition

crates/index-scheduler/src/batch.rs

@@ -104,7 +104,6 @@ pub(crate) enum IndexOperation {
         index_uid: String,
         primary_key: Option<String>,
         method: IndexDocumentsMethod,
-        documents_counts: Vec<u64>,
         operations: Vec<DocumentOperation>,
         tasks: Vec<Task>,
     },
@@ -130,19 +129,6 @@ pub(crate) enum IndexOperation {
         index_uid: String,
         cleared_tasks: Vec<Task>,
 
-        // The boolean indicates if it's a settings deletion or creation.
-        settings: Vec<(bool, Settings<Unchecked>)>,
-        settings_tasks: Vec<Task>,
-    },
-    SettingsAndDocumentOperation {
-        index_uid: String,
-
-        primary_key: Option<String>,
-        method: IndexDocumentsMethod,
-        documents_counts: Vec<u64>,
-        operations: Vec<DocumentOperation>,
-        document_import_tasks: Vec<Task>,
-
         // The boolean indicates if it's a settings deletion or creation.
         settings: Vec<(bool, Settings<Unchecked>)>,
         settings_tasks: Vec<Task>,
@@ -174,12 +160,7 @@ impl Batch {
                 IndexOperation::DocumentEdition { task, .. } => {
                     RoaringBitmap::from_sorted_iter(std::iter::once(task.uid)).unwrap()
                 }
-                IndexOperation::SettingsAndDocumentOperation {
-                    document_import_tasks: tasks,
-                    settings_tasks: other,
-                    ..
-                }
-                | IndexOperation::DocumentClearAndSetting {
+                IndexOperation::DocumentClearAndSetting {
                     cleared_tasks: tasks,
                     settings_tasks: other,
                     ..
@@ -239,8 +220,7 @@ impl IndexOperation {
             | IndexOperation::DocumentDeletion { index_uid, .. }
             | IndexOperation::DocumentClear { index_uid, .. }
             | IndexOperation::Settings { index_uid, .. }
-            | IndexOperation::DocumentClearAndSetting { index_uid, .. }
-            | IndexOperation::SettingsAndDocumentOperation { index_uid, .. } => index_uid,
+            | IndexOperation::DocumentClearAndSetting { index_uid, .. } => index_uid,
         }
     }
 }
@@ -262,9 +242,6 @@ impl fmt::Display for IndexOperation {
             IndexOperation::DocumentClearAndSetting { .. } => {
                 f.write_str("IndexOperation::DocumentClearAndSetting")
             }
-            IndexOperation::SettingsAndDocumentOperation { .. } => {
-                f.write_str("IndexOperation::SettingsAndDocumentOperation")
-            }
         }
     }
 }
@@ -330,21 +307,14 @@ impl IndexScheduler {
                     })
                     .flatten();
 
-                let mut documents_counts = Vec::new();
                 let mut operations = Vec::new();
 
                 for task in tasks.iter() {
                     match task.kind {
-                        KindWithContent::DocumentAdditionOrUpdate {
-                            content_file,
-                            documents_count,
-                            ..
-                        } => {
-                            documents_counts.push(documents_count);
+                        KindWithContent::DocumentAdditionOrUpdate { content_file, .. } => {
                             operations.push(DocumentOperation::Add(content_file));
                         }
                         KindWithContent::DocumentDeletion { ref documents_ids, .. } => {
-                            documents_counts.push(documents_ids.len() as u64);
                             operations.push(DocumentOperation::Delete(documents_ids.clone()));
                         }
                         _ => unreachable!(),
@@ -356,7 +326,6 @@ impl IndexScheduler {
                         index_uid,
                         primary_key,
                         method,
-                        documents_counts,
                         operations,
                         tasks,
                     },
@@ -441,67 +410,6 @@ impl IndexScheduler {
                     must_create_index,
                 }))
             }
-            BatchKind::SettingsAndDocumentOperation {
-                settings_ids,
-                method,
-                allow_index_creation,
-                primary_key,
-                operation_ids,
-            } => {
-                let settings = self.create_next_batch_index(
-                    rtxn,
-                    index_uid.clone(),
-                    BatchKind::Settings { settings_ids, allow_index_creation },
-                    current_batch,
-                    must_create_index,
-                )?;
-
-                let document_import = self.create_next_batch_index(
-                    rtxn,
-                    index_uid.clone(),
-                    BatchKind::DocumentOperation {
-                        method,
-                        allow_index_creation,
-                        primary_key,
-                        operation_ids,
-                    },
-                    current_batch,
-                    must_create_index,
-                )?;
-
-                match (document_import, settings) {
-                    (
-                        Some(Batch::IndexOperation {
-                            op:
-                                IndexOperation::DocumentOperation {
-                                    primary_key,
-                                    documents_counts,
-                                    operations,
-                                    tasks: document_import_tasks,
-                                    ..
-                                },
-                            ..
-                        }),
-                        Some(Batch::IndexOperation {
-                            op: IndexOperation::Settings { settings, tasks: settings_tasks, .. },
-                            ..
-                        }),
-                    ) => Ok(Some(Batch::IndexOperation {
-                        op: IndexOperation::SettingsAndDocumentOperation {
-                            index_uid,
-                            primary_key,
-                            method,
-                            documents_counts,
-                            operations,
-                            document_import_tasks,
-                            settings,
-                            settings_tasks,
-                        },
-                        must_create_index,
-                    })),
-                    _ => unreachable!(),
-                }
-            }
             BatchKind::IndexCreation { id } => {
                 let mut task = self.get_task(rtxn, id)?.ok_or(Error::CorruptedTaskQueue)?;
                 current_batch.processing(Some(&mut task));
@@ -589,7 +497,6 @@ impl IndexScheduler {
         // 5. We make a batch from the unprioritised tasks. Start by taking the next enqueued task.
         let task_id = if let Some(task_id) = enqueued.min() { task_id } else { return Ok(None) };
         let mut task = self.get_task(rtxn, task_id)?.ok_or(Error::CorruptedTaskQueue)?;
-        current_batch.processing(Some(&mut task));
 
         // If the task is not associated with any index, verify that it is an index swap and
         // create the batch directly. Otherwise, get the index name associated with the task
@@ -599,6 +506,7 @@ impl IndexScheduler {
             index_name
         } else {
             assert!(matches!(&task.kind, KindWithContent::IndexSwap { swaps } if swaps.is_empty()));
+            current_batch.processing(Some(&mut task));
             return Ok(Some((Batch::IndexSwap { task }, current_batch)));
         };
 
@@ -1304,7 +1212,6 @@ impl IndexScheduler {
                 index_uid: _,
                 primary_key,
                 method,
-                documents_counts: _,
                 operations,
                 mut tasks,
             } => {
@@ -1351,7 +1258,10 @@ impl IndexScheduler {
                 let pool = match &indexer_config.thread_pool {
                     Some(pool) => pool,
                     None => {
-                        local_pool = ThreadPoolNoAbortBuilder::new().build().unwrap();
+                        local_pool = ThreadPoolNoAbortBuilder::new()
+                            .thread_name(|i| format!("indexing-thread-{i}"))
+                            .build()
+                            .unwrap();
                         &local_pool
                     }
                 };
@@ -1399,21 +1309,19 @@ impl IndexScheduler {
                 }
 
                 if tasks.iter().any(|res| res.error.is_none()) {
-                    pool.install(|| {
-                        indexer::index(
-                            index_wtxn,
-                            index,
-                            indexer_config.grenad_parameters(),
-                            &db_fields_ids_map,
-                            new_fields_ids_map,
-                            primary_key,
-                            &document_changes,
-                            embedders,
-                            &|| must_stop_processing.get(),
-                            &send_progress,
-                        )
-                    })
-                    .unwrap()?;
+                    indexer::index(
+                        index_wtxn,
+                        index,
+                        pool,
+                        indexer_config.grenad_parameters(),
+                        &db_fields_ids_map,
+                        new_fields_ids_map,
+                        primary_key,
+                        &document_changes,
+                        embedders,
+                        &|| must_stop_processing.get(),
+                        &send_progress,
+                    )?;
 
                     tracing::info!(indexing_result = ?addition, processed_in = ?started_processing_at.elapsed(), "document indexing done");
                 }
@@ -1489,34 +1397,34 @@ impl IndexScheduler {
                     let pool = match &indexer_config.thread_pool {
                         Some(pool) => pool,
                         None => {
-                            local_pool = ThreadPoolNoAbortBuilder::new().build().unwrap();
+                            local_pool = ThreadPoolNoAbortBuilder::new()
+                                .thread_name(|i| format!("indexing-thread-{i}"))
+                                .build()
+                                .unwrap();
                             &local_pool
                         }
                     };
 
-                    pool.install(|| {
-                        let indexer =
-                            UpdateByFunction::new(candidates, context.clone(), code.clone());
-                        let document_changes = indexer.into_changes(&primary_key)?;
-                        let embedders = index.embedding_configs(index_wtxn)?;
-                        let embedders = self.embedders(embedders)?;
+                    let indexer = UpdateByFunction::new(candidates, context.clone(), code.clone());
+                    let document_changes =
+                        pool.install(|| indexer.into_changes(&primary_key)).unwrap()?;
 
-                        indexer::index(
-                            index_wtxn,
-                            index,
-                            indexer_config.grenad_parameters(),
-                            &db_fields_ids_map,
-                            new_fields_ids_map,
-                            None, // cannot change primary key in DocumentEdition
-                            &document_changes,
-                            embedders,
-                            &|| must_stop_processing.get(),
-                            &send_progress,
-                        )?;
+                    let embedders = index.embedding_configs(index_wtxn)?;
+                    let embedders = self.embedders(embedders)?;
 
-                        Result::Ok(())
-                    })
-                    .unwrap()?;
+                    indexer::index(
+                        index_wtxn,
+                        index,
+                        pool,
+                        indexer_config.grenad_parameters(),
+                        &db_fields_ids_map,
+                        new_fields_ids_map,
+                        None, // cannot change primary key in DocumentEdition
+                        &document_changes,
+                        embedders,
+                        &|| must_stop_processing.get(),
+                        &send_progress,
+                    )?;
 
                     // tracing::info!(indexing_result = ?addition, processed_in = ?started_processing_at.elapsed(), "document indexing done");
                 }
@@ -1641,7 +1549,10 @@ impl IndexScheduler {
                     let pool = match &indexer_config.thread_pool {
                         Some(pool) => pool,
                         None => {
-                            local_pool = ThreadPoolNoAbortBuilder::new().build().unwrap();
+                            local_pool = ThreadPoolNoAbortBuilder::new()
+                                .thread_name(|i| format!("indexing-thread-{i}"))
+                                .build()
+                                .unwrap();
                             &local_pool
                         }
                     };
@@ -1652,21 +1563,19 @@ impl IndexScheduler {
                     let embedders = index.embedding_configs(index_wtxn)?;
                     let embedders = self.embedders(embedders)?;
 
-                    pool.install(|| {
-                        indexer::index(
-                            index_wtxn,
-                            index,
-                            indexer_config.grenad_parameters(),
-                            &db_fields_ids_map,
-                            new_fields_ids_map,
-                            None, // document deletion never changes primary key
-                            &document_changes,
-                            embedders,
-                            &|| must_stop_processing.get(),
-                            &send_progress,
-                        )
-                    })
-                    .unwrap()?;
+                    indexer::index(
+                        index_wtxn,
+                        index,
+                        pool,
+                        indexer_config.grenad_parameters(),
+                        &db_fields_ids_map,
+                        new_fields_ids_map,
+                        None, // document deletion never changes primary key
+                        &document_changes,
+                        embedders,
+                        &|| must_stop_processing.get(),
+                        &send_progress,
+                    )?;
 
                     // tracing::info!(indexing_result = ?addition, processed_in = ?started_processing_at.elapsed(), "document indexing done");
                 }
@@ -1694,43 +1603,6 @@ impl IndexScheduler {
 
                 Ok(tasks)
             }
-            IndexOperation::SettingsAndDocumentOperation {
-                index_uid,
-                primary_key,
-                method,
-                documents_counts,
-                operations,
-                document_import_tasks,
-                settings,
-                settings_tasks,
-            } => {
-                let settings_tasks = self.apply_index_operation(
-                    index_wtxn,
-                    index,
-                    IndexOperation::Settings {
-                        index_uid: index_uid.clone(),
-                        settings,
-                        tasks: settings_tasks,
-                    },
-                )?;
-
-                let mut import_tasks = self.apply_index_operation(
-                    index_wtxn,
-                    index,
-                    IndexOperation::DocumentOperation {
-                        index_uid,
-                        primary_key,
-                        method,
-                        documents_counts,
-                        operations,
-                        tasks: document_import_tasks,
-                    },
-                )?;
-
-                let mut tasks = settings_tasks;
-                tasks.append(&mut import_tasks);
-                Ok(tasks)
-            }
             IndexOperation::DocumentClearAndSetting {
                 index_uid,
                 cleared_tasks,

crates/index-scheduler/src/lib.rs

@@ -407,7 +407,7 @@ pub struct IndexScheduler {
     ///
     /// See [self.breakpoint()](`IndexScheduler::breakpoint`) for an explanation.
     #[cfg(test)]
-    test_breakpoint_sdr: crossbeam::channel::Sender<(Breakpoint, bool)>,
+    test_breakpoint_sdr: crossbeam_channel::Sender<(Breakpoint, bool)>,
 
     /// A list of planned failures within the [`tick`](IndexScheduler::tick) method of the index scheduler.
     ///
@@ -476,7 +476,7 @@ impl IndexScheduler {
     /// Create an index scheduler and start its run loop.
     pub fn new(
         options: IndexSchedulerOptions,
-        #[cfg(test)] test_breakpoint_sdr: crossbeam::channel::Sender<(Breakpoint, bool)>,
+        #[cfg(test)] test_breakpoint_sdr: crossbeam_channel::Sender<(Breakpoint, bool)>,
         #[cfg(test)] planned_failures: Vec<(usize, tests::FailureLocation)>,
     ) -> Result<Self> {
         std::fs::create_dir_all(&options.tasks_path)?;
@@ -1440,7 +1440,7 @@ 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.info().map_size as u64 > 50
+            && (self.env.non_free_pages_size()? * 100) / self.env.info().map_size as u64 > 40
         {
             return Err(Error::NoSpaceLeftInTaskQueue);
         }
@@ -1738,11 +1738,8 @@ impl IndexScheduler {
             }
         }
 
-        self.processing_tasks.write().unwrap().stop_processing();
         // We must re-add the canceled task so they're part of the same batch.
-        // processed.processing |= canceled;
         ids |= canceled;
-
         self.write_batch(&mut wtxn, processing_batch, &ids)?;
 
         #[cfg(test)]
@@ -1750,6 +1747,10 @@ impl IndexScheduler {
 
         wtxn.commit().map_err(Error::HeedTransaction)?;
 
+        // We should stop processing AFTER everything is processed and written to disk otherwise, a batch (which only lives in RAM) may appear in the processing task
+        // and then become « not found » for some time until the commit everything is written and the final commit is made.
+        self.processing_tasks.write().unwrap().stop_processing();
+
         // Once the tasks are committed, we should delete all the update files associated ASAP to avoid leaking files in case of a restart
         tracing::debug!("Deleting the update files");
 
@@ -2237,7 +2238,7 @@ mod tests {
     use std::time::Instant;
 
     use big_s::S;
-    use crossbeam::channel::RecvTimeoutError;
+    use crossbeam_channel::RecvTimeoutError;
     use file_store::File;
     use insta::assert_json_snapshot;
     use maplit::btreeset;
@@ -2289,7 +2290,7 @@ mod tests {
             configuration: impl Fn(&mut IndexSchedulerOptions),
         ) -> (Self, IndexSchedulerHandle) {
             let tempdir = TempDir::new().unwrap();
-            let (sender, receiver) = crossbeam::channel::bounded(0);
+            let (sender, receiver) = crossbeam_channel::bounded(0);
 
             let indexer_config = IndexerConfig { skip_index_budget: true, ..Default::default() };
 
@@ -2421,7 +2422,7 @@ mod tests {
     pub struct IndexSchedulerHandle {
         _tempdir: TempDir,
         index_scheduler: IndexScheduler,
-        test_breakpoint_rcv: crossbeam::channel::Receiver<(Breakpoint, bool)>,
+        test_breakpoint_rcv: crossbeam_channel::Receiver<(Breakpoint, bool)>,
         last_breakpoint: Breakpoint,
     }
 
@@ -4318,10 +4319,35 @@ mod tests {
         let proc = index_scheduler.processing_tasks.read().unwrap().clone();
 
         let query = Query { statuses: Some(vec![Status::Processing]), ..Default::default() };
-        let (batches, _) = index_scheduler
-            .get_batch_ids_from_authorized_indexes(&rtxn, &proc, &query, &AuthFilter::default())
+        let (mut batches, _) = index_scheduler
+            .get_batches_from_authorized_indexes(query.clone(), &AuthFilter::default())
             .unwrap();
-        snapshot!(snapshot_bitmap(&batches), @"[0,]"); // only the processing batch in the first tick
+        assert_eq!(batches.len(), 1);
+        batches[0].started_at = OffsetDateTime::UNIX_EPOCH;
+        // Insta cannot snapshot our batches because the batch stats contains an enum as key: https://github.com/mitsuhiko/insta/issues/689
+        let batch = serde_json::to_string_pretty(&batches[0]).unwrap();
+        snapshot!(batch, @r#"
+        {
+          "uid": 0,
+          "details": {
+            "primaryKey": "mouse"
+          },
+          "stats": {
+            "totalNbTasks": 1,
+            "status": {
+              "processing": 1
+            },
+            "types": {
+              "indexCreation": 1
+            },
+            "indexUids": {
+              "catto": 1
+            }
+          },
+          "startedAt": "1970-01-01T00:00:00Z",
+          "finishedAt": null
+        }
+        "#);
 
         let query = Query { statuses: Some(vec![Status::Enqueued]), ..Default::default() };
         let (batches, _) = index_scheduler

crates/index-scheduler/src/snapshots/lib.rs/cancel_mix_of_tasks/aborted_indexation.snap

