rebasing composite aggregations

Closes #2849

Co-authored-by: Paul Masurel <paul.masurel@datadoghq.com>

.claude/skills/rationalize-deps/SKILL.md

@@ -0,0 +1,125 @@
+---
+name: rationalize-deps
+description: Analyze Cargo.toml dependencies and attempt to remove unused features to reduce compile times and binary size
+---
+
+# Rationalize Dependencies
+
+This skill analyzes Cargo.toml dependencies to identify and remove unused features.
+
+## Overview
+
+Many crates enable features by default that may not be needed. This skill:
+1. Identifies dependencies with default features enabled
+2. Tests if `default-features = false` works
+3. Identifies which specific features are actually needed
+4. Verifies compilation after changes
+
+## Step 1: Identify the target
+
+Ask the user which crate(s) to analyze:
+- A specific crate name (e.g., "tokio", "serde")
+- A specific workspace member (e.g., "quickwit-search")
+- "all" to scan the entire workspace
+
+## Step 2: Analyze current dependencies
+
+For the workspace Cargo.toml (`quickwit/Cargo.toml`), list dependencies that:
+- Do NOT have `default-features = false`
+- Have default features that might be unnecessary
+
+Run: `cargo tree -p <crate> -f "{p} {f}" --edges features` to see what features are actually used.
+
+## Step 3: For each candidate dependency
+
+### 3a: Check the crate's default features
+
+Look up the crate on crates.io or check its Cargo.toml to understand:
+- What features are enabled by default
+- What each feature provides
+
+Use: `cargo metadata --format-version=1 | jq '.packages[] | select(.name == "<crate>") | .features'`
+
+### 3b: Try disabling default features
+
+Modify the dependency in `quickwit/Cargo.toml`:
+
+From:
+```toml
+some-crate = { version = "1.0" }
+```
+
+To:
+```toml
+some-crate = { version = "1.0", default-features = false }
+```
+
+### 3c: Run cargo check
+
+Run: `cargo check --workspace` (or target specific packages for faster feedback)
+
+If compilation fails:
+1. Read the error messages to identify which features are needed
+2. Add only the required features explicitly:
+   ```toml
+   some-crate = { version = "1.0", default-features = false, features = ["needed-feature"] }
+   ```
+3. Re-run cargo check
+
+### 3d: Binary search for minimal features
+
+If there are many default features, use binary search:
+1. Start with no features
+2. If it fails, add half the default features
+3. Continue until you find the minimal set
+
+## Step 4: Document findings
+
+For each dependency analyzed, report:
+- Original configuration
+- New configuration (if changed)
+- Features that were removed
+- Any features that are required
+
+## Step 5: Verify full build
+
+After all changes, run:
+```bash
+cargo check --workspace --all-targets
+cargo test --workspace --no-run
+```
+
+## Common Patterns
+
+### Serde
+Often only needs `derive`:
+```toml
+serde = { version = "1.0", default-features = false, features = ["derive", "std"] }
+```
+
+### Tokio
+Identify which runtime features are actually used:
+```toml
+tokio = { version = "1.0", default-features = false, features = ["rt-multi-thread", "macros", "sync"] }
+```
+
+### Reqwest
+Often doesn't need all TLS backends:
+```toml
+reqwest = { version = "0.11", default-features = false, features = ["rustls-tls", "json"] }
+```
+
+## Rollback
+
+If changes cause issues:
+```bash
+git checkout quickwit/Cargo.toml
+cargo check --workspace
+```
+
+## Tips
+
+- Start with large crates that have many default features (tokio, reqwest, hyper)
+- Use `cargo bloat --crates` to identify large dependencies
+- Check `cargo tree -d` for duplicate dependencies that might indicate feature conflicts
+- Some features are needed only for tests - consider using `[dev-dependencies]` features

.claude/skills/simple-pr/SKILL.md

@@ -0,0 +1,60 @@
+---
+name: simple-pr
+description: Create a simple PR from staged changes with an auto-generated commit message
+disable-model-invocation: true
+---
+
+# Simple PR
+
+Follow these steps to create a simple PR from staged changes:
+
+## Step 1: Check workspace state
+
+Run: `git status`
+
+Verify that all changes have been staged (no unstaged changes). If there are unstaged changes, abort and ask the user to stage their changes first with `git add`.
+
+Also verify that we are on the `main` branch. If not, abort and ask the user to switch to main first.
+
+## Step 2: Ensure main is up to date
+
+Run: `git pull origin main`
+
+This ensures we're working from the latest code.
+
+## Step 3: Review staged changes
+
+Run: `git diff --cached`
+
+Review the staged changes to understand what the PR will contain.
+
+## Step 4: Generate commit message
+
+Based on the staged changes, generate a concise commit message (1-2 sentences) that describes the "why" rather than the "what".
+
+Display the proposed commit message to the user and ask for confirmation before proceeding.
+
+## Step 5: Create a new branch
+
+Get the git username: `git config user.name | tr ' ' '-' | tr '[:upper:]' '[:lower:]'`
+
+Create a short, descriptive branch name based on the changes (e.g., `fix-typo-in-readme`, `add-retry-logic`, `update-deps`).
+
+Create and checkout the branch: `git checkout -b {username}/{short-descriptive-name}`
+
+## Step 6: Commit changes
+
+Commit with the message from step 3:
+```
+git commit -m "{commit-message}"
+```
+
+## Step 7: Push and open a PR
+
+Push the branch and open a PR:
+```
+git push -u origin {branch-name}
+gh pr create --title "{commit-message-title}" --body "{longer-description-if-needed}"
+```
+
+Report the PR URL to the user when complete.

.github/workflows/coverage.yml

@@ -15,11 +15,11 @@ jobs:
     steps:
       - uses: actions/checkout@v4
       - name: Install Rust
-        run: rustup toolchain install nightly-2024-07-01 --profile minimal --component llvm-tools-preview
+        run: rustup toolchain install nightly-2025-12-01 --profile minimal --component llvm-tools-preview
       - uses: Swatinem/rust-cache@v2
       - uses: taiki-e/install-action@cargo-llvm-cov
       - name: Generate code coverage
-        run: cargo +nightly-2024-07-01 llvm-cov --all-features --workspace --doctests --lcov --output-path lcov.info
+        run: cargo +nightly-2025-12-01 llvm-cov --all-features --workspace --doctests --lcov --output-path lcov.info
       - name: Upload coverage to Codecov
         uses: codecov/codecov-action@v3
         continue-on-error: true

.github/workflows/test.yml

@@ -39,11 +39,11 @@ jobs:
 
     - name: Check Formatting
       run: cargo +nightly fmt --all -- --check
-    
+
     - name: Check Stable Compilation
       run: cargo build --all-features
 
-    
+
     - name: Check Bench Compilation
       run: cargo +nightly bench --no-run --profile=dev --all-features
 
@@ -59,10 +59,10 @@ jobs:
 
     strategy:
       matrix:
-        features: [
-            { label: "all", flags: "mmap,stopwords,lz4-compression,zstd-compression,failpoints" },
-            { label: "quickwit", flags: "mmap,quickwit,failpoints" }
-        ]
+        features:
+          - { label: "all", flags: "mmap,stopwords,lz4-compression,zstd-compression,failpoints,stemmer" }
+          - { label: "quickwit", flags: "mmap,quickwit,failpoints" }
+          - { label: "none", flags: "" }
 
     name: test-${{ matrix.features.label}}
 
@@ -80,7 +80,21 @@ jobs:
     - uses: Swatinem/rust-cache@v2
 
     - name: Run tests
-      run: cargo +stable nextest run --features ${{ matrix.features.flags }} --verbose --workspace
+      run: |
+        # if matrix.feature.flags is empty then run on --lib to avoid compiling examples
+        # (as most of them rely on mmap) otherwise run all
+        if [ -z "${{ matrix.features.flags }}" ]; then
+          cargo +stable nextest run --lib --no-default-features --verbose --workspace
+        else
+          cargo +stable nextest run --features ${{ matrix.features.flags }} --no-default-features --verbose --workspace
+        fi
 
     - name: Run doctests
-      run: cargo +stable test --doc --features ${{ matrix.features.flags }} --verbose --workspace
+      run: |
+        # if matrix.feature.flags is empty then run on --lib to avoid compiling examples
+        # (as most of them rely on mmap) otherwise run all
+        if [ -z "${{ matrix.features.flags }}" ]; then
+          echo "no doctest for no feature flag"
+        else
+          cargo +stable test --doc --features ${{ matrix.features.flags }} --verbose --workspace
+        fi

CHANGELOG.md

@@ -78,7 +78,7 @@ This will slightly increase space and access time. [#2439](https://github.com/qu
 
 - **Store DateTime as nanoseconds in doc store** DateTime in the doc store was truncated to microseconds previously. This removes this truncation, while still keeping backwards compatibility. [#2486](https://github.com/quickwit-oss/tantivy/pull/2486)(@PSeitz)
 
-- **Performace/Memory**
+- **Performance/Memory**
     - lift clauses in LogicalAst for optimized ast during execution [#2449](https://github.com/quickwit-oss/tantivy/pull/2449)(@PSeitz)
     - Use Vec instead of BTreeMap to back OwnedValue object [#2364](https://github.com/quickwit-oss/tantivy/pull/2364)(@fulmicoton)
     - Replace TantivyDocument with CompactDoc. CompactDoc is much smaller and provides similar performance. [#2402](https://github.com/quickwit-oss/tantivy/pull/2402)(@PSeitz)

Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "tantivy"
-version = "0.25.0"
+version = "0.26.0"
 authors = ["Paul Masurel <paul.masurel@gmail.com>"]
 license = "MIT"
 categories = ["database-implementations", "data-structures"]
@@ -15,7 +15,7 @@ rust-version = "1.85"
 exclude = ["benches/*.json", "benches/*.txt"]
 
 [dependencies]
-oneshot = "0.1.7"
+oneshot = "0.1.13"
 base64 = "0.22.0"
 byteorder = "1.4.3"
 crc32fast = "1.3.2"
@@ -27,7 +27,7 @@ regex = { version = "1.5.5", default-features = false, features = [
 aho-corasick = "1.0"
 tantivy-fst = "0.5"
 memmap2 = { version = "0.9.0", optional = true }
-lz4_flex = { version = "0.11", default-features = false, optional = true }
+lz4_flex = { version = "0.12", default-features = false, optional = true }
 zstd = { version = "0.13", optional = true, default-features = false }
 tempfile = { version = "3.12.0", optional = true }
 log = "0.4.16"
@@ -37,9 +37,9 @@ fs4 = { version = "0.13.1", optional = true }
 levenshtein_automata = "0.2.1"
 uuid = { version = "1.0.0", features = ["v4", "serde"] }
 crossbeam-channel = "0.5.4"
-rust-stemmers = "1.2.0"
+rust-stemmers = { version = "1.2.0", optional = true }
 downcast-rs = "2.0.1"
-bitpacking = { version = "0.9.2", default-features = false, features = [
+bitpacking = { version = "0.9.3", default-features = false, features = [
     "bitpacker4x",
 ] }
 census = "0.4.2"
@@ -47,10 +47,10 @@ rustc-hash = "2.0.0"
 thiserror = "2.0.1"
 htmlescape = "0.3.1"
 fail = { version = "0.5.0", optional = true }
-time = { version = "0.3.35", features = ["serde-well-known"] }
+time = { version = "0.3.47", features = ["serde-well-known"] }
 smallvec = "1.8.0"
 rayon = "1.5.2"
-lru = "0.12.0"
+lru = "0.16.3"
 fastdivide = "0.4.0"
 itertools = "0.14.0"
 measure_time = "0.9.0"
@@ -64,8 +64,8 @@ query-grammar = { version = "0.25.0", path = "./query-grammar", package = "tanti
 tantivy-bitpacker = { version = "0.9", path = "./bitpacker" }
 common = { version = "0.10", path = "./common/", package = "tantivy-common" }
 tokenizer-api = { version = "0.6", path = "./tokenizer-api", package = "tantivy-tokenizer-api" }
-sketches-ddsketch = { version = "0.3.0", features = ["use_serde"] }
-hyperloglogplus = { version = "0.4.1", features = ["const-loop"] }
+sketches-ddsketch = { git = "https://github.com/quickwit-oss/rust-sketches-ddsketch.git", rev = "555caf1", features = ["use_serde"] }
+datasketches = "0.2.0"
 futures-util = { version = "0.3.28", optional = true }
 futures-channel = { version = "0.3.28", optional = true }
 fnv = "1.0.7"
@@ -75,18 +75,18 @@ typetag = "0.2.21"
 winapi = "0.3.9"
 
 [dev-dependencies]
-binggan = "0.14.0"
-rand = "0.8.5"
+binggan = "0.14.2"
+rand = "0.9"
 maplit = "1.0.2"
 matches = "0.1.9"
 pretty_assertions = "1.2.1"
-proptest = "1.0.0"
+proptest = "1.7.0"
 test-log = "0.2.10"
 futures = "0.3.21"
 paste = "1.0.11"
 more-asserts = "0.3.1"
-rand_distr = "0.4.3"
-time = { version = "0.3.10", features = ["serde-well-known", "macros"] }
+rand_distr = "0.5"
+time = { version = "0.3.47", features = ["serde-well-known", "macros"] }
 postcard = { version = "1.0.4", features = [
     "use-std",
 ], default-features = false }
@@ -113,7 +113,8 @@ debug-assertions = true
 overflow-checks = true
 
 [features]
-default = ["mmap", "stopwords", "lz4-compression", "columnar-zstd-compression"]
+default = ["mmap", "stopwords", "lz4-compression", "columnar-zstd-compression", "stemmer"]
+stemmer = ["rust-stemmers"]
 mmap = ["fs4", "tempfile", "memmap2"]
 stopwords = []
 
@@ -173,6 +174,30 @@ harness = false
 name = "exists_json"
 harness = false
 
+[[bench]]
+name = "range_query"
+harness = false
+
 [[bench]]
 name = "and_or_queries"
 harness = false
+
+[[bench]]
+name = "range_queries"
+harness = false
+
+[[bench]]
+name = "bool_queries_with_range"
+harness = false
+
+[[bench]]
+name = "str_search_and_get"
+harness = false
+
+[[bench]]
+name = "merge_segments"
+harness = false
+
+[[bench]]
+name = "regex_all_terms"
+harness = false

README.md

@@ -123,6 +123,7 @@ You can also find other bindings on [GitHub](https://github.com/search?q=tantivy
 - [seshat](https://github.com/matrix-org/seshat/): A matrix message database/indexer
 - [tantiny](https://github.com/baygeldin/tantiny): Tiny full-text search for Ruby
 - [lnx](https://github.com/lnx-search/lnx): adaptable, typo tolerant search engine with a REST API
+- [Bichon](https://github.com/rustmailer/bichon): A lightweight, high-performance Rust email archiver with WebUI
 - and [more](https://github.com/search?q=tantivy)!
 
 ### On average, how much faster is Tantivy compared to Lucene?

TODO.txt

@@ -10,7 +10,7 @@ rename FastFieldReaders::open to load
 remove fast field reader
 
 find a way to unify the two DateTime.
-readd type check in the filter wrapper
+re-add type check in the filter wrapper
 
 add unit test on columnar list columns.
 

benches/agg_bench.rs

@@ -1,7 +1,8 @@
 use binggan::plugins::PeakMemAllocPlugin;
 use binggan::{black_box, InputGroup, PeakMemAlloc, INSTRUMENTED_SYSTEM};
-use rand::prelude::SliceRandom;
+use rand::distr::weighted::WeightedIndex;
 use rand::rngs::StdRng;
+use rand::seq::IndexedRandom;
 use rand::{Rng, SeedableRng};
 use rand_distr::Distribution;
 use serde_json::json;
@@ -9,7 +10,7 @@ use tantivy::aggregation::agg_req::Aggregations;
 use tantivy::aggregation::AggregationCollector;
 use tantivy::query::{AllQuery, TermQuery};
 use tantivy::schema::{IndexRecordOption, Schema, TextFieldIndexing, FAST, STRING};
-use tantivy::{doc, Index, Term};
+use tantivy::{doc, DateTime, Index, Term};
 
 #[global_allocator]
 pub static GLOBAL: &PeakMemAlloc<std::alloc::System> = &INSTRUMENTED_SYSTEM;
@@ -53,25 +54,39 @@ fn bench_agg(mut group: InputGroup<Index>) {
     register!(group, stats_f64);
     register!(group, extendedstats_f64);
     register!(group, percentiles_f64);
-    register!(group, terms_few);
-    register!(group, terms_many);
+    register!(group, terms_7);
+    register!(group, terms_all_unique);
+    register!(group, terms_150_000);
     register!(group, terms_many_top_1000);
     register!(group, terms_many_order_by_term);
     register!(group, terms_many_with_top_hits);
+    register!(group, terms_all_unique_with_avg_sub_agg);
     register!(group, terms_many_with_avg_sub_agg);
+    register!(group, terms_status_with_avg_sub_agg);
+    register!(group, terms_status_with_histogram);
+    register!(group, terms_zipf_1000);
+    register!(group, terms_zipf_1000_with_histogram);
+    register!(group, terms_zipf_1000_with_avg_sub_agg);
+
     register!(group, terms_many_json_mixed_type_with_avg_sub_agg);
 
+    register!(group, composite_term_many_page_1000);
+    register!(group, composite_term_many_page_1000_with_avg_sub_agg);
+    register!(group, composite_term_few);
+    register!(group, composite_histogram);
+    register!(group, composite_histogram_calendar);
+
     register!(group, cardinality_agg);
-    register!(group, terms_few_with_cardinality_agg);
+    register!(group, terms_status_with_cardinality_agg);
 
     register!(group, range_agg);
     register!(group, range_agg_with_avg_sub_agg);
-    register!(group, range_agg_with_term_agg_few);
+    register!(group, range_agg_with_term_agg_status);
     register!(group, range_agg_with_term_agg_many);
     register!(group, histogram);
     register!(group, histogram_hard_bounds);
     register!(group, histogram_with_avg_sub_agg);
-    register!(group, histogram_with_term_agg_few);
+    register!(group, histogram_with_term_agg_status);
     register!(group, avg_and_range_with_avg_sub_agg);
 
     // Filter aggregation benchmarks
@@ -130,12 +145,12 @@ fn extendedstats_f64(index: &Index) {
 }
 fn percentiles_f64(index: &Index) {
     let agg_req = json!({
-      "mypercentiles": {
-        "percentiles": {
-          "field": "score_f64",
-          "percents": [ 95, 99, 99.9 ]
+        "mypercentiles": {
+            "percentiles": {
+                "field": "score_f64",
+                "percents": [ 95, 99, 99.9 ]
+            }
         }
-      }
     });
     execute_agg(index, agg_req);
 }
@@ -150,10 +165,10 @@ fn cardinality_agg(index: &Index) {
     });
     execute_agg(index, agg_req);
 }
-fn terms_few_with_cardinality_agg(index: &Index) {
+fn terms_status_with_cardinality_agg(index: &Index) {
     let agg_req = json!({
         "my_texts": {
-            "terms": { "field": "text_few_terms" },
+            "terms": { "field": "text_few_terms_status" },
             "aggs": {
                 "cardinality": {
                     "cardinality": {
@@ -166,13 +181,20 @@ fn terms_few_with_cardinality_agg(index: &Index) {
     execute_agg(index, agg_req);
 }
 
-fn terms_few(index: &Index) {
+fn terms_7(index: &Index) {
     let agg_req = json!({
-        "my_texts": { "terms": { "field": "text_few_terms" } },
+        "my_texts": { "terms": { "field": "text_few_terms_status" } },
     });
     execute_agg(index, agg_req);
 }
-fn terms_many(index: &Index) {
+fn terms_all_unique(index: &Index) {
+    let agg_req = json!({
+        "my_texts": { "terms": { "field": "text_all_unique_terms" } },
+    });
+    execute_agg(index, agg_req);
+}
+
+fn terms_150_000(index: &Index) {
     let agg_req = json!({
         "my_texts": { "terms": { "field": "text_many_terms" } },
     });
@@ -220,6 +242,72 @@ fn terms_many_with_avg_sub_agg(index: &Index) {
     });
     execute_agg(index, agg_req);
 }
+fn terms_all_unique_with_avg_sub_agg(index: &Index) {
+    let agg_req = json!({
+        "my_texts": {
+            "terms": { "field": "text_all_unique_terms" },
+            "aggs": {
+                "average_f64": { "avg": { "field": "score_f64" } }
+            }
+        },
+    });
+    execute_agg(index, agg_req);
+}
+fn terms_status_with_histogram(index: &Index) {
+    let agg_req = json!({
+        "my_texts": {
+            "terms": { "field": "text_few_terms_status" },
+            "aggs": {
+                "histo": {"histogram": { "field": "score_f64", "interval": 10 }}
+            }
+        }
+    });
+    execute_agg(index, agg_req);
+}
+
+fn terms_zipf_1000_with_histogram(index: &Index) {
+    let agg_req = json!({
+        "my_texts": {
+            "terms": { "field": "text_1000_terms_zipf" },
+            "aggs": {
+                "histo": {"histogram": { "field": "score_f64", "interval": 10 }}
+            }
+        }
+    });
+    execute_agg(index, agg_req);
+}
+
+fn terms_status_with_avg_sub_agg(index: &Index) {
+    let agg_req = json!({
+        "my_texts": {
+            "terms": { "field": "text_few_terms_status" },
+            "aggs": {
+                "average_f64": { "avg": { "field": "score_f64" } }
+            }
+        },
+    });
+    execute_agg(index, agg_req);
+}
+
+fn terms_zipf_1000_with_avg_sub_agg(index: &Index) {
+    let agg_req = json!({
+        "my_texts": {
+            "terms": { "field": "text_1000_terms_zipf" },
+            "aggs": {
+                "average_f64": { "avg": { "field": "score_f64" } }
+            }
+        },
+    });
+    execute_agg(index, agg_req);
+}
+
+fn terms_zipf_1000(index: &Index) {
+    let agg_req = json!({
+        "my_texts": { "terms": { "field": "text_1000_terms_zipf" } },
+    });
+    execute_agg(index, agg_req);
+}
+
 fn terms_many_json_mixed_type_with_avg_sub_agg(index: &Index) {
     let agg_req = json!({
         "my_texts": {
@@ -232,6 +320,75 @@ fn terms_many_json_mixed_type_with_avg_sub_agg(index: &Index) {
     execute_agg(index, agg_req);
 }
 
+fn composite_term_few(index: &Index) {
+    let agg_req = json!({
+        "my_ctf": {
+            "composite": {
+                "sources": [
+                    { "text_few_terms": { "terms": { "field": "text_few_terms" } } }
+                ],
+                "size": 1000
+            }
+        },
+    });
+    execute_agg(index, agg_req);
+}
+fn composite_term_many_page_1000(index: &Index) {
+    let agg_req = json!({
+        "my_ctmp1000": {
+            "composite": {
+                "sources": [
+                    { "text_many_terms": { "terms": { "field": "text_many_terms" } } }
+                ],
+                "size": 1000
+            }
+        },
+    });
+    execute_agg(index, agg_req);
+}
+fn composite_term_many_page_1000_with_avg_sub_agg(index: &Index) {
+    let agg_req = json!({
+        "my_ctmp1000wasa": {
+            "composite": {
+                "sources": [
+                    { "text_many_terms": { "terms": { "field": "text_many_terms" } } }
+                ],
+                "size": 1000,
+            },
+            "aggs": {
+                "average_f64": { "avg": { "field": "score_f64" } }
+            }
+        },
+    });
+    execute_agg(index, agg_req);
+}
+fn composite_histogram(index: &Index) {
+    let agg_req = json!({
+        "my_ch": {
+            "composite": {
+                "sources": [
+                    { "f64_histogram": { "histogram": { "field": "score_f64", "interval": 1 } } }
+                ],
+                "size": 1000
+            }
+        },
+    });
+    execute_agg(index, agg_req);
+}
+fn composite_histogram_calendar(index: &Index) {
+    let agg_req = json!({
+        "my_chc": {
+            "composite": {
+                "sources": [
+                    { "time_histogram": { "date_histogram": { "field": "timestamp", "calendar_interval": "month" } } }
+                ],
+                "size": 1000
+            }
+        },
+    });
+    execute_agg(index, agg_req);
+}
+
 fn execute_agg(index: &Index, agg_req: serde_json::Value) {
     let agg_req: Aggregations = serde_json::from_value(agg_req).unwrap();
     let collector = get_collector(agg_req);
@@ -275,7 +432,7 @@ fn range_agg_with_avg_sub_agg(index: &Index) {
     execute_agg(index, agg_req);
 }
 
-fn range_agg_with_term_agg_few(index: &Index) {
+fn range_agg_with_term_agg_status(index: &Index) {
     let agg_req = json!({
         "rangef64": {
             "range": {
@@ -290,7 +447,7 @@ fn range_agg_with_term_agg_few(index: &Index) {
                 ]
             },
             "aggs": {
-                "my_texts": { "terms": { "field": "text_few_terms" } },
+                "my_texts": { "terms": { "field": "text_few_terms_status" } },
             }
         },
     });
@@ -346,12 +503,12 @@ fn histogram_with_avg_sub_agg(index: &Index) {
     });
     execute_agg(index, agg_req);
 }
-fn histogram_with_term_agg_few(index: &Index) {
+fn histogram_with_term_agg_status(index: &Index) {
     let agg_req = json!({
         "rangef64": {
             "histogram": { "field": "score_f64", "interval": 10 },
             "aggs": {
-                "my_texts": { "terms": { "field": "text_few_terms" } }
+                "my_texts": { "terms": { "field": "text_few_terms_status" } }
             }
         }
     });
@@ -396,6 +553,13 @@ fn get_collector(agg_req: Aggregations) -> AggregationCollector {
 }
 
 fn get_test_index_bench(cardinality: Cardinality) -> tantivy::Result<Index> {
+    // Flag to use existing index
+    let reuse_index = std::env::var("REUSE_AGG_BENCH_INDEX").is_ok();
+    if reuse_index && std::path::Path::new("agg_bench").exists() {
+        return Index::open_in_dir("agg_bench");
+    }
+    // crreate dir
+    std::fs::create_dir_all("agg_bench")?;
     let mut schema_builder = Schema::builder();
     let text_fieldtype = tantivy::schema::TextOptions::default()
         .set_indexing_options(
@@ -404,20 +568,50 @@ fn get_test_index_bench(cardinality: Cardinality) -> tantivy::Result<Index> {
         .set_stored();
     let text_field = schema_builder.add_text_field("text", text_fieldtype);
     let json_field = schema_builder.add_json_field("json", FAST);
+    let text_field_all_unique_terms =
+        schema_builder.add_text_field("text_all_unique_terms", STRING | FAST);
     let text_field_many_terms = schema_builder.add_text_field("text_many_terms", STRING | FAST);
     let text_field_few_terms = schema_builder.add_text_field("text_few_terms", STRING | FAST);
+    let text_field_few_terms_status =
+        schema_builder.add_text_field("text_few_terms_status", STRING | FAST);
+    let text_field_1000_terms_zipf =
+        schema_builder.add_text_field("text_1000_terms_zipf", STRING | FAST);
     let score_fieldtype = tantivy::schema::NumericOptions::default().set_fast();
     let score_field = schema_builder.add_u64_field("score", score_fieldtype.clone());
     let score_field_f64 = schema_builder.add_f64_field("score_f64", score_fieldtype.clone());
     let score_field_i64 = schema_builder.add_i64_field("score_i64", score_fieldtype);
-    let index = Index::create_from_tempdir(schema_builder.build())?;
-    let few_terms_data = ["INFO", "ERROR", "WARN", "DEBUG"];
+    let date_field = schema_builder.add_date_field("timestamp", FAST);
+    // use tmp dir
+    let index = if reuse_index {
+        Index::create_in_dir("agg_bench", schema_builder.build())?
+    } else {
+        Index::create_from_tempdir(schema_builder.build())?
+    };
+    // Approximate log proportions
+    let status_field_data = [
+        ("INFO", 8000),
+        ("ERROR", 300),
+        ("WARN", 1200),
+        ("DEBUG", 500),
+        ("OK", 500),
+        ("CRITICAL", 20),
+        ("EMERGENCY", 1),
+    ];
+    let log_level_distribution =
+        WeightedIndex::new(status_field_data.iter().map(|item| item.1)).unwrap();
 
+    let few_terms_data = ["INFO", "ERROR", "WARN", "DEBUG"];
     let lg_norm = rand_distr::LogNormal::new(2.996f64, 0.979f64).unwrap();
 
     let many_terms_data = (0..150_000)
         .map(|num| format!("author{num}"))
         .collect::<Vec<_>>();
+
+    // Prepare 1000 unique terms sampled using a Zipf distribution.
+    // Exponent ~1.1 approximates top-20 terms covering around ~20%.
+    let terms_1000: Vec<String> = (1..=1000).map(|i| format!("term_{i}")).collect();
+    let zipf_1000 = rand_distr::Zipf::new(1000.0, 1.1f64).unwrap();
+
     {
         let mut rng = StdRng::from_seed([1u8; 32]);
         let mut index_writer = index.writer_with_num_threads(1, 200_000_000)?;
@@ -427,15 +621,27 @@ fn get_test_index_bench(cardinality: Cardinality) -> tantivy::Result<Index> {
             index_writer.add_document(doc!())?;
         }
         if cardinality == Cardinality::Multivalued {
+            let log_level_sample_a = status_field_data[log_level_distribution.sample(&mut rng)].0;
+            let log_level_sample_b = status_field_data[log_level_distribution.sample(&mut rng)].0;
+            let idx_a = zipf_1000.sample(&mut rng) as usize - 1;
+            let idx_b = zipf_1000.sample(&mut rng) as usize - 1;
+            let term_1000_a = &terms_1000[idx_a];
+            let term_1000_b = &terms_1000[idx_b];
             index_writer.add_document(doc!(
                 json_field => json!({"mixed_type": 10.0}),
                 json_field => json!({"mixed_type": 10.0}),
                 text_field => "cool",
                 text_field => "cool",
+                text_field_all_unique_terms => "cool",
+                text_field_all_unique_terms => "coolo",
                 text_field_many_terms => "cool",
                 text_field_many_terms => "cool",
                 text_field_few_terms => "cool",
                 text_field_few_terms => "cool",
+                text_field_few_terms_status => log_level_sample_a,
+                text_field_few_terms_status => log_level_sample_b,
+                text_field_1000_terms_zipf => term_1000_a.as_str(),
+                text_field_1000_terms_zipf => term_1000_b.as_str(),
                 score_field => 1u64,
                 score_field => 1u64,
                 score_field_f64 => lg_norm.sample(&mut rng),
@@ -450,8 +656,8 @@ fn get_test_index_bench(cardinality: Cardinality) -> tantivy::Result<Index> {
         }
         let _val_max = 1_000_000.0;
         for _ in 0..doc_with_value {
-            let val: f64 = rng.gen_range(0.0..1_000_000.0);
-            let json = if rng.gen_bool(0.1) {
+            let val: f64 = rng.random_range(0.0..1_000_000.0);
+            let json = if rng.random_bool(0.1) {
                 // 10% are numeric values
                 json!({ "mixed_type": val })
             } else {
@@ -460,11 +666,15 @@ fn get_test_index_bench(cardinality: Cardinality) -> tantivy::Result<Index> {
             index_writer.add_document(doc!(
                 text_field => "cool",
                 json_field => json,
+                text_field_all_unique_terms => format!("unique_term_{}", rng.random::<u64>()),
                 text_field_many_terms => many_terms_data.choose(&mut rng).unwrap().to_string(),
                 text_field_few_terms => few_terms_data.choose(&mut rng).unwrap().to_string(),
+                text_field_few_terms_status => status_field_data[log_level_distribution.sample(&mut rng)].0,
+                text_field_1000_terms_zipf => terms_1000[zipf_1000.sample(&mut rng) as usize - 1].as_str(),
                 score_field => val as u64,
                 score_field_f64 => lg_norm.sample(&mut rng),
                 score_field_i64 => val as i64,
+                date_field => DateTime::from_timestamp_millis((val * 1_000_000.) as i64),
             ))?;
             if cardinality == Cardinality::OptionalSparse {
                 for _ in 0..20 {
@@ -513,7 +723,7 @@ fn filter_agg_all_query_with_sub_aggs(index: &Index) {
                 "avg_score": { "avg": { "field": "score" } },
                 "stats_score": { "stats": { "field": "score_f64" } },
                 "terms_text": {
-                    "terms": { "field": "text_few_terms" }
+                    "terms": { "field": "text_few_terms_status" }
                 }
             }
         }
@@ -529,7 +739,7 @@ fn filter_agg_term_query_with_sub_aggs(index: &Index) {
                 "avg_score": { "avg": { "field": "score" } },
                 "stats_score": { "stats": { "field": "score_f64" } },
                 "terms_text": {
-                    "terms": { "field": "text_few_terms" }
+                    "terms": { "field": "text_few_terms_status" }
                 }
             }
         }

benches/and_or_queries.rs

@@ -16,14 +16,15 @@
 // - This bench isolates boolean iteration speed and intersection/union cost.
 // - Use `cargo bench --bench boolean_conjunction` to run.
 