@@ -5,7 +5,7 @@ snapshot_kind: text
 ### Autobatching Enabled = true
 ### Processing batch Some(1):
 [1,]
-{uid: 1, details: {"receivedDocuments":2,"indexedDocuments":null}, stats: {"totalNbTasks":2,"status":{"enqueued":2},"types":{"documentAdditionOrUpdate":2},"indexUids":{"beavero":2}}, }
+{uid: 1, details: {"receivedDocuments":1,"indexedDocuments":null}, stats: {"totalNbTasks":1,"status":{"processing":1},"types":{"documentAdditionOrUpdate":1},"indexUids":{"beavero":1}}, }
 ----------------------------------------------------------------------
 ### All Tasks:
 0 {uid: 0, batch_uid: 0, status: succeeded, details: { received_documents: 1, indexed_documents: Some(1) }, kind: DocumentAdditionOrUpdate { index_uid: "catto", primary_key: None, method: ReplaceDocuments, content_file: 00000000-0000-0000-0000-000000000000, documents_count: 1, allow_index_creation: true }}

crates/index-scheduler/src/snapshots/lib.rs/cancel_mix_of_tasks/processing_second_task_cancel_enqueued.snap

@@ -5,7 +5,7 @@ snapshot_kind: text
 ### Autobatching Enabled = true
 ### Processing batch Some(1):
 [1,]
-{uid: 1, details: {"receivedDocuments":2,"indexedDocuments":null}, stats: {"totalNbTasks":2,"status":{"enqueued":2},"types":{"documentAdditionOrUpdate":2},"indexUids":{"beavero":2}}, }
+{uid: 1, details: {"receivedDocuments":1,"indexedDocuments":null}, stats: {"totalNbTasks":1,"status":{"processing":1},"types":{"documentAdditionOrUpdate":1},"indexUids":{"beavero":1}}, }
 ----------------------------------------------------------------------
 ### All Tasks:
 0 {uid: 0, batch_uid: 0, status: succeeded, details: { received_documents: 1, indexed_documents: Some(1) }, kind: DocumentAdditionOrUpdate { index_uid: "catto", primary_key: None, method: ReplaceDocuments, content_file: 00000000-0000-0000-0000-000000000000, documents_count: 1, allow_index_creation: true }}

crates/index-scheduler/src/snapshots/lib.rs/cancel_processing_dump/cancel_registered.snap

@@ -5,7 +5,7 @@ snapshot_kind: text
 ### Autobatching Enabled = true
 ### Processing batch Some(0):
 [0,]
-{uid: 0, details: {"dumpUid":null}, stats: {"totalNbTasks":1,"status":{"enqueued":1},"types":{"dumpCreation":1},"indexUids":{}}, }
+{uid: 0, details: {"dumpUid":null}, stats: {"totalNbTasks":1,"status":{"processing":1},"types":{"dumpCreation":1},"indexUids":{}}, }
 ----------------------------------------------------------------------
 ### All Tasks:
 0 {uid: 0, status: enqueued, details: { dump_uid: None }, kind: DumpCreation { keys: [], instance_uid: None }}

crates/index-scheduler/src/snapshots/lib.rs/cancel_processing_task/aborted_indexation.snap

@@ -5,7 +5,7 @@ snapshot_kind: text
 ### Autobatching Enabled = true
 ### Processing batch Some(0):
 [0,]
-{uid: 0, details: {"receivedDocuments":2,"indexedDocuments":null}, stats: {"totalNbTasks":2,"status":{"enqueued":2},"types":{"documentAdditionOrUpdate":2},"indexUids":{"catto":2}}, }
+{uid: 0, details: {"receivedDocuments":1,"indexedDocuments":null}, stats: {"totalNbTasks":1,"status":{"processing":1},"types":{"documentAdditionOrUpdate":1},"indexUids":{"catto":1}}, }
 ----------------------------------------------------------------------
 ### All Tasks:
 0 {uid: 0, status: enqueued, details: { received_documents: 1, indexed_documents: None }, kind: DocumentAdditionOrUpdate { index_uid: "catto", primary_key: None, method: ReplaceDocuments, content_file: 00000000-0000-0000-0000-000000000000, documents_count: 1, allow_index_creation: true }}

crates/index-scheduler/src/snapshots/lib.rs/cancel_processing_task/cancel_task_registered.snap

@@ -5,7 +5,7 @@ snapshot_kind: text
 ### Autobatching Enabled = true
 ### Processing batch Some(0):
 [0,]
-{uid: 0, details: {"receivedDocuments":2,"indexedDocuments":null}, stats: {"totalNbTasks":2,"status":{"enqueued":2},"types":{"documentAdditionOrUpdate":2},"indexUids":{"catto":2}}, }
+{uid: 0, details: {"receivedDocuments":1,"indexedDocuments":null}, stats: {"totalNbTasks":1,"status":{"processing":1},"types":{"documentAdditionOrUpdate":1},"indexUids":{"catto":1}}, }
 ----------------------------------------------------------------------
 ### All Tasks:
 0 {uid: 0, status: enqueued, details: { received_documents: 1, indexed_documents: None }, kind: DocumentAdditionOrUpdate { index_uid: "catto", primary_key: None, method: ReplaceDocuments, content_file: 00000000-0000-0000-0000-000000000000, documents_count: 1, allow_index_creation: true }}

crates/index-scheduler/src/snapshots/lib.rs/cancel_processing_task/initial_task_processing.snap

@@ -5,7 +5,7 @@ snapshot_kind: text
 ### Autobatching Enabled = true
 ### Processing batch Some(0):
 [0,]
-{uid: 0, details: {"receivedDocuments":2,"indexedDocuments":null}, stats: {"totalNbTasks":2,"status":{"enqueued":2},"types":{"documentAdditionOrUpdate":2},"indexUids":{"catto":2}}, }
+{uid: 0, details: {"receivedDocuments":1,"indexedDocuments":null}, stats: {"totalNbTasks":1,"status":{"processing":1},"types":{"documentAdditionOrUpdate":1},"indexUids":{"catto":1}}, }
 ----------------------------------------------------------------------
 ### All Tasks:
 0 {uid: 0, status: enqueued, details: { received_documents: 1, indexed_documents: None }, kind: DocumentAdditionOrUpdate { index_uid: "catto", primary_key: None, method: ReplaceDocuments, content_file: 00000000-0000-0000-0000-000000000000, documents_count: 1, allow_index_creation: true }}

crates/index-scheduler/src/snapshots/lib.rs/document_addition/after_the_batch_creation.snap

@@ -5,7 +5,7 @@ snapshot_kind: text
 ### Autobatching Enabled = true
 ### Processing batch Some(0):
 [0,]
-{uid: 0, details: {"receivedDocuments":2,"indexedDocuments":null}, stats: {"totalNbTasks":2,"status":{"enqueued":2},"types":{"documentAdditionOrUpdate":2},"indexUids":{"doggos":2}}, }
+{uid: 0, details: {"receivedDocuments":1,"indexedDocuments":null}, stats: {"totalNbTasks":1,"status":{"processing":1},"types":{"documentAdditionOrUpdate":1},"indexUids":{"doggos":1}}, }
 ----------------------------------------------------------------------
 ### All Tasks:
 0 {uid: 0, status: enqueued, details: { received_documents: 1, indexed_documents: None }, kind: DocumentAdditionOrUpdate { index_uid: "doggos", primary_key: Some("id"), method: ReplaceDocuments, content_file: 00000000-0000-0000-0000-000000000000, documents_count: 1, allow_index_creation: true }}

crates/index-scheduler/src/snapshots/lib.rs/fail_in_process_batch_for_document_addition/document_addition_batch_created.snap

@@ -5,7 +5,7 @@ snapshot_kind: text
 ### Autobatching Enabled = true
 ### Processing batch Some(0):
 [0,]
-{uid: 0, details: {"receivedDocuments":2,"indexedDocuments":null}, stats: {"totalNbTasks":2,"status":{"enqueued":2},"types":{"documentAdditionOrUpdate":2},"indexUids":{"doggos":2}}, }
+{uid: 0, details: {"receivedDocuments":1,"indexedDocuments":null}, stats: {"totalNbTasks":1,"status":{"processing":1},"types":{"documentAdditionOrUpdate":1},"indexUids":{"doggos":1}}, }
 ----------------------------------------------------------------------
 ### All Tasks:
 0 {uid: 0, status: enqueued, details: { received_documents: 1, indexed_documents: None }, kind: DocumentAdditionOrUpdate { index_uid: "doggos", primary_key: Some("id"), method: ReplaceDocuments, content_file: 00000000-0000-0000-0000-000000000000, documents_count: 1, allow_index_creation: true }}

crates/index-scheduler/src/snapshots/lib.rs/fail_in_update_task_after_process_batch_success_for_document_addition/after_batch_succeeded.snap

@@ -5,7 +5,7 @@ snapshot_kind: text
 ### Autobatching Enabled = true
 ### Processing batch Some(0):
 [0,]
-{uid: 0, details: {"receivedDocuments":2,"indexedDocuments":null}, stats: {"totalNbTasks":2,"status":{"enqueued":2},"types":{"documentAdditionOrUpdate":2},"indexUids":{"doggos":2}}, }
+{uid: 0, details: {"receivedDocuments":1,"indexedDocuments":null}, stats: {"totalNbTasks":1,"status":{"processing":1},"types":{"documentAdditionOrUpdate":1},"indexUids":{"doggos":1}}, }
 ----------------------------------------------------------------------
 ### All Tasks:
 0 {uid: 0, status: enqueued, details: { received_documents: 1, indexed_documents: None }, kind: DocumentAdditionOrUpdate { index_uid: "doggos", primary_key: Some("id"), method: ReplaceDocuments, content_file: 00000000-0000-0000-0000-000000000000, documents_count: 1, allow_index_creation: true }}

crates/index-scheduler/src/snapshots/lib.rs/fail_in_update_task_after_process_batch_success_for_document_addition/after_failing_to_commit.snap

@@ -5,7 +5,7 @@ snapshot_kind: text
 ### Autobatching Enabled = true
 ### Processing batch Some(0):
 [0,]
-{uid: 0, details: {"receivedDocuments":2,"indexedDocuments":null}, stats: {"totalNbTasks":2,"status":{"enqueued":2},"types":{"documentAdditionOrUpdate":2},"indexUids":{"doggos":2}}, }
+{uid: 0, details: {"receivedDocuments":1,"indexedDocuments":null}, stats: {"totalNbTasks":1,"status":{"processing":1},"types":{"documentAdditionOrUpdate":1},"indexUids":{"doggos":1}}, }
 ----------------------------------------------------------------------
 ### All Tasks:
 0 {uid: 0, status: enqueued, details: { received_documents: 1, indexed_documents: None }, kind: DocumentAdditionOrUpdate { index_uid: "doggos", primary_key: Some("id"), method: ReplaceDocuments, content_file: 00000000-0000-0000-0000-000000000000, documents_count: 1, allow_index_creation: true }}

crates/index-scheduler/src/snapshots/lib.rs/insert_task_while_another_task_is_processing/after_batch_creation.snap

@@ -5,7 +5,7 @@ snapshot_kind: text
 ### Autobatching Enabled = true
 ### Processing batch Some(0):
 [0,]
-{uid: 0, details: {"primaryKey":"id"}, stats: {"totalNbTasks":2,"status":{"enqueued":2},"types":{"indexCreation":2},"indexUids":{"index_a":2}}, }
+{uid: 0, details: {"primaryKey":"id"}, stats: {"totalNbTasks":1,"status":{"processing":1},"types":{"indexCreation":1},"indexUids":{"index_a":1}}, }
 ----------------------------------------------------------------------
 ### All Tasks:
 0 {uid: 0, status: enqueued, details: { primary_key: Some("id") }, kind: IndexCreation { index_uid: "index_a", primary_key: Some("id") }}

crates/index-scheduler/src/snapshots/lib.rs/insert_task_while_another_task_is_processing/registered_the_second_task.snap

@@ -5,7 +5,7 @@ snapshot_kind: text
 ### Autobatching Enabled = true
 ### Processing batch Some(0):
 [0,]
-{uid: 0, details: {"primaryKey":"id"}, stats: {"totalNbTasks":2,"status":{"enqueued":2},"types":{"indexCreation":2},"indexUids":{"index_a":2}}, }
+{uid: 0, details: {"primaryKey":"id"}, stats: {"totalNbTasks":1,"status":{"processing":1},"types":{"indexCreation":1},"indexUids":{"index_a":1}}, }
 ----------------------------------------------------------------------
 ### All Tasks:
 0 {uid: 0, status: enqueued, details: { primary_key: Some("id") }, kind: IndexCreation { index_uid: "index_a", primary_key: Some("id") }}

crates/index-scheduler/src/snapshots/lib.rs/insert_task_while_another_task_is_processing/registered_the_third_task.snap

@@ -5,7 +5,7 @@ snapshot_kind: text
 ### Autobatching Enabled = true
 ### Processing batch Some(0):
 [0,]
-{uid: 0, details: {"primaryKey":"id"}, stats: {"totalNbTasks":2,"status":{"enqueued":2},"types":{"indexCreation":2},"indexUids":{"index_a":2}}, }
+{uid: 0, details: {"primaryKey":"id"}, stats: {"totalNbTasks":1,"status":{"processing":1},"types":{"indexCreation":1},"indexUids":{"index_a":1}}, }
 ----------------------------------------------------------------------
 ### All Tasks:
 0 {uid: 0, status: enqueued, details: { primary_key: Some("id") }, kind: IndexCreation { index_uid: "index_a", primary_key: Some("id") }}

crates/index-scheduler/src/snapshots/lib.rs/query_batches_simple/after-advancing-a-bit.snap

@@ -5,7 +5,7 @@ snapshot_kind: text
 ### Autobatching Enabled = true
 ### Processing batch Some(1):
 [1,]
-{uid: 1, details: {"primaryKey":"sheep"}, stats: {"totalNbTasks":2,"status":{"enqueued":2},"types":{"indexCreation":2},"indexUids":{"doggo":2}}, }
+{uid: 1, details: {"primaryKey":"sheep"}, stats: {"totalNbTasks":1,"status":{"processing":1},"types":{"indexCreation":1},"indexUids":{"doggo":1}}, }
 ----------------------------------------------------------------------
 ### All Tasks:
 0 {uid: 0, batch_uid: 0, status: succeeded, details: { primary_key: Some("mouse") }, kind: IndexCreation { index_uid: "catto", primary_key: Some("mouse") }}

crates/index-scheduler/src/snapshots/lib.rs/task_deletion_undeleteable/task_deletion_processing.snap

@@ -5,7 +5,7 @@ snapshot_kind: text
 ### Autobatching Enabled = true
 ### Processing batch Some(0):
 [3,]
-{uid: 0, details: {"matchedTasks":2,"deletedTasks":null,"originalFilter":"test_query"}, stats: {"totalNbTasks":1,"status":{"enqueued":1},"types":{"taskDeletion":1},"indexUids":{}}, }
+{uid: 0, details: {"matchedTasks":2,"deletedTasks":null,"originalFilter":"test_query"}, stats: {"totalNbTasks":1,"status":{"processing":1},"types":{"taskDeletion":1},"indexUids":{}}, }
 ----------------------------------------------------------------------
 ### All Tasks:
 0 {uid: 0, status: enqueued, details: { primary_key: Some("mouse") }, kind: IndexCreation { index_uid: "catto", primary_key: Some("mouse") }}

crates/index-scheduler/src/utils.rs

@@ -67,7 +67,7 @@ impl ProcessingBatch {
             task.batch_uid = Some(self.uid);
             // We don't store the statuses in the map since they're all enqueued but we must
             // still store them in the stats since that can be displayed.
-            *self.stats.status.entry(task.status).or_default() += 1;
+            *self.stats.status.entry(Status::Processing).or_default() += 1;
 
             self.kinds.insert(task.kind.as_kind());
             *self.stats.types.entry(task.kind.as_kind()).or_default() += 1;
@@ -106,7 +106,7 @@ impl ProcessingBatch {
         self.stats.total_nb_tasks = 0;
     }
 
-    /// Update the timestamp of the tasks and the inner structure of this sturcture.
+    /// Update the timestamp of the tasks and the inner structure of this structure.
     pub fn update(&mut self, task: &mut Task) {
         // We must re-set this value in case we're dealing with a task that has been added between
         // the `processing` and `finished` state

crates/meilisearch-auth/Cargo.toml

@@ -17,7 +17,7 @@ hmac = "0.12.1"
 maplit = "1.0.2"
 meilisearch-types = { path = "../meilisearch-types" }
 rand = "0.8.5"
-roaring = { version = "0.10.6", features = ["serde"] }
+roaring = { version = "0.10.7", features = ["serde"] }
 serde = { version = "1.0.204", features = ["derive"] }
 serde_json = { version = "1.0.120", features = ["preserve_order"] }
 sha2 = "0.10.8"

crates/meilisearch-types/Cargo.toml

@@ -25,7 +25,7 @@ fst = "0.4.7"
 memmap2 = "0.9.4"
 milli = { path = "../milli" }
 raw-collections = { git = "https://github.com/meilisearch/raw-collections.git", version = "0.1.0" }
-roaring = { version = "0.10.6", features = ["serde"] }
+roaring = { version = "0.10.7", features = ["serde"] }
 serde = { version = "1.0.204", features = ["derive"] }
 serde-cs = "0.2.4"
 serde_json = "1.0.120"

crates/meilisearch-types/src/document_formats.rs

@@ -214,7 +214,7 @@ pub fn read_json(input: &File, output: impl io::Write) -> Result<u64> {
     // We memory map to be able to deserialize into a RawMap that
     // does not allocate when possible and only materialize the first/top level.
     let input = unsafe { Mmap::map(input).map_err(DocumentFormatError::Io)? };
-    let mut doc_alloc = Bump::with_capacity(1024 * 1024 * 1024); // 1MiB
+    let mut doc_alloc = Bump::with_capacity(1024 * 1024); // 1MiB
 
     let mut out = BufWriter::new(output);
     let mut deserializer = serde_json::Deserializer::from_slice(&input);

crates/meilisearch-types/src/error.rs

@@ -279,6 +279,7 @@ InvalidSearchPage                     , InvalidRequest       , BAD_REQUEST ;
 InvalidSearchQ                        , InvalidRequest       , BAD_REQUEST ;
 InvalidFacetSearchQuery               , InvalidRequest       , BAD_REQUEST ;
 InvalidFacetSearchName                , InvalidRequest       , BAD_REQUEST ;
+FacetSearchDisabled                   , InvalidRequest       , BAD_REQUEST ;
 InvalidSearchVector                   , InvalidRequest       , BAD_REQUEST ;
 InvalidSearchShowMatchesPosition      , InvalidRequest       , BAD_REQUEST ;
 InvalidSearchShowRankingScore         , InvalidRequest       , BAD_REQUEST ;

crates/meilisearch/Cargo.toml

@@ -103,7 +103,7 @@ tracing-subscriber = { version = "0.3.18", features = ["json"] }
 tracing-trace = { version = "0.1.0", path = "../tracing-trace" }
 tracing-actix-web = "0.7.11"
 build-info = { version = "1.7.0", path = "../build-info" }
-roaring = "0.10.2"
+roaring = "0.10.7"
 mopa-maintained = "0.2.3"
 
 [dev-dependencies]

crates/meilisearch/src/main.rs

@@ -20,14 +20,14 @@ use meilisearch::{
     LogStderrType, Opt, SubscriberForSecondLayer,
 };
 use meilisearch_auth::{generate_master_key, AuthController, MASTER_KEY_MIN_SIZE};
-use mimalloc::MiMalloc;
 use termcolor::{Color, ColorChoice, ColorSpec, StandardStream, WriteColor};
 use tracing::level_filters::LevelFilter;
 use tracing_subscriber::layer::SubscriberExt as _;
 use tracing_subscriber::Layer;
 
+#[cfg(not(windows))]
 #[global_allocator]
-static ALLOC: MiMalloc = MiMalloc;
+static ALLOC: mimalloc::MiMalloc = mimalloc::MiMalloc;
 
 fn default_log_route_layer() -> LogRouteType {
     None.with_filter(tracing_subscriber::filter::Targets::new().with_target("", LevelFilter::OFF))

crates/meilisearch/src/option.rs

@@ -654,8 +654,9 @@ impl Opt {
 
 #[derive(Debug, Default, Clone, Parser, Deserialize)]
 pub struct IndexerOpts {
-    /// Sets the maximum amount of RAM Meilisearch can use when indexing. By default, Meilisearch
-    /// uses no more than two thirds of available memory.
+    /// Specifies the maximum resident memory that Meilisearch can use for indexing.
+    /// By default, Meilisearch limits the RAM usage to 5% of the total available memory.
+    /// Note that the underlying store utilizes memory-mapping and makes use of the rest.
     #[clap(long, env = MEILI_MAX_INDEXING_MEMORY, default_value_t)]
     #[serde(default)]
     pub max_indexing_memory: MaxMemory,
@@ -714,7 +715,7 @@ impl TryFrom<&IndexerOpts> for IndexerConfig {
     }
 }
 
-/// A type used to detect the max memory available and use 2/3 of it.
+/// A type used to detect the max resident memory available and use 5% of it.
 #[derive(Debug, Clone, Copy, Deserialize, Serialize)]
 pub struct MaxMemory(Option<Byte>);
 
@@ -728,7 +729,7 @@ impl FromStr for MaxMemory {
 
 impl Default for MaxMemory {
     fn default() -> MaxMemory {
-        MaxMemory(total_memory_bytes().map(|bytes| bytes * 2 / 3).map(Byte::from_u64))
+        MaxMemory(total_memory_bytes().map(|bytes| bytes * 5 / 100).map(Byte::from_u64))
     }
 }
 

crates/meilisearch/src/search/mod.rs

@@ -1407,6 +1407,13 @@ pub fn perform_facet_search(
         None => TimeBudget::default(),
     };
 
+    if !index.facet_search(&rtxn)? {
+        return Err(ResponseError::from_msg(
+            "The facet search is disabled for this index".to_string(),
+            Code::FacetSearchDisabled,
+        ));
+    }
+
     // In the faceted search context, we want to use the intersection between the locales provided by the user
     // and the locales of the facet string.
     // If the facet string is not localized, we **ignore** the locales provided by the user because the facet data has no locale.

crates/meilisearch/tests/batches/mod.rs

@@ -224,7 +224,7 @@ async fn list_batches_status_and_type_filtered() {
 }
 
 #[actix_rt::test]
-async fn get_batch_filter_error() {
+async fn list_batch_filter_error() {
     let server = Server::new().await;
 
     let (response, code) = server.batches_filter("lol=pied").await;

crates/meilisearch/tests/common/mod.rs

@@ -52,6 +52,25 @@ impl Value {
         }
         self
     }
+
+    /// Return `true` if the `status` field is set to `failed`.
+    /// Panic if the `status` field doesn't exists.
+    #[track_caller]
+    pub fn is_fail(&self) -> bool {
+        if !self["status"].is_string() {
+            panic!("Called `is_fail` on {}", serde_json::to_string_pretty(&self.0).unwrap());
+        }
+        self["status"] == serde_json::Value::String(String::from("failed"))
+    }
+
+    // Panic if the json doesn't contain the `status` field set to "succeeded"
+    #[track_caller]
+    pub fn failed(&self) -> &Self {
+        if !self.is_fail() {
+            panic!("Called failed on {}", serde_json::to_string_pretty(&self.0).unwrap());
+        }
+        self
+    }
 }
 
 impl From<serde_json::Value> for Value {

crates/meilisearch/tests/search/facet_search.rs

@@ -221,8 +221,15 @@ async fn add_documents_and_deactivate_facet_search() {
     let (response, code) =
         index.facet_search(json!({"facetName": "genres", "facetQuery": "a"})).await;
 
-    assert_eq!(code, 200, "{}", response);
-    assert_eq!(dbg!(response)["facetHits"].as_array().unwrap().len(), 0);
+    assert_eq!(code, 400, "{}", response);
+    snapshot!(response, @r###"
+    {
+      "message": "The facet search is disabled for this index",
+      "code": "facet_search_disabled",
+      "type": "invalid_request",
+      "link": "https://docs.meilisearch.com/errors#facet_search_disabled"
+    }
+    "###);
 }
 
 #[actix_rt::test]
@@ -245,8 +252,15 @@ async fn deactivate_facet_search_and_add_documents() {
     let (response, code) =
         index.facet_search(json!({"facetName": "genres", "facetQuery": "a"})).await;
 
-    assert_eq!(code, 200, "{}", response);
-    assert_eq!(dbg!(response)["facetHits"].as_array().unwrap().len(), 0);
+    assert_eq!(code, 400, "{}", response);
+    snapshot!(response, @r###"
+    {
+      "message": "The facet search is disabled for this index",
+      "code": "facet_search_disabled",
+      "type": "invalid_request",
+      "link": "https://docs.meilisearch.com/errors#facet_search_disabled"
+    }
+    "###);
 }
 
 #[actix_rt::test]

crates/meilisearch/tests/snapshot/mod.rs

@@ -129,11 +129,11 @@ async fn perform_on_demand_snapshot() {
 
     index.load_test_set().await;
 
-    server.index("doggo").create(Some("bone")).await;
-    index.wait_task(2).await;
+    let (task, _) = server.index("doggo").create(Some("bone")).await;
+    index.wait_task(task.uid()).await.succeeded();
 
-    server.index("doggo").create(Some("bone")).await;
-    index.wait_task(2).await;
+    let (task, _) = server.index("doggo").create(Some("bone")).await;
+    index.wait_task(task.uid()).await.failed();
 
     let (task, code) = server.create_snapshot().await;
     snapshot!(code, @"202 Accepted");

crates/milli/Cargo.toml

@@ -42,7 +42,7 @@ obkv = "0.3.0"
 once_cell = "1.19.0"
 ordered-float = "4.2.1"
 rayon = "1.10.0"
-roaring = { version = "0.10.6", features = ["serde"] }
+roaring = { version = "0.10.7", features = ["serde"] }
 rstar = { version = "0.12.0", features = ["serde"] }
 serde = { version = "1.0.204", features = ["derive"] }
 serde_json = { version = "1.0.120", features = ["preserve_order", "raw_value"] }
@@ -98,6 +98,8 @@ allocator-api2 = "0.2.18"
 rustc-hash = "2.0.0"
 uell = "0.1.0"
 enum-iterator = "2.1.0"
+bbqueue = { git = "https://github.com/meilisearch/bbqueue" }
+flume = { version = "0.11.1", default-features = false }
 
 [dev-dependencies]
 mimalloc = { version = "0.1.43", default-features = false }

crates/milli/src/error.rs

@@ -3,6 +3,7 @@ use std::convert::Infallible;
 use std::fmt::Write;
 use std::{io, str};
 
+use bstr::BString;
 use heed::{Error as HeedError, MdbError};
 use rayon::ThreadPoolBuildError;
 use rhai::EvalAltResult;
@@ -62,9 +63,9 @@ pub enum InternalError {
     #[error(transparent)]
     Store(#[from] MdbError),
     #[error("Cannot delete {key:?} from database {database_name}: {error}")]
-    StoreDeletion { database_name: &'static str, key: Vec<u8>, error: heed::Error },
+    StoreDeletion { database_name: &'static str, key: BString, error: heed::Error },
     #[error("Cannot insert {key:?} and value with length {value_length} into database {database_name}: {error}")]
-    StorePut { database_name: &'static str, key: Vec<u8>, value_length: usize, error: heed::Error },
+    StorePut { database_name: &'static str, key: BString, value_length: usize, error: heed::Error },
     #[error(transparent)]
     Utf8(#[from] str::Utf8Error),
     #[error("An indexation process was explicitly aborted")]

crates/milli/src/heed_codec/facet/mod.rs

@@ -97,7 +97,7 @@ impl<'a> heed::BytesEncode<'a> for FacetGroupValueCodec {
 
     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);
+        CboRoaringBitmapCodec::serialize_into_vec(&value.bitmap, &mut v);
         Ok(Cow::Owned(v))
     }
 }

crates/milli/src/heed_codec/roaring_bitmap/cbo_roaring_bitmap_codec.rs

@@ -27,18 +27,27 @@ impl CboRoaringBitmapCodec {
         }
     }
 
-    pub fn serialize_into(roaring: &RoaringBitmap, vec: &mut Vec<u8>) {
+    pub fn serialize_into_vec(roaring: &RoaringBitmap, vec: &mut Vec<u8>) {
+        Self::serialize_into_writer(roaring, vec).unwrap()
+    }
+
+    pub fn serialize_into_writer<W: io::Write>(
+        roaring: &RoaringBitmap,
+        mut writer: W,
+    ) -> io::Result<()> {
         if roaring.len() <= THRESHOLD as u64 {
             // If the number of items (u32s) to encode is less than or equal to the threshold
             // it means that it would weigh the same or less than the RoaringBitmap
             // header, so we directly encode them using ByteOrder instead.
             for integer in roaring {
-                vec.write_u32::<NativeEndian>(integer).unwrap();
+                writer.write_u32::<NativeEndian>(integer)?;
             }
         } else {
             // Otherwise, we use the classic RoaringBitmapCodec that writes a header.
-            roaring.serialize_into(vec).unwrap();
+            roaring.serialize_into(writer)?;
         }
+
+        Ok(())
     }
 
     pub fn deserialize_from(mut bytes: &[u8]) -> io::Result<RoaringBitmap> {
@@ -143,7 +152,7 @@ impl CboRoaringBitmapCodec {
             return Ok(None);
         }
 
-        Self::serialize_into(&previous, buffer);
+        Self::serialize_into_vec(&previous, buffer);
         Ok(Some(&buffer[..]))
     }
 }
@@ -169,7 +178,7 @@ impl heed::BytesEncode<'_> for CboRoaringBitmapCodec {
 
     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);
+        Self::serialize_into_vec(item, &mut vec);
         Ok(Cow::Owned(vec))
     }
 }

crates/milli/src/index.rs

@@ -1821,6 +1821,7 @@ pub(crate) mod tests {
                 indexer::index(
                     wtxn,
                     &self.inner,
+                    &crate::ThreadPoolNoAbortBuilder::new().build().unwrap(),
                     indexer_config.grenad_parameters(),
                     &db_fields_ids_map,
                     new_fields_ids_map,
@@ -1911,6 +1912,7 @@ pub(crate) mod tests {
                 indexer::index(
                     wtxn,
                     &self.inner,
+                    &crate::ThreadPoolNoAbortBuilder::new().build().unwrap(),
                     indexer_config.grenad_parameters(),
                     &db_fields_ids_map,
                     new_fields_ids_map,
@@ -1991,6 +1993,7 @@ pub(crate) mod tests {
                 indexer::index(
                     &mut wtxn,
                     &index.inner,
+                    &crate::ThreadPoolNoAbortBuilder::new().build().unwrap(),
                     indexer_config.grenad_parameters(),
                     &db_fields_ids_map,
                     new_fields_ids_map,

crates/milli/src/lib.rs

@@ -1,6 +1,7 @@
 #![cfg_attr(all(test, fuzzing), feature(no_coverage))]
 #![allow(clippy::type_complexity)]
 
+#[cfg(not(windows))]
 #[cfg(test)]
 #[global_allocator]
 pub static ALLOC: mimalloc::MiMalloc = mimalloc::MiMalloc;

crates/milli/src/search/new/tests/integration.rs

@@ -83,6 +83,7 @@ pub fn setup_search_index_with_criteria(criteria: &[Criterion]) -> Index {
     indexer::index(
         &mut wtxn,
         &index,
+        &crate::ThreadPoolNoAbortBuilder::new().build().unwrap(),
         config.grenad_parameters(),
         &db_fields_ids_map,
         new_fields_ids_map,

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

@@ -2155,6 +2155,7 @@ mod tests {
         indexer::index(
             &mut wtxn,
             &index.inner,
+            &crate::ThreadPoolNoAbortBuilder::new().build().unwrap(),
             indexer_config.grenad_parameters(),
             &db_fields_ids_map,
             new_fields_ids_map,
@@ -2216,6 +2217,7 @@ mod tests {
         indexer::index(
             &mut wtxn,
             &index.inner,
+            &crate::ThreadPoolNoAbortBuilder::new().build().unwrap(),
             indexer_config.grenad_parameters(),
             &db_fields_ids_map,
             new_fields_ids_map,
@@ -2268,6 +2270,7 @@ mod tests {
         indexer::index(
             &mut wtxn,
             &index.inner,
+            &crate::ThreadPoolNoAbortBuilder::new().build().unwrap(),
             indexer_config.grenad_parameters(),
             &db_fields_ids_map,
             new_fields_ids_map,
@@ -2319,6 +2322,7 @@ mod tests {
         indexer::index(
             &mut wtxn,
             &index.inner,
+            &crate::ThreadPoolNoAbortBuilder::new().build().unwrap(),
             indexer_config.grenad_parameters(),
             &db_fields_ids_map,
             new_fields_ids_map,
@@ -2372,6 +2376,7 @@ mod tests {
         indexer::index(
             &mut wtxn,
             &index.inner,
+            &crate::ThreadPoolNoAbortBuilder::new().build().unwrap(),
             indexer_config.grenad_parameters(),
             &db_fields_ids_map,
             new_fields_ids_map,
@@ -2430,6 +2435,7 @@ mod tests {
         indexer::index(
             &mut wtxn,
             &index.inner,
+            &crate::ThreadPoolNoAbortBuilder::new().build().unwrap(),
             indexer_config.grenad_parameters(),
             &db_fields_ids_map,
             new_fields_ids_map,
@@ -2481,6 +2487,7 @@ mod tests {
         indexer::index(
             &mut wtxn,
             &index.inner,
+            &crate::ThreadPoolNoAbortBuilder::new().build().unwrap(),
             indexer_config.grenad_parameters(),
             &db_fields_ids_map,
             new_fields_ids_map,
@@ -2532,6 +2539,7 @@ mod tests {
         indexer::index(
             &mut wtxn,
             &index.inner,
+            &crate::ThreadPoolNoAbortBuilder::new().build().unwrap(),
             indexer_config.grenad_parameters(),
             &db_fields_ids_map,
             new_fields_ids_map,
@@ -2725,6 +2733,7 @@ mod tests {
         indexer::index(
             &mut wtxn,
             &index.inner,
+            &crate::ThreadPoolNoAbortBuilder::new().build().unwrap(),
             indexer_config.grenad_parameters(),
             &db_fields_ids_map,
             new_fields_ids_map,
@@ -2783,6 +2792,7 @@ mod tests {
         indexer::index(
             &mut wtxn,
             &index.inner,
+            &crate::ThreadPoolNoAbortBuilder::new().build().unwrap(),
             indexer_config.grenad_parameters(),
             &db_fields_ids_map,
             new_fields_ids_map,
@@ -2838,6 +2848,7 @@ mod tests {
         indexer::index(
             &mut wtxn,
             &index.inner,
+            &crate::ThreadPoolNoAbortBuilder::new().build().unwrap(),
             indexer_config.grenad_parameters(),
             &db_fields_ids_map,
             new_fields_ids_map,

crates/milli/src/update/new/document.rs

@@ -1,5 +1,6 @@
 use std::collections::{BTreeMap, BTreeSet};
 