-use binggan::{black_box, BenchRunner};
+use binggan::{black_box, BenchGroup, BenchRunner};
 use rand::prelude::*;
 use rand::rngs::StdRng;
 use rand::SeedableRng;
-use tantivy::collector::{Count, TopDocs};
-use tantivy::query::QueryParser;
-use tantivy::schema::{Schema, TEXT};
-use tantivy::{doc, Index, ReloadPolicy, Searcher};
+use tantivy::collector::sort_key::SortByStaticFastValue;
+use tantivy::collector::{Collector, Count, TopDocs};
+use tantivy::query::{Query, QueryParser};
+use tantivy::schema::{Schema, FAST, TEXT};
+use tantivy::{doc, Index, Order, ReloadPolicy, Searcher};
 
 #[derive(Clone)]
 struct BenchIndex {
@@ -33,23 +34,6 @@ struct BenchIndex {
     query_parser: QueryParser,
 }
 
-impl BenchIndex {
-    #[inline(always)]
-    fn count_query(&self, query_str: &str) -> usize {
-        let query = self.query_parser.parse_query(query_str).unwrap();
-        self.searcher.search(&query, &Count).unwrap()
-    }
-
-    #[inline(always)]
-    fn topk_len(&self, query_str: &str, k: usize) -> usize {
-        let query = self.query_parser.parse_query(query_str).unwrap();
-        self.searcher
-            .search(&query, &TopDocs::with_limit(k))
-            .unwrap()
-            .len()
-    }
-}
-
 /// Build a single index containing both fields (title, body) and
 /// return two BenchIndex views:
 /// - single_field: QueryParser defaults to only "body"
@@ -59,6 +43,8 @@ fn build_shared_indices(num_docs: usize, p_a: f32, p_b: f32, p_c: f32) -> (Bench
     let mut schema_builder = Schema::builder();
     let f_title = schema_builder.add_text_field("title", TEXT);
     let f_body = schema_builder.add_text_field("body", TEXT);
+    let f_score = schema_builder.add_u64_field("score", FAST);
+    let f_score2 = schema_builder.add_u64_field("score2", FAST);
     let schema = schema_builder.build();
     let index = Index::create_in_ram(schema.clone());
 
@@ -67,29 +53,31 @@ fn build_shared_indices(num_docs: usize, p_a: f32, p_b: f32, p_c: f32) -> (Bench
 
     // Populate: spread each present token 90/10 to body/title
     {
-        let mut writer = index.writer(500_000_000).unwrap();
+        let mut writer = index.writer_with_num_threads(1, 500_000_000).unwrap();
         for _ in 0..num_docs {
-            let has_a = rng.gen_bool(p_a as f64);
-            let has_b = rng.gen_bool(p_b as f64);
-            let has_c = rng.gen_bool(p_c as f64);
+            let has_a = rng.random_bool(p_a as f64);
+            let has_b = rng.random_bool(p_b as f64);
+            let has_c = rng.random_bool(p_c as f64);
+            let score = rng.random_range(0u64..100u64);
+            let score2 = rng.random_range(0u64..100_000u64);
             let mut title_tokens: Vec<&str> = Vec::new();
             let mut body_tokens: Vec<&str> = Vec::new();
             if has_a {
-                if rng.gen_bool(0.1) {
+                if rng.random_bool(0.1) {
                     title_tokens.push("a");
                 } else {
                     body_tokens.push("a");
                 }
             }
             if has_b {
-                if rng.gen_bool(0.1) {
+                if rng.random_bool(0.1) {
                     title_tokens.push("b");
                 } else {
                     body_tokens.push("b");
                 }
             }
             if has_c {
-                if rng.gen_bool(0.1) {
+                if rng.random_bool(0.1) {
                     title_tokens.push("c");
                 } else {
                     body_tokens.push("c");
@@ -101,7 +89,9 @@ fn build_shared_indices(num_docs: usize, p_a: f32, p_b: f32, p_c: f32) -> (Bench
             writer
                 .add_document(doc!(
                     f_title=>title_tokens.join(" "),
-                    f_body=>body_tokens.join(" ")
+                    f_body=>body_tokens.join(" "),
+                    f_score=>score,
+                    f_score2=>score2,
                 ))
                 .unwrap();
         }
@@ -153,72 +143,76 @@ fn main() {
         ),
     ];
 
+    let queries = &["a", "+a +b", "+a +b +c", "a OR b", "a OR b OR c"];
+
     let mut runner = BenchRunner::new();
     for (label, n, pa, pb, pc) in scenarios {
         let (single_view, multi_view) = build_shared_indices(n, pa, pb, pc);
 
-        // Single-field group: default field is body only
+        for (view_name, bench_index) in [("single_field", single_view), ("multi_field", multi_view)]
         {
+            // Single-field group: default field is body only
             let mut group = runner.new_group();
-            group.set_name(format!("single_field — {}", label));
-            group.register_with_input("+a_+b_count", &single_view, |benv: &BenchIndex| {
-                black_box(benv.count_query("+a +b"))
-            });
-            group.register_with_input("+a_+b_+c_count", &single_view, |benv: &BenchIndex| {
-                black_box(benv.count_query("+a +b +c"))
-            });
-            group.register_with_input("+a_+b_top10", &single_view, |benv: &BenchIndex| {
-                black_box(benv.topk_len("+a +b", 10))
-            });
-            group.register_with_input("+a_+b_+c_top10", &single_view, |benv: &BenchIndex| {
-                black_box(benv.topk_len("+a +b +c", 10))
-            });
-            // OR queries
-            group.register_with_input("a_OR_b_count", &single_view, |benv: &BenchIndex| {
-                black_box(benv.count_query("a OR b"))
-            });
-            group.register_with_input("a_OR_b_OR_c_count", &single_view, |benv: &BenchIndex| {
-                black_box(benv.count_query("a OR b OR c"))
-            });
-            group.register_with_input("a_OR_b_top10", &single_view, |benv: &BenchIndex| {
-                black_box(benv.topk_len("a OR b", 10))
-            });
-            group.register_with_input("a_OR_b_OR_c_top10", &single_view, |benv: &BenchIndex| {
-                black_box(benv.topk_len("a OR b OR c", 10))
-            });
-            group.run();
-        }
-
-        // Multi-field group: default fields are [title, body]
-        {
-            let mut group = runner.new_group();
-            group.set_name(format!("multi_field — {}", label));
-            group.register_with_input("+a_+b_count", &multi_view, |benv: &BenchIndex| {
-                black_box(benv.count_query("+a +b"))
-            });
-            group.register_with_input("+a_+b_+c_count", &multi_view, |benv: &BenchIndex| {
-                black_box(benv.count_query("+a +b +c"))
-            });
-            group.register_with_input("+a_+b_top10", &multi_view, |benv: &BenchIndex| {
-                black_box(benv.topk_len("+a +b", 10))
-            });
-            group.register_with_input("+a_+b_+c_top10", &multi_view, |benv: &BenchIndex| {
-                black_box(benv.topk_len("+a +b +c", 10))
-            });
-            // OR queries
-            group.register_with_input("a_OR_b_count", &multi_view, |benv: &BenchIndex| {
-                black_box(benv.count_query("a OR b"))
-            });
-            group.register_with_input("a_OR_b_OR_c_count", &multi_view, |benv: &BenchIndex| {
-                black_box(benv.count_query("a OR b OR c"))
-            });
-            group.register_with_input("a_OR_b_top10", &multi_view, |benv: &BenchIndex| {
-                black_box(benv.topk_len("a OR b", 10))
-            });
-            group.register_with_input("a_OR_b_OR_c_top10", &multi_view, |benv: &BenchIndex| {
-                black_box(benv.topk_len("a OR b OR c", 10))
-            });
+            group.set_name(format!("{} — {}", view_name, label));
+            for query_str in queries {
+                add_bench_task(&mut group, &bench_index, query_str, Count, "count");
+                add_bench_task(
+                    &mut group,
+                    &bench_index,
+                    query_str,
+                    TopDocs::with_limit(10).order_by_score(),
+                    "top10",
+                );
+                add_bench_task(
+                    &mut group,
+                    &bench_index,
+                    query_str,
+                    TopDocs::with_limit(10).order_by_fast_field::<u64>("score", Order::Asc),
+                    "top10_by_ff",
+                );
+                add_bench_task(
+                    &mut group,
+                    &bench_index,
+                    query_str,
+                    TopDocs::with_limit(10).order_by((
+                        SortByStaticFastValue::<u64>::for_field("score"),
+                        SortByStaticFastValue::<u64>::for_field("score2"),
+                    )),
+                    "top10_by_2ff",
+                );
+            }
             group.run();
         }
     }
 }
+
+fn add_bench_task<C: Collector + 'static>(
+    bench_group: &mut BenchGroup,
+    bench_index: &BenchIndex,
+    query_str: &str,
+    collector: C,
+    collector_name: &str,
+) {
+    let task_name = format!("{}_{}", query_str.replace(" ", "_"), collector_name);
+    let query = bench_index.query_parser.parse_query(query_str).unwrap();
+    let search_task = SearchTask {
+        searcher: bench_index.searcher.clone(),
+        collector,
+        query,
+    };
+    bench_group.register(task_name, move |_| black_box(search_task.run()));
+}
+
+struct SearchTask<C: Collector> {
+    searcher: Searcher,
+    collector: C,
+    query: Box<dyn Query>,
+}
+
+impl<C: Collector> SearchTask<C> {
+    #[inline(never)]
+    pub fn run(&self) -> usize {
+        self.searcher.search(&self.query, &self.collector).unwrap();
+        1
+    }
+}

benches/bool_queries_with_range.rs

@@ -0,0 +1,288 @@
+use binggan::{black_box, BenchGroup, BenchRunner};
+use rand::prelude::*;
+use rand::rngs::StdRng;
+use rand::SeedableRng;
+use tantivy::collector::{Collector, Count, DocSetCollector, TopDocs};
+use tantivy::query::{Query, QueryParser};
+use tantivy::schema::{Schema, FAST, INDEXED, TEXT};
+use tantivy::{doc, Index, Order, ReloadPolicy, Searcher};
+
+#[derive(Clone)]
+struct BenchIndex {
+    #[allow(dead_code)]
+    index: Index,
+    searcher: Searcher,
+    query_parser: QueryParser,
+}
+
+fn build_shared_indices(num_docs: usize, p_title_a: f32, distribution: &str) -> BenchIndex {
+    // Unified schema
+    let mut schema_builder = Schema::builder();
+    let f_title = schema_builder.add_text_field("title", TEXT);
+    let f_num_rand = schema_builder.add_u64_field("num_rand", INDEXED);
+    let f_num_asc = schema_builder.add_u64_field("num_asc", INDEXED);
+    let f_num_rand_fast = schema_builder.add_u64_field("num_rand_fast", INDEXED | FAST);
+    let f_num_asc_fast = schema_builder.add_u64_field("num_asc_fast", INDEXED | FAST);
+    let schema = schema_builder.build();
+    let index = Index::create_in_ram(schema.clone());
+
+    // Populate index with stable RNG for reproducibility.
+    let mut rng = StdRng::from_seed([7u8; 32]);
+
+    {
+        let mut writer = index.writer_with_num_threads(1, 4_000_000_000).unwrap();
+
+        match distribution {
+            "dense" => {
+                for doc_id in 0..num_docs {
+                    // Always add title to avoid empty documents
+                    let title_token = if rng.random_bool(p_title_a as f64) {
+                        "a"
+                    } else {
+                        "b"
+                    };
+
+                    let num_rand = rng.random_range(0u64..1000u64);
+
+                    let num_asc = (doc_id / 10000) as u64;
+
+                    writer
+                        .add_document(doc!(
+                            f_title=>title_token,
+                            f_num_rand=>num_rand,
+                            f_num_asc=>num_asc,
+                            f_num_rand_fast=>num_rand,
+                            f_num_asc_fast=>num_asc,
+                        ))
+                        .unwrap();
+                }
+            }
+            "sparse" => {
+                for doc_id in 0..num_docs {
+                    // Always add title to avoid empty documents
+                    let title_token = if rng.random_bool(p_title_a as f64) {
+                        "a"
+                    } else {
+                        "b"
+                    };
+
+                    let num_rand = rng.random_range(0u64..10000000u64);
+
+                    let num_asc = doc_id as u64;
+
+                    writer
+                        .add_document(doc!(
+                            f_title=>title_token,
+                            f_num_rand=>num_rand,
+                            f_num_asc=>num_asc,
+                            f_num_rand_fast=>num_rand,
+                            f_num_asc_fast=>num_asc,
+                        ))
+                        .unwrap();
+                }
+            }
+            _ => {
+                panic!("Unsupported distribution type");
+            }
+        }
+        writer.commit().unwrap();
+    }
+
+    // Prepare reader/searcher once.
+    let reader = index
+        .reader_builder()
+        .reload_policy(ReloadPolicy::Manual)
+        .try_into()
+        .unwrap();
+    let searcher = reader.searcher();
+
+    // Build query parser for title field
+    let qp_title = QueryParser::for_index(&index, vec![f_title]);
+
+    BenchIndex {
+        index,
+        searcher,
+        query_parser: qp_title,
+    }
+}
+
+fn main() {
+    // Prepare corpora with varying scenarios
+    let scenarios = vec![
+        (
+            "dense and 99% a".to_string(),
+            10_000_000,
+            0.99,
+            "dense",
+            0,
+            9,
+        ),
+        (
+            "dense and 99% a".to_string(),
+            10_000_000,
+            0.99,
+            "dense",
+            990,
+            999,
+        ),
+        (
+            "sparse and 99% a".to_string(),
+            10_000_000,
+            0.99,
+            "sparse",
+            0,
+            9,
+        ),
+        (
+            "sparse and 99% a".to_string(),
+            10_000_000,
+            0.99,
+            "sparse",
+            9_999_990,
+            9_999_999,
+        ),
+    ];
+
+    let mut runner = BenchRunner::new();
+    for (scenario_id, n, p_title_a, num_rand_distribution, range_low, range_high) in scenarios {
+        // Build index for this scenario
+        let bench_index = build_shared_indices(n, p_title_a, num_rand_distribution);
+
+        // Create benchmark group
+        let mut group = runner.new_group();
+
+        // Now set the name (this moves scenario_id)
+        group.set_name(scenario_id);
+
+        // Define all four field types
+        let field_names = ["num_rand", "num_asc", "num_rand_fast", "num_asc_fast"];
+
+        // Define the three terms we want to test with
+        let terms = ["a", "b", "z"];
+
+        // Generate all combinations of terms and field names
+        let mut queries = Vec::new();
+        for &term in &terms {
+            for &field_name in &field_names {
+                let query_str = format!(
+                    "{} AND {}:[{} TO {}]",
+                    term, field_name, range_low, range_high
+                );
+                queries.push((query_str, field_name.to_string()));
+            }
+        }
+
+        let query_str = format!(
+            "{}:[{} TO {}] AND {}:[{} TO {}]",
+            "num_rand_fast", range_low, range_high, "num_asc_fast", range_low, range_high
+        );
+        queries.push((query_str, "num_asc_fast".to_string()));
+
+        // Run all benchmark tasks for each query and its corresponding field name
+        for (query_str, field_name) in queries {
+            run_benchmark_tasks(&mut group, &bench_index, &query_str, &field_name);
+        }
+
+        group.run();
+    }
+}
+
+/// Run all benchmark tasks for a given query string and field name
+fn run_benchmark_tasks(
+    bench_group: &mut BenchGroup,
+    bench_index: &BenchIndex,
+    query_str: &str,
+    field_name: &str,
+) {
+    // Test count
+    add_bench_task(bench_group, bench_index, query_str, Count, "count");
+
+    // Test all results
+    add_bench_task(
+        bench_group,
+        bench_index,
+        query_str,
+        DocSetCollector,
+        "all results",
+    );
+
+    // Test top 100 by the field (if it's a FAST field)
+    if field_name.ends_with("_fast") {
+        // Ascending order
+        {
+            let collector_name = format!("top100_by_{}_asc", field_name);
+            let field_name_owned = field_name.to_string();
+            add_bench_task(
+                bench_group,
+                bench_index,
+                query_str,
+                TopDocs::with_limit(100).order_by_fast_field::<u64>(field_name_owned, Order::Asc),
+                &collector_name,
+            );
+        }
+
+        // Descending order
+        {
+            let collector_name = format!("top100_by_{}_desc", field_name);
+            let field_name_owned = field_name.to_string();
+            add_bench_task(
+                bench_group,
+                bench_index,
+                query_str,
+                TopDocs::with_limit(100).order_by_fast_field::<u64>(field_name_owned, Order::Desc),
+                &collector_name,
+            );
+        }
+    }
+}
+
+fn add_bench_task<C: Collector + 'static>(
+    bench_group: &mut BenchGroup,
+    bench_index: &BenchIndex,
+    query_str: &str,
+    collector: C,
+    collector_name: &str,
+) {
+    let task_name = format!("{}_{}", query_str.replace(" ", "_"), collector_name);
+    let query = bench_index.query_parser.parse_query(query_str).unwrap();
+    let search_task = SearchTask {
+        searcher: bench_index.searcher.clone(),
+        collector,
+        query,
+    };
+    bench_group.register(task_name, move |_| black_box(search_task.run()));
+}
+
+struct SearchTask<C: Collector> {
+    searcher: Searcher,
+    collector: C,
+    query: Box<dyn Query>,
+}
+
+impl<C: Collector> SearchTask<C> {
+    #[inline(never)]
+    pub fn run(&self) -> usize {
+        let result = self.searcher.search(&self.query, &self.collector).unwrap();
+        if let Some(count) = (&result as &dyn std::any::Any).downcast_ref::<usize>() {
+            *count
+        } else if let Some(top_docs) = (&result as &dyn std::any::Any)
+            .downcast_ref::<Vec<(Option<u64>, tantivy::DocAddress)>>()
+        {
+            top_docs.len()
+        } else if let Some(top_docs) =
+            (&result as &dyn std::any::Any).downcast_ref::<Vec<(u64, tantivy::DocAddress)>>()
+        {
+            top_docs.len()
+        } else if let Some(doc_set) = (&result as &dyn std::any::Any)
+            .downcast_ref::<std::collections::HashSet<tantivy::DocAddress>>()
+        {
+            doc_set.len()
+        } else {
+            eprintln!(
+                "Unknown collector result type: {:?}",
+                std::any::type_name::<C::Fruit>()
+            );
+            0
+        }
+    }
+}

benches/merge_segments.rs

@@ -0,0 +1,224 @@
+// Benchmarks segment merging
+//
+// Notes:
+// - Input segments are kept intact (no deletes / no IndexWriter merge).
+// - Output is written to a `NullDirectory` that discards all files except
+//  fieldnorms (needed for merging).
+
+use std::collections::HashMap;
+use std::io::{self, Write};
+use std::path::{Path, PathBuf};
+use std::sync::{Arc, RwLock};
+
+use binggan::{black_box, BenchRunner};
+use rand::prelude::*;
+use rand::rngs::StdRng;
+use rand::SeedableRng;
+use tantivy::directory::error::{DeleteError, OpenReadError, OpenWriteError};
+use tantivy::directory::{
+    AntiCallToken, Directory, FileHandle, OwnedBytes, TerminatingWrite, WatchCallback, WatchHandle,
+    WritePtr,
+};
+use tantivy::indexer::{merge_filtered_segments, NoMergePolicy};
+use tantivy::schema::{Schema, TEXT};
+use tantivy::{doc, HasLen, Index, IndexSettings, Segment};
+
+#[derive(Clone, Default, Debug)]
+struct NullDirectory {
+    blobs: Arc<RwLock<HashMap<PathBuf, OwnedBytes>>>,
+}
+
+struct NullWriter;
+
+impl Write for NullWriter {
+    fn write(&mut self, buf: &[u8]) -> io::Result<usize> {
+        Ok(buf.len())
+    }
+
+    fn flush(&mut self) -> io::Result<()> {
+        Ok(())
+    }
+}
+
+impl TerminatingWrite for NullWriter {
+    fn terminate_ref(&mut self, _token: AntiCallToken) -> io::Result<()> {
+        Ok(())
+    }
+}
+
+struct InMemoryWriter {
+    path: PathBuf,
+    buffer: Vec<u8>,
+    blobs: Arc<RwLock<HashMap<PathBuf, OwnedBytes>>>,
+}
+
+impl Write for InMemoryWriter {
+    fn write(&mut self, buf: &[u8]) -> io::Result<usize> {
+        self.buffer.extend_from_slice(buf);
+        Ok(buf.len())
+    }
+
+    fn flush(&mut self) -> io::Result<()> {
+        Ok(())
+    }
+}
+
+impl TerminatingWrite for InMemoryWriter {
+    fn terminate_ref(&mut self, _token: AntiCallToken) -> io::Result<()> {
+        let bytes = OwnedBytes::new(std::mem::take(&mut self.buffer));
+        self.blobs.write().unwrap().insert(self.path.clone(), bytes);
+        Ok(())
+    }
+}
+
+#[derive(Debug, Default)]
+struct NullFileHandle;
+impl HasLen for NullFileHandle {
+    fn len(&self) -> usize {
+        0
+    }
+}
+impl FileHandle for NullFileHandle {
+    fn read_bytes(&self, _range: std::ops::Range<usize>) -> io::Result<OwnedBytes> {
+        unimplemented!()
+    }
+}
+
+impl Directory for NullDirectory {
+    fn get_file_handle(&self, path: &Path) -> Result<Arc<dyn FileHandle>, OpenReadError> {
+        if let Some(bytes) = self.blobs.read().unwrap().get(path) {
+            return Ok(Arc::new(bytes.clone()));
+        }
+        Ok(Arc::new(NullFileHandle))
+    }
+
+    fn delete(&self, _path: &Path) -> Result<(), DeleteError> {
+        Ok(())
+    }
+
+    fn exists(&self, _path: &Path) -> Result<bool, OpenReadError> {
+        Ok(true)
+    }
+
+    fn open_write(&self, path: &Path) -> Result<WritePtr, OpenWriteError> {
+        let path_buf = path.to_path_buf();
+        if path.to_string_lossy().ends_with(".fieldnorm") {
+            let writer = InMemoryWriter {
+                path: path_buf,
+                buffer: Vec::new(),
+                blobs: Arc::clone(&self.blobs),
+            };
+            Ok(io::BufWriter::new(Box::new(writer)))
+        } else {
+            Ok(io::BufWriter::new(Box::new(NullWriter)))
+        }
+    }
+
+    fn atomic_read(&self, path: &Path) -> Result<Vec<u8>, OpenReadError> {
+        if let Some(bytes) = self.blobs.read().unwrap().get(path) {
+            return Ok(bytes.as_slice().to_vec());
+        }
+        Err(OpenReadError::FileDoesNotExist(path.to_path_buf()))
+    }
+
+    fn atomic_write(&self, _path: &Path, _data: &[u8]) -> io::Result<()> {
+        Ok(())
+    }
+
+    fn sync_directory(&self) -> io::Result<()> {
+        Ok(())
+    }
+
+    fn watch(&self, _watch_callback: WatchCallback) -> tantivy::Result<WatchHandle> {
+        Ok(WatchHandle::empty())
+    }
+}
+
+struct MergeScenario {
+    #[allow(dead_code)]
+    index: Index,
+    segments: Vec<Segment>,
+    settings: IndexSettings,
+    label: String,
+}
+
+fn build_index(
+    num_segments: usize,
+    docs_per_segment: usize,
+    tokens_per_doc: usize,
+    vocab_size: usize,
+) -> MergeScenario {
+    let mut schema_builder = Schema::builder();
+    let body = schema_builder.add_text_field("body", TEXT);
+    let schema = schema_builder.build();
+    let index = Index::create_in_ram(schema.clone());
+
+    assert!(vocab_size > 0);
+    let total_tokens = num_segments * docs_per_segment * tokens_per_doc;
+    let use_unique_terms = vocab_size >= total_tokens;
+    let mut rng = StdRng::from_seed([7u8; 32]);
+    let mut next_token_id: u64 = 0;
+
+    {
+        let mut writer = index.writer_with_num_threads(1, 256_000_000).unwrap();
+        writer.set_merge_policy(Box::new(NoMergePolicy));
+        for _ in 0..num_segments {
+            for _ in 0..docs_per_segment {
+                let mut tokens = Vec::with_capacity(tokens_per_doc);
+                for _ in 0..tokens_per_doc {
+                    let token_id = if use_unique_terms {
+                        let id = next_token_id;
+                        next_token_id += 1;
+                        id
+                    } else {
+                        rng.random_range(0..vocab_size as u64)
+                    };
+                    tokens.push(format!("term_{token_id}"));
+                }
+                writer.add_document(doc!(body => tokens.join(" "))).unwrap();
+            }
+            writer.commit().unwrap();
+        }
+    }
+
+    let segments = index.searchable_segments().unwrap();
+    let settings = index.settings().clone();
+    let label = format!(
+        "segments={}, docs/seg={}, tokens/doc={}, vocab={}",
+        num_segments, docs_per_segment, tokens_per_doc, vocab_size
+    );
+
+    MergeScenario {
+        index,
+        segments,
+        settings,
+        label,
+    }
+}
+
+fn main() {
+    let scenarios = vec![
+        build_index(8, 50_000, 12, 8),
+        build_index(16, 50_000, 12, 8),
+        build_index(16, 100_000, 12, 8),
+        build_index(8, 50_000, 8, 8 * 50_000 * 8),
+    ];
+
+    let mut runner = BenchRunner::new();
+    for scenario in scenarios {
+        let mut group = runner.new_group();
+        group.set_name(format!("merge_segments inv_index — {}", scenario.label));
+        let segments = scenario.segments.clone();
+        let settings = scenario.settings.clone();
+        group.register("merge", move |_| {
+            let output_dir = NullDirectory::default();
+            let filter_doc_ids = vec![None; segments.len()];
+            let merged_index =
+                merge_filtered_segments(&segments, settings.clone(), filter_doc_ids, output_dir)
+                    .unwrap();
+            black_box(merged_index);
+        });
+
+        group.run();
+    }
+}