+use either::Either;
 use heed::RoTxn;
 use raw_collections::RawMap;
 use serde_json::value::RawValue;
@@ -209,29 +210,34 @@ impl<'d, 'doc: 'd, 't: 'd, Mapper: FieldIdMapper> Document<'d>
     for MergedDocument<'d, 'doc, 't, Mapper>
 {
     fn iter_top_level_fields(&self) -> impl Iterator<Item = Result<(&'d str, &'d RawValue)>> {
-        let mut new_doc_it = self.new_doc.iter_top_level_fields();
-        let mut db_it = self.db.iter().flat_map(|db| db.iter_top_level_fields());
-        let mut seen_fields = BTreeSet::new();
+        match &self.db {
+            Some(db) => {
+                let mut new_doc_it = self.new_doc.iter_top_level_fields();
+                let mut db_it = db.iter_top_level_fields();
+                let mut seen_fields = BTreeSet::new();
 
-        std::iter::from_fn(move || {
-            if let Some(next) = new_doc_it.next() {
-                if let Ok((name, _)) = next {
-                    seen_fields.insert(name);
-                }
-                return Some(next);
-            }
-            loop {
-                match db_it.next()? {
-                    Ok((name, value)) => {
-                        if seen_fields.contains(name) {
-                            continue;
+                Either::Left(std::iter::from_fn(move || {
+                    if let Some(next) = new_doc_it.next() {
+                        if let Ok((name, _)) = next {
+                            seen_fields.insert(name);
                         }
-                        return Some(Ok((name, value)));
+                        return Some(next);
                     }
-                    Err(err) => return Some(Err(err)),
-                }
+                    loop {
+                        match db_it.next()? {
+                            Ok((name, value)) => {
+                                if seen_fields.contains(name) {
+                                    continue;
+                                }
+                                return Some(Ok((name, value)));
+                            }
+                            Err(err) => return Some(Err(err)),
+                        }
+                    }
+                }))
             }
-        })
+            None => Either::Right(self.new_doc.iter_top_level_fields()),
+        }
     }
 
     fn vectors_field(&self) -> Result<Option<&'d RawValue>> {

crates/milli/src/update/new/document_change.rs

@@ -1,7 +1,10 @@
 use bumpalo::Bump;
 use heed::RoTxn;
 
-use super::document::{DocumentFromDb, DocumentFromVersions, MergedDocument, Versions};
+use super::document::{
+    Document as _, DocumentFromDb, DocumentFromVersions, MergedDocument, Versions,
+};
+use super::extract::perm_json_p;
 use super::vector_document::{
     MergedVectorDocument, VectorDocumentFromDb, VectorDocumentFromVersions,
 };
@@ -164,6 +167,80 @@ impl<'doc> Update<'doc> {
         }
     }
 
+    /// Returns whether the updated version of the document is different from the current version for the passed subset of fields.
+    ///
+    /// `true` if at least one top-level-field that is a exactly a member of field or a parent of a member of field changed.
+    /// Otherwise `false`.
+    pub fn has_changed_for_fields<'t, Mapper: FieldIdMapper>(
+        &self,
+        fields: Option<&[&str]>,
+        rtxn: &'t RoTxn,
+        index: &'t Index,
+        mapper: &'t Mapper,
+    ) -> Result<bool> {
+        let mut changed = false;
+        let mut cached_current = None;
+        let mut updated_selected_field_count = 0;
+
+        for entry in self.updated().iter_top_level_fields() {
+            let (key, updated_value) = entry?;
+
+            if perm_json_p::select_field(key, fields, &[]) == perm_json_p::Selection::Skip {
+                continue;
+            }
+
+            updated_selected_field_count += 1;
+            let current = match cached_current {
+                Some(current) => current,
+                None => self.current(rtxn, index, mapper)?,
+            };
+            let current_value = current.top_level_field(key)?;
+            let Some(current_value) = current_value else {
+                changed = true;
+                break;
+            };
+
+            if current_value.get() != updated_value.get() {
+                changed = true;
+                break;
+            }
+            cached_current = Some(current);
+        }
+
+        if !self.has_deletion {
+            // no field deletion, so fields that don't appear in `updated` cannot have changed
+            return Ok(changed);
+        }
+
+        if changed {
+            return Ok(true);
+        }
+
+        // we saw all updated fields, and set `changed` if any field wasn't in `current`.
+        // so if there are as many fields in `current` as in `updated`, then nothing changed.
+        // If there is any more fields in `current`, then they are missing in `updated`.
+        let has_deleted_fields = {
+            let current = match cached_current {
+                Some(current) => current,
+                None => self.current(rtxn, index, mapper)?,
+            };
+
+            let mut current_selected_field_count = 0;
+            for entry in current.iter_top_level_fields() {
+                let (key, _) = entry?;
+
+                if perm_json_p::select_field(key, fields, &[]) == perm_json_p::Selection::Skip {
+                    continue;
+                }
+                current_selected_field_count += 1;
+            }
+
+            current_selected_field_count != updated_selected_field_count
+        };
+
+        Ok(has_deleted_fields)
+    }
+
     pub fn updated_vectors(
         &self,
         doc_alloc: &'doc Bump,

crates/milli/src/update/new/extract/cache.rs

@@ -415,21 +415,21 @@ fn spill_entry_to_sorter(
     match deladd {
         DelAddRoaringBitmap { del: Some(del), add: None } => {
             cbo_buffer.clear();
-            CboRoaringBitmapCodec::serialize_into(&del, cbo_buffer);
+            CboRoaringBitmapCodec::serialize_into_vec(&del, cbo_buffer);
             value_writer.insert(DelAdd::Deletion, &cbo_buffer)?;
         }
         DelAddRoaringBitmap { del: None, add: Some(add) } => {
             cbo_buffer.clear();
-            CboRoaringBitmapCodec::serialize_into(&add, cbo_buffer);
+            CboRoaringBitmapCodec::serialize_into_vec(&add, cbo_buffer);
             value_writer.insert(DelAdd::Addition, &cbo_buffer)?;
         }
         DelAddRoaringBitmap { del: Some(del), add: Some(add) } => {
             cbo_buffer.clear();
-            CboRoaringBitmapCodec::serialize_into(&del, cbo_buffer);
+            CboRoaringBitmapCodec::serialize_into_vec(&del, cbo_buffer);
             value_writer.insert(DelAdd::Deletion, &cbo_buffer)?;
 
             cbo_buffer.clear();
-            CboRoaringBitmapCodec::serialize_into(&add, cbo_buffer);
+            CboRoaringBitmapCodec::serialize_into_vec(&add, cbo_buffer);
             value_writer.insert(DelAdd::Addition, &cbo_buffer)?;
         }
         DelAddRoaringBitmap { del: None, add: None } => return Ok(()),
@@ -466,12 +466,13 @@ pub fn transpose_and_freeze_caches<'a, 'extractor>(
     Ok(bucket_caches)
 }
 
-/// Merges the caches that must be all associated to the same bucket.
+/// Merges the caches that must be all associated to the same bucket
+/// but make sure to sort the different buckets before performing the merges.
 ///
 /// # Panics
 ///
 /// - If the bucket IDs in these frozen caches are not exactly the same.
-pub fn merge_caches<F>(frozen: Vec<FrozenCache>, mut f: F) -> Result<()>
+pub fn merge_caches_sorted<F>(frozen: Vec<FrozenCache>, mut f: F) -> Result<()>
 where
     F: for<'a> FnMut(&'a [u8], DelAddRoaringBitmap) -> Result<()>,
 {
@@ -543,12 +544,12 @@ where
 
     // Then manage the content on the HashMap entries that weren't taken (mem::take).
     while let Some(mut map) = maps.pop() {
-        for (key, bbbul) in map.iter_mut() {
-            // Make sure we don't try to work with entries already managed by the spilled
-            if bbbul.is_empty() {
-                continue;
-            }
+        // Make sure we don't try to work with entries already managed by the spilled
+        let mut ordered_entries: Vec<_> =
+            map.iter_mut().filter(|(_, bbbul)| !bbbul.is_empty()).collect();
+        ordered_entries.sort_unstable_by_key(|(key, _)| *key);
 
+        for (key, bbbul) in ordered_entries {
             let mut output = DelAddRoaringBitmap::empty();
             output.union_and_clear_bbbul(bbbul);
 

crates/milli/src/update/new/extract/documents.rs

@@ -12,13 +12,14 @@ use crate::update::new::thread_local::FullySend;
 use crate::update::new::DocumentChange;
 use crate::vector::EmbeddingConfigs;
 use crate::Result;
-pub struct DocumentsExtractor<'a> {
-    document_sender: &'a DocumentsSender<'a>,
+
+pub struct DocumentsExtractor<'a, 'b> {
+    document_sender: DocumentsSender<'a, 'b>,
     embedders: &'a EmbeddingConfigs,
 }
 
-impl<'a> DocumentsExtractor<'a> {
-    pub fn new(document_sender: &'a DocumentsSender<'a>, embedders: &'a EmbeddingConfigs) -> Self {
+impl<'a, 'b> DocumentsExtractor<'a, 'b> {
+    pub fn new(document_sender: DocumentsSender<'a, 'b>, embedders: &'a EmbeddingConfigs) -> Self {
         Self { document_sender, embedders }
     }
 }
@@ -29,7 +30,7 @@ pub struct DocumentExtractorData {
     pub field_distribution_delta: HashMap<String, i64>,
 }
 
-impl<'a, 'extractor> Extractor<'extractor> for DocumentsExtractor<'a> {
+impl<'a, 'b, 'extractor> Extractor<'extractor> for DocumentsExtractor<'a, 'b> {
     type Data = FullySend<RefCell<DocumentExtractorData>>;
 
     fn init_data(&self, _extractor_alloc: &'extractor Bump) -> Result<Self::Data> {

crates/milli/src/update/new/extract/faceted/extract_facets.rs

@@ -25,14 +25,14 @@ use crate::update::new::DocumentChange;
 use crate::update::GrenadParameters;
 use crate::{DocumentId, FieldId, Index, Result, MAX_FACET_VALUE_LENGTH};
 
-pub struct FacetedExtractorData<'a> {
+pub struct FacetedExtractorData<'a, 'b> {
     attributes_to_extract: &'a [&'a str],
-    sender: &'a FieldIdDocidFacetSender<'a>,
+    sender: &'a FieldIdDocidFacetSender<'a, 'b>,
     grenad_parameters: GrenadParameters,
     buckets: usize,
 }
 
-impl<'a, 'extractor> Extractor<'extractor> for FacetedExtractorData<'a> {
+impl<'a, 'b, 'extractor> Extractor<'extractor> for FacetedExtractorData<'a, 'b> {
     type Data = RefCell<BalancedCaches<'extractor>>;
 
     fn init_data(&self, extractor_alloc: &'extractor Bump) -> Result<Self::Data> {
@@ -97,6 +97,15 @@ impl FacetedDocidsExtractor {
                 },
             ),
             DocumentChange::Update(inner) => {
+                if !inner.has_changed_for_fields(
+                    Some(attributes_to_extract),
+                    rtxn,
+                    index,
+                    context.db_fields_ids_map,
+                )? {
+                    return Ok(());
+                }
+
                 extract_document_facets(
                     attributes_to_extract,
                     inner.current(rtxn, index, context.db_fields_ids_map)?,
@@ -318,7 +327,7 @@ impl<'doc> DelAddFacetValue<'doc> {
         docid: DocumentId,
         sender: &FieldIdDocidFacetSender,
         doc_alloc: &Bump,
-    ) -> std::result::Result<(), crossbeam_channel::SendError<()>> {
+    ) -> crate::Result<()> {
         let mut buffer = bumpalo::collections::Vec::new_in(doc_alloc);
         for ((fid, value), deladd) in self.strings {
             if let Ok(s) = std::str::from_utf8(&value) {

crates/milli/src/update/new/extract/geo/mod.rs

@@ -1,6 +1,6 @@
 use std::cell::RefCell;
 use std::fs::File;
-use std::io::{self, BufReader, BufWriter, ErrorKind, Read, Write as _};
+use std::io::{self, BufReader, BufWriter, ErrorKind, Read, Seek as _, Write as _};
 use std::{iter, mem, result};
 
 use bumpalo::Bump;
@@ -97,30 +97,34 @@ pub struct FrozenGeoExtractorData<'extractor> {
 impl<'extractor> FrozenGeoExtractorData<'extractor> {
     pub fn iter_and_clear_removed(
         &mut self,
-    ) -> impl IntoIterator<Item = io::Result<ExtractedGeoPoint>> + '_ {
-        mem::take(&mut self.removed)
+    ) -> io::Result<impl IntoIterator<Item = io::Result<ExtractedGeoPoint>> + '_> {
+        Ok(mem::take(&mut self.removed)
             .iter()
             .copied()
             .map(Ok)
-            .chain(iterator_over_spilled_geopoints(&mut self.spilled_removed))
+            .chain(iterator_over_spilled_geopoints(&mut self.spilled_removed)?))
     }
 
     pub fn iter_and_clear_inserted(
         &mut self,
-    ) -> impl IntoIterator<Item = io::Result<ExtractedGeoPoint>> + '_ {
-        mem::take(&mut self.inserted)
+    ) -> io::Result<impl IntoIterator<Item = io::Result<ExtractedGeoPoint>> + '_> {
+        Ok(mem::take(&mut self.inserted)
             .iter()
             .copied()
             .map(Ok)
-            .chain(iterator_over_spilled_geopoints(&mut self.spilled_inserted))
+            .chain(iterator_over_spilled_geopoints(&mut self.spilled_inserted)?))
     }
 }
 
 fn iterator_over_spilled_geopoints(
     spilled: &mut Option<BufReader<File>>,
-) -> impl IntoIterator<Item = io::Result<ExtractedGeoPoint>> + '_ {
+) -> io::Result<impl IntoIterator<Item = io::Result<ExtractedGeoPoint>> + '_> {
     let mut spilled = spilled.take();
-    iter::from_fn(move || match &mut spilled {
+    if let Some(spilled) = &mut spilled {
+        spilled.rewind()?;
+    }
+
+    Ok(iter::from_fn(move || match &mut spilled {
         Some(file) => {
             let geopoint_bytes = &mut [0u8; mem::size_of::<ExtractedGeoPoint>()];
             match file.read_exact(geopoint_bytes) {
@@ -130,7 +134,7 @@ fn iterator_over_spilled_geopoints(
             }
         }
         None => None,
-    })
+    }))
 }
 
 impl<'extractor> Extractor<'extractor> for GeoExtractor {
@@ -157,7 +161,9 @@ impl<'extractor> Extractor<'extractor> for GeoExtractor {
         let mut data_ref = context.data.borrow_mut_or_yield();
 
         for change in changes {
-            if max_memory.map_or(false, |mm| context.extractor_alloc.allocated_bytes() >= mm) {
+            if data_ref.spilled_removed.is_none()
+                && max_memory.map_or(false, |mm| context.extractor_alloc.allocated_bytes() >= mm)
+            {
                 // We must spill as we allocated too much memory
                 data_ref.spilled_removed = tempfile::tempfile().map(BufWriter::new).map(Some)?;
                 data_ref.spilled_inserted = tempfile::tempfile().map(BufWriter::new).map(Some)?;

crates/milli/src/update/new/extract/mod.rs

@@ -6,7 +6,9 @@ mod searchable;
 mod vectors;
 
 use bumpalo::Bump;
-pub use cache::{merge_caches, transpose_and_freeze_caches, BalancedCaches, DelAddRoaringBitmap};
+pub use cache::{
+    merge_caches_sorted, transpose_and_freeze_caches, BalancedCaches, DelAddRoaringBitmap,
+};
 pub use documents::*;
 pub use faceted::*;
 pub use geo::*;

crates/milli/src/update/new/extract/searchable/extract_word_docids.rs

@@ -351,6 +351,15 @@ impl WordDocidsExtractors {
                 )?;
             }
             DocumentChange::Update(inner) => {
+                if !inner.has_changed_for_fields(
+                    document_tokenizer.attribute_to_extract,
+                    &context.rtxn,
+                    context.index,
+                    context.db_fields_ids_map,
+                )? {
+                    return Ok(());
+                }
+
                 let mut token_fn = |fname: &str, fid, pos, word: &str| {
                     cached_sorter.insert_del_u32(
                         fid,

crates/milli/src/update/new/extract/searchable/extract_word_pair_proximity_docids.rs

@@ -70,6 +70,15 @@ impl SearchableExtractor for WordPairProximityDocidsExtractor {
                 )?;
             }
             DocumentChange::Update(inner) => {
+                if !inner.has_changed_for_fields(
+                    document_tokenizer.attribute_to_extract,
+                    rtxn,
+                    index,
+                    context.db_fields_ids_map,
+                )? {
+                    return Ok(());
+                }
+
                 let document = inner.current(rtxn, index, context.db_fields_ids_map)?;
                 process_document_tokens(
                     document,

crates/milli/src/update/new/extract/vectors/mod.rs

@@ -18,17 +18,17 @@ use crate::vector::error::{
 use crate::vector::{Embedder, Embedding, EmbeddingConfigs};
 use crate::{DocumentId, FieldDistribution, InternalError, Result, ThreadPoolNoAbort, UserError};
 
-pub struct EmbeddingExtractor<'a> {
+pub struct EmbeddingExtractor<'a, 'b> {
     embedders: &'a EmbeddingConfigs,
-    sender: &'a EmbeddingSender<'a>,
+    sender: EmbeddingSender<'a, 'b>,
     possible_embedding_mistakes: PossibleEmbeddingMistakes,
     threads: &'a ThreadPoolNoAbort,
 }
 
-impl<'a> EmbeddingExtractor<'a> {
+impl<'a, 'b> EmbeddingExtractor<'a, 'b> {
     pub fn new(
         embedders: &'a EmbeddingConfigs,
-        sender: &'a EmbeddingSender<'a>,
+        sender: EmbeddingSender<'a, 'b>,
         field_distribution: &'a FieldDistribution,
         threads: &'a ThreadPoolNoAbort,
     ) -> Self {
@@ -43,7 +43,7 @@ pub struct EmbeddingExtractorData<'extractor>(
 
 unsafe impl MostlySend for EmbeddingExtractorData<'_> {}
 
-impl<'a, 'extractor> Extractor<'extractor> for EmbeddingExtractor<'a> {
+impl<'a, 'b, 'extractor> Extractor<'extractor> for EmbeddingExtractor<'a, 'b> {
     type Data = RefCell<EmbeddingExtractorData<'extractor>>;
 
     fn init_data<'doc>(&'doc self, extractor_alloc: &'extractor Bump) -> crate::Result<Self::Data> {
@@ -259,7 +259,7 @@ impl<'a, 'extractor> Extractor<'extractor> for EmbeddingExtractor<'a> {
 // Currently this is the case as:
 // 1. BVec are inside of the bumaplo
 // 2. All other fields are either trivial (u8) or references.
-struct Chunks<'a, 'extractor> {
+struct Chunks<'a, 'b, 'extractor> {
     texts: BVec<'a, &'a str>,
     ids: BVec<'a, DocumentId>,
 
@@ -270,11 +270,11 @@ struct Chunks<'a, 'extractor> {
     possible_embedding_mistakes: &'a PossibleEmbeddingMistakes,
     user_provided: &'a RefCell<EmbeddingExtractorData<'extractor>>,
     threads: &'a ThreadPoolNoAbort,
-    sender: &'a EmbeddingSender<'a>,
+    sender: EmbeddingSender<'a, 'b>,
     has_manual_generation: Option<&'a str>,
 }
 
-impl<'a, 'extractor> Chunks<'a, 'extractor> {
+impl<'a, 'b, 'extractor> Chunks<'a, 'b, 'extractor> {
     #[allow(clippy::too_many_arguments)]
     pub fn new(
         embedder: &'a Embedder,
@@ -284,7 +284,7 @@ impl<'a, 'extractor> Chunks<'a, 'extractor> {
         user_provided: &'a RefCell<EmbeddingExtractorData<'extractor>>,
         possible_embedding_mistakes: &'a PossibleEmbeddingMistakes,
         threads: &'a ThreadPoolNoAbort,
-        sender: &'a EmbeddingSender<'a>,
+        sender: EmbeddingSender<'a, 'b>,
         doc_alloc: &'a Bump,
     ) -> Self {
         let capacity = embedder.prompt_count_in_chunk_hint() * embedder.chunk_count_hint();
@@ -368,7 +368,7 @@ impl<'a, 'extractor> Chunks<'a, 'extractor> {
         possible_embedding_mistakes: &PossibleEmbeddingMistakes,
         unused_vectors_distribution: &UnusedVectorsDistributionBump,
         threads: &ThreadPoolNoAbort,
-        sender: &EmbeddingSender<'a>,
+        sender: EmbeddingSender<'a, 'b>,
         has_manual_generation: Option<&'a str>,
     ) -> Result<()> {
         if let Some(external_docid) = has_manual_generation {

crates/milli/src/update/new/indexer/document_changes.rs

@@ -70,7 +70,7 @@ impl<
         F: FnOnce(&'extractor Bump) -> Result<T>,
     {
         let doc_alloc =
-            doc_allocs.get_or(|| FullySend(Cell::new(Bump::with_capacity(1024 * 1024 * 1024))));
+            doc_allocs.get_or(|| FullySend(Cell::new(Bump::with_capacity(1024 * 1024))));
         let doc_alloc = doc_alloc.0.take();
         let fields_ids_map = fields_ids_map_store
             .get_or(|| RefCell::new(GlobalFieldsIdsMap::new(new_fields_ids_map)).into());

crates/milli/src/update/new/indexer/mod.rs

@@ -62,6 +62,7 @@ mod update_by_function;
 pub fn index<'pl, 'indexer, 'index, DC, MSP, SP>(
     wtxn: &mut RwTxn,
     index: &'index Index,
+    pool: &ThreadPoolNoAbort,
     grenad_parameters: GrenadParameters,
     db_fields_ids_map: &'indexer FieldsIdsMap,
     new_fields_ids_map: FieldsIdsMap,
@@ -76,9 +77,30 @@ where
     MSP: Fn() -> bool + Sync,
     SP: Fn(Progress) + Sync,
 {
-    let (extractor_sender, writer_receiver) = extractor_writer_channel(10_000);
+    let mut bbbuffers = Vec::new();
     let finished_extraction = AtomicBool::new(false);
 
+    // We compute and remove the allocated BBQueues buffers capacity from the indexing memory.
+    let minimum_capacity = 50 * 1024 * 1024 * pool.current_num_threads(); // 50 MiB
+    let (grenad_parameters, total_bbbuffer_capacity) = grenad_parameters.max_memory.map_or(
+        (grenad_parameters, 2 * minimum_capacity), // 100 MiB by thread by default
+        |max_memory| {
+            // 2% of the indexing memory
+            let total_bbbuffer_capacity = (max_memory / 100 / 2).max(minimum_capacity);
+            let new_grenad_parameters = GrenadParameters {
+                max_memory: Some(
+                    max_memory.saturating_sub(total_bbbuffer_capacity).max(100 * 1024 * 1024),
+                ),
+                ..grenad_parameters
+            };
+            (new_grenad_parameters, total_bbbuffer_capacity)
+        },
+    );
+
+    let (extractor_sender, mut writer_receiver) = pool
+        .install(|| extractor_writer_bbqueue(&mut bbbuffers, total_bbbuffer_capacity, 1000))
+        .unwrap();
+
     let metadata_builder = MetadataBuilder::from_index(index, wtxn)?;
     let new_fields_ids_map = FieldIdMapWithMetadata::new(new_fields_ids_map, metadata_builder);
     let new_fields_ids_map = RwLock::new(new_fields_ids_map);
@@ -96,6 +118,7 @@ where
         send_progress,
     };
 
+    let mut index_embeddings = index.embedding_configs(wtxn)?;
     let mut field_distribution = index.field_distribution(wtxn)?;
     let mut document_ids = index.documents_ids(wtxn)?;
 
@@ -107,261 +130,261 @@ where
         let field_distribution = &mut field_distribution;
         let document_ids = &mut document_ids;
         let extractor_handle = Builder::new().name(S("indexer-extractors")).spawn_scoped(s, move || {
-            let span = tracing::trace_span!(target: "indexing::documents", parent: &indexer_span, "extract");
-            let _entered = span.enter();
-
-            let rtxn = index.read_txn()?;
-
-
-            // document but we need to create a function that collects and compresses documents.
-            let document_sender = extractor_sender.documents();
-            let document_extractor = DocumentsExtractor::new(&document_sender, embedders);
-            let datastore = ThreadLocal::with_capacity(rayon::current_num_threads());
-            {
-                let span = tracing::trace_span!(target: "indexing::documents::extract", parent: &indexer_span, "documents");
+            pool.install(move || {
+                let span = tracing::trace_span!(target: "indexing::documents", parent: &indexer_span, "extract");
                 let _entered = span.enter();
-                extract(document_changes,
-                    &document_extractor,
-                    indexing_context,
-                    &mut extractor_allocs,
-                    &datastore,
-                    Step::ExtractingDocuments,
-                )?;
-            }
-            {
-                let span = tracing::trace_span!(target: "indexing::documents::merge", parent: &indexer_span, "documents");
-                let _entered = span.enter();
-                for document_extractor_data in datastore {
-                    let document_extractor_data = document_extractor_data.0.into_inner();
-                    for (field, delta) in document_extractor_data.field_distribution_delta {
-                        let current = field_distribution.entry(field).or_default();
-                        // adding the delta should never cause a negative result, as we are removing fields that previously existed.
-                        *current = current.saturating_add_signed(delta);
+
+                let rtxn = index.read_txn()?;
+
+                // document but we need to create a function that collects and compresses documents.
+                let document_sender = extractor_sender.documents();
+                let document_extractor = DocumentsExtractor::new(document_sender, embedders);
+                let datastore = ThreadLocal::with_capacity(rayon::current_num_threads());
+                {
+                    let span = tracing::trace_span!(target: "indexing::documents::extract", parent: &indexer_span, "documents");
+                    let _entered = span.enter();
+                    extract(
+                        document_changes,
+                        &document_extractor,
+                        indexing_context,
+                        &mut extractor_allocs,
+                        &datastore,
+                        Step::ExtractingDocuments,
+                    )?;
+                }
+                {
+                    let span = tracing::trace_span!(target: "indexing::documents::merge", parent: &indexer_span, "documents");
+                    let _entered = span.enter();
+                    for document_extractor_data in datastore {
+                        let document_extractor_data = document_extractor_data.0.into_inner();
+                        for (field, delta) in document_extractor_data.field_distribution_delta {
+                            let current = field_distribution.entry(field).or_default();
+                            // adding the delta should never cause a negative result, as we are removing fields that previously existed.
+                            *current = current.saturating_add_signed(delta);
+                        }
+                        document_extractor_data.docids_delta.apply_to(document_ids);
                     }
-                    document_extractor_data.docids_delta.apply_to(document_ids);
+
+                    field_distribution.retain(|_, v| *v != 0);
                 }
 
-                field_distribution.retain(|_, v| *v != 0);
-            }
+                let facet_field_ids_delta;
 
-            let facet_field_ids_delta;
+                {
+                    let caches = {
+                        let span = tracing::trace_span!(target: "indexing::documents::extract", parent: &indexer_span, "faceted");
+                        let _entered = span.enter();
 
-            {
-                let caches = {
-                    let span = tracing::trace_span!(target: "indexing::documents::extract", parent: &indexer_span, "faceted");
-                    let _entered = span.enter();
+                        FacetedDocidsExtractor::run_extraction(
+                                grenad_parameters,
+                                document_changes,
+                                indexing_context,
+                                &mut extractor_allocs,
+                                &extractor_sender.field_id_docid_facet_sender(),
+                                Step::ExtractingFacets
+                            )?
+                    };
 
-                    FacetedDocidsExtractor::run_extraction(
+                    {
+                        let span = tracing::trace_span!(target: "indexing::documents::merge", parent: &indexer_span, "faceted");
+                        let _entered = span.enter();
+
+                        facet_field_ids_delta = merge_and_send_facet_docids(
+                            caches,
+                            FacetDatabases::new(index),
+                            index,
+                            extractor_sender.facet_docids(),
+                        )?;
+                    }
+                }
+
+                {
+                    let WordDocidsCaches {
+                        word_docids,
+                        word_fid_docids,
+                        exact_word_docids,
+                        word_position_docids,
+                        fid_word_count_docids,
+                    } = {
+                        let span = tracing::trace_span!(target: "indexing::documents::extract", "word_docids");
+                        let _entered = span.enter();
+
+                        WordDocidsExtractors::run_extraction(
                             grenad_parameters,
                             document_changes,
                             indexing_context,
                             &mut extractor_allocs,
-                            &extractor_sender.field_id_docid_facet_sender(),
-                            Step::ExtractingFacets
+                            Step::ExtractingWords
                         )?
-                };
+                    };
 
-                {
-                    let span = tracing::trace_span!(target: "indexing::documents::merge", parent: &indexer_span, "faceted");
-                    let _entered = span.enter();
+                    {
+                        let span = tracing::trace_span!(target: "indexing::documents::merge", "word_docids");
+                        let _entered = span.enter();
+                        merge_and_send_docids(
+                            word_docids,
+                            index.word_docids.remap_types(),
+                            index,
+                            extractor_sender.docids::<WordDocids>(),
+                            &indexing_context.must_stop_processing,
+                        )?;
+                    }
 
-                    facet_field_ids_delta = merge_and_send_facet_docids(
-                        caches,
-                        FacetDatabases::new(index),
-                        index,
-                        extractor_sender.facet_docids(),
-                    )?;
-                }
-            }
+                    {
+                        let span = tracing::trace_span!(target: "indexing::documents::merge", "word_fid_docids");
+                        let _entered = span.enter();
+                        merge_and_send_docids(
+                            word_fid_docids,
+                            index.word_fid_docids.remap_types(),
+                            index,
+                            extractor_sender.docids::<WordFidDocids>(),
+                            &indexing_context.must_stop_processing,
+                        )?;
+                    }
 
-            {
+                    {
+                        let span = tracing::trace_span!(target: "indexing::documents::merge", "exact_word_docids");
+                        let _entered = span.enter();
+                        merge_and_send_docids(
+                            exact_word_docids,
+                            index.exact_word_docids.remap_types(),
+                            index,
+                            extractor_sender.docids::<ExactWordDocids>(),
+                            &indexing_context.must_stop_processing,
+                        )?;
+                    }
 
+                    {
+                        let span = tracing::trace_span!(target: "indexing::documents::merge", "word_position_docids");
+                        let _entered = span.enter();
+                        merge_and_send_docids(
+                            word_position_docids,
+                            index.word_position_docids.remap_types(),
+                            index,
+                            extractor_sender.docids::<WordPositionDocids>(),
+                            &indexing_context.must_stop_processing,
+                        )?;
+                    }
 
-
-
-                let WordDocidsCaches {
-                    word_docids,
-                    word_fid_docids,
-                    exact_word_docids,
-                    word_position_docids,
-                    fid_word_count_docids,
-                } = {
-                    let span = tracing::trace_span!(target: "indexing::documents::extract", "word_docids");
-                    let _entered = span.enter();
-
-                    WordDocidsExtractors::run_extraction(
-                        grenad_parameters,
-                        document_changes,
-                        indexing_context,
-                        &mut extractor_allocs,
-                        Step::ExtractingWords
-                    )?
-                };
-
-                {
-                    let span = tracing::trace_span!(target: "indexing::documents::merge", "word_docids");
-                    let _entered = span.enter();
-                    merge_and_send_docids(
-                        word_docids,
-                        index.word_docids.remap_types(),
-                        index,
-                        extractor_sender.docids::<WordDocids>(),
-                        &indexing_context.must_stop_processing,
-                    )?;
+                    {
+                        let span = tracing::trace_span!(target: "indexing::documents::merge", "fid_word_count_docids");
+                        let _entered = span.enter();
+                        merge_and_send_docids(
+                            fid_word_count_docids,
+                            index.field_id_word_count_docids.remap_types(),
+                            index,
+                            extractor_sender.docids::<FidWordCountDocids>(),
+                            &indexing_context.must_stop_processing,
+                        )?;
+                    }
                 }
 
-                {
-                    let span = tracing::trace_span!(target: "indexing::documents::merge", "word_fid_docids");
-                    let _entered = span.enter();
-                    merge_and_send_docids(
-                        word_fid_docids,
-                        index.word_fid_docids.remap_types(),
-                        index,
-                        extractor_sender.docids::<WordFidDocids>(),
-                        &indexing_context.must_stop_processing,
-                    )?;
+                // run the proximity extraction only if the precision is by word
+                // this works only if the settings didn't change during this transaction.
+                let proximity_precision = index.proximity_precision(&rtxn)?.unwrap_or_default();
+                if proximity_precision == ProximityPrecision::ByWord {
+                    let caches = {
+                        let span = tracing::trace_span!(target: "indexing::documents::extract", "word_pair_proximity_docids");
+                        let _entered = span.enter();
+
+                        <WordPairProximityDocidsExtractor as DocidsExtractor>::run_extraction(
+                            grenad_parameters,
+                            document_changes,
+                            indexing_context,
+                            &mut extractor_allocs,
+                            Step::ExtractingWordProximity,
+                        )?
+                    };
+
+                    {
+                        let span = tracing::trace_span!(target: "indexing::documents::merge", "word_pair_proximity_docids");
+                        let _entered = span.enter();
+
+                        merge_and_send_docids(
+                            caches,
+                            index.word_pair_proximity_docids.remap_types(),
+                            index,
+                            extractor_sender.docids::<WordPairProximityDocids>(),
+                            &indexing_context.must_stop_processing,
+                        )?;
+                    }
                 }
 