benches/range_queries.rs

@@ -0,0 +1,365 @@
+use std::ops::Bound;
+
+use binggan::{black_box, BenchGroup, BenchRunner};
+use rand::prelude::*;
+use rand::rngs::StdRng;
+use rand::SeedableRng;
+use tantivy::collector::{Count, DocSetCollector, TopDocs};
+use tantivy::query::RangeQuery;
+use tantivy::schema::{Schema, FAST, INDEXED};
+use tantivy::{doc, Index, Order, ReloadPolicy, Searcher, Term};
+
+#[derive(Clone)]
+struct BenchIndex {
+    #[allow(dead_code)]
+    index: Index,
+    searcher: Searcher,
+}
+
+fn build_shared_indices(num_docs: usize, distribution: &str) -> BenchIndex {
+    // Schema with fast fields only
+    let mut schema_builder = Schema::builder();
+    let f_num_rand_fast = schema_builder.add_u64_field("num_rand_fast", INDEXED | FAST);
+    let f_num_asc_fast = schema_builder.add_u64_field("num_asc_fast", INDEXED | FAST);
+    let schema = schema_builder.build();
+    let index = Index::create_in_ram(schema.clone());
+
+    // Populate index with stable RNG for reproducibility.
+    let mut rng = StdRng::from_seed([7u8; 32]);
+
+    {
+        let mut writer = index.writer_with_num_threads(1, 4_000_000_000).unwrap();
+
+        match distribution {
+            "dense" => {
+                for doc_id in 0..num_docs {
+                    let num_rand = rng.random_range(0u64..1000u64);
+                    let num_asc = (doc_id / 10000) as u64;
+
+                    writer
+                        .add_document(doc!(
+                            f_num_rand_fast=>num_rand,
+                            f_num_asc_fast=>num_asc,
+                        ))
+                        .unwrap();
+                }
+            }
+            "sparse" => {
+                for doc_id in 0..num_docs {
+                    let num_rand = rng.random_range(0u64..10000000u64);
+                    let num_asc = doc_id as u64;
+
+                    writer
+                        .add_document(doc!(
+                            f_num_rand_fast=>num_rand,
+                            f_num_asc_fast=>num_asc,
+                        ))
+                        .unwrap();
+                }
+            }
+            _ => {
+                panic!("Unsupported distribution type");
+            }
+        }
+        writer.commit().unwrap();
+    }
+
+    // Prepare reader/searcher once.
+    let reader = index
+        .reader_builder()
+        .reload_policy(ReloadPolicy::Manual)
+        .try_into()
+        .unwrap();
+    let searcher = reader.searcher();
+
+    BenchIndex { index, searcher }
+}
+
+fn main() {
+    // Prepare corpora with varying scenarios
+    let scenarios = vec![
+        // Dense distribution - random values in small range (0-999)
+        (
+            "dense_values_search_low_value_range".to_string(),
+            10_000_000,
+            "dense",
+            0,
+            9,
+        ),
+        (
+            "dense_values_search_high_value_range".to_string(),
+            10_000_000,
+            "dense",
+            990,
+            999,
+        ),
+        (
+            "dense_values_search_out_of_range".to_string(),
+            10_000_000,
+            "dense",
+            1000,
+            1002,
+        ),
+        (
+            "sparse_values_search_low_value_range".to_string(),
+            10_000_000,
+            "sparse",
+            0,
+            9,
+        ),
+        (
+            "sparse_values_search_high_value_range".to_string(),
+            10_000_000,
+            "sparse",
+            9_999_990,
+            9_999_999,
+        ),
+        (
+            "sparse_values_search_out_of_range".to_string(),
+            10_000_000,
+            "sparse",
+            10_000_000,
+            10_000_002,
+        ),
+    ];
+
+    let mut runner = BenchRunner::new();
+    for (scenario_id, n, num_rand_distribution, range_low, range_high) in scenarios {
+        // Build index for this scenario
+        let bench_index = build_shared_indices(n, num_rand_distribution);
+
+        // Create benchmark group
+        let mut group = runner.new_group();
+
+        // Now set the name (this moves scenario_id)
+        group.set_name(scenario_id);
+
+        // Define fast field types
+        let field_names = ["num_rand_fast", "num_asc_fast"];
+
+        // Generate range queries for fast fields
+        for &field_name in &field_names {
+            // Create the range query
+            let field = bench_index.searcher.schema().get_field(field_name).unwrap();
+            let lower_term = Term::from_field_u64(field, range_low);
+            let upper_term = Term::from_field_u64(field, range_high);
+
+            let query = RangeQuery::new(Bound::Included(lower_term), Bound::Included(upper_term));
+
+            run_benchmark_tasks(
+                &mut group,
+                &bench_index,
+                query,
+                field_name,
+                range_low,
+                range_high,
+            );
+        }
+
+        group.run();
+    }
+}
+
+/// Run all benchmark tasks for a given range query and field name
+fn run_benchmark_tasks(
+    bench_group: &mut BenchGroup,
+    bench_index: &BenchIndex,
+    query: RangeQuery,
+    field_name: &str,
+    range_low: u64,
+    range_high: u64,
+) {
+    // Test count
+    add_bench_task_count(
+        bench_group,
+        bench_index,
+        query.clone(),
+        "count",
+        field_name,
+        range_low,
+        range_high,
+    );
+
+    // Test top 100 by the field (ascending order)
+    {
+        let collector_name = format!("top100_by_{}_asc", field_name);
+        let field_name_owned = field_name.to_string();
+        add_bench_task_top100_asc(
+            bench_group,
+            bench_index,
+            query.clone(),
+            &collector_name,
+            field_name,
+            range_low,
+            range_high,
+            field_name_owned,
+        );
+    }
+
+    // Test top 100 by the field (descending order)
+    {
+        let collector_name = format!("top100_by_{}_desc", field_name);
+        let field_name_owned = field_name.to_string();
+        add_bench_task_top100_desc(
+            bench_group,
+            bench_index,
+            query,
+            &collector_name,
+            field_name,
+            range_low,
+            range_high,
+            field_name_owned,
+        );
+    }
+}
+
+fn add_bench_task_count(
+    bench_group: &mut BenchGroup,
+    bench_index: &BenchIndex,
+    query: RangeQuery,
+    collector_name: &str,
+    field_name: &str,
+    range_low: u64,
+    range_high: u64,
+) {
+    let task_name = format!(
+        "range_{}_[{} TO {}]_{}",
+        field_name, range_low, range_high, collector_name
+    );
+
+    let search_task = CountSearchTask {
+        searcher: bench_index.searcher.clone(),
+        query,
+    };
+    bench_group.register(task_name, move |_| black_box(search_task.run()));
+}
+
+fn add_bench_task_docset(
+    bench_group: &mut BenchGroup,
+    bench_index: &BenchIndex,
+    query: RangeQuery,
+    collector_name: &str,
+    field_name: &str,
+    range_low: u64,
+    range_high: u64,
+) {
+    let task_name = format!(
+        "range_{}_[{} TO {}]_{}",
+        field_name, range_low, range_high, collector_name
+    );
+
+    let search_task = DocSetSearchTask {
+        searcher: bench_index.searcher.clone(),
+        query,
+    };
+    bench_group.register(task_name, move |_| black_box(search_task.run()));
+}
+
+fn add_bench_task_top100_asc(
+    bench_group: &mut BenchGroup,
+    bench_index: &BenchIndex,
+    query: RangeQuery,
+    collector_name: &str,
+    field_name: &str,
+    range_low: u64,
+    range_high: u64,
+    field_name_owned: String,
+) {
+    let task_name = format!(
+        "range_{}_[{} TO {}]_{}",
+        field_name, range_low, range_high, collector_name
+    );
+
+    let search_task = Top100AscSearchTask {
+        searcher: bench_index.searcher.clone(),
+        query,
+        field_name: field_name_owned,
+    };
+    bench_group.register(task_name, move |_| black_box(search_task.run()));
+}
+
+fn add_bench_task_top100_desc(
+    bench_group: &mut BenchGroup,
+    bench_index: &BenchIndex,
+    query: RangeQuery,
+    collector_name: &str,
+    field_name: &str,
+    range_low: u64,
+    range_high: u64,
+    field_name_owned: String,
+) {
+    let task_name = format!(
+        "range_{}_[{} TO {}]_{}",
+        field_name, range_low, range_high, collector_name
+    );
+
+    let search_task = Top100DescSearchTask {
+        searcher: bench_index.searcher.clone(),
+        query,
+        field_name: field_name_owned,
+    };
+    bench_group.register(task_name, move |_| black_box(search_task.run()));
+}
+
+struct CountSearchTask {
+    searcher: Searcher,
+    query: RangeQuery,
+}
+
+impl CountSearchTask {
+    #[inline(never)]
+    pub fn run(&self) -> usize {
+        self.searcher.search(&self.query, &Count).unwrap()
+    }
+}
+
+struct DocSetSearchTask {
+    searcher: Searcher,
+    query: RangeQuery,
+}
+
+impl DocSetSearchTask {
+    #[inline(never)]
+    pub fn run(&self) -> usize {
+        let result = self.searcher.search(&self.query, &DocSetCollector).unwrap();
+        result.len()
+    }
+}
+
+struct Top100AscSearchTask {
+    searcher: Searcher,
+    query: RangeQuery,
+    field_name: String,
+}
+
+impl Top100AscSearchTask {
+    #[inline(never)]
+    pub fn run(&self) -> usize {
+        let collector =
+            TopDocs::with_limit(100).order_by_fast_field::<u64>(&self.field_name, Order::Asc);
+        let result = self.searcher.search(&self.query, &collector).unwrap();
+        for (_score, doc_address) in &result {
+            let _doc: tantivy::TantivyDocument = self.searcher.doc(*doc_address).unwrap();
+        }
+        result.len()
+    }
+}
+
+struct Top100DescSearchTask {
+    searcher: Searcher,
+    query: RangeQuery,
+    field_name: String,
+}
+
+impl Top100DescSearchTask {
+    #[inline(never)]
+    pub fn run(&self) -> usize {
+        let collector =
+            TopDocs::with_limit(100).order_by_fast_field::<u64>(&self.field_name, Order::Desc);
+        let result = self.searcher.search(&self.query, &collector).unwrap();
+        for (_score, doc_address) in &result {
+            let _doc: tantivy::TantivyDocument = self.searcher.doc(*doc_address).unwrap();
+        }
+        result.len()
+    }
+}

benches/range_query.rs

@@ -0,0 +1,260 @@
+use std::fmt::Display;
+use std::net::Ipv6Addr;
+use std::ops::RangeInclusive;
+
+use binggan::plugins::PeakMemAllocPlugin;
+use binggan::{black_box, BenchRunner, OutputValue, PeakMemAlloc, INSTRUMENTED_SYSTEM};
+use columnar::MonotonicallyMappableToU128;
+use rand::rngs::StdRng;
+use rand::{Rng, SeedableRng};
+use tantivy::collector::{Count, TopDocs};
+use tantivy::query::QueryParser;
+use tantivy::schema::*;
+use tantivy::{doc, Index};
+
+#[global_allocator]
+pub static GLOBAL: &PeakMemAlloc<std::alloc::System> = &INSTRUMENTED_SYSTEM;
+
+fn main() {
+    bench_range_query();
+}
+
+fn bench_range_query() {
+    let index = get_index_0_to_100();
+    let mut runner = BenchRunner::new();
+    runner.add_plugin(PeakMemAllocPlugin::new(GLOBAL));
+
+    runner.set_name("range_query on u64");
+    let field_name_and_descr: Vec<_> = vec![
+        ("id", "Single Valued Range Field"),
+        ("ids", "Multi Valued Range Field"),
+    ];
+    let range_num_hits = vec![
+        ("90_percent", get_90_percent()),
+        ("10_percent", get_10_percent()),
+        ("1_percent", get_1_percent()),
+    ];
+
+    test_range(&mut runner, &index, &field_name_and_descr, range_num_hits);
+
+    runner.set_name("range_query on ip");
+    let field_name_and_descr: Vec<_> = vec![
+        ("ip", "Single Valued Range Field"),
+        ("ips", "Multi Valued Range Field"),
+    ];
+    let range_num_hits = vec![
+        ("90_percent", get_90_percent_ip()),
+        ("10_percent", get_10_percent_ip()),
+        ("1_percent", get_1_percent_ip()),
+    ];
+
+    test_range(&mut runner, &index, &field_name_and_descr, range_num_hits);
+}
+
+fn test_range<T: Display>(
+    runner: &mut BenchRunner,
+    index: &Index,
+    field_name_and_descr: &[(&str, &str)],
+    range_num_hits: Vec<(&str, RangeInclusive<T>)>,
+) {
+    for (field, suffix) in field_name_and_descr {
+        let term_num_hits = vec![
+            ("", ""),
+            ("1_percent", "veryfew"),
+            ("10_percent", "few"),
+            ("90_percent", "most"),
+        ];
+        let mut group = runner.new_group();
+        group.set_name(suffix);
+        // all intersect combinations
+        for (range_name, range) in &range_num_hits {
+            for (term_name, term) in &term_num_hits {
+                let index = &index;
+                let test_name = if term_name.is_empty() {
+                    format!("id_range_hit_{}", range_name)
+                } else {
+                    format!(
+                        "id_range_hit_{}_intersect_with_term_{}",
+                        range_name, term_name
+                    )
+                };
+                group.register(test_name, move |_| {
+                    let query = if term_name.is_empty() {
+                        "".to_string()
+                    } else {
+                        format!("AND id_name:{}", term)
+                    };
+                    black_box(execute_query(field, range, &query, index));
+                });
+            }
+        }
+        group.run();
+    }
+}
+
+fn get_index_0_to_100() -> Index {
+    let mut rng = StdRng::from_seed([1u8; 32]);
+    let num_vals = 100_000;
+    let docs: Vec<_> = (0..num_vals)
+        .map(|_i| {
+            let id_name = if rng.random_bool(0.01) {
+                "veryfew".to_string() // 1%
+            } else if rng.random_bool(0.1) {
+                "few".to_string() // 9%
+            } else {
+                "most".to_string() // 90%
+            };
+            Doc {
+                id_name,
+                id: rng.random_range(0..100),
+                // Multiply by 1000, so that we create most buckets in the compact space
+                // The benches depend on this range to select n-percent of elements with the
+                // methods below.
+                ip: Ipv6Addr::from_u128(rng.random_range(0..100) * 1000),
+            }
+        })
+        .collect();
+
+    create_index_from_docs(&docs)
+}
+
+#[derive(Clone, Debug)]
+pub struct Doc {
+    pub id_name: String,
+    pub id: u64,
+    pub ip: Ipv6Addr,
+}
+
+pub fn create_index_from_docs(docs: &[Doc]) -> Index {
+    let mut schema_builder = Schema::builder();
+    let id_u64_field = schema_builder.add_u64_field("id", INDEXED | STORED | FAST);
+    let ids_u64_field =
+        schema_builder.add_u64_field("ids", NumericOptions::default().set_fast().set_indexed());
+
+    let id_f64_field = schema_builder.add_f64_field("id_f64", INDEXED | STORED | FAST);
+    let ids_f64_field = schema_builder.add_f64_field(
+        "ids_f64",
+        NumericOptions::default().set_fast().set_indexed(),
+    );
+
+    let id_i64_field = schema_builder.add_i64_field("id_i64", INDEXED | STORED | FAST);
+    let ids_i64_field = schema_builder.add_i64_field(
+        "ids_i64",
+        NumericOptions::default().set_fast().set_indexed(),
+    );
+
+    let text_field = schema_builder.add_text_field("id_name", STRING | STORED);
+    let text_field2 = schema_builder.add_text_field("id_name_fast", STRING | STORED | FAST);
+
+    let ip_field = schema_builder.add_ip_addr_field("ip", FAST);
+    let ips_field = schema_builder.add_ip_addr_field("ips", FAST);
+
+    let schema = schema_builder.build();
+
+    let index = Index::create_in_ram(schema);
+
+    {
+        let mut index_writer = index.writer_with_num_threads(1, 50_000_000).unwrap();
+        for doc in docs.iter() {
+            index_writer
+                .add_document(doc!(
+                    ids_i64_field => doc.id as i64,
+                    ids_i64_field => doc.id as i64,
+                    ids_f64_field => doc.id as f64,
+                    ids_f64_field => doc.id as f64,
+                    ids_u64_field => doc.id,
+                    ids_u64_field => doc.id,
+                    id_u64_field => doc.id,
+                    id_f64_field => doc.id as f64,
+                    id_i64_field => doc.id as i64,
+                    text_field => doc.id_name.to_string(),
+                    text_field2 => doc.id_name.to_string(),
+                    ips_field => doc.ip,
+                    ips_field => doc.ip,
+                    ip_field => doc.ip,
+                ))
+                .unwrap();
+        }
+
+        index_writer.commit().unwrap();
+    }
+    index
+}
+
+fn get_90_percent() -> RangeInclusive<u64> {
+    0..=90
+}
+
+fn get_10_percent() -> RangeInclusive<u64> {
+    0..=10
+}
+
+fn get_1_percent() -> RangeInclusive<u64> {
+    10..=10
+}
+
+fn get_90_percent_ip() -> RangeInclusive<Ipv6Addr> {
+    let start = Ipv6Addr::from_u128(0);
+    let end = Ipv6Addr::from_u128(90 * 1000);
+    start..=end
+}
+
+fn get_10_percent_ip() -> RangeInclusive<Ipv6Addr> {
+    let start = Ipv6Addr::from_u128(0);
+    let end = Ipv6Addr::from_u128(10 * 1000);
+    start..=end
+}
+
+fn get_1_percent_ip() -> RangeInclusive<Ipv6Addr> {
+    let start = Ipv6Addr::from_u128(10 * 1000);
+    let end = Ipv6Addr::from_u128(10 * 1000);
+    start..=end
+}
+
+struct NumHits {
+    count: usize,
+}
+impl OutputValue for NumHits {
+    fn column_title() -> &'static str {
+        "NumHits"
+    }
+    fn format(&self) -> Option<String> {
+        Some(self.count.to_string())
+    }
+}
+
+fn execute_query<T: Display>(
+    field: &str,
+    id_range: &RangeInclusive<T>,
+    suffix: &str,
+    index: &Index,
+) -> NumHits {
+    let gen_query_inclusive = |from: &T, to: &T| {
+        format!(
+            "{}:[{} TO {}] {}",
+            field,
+            &from.to_string(),
+            &to.to_string(),
+            suffix
+        )
+    };
+
+    let query = gen_query_inclusive(id_range.start(), id_range.end());
+    execute_query_(&query, index)
+}
+
+fn execute_query_(query: &str, index: &Index) -> NumHits {
+    let query_from_text = |text: &str| {
+        QueryParser::for_index(index, vec![])
+            .parse_query(text)
+            .unwrap()
+    };
+    let query = query_from_text(query);
+    let reader = index.reader().unwrap();
+    let searcher = reader.searcher();
+    let num_hits = searcher
+        .search(&query, &(TopDocs::with_limit(10).order_by_score(), Count))
+        .unwrap()
+        .1;
+    NumHits { count: num_hits }
+}

benches/regex_all_terms.rs

@@ -0,0 +1,113 @@
+// Benchmarks regex query that matches all terms in a synthetic index.
+//
+// Corpus model:
+// - N unique terms: t000000, t000001, ...
+// - M docs
+// - K tokens per doc: doc i gets terms derived from (i, token_index)
+//
+// Query:
+// - Regex "t.*" to match all terms
+//
+// Run with:
+// - cargo bench --bench regex_all_terms
+//
+
+use std::fmt::Write;
+
+use binggan::{black_box, BenchRunner};
+use tantivy::collector::Count;
+use tantivy::query::RegexQuery;
+use tantivy::schema::{Schema, TEXT};
+use tantivy::{doc, Index, ReloadPolicy};
+
+const HEAP_SIZE_BYTES: usize = 200_000_000;
+
+#[derive(Clone, Copy)]
+struct BenchConfig {
+    num_terms: usize,
+    num_docs: usize,
+    tokens_per_doc: usize,
+}
+
+fn main() {
+    let configs = default_configs();
+
+    let mut runner = BenchRunner::new();
+    for config in configs {
+        let (index, text_field) = build_index(config, HEAP_SIZE_BYTES);
+        let reader = index
+            .reader_builder()
+            .reload_policy(ReloadPolicy::Manual)
+            .try_into()
+            .expect("reader");
+        let searcher = reader.searcher();
+        let query = RegexQuery::from_pattern("t.*", text_field).expect("regex query");
+
+        let mut group = runner.new_group();
+        group.set_name(format!(
+            "regex_all_terms_t{}_d{}_k{}",
+            config.num_terms, config.num_docs, config.tokens_per_doc
+        ));
+        group.register("regex_count", move |_| {
+            let count = searcher.search(&query, &Count).expect("search");
+            black_box(count);
+        });
+        group.run();
+    }
+}
+
+fn default_configs() -> Vec<BenchConfig> {
+    vec![
+        BenchConfig {
+            num_terms: 10_000,
+            num_docs: 100_000,
+            tokens_per_doc: 1,
+        },
+        BenchConfig {
+            num_terms: 10_000,
+            num_docs: 100_000,
+            tokens_per_doc: 8,
+        },
+        BenchConfig {
+            num_terms: 100_000,
+            num_docs: 100_000,
+            tokens_per_doc: 1,
+        },
+        BenchConfig {
+            num_terms: 100_000,
+            num_docs: 100_000,
+            tokens_per_doc: 8,
+        },
+    ]
+}
+
+fn build_index(config: BenchConfig, heap_size_bytes: usize) -> (Index, tantivy::schema::Field) {
+    let mut schema_builder = Schema::builder();
+    let text_field = schema_builder.add_text_field("text", TEXT);
+    let schema = schema_builder.build();
+    let index = Index::create_in_ram(schema);
+
+    let term_width = config.num_terms.to_string().len();
+    {
+        let mut writer = index
+            .writer_with_num_threads(1, heap_size_bytes)
+            .expect("writer");
+        let mut buffer = String::new();
+        for doc_id in 0..config.num_docs {
+            buffer.clear();
+            for token_idx in 0..config.tokens_per_doc {
+                if token_idx > 0 {
+                    buffer.push(' ');
+                }
+                let term_id = (doc_id * config.tokens_per_doc + token_idx) % config.num_terms;
+                write!(&mut buffer, "t{term_id:0term_width$}").expect("write token");
+            }
+            writer
+                .add_document(doc!(text_field => buffer.as_str()))
+                .expect("add_document");
+        }
+        writer.commit().expect("commit");
+    }
+
+    (index, text_field)
+}

benches/str_search_and_get.rs

@@ -0,0 +1,421 @@
+// This benchmark compares different approaches for retrieving string values:
+//
+// 1. Fast Field Approach: retrieves string values via term_ords() and ord_to_str()
+//
+// 2. Doc Store Approach: retrieves string values via searcher.doc() and field extraction
+//
+// The benchmark includes various data distributions:
+// - Dense Sequential: Sequential document IDs with dense data
+// - Dense Random: Random document IDs with dense data
+// - Sparse Sequential: Sequential document IDs with sparse data
+// - Sparse Random: Random document IDs with sparse data
+use std::ops::Bound;
+
+use binggan::{black_box, BenchGroup, BenchRunner};
+use rand::prelude::*;
+use rand::rngs::StdRng;
+use rand::SeedableRng;
+use tantivy::collector::{Count, DocSetCollector};
+use tantivy::query::RangeQuery;
+use tantivy::schema::document::TantivyDocument;
+use tantivy::schema::{Schema, Value, FAST, STORED, STRING};
+use tantivy::{doc, Index, ReloadPolicy, Searcher, Term};
+
+#[derive(Clone)]
+struct BenchIndex {
+    #[allow(dead_code)]
+    index: Index,
+    searcher: Searcher,
+}
+
+fn build_shared_indices(num_docs: usize, distribution: &str) -> BenchIndex {
+    // Schema with string fast field and stored field for doc access
+    let mut schema_builder = Schema::builder();
+    let f_str_fast = schema_builder.add_text_field("str_fast", STRING | STORED | FAST);
+    let f_str_stored = schema_builder.add_text_field("str_stored", STRING | STORED);
+    let schema = schema_builder.build();
+    let index = Index::create_in_ram(schema.clone());
+
+    // Populate index with stable RNG for reproducibility.
+    let mut rng = StdRng::from_seed([7u8; 32]);
+
+    {
+        let mut writer = index.writer_with_num_threads(1, 4_000_000_000).unwrap();
+
+        match distribution {
+            "dense_random" => {
+                for _doc_id in 0..num_docs {
+                    let suffix = rng.gen_range(0u64..1000u64);
+                    let str_val = format!("str_{:03}", suffix);
+
+                    writer
+                        .add_document(doc!(
+                            f_str_fast=>str_val.clone(),
+                            f_str_stored=>str_val,
+                        ))
+                        .unwrap();
+                }
+            }
+            "dense_sequential" => {
+                for doc_id in 0..num_docs {
+                    let suffix = doc_id as u64 % 1000;
+                    let str_val = format!("str_{:03}", suffix);
+
+                    writer
+                        .add_document(doc!(
+                            f_str_fast=>str_val.clone(),
+                            f_str_stored=>str_val,
+                        ))
+                        .unwrap();
+                }
+            }
+            "sparse_random" => {
+                for _doc_id in 0..num_docs {
+                    let suffix = rng.gen_range(0u64..1000000u64);
+                    let str_val = format!("str_{:07}", suffix);
+
+                    writer
+                        .add_document(doc!(
+                            f_str_fast=>str_val.clone(),
+                            f_str_stored=>str_val,
+                        ))
+                        .unwrap();
+                }
+            }
+            "sparse_sequential" => {
+                for doc_id in 0..num_docs {
+                    let suffix = doc_id as u64;
+                    let str_val = format!("str_{:07}", suffix);
+
+                    writer
+                        .add_document(doc!(
+                            f_str_fast=>str_val.clone(),
+                            f_str_stored=>str_val,
+                        ))
+                        .unwrap();
+                }
+            }
+            _ => {
+                panic!("Unsupported distribution type");
+            }
+        }
+        writer.commit().unwrap();
+    }
+
+    // Prepare reader/searcher once.
+    let reader = index
+        .reader_builder()
+        .reload_policy(ReloadPolicy::Manual)
+        .try_into()
+        .unwrap();
+    let searcher = reader.searcher();
+
+    BenchIndex { index, searcher }
+}
+
+fn main() {
+    // Prepare corpora with varying scenarios
+    let scenarios = vec![
+        (
+            "dense_random_search_low_range".to_string(),
+            1_000_000,
+            "dense_random",
+            0,
+            9,
+        ),
+        (
+            "dense_random_search_high_range".to_string(),
+            1_000_000,
+            "dense_random",
+            990,
+            999,
+        ),
+        (
+            "dense_sequential_search_low_range".to_string(),
+            1_000_000,
+            "dense_sequential",
+            0,
+            9,
+        ),
+        (
+            "dense_sequential_search_high_range".to_string(),
+            1_000_000,
+            "dense_sequential",
+            990,
+            999,
+        ),
+        (
+            "sparse_random_search_low_range".to_string(),
+            1_000_000,
+            "sparse_random",
+            0,
+            9999,
+        ),
+        (
+            "sparse_random_search_high_range".to_string(),
+            1_000_000,
+            "sparse_random",
+            990_000,
+            999_999,
+        ),
+        (
+            "sparse_sequential_search_low_range".to_string(),
+            1_000_000,
+            "sparse_sequential",
+            0,
+            9999,
+        ),
+        (
+            "sparse_sequential_search_high_range".to_string(),
+            1_000_000,
+            "sparse_sequential",
+            990_000,
+            999_999,
+        ),
+    ];
+
+    let mut runner = BenchRunner::new();
+    for (scenario_id, n, distribution, range_low, range_high) in scenarios {
+        let bench_index = build_shared_indices(n, distribution);
+        let mut group = runner.new_group();
+        group.set_name(scenario_id);
+
+        let field = bench_index.searcher.schema().get_field("str_fast").unwrap();
+
+        let (lower_str, upper_str) =
+            if distribution == "dense_sequential" || distribution == "dense_random" {
+                (
+                    format!("str_{:03}", range_low),
+                    format!("str_{:03}", range_high),
+                )
+            } else {
+                (
+                    format!("str_{:07}", range_low),
+                    format!("str_{:07}", range_high),
+                )
+            };
+
+        let lower_term = Term::from_field_text(field, &lower_str);
+        let upper_term = Term::from_field_text(field, &upper_str);
+
+        let query = RangeQuery::new(Bound::Included(lower_term), Bound::Included(upper_term));
+
+        run_benchmark_tasks(&mut group, &bench_index, query, range_low, range_high);
+
+        group.run();
+    }
+}
+
+/// Run all benchmark tasks for a given range query
+fn run_benchmark_tasks(
+    bench_group: &mut BenchGroup,
+    bench_index: &BenchIndex,
+    query: RangeQuery,
+    range_low: u64,
+    range_high: u64,
+) {
+    // Test count of matching documents
+    add_bench_task_count(
+        bench_group,
+        bench_index,
+        query.clone(),
+        range_low,
+        range_high,
+    );
+
+    // Test fetching all DocIds of matching documents
+    add_bench_task_docset(
+        bench_group,
+        bench_index,
+        query.clone(),
+        range_low,
+        range_high,
+    );
+
+    // Test fetching all string fast field values of matching documents
+    add_bench_task_fetch_all_strings(
+        bench_group,
+        bench_index,
+        query.clone(),
+        range_low,
+        range_high,
+    );
+
+    // Test fetching all string values of matching documents through doc() method
+    add_bench_task_fetch_all_strings_from_doc(
+        bench_group,
+        bench_index,
+        query,
+        range_low,
+        range_high,
+    );
+}
+
+fn add_bench_task_count(
+    bench_group: &mut BenchGroup,
+    bench_index: &BenchIndex,
+    query: RangeQuery,
+    range_low: u64,
+    range_high: u64,
+) {
+    let task_name = format!("string_search_count_[{}-{}]", range_low, range_high);
+
+    let search_task = CountSearchTask {
+        searcher: bench_index.searcher.clone(),
+        query,
+    };
+    bench_group.register(task_name, move |_| black_box(search_task.run()));
+}
+
+fn add_bench_task_docset(
+    bench_group: &mut BenchGroup,
+    bench_index: &BenchIndex,
+    query: RangeQuery,
+    range_low: u64,
+    range_high: u64,
+) {
+    let task_name = format!("string_fetch_all_docset_[{}-{}]", range_low, range_high);
+
+    let search_task = DocSetSearchTask {
+        searcher: bench_index.searcher.clone(),
+        query,
+    };
+    bench_group.register(task_name, move |_| black_box(search_task.run()));
+}
+
+fn add_bench_task_fetch_all_strings(
+    bench_group: &mut BenchGroup,
+    bench_index: &BenchIndex,
+    query: RangeQuery,
+    range_low: u64,
+    range_high: u64,
+) {
+    let task_name = format!(
+        "string_fastfield_fetch_all_strings_[{}-{}]",
+        range_low, range_high
+    );
+
+    let search_task = FetchAllStringsSearchTask {
+        searcher: bench_index.searcher.clone(),
+        query,
+    };
+
+    bench_group.register(task_name, move |_| {
+        let result = black_box(search_task.run());
+        result.len()
+    });
+}
+
+fn add_bench_task_fetch_all_strings_from_doc(
+    bench_group: &mut BenchGroup,
+    bench_index: &BenchIndex,
+    query: RangeQuery,
+    range_low: u64,
+    range_high: u64,
+) {
+    let task_name = format!(
+        "string_doc_fetch_all_strings_[{}-{}]",
+        range_low, range_high
+    );
+
+    let search_task = FetchAllStringsFromDocTask {
+        searcher: bench_index.searcher.clone(),
+        query,
+    };
+
+    bench_group.register(task_name, move |_| {
+        let result = black_box(search_task.run());
+        result.len()
+    });
+}
+
+struct CountSearchTask {
+    searcher: Searcher,
+    query: RangeQuery,
+}
+
+impl CountSearchTask {
+    #[inline(never)]
+    pub fn run(&self) -> usize {
+        self.searcher.search(&self.query, &Count).unwrap()
+    }
+}
+
+struct DocSetSearchTask {
+    searcher: Searcher,
+    query: RangeQuery,
+}
+
+impl DocSetSearchTask {
+    #[inline(never)]
+    pub fn run(&self) -> usize {
+        let result = self.searcher.search(&self.query, &DocSetCollector).unwrap();
+        result.len()
+    }
+}
+
+struct FetchAllStringsSearchTask {
+    searcher: Searcher,
+    query: RangeQuery,
+}
+
+impl FetchAllStringsSearchTask {
+    #[inline(never)]
+    pub fn run(&self) -> Vec<String> {
+        let doc_addresses = self.searcher.search(&self.query, &DocSetCollector).unwrap();
+        let mut docs = doc_addresses.into_iter().collect::<Vec<_>>();
+        docs.sort();
+        let mut strings = Vec::with_capacity(docs.len());
+
+        for doc_address in docs {
+            let segment_reader = &self.searcher.segment_readers()[doc_address.segment_ord as usize];
+            let str_column_opt = segment_reader.fast_fields().str("str_fast");
+
+            if let Ok(Some(str_column)) = str_column_opt {
+                let doc_id = doc_address.doc_id;
+                let term_ord = str_column.term_ords(doc_id).next().unwrap();
+                let mut str_buffer = String::new();
+                if str_column.ord_to_str(term_ord, &mut str_buffer).is_ok() {
+                    strings.push(str_buffer);
+                }
+            }
+        }
+
+        strings
+    }
+}
+
+struct FetchAllStringsFromDocTask {
+    searcher: Searcher,
+    query: RangeQuery,
+}
+
+impl FetchAllStringsFromDocTask {
+    #[inline(never)]
+    pub fn run(&self) -> Vec<String> {
+        let doc_addresses = self.searcher.search(&self.query, &DocSetCollector).unwrap();
+        let mut docs = doc_addresses.into_iter().collect::<Vec<_>>();
+        docs.sort();
+        let mut strings = Vec::with_capacity(docs.len());
+
+        let str_stored_field = self
+            .searcher
+            .schema()
+            .get_field("str_stored")
+            .expect("str_stored field should exist");
+
+        for doc_address in docs {
+            // Get the document from the doc store (row store access)
+            if let Ok(doc) = self.searcher.doc::<TantivyDocument>(doc_address) {
+                // Extract string values from the stored field
+                if let Some(field_value) = doc.get_first(str_stored_field) {
+                    if let Some(text) = field_value.as_value().as_str() {
+                        strings.push(text.to_string());
+                    }
+                }
+            }
+        }
+
+        strings
+    }
+}

bitpacker/Cargo.toml

@@ -18,5 +18,5 @@ homepage = "https://github.com/quickwit-oss/tantivy"
 bitpacking = { version = "0.9.2", default-features = false, features = ["bitpacker1x"] }
 
 [dev-dependencies]
-rand = "0.8"
+rand = "0.9"
 proptest = "1"

bitpacker/benches/bench.rs

@@ -4,8 +4,8 @@ extern crate test;
 
 #[cfg(test)]
 mod tests {
+    use rand::rng;
     use rand::seq::IteratorRandom;
-    use rand::thread_rng;
     use tantivy_bitpacker::{BitPacker, BitUnpacker, BlockedBitpacker};
     use test::Bencher;
 
@@ -27,7 +27,7 @@ mod tests {
         let num_els = 1_000_000u32;
         let bit_unpacker = BitUnpacker::new(bit_width);
         let data = create_bitpacked_data(bit_width, num_els);
-        let idxs: Vec<u32> = (0..num_els).choose_multiple(&mut thread_rng(), 100_000);
+        let idxs: Vec<u32> = (0..num_els).choose_multiple(&mut rng(), 100_000);
         b.iter(|| {
             let mut out = 0u64;
             for &idx in &idxs {

bitpacker/src/bitpacker.rs

@@ -258,7 +258,7 @@ mod test {
             bitpacker.write(val, num_bits, &mut data).unwrap();
         }
         bitpacker.close(&mut data).unwrap();
-        assert_eq!(data.len(), ((num_bits as usize) * len + 7) / 8);
+        assert_eq!(data.len(), ((num_bits as usize) * len).div_ceil(8));
         let bitunpacker = BitUnpacker::new(num_bits);
         (bitunpacker, vals, data)
     }
@@ -304,7 +304,7 @@ mod test {
             bitpacker.write(val, num_bits, &mut buffer).unwrap();
         }
         bitpacker.flush(&mut buffer).unwrap();
-        assert_eq!(buffer.len(), (vals.len() * num_bits as usize + 7) / 8);
+        assert_eq!(buffer.len(), (vals.len() * num_bits as usize).div_ceil(8));
         let bitunpacker = BitUnpacker::new(num_bits);
         let max_val = if num_bits == 64 {
             u64::MAX

bitpacker/src/filter_vec/avx2.rs

@@ -19,7 +19,7 @@ fn u32_to_i32(val: u32) -> i32 {
 #[inline]
 unsafe fn u32_to_i32_avx2(vals_u32x8s: DataType) -> DataType {
     const HIGHEST_BIT_MASK: DataType = from_u32x8([HIGHEST_BIT; NUM_LANES]);
-    op_xor(vals_u32x8s, HIGHEST_BIT_MASK)
+    unsafe { op_xor(vals_u32x8s, HIGHEST_BIT_MASK) }
 }
 
 pub fn filter_vec_in_place(range: RangeInclusive<u32>, offset: u32, output: &mut Vec<u32>) {
@@ -66,17 +66,19 @@ unsafe fn filter_vec_avx2_aux(
     ]);
     const SHIFT: __m256i = from_u32x8([NUM_LANES as u32; NUM_LANES]);
     for _ in 0..num_words {
-        let word = load_unaligned(input);
-        let word = u32_to_i32_avx2(word);
-        let keeper_bitset = compute_filter_bitset(word, range_simd.clone());
-        let added_len = keeper_bitset.count_ones();
-        let filtered_doc_ids = compact(ids, keeper_bitset);
-        store_unaligned(output_tail as *mut __m256i, filtered_doc_ids);
-        output_tail = output_tail.offset(added_len as isize);
-        ids = op_add(ids, SHIFT);
-        input = input.offset(1);
+        unsafe {
+            let word = load_unaligned(input);
+            let word = u32_to_i32_avx2(word);
+            let keeper_bitset = compute_filter_bitset(word, range_simd.clone());
+            let added_len = keeper_bitset.count_ones();
+            let filtered_doc_ids = compact(ids, keeper_bitset);
+            store_unaligned(output_tail as *mut __m256i, filtered_doc_ids);
+            output_tail = output_tail.offset(added_len as isize);
+            ids = op_add(ids, SHIFT);
+            input = input.offset(1);
+        }
     }
-    output_tail.offset_from(output) as usize
+    unsafe { output_tail.offset_from(output) as usize }
 }
 
 #[inline]
@@ -92,8 +94,7 @@ unsafe fn compute_filter_bitset(val: __m256i, range: std::ops::RangeInclusive<__
     let too_low = op_greater(*range.start(), val);
     let too_high = op_greater(val, *range.end());
     let inside = op_or(too_low, too_high);
-    255 - std::arch::x86_64::_mm256_movemask_ps(std::mem::transmute::<DataType, __m256>(inside))
-        as u8
+    255 - std::arch::x86_64::_mm256_movemask_ps(_mm256_castsi256_ps(inside)) as u8
 }
 
 union U8x32 {

columnar/Cargo.toml

@@ -22,7 +22,7 @@ downcast-rs = "2.0.1"
 [dev-dependencies]
 proptest = "1"
 more-asserts = "0.3.1"
-rand = "0.8"
+rand = "0.9"
 binggan = "0.14.0"
 
 [[bench]]

columnar/README.md

@@ -73,7 +73,7 @@ The crate introduces the following concepts.
 `Columnar` is an equivalent of a dataframe.
 It maps `column_key` to `Column`.
 
-A `Column<T>` asssociates a `RowId` (u32) to any
+A `Column<T>` associates a `RowId` (u32) to any
 number of values.
 
 This is made possible by wrapping a `ColumnIndex` and a `ColumnValue` object.

columnar/benches/bench_column_values_get.rs

@@ -9,7 +9,7 @@ use tantivy_columnar::column_values::{CodecType, serialize_and_load_u64_based_co
 fn get_data() -> Vec<u64> {
     let mut rng = StdRng::seed_from_u64(2u64);
     let mut data: Vec<_> = (100..55_000_u64)
-        .map(|num| num + rng.r#gen::<u8>() as u64)
+        .map(|num| num + rng.random::<u8>() as u64)
         .collect();
     data.push(99_000);
     data.insert(1000, 2000);

columnar/benches/bench_create_column_values.rs

@@ -6,7 +6,7 @@ use tantivy_columnar::column_values::{CodecType, serialize_u64_based_column_valu
 fn get_data() -> Vec<u64> {
     let mut rng = StdRng::seed_from_u64(2u64);
     let mut data: Vec<_> = (100..55_000_u64)
-        .map(|num| num + rng.r#gen::<u8>() as u64)
+        .map(|num| num + rng.random::<u8>() as u64)
         .collect();
     data.push(99_000);
     data.insert(1000, 2000);

columnar/benches/bench_first_vals.rs

@@ -89,13 +89,6 @@ fn main() {
         black_box(sum);
     });
 
-    group.register("first_block_fetch", |column| {
-        let mut block: Vec<Option<u64>> = vec![None; 64];
-        let fetch_docids = (0..64).collect::<Vec<_>>();
-        column.first_vals(&fetch_docids, &mut block);
-        black_box(block[0]);
-    });
-
     group.register("first_block_single_calls", |column| {
         let mut block: Vec<Option<u64>> = vec![None; 64];
         let fetch_docids = (0..64).collect::<Vec<_>>();

columnar/benches/bench_optional_index.rs

@@ -8,7 +8,7 @@ const TOTAL_NUM_VALUES: u32 = 1_000_000;
 fn gen_optional_index(fill_ratio: f64) -> OptionalIndex {
     let mut rng: StdRng = StdRng::from_seed([1u8; 32]);
     let vals: Vec<u32> = (0..TOTAL_NUM_VALUES)
-        .map(|_| rng.gen_bool(fill_ratio))
+        .map(|_| rng.random_bool(fill_ratio))
         .enumerate()
         .filter(|(_pos, val)| *val)
         .map(|(pos, _)| pos as u32)
@@ -25,7 +25,7 @@ fn random_range_iterator(
     let mut rng: StdRng = StdRng::from_seed([1u8; 32]);
     let mut current = start;
     std::iter::from_fn(move || {
-        current += rng.gen_range(avg_step_size - avg_deviation..=avg_step_size + avg_deviation);
+        current += rng.random_range(avg_step_size - avg_deviation..=avg_step_size + avg_deviation);
         if current >= end { None } else { Some(current) }
     })
 }

columnar/benches/bench_values_u128.rs

@@ -39,7 +39,7 @@ fn get_data_50percent_item() -> Vec<u128> {
 
     let mut data = vec![];
     for _ in 0..300_000 {
-        let val = rng.gen_range(1..=100);
+        let val = rng.random_range(1..=100);
         data.push(val);
     }
     data.push(SINGLE_ITEM);

columnar/benches/bench_values_u64.rs

@@ -34,7 +34,7 @@ fn get_data_50percent_item() -> Vec<u128> {
 
     let mut data = vec![];
     for _ in 0..300_000 {
-        let val = rng.gen_range(1..=100);
+        let val = rng.random_range(1..=100);
         data.push(val);
     }
     data.push(SINGLE_ITEM);

columnar/src/block_accessor.rs

@@ -29,12 +29,20 @@ impl<T: PartialOrd + Copy + std::fmt::Debug + Send + Sync + 'static + Default>
         }
     }
     #[inline]
-    pub fn fetch_block_with_missing(&mut self, docs: &[u32], accessor: &Column<T>, missing: T) {
+    pub fn fetch_block_with_missing(
+        &mut self,
+        docs: &[u32],
+        accessor: &Column<T>,
+        missing: Option<T>,
+    ) {
         self.fetch_block(docs, accessor);
         // no missing values
         if accessor.index.get_cardinality().is_full() {
             return;
         }
+        let Some(missing) = missing else {
+            return;
+        };
 
         // We can compare docid_cache length with docs to find missing docs
         // For multi value columns we can't rely on the length and always need to scan

columnar/src/column/mod.rs

@@ -85,8 +85,8 @@ impl<T: PartialOrd + Copy + Debug + Send + Sync + 'static> Column<T> {
     }
 
     #[inline]
-    pub fn first(&self, row_id: RowId) -> Option<T> {
-        self.values_for_doc(row_id).next()
+    pub fn first(&self, doc_id: DocId) -> Option<T> {
+        self.values_for_doc(doc_id).next()
     }
 
     /// Load the first value for each docid in the provided slice.
@@ -131,6 +131,8 @@ impl<T: PartialOrd + Copy + Debug + Send + Sync + 'static> Column<T> {
         self.index.docids_to_rowids(doc_ids, doc_ids_out, row_ids)
     }
 
+    /// Get an iterator over the values for the provided docid.
+    #[inline]
     pub fn values_for_doc(&self, doc_id: DocId) -> impl Iterator<Item = T> + '_ {
         self.index
             .value_row_ids(doc_id)
@@ -158,15 +160,6 @@ impl<T: PartialOrd + Copy + Debug + Send + Sync + 'static> Column<T> {
             .select_batch_in_place(selected_docid_range.start, doc_ids);
     }
 
-    /// Fills the output vector with the (possibly multiple values that are associated_with
-    /// `row_id`.
-    ///
-    /// This method clears the `output` vector.
-    pub fn fill_vals(&self, row_id: RowId, output: &mut Vec<T>) {
-        output.clear();
-        output.extend(self.values_for_doc(row_id));
-    }
-
     pub fn first_or_default_col(self, default_value: T) -> Arc<dyn ColumnValues<T>> {
         Arc::new(FirstValueWithDefault {
             column: self,

columnar/src/column_values/monotonic_mapping_u128.rs

@@ -1,7 +1,7 @@
 use std::fmt::Debug;
 use std::net::Ipv6Addr;
 
-/// Montonic maps a value to u128 value space
+/// Monotonic maps a value to u128 value space
 /// Monotonic mapping enables `PartialOrd` on u128 space without conversion to original space.
 pub trait MonotonicallyMappableToU128: 'static + PartialOrd + Copy + Debug + Send + Sync {
     /// Converts a value to u128.

columnar/src/column_values/u64_based/bitpacked.rs

@@ -41,12 +41,6 @@ fn transform_range_before_linear_transformation(
     if range.is_empty() {
         return None;
     }
-    if stats.min_value > *range.end() {
-        return None;
-    }
-    if stats.max_value < *range.start() {
-        return None;
-    }
     let shifted_range =
         range.start().saturating_sub(stats.min_value)..=range.end().saturating_sub(stats.min_value);
     let start_before_gcd_multiplication: u64 = div_ceil(*shifted_range.start(), stats.gcd);

columnar/src/column_values/u64_based/line.rs

@@ -8,7 +8,7 @@ use crate::column_values::ColumnValues;
 const MID_POINT: u64 = (1u64 << 32) - 1u64;
 
 /// `Line` describes a line function `y: ax + b` using integer
-/// arithmetics.
+/// arithmetic.
 ///
 /// The slope is in fact a decimal split into a 32 bit integer value,
 /// and a 32-bit decimal value.
@@ -94,7 +94,7 @@ impl Line {
         // `(i, ys[])`.
         //
         // The best intercept therefore has the form
-        // `y[i] - line.eval(i)` (using wrapping arithmetics).
+        // `y[i] - line.eval(i)` (using wrapping arithmetic).
         // In other words, the best intercept is one of the `y - Line::eval(ys[i])`
         // and our task is just to pick the one that minimizes our error.
         //

columnar/src/column_values/u64_based/linear.rs

@@ -268,7 +268,7 @@ mod tests {
 
     #[test]
     fn linear_interpol_fast_field_rand() {
-        let mut rng = rand::thread_rng();
+        let mut rng = rand::rng();
         for _ in 0..50 {
             let mut data = (0..10_000).map(|_| rng.next_u64()).collect::<Vec<_>>();
             create_and_validate::<LinearCodec>(&data, "random");

columnar/src/column_values/u64_based/mod.rs

@@ -52,7 +52,7 @@ pub trait ColumnCodecEstimator<T = u64>: 'static {
     ) -> io::Result<()>;
 }
 
-/// A column codec describes a colunm serialization format.
+/// A column codec describes a column serialization format.
 pub trait ColumnCodec<T: PartialOrd = u64> {
     /// Specialized `ColumnValues` type.
     type ColumnValues: ColumnValues<T> + 'static;

columnar/src/column_values/u64_based/tests.rs

@@ -122,7 +122,7 @@ pub(crate) fn create_and_validate<TColumnCodec: ColumnCodec>(
     assert_eq!(vals, buffer);
 
     if !vals.is_empty() {
-        let test_rand_idx = rand::thread_rng().gen_range(0..=vals.len() - 1);
+        let test_rand_idx = rand::rng().random_range(0..=vals.len() - 1);
         let expected_positions: Vec<u32> = vals
             .iter()
             .enumerate()

columnar/src/dynamic_column.rs

@@ -3,7 +3,8 @@ use std::sync::Arc;
 use std::{fmt, io};
 
 use common::file_slice::FileSlice;
-use common::{ByteCount, DateTime, HasLen, OwnedBytes};
+use common::{ByteCount, DateTime, OwnedBytes};
+use serde::{Deserialize, Serialize};
 
 use crate::column::{BytesColumn, Column, StrColumn};
 use crate::column_values::{StrictlyMonotonicFn, monotonic_map_column};
@@ -317,10 +318,89 @@ impl DynamicColumnHandle {
     }
 
     pub fn num_bytes(&self) -> ByteCount {
-        self.file_slice.len().into()
+        self.file_slice.num_bytes()
+    }
+
+    /// Legacy helper returning the column space usage.
+    pub fn column_and_dictionary_num_bytes(&self) -> io::Result<ColumnSpaceUsage> {
+        self.space_usage()
+    }
+
+    /// Return the space usage of the column, optionally broken down by dictionary and column
+    /// values.
+    ///
+    /// For dictionary encoded columns (strings and bytes), this splits the total footprint into
+    /// the dictionary and the remaining column data (including index and values).
+    /// For all other column types, the dictionary size is `None` and the column size
+    /// equals the total bytes.
+    pub fn space_usage(&self) -> io::Result<ColumnSpaceUsage> {
+        let total_num_bytes = self.num_bytes();
+        let dynamic_column = self.open()?;
+        let dictionary_num_bytes = match &dynamic_column {
+            DynamicColumn::Bytes(bytes_column) => bytes_column.dictionary().num_bytes(),
+            DynamicColumn::Str(str_column) => str_column.dictionary().num_bytes(),
+            _ => {
+                return Ok(ColumnSpaceUsage::new(self.num_bytes(), None));
+            }
+        };
+        assert!(dictionary_num_bytes <= total_num_bytes);
+        let column_num_bytes =
+            ByteCount::from(total_num_bytes.get_bytes() - dictionary_num_bytes.get_bytes());
+        Ok(ColumnSpaceUsage::new(
+            column_num_bytes,
+            Some(dictionary_num_bytes),
+        ))
     }
 
     pub fn column_type(&self) -> ColumnType {
         self.column_type
     }
 }
+
+/// Represents space usage of a column.
+///
+/// `column_num_bytes` tracks the column payload (index, values and footer).
+/// For dictionary encoded columns, `dictionary_num_bytes` captures the dictionary footprint.
+/// [`ColumnSpaceUsage::total_num_bytes`] returns the sum of both parts.
+#[derive(Clone, Debug, Serialize, Deserialize)]
+pub struct ColumnSpaceUsage {
+    column_num_bytes: ByteCount,
+    dictionary_num_bytes: Option<ByteCount>,
+}
+
+impl ColumnSpaceUsage {
+    pub(crate) fn new(
+        column_num_bytes: ByteCount,
+        dictionary_num_bytes: Option<ByteCount>,
+    ) -> Self {
+        ColumnSpaceUsage {
+            column_num_bytes,
+            dictionary_num_bytes,
+        }
+    }
+
+    pub fn column_num_bytes(&self) -> ByteCount {
+        self.column_num_bytes
+    }
+
+    pub fn dictionary_num_bytes(&self) -> Option<ByteCount> {
+        self.dictionary_num_bytes
+    }
+
+    pub fn total_num_bytes(&self) -> ByteCount {
+        self.column_num_bytes + self.dictionary_num_bytes.unwrap_or_default()
+    }
+
+    /// Merge two space usage values by summing their components.
+    pub fn merge(&self, other: &ColumnSpaceUsage) -> ColumnSpaceUsage {
+        let dictionary_num_bytes = match (self.dictionary_num_bytes, other.dictionary_num_bytes) {
+            (Some(lhs), Some(rhs)) => Some(lhs + rhs),
+            (Some(val), None) | (None, Some(val)) => Some(val),
+            (None, None) => None,
+        };
+        ColumnSpaceUsage {
+            column_num_bytes: self.column_num_bytes + other.column_num_bytes,
+            dictionary_num_bytes,
+        }
+    }
+}

columnar/src/lib.rs

@@ -48,7 +48,7 @@ pub use columnar::{
 use sstable::VoidSSTable;
 pub use value::{NumericalType, NumericalValue};
 
-pub use self::dynamic_column::{DynamicColumn, DynamicColumnHandle};
+pub use self::dynamic_column::{ColumnSpaceUsage, DynamicColumn, DynamicColumnHandle};
 
 pub type RowId = u32;
 pub type DocId = u32;
@@ -59,7 +59,7 @@ pub struct RowAddr {
     pub row_id: RowId,
 }
 
-pub use sstable::Dictionary;
+pub use sstable::{Dictionary, TermOrdHit};
 pub type Streamer<'a> = sstable::Streamer<'a, VoidSSTable>;
 
 pub use common::DateTime;

columnar/src/tests.rs

@@ -60,7 +60,7 @@ fn test_dataframe_writer_bool() {
     let DynamicColumn::Bool(bool_col) = dyn_bool_col else {
         panic!();
     };
-    let vals: Vec<Option<bool>> = (0..5).map(|row_id| bool_col.first(row_id)).collect();
+    let vals: Vec<Option<bool>> = (0..5).map(|doc_id| bool_col.first(doc_id)).collect();
     assert_eq!(&vals, &[None, Some(false), None, Some(true), None,]);
 }
 
@@ -108,7 +108,7 @@ fn test_dataframe_writer_ip_addr() {
     let DynamicColumn::IpAddr(ip_col) = dyn_bool_col else {
         panic!();
     };
-    let vals: Vec<Option<Ipv6Addr>> = (0..5).map(|row_id| ip_col.first(row_id)).collect();
+    let vals: Vec<Option<Ipv6Addr>> = (0..5).map(|doc_id| ip_col.first(doc_id)).collect();
     assert_eq!(
         &vals,
         &[
@@ -169,7 +169,7 @@ fn test_dictionary_encoded_str() {
     let DynamicColumn::Str(str_col) = col_handles[0].open().unwrap() else {
         panic!();
     };
-    let index: Vec<Option<u64>> = (0..5).map(|row_id| str_col.ords().first(row_id)).collect();
+    let index: Vec<Option<u64>> = (0..5).map(|doc_id| str_col.ords().first(doc_id)).collect();
     assert_eq!(index, &[None, Some(0), None, Some(2), Some(1)]);
     assert_eq!(str_col.num_rows(), 5);
     let mut term_buffer = String::new();
@@ -204,7 +204,7 @@ fn test_dictionary_encoded_bytes() {
         panic!();
     };
     let index: Vec<Option<u64>> = (0..5)
-        .map(|row_id| bytes_col.ords().first(row_id))
+        .map(|doc_id| bytes_col.ords().first(doc_id))
         .collect();
     assert_eq!(index, &[None, Some(0), None, Some(2), Some(1)]);
     assert_eq!(bytes_col.num_rows(), 5);

common/Cargo.toml

@@ -15,11 +15,10 @@ repository = "https://github.com/quickwit-oss/tantivy"
 byteorder = "1.4.3"
 ownedbytes = { version= "0.9", path="../ownedbytes" }
 async-trait = "0.1"
-time = { version = "0.3.10", features = ["serde-well-known"] }
+time = { version = "0.3.47", features = ["serde-well-known"] }
 serde = { version = "1.0.136", features = ["derive"] }
 
 [dev-dependencies]
 binggan = "0.14.0"
 proptest = "1.0.0"
-rand = "0.8.4"
-
+rand = "0.9"

common/benches/bench.rs

@@ -1,6 +1,6 @@
 use binggan::{BenchRunner, black_box};
+use rand::rng;
 use rand::seq::IteratorRandom;
-use rand::thread_rng;
 use tantivy_common::{BitSet, TinySet, serialize_vint_u32};
 
 fn bench_vint() {
@@ -17,7 +17,7 @@ fn bench_vint() {
         black_box(out);
     });
 
-    let vals: Vec<u32> = (0..20_000).choose_multiple(&mut thread_rng(), 100_000);
+    let vals: Vec<u32> = (0..20_000).choose_multiple(&mut rng(), 100_000);
     runner.bench_function("bench_vint_rand", move |_| {
         let mut out = 0u64;
         for val in vals.iter().cloned() {

common/src/bitset.rs

@@ -181,6 +181,14 @@ pub struct BitSet {
     len: u64,
     max_value: u32,
 }
+impl std::fmt::Debug for BitSet {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        f.debug_struct("BitSet")
+            .field("len", &self.len)
+            .field("max_value", &self.max_value)
+            .finish()
+    }
+}
 
 fn num_buckets(max_val: u32) -> u32 {
     max_val.div_ceil(64u32)
@@ -408,7 +416,7 @@ mod tests {
     use std::collections::HashSet;
 
     use ownedbytes::OwnedBytes;
-    use rand::distributions::Bernoulli;
+    use rand::distr::Bernoulli;
     use rand::rngs::StdRng;
     use rand::{Rng, SeedableRng};
 

common/src/vint.rs

@@ -28,7 +28,9 @@ impl BinarySerializable for VIntU128 {
         writer.write_all(&buffer)
     }
 
+    #[allow(clippy::unbuffered_bytes)]
     fn deserialize<R: Read>(reader: &mut R) -> io::Result<Self> {
+        #[allow(clippy::unbuffered_bytes)]
         let mut bytes = reader.bytes();
         let mut result = 0u128;
         let mut shift = 0u64;
@@ -195,7 +197,9 @@ impl BinarySerializable for VInt {
         writer.write_all(&buffer[0..num_bytes])
     }
 
+    #[allow(clippy::unbuffered_bytes)]
     fn deserialize<R: Read>(reader: &mut R) -> io::Result<Self> {
+        #[allow(clippy::unbuffered_bytes)]
         let mut bytes = reader.bytes();
         let mut result = 0u64;
         let mut shift = 0u64;

common/src/writer.rs

@@ -62,7 +62,9 @@ impl<W: TerminatingWrite> TerminatingWrite for CountingWriter<W> {
 pub struct AntiCallToken(());
 
 /// Trait used to indicate when no more write need to be done on a writer
-pub trait TerminatingWrite: Write + Send + Sync {
+///
+/// Thread-safety is enforced at the call sites that require it.
+pub trait TerminatingWrite: Write {
     /// Indicate that the writer will no longer be used. Internally call terminate_ref.
     fn terminate(mut self) -> io::Result<()>
     where Self: Sized {

doc/src/json.md

@@ -60,7 +60,7 @@ At indexing, tantivy will try to interpret number and strings as different type
 priority order.
 
 Numbers will be interpreted as u64, i64 and f64 in that order.
-Strings will be interpreted as rfc3999 dates or simple strings.
+Strings will be interpreted as rfc3339 dates or simple strings.
 