-                {
-                    let span = tracing::trace_span!(target: "indexing::documents::merge", "exact_word_docids");
-                    let _entered = span.enter();
-                    merge_and_send_docids(
-                        exact_word_docids,
-                        index.exact_word_docids.remap_types(),
-                        index,
-                        extractor_sender.docids::<ExactWordDocids>(),
-                        &indexing_context.must_stop_processing,
-                    )?;
-                }
+                'vectors: {
+                    if index_embeddings.is_empty() {
+                        break 'vectors;
+                    }
 
-                {
-                    let span = tracing::trace_span!(target: "indexing::documents::merge", "word_position_docids");
-                    let _entered = span.enter();
-                    merge_and_send_docids(
-                        word_position_docids,
-                        index.word_position_docids.remap_types(),
-                        index,
-                        extractor_sender.docids::<WordPositionDocids>(),
-                        &indexing_context.must_stop_processing,
-                    )?;
-                }
+                    let embedding_sender = extractor_sender.embeddings();
+                    let extractor = EmbeddingExtractor::new(embedders, embedding_sender, field_distribution, request_threads());
+                    let mut datastore = ThreadLocal::with_capacity(rayon::current_num_threads());
+                    {
+                        let span = tracing::trace_span!(target: "indexing::documents::extract", "vectors");
+                        let _entered = span.enter();
 
-                {
-                    let span = tracing::trace_span!(target: "indexing::documents::merge", "fid_word_count_docids");
-                    let _entered = span.enter();
-                    merge_and_send_docids(
-                        fid_word_count_docids,
-                        index.field_id_word_count_docids.remap_types(),
-                        index,
-                        extractor_sender.docids::<FidWordCountDocids>(),
-                        &indexing_context.must_stop_processing,
-                    )?;
-                }
-            }
+                        extract(
+                            document_changes,
+                            &extractor,
+                            indexing_context,
+                            &mut extractor_allocs,
+                            &datastore,
+                            Step::ExtractingEmbeddings,
+                        )?;
+                    }
+                    {
+                        let span = tracing::trace_span!(target: "indexing::documents::merge", "vectors");
+                        let _entered = span.enter();
 
-            // run the proximity extraction only if the precision is by word
-            // this works only if the settings didn't change during this transaction.
-            let proximity_precision = index.proximity_precision(&rtxn)?.unwrap_or_default();
-            if proximity_precision == ProximityPrecision::ByWord {
-                let caches = {
-                    let span = tracing::trace_span!(target: "indexing::documents::extract", "word_pair_proximity_docids");
-                    let _entered = span.enter();
-
-                    <WordPairProximityDocidsExtractor as DocidsExtractor>::run_extraction(
-                        grenad_parameters,
-                        document_changes,
-                        indexing_context,
-                        &mut extractor_allocs,
-                        Step::ExtractingWordProximity,
-                    )?
-                };
-
-                {
-                    let span = tracing::trace_span!(target: "indexing::documents::merge", "word_pair_proximity_docids");
-                    let _entered = span.enter();
-
-                    merge_and_send_docids(
-                        caches,
-                        index.word_pair_proximity_docids.remap_types(),
-                        index,
-                        extractor_sender.docids::<WordPairProximityDocids>(),
-                        &indexing_context.must_stop_processing,
-                    )?;
-                }
-            }
-
-            'vectors: {
-
-                let mut index_embeddings = index.embedding_configs(&rtxn)?;
-                if index_embeddings.is_empty() {
-                    break 'vectors;
-                }
-
-                let embedding_sender = extractor_sender.embeddings();
-                let extractor = EmbeddingExtractor::new(embedders, &embedding_sender, field_distribution, request_threads());
-                let mut datastore = ThreadLocal::with_capacity(rayon::current_num_threads());
-                {
-                    let span = tracing::trace_span!(target: "indexing::documents::extract", "vectors");
-                    let _entered = span.enter();
-
-                    extract(document_changes, &extractor, indexing_context, &mut extractor_allocs, &datastore, Step::ExtractingEmbeddings)?;
-                }
-                {
-                    let span = tracing::trace_span!(target: "indexing::documents::merge", "vectors");
-                    let _entered = span.enter();
-
-                    for config in &mut index_embeddings {
-                        'data: for data in datastore.iter_mut() {
-                            let data = &mut data.get_mut().0;
-                            let Some(deladd) = data.remove(&config.name) else { continue 'data; };
-                            deladd.apply_to(&mut config.user_provided);
+                        for config in &mut index_embeddings {
+                            'data: for data in datastore.iter_mut() {
+                                let data = &mut data.get_mut().0;
+                                let Some(deladd) = data.remove(&config.name) else { continue 'data; };
+                                deladd.apply_to(&mut config.user_provided);
+                            }
                         }
                     }
                 }
 
-                embedding_sender.finish(index_embeddings).unwrap();
-            }
+                'geo: {
+                    let Some(extractor) = GeoExtractor::new(&rtxn, index, grenad_parameters)? else {
+                        break 'geo;
+                    };
+                    let datastore = ThreadLocal::with_capacity(rayon::current_num_threads());
 
-            'geo: {
-                let Some(extractor) = GeoExtractor::new(&rtxn, index, grenad_parameters)? else {
-                    break 'geo;
-                };
-                let datastore = ThreadLocal::with_capacity(rayon::current_num_threads());
+                    {
+                        let span = tracing::trace_span!(target: "indexing::documents::extract", "geo");
+                        let _entered = span.enter();
 
-                {
-                    let span = tracing::trace_span!(target: "indexing::documents::extract", "geo");
-                    let _entered = span.enter();
+                        extract(
+                            document_changes,
+                            &extractor,
+                            indexing_context,
+                            &mut extractor_allocs,
+                            &datastore,
+                            Step::WritingGeoPoints
+                        )?;
+                    }
 
-                    extract(
-                        document_changes,
-                        &extractor,
-                        indexing_context,
-                        &mut extractor_allocs,
-                        &datastore,
-                        Step::WritingGeoPoints
+                    merge_and_send_rtree(
+                        datastore,
+                        &rtxn,
+                        index,
+                        extractor_sender.geo(),
+                        &indexing_context.must_stop_processing,
                     )?;
                 }
 
-                merge_and_send_rtree(
-                    datastore,
-                    &rtxn,
-                    index,
-                    extractor_sender.geo(),
-                    &indexing_context.must_stop_processing,
-                )?;
-            }
+                (indexing_context.send_progress)(Progress::from_step(Step::WritingToDatabase));
 
-            (indexing_context.send_progress)(Progress::from_step(Step::WritingToDatabase));
+                finished_extraction.store(true, std::sync::atomic::Ordering::Relaxed);
 
-            finished_extraction.store(true, std::sync::atomic::Ordering::Relaxed);
-
-            Result::Ok(facet_field_ids_delta)
+                Result::Ok((facet_field_ids_delta, index_embeddings))
+            }).unwrap()
         })?;
 
         let global_fields_ids_map = GlobalFieldsIdsMap::new(&new_fields_ids_map);
 
         let vector_arroy = index.vector_arroy;
-        let mut rng = rand::rngs::StdRng::seed_from_u64(42);
         let indexer_span = tracing::Span::current();
         let arroy_writers: Result<HashMap<_, _>> = embedders
             .inner_as_ref()
@@ -384,7 +407,11 @@ where
             })
             .collect();
 
+        // Used by by the ArroySetVector to copy the embedding into an
+        // aligned memory area, required by arroy to accept a new vector.
+        let mut aligned_embedding = Vec::new();
         let mut arroy_writers = arroy_writers?;
+
         {
             let span = tracing::trace_span!(target: "indexing::write_db", "all");
             let _entered = span.enter();
@@ -392,110 +419,93 @@ where
             let span = tracing::trace_span!(target: "indexing::write_db", "post_merge");
             let mut _entered_post_merge = None;
 
-            for operation in writer_receiver {
+            while let Some(action) = writer_receiver.recv_action() {
                 if _entered_post_merge.is_none()
                     && finished_extraction.load(std::sync::atomic::Ordering::Relaxed)
                 {
                     _entered_post_merge = Some(span.enter());
                 }
-                match operation {
-                    WriterOperation::DbOperation(db_operation) => {
-                        let database = db_operation.database(index);
-                        let database_name = db_operation.database_name();
-                        match db_operation.entry() {
-                            EntryOperation::Delete(e) => match database.delete(wtxn, e.entry()) {
-                                Ok(false) => unreachable!("We tried to delete an unknown key"),
-                                Ok(_) => (),
-                                Err(error) => {
-                                    return Err(Error::InternalError(
-                                        InternalError::StoreDeletion {
-                                            database_name,
-                                            key: e.entry().to_owned(),
-                                            error,
-                                        },
-                                    ));
-                                }
-                            },
-                            EntryOperation::Write(e) => {
-                                if let Err(error) = database.put(wtxn, e.key(), e.value()) {
-                                    return Err(Error::InternalError(InternalError::StorePut {
-                                        database_name,
-                                        key: e.key().to_owned(),
-                                        value_length: e.value().len(),
-                                        error,
-                                    }));
-                                }
-                            }
+
+                match action {
+                    ReceiverAction::WakeUp => (),
+                    ReceiverAction::LargeEntry(LargeEntry { database, key, value }) => {
+                        let database_name = database.database_name();
+                        let database = database.database(index);
+                        if let Err(error) = database.put(wtxn, &key, &value) {
+                            return Err(Error::InternalError(InternalError::StorePut {
+                                database_name,
+                                key: bstr::BString::from(&key[..]),
+                                value_length: value.len(),
+                                error,
+                            }));
                         }
                     }
-                    WriterOperation::ArroyOperation(arroy_operation) => match arroy_operation {
-                        ArroyOperation::DeleteVectors { docid } => {
-                            for (
-                                _embedder_index,
-                                (_embedder_name, _embedder, writer, dimensions),
-                            ) in &mut arroy_writers
-                            {
-                                let dimensions = *dimensions;
-                                writer.del_items(wtxn, dimensions, docid)?;
-                            }
+                    ReceiverAction::LargeVectors(large_vectors) => {
+                        let LargeVectors { docid, embedder_id, .. } = large_vectors;
+                        let (_, _, writer, dimensions) =
+                            arroy_writers.get(&embedder_id).expect("requested a missing embedder");
+                        let mut embeddings = Embeddings::new(*dimensions);
+                        for embedding in large_vectors.read_embeddings(*dimensions) {
+                            embeddings.push(embedding.to_vec()).unwrap();
                         }
-                        ArroyOperation::SetVectors {
-                            docid,
-                            embedder_id,
-                            embeddings: raw_embeddings,
-                        } => {
-                            let (_, _, writer, dimensions) = arroy_writers
-                                .get(&embedder_id)
-                                .expect("requested a missing embedder");
-
-                            let mut embeddings = Embeddings::new(*dimensions);
-                            for embedding in raw_embeddings {
-                                embeddings.append(embedding).unwrap();
-                            }
-
-                            writer.del_items(wtxn, *dimensions, docid)?;
-                            writer.add_items(wtxn, docid, &embeddings)?;
-                        }
-                        ArroyOperation::SetVector { docid, embedder_id, embedding } => {
-                            let (_, _, writer, dimensions) = arroy_writers
-                                .get(&embedder_id)
-                                .expect("requested a missing embedder");
-                            writer.del_items(wtxn, *dimensions, docid)?;
-                            writer.add_item(wtxn, docid, &embedding)?;
-                        }
-                        ArroyOperation::Finish { configs } => {
-                            let span = tracing::trace_span!(target: "indexing::vectors", parent: &indexer_span, "build");
-                            let _entered = span.enter();
-
-                            (indexing_context.send_progress)(Progress::from_step(
-                                Step::WritingEmbeddingsToDatabase,
-                            ));
-
-                            for (
-                                _embedder_index,
-                                (_embedder_name, _embedder, writer, dimensions),
-                            ) in &mut arroy_writers
-                            {
-                                let dimensions = *dimensions;
-                                writer.build_and_quantize(
-                                    wtxn,
-                                    &mut rng,
-                                    dimensions,
-                                    false,
-                                    &indexing_context.must_stop_processing,
-                                )?;
-                            }
-
-                            index.put_embedding_configs(wtxn, configs)?;
-                        }
-                    },
+                        writer.del_items(wtxn, *dimensions, docid)?;
+                        writer.add_items(wtxn, docid, &embeddings)?;
+                    }
                 }
+
+                // Every time the is a message in the channel we search
+                // for new entries in the BBQueue buffers.
+                write_from_bbqueue(
+                    &mut writer_receiver,
+                    index,
+                    wtxn,
+                    &arroy_writers,
+                    &mut aligned_embedding,
+                )?;
             }
+
+            // Once the extractor/writer channel is closed
+            // we must process the remaining BBQueue messages.
+            write_from_bbqueue(
+                &mut writer_receiver,
+                index,
+                wtxn,
+                &arroy_writers,
+                &mut aligned_embedding,
+            )?;
         }
 
         (indexing_context.send_progress)(Progress::from_step(Step::WaitingForExtractors));
 
-        let facet_field_ids_delta = extractor_handle.join().unwrap()?;
+        let (facet_field_ids_delta, index_embeddings) = extractor_handle.join().unwrap()?;
+
+        'vectors: {
+            let span =
+                tracing::trace_span!(target: "indexing::vectors", parent: &indexer_span, "build");
+            let _entered = span.enter();
+
+            if index_embeddings.is_empty() {
+                break 'vectors;
+            }
+
+            (indexing_context.send_progress)(Progress::from_step(
+                Step::WritingEmbeddingsToDatabase,
+            ));
+
+            let mut rng = rand::rngs::StdRng::seed_from_u64(42);
+            for (_index, (_embedder_name, _embedder, writer, dimensions)) in &mut arroy_writers {
+                let dimensions = *dimensions;
+                writer.build_and_quantize(
+                    wtxn,
+                    &mut rng,
+                    dimensions,
+                    false,
+                    &indexing_context.must_stop_processing,
+                )?;
+            }
+
+            index.put_embedding_configs(wtxn, index_embeddings)?;
+        }
 
         (indexing_context.send_progress)(Progress::from_step(Step::PostProcessingFacets));
 
@@ -537,6 +547,72 @@ where
     Ok(())
 }
 
+/// A function dedicated to manage all the available BBQueue frames.
+///
+/// It reads all the available frames, do the corresponding database operations
+/// and stops when no frame are available.
+fn write_from_bbqueue(
+    writer_receiver: &mut WriterBbqueueReceiver<'_>,
+    index: &Index,
+    wtxn: &mut RwTxn<'_>,
+    arroy_writers: &HashMap<u8, (&str, &crate::vector::Embedder, ArroyWrapper, usize)>,
+    aligned_embedding: &mut Vec<f32>,
+) -> crate::Result<()> {
+    while let Some(frame_with_header) = writer_receiver.recv_frame() {
+        match frame_with_header.header() {
+            EntryHeader::DbOperation(operation) => {
+                let database_name = operation.database.database_name();
+                let database = operation.database.database(index);
+                let frame = frame_with_header.frame();
+                match operation.key_value(frame) {
+                    (key, Some(value)) => {
+                        if let Err(error) = database.put(wtxn, key, value) {
+                            return Err(Error::InternalError(InternalError::StorePut {
+                                database_name,
+                                key: key.into(),
+                                value_length: value.len(),
+                                error,
+                            }));
+                        }
+                    }
+                    (key, None) => match database.delete(wtxn, key) {
+                        Ok(false) => {
+                            unreachable!("We tried to delete an unknown key: {key:?}")
+                        }
+                        Ok(_) => (),
+                        Err(error) => {
+                            return Err(Error::InternalError(InternalError::StoreDeletion {
+                                database_name,
+                                key: key.into(),
+                                error,
+                            }));
+                        }
+                    },
+                }
+            }
+            EntryHeader::ArroyDeleteVector(ArroyDeleteVector { docid }) => {
+                for (_index, (_name, _embedder, writer, dimensions)) in arroy_writers {
+                    let dimensions = *dimensions;
+                    writer.del_items(wtxn, dimensions, docid)?;
+                }
+            }
+            EntryHeader::ArroySetVectors(asvs) => {
+                let ArroySetVectors { docid, embedder_id, .. } = asvs;
+                let frame = frame_with_header.frame();
+                let (_, _, writer, dimensions) =
+                    arroy_writers.get(&embedder_id).expect("requested a missing embedder");
+                let mut embeddings = Embeddings::new(*dimensions);
+                let all_embeddings = asvs.read_all_embeddings_into_vec(frame, aligned_embedding);
+                embeddings.append(all_embeddings.to_vec()).unwrap();
+                writer.del_items(wtxn, *dimensions, docid)?;
+                writer.add_items(wtxn, docid, &embeddings)?;
+            }
+        }
+    }
+
+    Ok(())
+}
+
 #[tracing::instrument(level = "trace", skip_all, target = "indexing::prefix")]
 fn compute_prefix_database(
     index: &Index,

crates/milli/src/update/new/merger.rs

@@ -9,8 +9,8 @@ use roaring::RoaringBitmap;
 
 use super::channel::*;
 use super::extract::{
-    merge_caches, transpose_and_freeze_caches, BalancedCaches, DelAddRoaringBitmap, FacetKind,
-    GeoExtractorData,
+    merge_caches_sorted, transpose_and_freeze_caches, BalancedCaches, DelAddRoaringBitmap,
+    FacetKind, GeoExtractorData,
 };
 use crate::{CboRoaringBitmapCodec, FieldId, GeoPoint, Index, InternalError, Result};
 