 The first working type is picked and is the only term that is emitted for indexing.
 Note this interpretation happens on a per-document basis, and there is no effort to try to sniff
@@ -81,7 +81,7 @@ Will be interpreted as
 (my_path.my_segment, String, 233) or (my_path.my_segment, u64, 233)
 ```
 
-Likewise, we need to emit two tokens if the query contains an rfc3999 date.
+Likewise, we need to emit two tokens if the query contains an rfc3339 date.
 Indeed the date could have been actually a single token inside the text of a document at ingestion time. Generally speaking, we will always at least emit a string token in query parsing, and sometimes more.
 
 If one more json field is defined, things get even more complicated.

examples/basic_search.rs

@@ -208,7 +208,7 @@ fn main() -> tantivy::Result<()> {
     // is the role of the `TopDocs` collector.
 
     // We can now perform our query.
-    let top_docs = searcher.search(&query, &TopDocs::with_limit(10))?;
+    let top_docs = searcher.search(&query, &TopDocs::with_limit(10).order_by_score())?;
 
     // The actual documents still need to be
     // retrieved from Tantivy's store.
@@ -226,7 +226,7 @@ fn main() -> tantivy::Result<()> {
     let query = query_parser.parse_query("title:sea^20 body:whale^70")?;
 
     let (_score, doc_address) = searcher
-        .search(&query, &TopDocs::with_limit(1))?
+        .search(&query, &TopDocs::with_limit(1).order_by_score())?
         .into_iter()
         .next()
         .unwrap();

examples/custom_tokenizer.rs

@@ -100,7 +100,7 @@ fn main() -> tantivy::Result<()> {
     // here we want to get a hit on the 'ken' in Frankenstein
     let query = query_parser.parse_query("ken")?;
 
-    let top_docs = searcher.search(&query, &TopDocs::with_limit(10))?;
+    let top_docs = searcher.search(&query, &TopDocs::with_limit(10).order_by_score())?;
 
     for (_, doc_address) in top_docs {
         let retrieved_doc: TantivyDocument = searcher.doc(doc_address)?;

examples/date_time_field.rs

@@ -50,14 +50,14 @@ fn main() -> tantivy::Result<()> {
     {
         // Simple exact search on the date
         let query = query_parser.parse_query("occurred_at:\"2022-06-22T12:53:50.53Z\"")?;
-        let count_docs = searcher.search(&*query, &TopDocs::with_limit(5))?;
+        let count_docs = searcher.search(&*query, &TopDocs::with_limit(5).order_by_score())?;
         assert_eq!(count_docs.len(), 1);
     }
     {
         // Range query on the date field
         let query = query_parser
             .parse_query(r#"occurred_at:[2022-06-22T12:58:00Z TO 2022-06-23T00:00:00Z}"#)?;
-        let count_docs = searcher.search(&*query, &TopDocs::with_limit(4))?;
+        let count_docs = searcher.search(&*query, &TopDocs::with_limit(4).order_by_score())?;
         assert_eq!(count_docs.len(), 1);
         for (_score, doc_address) in count_docs {
             let retrieved_doc = searcher.doc::<TantivyDocument>(doc_address)?;

examples/deleting_updating_documents.rs

@@ -28,7 +28,7 @@ fn extract_doc_given_isbn(
     // The second argument is here to tell we don't care about decoding positions,
     // or term frequencies.
     let term_query = TermQuery::new(isbn_term.clone(), IndexRecordOption::Basic);
-    let top_docs = searcher.search(&term_query, &TopDocs::with_limit(1))?;
+    let top_docs = searcher.search(&term_query, &TopDocs::with_limit(1).order_by_score())?;
 
     if let Some((_score, doc_address)) = top_docs.first() {
         let doc = searcher.doc(*doc_address)?;

examples/fuzzy_search.rs

@@ -145,7 +145,7 @@ fn main() -> tantivy::Result<()> {
         let query = FuzzyTermQuery::new(term, 2, true);
 
         let (top_docs, count) = searcher
-            .search(&query, &(TopDocs::with_limit(5), Count))
+            .search(&query, &(TopDocs::with_limit(5).order_by_score(), Count))
             .unwrap();
         assert_eq!(count, 3);
         assert_eq!(top_docs.len(), 3);

examples/ip_field.rs

@@ -69,25 +69,25 @@ fn main() -> tantivy::Result<()> {
     {
         // Inclusive range queries
         let query = query_parser.parse_query("ip:[192.168.0.80 TO 192.168.0.100]")?;
-        let count_docs = searcher.search(&*query, &TopDocs::with_limit(5))?;
+        let count_docs = searcher.search(&*query, &TopDocs::with_limit(5).order_by_score())?;
         assert_eq!(count_docs.len(), 1);
     }
     {
         // Exclusive range queries
         let query = query_parser.parse_query("ip:{192.168.0.80 TO 192.168.1.100]")?;
-        let count_docs = searcher.search(&*query, &TopDocs::with_limit(2))?;
+        let count_docs = searcher.search(&*query, &TopDocs::with_limit(2).order_by_score())?;
         assert_eq!(count_docs.len(), 0);
     }
     {
         // Find docs with IP addresses smaller equal 192.168.1.100
         let query = query_parser.parse_query("ip:[* TO 192.168.1.100]")?;
-        let count_docs = searcher.search(&*query, &TopDocs::with_limit(2))?;
+        let count_docs = searcher.search(&*query, &TopDocs::with_limit(2).order_by_score())?;
         assert_eq!(count_docs.len(), 2);
     }
     {
         // Find docs with IP addresses smaller than 192.168.1.100
         let query = query_parser.parse_query("ip:[* TO 192.168.1.100}")?;
-        let count_docs = searcher.search(&*query, &TopDocs::with_limit(2))?;
+        let count_docs = searcher.search(&*query, &TopDocs::with_limit(2).order_by_score())?;
         assert_eq!(count_docs.len(), 2);
     }
 

examples/json_field.rs

@@ -59,12 +59,12 @@ fn main() -> tantivy::Result<()> {
     let query_parser = QueryParser::for_index(&index, vec![event_type, attributes]);
     {
         let query = query_parser.parse_query("target:submit-button")?;
-        let count_docs = searcher.search(&*query, &TopDocs::with_limit(2))?;
+        let count_docs = searcher.search(&*query, &TopDocs::with_limit(2).order_by_score())?;
         assert_eq!(count_docs.len(), 2);
     }
     {
         let query = query_parser.parse_query("target:submit")?;
-        let count_docs = searcher.search(&*query, &TopDocs::with_limit(2))?;
+        let count_docs = searcher.search(&*query, &TopDocs::with_limit(2).order_by_score())?;
         assert_eq!(count_docs.len(), 2);
     }
     {
@@ -74,33 +74,33 @@ fn main() -> tantivy::Result<()> {
     }
     {
         let query = query_parser.parse_query("click AND cart.product_id:133")?;
-        let hits = searcher.search(&*query, &TopDocs::with_limit(2))?;
+        let hits = searcher.search(&*query, &TopDocs::with_limit(2).order_by_score())?;
         assert_eq!(hits.len(), 1);
     }
     {
         // The sub-fields in the json field marked as default field still need to be explicitly
         // addressed
         let query = query_parser.parse_query("click AND 133")?;
-        let hits = searcher.search(&*query, &TopDocs::with_limit(2))?;
+        let hits = searcher.search(&*query, &TopDocs::with_limit(2).order_by_score())?;
         assert_eq!(hits.len(), 0);
     }
     {
         // Default json fields are ignored if they collide with the schema
         let query = query_parser.parse_query("event_type:holiday-sale")?;
-        let hits = searcher.search(&*query, &TopDocs::with_limit(2))?;
+        let hits = searcher.search(&*query, &TopDocs::with_limit(2).order_by_score())?;
         assert_eq!(hits.len(), 0);
     }
     // # Query via full attribute path
     {
         // This only searches in our schema's `event_type` field
         let query = query_parser.parse_query("event_type:click")?;
-        let hits = searcher.search(&*query, &TopDocs::with_limit(2))?;
+        let hits = searcher.search(&*query, &TopDocs::with_limit(2).order_by_score())?;
         assert_eq!(hits.len(), 2);
     }
     {
         // Default json fields can still be accessed by full path
         let query = query_parser.parse_query("attributes.event_type:holiday-sale")?;
-        let hits = searcher.search(&*query, &TopDocs::with_limit(2))?;
+        let hits = searcher.search(&*query, &TopDocs::with_limit(2).order_by_score())?;
         assert_eq!(hits.len(), 1);
     }
     Ok(())

examples/phrase_prefix_search.rs

@@ -63,7 +63,7 @@ fn main() -> Result<()> {
     // but not "in the Gulf Stream".
     let query = query_parser.parse_query("\"in the su\"*")?;
 
-    let top_docs = searcher.search(&query, &TopDocs::with_limit(10))?;
+    let top_docs = searcher.search(&query, &TopDocs::with_limit(10).order_by_score())?;
     let mut titles = top_docs
         .into_iter()
         .map(|(_score, doc_address)| {

examples/pre_tokenized_text.rs

@@ -107,7 +107,8 @@ fn main() -> tantivy::Result<()> {
         IndexRecordOption::Basic,
     );
 
-    let (top_docs, count) = searcher.search(&query, &(TopDocs::with_limit(2), Count))?;
+    let (top_docs, count) =
+        searcher.search(&query, &(TopDocs::with_limit(2).order_by_score(), Count))?;
 
     assert_eq!(count, 2);
 
@@ -128,7 +129,8 @@ fn main() -> tantivy::Result<()> {
         IndexRecordOption::Basic,
     );
 
-    let (_top_docs, count) = searcher.search(&query, &(TopDocs::with_limit(2), Count))?;
+    let (_top_docs, count) =
+        searcher.search(&query, &(TopDocs::with_limit(2).order_by_score(), Count))?;
 
     assert_eq!(count, 0);
 

examples/snippet.rs

@@ -50,7 +50,7 @@ fn main() -> tantivy::Result<()> {
     let query_parser = QueryParser::for_index(&index, vec![title, body]);
     let query = query_parser.parse_query("sycamore spring")?;
 
-    let top_docs = searcher.search(&query, &TopDocs::with_limit(10))?;
+    let top_docs = searcher.search(&query, &TopDocs::with_limit(10).order_by_score())?;
 
     let snippet_generator = SnippetGenerator::create(&searcher, &*query, body)?;
 

examples/stop_words.rs

@@ -102,7 +102,7 @@ fn main() -> tantivy::Result<()> {
     // stop words are applied on the query as well.
     // The following will be equivalent to `title:frankenstein`
     let query = query_parser.parse_query("title:\"the Frankenstein\"")?;
-    let top_docs = searcher.search(&query, &TopDocs::with_limit(10))?;
+    let top_docs = searcher.search(&query, &TopDocs::with_limit(10).order_by_score())?;
 
     for (score, doc_address) in top_docs {
         let retrieved_doc: TantivyDocument = searcher.doc(doc_address)?;

examples/warmer.rs

@@ -164,7 +164,7 @@ fn main() -> tantivy::Result<()> {
         move |doc_id: DocId| Reverse(price[doc_id as usize])
     };
 
-    let most_expensive_first = TopDocs::with_limit(10).custom_score(score_by_price);
+    let most_expensive_first = TopDocs::with_limit(10).order_by(score_by_price);
 
     let hits = searcher.search(&query, &most_expensive_first)?;
     assert_eq!(

query-grammar/src/query_grammar.rs

@@ -560,7 +560,7 @@ fn range_infallible(inp: &str) -> JResult<&str, UserInputLeaf> {
             (
                 (
                     value((), tag(">=")),
-                    map(word_infallible("", false), |(bound, err)| {
+                    map(word_infallible(")", false), |(bound, err)| {
                         (
                             (
                                 bound
@@ -574,7 +574,7 @@ fn range_infallible(inp: &str) -> JResult<&str, UserInputLeaf> {
                 ),
                 (
                     value((), tag("<=")),
-                    map(word_infallible("", false), |(bound, err)| {
+                    map(word_infallible(")", false), |(bound, err)| {
                         (
                             (
                                 UserInputBound::Unbounded,
@@ -588,7 +588,7 @@ fn range_infallible(inp: &str) -> JResult<&str, UserInputLeaf> {
                 ),
                 (
                     value((), tag(">")),
-                    map(word_infallible("", false), |(bound, err)| {
+                    map(word_infallible(")", false), |(bound, err)| {
                         (
                             (
                                 bound
@@ -602,7 +602,7 @@ fn range_infallible(inp: &str) -> JResult<&str, UserInputLeaf> {
                 ),
                 (
                     value((), tag("<")),
-                    map(word_infallible("", false), |(bound, err)| {
+                    map(word_infallible(")", false), |(bound, err)| {
                         (
                             (
                                 UserInputBound::Unbounded,
@@ -704,7 +704,11 @@ fn regex(inp: &str) -> IResult<&str, UserInputLeaf> {
                 many1(alt((preceded(char('\\'), char('/')), none_of("/")))),
                 char('/'),
             ),
-            peek(alt((multispace1, eof))),
+            peek(alt((
+                value((), multispace1),
+                value((), char(')')),
+                value((), eof),
+            ))),
         ),
         |elements| UserInputLeaf::Regex {
             field: None,
@@ -721,8 +725,12 @@ fn regex_infallible(inp: &str) -> JResult<&str, UserInputLeaf> {
             opt_i_err(char('/'), "missing delimiter /"),
         ),
         opt_i_err(
-            peek(alt((multispace1, eof))),
-            "expected whitespace or end of input",
+            peek(alt((
+                value((), multispace1),
+                value((), char(')')),
+                value((), eof),
+            ))),
+            "expected whitespace, closing parenthesis, or end of input",
         ),
     )(inp)
     {
@@ -758,7 +766,17 @@ fn negate(expr: UserInputAst) -> UserInputAst {
 fn leaf(inp: &str) -> IResult<&str, UserInputAst> {
     alt((
         delimited(char('('), ast, char(')')),
-        map(char('*'), |_| UserInputAst::from(UserInputLeaf::All)),
+        map(
+            terminated(
+                char('*'),
+                peek(alt((
+                    value((), multispace1),
+                    value((), char(')')),
+                    value((), eof),
+                ))),
+            ),
+            |_| UserInputAst::from(UserInputLeaf::All),
+        ),
         map(preceded(tuple((tag("NOT"), multispace1)), leaf), negate),
         literal,
     ))(inp)
@@ -779,7 +797,17 @@ fn leaf_infallible(inp: &str) -> JResult<&str, Option<UserInputAst>> {
                 ),
             ),
             (
-                value((), char('*')),
+                value(
+                    (),
+                    terminated(
+                        char('*'),
+                        peek(alt((
+                            value((), multispace1),
+                            value((), char(')')),
+                            value((), eof),
+                        ))),
+                    ),
+                ),
                 map(nothing, |_| {
                     (Some(UserInputAst::from(UserInputLeaf::All)), Vec::new())
                 }),
@@ -1303,6 +1331,14 @@ mod test {
         test_parse_query_to_ast_helper("<a", "{\"*\" TO \"a\"}");
         test_parse_query_to_ast_helper("<=a", "{\"*\" TO \"a\"]");
         test_parse_query_to_ast_helper("<=bsd", "{\"*\" TO \"bsd\"]");
+
+        test_parse_query_to_ast_helper("(<=42)", "{\"*\" TO \"42\"]");
+        test_parse_query_to_ast_helper("(<=42 )", "{\"*\" TO \"42\"]");
+        test_parse_query_to_ast_helper("(age:>5)", "\"age\":{\"5\" TO \"*\"}");
+        test_parse_query_to_ast_helper(
+            "(title:bar AND age:>12)",
+            "(+\"title\":bar +\"age\":{\"12\" TO \"*\"})",
+        );
     }
 
     #[test]
@@ -1671,6 +1707,25 @@ mod test {
         test_parse_query_to_ast_helper("abc:a b", "(*\"abc\":a *b)");
         test_parse_query_to_ast_helper("abc:\"a b\"", "\"abc\":\"a b\"");
         test_parse_query_to_ast_helper("foo:[1 TO 5]", "\"foo\":[\"1\" TO \"5\"]");
+
+        // Phrase prefixed with *
+        test_parse_query_to_ast_helper("foo:(*A)", "\"foo\":*A");
+        test_parse_query_to_ast_helper("*A", "*A");
+        test_parse_query_to_ast_helper("(*A)", "*A");
+        test_parse_query_to_ast_helper("foo:(A OR B)", "(?\"foo\":A ?\"foo\":B)");
+        test_parse_query_to_ast_helper("foo:(A* OR B*)", "(?\"foo\":A* ?\"foo\":B*)");
+        test_parse_query_to_ast_helper("foo:(*A OR *B)", "(?\"foo\":*A ?\"foo\":*B)");
+
+        // Regexes between parentheses
+        test_parse_query_to_ast_helper("foo:(/A.*/)", "\"foo\":/A.*/");
+        test_parse_query_to_ast_helper("foo:(/A.*/ OR /B.*/)", "(?\"foo\":/A.*/ ?\"foo\":/B.*/)");
+    }
+
+    #[test]
+    fn test_parse_query_all() {
+        test_parse_query_to_ast_helper("*", "*");
+        test_parse_query_to_ast_helper("(*)", "*");
+        test_parse_query_to_ast_helper("(* )", "*");
     }
 
     #[test]

query-grammar/src/user_input_ast.rs

@@ -66,6 +66,7 @@ impl UserInputLeaf {
             }
             UserInputLeaf::Range { field, .. } if field.is_none() => *field = Some(default_field),
             UserInputLeaf::Set { field, .. } if field.is_none() => *field = Some(default_field),
+            UserInputLeaf::Regex { field, .. } if field.is_none() => *field = Some(default_field),
             _ => (), // field was already set, do nothing
         }
     }

src/aggregation/accessor_helpers.rs

@@ -16,15 +16,16 @@ use crate::index::SegmentReader;
 /// That way we can use it the same way as if it would come from the fastfield.
 pub(crate) fn get_missing_val_as_u64_lenient(
     column_type: ColumnType,
+    column_max_value: u64,
     missing: &Key,
     field_name: &str,
 ) -> crate::Result<Option<u64>> {
     let missing_val = match missing {
-        Key::Str(_) if column_type == ColumnType::Str => Some(u64::MAX),
+        Key::Str(_) if column_type == ColumnType::Str => Some(column_max_value + 1),
         // Allow fallback to number on text fields
-        Key::F64(_) if column_type == ColumnType::Str => Some(u64::MAX),
-        Key::U64(_) if column_type == ColumnType::Str => Some(u64::MAX),
-        Key::I64(_) if column_type == ColumnType::Str => Some(u64::MAX),
+        Key::F64(_) if column_type == ColumnType::Str => Some(column_max_value + 1),
+        Key::U64(_) if column_type == ColumnType::Str => Some(column_max_value + 1),
+        Key::I64(_) if column_type == ColumnType::Str => Some(column_max_value + 1),
         Key::F64(val) if column_type.numerical_type().is_some() => {
             f64_to_fastfield_u64(*val, &column_type)
         }

src/aggregation/agg_data.rs

@@ -1,4 +1,4 @@
-use columnar::{Column, ColumnType, StrColumn};
+use columnar::{Column, ColumnBlockAccessor, ColumnType, StrColumn};
 use common::BitSet;
 use rustc_hash::FxHashSet;
 use serde::Serialize;
@@ -10,16 +10,17 @@ use crate::aggregation::accessor_helpers::{
 };
 use crate::aggregation::agg_req::{Aggregation, AggregationVariants, Aggregations};
 use crate::aggregation::bucket::{
-    build_segment_aggregation_collector, FilterAggReqData, HistogramAggReqData, HistogramBounds,
-    IncludeExcludeParam, MissingTermAggReqData, RangeAggReqData, SegmentFilterCollector,
-    SegmentHistogramCollector, SegmentRangeCollector, TermMissingAgg, TermsAggReqData,
-    TermsAggregation, TermsAggregationInternal,
+    build_segment_filter_collector, build_segment_range_collector, CompositeAggReqData,
+    CompositeAggregation, CompositeSourceAccessors, FilterAggReqData, HistogramAggReqData,
+    HistogramBounds, IncludeExcludeParam, MissingTermAggReqData, RangeAggReqData,
+    SegmentHistogramCollector, TermMissingAgg, TermsAggReqData, TermsAggregation,
+    TermsAggregationInternal,
 };
 use crate::aggregation::metric::{
-    AverageAggregation, CardinalityAggReqData, CardinalityAggregationReq, CountAggregation,
-    ExtendedStatsAggregation, MaxAggregation, MetricAggReqData, MinAggregation,
-    SegmentCardinalityCollector, SegmentExtendedStatsCollector, SegmentPercentilesCollector,
-    SegmentStatsCollector, StatsAggregation, StatsType, SumAggregation, TopHitsAggReqData,
+    build_segment_stats_collector, AverageAggregation, CardinalityAggReqData,
+    CardinalityAggregationReq, CountAggregation, ExtendedStatsAggregation, MaxAggregation,
+    MetricAggReqData, MinAggregation, SegmentCardinalityCollector, SegmentExtendedStatsCollector,
+    SegmentPercentilesCollector, StatsAggregation, StatsType, SumAggregation, TopHitsAggReqData,
     TopHitsSegmentCollector,
 };
 use crate::aggregation::segment_agg_result::{
@@ -35,6 +36,7 @@ pub struct AggregationsSegmentCtx {
     /// Request data for each aggregation type.
     pub per_request: PerRequestAggSegCtx,
     pub context: AggContextParams,
+    pub column_block_accessor: ColumnBlockAccessor<u64>,
 }
 
 impl AggregationsSegmentCtx {
@@ -72,6 +74,12 @@ impl AggregationsSegmentCtx {
         self.per_request.filter_req_data.push(Some(Box::new(data)));
         self.per_request.filter_req_data.len() - 1
     }
+    pub(crate) fn push_composite_req_data(&mut self, data: CompositeAggReqData) -> usize {
+        self.per_request
+            .composite_req_data
+            .push(Some(Box::new(data)));
+        self.per_request.composite_req_data.len() - 1
+    }
 
     #[inline]
     pub(crate) fn get_term_req_data(&self, idx: usize) -> &TermsAggReqData {
@@ -108,20 +116,19 @@ impl AggregationsSegmentCtx {
             .expect("range_req_data slot is empty (taken)")
     }
     #[inline]
-    pub(crate) fn get_filter_req_data(&self, idx: usize) -> &FilterAggReqData {
-        self.per_request.filter_req_data[idx]
+    pub(crate) fn get_composite_req_data(&self, idx: usize) -> &CompositeAggReqData {
+        self.per_request.composite_req_data[idx]
             .as_deref()
-            .expect("filter_req_data slot is empty (taken)")
+            .expect("composite_req_data slot is empty (taken)")
     }
 
     // ---------- mutable getters ----------
 
     #[inline]
-    pub(crate) fn get_term_req_data_mut(&mut self, idx: usize) -> &mut TermsAggReqData {
-        self.per_request.term_req_data[idx]
-            .as_deref_mut()
-            .expect("term_req_data slot is empty (taken)")
+    pub(crate) fn get_metric_req_data_mut(&mut self, idx: usize) -> &mut MetricAggReqData {
+        &mut self.per_request.stats_metric_req_data[idx]
     }
+
     #[inline]
     pub(crate) fn get_cardinality_req_data_mut(
         &mut self,
@@ -129,10 +136,7 @@ impl AggregationsSegmentCtx {
     ) -> &mut CardinalityAggReqData {
         &mut self.per_request.cardinality_req_data[idx]
     }
-    #[inline]
-    pub(crate) fn get_metric_req_data_mut(&mut self, idx: usize) -> &mut MetricAggReqData {
-        &mut self.per_request.stats_metric_req_data[idx]
-    }
+
     #[inline]
     pub(crate) fn get_histogram_req_data_mut(&mut self, idx: usize) -> &mut HistogramAggReqData {
         self.per_request.histogram_req_data[idx]
@@ -142,21 +146,6 @@ impl AggregationsSegmentCtx {
 
     // ---------- take / put (terms, histogram, range) ----------
 
-    /// Move out the boxed Terms request at `idx`, leaving `None`.
-    #[inline]
-    pub(crate) fn take_term_req_data(&mut self, idx: usize) -> Box<TermsAggReqData> {
-        self.per_request.term_req_data[idx]
-            .take()
-            .expect("term_req_data slot is empty (taken)")
-    }
-
-    /// Put back a Terms request into an empty slot at `idx`.
-    #[inline]
-    pub(crate) fn put_back_term_req_data(&mut self, idx: usize, value: Box<TermsAggReqData>) {
-        debug_assert!(self.per_request.term_req_data[idx].is_none());
-        self.per_request.term_req_data[idx] = Some(value);
-    }
-
     /// Move out the boxed Histogram request at `idx`, leaving `None`.
     #[inline]
     pub(crate) fn take_histogram_req_data(&mut self, idx: usize) -> Box<HistogramAggReqData> {
@@ -205,6 +194,25 @@ impl AggregationsSegmentCtx {
         debug_assert!(self.per_request.filter_req_data[idx].is_none());
         self.per_request.filter_req_data[idx] = Some(value);
     }
+
+    /// Move out the Composite request at `idx`.
+    #[inline]
+    pub(crate) fn take_composite_req_data(&mut self, idx: usize) -> Box<CompositeAggReqData> {
+        self.per_request.composite_req_data[idx]
+            .take()
+            .expect("composite_req_data slot is empty (taken)")
+    }
+
+    /// Put back a Composite request into an empty slot at `idx`.
+    #[inline]
+    pub(crate) fn put_back_composite_req_data(
+        &mut self,
+        idx: usize,
+        value: Box<CompositeAggReqData>,
+    ) {
+        debug_assert!(self.per_request.composite_req_data[idx].is_none());
+        self.per_request.composite_req_data[idx] = Some(value);
+    }
 }
 
 /// Each type of aggregation has its own request data struct. This struct holds
@@ -232,6 +240,8 @@ pub struct PerRequestAggSegCtx {
     pub top_hits_req_data: Vec<TopHitsAggReqData>,
     /// MissingTermAggReqData contains the request data for a missing term aggregation.
     pub missing_term_req_data: Vec<MissingTermAggReqData>,
+    /// CompositeAggReqData contains the request data for a composite aggregation.
+    pub composite_req_data: Vec<Option<Box<CompositeAggReqData>>>,
 
     /// Request tree used to build collectors.
     pub agg_tree: Vec<AggRefNode>,
@@ -279,6 +289,11 @@ impl PerRequestAggSegCtx {
                 .iter()
                 .map(|t| t.get_memory_consumption())
                 .sum::<usize>()
+            + self
+                .composite_req_data
+                .iter()
+                .map(|b| b.as_ref().map(|d| d.get_memory_consumption()).unwrap_or(0))
+                .sum::<usize>()
             + self.agg_tree.len() * std::mem::size_of::<AggRefNode>()
     }
 
@@ -315,11 +330,17 @@ impl PerRequestAggSegCtx {
                 .expect("filter_req_data slot is empty (taken)")
                 .name
                 .as_str(),
+            AggKind::Composite => self.composite_req_data[idx]
+                .as_deref()
+                .expect("composite_req_data slot is empty (taken)")
+                .name
+                .as_str(),
         }
     }
 
     /// Convert the aggregation tree into a serializable struct representation.
     /// Each node contains: { name, kind, children }.
+    #[allow(dead_code)]
     pub fn get_view_tree(&self) -> Vec<AggTreeViewNode> {
         fn node_to_view(node: &AggRefNode, pr: &PerRequestAggSegCtx) -> AggTreeViewNode {
             let mut children: Vec<AggTreeViewNode> =
@@ -345,12 +366,19 @@ impl PerRequestAggSegCtx {
 pub(crate) fn build_segment_agg_collectors_root(
     req: &mut AggregationsSegmentCtx,
 ) -> crate::Result<Box<dyn SegmentAggregationCollector>> {
-    build_segment_agg_collectors(req, &req.per_request.agg_tree.clone())
+    build_segment_agg_collectors_generic(req, &req.per_request.agg_tree.clone())
 }
 
 pub(crate) fn build_segment_agg_collectors(
     req: &mut AggregationsSegmentCtx,
     nodes: &[AggRefNode],
+) -> crate::Result<Box<dyn SegmentAggregationCollector>> {
+    build_segment_agg_collectors_generic(req, nodes)
+}
+
+fn build_segment_agg_collectors_generic(
+    req: &mut AggregationsSegmentCtx,
+    nodes: &[AggRefNode],
 ) -> crate::Result<Box<dyn SegmentAggregationCollector>> {
     let mut collectors = Vec::new();
     for node in nodes.iter() {
@@ -373,7 +401,7 @@ pub(crate) fn build_segment_agg_collector(
     node: &AggRefNode,
 ) -> crate::Result<Box<dyn SegmentAggregationCollector>> {
     match node.kind {
-        AggKind::Terms => build_segment_aggregation_collector(req, node),
+        AggKind::Terms => crate::aggregation::bucket::build_segment_term_collector(req, node),
         AggKind::MissingTerm => {
             let req_data = &mut req.per_request.missing_term_req_data[node.idx_in_req_data];
             if req_data.accessors.is_empty() {
@@ -388,6 +416,8 @@ pub(crate) fn build_segment_agg_collector(
             Ok(Box::new(SegmentCardinalityCollector::from_req(
                 req_data.column_type,
                 node.idx_in_req_data,
+                req_data.accessor.clone(),
+                req_data.missing_value_for_accessor,
             )))
         }
         AggKind::StatsKind(stats_type) => {
@@ -398,20 +428,21 @@ pub(crate) fn build_segment_agg_collector(
                 | StatsType::Count
                 | StatsType::Max
                 | StatsType::Min
-                | StatsType::Stats => Ok(Box::new(SegmentStatsCollector::from_req(
-                    node.idx_in_req_data,
-                ))),
-                StatsType::ExtendedStats(sigma) => {
-                    Ok(Box::new(SegmentExtendedStatsCollector::from_req(
-                        req_data.field_type,
-                        sigma,
-                        node.idx_in_req_data,
-                        req_data.missing,
-                    )))
-                }
-                StatsType::Percentiles => Ok(Box::new(
-                    SegmentPercentilesCollector::from_req_and_validate(node.idx_in_req_data)?,
+                | StatsType::Stats => build_segment_stats_collector(req_data),
+                StatsType::ExtendedStats(sigma) => Ok(Box::new(
+                    SegmentExtendedStatsCollector::from_req(req_data, sigma),
                 )),
+                StatsType::Percentiles => {
+                    let req_data = req.get_metric_req_data_mut(node.idx_in_req_data);
+                    Ok(Box::new(
+                        SegmentPercentilesCollector::from_req_and_validate(
+                            req_data.field_type,
+                            req_data.missing_u64,
+                            req_data.accessor.clone(),
+                            node.idx_in_req_data,
+                        ),
+                    ))
+                }
             }
         }
         AggKind::TopHits => {
@@ -428,12 +459,13 @@ pub(crate) fn build_segment_agg_collector(
         AggKind::DateHistogram => Ok(Box::new(SegmentHistogramCollector::from_req_and_validate(
             req, node,
         )?)),
-        AggKind::Range => Ok(Box::new(SegmentRangeCollector::from_req_and_validate(
-            req, node,
-        )?)),
-        AggKind::Filter => Ok(Box::new(SegmentFilterCollector::from_req_and_validate(
-            req, node,
-        )?)),
+        AggKind::Range => Ok(build_segment_range_collector(req, node)?),
+        AggKind::Filter => build_segment_filter_collector(req, node),
+        AggKind::Composite => Ok(Box::new(
+            crate::aggregation::bucket::SegmentCompositeCollector::from_req_and_validate(
+                req, node,
+            )?,
+        )),
     }
 }
 
@@ -464,6 +496,7 @@ pub enum AggKind {
     DateHistogram,
     Range,
     Filter,
+    Composite,
 }
 
 impl AggKind {
@@ -479,6 +512,7 @@ impl AggKind {
             AggKind::DateHistogram => "DateHistogram",
             AggKind::Range => "Range",
             AggKind::Filter => "Filter",
+            AggKind::Composite => "Composite",
         }
     }
 }
@@ -493,10 +527,11 @@ pub(crate) fn build_aggregations_data_from_req(
     let mut data = AggregationsSegmentCtx {
         per_request: Default::default(),
         context,
+        column_block_accessor: ColumnBlockAccessor::default(),
     };
 
     for (name, agg) in aggs.iter() {
-        let nodes = build_nodes(name, agg, reader, segment_ordinal, &mut data)?;
+        let nodes = build_nodes(name, agg, reader, segment_ordinal, &mut data, true)?;
         data.per_request.agg_tree.extend(nodes);
     }
     Ok(data)
@@ -508,6 +543,7 @@ fn build_nodes(
     reader: &SegmentReader,
     segment_ordinal: SegmentOrdinal,
     data: &mut AggregationsSegmentCtx,
+    is_top_level: bool,
 ) -> crate::Result<Vec<AggRefNode>> {
     use AggregationVariants::*;
     match &req.agg {
@@ -520,9 +556,9 @@ fn build_nodes(
             let idx_in_req_data = data.push_range_req_data(RangeAggReqData {
                 accessor,
                 field_type,
-                column_block_accessor: Default::default(),
                 name: agg_name.to_string(),
                 req: range_req.clone(),
+                is_top_level,
             });
             let children = build_children(&req.sub_aggregation, reader, segment_ordinal, data)?;
             Ok(vec![AggRefNode {
@@ -540,9 +576,7 @@ fn build_nodes(
             let idx_in_req_data = data.push_histogram_req_data(HistogramAggReqData {
                 accessor,
                 field_type,
-                column_block_accessor: Default::default(),
                 name: agg_name.to_string(),
-                sub_aggregation_blueprint: None,
                 req: histo_req.clone(),
                 is_date_histogram: false,
                 bounds: HistogramBounds {
@@ -567,9 +601,7 @@ fn build_nodes(
             let idx_in_req_data = data.push_histogram_req_data(HistogramAggReqData {
                 accessor,
                 field_type,
-                column_block_accessor: Default::default(),
                 name: agg_name.to_string(),
-                sub_aggregation_blueprint: None,
                 req: histo_req,
                 is_date_histogram: true,
                 bounds: HistogramBounds {
@@ -594,6 +626,7 @@ fn build_nodes(
             data,
             &req.sub_aggregation,
             TermsOrCardinalityRequest::Terms(terms_req.clone()),
+            is_top_level,
         ),
         Cardinality(card_req) => build_terms_or_cardinality_nodes(
             agg_name,
@@ -604,6 +637,7 @@ fn build_nodes(
             data,
             &req.sub_aggregation,
             TermsOrCardinalityRequest::Cardinality(card_req.clone()),
+            is_top_level,
         ),
         Average(AverageAggregation { field, missing, .. })
         | Max(MaxAggregation { field, missing, .. })
@@ -647,7 +681,6 @@ fn build_nodes(
             let idx_in_req_data = data.push_metric_req_data(MetricAggReqData {
                 accessor,
                 field_type,
-                column_block_accessor: Default::default(),
                 name: agg_name.to_string(),
                 collecting_for,
                 missing: *missing,
@@ -675,7 +708,6 @@ fn build_nodes(
             let idx_in_req_data = data.push_metric_req_data(MetricAggReqData {
                 accessor,
                 field_type,
-                column_block_accessor: Default::default(),
                 name: agg_name.to_string(),
                 collecting_for: StatsType::Percentiles,
                 missing: percentiles_req.missing,
@@ -728,11 +760,19 @@ fn build_nodes(
                 children,
             }])
         }
+        AggregationVariants::Composite(composite_req) => Ok(vec![build_composite_node(
+            agg_name,
+            reader,
+            segment_ordinal,
+            data,
+            &req.sub_aggregation,
+            composite_req,
+        )?]),
         AggregationVariants::Filter(filter_req) => {
             // Build the query and evaluator upfront
             let schema = reader.schema();
             let tokenizers = &data.context.tokenizers;
-            let query = filter_req.parse_query(&schema, tokenizers)?;
+            let query = filter_req.parse_query(schema, tokenizers)?;
             let evaluator = crate::aggregation::bucket::DocumentQueryEvaluator::new(
                 query,
                 schema.clone(),
@@ -750,6 +790,7 @@ fn build_nodes(
                 segment_reader: reader.clone(),
                 evaluator,
                 matching_docs_buffer,
+                is_top_level,
             });
             let children = build_children(&req.sub_aggregation, reader, segment_ordinal, data)?;
             Ok(vec![AggRefNode {
@@ -761,6 +802,35 @@ fn build_nodes(
     }
 }
 
+fn build_composite_node(
+    agg_name: &str,
+    reader: &SegmentReader,
+    _segment_ordinal: SegmentOrdinal,
+    data: &mut AggregationsSegmentCtx,
+    sub_aggs: &Aggregations,
+    req: &CompositeAggregation,
+) -> crate::Result<AggRefNode> {
+    let mut composite_accessors = Vec::with_capacity(req.sources.len());
+    for source in &req.sources {
+        let source_after_key_opt = req.after.get(source.name()).map(|k| &k.0);
+        let source_accessor =
+            CompositeSourceAccessors::build_for_source(reader, source, source_after_key_opt)?;
+        composite_accessors.push(source_accessor);
+    }
+    let agg = CompositeAggReqData {
+        name: agg_name.to_string(),
+        req: req.clone(),
+        composite_accessors,
+    };
+    let idx = data.push_composite_req_data(agg);
+    let children = build_children(sub_aggs, reader, _segment_ordinal, data)?;
+    Ok(AggRefNode {
+        kind: AggKind::Composite,
+        idx_in_req_data: idx,
+        children,
+    })
+}
+
 fn build_children(
     aggs: &Aggregations,
     reader: &SegmentReader,
@@ -769,7 +839,14 @@ fn build_children(
 ) -> crate::Result<Vec<AggRefNode>> {
     let mut children = Vec::new();
     for (name, agg) in aggs.iter() {
-        children.extend(build_nodes(name, agg, reader, segment_ordinal, data)?);
+        children.extend(build_nodes(
+            name,
+            agg,
+            reader,
+            segment_ordinal,
+            data,
+            false,
+        )?);
     }
     Ok(children)
 }
@@ -833,6 +910,7 @@ fn build_terms_or_cardinality_nodes(
     data: &mut AggregationsSegmentCtx,
     sub_aggs: &Aggregations,
     req: TermsOrCardinalityRequest,
+    is_top_level: bool,
 ) -> crate::Result<Vec<AggRefNode>> {
     let mut nodes = Vec::new();
 
@@ -884,12 +962,12 @@ fn build_terms_or_cardinality_nodes(
         });
     }
 
-    // Add one node per accessor to mirror previous behavior and allow per-type missing handling.
+    // Add one node per accessor
     for (accessor, column_type) in column_and_types {
         let missing_value_for_accessor = if use_special_missing_agg {
             None
         } else if let Some(m) = missing.as_ref() {
-            get_missing_val_as_u64_lenient(column_type, m, field_name)?
+            get_missing_val_as_u64_lenient(column_type, accessor.max_value(), m, field_name)?
         } else {
             None
         };
@@ -915,13 +993,11 @@ fn build_terms_or_cardinality_nodes(
                     column_type,
                     str_dict_column: str_dict_column.clone(),
                     missing_value_for_accessor,
-                    column_block_accessor: Default::default(),
                     name: agg_name.to_string(),
                     req: TermsAggregationInternal::from_req(req),
-                    // Will be filled later when building collectors
-                    sub_aggregation_blueprint: None,
                     sug_aggregations: sub_aggs.clone(),
                     allowed_term_ids,
+                    is_top_level,
                 });
                 (idx_in_req_data, AggKind::Terms)
             }
@@ -931,7 +1007,6 @@ fn build_terms_or_cardinality_nodes(
                     column_type,
                     str_dict_column: str_dict_column.clone(),
                     missing_value_for_accessor,
-                    column_block_accessor: Default::default(),
                     name: agg_name.to_string(),
                     req: req.clone(),
                 });

src/aggregation/agg_limits.rs

@@ -35,6 +35,7 @@ pub struct AggregationLimitsGuard {
     /// Allocated memory with this guard.
     allocated_with_the_guard: u64,
 }
+
 impl Clone for AggregationLimitsGuard {
     fn clone(&self) -> Self {
         Self {

src/aggregation/agg_req.rs

@@ -32,8 +32,8 @@ use rustc_hash::FxHashMap;
 use serde::{Deserialize, Serialize};
 
 use super::bucket::{
-    DateHistogramAggregationReq, FilterAggregation, HistogramAggregation, RangeAggregation,
-    TermsAggregation,
+    CompositeAggregation, DateHistogramAggregationReq, FilterAggregation, HistogramAggregation,
+    RangeAggregation, TermsAggregation,
 };
 use super::metric::{
     AverageAggregation, CardinalityAggregationReq, CountAggregation, ExtendedStatsAggregation,
@@ -134,6 +134,9 @@ pub enum AggregationVariants {
     /// Filter documents into a single bucket.
     #[serde(rename = "filter")]
     Filter(FilterAggregation),
+    /// Multi-dimensional, paginable bucket aggregation.
+    #[serde(rename = "composite")]
+    Composite(CompositeAggregation),
 
     // Metric aggregation types
     /// Computes the average of the extracted values.
@@ -180,6 +183,11 @@ impl AggregationVariants {
             AggregationVariants::Histogram(histogram) => vec![histogram.field.as_str()],
             AggregationVariants::DateHistogram(histogram) => vec![histogram.field.as_str()],
             AggregationVariants::Filter(filter) => filter.get_fast_field_names(),
+            AggregationVariants::Composite(composite) => composite
+                .sources
+                .iter()
+                .map(|source| source.field())
+                .collect(),
             AggregationVariants::Average(avg) => vec![avg.field_name()],
             AggregationVariants::Count(count) => vec![count.field_name()],
             AggregationVariants::Max(max) => vec![max.field_name()],
@@ -214,6 +222,12 @@ impl AggregationVariants {
             _ => None,
         }
     }
+    pub(crate) fn as_composite(&self) -> Option<&CompositeAggregation> {
+        match &self {
+            AggregationVariants::Composite(composite) => Some(composite),
+            _ => None,
+        }
+    }
     pub(crate) fn as_percentile(&self) -> Option<&PercentilesAggregationReq> {
         match &self {
             AggregationVariants::Percentiles(percentile_req) => Some(percentile_req),

src/aggregation/agg_result.rs

@@ -9,14 +9,16 @@ use rustc_hash::FxHashMap;
 use serde::{Deserialize, Serialize};
 
 use super::bucket::GetDocCount;
+use super::intermediate_agg_result::CompositeIntermediateKey;
 use super::metric::{
     ExtendedStats, PercentilesMetricResult, SingleMetricResult, Stats, TopHitsMetricResult,
 };
 use super::{AggregationError, Key};
+use crate::aggregation::bucket::AfterKey;
 use crate::TantivyError;
 
 #[derive(Clone, Default, Debug, PartialEq, Serialize, Deserialize)]
-/// The final aggegation result.
+/// The final aggregation result.
 pub struct AggregationResults(pub FxHashMap<String, AggregationResult>);
 
 impl AggregationResults {
@@ -158,6 +160,14 @@ pub enum BucketResult {
     },
     /// This is the filter result - a single bucket with sub-aggregations
     Filter(FilterBucketResult),
+    /// This is the composite result
+    Composite {
+        /// The buckets
+        buckets: Vec<CompositeBucketEntry>,
+        /// The key to start after when paginating
+        #[serde(skip_serializing_if = "FxHashMap::is_empty")]
+        after_key: FxHashMap<String, AfterKey>,
+    },
 }
 
 impl BucketResult {
@@ -179,6 +189,9 @@ impl BucketResult {
                 // Only count sub-aggregation buckets
                 filter_result.sub_aggregations.get_bucket_count()
             }
+            BucketResult::Composite { buckets, .. } => {
+                buckets.iter().map(|bucket| bucket.get_bucket_count()).sum()
+            }
         }
     }
 }
@@ -337,3 +350,87 @@ pub struct FilterBucketResult {
     #[serde(flatten)]
     pub sub_aggregations: AggregationResults,
 }
+
+/// Note the type information loss compared to `CompositeIntermediateKey`.
+/// Pagination is performed using `AfterKey`, which encodes type information.
+#[derive(Clone, Debug, Serialize, Deserialize)]
+#[serde(untagged)]
+pub enum CompositeKey {
+    /// Boolean key
+    Bool(bool),
+    /// String key
+    Str(String),
+    /// `i64` key
+    I64(i64),
+    /// `u64` key
+    U64(u64),
+    /// `f64` key
+    F64(f64),
+    /// Null key
+    Null,
+}
+impl Eq for CompositeKey {}
+impl std::hash::Hash for CompositeKey {
+    fn hash<H: std::hash::Hasher>(&self, state: &mut H) {
+        core::mem::discriminant(self).hash(state);
+        match self {
+            Self::Bool(val) => val.hash(state),
+            Self::Str(text) => text.hash(state),
+            Self::F64(val) => val.to_bits().hash(state),
+            Self::U64(val) => val.hash(state),
+            Self::I64(val) => val.hash(state),
+            Self::Null => {}
+        }
+    }
+}
+impl PartialEq for CompositeKey {
+    fn eq(&self, other: &Self) -> bool {
+        match (self, other) {
+            (Self::Bool(l), Self::Bool(r)) => l == r,
+            (Self::Str(l), Self::Str(r)) => l == r,
+            (Self::F64(l), Self::F64(r)) => l.to_bits() == r.to_bits(),
+            (Self::I64(l), Self::I64(r)) => l == r,
+            (Self::U64(l), Self::U64(r)) => l == r,
+            (Self::Null, Self::Null) => true,
+            _ => false,
+        }
+    }
+}
+impl From<CompositeIntermediateKey> for CompositeKey {
+    fn from(value: CompositeIntermediateKey) -> Self {
+        match value {
+            CompositeIntermediateKey::Str(s) => Self::Str(s),
+            CompositeIntermediateKey::IpAddr(s) => {
+                if let Some(ip) = s.to_ipv4_mapped() {
+                    Self::Str(ip.to_string())
+                } else {
+                    Self::Str(s.to_string())
+                }
+            }
+            CompositeIntermediateKey::F64(f) => Self::F64(f),
+            CompositeIntermediateKey::Bool(f) => Self::Bool(f),
+            CompositeIntermediateKey::U64(f) => Self::U64(f),
+            CompositeIntermediateKey::I64(f) => Self::I64(f),
+            CompositeIntermediateKey::DateTime(f) => Self::I64(f / 1_000_000), // ns to ms
+            CompositeIntermediateKey::Null => Self::Null,
+        }
+    }
+}
+
+/// Composite bucket entry with a multi-dimensional key.
+#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
+pub struct CompositeBucketEntry {
+    /// The identifier of the bucket.
+    pub key: FxHashMap<String, CompositeKey>,
+    /// Number of documents in the bucket.
+    pub doc_count: u64,
+    #[serde(flatten)]
+    /// Sub-aggregations in this bucket.
+    pub sub_aggregation: AggregationResults,
+}
+
+impl CompositeBucketEntry {
+    pub(crate) fn get_bucket_count(&self) -> u64 {
+        1 + self.sub_aggregation.get_bucket_count()
+    }
+}

src/aggregation/agg_tests.rs

@@ -2,15 +2,441 @@ use serde_json::Value;
 
 use crate::aggregation::agg_req::{Aggregation, Aggregations};
 use crate::aggregation::agg_result::AggregationResults;
-use crate::aggregation::buf_collector::DOC_BLOCK_SIZE;
 use crate::aggregation::collector::AggregationCollector;
 use crate::aggregation::intermediate_agg_result::IntermediateAggregationResults;
 use crate::aggregation::tests::{get_test_index_2_segments, get_test_index_from_values_and_terms};
 use crate::aggregation::DistributedAggregationCollector;
+use crate::docset::COLLECT_BLOCK_BUFFER_LEN;
 use crate::query::{AllQuery, TermQuery};
 use crate::schema::{IndexRecordOption, Schema, FAST};
 use crate::{Index, IndexWriter, Term};
 
+// The following tests ensure that each bucket aggregation type correctly functions as a
+// sub-aggregation of another bucket aggregation in two scenarios:
+// 1) The parent has more buckets than the child sub-aggregation
+// 2) The child sub-aggregation has more buckets than the parent
+//
+// These scenarios exercise the bucket id mapping and sub-aggregation routing logic.
+
+#[test]
+fn test_terms_as_subagg_parent_more_vs_child_more() -> crate::Result<()> {
+    let index = get_test_index_2_segments(false)?;
+
+    // Case A: parent has more buckets than child
+    // Parent: range with 4 buckets
+    // Child: terms on text -> 2 buckets
+    let agg_parent_more: Aggregations = serde_json::from_value(json!({
+        "parent_range": {
+            "range": {
+                "field": "score",
+                "ranges": [
+                    {"to": 3.0},
+                    {"from": 3.0, "to": 7.0},
+                    {"from": 7.0, "to": 20.0},
+                    {"from": 20.0}
+                ]
+            },
+            "aggs": {
+                "child_terms": {"terms": {"field": "text", "order": {"_key": "asc"}}}
+            }
+        }
+    }))
+    .unwrap();
+
+    let res = crate::aggregation::tests::exec_request(agg_parent_more, &index)?;
+    // Exact expected structure and counts
+    assert_eq!(
+        res["parent_range"]["buckets"],
+        json!([
+            {
+                "key": "*-3",
+                "doc_count": 1,
+                "to": 3.0,
+                "child_terms": {
+                    "buckets": [
+                        {"doc_count": 1, "key": "cool"}
+                    ],
+                    "sum_other_doc_count": 0
+                }
+            },
+            {
+                "key": "3-7",
+                "doc_count": 3,
+                "from": 3.0,
+                "to": 7.0,
+                "child_terms": {
+                    "buckets": [
+                        {"doc_count": 2, "key": "cool"},
+                        {"doc_count": 1, "key": "nohit"}
+                    ],
+                    "sum_other_doc_count": 0
+                }
+            },
+            {
+                "key": "7-20",
+                "doc_count": 3,
+                "from": 7.0,
+                "to": 20.0,
+                "child_terms": {
+                    "buckets": [
+                        {"doc_count": 3, "key": "cool"}
+                    ],
+                    "sum_other_doc_count": 0
+                }
+            },
+            {
+                "key": "20-*",
+                "doc_count": 2,
+                "from": 20.0,
+                "child_terms": {
+                    "buckets": [
+                        {"doc_count": 1, "key": "cool"},
+                        {"doc_count": 1, "key": "nohit"}
+                    ],
+                    "sum_other_doc_count": 0
+                }
+            }
+        ])
+    );
+
+    // Case B: child has more buckets than parent
+    // Parent: histogram on score with large interval -> 1 bucket
+    // Child: terms on text -> 2 buckets (cool/nohit)
+    let agg_child_more: Aggregations = serde_json::from_value(json!({
+        "parent_hist": {
+            "histogram": {"field": "score", "interval": 100.0},
+            "aggs": {
+                "child_terms": {"terms": {"field": "text", "order": {"_key": "asc"}}}
+            }
+        }
+    }))
+    .unwrap();
+
+    let res = crate::aggregation::tests::exec_request(agg_child_more, &index)?;
+    assert_eq!(
+        res["parent_hist"],
+        json!({
+            "buckets": [
+                {
+                    "key": 0.0,
+                    "doc_count": 9,
+                    "child_terms": {
+                        "buckets": [
+                            {"doc_count": 7, "key": "cool"},
+                            {"doc_count": 2, "key": "nohit"}
+                        ],
+                        "sum_other_doc_count": 0
+                    }
+                }
+            ]
+        })
+    );
+
+    Ok(())
+}
+
+#[test]
+fn test_range_as_subagg_parent_more_vs_child_more() -> crate::Result<()> {
+    let index = get_test_index_2_segments(false)?;
+
+    // Case A: parent has more buckets than child
+    // Parent: range with 5 buckets
+    // Child: coarse range with 3 buckets
+    let agg_parent_more: Aggregations = serde_json::from_value(json!({
+        "parent_range": {
+            "range": {
+                "field": "score",
+                "ranges": [
+                    {"to": 3.0},
+                    {"from": 3.0, "to": 7.0},
+                    {"from": 7.0, "to": 11.0},
+                    {"from": 11.0, "to": 20.0},
+                    {"from": 20.0}
+                ]
+            },
+            "aggs": {
+                "child_range": {
+                    "range": {
+                        "field": "score",
+                        "ranges": [
+                            {"to": 3.0},
+                            {"from": 3.0, "to": 20.0}
+                        ]
+                    }
+                }
+            }
+        }
+    }))
+    .unwrap();
+    let res = crate::aggregation::tests::exec_request(agg_parent_more, &index)?;
+    assert_eq!(
+        res["parent_range"]["buckets"],
+        json!([
+            {"key": "*-3", "doc_count": 1, "to": 3.0,
+                "child_range": {"buckets": [
+                    {"key": "*-3", "doc_count": 1, "to": 3.0},
+                    {"key": "3-20", "doc_count": 0, "from": 3.0, "to": 20.0},
+                    {"key": "20-*", "doc_count": 0, "from": 20.0}
+                ]}
+            },
+            {"key": "3-7", "doc_count": 3, "from": 3.0, "to": 7.0,
+                "child_range": {"buckets": [
+                    {"key": "*-3", "doc_count": 0, "to": 3.0},
+                    {"key": "3-20", "doc_count": 3, "from": 3.0, "to": 20.0},
+                    {"key": "20-*", "doc_count": 0, "from": 20.0}
+                ]}
+            },
+            {"key": "7-11", "doc_count": 1, "from": 7.0, "to": 11.0,
+                "child_range": {"buckets": [
+                    {"key": "*-3", "doc_count": 0, "to": 3.0},
+                    {"key": "3-20", "doc_count": 1, "from": 3.0, "to": 20.0},
+                    {"key": "20-*", "doc_count": 0, "from": 20.0}
+                ]}
+            },
+            {"key": "11-20", "doc_count": 2, "from": 11.0, "to": 20.0,
+                "child_range": {"buckets": [
+                    {"key": "*-3", "doc_count": 0, "to": 3.0},
+                    {"key": "3-20", "doc_count": 2, "from": 3.0, "to": 20.0},
+                    {"key": "20-*", "doc_count": 0, "from": 20.0}
+                ]}
+            },
+            {"key": "20-*", "doc_count": 2, "from": 20.0,
+                "child_range": {"buckets": [
+                    {"key": "*-3", "doc_count": 0, "to": 3.0},
+                    {"key": "3-20", "doc_count": 0, "from": 3.0, "to": 20.0},
+                    {"key": "20-*", "doc_count": 2, "from": 20.0}
+                ]}
+            }
+        ])
+    );
+
+    // Case B: child has more buckets than parent
+    // Parent: terms on text (2 buckets)
+    // Child: range with 4 buckets
+    let agg_child_more: Aggregations = serde_json::from_value(json!({
+        "parent_terms": {
+            "terms": {"field": "text"},
+            "aggs": {
+                "child_range": {
+                    "range": {
+                        "field": "score",
+                        "ranges": [
+                            {"to": 3.0},
+                            {"from": 3.0, "to": 7.0},
+                            {"from": 7.0, "to": 20.0}
+                        ]
+                    }
+                }
+            }
+        }
+    }))
+    .unwrap();
+    let res = crate::aggregation::tests::exec_request(agg_child_more, &index)?;
+
+    assert_eq!(
+        res["parent_terms"],
+        json!({
+            "buckets": [
+                {
+                    "key": "cool",
+                    "doc_count": 7,
+                    "child_range": {
+                        "buckets": [
+                            {"key": "*-3", "doc_count": 1, "to": 3.0},
+                            {"key": "3-7", "doc_count": 2, "from": 3.0, "to": 7.0},
+                            {"key": "7-20", "doc_count": 3, "from": 7.0, "to": 20.0},
+                            {"key": "20-*", "doc_count": 1, "from": 20.0}
+                        ]
+                    }
+                },
+                {
+                    "key": "nohit",
+                    "doc_count": 2,
+                    "child_range": {
+                        "buckets": [
+                            {"key": "*-3", "doc_count": 0, "to": 3.0},
+                            {"key": "3-7", "doc_count": 1, "from": 3.0, "to": 7.0},
+                            {"key": "7-20", "doc_count": 0, "from": 7.0, "to": 20.0},
+                            {"key": "20-*", "doc_count": 1, "from": 20.0}
+                        ]
+                    }
+                }
+            ],
+            "doc_count_error_upper_bound": 0,
+            "sum_other_doc_count": 0
+        })
+    );
+
+    Ok(())
+}
+
+#[test]
+fn test_histogram_as_subagg_parent_more_vs_child_more() -> crate::Result<()> {
+    let index = get_test_index_2_segments(false)?;
+
+    // Case A: parent has more buckets than child
+    // Parent: range with several ranges
+    // Child: histogram with large interval (single bucket per parent)
+    let agg_parent_more: Aggregations = serde_json::from_value(json!({
+        "parent_range": {
+            "range": {
+                "field": "score",
+                "ranges": [
+                    {"to": 3.0},
+                    {"from": 3.0, "to": 7.0},
+                    {"from": 7.0, "to": 11.0},
+                    {"from": 11.0, "to": 20.0},
+                    {"from": 20.0}
+                ]
+            },
+            "aggs": {
+                "child_hist": {"histogram": {"field": "score", "interval": 100.0}}
+            }
+        }
+    }))
+    .unwrap();
+    let res = crate::aggregation::tests::exec_request(agg_parent_more, &index)?;
+    assert_eq!(
+        res["parent_range"]["buckets"],
+        json!([
+            {"key": "*-3", "doc_count": 1, "to": 3.0,
+                "child_hist": {"buckets": [ {"key": 0.0, "doc_count": 1} ]}
+            },
+            {"key": "3-7", "doc_count": 3, "from": 3.0, "to": 7.0,
+                "child_hist": {"buckets": [ {"key": 0.0, "doc_count": 3} ]}
+            },
+            {"key": "7-11", "doc_count": 1, "from": 7.0, "to": 11.0,
+                "child_hist": {"buckets": [ {"key": 0.0, "doc_count": 1} ]}
+            },
+            {"key": "11-20", "doc_count": 2, "from": 11.0, "to": 20.0,
+                "child_hist": {"buckets": [ {"key": 0.0, "doc_count": 2} ]}
+            },
+            {"key": "20-*", "doc_count": 2, "from": 20.0,
+                "child_hist": {"buckets": [ {"key": 0.0, "doc_count": 2} ]}
+            }
+        ])
+    );
+
+    // Case B: child has more buckets than parent
+    // Parent: terms on text -> 2 buckets
+    // Child: histogram with small interval -> multiple buckets including empties
+    let agg_child_more: Aggregations = serde_json::from_value(json!({
+        "parent_terms": {
+            "terms": {"field": "text"},
+            "aggs": {
+                "child_hist": {"histogram": {"field": "score", "interval": 10.0}}
+            }
+        }
+    }))
+    .unwrap();
+    let res = crate::aggregation::tests::exec_request(agg_child_more, &index)?;
+    assert_eq!(
+        res["parent_terms"],
+        json!({
+            "buckets": [
+                {
+                    "key": "cool",
+                    "doc_count": 7,
+                    "child_hist": {
+                        "buckets": [
+                            {"key": 0.0, "doc_count": 4},
+                            {"key": 10.0, "doc_count": 2},
+                            {"key": 20.0, "doc_count": 0},
+                            {"key": 30.0, "doc_count": 0},
+                            {"key": 40.0, "doc_count": 1}
+                        ]
+                    }
+                },
+                {
+                    "key": "nohit",
+                    "doc_count": 2,
+                    "child_hist": {
+                        "buckets": [
+                            {"key": 0.0, "doc_count": 1},
+                            {"key": 10.0, "doc_count": 0},
+                            {"key": 20.0, "doc_count": 0},
+                            {"key": 30.0, "doc_count": 0},
+                            {"key": 40.0, "doc_count": 1}
+                        ]
+                    }
+                }
+            ],
+            "doc_count_error_upper_bound": 0,
+            "sum_other_doc_count": 0
+        })
+    );
+
+    Ok(())
+}
+
+#[test]
+fn test_date_histogram_as_subagg_parent_more_vs_child_more() -> crate::Result<()> {
+    let index = get_test_index_2_segments(false)?;
+
+    // Case A: parent has more buckets than child
+    // Parent: range with several buckets
+    // Child: date_histogram with 30d -> single bucket per parent
+    let agg_parent_more: Aggregations = serde_json::from_value(json!({
+        "parent_range": {
+            "range": {
+                "field": "score",
+                "ranges": [
+                    {"to": 3.0},
+                    {"from": 3.0, "to": 7.0},
+                    {"from": 7.0, "to": 11.0},
+                    {"from": 11.0, "to": 20.0},
+                    {"from": 20.0}
+                ]
+            },
+            "aggs": {
+                "child_date_hist": {"date_histogram": {"field": "date", "fixed_interval": "30d"}}
+            }
+        }
+    }))
+    .unwrap();
+    let res = crate::aggregation::tests::exec_request(agg_parent_more, &index)?;
+    let buckets = res["parent_range"]["buckets"].as_array().unwrap();
+    // Verify each parent bucket has exactly one child date bucket with matching doc_count
+    for bucket in buckets {
+        let parent_count = bucket["doc_count"].as_u64().unwrap();
+        let child_buckets = bucket["child_date_hist"]["buckets"].as_array().unwrap();
+        assert_eq!(child_buckets.len(), 1);
+        assert_eq!(child_buckets[0]["doc_count"], parent_count);
+    }
+
+    // Case B: child has more buckets than parent
+    // Parent: terms on text (2 buckets)
+    // Child: date_histogram with 1d -> multiple buckets
+    let agg_child_more: Aggregations = serde_json::from_value(json!({
+        "parent_terms": {
+            "terms": {"field": "text"},
+            "aggs": {
+                "child_date_hist": {"date_histogram": {"field": "date", "fixed_interval": "1d"}}
+            }
+        }
+    }))
+    .unwrap();
+    let res = crate::aggregation::tests::exec_request(agg_child_more, &index)?;
+    let buckets = res["parent_terms"]["buckets"].as_array().unwrap();
+
+    // cool bucket
+    assert_eq!(buckets[0]["key"], "cool");
+    let cool_buckets = buckets[0]["child_date_hist"]["buckets"].as_array().unwrap();
+    assert_eq!(cool_buckets.len(), 3);
+    assert_eq!(cool_buckets[0]["doc_count"], 1); // day 0
+    assert_eq!(cool_buckets[1]["doc_count"], 4); // day 1
+    assert_eq!(cool_buckets[2]["doc_count"], 2); // day 2
+
+    // nohit bucket
+    assert_eq!(buckets[1]["key"], "nohit");
+    let nohit_buckets = buckets[1]["child_date_hist"]["buckets"].as_array().unwrap();
+    assert_eq!(nohit_buckets.len(), 2);
+    assert_eq!(nohit_buckets[0]["doc_count"], 1); // day 1
+    assert_eq!(nohit_buckets[1]["doc_count"], 1); // day 2
+
+    Ok(())
+}
+
 fn get_avg_req(field_name: &str) -> Aggregation {
     serde_json::from_value(json!({
         "avg": {
@@ -25,6 +451,10 @@ fn get_collector(agg_req: Aggregations) -> AggregationCollector {
 }
 
 // *** EVERY BUCKET-TYPE SHOULD BE TESTED HERE ***
+// Note: The flushng part of these  tests are outdated, since the buffering change after converting
+// the collection into one collector per request instead of per bucket.
+//
+// However they are useful as they test a complex aggregation requests.
 fn test_aggregation_flushing(
     merge_segments: bool,
     use_distributed_collector: bool,
@@ -37,8 +467,9 @@ fn test_aggregation_flushing(
 
     let reader = index.reader()?;
 
-    assert_eq!(DOC_BLOCK_SIZE, 64);
-    // In the tree we cache Documents of DOC_BLOCK_SIZE, before passing them down as one block.
+    assert_eq!(COLLECT_BLOCK_BUFFER_LEN, 64);
+    // In the tree we cache documents of COLLECT_BLOCK_BUFFER_LEN before passing them down as one
+    // block.
     //
     // Build a request so that on the first level we have one full cache, which is then flushed.
     // The same cache should have some residue docs at the end, which are flushed (Range 0-70)

src/aggregation/bucket/composite/accessors.rs

@@ -0,0 +1,548 @@
+use std::net::Ipv6Addr;
+
+use columnar::column_values::CompactSpaceU64Accessor;
+use columnar::{Column, ColumnType, MonotonicallyMappableToU64, StrColumn, TermOrdHit};
+
+use crate::aggregation::accessor_helpers::get_numeric_or_date_column_types;
+use crate::aggregation::bucket::composite::numeric_types::num_proj;
+use crate::aggregation::bucket::composite::numeric_types::num_proj::ProjectedNumber;
+use crate::aggregation::bucket::composite::ToTypePaginationOrder;
+use crate::aggregation::bucket::{
+    parse_into_milliseconds, CalendarInterval, CompositeAggregation, CompositeAggregationSource,
+    MissingOrder, Order,
+};
+use crate::aggregation::intermediate_agg_result::CompositeIntermediateKey;
+use crate::{SegmentReader, TantivyError};
+
+/// Contains all information required by the SegmentCompositeCollector to perform the
+/// composite aggregation on a segment.
+pub struct CompositeAggReqData {
+    /// The name of the aggregation.
+    pub name: String,
+    /// The normalized term aggregation request.
+    pub req: CompositeAggregation,
+    /// Accessors for each source, each source can have multiple accessors (columns).
+    pub composite_accessors: Vec<CompositeSourceAccessors>,
+}
+
+impl CompositeAggReqData {
+    /// Estimate the memory consumption of this struct in bytes.
+    pub fn get_memory_consumption(&self) -> usize {
+        std::mem::size_of::<Self>()
+            + self.composite_accessors.len() * std::mem::size_of::<CompositeSourceAccessors>()
+    }
+}
+
+/// Accessors for a single column in a composite source.
+pub struct CompositeAccessor {
+    /// The fast field column
+    pub column: Column<u64>,
+    /// The column type
+    pub column_type: ColumnType,
+    /// Term dictionary if the column type is Str
+    ///
+    /// Only used by term sources
+    pub str_dict_column: Option<StrColumn>,
+    /// Parsed date interval for date histogram sources
+    pub date_histogram_interval: PrecomputedDateInterval,
+}
+
+/// Accessors to all the columns that belong to the field of a composite source.
+pub struct CompositeSourceAccessors {
+    /// The accessors for this source
+    pub accessors: Vec<CompositeAccessor>,
+    /// The key after which to start collecting results. Applies to the first