@@ -19,7 +19,7 @@ pub fn merge_and_send_rtree<'extractor, MSP>(
     datastore: impl IntoIterator<Item = RefCell<GeoExtractorData<'extractor>>>,
     rtxn: &RoTxn,
     index: &Index,
-    geo_sender: GeoSender<'_>,
+    geo_sender: GeoSender<'_, '_>,
     must_stop_processing: &MSP,
 ) -> Result<()>
 where
@@ -34,7 +34,7 @@ where
         }
 
         let mut frozen = data.into_inner().freeze()?;
-        for result in frozen.iter_and_clear_removed() {
+        for result in frozen.iter_and_clear_removed()? {
             let extracted_geo_point = result?;
             let removed = rtree.remove(&GeoPoint::from(extracted_geo_point));
             debug_assert!(removed.is_some());
@@ -42,7 +42,7 @@ where
             debug_assert!(removed);
         }
 
-        for result in frozen.iter_and_clear_inserted() {
+        for result in frozen.iter_and_clear_inserted()? {
             let extracted_geo_point = result?;
             rtree.insert(GeoPoint::from(extracted_geo_point));
             let inserted = faceted.insert(extracted_geo_point.docid);
@@ -56,38 +56,37 @@ where
 
     let rtree_mmap = unsafe { Mmap::map(&file)? };
     geo_sender.set_rtree(rtree_mmap).unwrap();
-    geo_sender.set_geo_faceted(&faceted).unwrap();
+    geo_sender.set_geo_faceted(&faceted)?;
 
     Ok(())
 }
 
 #[tracing::instrument(level = "trace", skip_all, target = "indexing::merge")]
-pub fn merge_and_send_docids<'extractor, MSP>(
+pub fn merge_and_send_docids<'extractor, MSP, D>(
     mut caches: Vec<BalancedCaches<'extractor>>,
     database: Database<Bytes, Bytes>,
     index: &Index,
-    docids_sender: impl DocidsSender + Sync,
+    docids_sender: WordDocidsSender<D>,
     must_stop_processing: &MSP,
 ) -> Result<()>
 where
     MSP: Fn() -> bool + Sync,
+    D: DatabaseType + Sync,
 {
     transpose_and_freeze_caches(&mut caches)?.into_par_iter().try_for_each(|frozen| {
         let rtxn = index.read_txn()?;
-        let mut buffer = Vec::new();
         if must_stop_processing() {
             return Err(InternalError::AbortedIndexation.into());
         }
-        merge_caches(frozen, |key, DelAddRoaringBitmap { del, add }| {
+        merge_caches_sorted(frozen, |key, DelAddRoaringBitmap { del, add }| {
             let current = database.get(&rtxn, key)?;
             match merge_cbo_bitmaps(current, del, add)? {
                 Operation::Write(bitmap) => {
-                    let value = cbo_bitmap_serialize_into_vec(&bitmap, &mut buffer);
-                    docids_sender.write(key, value).unwrap();
+                    docids_sender.write(key, &bitmap)?;
                     Ok(())
                 }
                 Operation::Delete => {
-                    docids_sender.delete(key).unwrap();
+                    docids_sender.delete(key)?;
                     Ok(())
                 }
                 Operation::Ignore => Ok(()),
@@ -101,26 +100,24 @@ pub fn merge_and_send_facet_docids<'extractor>(
     mut caches: Vec<BalancedCaches<'extractor>>,
     database: FacetDatabases,
     index: &Index,
-    docids_sender: impl DocidsSender + Sync,
+    docids_sender: FacetDocidsSender,
 ) -> Result<FacetFieldIdsDelta> {
     transpose_and_freeze_caches(&mut caches)?
         .into_par_iter()
         .map(|frozen| {
             let mut facet_field_ids_delta = FacetFieldIdsDelta::default();
             let rtxn = index.read_txn()?;
-            let mut buffer = Vec::new();
-            merge_caches(frozen, |key, DelAddRoaringBitmap { del, add }| {
+            merge_caches_sorted(frozen, |key, DelAddRoaringBitmap { del, add }| {
                 let current = database.get_cbo_roaring_bytes_value(&rtxn, key)?;
                 match merge_cbo_bitmaps(current, del, add)? {
                     Operation::Write(bitmap) => {
                         facet_field_ids_delta.register_from_key(key);
-                        let value = cbo_bitmap_serialize_into_vec(&bitmap, &mut buffer);
-                        docids_sender.write(key, value).unwrap();
+                        docids_sender.write(key, &bitmap)?;
                         Ok(())
                     }
                     Operation::Delete => {
                         facet_field_ids_delta.register_from_key(key);
-                        docids_sender.delete(key).unwrap();
+                        docids_sender.delete(key)?;
                         Ok(())
                     }
                     Operation::Ignore => Ok(()),
@@ -252,10 +249,3 @@ fn merge_cbo_bitmaps(
         }
     }
 }
-
-/// TODO Return the slice directly from the serialize_into method
-fn cbo_bitmap_serialize_into_vec<'b>(bitmap: &RoaringBitmap, buffer: &'b mut Vec<u8>) -> &'b [u8] {
-    buffer.clear();
-    CboRoaringBitmapCodec::serialize_into(bitmap, buffer);
-    buffer.as_slice()
-}

crates/milli/src/update/new/ref_cell_ext.rs

@@ -5,6 +5,7 @@ pub trait RefCellExt<T: ?Sized> {
         &self,
     ) -> std::result::Result<RefMut<'_, T>, std::cell::BorrowMutError>;
 
+    #[track_caller]
     fn borrow_mut_or_yield(&self) -> RefMut<'_, T> {
         self.try_borrow_mut_or_yield().unwrap()
     }

crates/milli/src/update/new/steps.rs

@@ -11,8 +11,8 @@ pub enum Step {
     ExtractingEmbeddings,
     WritingGeoPoints,
     WritingToDatabase,
-    WritingEmbeddingsToDatabase,
     WaitingForExtractors,
+    WritingEmbeddingsToDatabase,
     PostProcessingFacets,
     PostProcessingWords,
     Finalizing,
@@ -29,8 +29,8 @@ impl Step {
             Step::ExtractingEmbeddings => "extracting embeddings",
             Step::WritingGeoPoints => "writing geo points",
             Step::WritingToDatabase => "writing to database",
-            Step::WritingEmbeddingsToDatabase => "writing embeddings to database",
             Step::WaitingForExtractors => "waiting for extractors",
+            Step::WritingEmbeddingsToDatabase => "writing embeddings to database",
             Step::PostProcessingFacets => "post-processing facets",
             Step::PostProcessingWords => "post-processing words",
             Step::Finalizing => "finalizing",

crates/milli/src/update/new/word_fst_builder.rs

@@ -1,4 +1,4 @@
-use std::collections::HashSet;
+use std::collections::BTreeSet;
 use std::io::BufWriter;
 
 use fst::{Set, SetBuilder, Streamer};
@@ -75,8 +75,8 @@ pub struct PrefixData {
 
 #[derive(Debug)]
 pub struct PrefixDelta {
-    pub modified: HashSet<Prefix>,
-    pub deleted: HashSet<Prefix>,
+    pub modified: BTreeSet<Prefix>,
+    pub deleted: BTreeSet<Prefix>,
 }
 
 struct PrefixFstBuilder {
@@ -86,7 +86,7 @@ struct PrefixFstBuilder {
     prefix_fst_builders: Vec<SetBuilder<Vec<u8>>>,
     current_prefix: Vec<Prefix>,
     current_prefix_count: Vec<usize>,
-    modified_prefixes: HashSet<Prefix>,
+    modified_prefixes: BTreeSet<Prefix>,
     current_prefix_is_modified: Vec<bool>,
 }
 
@@ -110,7 +110,7 @@ impl PrefixFstBuilder {
             prefix_fst_builders,
             current_prefix: vec![Prefix::new(); max_prefix_length],
             current_prefix_count: vec![0; max_prefix_length],
-            modified_prefixes: HashSet::new(),
+            modified_prefixes: BTreeSet::new(),
             current_prefix_is_modified: vec![false; max_prefix_length],
         })
     }
@@ -180,7 +180,7 @@ impl PrefixFstBuilder {
         let prefix_fst_mmap = unsafe { Mmap::map(&prefix_fst_file)? };
         let new_prefix_fst = Set::new(&prefix_fst_mmap)?;
         let old_prefix_fst = index.words_prefixes_fst(rtxn)?;
-        let mut deleted_prefixes = HashSet::new();
+        let mut deleted_prefixes = BTreeSet::new();
         {
             let mut deleted_prefixes_stream = old_prefix_fst.op().add(&new_prefix_fst).difference();
             while let Some(prefix) = deleted_prefixes_stream.next() {

crates/milli/src/update/new/words_prefix_docids.rs

@@ -1,5 +1,5 @@
 use std::cell::RefCell;
-use std::collections::HashSet;
+use std::collections::BTreeSet;
 use std::io::{BufReader, BufWriter, Read, Seek, Write};
 
 use hashbrown::HashMap;
@@ -37,8 +37,8 @@ impl WordPrefixDocids {
     fn execute(
         self,
         wtxn: &mut heed::RwTxn,
-        prefix_to_compute: &HashSet<Prefix>,
-        prefix_to_delete: &HashSet<Prefix>,
+        prefix_to_compute: &BTreeSet<Prefix>,
+        prefix_to_delete: &BTreeSet<Prefix>,
     ) -> Result<()> {
         delete_prefixes(wtxn, &self.prefix_database, prefix_to_delete)?;
         self.recompute_modified_prefixes(wtxn, prefix_to_compute)
@@ -48,7 +48,7 @@ impl WordPrefixDocids {
     fn recompute_modified_prefixes(
         &self,
         wtxn: &mut RwTxn,
-        prefixes: &HashSet<Prefix>,
+        prefixes: &BTreeSet<Prefix>,
     ) -> Result<()> {
         // We fetch the docids associated to the newly added word prefix fst only.
         // And collect the CboRoaringBitmaps pointers in an HashMap.
@@ -76,7 +76,7 @@ impl WordPrefixDocids {
                 .union()?;
 
             buffer.clear();
-            CboRoaringBitmapCodec::serialize_into(&output, buffer);
+            CboRoaringBitmapCodec::serialize_into_vec(&output, buffer);
             index.push(PrefixEntry { prefix, serialized_length: buffer.len() });
             file.write_all(buffer)
         })?;
@@ -127,7 +127,7 @@ impl<'a, 'rtxn> FrozenPrefixBitmaps<'a, 'rtxn> {
     pub fn from_prefixes(
         database: Database<Bytes, CboRoaringBitmapCodec>,
         rtxn: &'rtxn RoTxn,
-        prefixes: &'a HashSet<Prefix>,
+        prefixes: &'a BTreeSet<Prefix>,
     ) -> heed::Result<Self> {
         let database = database.remap_data_type::<Bytes>();
 
@@ -173,8 +173,8 @@ impl WordPrefixIntegerDocids {
     fn execute(
         self,
         wtxn: &mut heed::RwTxn,
-        prefix_to_compute: &HashSet<Prefix>,
-        prefix_to_delete: &HashSet<Prefix>,
+        prefix_to_compute: &BTreeSet<Prefix>,
+        prefix_to_delete: &BTreeSet<Prefix>,
     ) -> Result<()> {
         delete_prefixes(wtxn, &self.prefix_database, prefix_to_delete)?;
         self.recompute_modified_prefixes(wtxn, prefix_to_compute)
@@ -184,7 +184,7 @@ impl WordPrefixIntegerDocids {
     fn recompute_modified_prefixes(
         &self,
         wtxn: &mut RwTxn,
-        prefixes: &HashSet<Prefix>,
+        prefixes: &BTreeSet<Prefix>,
     ) -> Result<()> {
         // We fetch the docids associated to the newly added word prefix fst only.
         // And collect the CboRoaringBitmaps pointers in an HashMap.
@@ -211,7 +211,7 @@ impl WordPrefixIntegerDocids {
                     .union()?;
 
                 buffer.clear();
-                CboRoaringBitmapCodec::serialize_into(&output, buffer);
+                CboRoaringBitmapCodec::serialize_into_vec(&output, buffer);
                 index.push(PrefixIntegerEntry { prefix, pos, serialized_length: buffer.len() });
                 file.write_all(buffer)?;
             }
@@ -262,7 +262,7 @@ impl<'a, 'rtxn> FrozenPrefixIntegerBitmaps<'a, 'rtxn> {
     pub fn from_prefixes(
         database: Database<Bytes, CboRoaringBitmapCodec>,
         rtxn: &'rtxn RoTxn,
-        prefixes: &'a HashSet<Prefix>,
+        prefixes: &'a BTreeSet<Prefix>,
     ) -> heed::Result<Self> {
         let database = database.remap_data_type::<Bytes>();
 
@@ -291,7 +291,7 @@ unsafe impl<'a, 'rtxn> Sync for FrozenPrefixIntegerBitmaps<'a, 'rtxn> {}
 fn delete_prefixes(
     wtxn: &mut RwTxn,
     prefix_database: &Database<Bytes, CboRoaringBitmapCodec>,
-    prefixes: &HashSet<Prefix>,
+    prefixes: &BTreeSet<Prefix>,
 ) -> Result<()> {
     // We remove all the entries that are no more required in this word prefix docids database.
     for prefix in prefixes {
@@ -309,8 +309,8 @@ fn delete_prefixes(
 pub fn compute_word_prefix_docids(
     wtxn: &mut RwTxn,
     index: &Index,
-    prefix_to_compute: &HashSet<Prefix>,
-    prefix_to_delete: &HashSet<Prefix>,
+    prefix_to_compute: &BTreeSet<Prefix>,
+    prefix_to_delete: &BTreeSet<Prefix>,
     grenad_parameters: GrenadParameters,
 ) -> Result<()> {
     WordPrefixDocids::new(
@@ -325,8 +325,8 @@ pub fn compute_word_prefix_docids(
 pub fn compute_exact_word_prefix_docids(
     wtxn: &mut RwTxn,
     index: &Index,
-    prefix_to_compute: &HashSet<Prefix>,
-    prefix_to_delete: &HashSet<Prefix>,
+    prefix_to_compute: &BTreeSet<Prefix>,
+    prefix_to_delete: &BTreeSet<Prefix>,
     grenad_parameters: GrenadParameters,
 ) -> Result<()> {
     WordPrefixDocids::new(
@@ -341,8 +341,8 @@ pub fn compute_exact_word_prefix_docids(
 pub fn compute_word_prefix_fid_docids(
     wtxn: &mut RwTxn,
     index: &Index,
-    prefix_to_compute: &HashSet<Prefix>,
-    prefix_to_delete: &HashSet<Prefix>,
+    prefix_to_compute: &BTreeSet<Prefix>,
+    prefix_to_delete: &BTreeSet<Prefix>,
     grenad_parameters: GrenadParameters,
 ) -> Result<()> {
     WordPrefixIntegerDocids::new(
@@ -357,8 +357,8 @@ pub fn compute_word_prefix_fid_docids(
 pub fn compute_word_prefix_position_docids(
     wtxn: &mut RwTxn,
     index: &Index,
-    prefix_to_compute: &HashSet<Prefix>,
-    prefix_to_delete: &HashSet<Prefix>,
+    prefix_to_compute: &BTreeSet<Prefix>,
+    prefix_to_delete: &BTreeSet<Prefix>,
     grenad_parameters: GrenadParameters,
 ) -> Result<()> {
     WordPrefixIntegerDocids::new(

crates/milli/src/vector/mod.rs

@@ -475,7 +475,7 @@ impl<F> Embeddings<F> {
         Ok(())
     }
 
-    /// Append a flat vector of embeddings a the end of the embeddings.
+    /// Append a flat vector of embeddings at the end of the embeddings.
     ///
     /// If `embeddings.len() % self.dimension != 0`, then the append operation fails.
     pub fn append(&mut self, mut embeddings: Vec<F>) -> Result<(), Vec<F>> {

crates/milli/tests/search/facet_distribution.rs

@@ -64,6 +64,7 @@ fn test_facet_distribution_with_no_facet_values() {
     indexer::index(
         &mut wtxn,
         &index,
+        &milli::ThreadPoolNoAbortBuilder::new().build().unwrap(),
         config.grenad_parameters(),
         &db_fields_ids_map,
         new_fields_ids_map,

crates/milli/tests/search/mod.rs

@@ -101,6 +101,7 @@ pub fn setup_search_index_with_criteria(criteria: &[Criterion]) -> Index {
     indexer::index(
         &mut wtxn,
         &index,
+        &milli::ThreadPoolNoAbortBuilder::new().build().unwrap(),
         config.grenad_parameters(),
         &db_fields_ids_map,
         new_fields_ids_map,

crates/milli/tests/search/query_criteria.rs

@@ -333,6 +333,7 @@ fn criteria_ascdesc() {
     indexer::index(
         &mut wtxn,
         &index,
+        &milli::ThreadPoolNoAbortBuilder::new().build().unwrap(),
         config.grenad_parameters(),
         &db_fields_ids_map,
         new_fields_ids_map,

crates/milli/tests/search/typo_tolerance.rs

@@ -142,6 +142,7 @@ fn test_typo_disabled_on_word() {
     indexer::index(
         &mut wtxn,
         &index,
+        &milli::ThreadPoolNoAbortBuilder::new().build().unwrap(),
         config.grenad_parameters(),
         &db_fields_ids_map,
         new_fields_ids_map,

crates/xtask/src/bench/mod.rs

@@ -82,6 +82,10 @@ pub struct BenchDeriveArgs {
     /// Reason for the benchmark invocation
     #[arg(short, long)]
     reason: Option<String>,
+
+    /// The maximum time in seconds we allow for fetching the task queue before timing out.
+    #[arg(long, default_value_t = 60)]
+    tasks_queue_timeout_secs: u64,
 }
 
 pub fn run(args: BenchDeriveArgs) -> anyhow::Result<()> {
@@ -127,7 +131,7 @@ pub fn run(args: BenchDeriveArgs) -> anyhow::Result<()> {
     let meili_client = Client::new(
         Some("http://127.0.0.1:7700".into()),
         args.master_key.as_deref(),
-        Some(std::time::Duration::from_secs(60)),
+        Some(std::time::Duration::from_secs(args.tasks_queue_timeout_secs)),
     )?;
 