+    /// column of the source.
+    pub after_key: PrecomputedAfterKey,
+
+    /// The column index the after_key applies to. The after_key only applies to
+    /// one column. Columns before should be skipped. Columns after should be
+    /// kept without comparison to the after_key.
+    pub after_key_accessor_idx: usize,
+
+    /// Whether to skip missing values because of the after_key. Skipping only
+    /// applies if the value for previous columns were exactly equal to the
+    /// corresponding after keys (is_on_after_key).
+    pub skip_missing: bool,
+
+    /// The after key was set to null to indicate that the last collected key
+    /// was a missing value.
+    pub is_after_key_explicit_missing: bool,
+}
+
+impl CompositeSourceAccessors {
+    /// Creates a new set of accessors for the composite source.
+    ///
+    /// Precomputes some values to make collection faster.
+    pub fn build_for_source(
+        reader: &SegmentReader,
+        source: &CompositeAggregationSource,
+        // First option is None when no after key was set in the query, the
+        // second option is None when the after key was set but its value for
+        // this source was set to `null`
+        source_after_key_opt: Option<&CompositeIntermediateKey>,
+    ) -> crate::Result<Self> {
+        let is_after_key_explicit_missing = source_after_key_opt
+            .map(|after_key| matches!(after_key, CompositeIntermediateKey::Null))
+            .unwrap_or(false);
+        let mut skip_missing = false;
+        if let Some(CompositeIntermediateKey::Null) = source_after_key_opt {
+            if !source.missing_bucket() {
+                return Err(TantivyError::InvalidArgument(
+                    "the 'after' key for a source cannot be null when 'missing_bucket' is false"
+                        .to_string(),
+                ));
+            }
+        } else if source_after_key_opt.is_some() {
+            // if missing buckets come first and we have a non null after key, we skip missing
+            if MissingOrder::First == source.missing_order() {
+                skip_missing = true;
+            }
+            if MissingOrder::Default == source.missing_order() && Order::Asc == source.order() {
+                skip_missing = true;
+            }
+        };
+
+        match source {
+            CompositeAggregationSource::Terms(source) => {
+                let allowed_column_types = [
+                    ColumnType::I64,
+                    ColumnType::U64,
+                    ColumnType::F64,
+                    ColumnType::Str,
+                    ColumnType::DateTime,
+                    ColumnType::Bool,
+                    ColumnType::IpAddr,
+                    // ColumnType::Bytes Unsupported
+                ];
+                let mut columns_and_types = reader
+                    .fast_fields()
+                    .u64_lenient_for_type_all(Some(&allowed_column_types), &source.field)?;
+
+                // Sort columns by their pagination order and determine which to skip
+                columns_and_types.sort_by_key(|(_, col_type): &(Column, ColumnType)| {
+                    col_type.column_pagination_order()
+                });
+                if source.order == Order::Desc {
+                    columns_and_types.reverse();
+                }
+                let after_key_accessor_idx = find_first_column_to_collect(
+                    &columns_and_types,
+                    source_after_key_opt,
+                    source.missing_order,
+                    source.order,
+                )?;
+
+                let source_collectors: Vec<CompositeAccessor> = columns_and_types
+                    .into_iter()
+                    .map(|(column, column_type)| {
+                        Ok(CompositeAccessor {
+                            column,
+                            column_type,
+                            str_dict_column: reader.fast_fields().str(&source.field)?,
+                            date_histogram_interval: PrecomputedDateInterval::NotApplicable,
+                        })
+                    })
+                    .collect::<crate::Result<_>>()?;
+
+                let after_key = if let Some(first_col) =
+                    source_collectors.get(after_key_accessor_idx)
+                {
+                    match source_after_key_opt {
+                        Some(after_key) => PrecomputedAfterKey::precompute(
+                            &first_col,
+                            after_key,
+                            &source.field,
+                            source.missing_order,
+                            source.order,
+                        )?,
+                        None => {
+                            precompute_missing_after_key(false, source.missing_order, source.order)
+                        }
+                    }
+                } else {
+                    // if no columns, we don't care about the after_key
+                    PrecomputedAfterKey::Next(0)
+                };
+
+                Ok(CompositeSourceAccessors {
+                    accessors: source_collectors,
+                    is_after_key_explicit_missing,
+                    skip_missing,
+                    after_key,
+                    after_key_accessor_idx,
+                })
+            }
+            CompositeAggregationSource::Histogram(source) => {
+                let column_and_types: Vec<(Column, ColumnType)> =
+                    reader.fast_fields().u64_lenient_for_type_all(
+                        Some(get_numeric_or_date_column_types()),
+                        &source.field,
+                    )?;
+                let source_collectors: Vec<CompositeAccessor> = column_and_types
+                    .into_iter()
+                    .map(|(column, column_type)| {
+                        Ok(CompositeAccessor {
+                            column,
+                            column_type,
+                            str_dict_column: None,
+                            date_histogram_interval: PrecomputedDateInterval::NotApplicable,
+                        })
+                    })
+                    .collect::<crate::Result<_>>()?;
+                let after_key = match source_after_key_opt {
+                    Some(CompositeIntermediateKey::F64(key)) => {
+                        let normalized_key = *key / source.interval;
+                        num_proj::f64_to_i64(normalized_key).into()
+                    }
+                    Some(CompositeIntermediateKey::Null) => {
+                        precompute_missing_after_key(true, source.missing_order, source.order)
+                    }
+                    None => precompute_missing_after_key(true, source.missing_order, source.order),
+                    _ => {
+                        return Err(crate::TantivyError::InvalidArgument(
+                            "After key type invalid for interval composite source".to_string(),
+                        ));
+                    }
+                };
+                Ok(CompositeSourceAccessors {
+                    accessors: source_collectors,
+                    is_after_key_explicit_missing,
+                    skip_missing,
+                    after_key,
+                    after_key_accessor_idx: 0,
+                })
+            }
+            CompositeAggregationSource::DateHistogram(source) => {
+                let column_and_types = reader
+                    .fast_fields()
+                    .u64_lenient_for_type_all(Some(&[ColumnType::DateTime]), &source.field)?;
+                let date_histogram_interval =
+                    PrecomputedDateInterval::from_date_histogram_source_intervals(
+                        &source.fixed_interval,
+                        source.calendar_interval,
+                    )?;
+                let source_collectors: Vec<CompositeAccessor> = column_and_types
+                    .into_iter()
+                    .map(|(column, column_type)| {
+                        Ok(CompositeAccessor {
+                            column,
+                            column_type,
+                            str_dict_column: None,
+                            date_histogram_interval,
+                        })
+                    })
+                    .collect::<crate::Result<_>>()?;
+                let after_key = match source_after_key_opt {
+                    Some(CompositeIntermediateKey::DateTime(key)) => {
+                        PrecomputedAfterKey::Exact(key.to_u64())
+                    }
+                    Some(CompositeIntermediateKey::Null) => {
+                        precompute_missing_after_key(true, source.missing_order, source.order)
+                    }
+                    None => precompute_missing_after_key(true, source.missing_order, source.order),
+                    _ => {
+                        return Err(crate::TantivyError::InvalidArgument(
+                            "After key type invalid for interval composite source".to_string(),
+                        ));
+                    }
+                };
+                Ok(CompositeSourceAccessors {
+                    accessors: source_collectors,
+                    is_after_key_explicit_missing,
+                    skip_missing,
+                    after_key,
+                    after_key_accessor_idx: 0,
+                })
+            }
+        }
+    }
+}
+
+/// Finds the index of the first column we should start collecting from to
+/// resume the pagination from the after_key.
+fn find_first_column_to_collect<T>(
+    sorted_columns: &[(T, ColumnType)],
+    after_key_opt: Option<&CompositeIntermediateKey>,
+    missing_order: MissingOrder,
+    order: Order,
+) -> crate::Result<usize> {
+    let after_key = match after_key_opt {
+        None => return Ok(0), // No pagination, start from beginning
+        Some(key) => key,
+    };
+    // Handle null after_key (we were on a missing value last time)
+    if matches!(after_key, CompositeIntermediateKey::Null) {
+        return match (missing_order, order) {
+            // Missing values come first, so all columns remain
+            (MissingOrder::First, _) | (MissingOrder::Default, Order::Asc) => Ok(0),
+            // Missing values come last, so all columns are done
+            (MissingOrder::Last, _) | (MissingOrder::Default, Order::Desc) => {
+                Ok(sorted_columns.len())
+            }
+        };
+    }
+    // Find the first column whose type order matches or follows the after_key's
+    // type in the pagination sequence
+    let after_key_column_order = after_key.column_pagination_order();
+    for (idx, (_, col_type)) in sorted_columns.iter().enumerate() {
+        let col_order = col_type.column_pagination_order();
+        let is_first_to_collect = match order {
+            Order::Asc => col_order >= after_key_column_order,
+            Order::Desc => col_order <= after_key_column_order,
+        };
+        if is_first_to_collect {
+            return Ok(idx);
+        }
+    }
+    // All columns are before the after_key, nothing left to collect
+    Ok(sorted_columns.len())
+}
+
+fn precompute_missing_after_key(
+    is_after_key_explicit_missing: bool,
+    missing_order: MissingOrder,
+    order: Order,
+) -> PrecomputedAfterKey {
+    let after_last = PrecomputedAfterKey::AfterLast;
+    let before_first = PrecomputedAfterKey::Next(0);
+    match (is_after_key_explicit_missing, missing_order, order) {
+        (true, MissingOrder::First, Order::Asc) => before_first,
+        (true, MissingOrder::First, Order::Desc) => after_last,
+        (true, MissingOrder::Last, Order::Asc) => after_last,
+        (true, MissingOrder::Last, Order::Desc) => before_first,
+        (true, MissingOrder::Default, Order::Asc) => before_first,
+        (true, MissingOrder::Default, Order::Desc) => after_last,
+        (false, _, Order::Asc) => before_first,
+        (false, _, Order::Desc) => after_last,
+    }
+}
+
+/// A parsed representation of the date interval for date histogram sources
+#[derive(Clone, Copy, Debug)]
+pub enum PrecomputedDateInterval {
+    /// This is not a date histogram source
+    NotApplicable,
+    /// Source was configured with a fixed interval
+    FixedNanoseconds(i64),
+    /// Source was configured with a calendar interval
+    Calendar(CalendarInterval),
+}
+
+impl PrecomputedDateInterval {
+    /// Validates the date histogram source interval fields and parses a date interval from them.
+    pub fn from_date_histogram_source_intervals(
+        fixed_interval: &Option<String>,
+        calendar_interval: Option<CalendarInterval>,
+    ) -> crate::Result<Self> {
+        match (fixed_interval, calendar_interval) {
+            (Some(_), Some(_)) | (None, None) => Err(TantivyError::InvalidArgument(
+                "date histogram source must one and only one of fixed_interval or \
+                 calendar_interval set"
+                    .to_string(),
+            )),
+            (Some(fixed_interval), None) => {
+                let fixed_interval_ms = parse_into_milliseconds(&fixed_interval)?;
+                Ok(PrecomputedDateInterval::FixedNanoseconds(
+                    fixed_interval_ms * 1_000_000,
+                ))
+            }
+            (None, Some(calendar_interval)) => {
+                Ok(PrecomputedDateInterval::Calendar(calendar_interval))
+            }
+        }
+    }
+}
+
+/// The after key projected to the u64 column space
+///
+/// Some column types (term, IP) might not have an exact representation of the
+/// specified after key
+#[derive(Debug)]
+pub enum PrecomputedAfterKey {
+    /// The after key could be exactly represented in the column space.
+    Exact(u64),
+    /// The after key could not be exactly represented exactly represented, so
+    /// this is the next closest one.
+    Next(u64),
+    /// The after key could not be represented in the column space, it is
+    /// greater than all value
+    AfterLast,
+}
+
+impl From<TermOrdHit> for PrecomputedAfterKey {
+    fn from(hit: TermOrdHit) -> Self {
+        match hit {
+            TermOrdHit::Exact(ord) => PrecomputedAfterKey::Exact(ord),
+            // TermOrdHit represents AfterLast as Next(u64::MAX), we keep it as is
+            TermOrdHit::Next(ord) => PrecomputedAfterKey::Next(ord),
+        }
+    }
+}
+
+impl<T: MonotonicallyMappableToU64> From<ProjectedNumber<T>> for PrecomputedAfterKey {
+    fn from(num: ProjectedNumber<T>) -> Self {
+        match num {
+            ProjectedNumber::Exact(number) => PrecomputedAfterKey::Exact(number.to_u64()),
+            ProjectedNumber::Next(number) => PrecomputedAfterKey::Next(number.to_u64()),
+            ProjectedNumber::AfterLast => PrecomputedAfterKey::AfterLast,
+        }
+    }
+}
+
+// /!\ These operators only makes sense if both values are in the same column space
+impl PrecomputedAfterKey {
+    pub fn equals(&self, column_value: u64) -> bool {
+        match self {
+            PrecomputedAfterKey::Exact(v) => *v == column_value,
+            PrecomputedAfterKey::Next(_) => false,
+            PrecomputedAfterKey::AfterLast => false,
+        }
+    }
+
+    pub fn gt(&self, column_value: u64) -> bool {
+        match self {
+            PrecomputedAfterKey::Exact(v) => *v > column_value,
+            PrecomputedAfterKey::Next(v) => *v > column_value,
+            PrecomputedAfterKey::AfterLast => true,
+        }
+    }
+
+    pub fn lt(&self, column_value: u64) -> bool {
+        match self {
+            PrecomputedAfterKey::Exact(v) => *v < column_value,
+            // a value equal to the next is greater than the after key
+            PrecomputedAfterKey::Next(v) => *v <= column_value,
+            PrecomputedAfterKey::AfterLast => false,
+        }
+    }
+
+    fn precompute_ip_addr(column: &Column<u64>, key: &Ipv6Addr) -> crate::Result<Self> {
+        // For IP addresses we need to find the compact space value.
+        // We try to convert via the column's min/max range scan.
+        // Since CompactSpaceU64Accessor::u128_to_compact is not public,
+        // we search linearly for the exact u64 value by scanning column values.
+        let ip_u128 = key.to_bits();
+
+        // Scan for matching value - IP columns are typically small
+        let num_vals = column.values.num_vals();
+        let mut found_exact = false;
+        let mut exact_compact = 0u64;
+        let mut best_next: Option<u64> = None;
+
+        for doc_id in 0..num_vals {
+            let val = column.values.get_val(doc_id);
+            // We need the CompactSpaceU64Accessor to convert compact to u128
+            let compact_accessor = column
+                .values
+                .clone()
+                .downcast_arc::<CompactSpaceU64Accessor>()
+                .map_err(|_| {
+                    TantivyError::AggregationError(
+                        crate::aggregation::AggregationError::InternalError(
+                            "type mismatch: could not downcast to CompactSpaceU64Accessor"
+                                .to_string(),
+                        ),
+                    )
+                })?;
+            let val_u128 = compact_accessor.compact_to_u128(val as u32);
+            if val_u128 == ip_u128 {
+                found_exact = true;
+                exact_compact = val;
+                break;
+            } else if val_u128 > ip_u128 {
+                match best_next {
+                    None => best_next = Some(val),
+                    Some(current_best) => {
+                        let current_u128 = compact_accessor.compact_to_u128(current_best as u32);
+                        if val_u128 < current_u128 {
+                            best_next = Some(val);
+                        }
+                    }
+                }
+            }
+        }
+
+        if found_exact {
+            Ok(PrecomputedAfterKey::Exact(exact_compact))
+        } else if let Some(next) = best_next {
+            Ok(PrecomputedAfterKey::Next(next))
+        } else {
+            Ok(PrecomputedAfterKey::AfterLast)
+        }
+    }
+
+    fn precompute_term_ord(
+        str_dict_column: &Option<StrColumn>,
+        key: &str,
+        field: &str,
+    ) -> crate::Result<Self> {
+        let dict = str_dict_column
+            .as_ref()
+            .expect("dictionary missing for str accessor")
+            .dictionary();
+        let next_ord = dict.term_ord_or_next(key).map_err(|_| {
+            TantivyError::InvalidArgument(format!(
+                "failed to lookup after_key '{}' for field '{}'",
+                key, field
+            ))
+        })?;
+        Ok(next_ord.into())
+    }
+
+    /// Projects the after key into the column space of the given accessor.
+    ///
+    /// The computed after key will not take care of skipping entire columns
+    /// when the after key type is ordered after the accessor's type, that
+    /// should be performed earlier.
+    pub fn precompute(
+        composite_accessor: &CompositeAccessor,
+        source_after_key: &CompositeIntermediateKey,
+        field: &str,
+        missing_order: MissingOrder,
+        order: Order,
+    ) -> crate::Result<Self> {
+        use CompositeIntermediateKey as CIKey;
+        let precomputed_key = match (composite_accessor.column_type, source_after_key) {
+            (ColumnType::Bytes, _) => panic!("unsupported"),
+            // null after key
+            (_, CIKey::Null) => precompute_missing_after_key(false, missing_order, order),
+            // numerical
+            (ColumnType::I64, CIKey::I64(k)) => PrecomputedAfterKey::Exact(k.to_u64()),
+            (ColumnType::I64, CIKey::U64(k)) => num_proj::u64_to_i64(*k).into(),
+            (ColumnType::I64, CIKey::F64(k)) => num_proj::f64_to_i64(*k).into(),
+            (ColumnType::U64, CIKey::I64(k)) => num_proj::i64_to_u64(*k).into(),
+            (ColumnType::U64, CIKey::U64(k)) => PrecomputedAfterKey::Exact(*k),
+            (ColumnType::U64, CIKey::F64(k)) => num_proj::f64_to_u64(*k).into(),
+            (ColumnType::F64, CIKey::I64(k)) => num_proj::i64_to_f64(*k).into(),
+            (ColumnType::F64, CIKey::U64(k)) => num_proj::u64_to_f64(*k).into(),
+            (ColumnType::F64, CIKey::F64(k)) => PrecomputedAfterKey::Exact(k.to_u64()),
+            // boolean
+            (ColumnType::Bool, CIKey::Bool(key)) => PrecomputedAfterKey::Exact(key.to_u64()),
+            // string
+            (ColumnType::Str, CIKey::Str(key)) => PrecomputedAfterKey::precompute_term_ord(
+                &composite_accessor.str_dict_column,
+                key,
+                field,
+            )?,
+            // date time
+            (ColumnType::DateTime, CIKey::DateTime(key)) => {
+                PrecomputedAfterKey::Exact(key.to_u64())
+            }
+            // ip address
+            (ColumnType::IpAddr, CIKey::IpAddr(key)) => {
+                PrecomputedAfterKey::precompute_ip_addr(&composite_accessor.column, key)?
+            }
+            // assume the column's type is ordered after the after_key's type
+            _ => PrecomputedAfterKey::keep_all(order),
+        };
+        Ok(precomputed_key)
+    }
+
+    fn keep_all(order: Order) -> Self {
+        match order {
+            Order::Asc => PrecomputedAfterKey::Next(0),
+            Order::Desc => PrecomputedAfterKey::Next(u64::MAX),
+        }
+    }
+}

src/aggregation/bucket/composite/calendar_interval.rs

@@ -0,0 +1,140 @@
+use time::convert::{Day, Nanosecond};
+use time::{Time, UtcDateTime};
+
+const NS_IN_DAY: i64 = Nanosecond::per_t::<i128>(Day) as i64;
+
+/// Computes the timestamp in nanoseconds corresponding to the beginning of the
+/// year (January 1st at midnight UTC).
+pub(super) fn try_year_bucket(timestamp_ns: i64) -> crate::Result<i64> {
+    year_bucket_using_time_crate(timestamp_ns).map_err(|e| {
+        crate::TantivyError::InvalidArgument(format!(
+            "Failed to compute year bucket for timestamp {}: {}",
+            timestamp_ns,
+            e.to_string()
+        ))
+    })
+}
+
+/// Computes the timestamp in nanoseconds corresponding to the beginning of the
+/// month (1st at midnight UTC).
+pub(super) fn try_month_bucket(timestamp_ns: i64) -> crate::Result<i64> {
+    month_bucket_using_time_crate(timestamp_ns).map_err(|e| {
+        crate::TantivyError::InvalidArgument(format!(
+            "Failed to compute month bucket for timestamp {}: {}",
+            timestamp_ns,
+            e.to_string()
+        ))
+    })
+}
+
+/// Computes the timestamp in nanoseconds corresponding to the beginning of the
+/// week (Monday at midnight UTC).
+pub(super) fn week_bucket(timestamp_ns: i64) -> i64 {
+    // 1970-01-01 was a Thursday (weekday = 4)
+    let days_since_epoch = timestamp_ns.div_euclid(NS_IN_DAY);
+    // Find the weekday: 0=Monday, ..., 6=Sunday
+    let weekday = (days_since_epoch + 3).rem_euclid(7);
+    let monday_days_since_epoch = days_since_epoch - weekday;
+    monday_days_since_epoch * NS_IN_DAY
+}
+
+fn year_bucket_using_time_crate(timestamp_ns: i64) -> Result<i64, time::Error> {
+    let timestamp_ns = UtcDateTime::from_unix_timestamp_nanos(timestamp_ns as i128)?
+        .replace_ordinal(1)?
+        .replace_time(Time::MIDNIGHT)
+        .unix_timestamp_nanos();
+    Ok(timestamp_ns as i64)
+}
+
+fn month_bucket_using_time_crate(timestamp_ns: i64) -> Result<i64, time::Error> {
+    let timestamp_ns = UtcDateTime::from_unix_timestamp_nanos(timestamp_ns as i128)?
+        .replace_day(1)?
+        .replace_time(Time::MIDNIGHT)
+        .unix_timestamp_nanos();
+    Ok(timestamp_ns as i64)
+}
+
+#[cfg(test)]
+mod tests {
+    use std::i64;
+
+    use time::format_description::well_known::Iso8601;
+    use time::UtcDateTime;
+
+    use super::*;
+
+    fn ts_ns(iso: &str) -> i64 {
+        UtcDateTime::parse(iso, &Iso8601::DEFAULT)
+            .unwrap()
+            .unix_timestamp_nanos() as i64
+    }
+
+    #[test]
+    fn test_year_bucket() {
+        let ts = ts_ns("1970-01-01T00:00:00Z");
+        let res = try_year_bucket(ts).unwrap();
+        assert_eq!(res, ts_ns("1970-01-01T00:00:00Z"));
+
+        let ts = ts_ns("1970-06-01T10:00:01.010Z");
+        let res = try_year_bucket(ts).unwrap();
+        assert_eq!(res, ts_ns("1970-01-01T00:00:00Z"));
+
+        let ts = ts_ns("2008-12-31T23:59:59.999999999Z"); // leap year
+        let res = try_year_bucket(ts).unwrap();
+        assert_eq!(res, ts_ns("2008-01-01T00:00:00Z"));
+
+        let ts = ts_ns("2008-01-01T00:00:00Z"); // leap year
+        let res = try_year_bucket(ts).unwrap();
+        assert_eq!(res, ts_ns("2008-01-01T00:00:00Z"));
+
+        let ts = ts_ns("2010-12-31T23:59:59.999999999Z");
+        let res = try_year_bucket(ts).unwrap();
+        assert_eq!(res, ts_ns("2010-01-01T00:00:00Z"));
+
+        let ts = ts_ns("1972-06-01T00:10:00Z");
+        let res = try_year_bucket(ts).unwrap();
+        assert_eq!(res, ts_ns("1972-01-01T00:00:00Z"));
+    }
+
+    #[test]
+    fn test_month_bucket() {
+        let ts = ts_ns("1970-01-15T00:00:00Z");
+        let res = try_month_bucket(ts).unwrap();
+        assert_eq!(res, ts_ns("1970-01-01T00:00:00Z"));
+
+        let ts = ts_ns("1970-02-01T00:00:00Z");
+        let res = try_month_bucket(ts).unwrap();
+        assert_eq!(res, ts_ns("1970-02-01T00:00:00Z"));
+
+        let ts = ts_ns("2000-01-31T23:59:59.999999999Z");
+        let res = try_month_bucket(ts).unwrap();
+        assert_eq!(res, ts_ns("2000-01-01T00:00:00Z"));
+    }
+
+    #[test]
+    fn test_week_bucket() {
+        let ts = ts_ns("1970-01-05T00:00:00Z"); // Monday
+        let res = week_bucket(ts);
+        assert_eq!(res, ts_ns("1970-01-05T00:00:00Z"));
+
+        let ts = ts_ns("1970-01-05T23:59:59Z"); // Monday
+        let res = week_bucket(ts);
+        assert_eq!(res, ts_ns("1970-01-05T00:00:00Z"));
+
+        let ts = ts_ns("1970-01-07T01:13:00Z"); // Wednesday
+        let res = week_bucket(ts);
+        assert_eq!(res, ts_ns("1970-01-05T00:00:00Z"));
+
+        let ts = ts_ns("1970-01-11T23:59:59.999999999Z"); // Sunday
+        let res = week_bucket(ts);
+        assert_eq!(res, ts_ns("1970-01-05T00:00:00Z"));
+
+        let ts = ts_ns("2025-10-16T10:41:59.010Z"); // Thursday
+        let res = week_bucket(ts);
+        assert_eq!(res, ts_ns("2025-10-13T00:00:00Z"));
+
+        let ts = ts_ns("1970-01-01T00:00:00Z"); // Thursday
+        let res = week_bucket(ts);
+        assert_eq!(res, ts_ns("1969-12-29T00:00:00Z")); // Negative
+    }
+}

src/aggregation/bucket/composite/collector.rs

@@ -0,0 +1,674 @@
+use std::fmt::Debug;
+use std::mem;
+use std::net::Ipv6Addr;
+
+use columnar::column_values::CompactSpaceU64Accessor;
+use columnar::{
+    Column, ColumnType, Dictionary, MonotonicallyMappableToU128, MonotonicallyMappableToU64,
+    NumericalValue, StrColumn,
+};
+use rustc_hash::FxHashMap;
+use smallvec::SmallVec;
+
+use crate::aggregation::agg_data::{
+    build_segment_agg_collectors, AggRefNode, AggregationsSegmentCtx,
+};
+use crate::aggregation::bucket::composite::accessors::{
+    CompositeAccessor, CompositeAggReqData, PrecomputedDateInterval,
+};
+use crate::aggregation::bucket::composite::calendar_interval;
+use crate::aggregation::bucket::composite::map::{DynArrayHeapMap, MAX_DYN_ARRAY_SIZE};
+use crate::aggregation::bucket::{
+    CalendarInterval, CompositeAggregationSource, MissingOrder, Order,
+};
+use crate::aggregation::cached_sub_aggs::{CachedSubAggs, HighCardSubAggCache};
+use crate::aggregation::intermediate_agg_result::{
+    CompositeIntermediateKey, IntermediateAggregationResult, IntermediateAggregationResults,
+    IntermediateBucketResult, IntermediateCompositeBucketEntry, IntermediateCompositeBucketResult,
+};
+use crate::aggregation::segment_agg_result::{BucketIdProvider, SegmentAggregationCollector};
+use crate::aggregation::BucketId;
+use crate::TantivyError;
+
+#[derive(Clone, Debug)]
+struct CompositeBucketCollector {
+    count: u32,
+    bucket_id: BucketId,
+}
+
+/// Compact sortable representation of a single source value within a composite key.
+///
+/// The struct encodes both the column identity and the fast field value in a way
+/// that preserves the desired sort order via the derived `Ord` implementation
+/// (fields are compared top-to-bottom: `sort_key` first, then `encoded_value`).
+///
+/// ## `sort_key` encoding
+/// - `0` — missing value, sorted first
+/// - `1..=254` — present value; the original accessor index is `sort_key - 1`
+/// - `u8::MAX` (255) — missing value, sorted last
+///
+/// ## `encoded_value` encoding
+/// - `0` when the field is missing
+/// - The raw u64 fast-field representation when order is ascending
+/// - Bitwise NOT of the raw u64 when order is descending
+#[derive(Clone, Copy, Debug, PartialEq, Eq, PartialOrd, Ord, Default, Hash)]
+struct InternalValueRepr {
+    /// Column index biased by +1 (so 0 and u8::MAX are reserved for missing sentinels).
+    sort_key: u8,
+    /// Fast field value, possibly bit-flipped for descending order.
+    encoded_value: u64,
+}
+
+impl InternalValueRepr {
+    #[inline]
+    fn new_term(raw: u64, accessor_idx: u8, order: Order) -> Self {
+        let encoded_value = match order {
+            Order::Asc => raw,
+            Order::Desc => !raw,
+        };
+        InternalValueRepr {
+            sort_key: accessor_idx + 1,
+            encoded_value,
+        }
+    }
+
+    /// For histogram sources the column index is irrelevant (always 1).
+    #[inline]
+    fn new_histogram(raw: u64, order: Order) -> Self {
+        let encoded_value = match order {
+            Order::Asc => raw,
+            Order::Desc => !raw,
+        };
+        InternalValueRepr {
+            sort_key: 1,
+            encoded_value,
+        }
+    }
+
+    #[inline]
+    fn new_missing(order: Order, missing_order: MissingOrder) -> Self {
+        let sort_key = match (missing_order, order) {
+            (MissingOrder::First, _) | (MissingOrder::Default, Order::Asc) => 0,
+            (MissingOrder::Last, _) | (MissingOrder::Default, Order::Desc) => u8::MAX,
+        };
+        InternalValueRepr {
+            sort_key,
+            encoded_value: 0,
+        }
+    }
+
+    /// Decode back to `(accessor_idx, raw_value)`.
+    /// Returns `None` when the value represents a missing field.
+    #[inline]
+    fn decode(self, order: Order) -> Option<(u8, u64)> {
+        if self.sort_key == 0 || self.sort_key == u8::MAX {
+            return None;
+        }
+        let raw = match order {
+            Order::Asc => self.encoded_value,
+            Order::Desc => !self.encoded_value,
+        };
+        Some((self.sort_key - 1, raw))
+    }
+}
+
+/// The collector puts values from the fast field into the correct buckets and
+/// does a conversion to the correct datatype.
+#[derive(Debug)]
+pub struct SegmentCompositeCollector {
+    /// One DynArrayHeapMap per parent bucket.
+    parent_buckets: Vec<DynArrayHeapMap<InternalValueRepr, CompositeBucketCollector>>,
+    accessor_idx: usize,
+    sub_agg: Option<CachedSubAggs<HighCardSubAggCache>>,
+    bucket_id_provider: BucketIdProvider,
+    /// Number of sources, needed when creating new DynArrayHeapMaps.
+    num_sources: usize,
+}
+
+impl SegmentAggregationCollector for SegmentCompositeCollector {
+    fn add_intermediate_aggregation_result(
+        &mut self,
+        agg_data: &AggregationsSegmentCtx,
+        results: &mut IntermediateAggregationResults,
+        parent_bucket_id: BucketId,
+    ) -> crate::Result<()> {
+        let name = agg_data
+            .get_composite_req_data(self.accessor_idx)
+            .name
+            .clone();
+
+        let buckets = self.into_intermediate_bucket_result(agg_data, parent_bucket_id)?;
+        results.push(
+            name,
+            IntermediateAggregationResult::Bucket(IntermediateBucketResult::Composite { buckets }),
+        )?;
+
+        Ok(())
+    }
+
+    fn collect(
+        &mut self,
+        parent_bucket_id: BucketId,
+        docs: &[crate::DocId],
+        agg_data: &mut AggregationsSegmentCtx,
+    ) -> crate::Result<()> {
+        let mem_pre = self.get_memory_consumption();
+        let composite_agg_data = agg_data.take_composite_req_data(self.accessor_idx);
+
+        for doc in docs {
+            let mut sub_level_values = SmallVec::new();
+            recursive_key_visitor(
+                *doc,
+                &composite_agg_data,
+                0,
+                &mut sub_level_values,
+                &mut self.parent_buckets[parent_bucket_id as usize],
+                true,
+                &mut self.sub_agg,
+                &mut self.bucket_id_provider,
+            )?;
+        }
+        agg_data.put_back_composite_req_data(self.accessor_idx, composite_agg_data);
+
+        if let Some(sub_agg) = &mut self.sub_agg {