     // enter runtime

crates/xtask/src/main.rs

@@ -16,6 +16,7 @@ struct ListFeaturesDeriveArgs {
 #[command(author, version, about, long_about)]
 #[command(name = "cargo xtask")]
 #[command(bin_name = "cargo xtask")]
+#[allow(clippy::large_enum_variant)] // please, that's enough...
 enum Command {
     ListFeatures(ListFeaturesDeriveArgs),
     Bench(BenchDeriveArgs),

workloads/hackernews-add-new-documents.json

@@ -0,0 +1,105 @@
+{
+  "name": "hackernews.add_new_documents",
+  "run_count": 3,
+  "extra_cli_args": [],
+  "assets": {
+      "hackernews-01.ndjson": {
+        "local_location": null,
+        "remote_location": "https://milli-benchmarks.fra1.digitaloceanspaces.com/bench/datasets/hackernews/modification/01.ndjson",
+        "sha256": "cd3627b86c064d865b6754848ed0e73ef1d8142752a25e5f0765c3a1296dd3ae"
+      },
+      "hackernews-02.ndjson": {
+        "local_location": null,
+        "remote_location": "https://milli-benchmarks.fra1.digitaloceanspaces.com/bench/datasets/hackernews/modification/02.ndjson",
+        "sha256": "5d533b83bcf992201dace88b4d0c0be8b4df5225c6c4b763582d986844bcc23b"
+      },
+      "hackernews-03.ndjson": {
+        "local_location": null,
+        "remote_location": "https://milli-benchmarks.fra1.digitaloceanspaces.com/bench/datasets/hackernews/modification/03.ndjson",
+        "sha256": "f5f351a0d04a8a83643ace12cafa2b7ec8ca8cb7d46fd268e5126492a6c66f2a"
+      },
+      "hackernews-04.ndjson": {
+        "local_location": null,
+        "remote_location": "https://milli-benchmarks.fra1.digitaloceanspaces.com/bench/datasets/hackernews/modification/04.ndjson",
+        "sha256": "ac1915ee7ce53a6718548c255a6cc59969784b2570745dc5b739f714beda291a"
+      },
+      "hackernews-05.ndjson": {
+        "local_location": null,
+        "remote_location": "https://milli-benchmarks.fra1.digitaloceanspaces.com/bench/datasets/hackernews/modification/05.ndjson",
+        "sha256": "be31d5632602f798e62d1c10c83bdfda2b4deaa068477eacde05fdd247572b82"
+      }
+  },
+  "precommands": [
+    {
+      "route": "indexes/movies/settings",
+      "method": "PATCH",
+      "body": {
+        "inline": {
+          "displayedAttributes": [
+            "title",
+            "by",
+            "score",
+            "time",
+            "text"
+          ],
+          "searchableAttributes": [
+            "title",
+            "text"
+          ],
+          "filterableAttributes": [
+            "by",
+            "kids",
+            "parent"
+          ],
+          "sortableAttributes": [
+            "score",
+            "time"
+          ]
+        }
+      },
+      "synchronous": "WaitForTask"
+    },
+    {
+      "route": "indexes/movies/documents",
+      "method": "POST",
+      "body": {
+        "asset": "hackernews-01.ndjson"
+      },
+      "synchronous": "WaitForResponse"
+    },
+    {
+      "route": "indexes/movies/documents",
+      "method": "POST",
+      "body": {
+        "asset": "hackernews-02.ndjson"
+      },
+      "synchronous": "WaitForResponse"
+    },
+    {
+      "route": "indexes/movies/documents",
+      "method": "POST",
+      "body": {
+        "asset": "hackernews-03.ndjson"
+      },
+      "synchronous": "WaitForResponse"
+    },
+    {
+      "route": "indexes/movies/documents",
+      "method": "POST",
+      "body": {
+        "asset": "hackernews-04.ndjson"
+      },
+      "synchronous": "WaitForTask"
+    }
+  ],
+  "commands": [
+      {
+        "route": "indexes/movies/documents",
+        "method": "POST",
+        "body": {
+          "asset": "hackernews-05.ndjson"
+        },
+        "synchronous": "WaitForTask"
+      }
+  ]
+}

workloads/hackernews-modify-facet-numbers.json

@@ -0,0 +1,111 @@
+{
+  "name": "hackernews.modify_facet_numbers",
+  "run_count": 3,
+  "extra_cli_args": [],
+  "assets": {
+    "hackernews-01.ndjson": {
+      "local_location": null,
+      "remote_location": "https://milli-benchmarks.fra1.digitaloceanspaces.com/bench/datasets/hackernews/modification/01.ndjson",
+      "sha256": "cd3627b86c064d865b6754848ed0e73ef1d8142752a25e5f0765c3a1296dd3ae"
+    },
+    "hackernews-02.ndjson": {
+      "local_location": null,
+      "remote_location": "https://milli-benchmarks.fra1.digitaloceanspaces.com/bench/datasets/hackernews/modification/02.ndjson",
+      "sha256": "5d533b83bcf992201dace88b4d0c0be8b4df5225c6c4b763582d986844bcc23b"
+    },
+    "hackernews-03.ndjson": {
+      "local_location": null,
+      "remote_location": "https://milli-benchmarks.fra1.digitaloceanspaces.com/bench/datasets/hackernews/modification/03.ndjson",
+      "sha256": "f5f351a0d04a8a83643ace12cafa2b7ec8ca8cb7d46fd268e5126492a6c66f2a"
+    },
+    "hackernews-04.ndjson": {
+      "local_location": null,
+      "remote_location": "https://milli-benchmarks.fra1.digitaloceanspaces.com/bench/datasets/hackernews/modification/04.ndjson",
+      "sha256": "ac1915ee7ce53a6718548c255a6cc59969784b2570745dc5b739f714beda291a"
+    },
+    "hackernews-05.ndjson": {
+      "local_location": null,
+      "remote_location": "https://milli-benchmarks.fra1.digitaloceanspaces.com/bench/datasets/hackernews/modification/05.ndjson",
+      "sha256": "be31d5632602f798e62d1c10c83bdfda2b4deaa068477eacde05fdd247572b82"
+    },
+    "hackernews-02-modified-filters.ndjson": {
+      "local_location": null,
+      "remote_location": "https://milli-benchmarks.fra1.digitaloceanspaces.com/bench/datasets/hackernews/modification/02-modified-filters.ndjson",
+      "sha256": "7272cbfd41110d32d7fe168424a0000f07589bfe40f664652b34f4f20aaf3802"
+    }
+  },
+  "precommands": [
+    {
+      "route": "indexes/movies/settings",
+      "method": "PATCH",
+      "body": {
+        "inline": {
+          "displayedAttributes": [
+            "title",
+            "by",
+            "score",
+            "time",
+            "text"
+          ],
+          "searchableAttributes": [
+            "title",
+            "text"
+          ],
+          "filterableAttributes": [
+            "by",
+            "kids",
+            "parent"
+          ],
+          "sortableAttributes": [
+            "score",
+            "time"
+          ]
+        }
+      },
+      "synchronous": "WaitForTask"
+    },
+    {
+      "route": "indexes/movies/documents",
+      "method": "POST",
+      "body": {
+        "asset": "hackernews-01.ndjson"
+      },
+      "synchronous": "WaitForResponse"
+    },
+    {
+      "route": "indexes/movies/documents",
+      "method": "POST",
+      "body": {
+        "asset": "hackernews-02.ndjson"
+      },
+      "synchronous": "WaitForResponse"
+    },
+    {
+      "route": "indexes/movies/documents",
+      "method": "POST",
+      "body": {
+        "asset": "hackernews-03.ndjson"
+      },
+      "synchronous": "WaitForResponse"
+    },
+    {
+      "route": "indexes/movies/documents",
+      "method": "POST",
+      "body": {
+        "asset": "hackernews-04.ndjson"
+      },
+      "synchronous": "WaitForTask"
+    }
+  ],
+  "commands": [
+      {
+        "route": "indexes/movies/documents",
+        "method": "POST",
+        "body": {
+          "asset": "hackernews-02-modified-filters.ndjson"
+        },
+        "synchronous": "WaitForTask"
+      }
+  ]
+}
+  

workloads/hackernews-modify-facet-strings.json

@@ -0,0 +1,111 @@
+{
+  "name": "hackernews.modify_facet_strings",
+  "run_count": 3,
+  "extra_cli_args": [],
+  "assets": {
+    "hackernews-01.ndjson": {
+      "local_location": null,
+      "remote_location": "https://milli-benchmarks.fra1.digitaloceanspaces.com/bench/datasets/hackernews/modification/01.ndjson",
+      "sha256": "cd3627b86c064d865b6754848ed0e73ef1d8142752a25e5f0765c3a1296dd3ae"
+    },
+    "hackernews-02.ndjson": {
+      "local_location": null,
+      "remote_location": "https://milli-benchmarks.fra1.digitaloceanspaces.com/bench/datasets/hackernews/modification/02.ndjson",
+      "sha256": "5d533b83bcf992201dace88b4d0c0be8b4df5225c6c4b763582d986844bcc23b"
+    },
+    "hackernews-03.ndjson": {
+      "local_location": null,
+      "remote_location": "https://milli-benchmarks.fra1.digitaloceanspaces.com/bench/datasets/hackernews/modification/03.ndjson",
+      "sha256": "f5f351a0d04a8a83643ace12cafa2b7ec8ca8cb7d46fd268e5126492a6c66f2a"
+    },
+    "hackernews-04.ndjson": {
+      "local_location": null,
+      "remote_location": "https://milli-benchmarks.fra1.digitaloceanspaces.com/bench/datasets/hackernews/modification/04.ndjson",
+      "sha256": "ac1915ee7ce53a6718548c255a6cc59969784b2570745dc5b739f714beda291a"
+    },
+    "hackernews-05.ndjson": {
+      "local_location": null,
+      "remote_location": "https://milli-benchmarks.fra1.digitaloceanspaces.com/bench/datasets/hackernews/modification/05.ndjson",
+      "sha256": "be31d5632602f798e62d1c10c83bdfda2b4deaa068477eacde05fdd247572b82"
+    },
+    "hackernews-01-modified-filters.ndjson": {
+      "local_location": null,
+      "remote_location": "https://milli-benchmarks.fra1.digitaloceanspaces.com/bench/datasets/hackernews/modification/01-modified-filters.ndjson",
+      "sha256": "b80c245ce1b1df80b9b38800f677f3bd11947ebc62716fb108269d50e796c35c"
+    }
+  },
+  "precommands": [
+    {
+      "route": "indexes/movies/settings",
+      "method": "PATCH",
+      "body": {
+        "inline": {
+          "displayedAttributes": [
+            "title",
+            "by",
+            "score",
+            "time",
+            "text"
+          ],
+          "searchableAttributes": [
+            "title",
+            "text"
+          ],
+          "filterableAttributes": [
+            "by",
+            "kids",
+            "parent"
+          ],
+          "sortableAttributes": [
+            "score",
+            "time"
+          ]
+        }
+      },
+      "synchronous": "WaitForTask"
+    },
+    {
+      "route": "indexes/movies/documents",
+      "method": "POST",
+      "body": {
+        "asset": "hackernews-01.ndjson"
+      },
+      "synchronous": "WaitForResponse"
+    },
+    {
+      "route": "indexes/movies/documents",
+      "method": "POST",
+      "body": {
+        "asset": "hackernews-02.ndjson"
+      },
+      "synchronous": "WaitForResponse"
+    },
+    {
+      "route": "indexes/movies/documents",
+      "method": "POST",
+      "body": {
+        "asset": "hackernews-03.ndjson"
+      },
+      "synchronous": "WaitForResponse"
+    },
+    {
+      "route": "indexes/movies/documents",
+      "method": "POST",
+      "body": {
+        "asset": "hackernews-04.ndjson"
+      },
+      "synchronous": "WaitForTask"
+    }
+  ],
+  "commands": [
+      {
+        "route": "indexes/movies/documents",
+        "method": "POST",
+        "body": {
+          "asset": "hackernews-01-modified-filters.ndjson"
+        },
+        "synchronous": "WaitForTask"
+      }
+  ]
+}
+ 

workloads/hackernews-modify-searchables.json

@@ -0,0 +1,123 @@
+{
+  "name": "hackernews.modify_searchables",
+  "run_count": 3,
+  "extra_cli_args": [],
+  "assets": {
+    "hackernews-01.ndjson": {
+      "local_location": null,
+      "remote_location": "https://milli-benchmarks.fra1.digitaloceanspaces.com/bench/datasets/hackernews/modification/01.ndjson",
+      "sha256": "cd3627b86c064d865b6754848ed0e73ef1d8142752a25e5f0765c3a1296dd3ae"
+    },
+    "hackernews-02.ndjson": {
+      "local_location": null,
+      "remote_location": "https://milli-benchmarks.fra1.digitaloceanspaces.com/bench/datasets/hackernews/modification/02.ndjson",
+      "sha256": "5d533b83bcf992201dace88b4d0c0be8b4df5225c6c4b763582d986844bcc23b"
+    },
+    "hackernews-03.ndjson": {
+      "local_location": null,
+      "remote_location": "https://milli-benchmarks.fra1.digitaloceanspaces.com/bench/datasets/hackernews/modification/03.ndjson",
+      "sha256": "f5f351a0d04a8a83643ace12cafa2b7ec8ca8cb7d46fd268e5126492a6c66f2a"
+    },
+    "hackernews-04.ndjson": {
+      "local_location": null,
+      "remote_location": "https://milli-benchmarks.fra1.digitaloceanspaces.com/bench/datasets/hackernews/modification/04.ndjson",
+      "sha256": "ac1915ee7ce53a6718548c255a6cc59969784b2570745dc5b739f714beda291a"
+    },
+    "hackernews-05.ndjson": {
+      "local_location": null,
+      "remote_location": "https://milli-benchmarks.fra1.digitaloceanspaces.com/bench/datasets/hackernews/modification/05.ndjson",
+      "sha256": "be31d5632602f798e62d1c10c83bdfda2b4deaa068477eacde05fdd247572b82"
+    },
+    "hackernews-01-modified-searchables.ndjson": {
+      "local_location": null,
+      "remote_location": "https://milli-benchmarks.fra1.digitaloceanspaces.com/bench/datasets/hackernews/modification/01-modified-searchables.ndjson",
+      "sha256": "e5c08710c6af70031ac7212e0ba242c72ef29c8d4e1fce66c789544641452a7c"
+    },
+    "hackernews-02-modified-searchables.ndjson": {
+      "local_location": null,
+      "remote_location": "https://milli-benchmarks.fra1.digitaloceanspaces.com/bench/datasets/hackernews/modification/02-modified-searchables.ndjson",
+      "sha256": "098b029851117087b1e26ccb7ac408eda9bba54c3008213a2880d6fab607346e"
+    }
+  },
+  "precommands": [
+    {
+      "route": "indexes/movies/settings",
+      "method": "PATCH",
+      "body": {
+        "inline": {
+          "displayedAttributes": [
+            "title",
+            "by",
+            "score",
+            "time",
+            "text"
+          ],
+          "searchableAttributes": [
+            "title",
+            "text"
+          ],
+          "filterableAttributes": [
+            "by",
+            "kids",
+            "parent"
+          ],
+          "sortableAttributes": [
+            "score",
+            "time"
+          ]
+        }
+      },
+      "synchronous": "WaitForTask"
+    },
+    {
+      "route": "indexes/movies/documents",
+      "method": "POST",
+      "body": {
+        "asset": "hackernews-01.ndjson"
+      },
+      "synchronous": "WaitForResponse"
+    },
+    {
+      "route": "indexes/movies/documents",
+      "method": "POST",
+      "body": {
+        "asset": "hackernews-02.ndjson"
+      },
+      "synchronous": "WaitForResponse"
+    },
+    {
+      "route": "indexes/movies/documents",
+      "method": "POST",
+      "body": {
+        "asset": "hackernews-03.ndjson"
+      },
+      "synchronous": "WaitForResponse"
+    },
+    {
+      "route": "indexes/movies/documents",
+      "method": "POST",
+      "body": {
+        "asset": "hackernews-04.ndjson"
+      },
+      "synchronous": "WaitForTask"
+    }
+  ],
+  "commands": [
+      {
+        "route": "indexes/movies/documents",
+        "method": "POST",
+        "body": {
+          "asset": "hackernews-01-modified-searchables.ndjson"
+        },
+        "synchronous": "WaitForTask"
+      },
+      {
+        "route": "indexes/movies/documents",
+        "method": "POST",
+        "body": {
+          "asset": "hackernews-02-modified-searchables.ndjson"
+        },
+        "synchronous": "WaitForTask"
+      }
+  ]
+}