+            sub_agg.check_flush_local(agg_data)?;
+        }
+
+        let mem_delta = self.get_memory_consumption() - mem_pre;
+        if mem_delta > 0 {
+            agg_data.context.limits.add_memory_consumed(mem_delta)?;
+        }
+
+        Ok(())
+    }
+
+    fn flush(&mut self, agg_data: &mut AggregationsSegmentCtx) -> crate::Result<()> {
+        if let Some(sub_agg) = &mut self.sub_agg {
+            sub_agg.flush(agg_data)?;
+        }
+        Ok(())
+    }
+
+    fn prepare_max_bucket(
+        &mut self,
+        max_bucket: BucketId,
+        _agg_data: &AggregationsSegmentCtx,
+    ) -> crate::Result<()> {
+        let required_len = max_bucket as usize + 1;
+        while self.parent_buckets.len() < required_len {
+            let map = DynArrayHeapMap::try_new(self.num_sources)?;
+            self.parent_buckets.push(map);
+        }
+        Ok(())
+    }
+}
+
+impl SegmentCompositeCollector {
+    fn get_memory_consumption(&self) -> u64 {
+        self.parent_buckets
+            .iter()
+            .map(|m| m.memory_consumption())
+            .sum()
+    }
+
+    pub(crate) fn from_req_and_validate(
+        req_data: &mut AggregationsSegmentCtx,
+        node: &AggRefNode,
+    ) -> crate::Result<Self> {
+        validate_req(req_data, node.idx_in_req_data)?;
+
+        let has_sub_aggregations = !node.children.is_empty();
+        let sub_agg = if has_sub_aggregations {
+            let sub_agg_collector = build_segment_agg_collectors(req_data, &node.children)?;
+            Some(CachedSubAggs::new(sub_agg_collector))
+        } else {
+            None
+        };
+
+        let composite_req_data = req_data.get_composite_req_data(node.idx_in_req_data);
+        let num_sources = composite_req_data.req.sources.len();
+
+        Ok(SegmentCompositeCollector {
+            parent_buckets: vec![DynArrayHeapMap::try_new(num_sources)?],
+            accessor_idx: node.idx_in_req_data,
+            sub_agg,
+            bucket_id_provider: BucketIdProvider::default(),
+            num_sources,
+        })
+    }
+
+    #[inline]
+    fn into_intermediate_bucket_result(
+        &mut self,
+        agg_data: &AggregationsSegmentCtx,
+        parent_bucket_id: BucketId,
+    ) -> crate::Result<IntermediateCompositeBucketResult> {
+        let empty_map = DynArrayHeapMap::try_new(self.num_sources)?;
+        let heap_map = mem::replace(
+            &mut self.parent_buckets[parent_bucket_id as usize],
+            empty_map,
+        );
+
+        let mut dict: FxHashMap<Vec<CompositeIntermediateKey>, IntermediateCompositeBucketEntry> =
+            Default::default();
+        dict.reserve(heap_map.size());
+        let composite_data = agg_data.get_composite_req_data(self.accessor_idx);
+        for (key_internal_repr, agg) in heap_map.into_iter() {
+            let key = resolve_key(&key_internal_repr, composite_data)?;
+            let mut sub_aggregation_res = IntermediateAggregationResults::default();
+            if let Some(sub_agg) = &mut self.sub_agg {
+                sub_agg
+                    .get_sub_agg_collector()
+                    .add_intermediate_aggregation_result(
+                        agg_data,
+                        &mut sub_aggregation_res,
+                        agg.bucket_id,
+                    )?;
+            }
+
+            dict.insert(
+                key,
+                IntermediateCompositeBucketEntry {
+                    doc_count: agg.count,
+                    sub_aggregation: sub_aggregation_res,
+                },
+            );
+        }
+
+        Ok(IntermediateCompositeBucketResult {
+            entries: dict,
+            target_size: composite_data.req.size,
+            orders: composite_data
+                .req
+                .sources
+                .iter()
+                .map(|source| match source {
+                    CompositeAggregationSource::Terms(t) => (t.order, t.missing_order),
+                    CompositeAggregationSource::Histogram(h) => (h.order, h.missing_order),
+                    CompositeAggregationSource::DateHistogram(d) => (d.order, d.missing_order),
+                })
+                .collect(),
+        })
+    }
+}
+
+fn validate_req(req_data: &mut AggregationsSegmentCtx, accessor_idx: usize) -> crate::Result<()> {
+    let composite_data = req_data.get_composite_req_data(accessor_idx);
+    let req = &composite_data.req;
+    if req.sources.is_empty() {
+        return Err(TantivyError::InvalidArgument(
+            "composite aggregation must have at least one source".to_string(),
+        ));
+    }
+    if req.size == 0 {
+        return Err(TantivyError::InvalidArgument(
+            "composite aggregation 'size' must be > 0".to_string(),
+        ));
+    }
+    let column_types_for_sources = composite_data.composite_accessors.iter().map(|item| {
+        item.accessors
+            .iter()
+            .map(|a| a.column_type)
+            .collect::<Vec<_>>()
+    });
+
+    for column_types in column_types_for_sources {
+        if column_types.len() > MAX_DYN_ARRAY_SIZE {
+            return Err(TantivyError::InvalidArgument(format!(
+                "composite aggregation source supports maximum {MAX_DYN_ARRAY_SIZE} sources",
+            )));
+        }
+        if column_types.contains(&ColumnType::Bytes) {
+            return Err(TantivyError::InvalidArgument(
+                "composite aggregation does not support 'bytes' field type".to_string(),
+            ));
+        }
+    }
+    Ok(())
+}
+
+fn collect_bucket_with_limit(
+    doc_id: crate::DocId,
+    limit_num_buckets: usize,
+    buckets: &mut DynArrayHeapMap<InternalValueRepr, CompositeBucketCollector>,
+    key: &[InternalValueRepr],
+    sub_agg: &mut Option<CachedSubAggs<HighCardSubAggCache>>,
+    bucket_id_provider: &mut BucketIdProvider,
+) {
+    let mut record_in_bucket = |bucket: &mut CompositeBucketCollector| {
+        bucket.count += 1;
+        if let Some(sub_agg) = sub_agg {
+            sub_agg.push(bucket.bucket_id, doc_id);
+        }
+    };
+
+    // We still have room for buckets, just insert
+    if buckets.size() < limit_num_buckets {
+        let bucket = buckets.get_or_insert_with(key, || CompositeBucketCollector {
+            count: 0,
+            bucket_id: bucket_id_provider.next_bucket_id(),
+        });
+        record_in_bucket(bucket);
+        return;
+    }
+
+    // Map is full, but we can still update the bucket if it already exists
+    if let Some(bucket) = buckets.get_mut(key) {
+        record_in_bucket(bucket);
+        return;
+    }
+
+    // Check if the item qualifies to enter the top-k, and evict the highest if it does
+    if let Some(highest_key) = buckets.peek_highest() {
+        if key < highest_key {
+            buckets.evict_highest();
+            let bucket = buckets.get_or_insert_with(key, || CompositeBucketCollector {
+                count: 0,
+                bucket_id: bucket_id_provider.next_bucket_id(),
+            });
+            record_in_bucket(bucket);
+        }
+    }
+}
+
+/// Converts the composite key from its internal column space representation
+/// (segment specific) into its intermediate form.
+fn resolve_key(
+    internal_key: &[InternalValueRepr],
+    agg_data: &CompositeAggReqData,
+) -> crate::Result<Vec<CompositeIntermediateKey>> {
+    internal_key
+        .iter()
+        .enumerate()
+        .map(|(idx, val)| {
+            resolve_internal_value_repr(
+                *val,
+                &agg_data.req.sources[idx],
+                &agg_data.composite_accessors[idx].accessors,
+            )
+        })
+        .collect()
+}
+
+fn resolve_internal_value_repr(
+    internal_value_repr: InternalValueRepr,
+    source: &CompositeAggregationSource,
+    composite_accessors: &[CompositeAccessor],
+) -> crate::Result<CompositeIntermediateKey> {
+    let decoded_value_opt = match source {
+        CompositeAggregationSource::Terms(source) => internal_value_repr.decode(source.order),
+        CompositeAggregationSource::Histogram(source) => internal_value_repr.decode(source.order),
+        CompositeAggregationSource::DateHistogram(source) => {
+            internal_value_repr.decode(source.order)
+        }
+    };
+    let Some((decoded_accessor_idx, val)) = decoded_value_opt else {
+        return Ok(CompositeIntermediateKey::Null);
+    };
+    let key = match source {
+        CompositeAggregationSource::Terms(_) => {
+            let CompositeAccessor {
+                column_type,
+                str_dict_column,
+                column,
+                ..
+            } = &composite_accessors[decoded_accessor_idx as usize];
+            resolve_term(val, column_type, str_dict_column, column)?
+        }
+        CompositeAggregationSource::Histogram(source) => {
+            CompositeIntermediateKey::F64(i64::from_u64(val) as f64 * source.interval)
+        }
+        CompositeAggregationSource::DateHistogram(_) => {
+            CompositeIntermediateKey::DateTime(i64::from_u64(val))
+        }
+    };
+
+    Ok(key)
+}
+
+fn resolve_term(
+    val: u64,
+    column_type: &ColumnType,
+    str_dict_column: &Option<StrColumn>,
+    column: &Column,
+) -> crate::Result<CompositeIntermediateKey> {
+    let key = if *column_type == ColumnType::Str {
+        let fallback_dict = Dictionary::empty();
+        let term_dict = str_dict_column
+            .as_ref()
+            .map(|el| el.dictionary())
+            .unwrap_or_else(|| &fallback_dict);
+
+        let mut buffer = Vec::new();
+        term_dict.ord_to_term(val, &mut buffer)?;
+        CompositeIntermediateKey::Str(
+            String::from_utf8(buffer.to_vec()).expect("could not convert to String"),
+        )
+    } else if *column_type == ColumnType::DateTime {
+        let val = i64::from_u64(val);
+        CompositeIntermediateKey::DateTime(val)
+    } else if *column_type == ColumnType::Bool {
+        let val = bool::from_u64(val);
+        CompositeIntermediateKey::Bool(val)
+    } else if *column_type == ColumnType::IpAddr {
+        let compact_space_accessor = column
+            .values
+            .clone()
+            .downcast_arc::<CompactSpaceU64Accessor>()
+            .map_err(|_| {
+                TantivyError::AggregationError(crate::aggregation::AggregationError::InternalError(
+                    "Type mismatch: Could not downcast to CompactSpaceU64Accessor".to_string(),
+                ))
+            })?;
+        let val: u128 = compact_space_accessor.compact_to_u128(val as u32);
+        let val = Ipv6Addr::from_u128(val);
+        CompositeIntermediateKey::IpAddr(val)
+    } else if *column_type == ColumnType::U64 {
+        CompositeIntermediateKey::U64(val)
+    } else if *column_type == ColumnType::I64 {
+        CompositeIntermediateKey::I64(i64::from_u64(val))
+    } else {
+        let val = f64::from_u64(val);
+        let val: NumericalValue = val.into();
+
+        match val.normalize() {
+            NumericalValue::U64(val) => CompositeIntermediateKey::U64(val),
+            NumericalValue::I64(val) => CompositeIntermediateKey::I64(val),
+            NumericalValue::F64(val) => CompositeIntermediateKey::F64(val),
+        }
+    };
+    Ok(key)
+}
+
+/// Depth-first walk of the accessors to build the composite key combinations
+/// and update the buckets.
+fn recursive_key_visitor(
+    doc_id: crate::DocId,
+    composite_agg_data: &CompositeAggReqData,
+    source_idx_for_recursion: usize,
+    sub_level_values: &mut SmallVec<[InternalValueRepr; MAX_DYN_ARRAY_SIZE]>,
+    buckets: &mut DynArrayHeapMap<InternalValueRepr, CompositeBucketCollector>,
+    // whether we need to consider the after_key in the following levels
+    is_on_after_key: bool,
+    sub_agg: &mut Option<CachedSubAggs<HighCardSubAggCache>>,
+    bucket_id_provider: &mut BucketIdProvider,
+) -> crate::Result<()> {
+    if source_idx_for_recursion == composite_agg_data.req.sources.len() {
+        if !is_on_after_key {
+            collect_bucket_with_limit(
+                doc_id,
+                composite_agg_data.req.size as usize,
+                buckets,
+                sub_level_values,
+                sub_agg,
+                bucket_id_provider,
+            );
+        }
+        return Ok(());
+    }
+
+    let current_level_accessors = &composite_agg_data.composite_accessors[source_idx_for_recursion];
+    let current_level_source = &composite_agg_data.req.sources[source_idx_for_recursion];
+    let mut missing = true;
+    for (accessor_idx, accessor) in current_level_accessors.accessors.iter().enumerate() {
+        let values = accessor.column.values_for_doc(doc_id);
+        for value in values {
+            missing = false;
+            match current_level_source {
+                CompositeAggregationSource::Terms(_) => {
+                    let preceeds_after_key_type =
+                        accessor_idx < current_level_accessors.after_key_accessor_idx;
+                    if is_on_after_key && preceeds_after_key_type {
+                        break;
+                    }
+                    let matches_after_key_type =
+                        accessor_idx == current_level_accessors.after_key_accessor_idx;
+
+                    if matches_after_key_type && is_on_after_key {
+                        let should_skip = match current_level_source.order() {
+                            Order::Asc => current_level_accessors.after_key.gt(value),
+                            Order::Desc => current_level_accessors.after_key.lt(value),
+                        };
+                        if should_skip {
+                            continue;
+                        }
+                    }
+                    sub_level_values.push(InternalValueRepr::new_term(
+                        value,
+                        accessor_idx as u8,
+                        current_level_source.order(),
+                    ));
+                    let still_on_after_key =
+                        matches_after_key_type && current_level_accessors.after_key.equals(value);
+                    recursive_key_visitor(
+                        doc_id,
+                        composite_agg_data,
+                        source_idx_for_recursion + 1,
+                        sub_level_values,
+                        buckets,
+                        is_on_after_key && still_on_after_key,
+                        sub_agg,
+                        bucket_id_provider,
+                    )?;
+                    sub_level_values.pop();
+                }
+                CompositeAggregationSource::Histogram(source) => {
+                    let float_value = match accessor.column_type {
+                        ColumnType::U64 => value as f64,
+                        ColumnType::I64 => i64::from_u64(value) as f64,
+                        ColumnType::DateTime => i64::from_u64(value) as f64 / 1_000_000.,
+                        ColumnType::F64 => f64::from_u64(value),
+                        _ => {
+                            panic!(
+                                "unexpected type {:?}. This should not happen",
+                                accessor.column_type
+                            )
+                        }
+                    };
+                    let bucket_index = (float_value / source.interval).floor() as i64;
+                    let bucket_value = i64::to_u64(bucket_index);
+                    if is_on_after_key {
+                        let should_skip = match current_level_source.order() {
+                            Order::Asc => current_level_accessors.after_key.gt(bucket_value),
+                            Order::Desc => current_level_accessors.after_key.lt(bucket_value),
+                        };
+                        if should_skip {
+                            continue;
+                        }
+                    }
+                    sub_level_values.push(InternalValueRepr::new_histogram(
+                        bucket_value,
+                        current_level_source.order(),
+                    ));
+                    let still_on_after_key = current_level_accessors.after_key.equals(bucket_value);
+                    recursive_key_visitor(
+                        doc_id,
+                        composite_agg_data,
+                        source_idx_for_recursion + 1,
+                        sub_level_values,
+                        buckets,
+                        is_on_after_key && still_on_after_key,
+                        sub_agg,
+                        bucket_id_provider,
+                    )?;
+                    sub_level_values.pop();
+                }
+                CompositeAggregationSource::DateHistogram(_) => {
+                    let value_ns = match accessor.column_type {
+                        ColumnType::DateTime => i64::from_u64(value),
+                        _ => {
+                            panic!(
+                                "unexpected type {:?}. This should not happen",
+                                accessor.column_type
+                            )
+                        }
+                    };
+                    let bucket_index = match accessor.date_histogram_interval {
+                        PrecomputedDateInterval::FixedNanoseconds(fixed_interval_ns) => {
+                            (value_ns / fixed_interval_ns) * fixed_interval_ns
+                        }
+                        PrecomputedDateInterval::Calendar(CalendarInterval::Year) => {
+                            calendar_interval::try_year_bucket(value_ns)?
+                        }
+                        PrecomputedDateInterval::Calendar(CalendarInterval::Month) => {
+                            calendar_interval::try_month_bucket(value_ns)?
+                        }
+                        PrecomputedDateInterval::Calendar(CalendarInterval::Week) => {
+                            calendar_interval::week_bucket(value_ns)
+                        }
+                        PrecomputedDateInterval::NotApplicable => {
+                            panic!("interval not precomputed for date histogram source")
+                        }
+                    };
+                    let bucket_value = i64::to_u64(bucket_index);
+                    if is_on_after_key {
+                        let should_skip = match current_level_source.order() {
+                            Order::Asc => current_level_accessors.after_key.gt(bucket_value),
+                            Order::Desc => current_level_accessors.after_key.lt(bucket_value),
+                        };
+                        if should_skip {
+                            continue;
+                        }
+                    }
+                    sub_level_values.push(InternalValueRepr::new_histogram(
+                        bucket_value,
+                        current_level_source.order(),
+                    ));
+                    let still_on_after_key = current_level_accessors.after_key.equals(bucket_value);
+                    recursive_key_visitor(
+                        doc_id,
+                        composite_agg_data,
+                        source_idx_for_recursion + 1,
+                        sub_level_values,
+                        buckets,
+                        is_on_after_key && still_on_after_key,
+                        sub_agg,
+                        bucket_id_provider,
+                    )?;
+                    sub_level_values.pop();
+                }
+            };
+        }
+    }
+    if missing && current_level_source.missing_bucket() {
+        if is_on_after_key && current_level_accessors.skip_missing {
+            return Ok(());
+        }
+        sub_level_values.push(InternalValueRepr::new_missing(
+            current_level_source.order(),
+            current_level_source.missing_order(),
+        ));
+        recursive_key_visitor(
+            doc_id,
+            composite_agg_data,
+            source_idx_for_recursion + 1,
+            sub_level_values,
+            buckets,
+            is_on_after_key && current_level_accessors.is_after_key_explicit_missing,
+            sub_agg,
+            bucket_id_provider,
+        )?;
+        sub_level_values.pop();
+    }
+    Ok(())
+}

src/aggregation/bucket/composite/map.rs

@@ -0,0 +1,329 @@
+use std::collections::BinaryHeap;
+use std::fmt::Debug;
+use std::hash::Hash;
+
+use rustc_hash::FxHashMap;
+use smallvec::SmallVec;
+
+use crate::TantivyError;
+
+/// Map backed by a hash map for fast access and a binary heap to track the
+/// highest key. The key is an array of fixed size S.
+#[derive(Clone, Debug)]
+struct ArrayHeapMap<K: Ord, V, const S: usize> {
+    pub(crate) buckets: FxHashMap<[K; S], V>,
+    pub(crate) heap: BinaryHeap<[K; S]>,
+}
+
+impl<K: Ord, V, const S: usize> Default for ArrayHeapMap<K, V, S> {
+    fn default() -> Self {
+        ArrayHeapMap {
+            buckets: FxHashMap::default(),
+            heap: BinaryHeap::default(),
+        }
+    }
+}
+
+impl<K: Eq + Hash + Clone + Ord, V, const S: usize> ArrayHeapMap<K, V, S> {
+    /// Panics if the length of `key` is not S.
+    fn get_or_insert_with<F: FnOnce() -> V>(&mut self, key: &[K], f: F) -> &mut V {
+        let key_array: &[K; S] = key.try_into().expect("Key length mismatch");
+        self.buckets.entry(key_array.clone()).or_insert_with(|| {
+            self.heap.push(key_array.clone());
+            f()
+        })
+    }
+
+    /// Panics if the length of `key` is not S.
+    fn get_mut(&mut self, key: &[K]) -> Option<&mut V> {
+        let key_array: &[K; S] = key.try_into().expect("Key length mismatch");
+        self.buckets.get_mut(key_array)
+    }
+
+    fn peek_highest(&self) -> Option<&[K]> {
+        self.heap.peek().map(|k_array| k_array.as_slice())
+    }
+
+    fn evict_highest(&mut self) {
+        if let Some(highest) = self.heap.pop() {
+            self.buckets.remove(&highest);
+        }
+    }
+
+    fn memory_consumption(&self) -> u64 {
+        let key_size = std::mem::size_of::<[K; S]>();
+        let map_size = (key_size + std::mem::size_of::<V>()) * self.buckets.capacity();
+        let heap_size = key_size * self.heap.capacity();
+        (map_size + heap_size) as u64
+    }
+}
+
+impl<K: Copy + Ord + Clone + 'static, V: 'static, const S: usize> ArrayHeapMap<K, V, S> {
+    fn into_iter(self) -> Box<dyn Iterator<Item = (SmallVec<[K; MAX_DYN_ARRAY_SIZE]>, V)>> {
+        Box::new(
+            self.buckets
+                .into_iter()
+                .map(|(k, v)| (SmallVec::from_slice(&k), v)),
+        )
+    }
+}
+
+pub(super) const MAX_DYN_ARRAY_SIZE: usize = 16;
+const MAX_DYN_ARRAY_SIZE_PLUS_ONE: usize = MAX_DYN_ARRAY_SIZE + 1;
+
+/// A map optimized for memory footprint, fast access and efficient eviction of
+/// the highest key.
+///
+/// Keys are inlined arrays of size 1 to [MAX_DYN_ARRAY_SIZE] but for a given
+/// instance the key size is fixed. This allows to avoid heap allocations for the
+/// keys.
+#[derive(Clone, Debug)]
+pub(super) struct DynArrayHeapMap<K: Ord, V>(DynArrayHeapMapInner<K, V>);
+
+/// Wrapper around ArrayHeapMap to dynamically dispatch on the array size.
+#[derive(Clone, Debug)]
+enum DynArrayHeapMapInner<K: Ord, V> {
+    Dim1(ArrayHeapMap<K, V, 1>),
+    Dim2(ArrayHeapMap<K, V, 2>),
+    Dim3(ArrayHeapMap<K, V, 3>),
+    Dim4(ArrayHeapMap<K, V, 4>),
+    Dim5(ArrayHeapMap<K, V, 5>),
+    Dim6(ArrayHeapMap<K, V, 6>),
+    Dim7(ArrayHeapMap<K, V, 7>),
+    Dim8(ArrayHeapMap<K, V, 8>),
+    Dim9(ArrayHeapMap<K, V, 9>),
+    Dim10(ArrayHeapMap<K, V, 10>),
+    Dim11(ArrayHeapMap<K, V, 11>),
+    Dim12(ArrayHeapMap<K, V, 12>),
+    Dim13(ArrayHeapMap<K, V, 13>),
+    Dim14(ArrayHeapMap<K, V, 14>),
+    Dim15(ArrayHeapMap<K, V, 15>),
+    Dim16(ArrayHeapMap<K, V, 16>),
+}
+
+impl<K: Ord, V> DynArrayHeapMap<K, V> {
+    /// Creates a new heap map with dynamic array keys of size `key_dimension`.
+    pub(super) fn try_new(key_dimension: usize) -> crate::Result<Self> {
+        let inner = match key_dimension {
+            0 => {
+                return Err(TantivyError::InvalidArgument(
+                    "DynArrayHeapMap dimension must be at least 1".to_string(),
+                ))
+            }
+            1 => DynArrayHeapMapInner::Dim1(ArrayHeapMap::default()),
+            2 => DynArrayHeapMapInner::Dim2(ArrayHeapMap::default()),
+            3 => DynArrayHeapMapInner::Dim3(ArrayHeapMap::default()),
+            4 => DynArrayHeapMapInner::Dim4(ArrayHeapMap::default()),
+            5 => DynArrayHeapMapInner::Dim5(ArrayHeapMap::default()),
+            6 => DynArrayHeapMapInner::Dim6(ArrayHeapMap::default()),
+            7 => DynArrayHeapMapInner::Dim7(ArrayHeapMap::default()),
+            8 => DynArrayHeapMapInner::Dim8(ArrayHeapMap::default()),
+            9 => DynArrayHeapMapInner::Dim9(ArrayHeapMap::default()),
+            10 => DynArrayHeapMapInner::Dim10(ArrayHeapMap::default()),
+            11 => DynArrayHeapMapInner::Dim11(ArrayHeapMap::default()),
+            12 => DynArrayHeapMapInner::Dim12(ArrayHeapMap::default()),
+            13 => DynArrayHeapMapInner::Dim13(ArrayHeapMap::default()),
+            14 => DynArrayHeapMapInner::Dim14(ArrayHeapMap::default()),
+            15 => DynArrayHeapMapInner::Dim15(ArrayHeapMap::default()),
+            16 => DynArrayHeapMapInner::Dim16(ArrayHeapMap::default()),
+            MAX_DYN_ARRAY_SIZE_PLUS_ONE.. => {
+                return Err(TantivyError::InvalidArgument(format!(
+                    "DynArrayHeapMap supports maximum {MAX_DYN_ARRAY_SIZE} dimensions, got \
+                     {key_dimension}",
+                )))
+            }
+        };
+        Ok(DynArrayHeapMap(inner))
+    }
+
+    /// Number of elements in the map. This is not the dimension of the keys.
+    pub(super) fn size(&self) -> usize {
+        match &self.0 {
+            DynArrayHeapMapInner::Dim1(map) => map.buckets.len(),
+            DynArrayHeapMapInner::Dim2(map) => map.buckets.len(),
+            DynArrayHeapMapInner::Dim3(map) => map.buckets.len(),
+            DynArrayHeapMapInner::Dim4(map) => map.buckets.len(),
+            DynArrayHeapMapInner::Dim5(map) => map.buckets.len(),
+            DynArrayHeapMapInner::Dim6(map) => map.buckets.len(),
+            DynArrayHeapMapInner::Dim7(map) => map.buckets.len(),
+            DynArrayHeapMapInner::Dim8(map) => map.buckets.len(),
+            DynArrayHeapMapInner::Dim9(map) => map.buckets.len(),
+            DynArrayHeapMapInner::Dim10(map) => map.buckets.len(),
+            DynArrayHeapMapInner::Dim11(map) => map.buckets.len(),
+            DynArrayHeapMapInner::Dim12(map) => map.buckets.len(),
+            DynArrayHeapMapInner::Dim13(map) => map.buckets.len(),
+            DynArrayHeapMapInner::Dim14(map) => map.buckets.len(),
+            DynArrayHeapMapInner::Dim15(map) => map.buckets.len(),
+            DynArrayHeapMapInner::Dim16(map) => map.buckets.len(),
+        }
+    }
+}
+
+impl<K: Ord + Hash + Clone, V> DynArrayHeapMap<K, V> {
+    /// Get a mutable reference to the value corresponding to `key` or inserts a new
+    /// value created by calling `f`.
+    ///
+    /// Panics if the length of `key` does not match the key dimension of the map.
+    pub(super) fn get_or_insert_with<F: FnOnce() -> V>(&mut self, key: &[K], f: F) -> &mut V {
+        match &mut self.0 {
+            DynArrayHeapMapInner::Dim1(map) => map.get_or_insert_with(key, f),
+            DynArrayHeapMapInner::Dim2(map) => map.get_or_insert_with(key, f),
+            DynArrayHeapMapInner::Dim3(map) => map.get_or_insert_with(key, f),
+            DynArrayHeapMapInner::Dim4(map) => map.get_or_insert_with(key, f),
+            DynArrayHeapMapInner::Dim5(map) => map.get_or_insert_with(key, f),
+            DynArrayHeapMapInner::Dim6(map) => map.get_or_insert_with(key, f),
+            DynArrayHeapMapInner::Dim7(map) => map.get_or_insert_with(key, f),
+            DynArrayHeapMapInner::Dim8(map) => map.get_or_insert_with(key, f),
+            DynArrayHeapMapInner::Dim9(map) => map.get_or_insert_with(key, f),
+            DynArrayHeapMapInner::Dim10(map) => map.get_or_insert_with(key, f),
+            DynArrayHeapMapInner::Dim11(map) => map.get_or_insert_with(key, f),
+            DynArrayHeapMapInner::Dim12(map) => map.get_or_insert_with(key, f),
+            DynArrayHeapMapInner::Dim13(map) => map.get_or_insert_with(key, f),
+            DynArrayHeapMapInner::Dim14(map) => map.get_or_insert_with(key, f),
+            DynArrayHeapMapInner::Dim15(map) => map.get_or_insert_with(key, f),
+            DynArrayHeapMapInner::Dim16(map) => map.get_or_insert_with(key, f),
+        }
+    }
+
+    /// Returns a mutable reference to the value corresponding to `key`.
+    ///
+    /// Panics if the length of `key` does not match the key dimension of the map.
+    pub fn get_mut(&mut self, key: &[K]) -> Option<&mut V> {
+        match &mut self.0 {
+            DynArrayHeapMapInner::Dim1(map) => map.get_mut(key),
+            DynArrayHeapMapInner::Dim2(map) => map.get_mut(key),
+            DynArrayHeapMapInner::Dim3(map) => map.get_mut(key),
+            DynArrayHeapMapInner::Dim4(map) => map.get_mut(key),
+            DynArrayHeapMapInner::Dim5(map) => map.get_mut(key),
+            DynArrayHeapMapInner::Dim6(map) => map.get_mut(key),
+            DynArrayHeapMapInner::Dim7(map) => map.get_mut(key),
+            DynArrayHeapMapInner::Dim8(map) => map.get_mut(key),
+            DynArrayHeapMapInner::Dim9(map) => map.get_mut(key),
+            DynArrayHeapMapInner::Dim10(map) => map.get_mut(key),
+            DynArrayHeapMapInner::Dim11(map) => map.get_mut(key),
+            DynArrayHeapMapInner::Dim12(map) => map.get_mut(key),
+            DynArrayHeapMapInner::Dim13(map) => map.get_mut(key),
+            DynArrayHeapMapInner::Dim14(map) => map.get_mut(key),
+            DynArrayHeapMapInner::Dim15(map) => map.get_mut(key),
+            DynArrayHeapMapInner::Dim16(map) => map.get_mut(key),
+        }
+    }
+
+    /// Returns a reference to the highest key in the map.
+    pub(super) fn peek_highest(&self) -> Option<&[K]> {
+        match &self.0 {
+            DynArrayHeapMapInner::Dim1(map) => map.peek_highest(),
+            DynArrayHeapMapInner::Dim2(map) => map.peek_highest(),
+            DynArrayHeapMapInner::Dim3(map) => map.peek_highest(),
+            DynArrayHeapMapInner::Dim4(map) => map.peek_highest(),
+            DynArrayHeapMapInner::Dim5(map) => map.peek_highest(),
+            DynArrayHeapMapInner::Dim6(map) => map.peek_highest(),
+            DynArrayHeapMapInner::Dim7(map) => map.peek_highest(),
+            DynArrayHeapMapInner::Dim8(map) => map.peek_highest(),
+            DynArrayHeapMapInner::Dim9(map) => map.peek_highest(),
+            DynArrayHeapMapInner::Dim10(map) => map.peek_highest(),
+            DynArrayHeapMapInner::Dim11(map) => map.peek_highest(),
+            DynArrayHeapMapInner::Dim12(map) => map.peek_highest(),
+            DynArrayHeapMapInner::Dim13(map) => map.peek_highest(),
+            DynArrayHeapMapInner::Dim14(map) => map.peek_highest(),
+            DynArrayHeapMapInner::Dim15(map) => map.peek_highest(),
+            DynArrayHeapMapInner::Dim16(map) => map.peek_highest(),
+        }
+    }
+
+    /// Removes the entry with the highest key from the map.
+    pub(super) fn evict_highest(&mut self) {
+        match &mut self.0 {
+            DynArrayHeapMapInner::Dim1(map) => map.evict_highest(),
+            DynArrayHeapMapInner::Dim2(map) => map.evict_highest(),
+            DynArrayHeapMapInner::Dim3(map) => map.evict_highest(),
+            DynArrayHeapMapInner::Dim4(map) => map.evict_highest(),
+            DynArrayHeapMapInner::Dim5(map) => map.evict_highest(),
+            DynArrayHeapMapInner::Dim6(map) => map.evict_highest(),
+            DynArrayHeapMapInner::Dim7(map) => map.evict_highest(),
+            DynArrayHeapMapInner::Dim8(map) => map.evict_highest(),
+            DynArrayHeapMapInner::Dim9(map) => map.evict_highest(),
+            DynArrayHeapMapInner::Dim10(map) => map.evict_highest(),
+            DynArrayHeapMapInner::Dim11(map) => map.evict_highest(),
+            DynArrayHeapMapInner::Dim12(map) => map.evict_highest(),
+            DynArrayHeapMapInner::Dim13(map) => map.evict_highest(),
+            DynArrayHeapMapInner::Dim14(map) => map.evict_highest(),
+            DynArrayHeapMapInner::Dim15(map) => map.evict_highest(),
+            DynArrayHeapMapInner::Dim16(map) => map.evict_highest(),
+        }
+    }
+
+    pub(crate) fn memory_consumption(&self) -> u64 {
+        match &self.0 {
+            DynArrayHeapMapInner::Dim1(map) => map.memory_consumption(),
+            DynArrayHeapMapInner::Dim2(map) => map.memory_consumption(),
+            DynArrayHeapMapInner::Dim3(map) => map.memory_consumption(),
+            DynArrayHeapMapInner::Dim4(map) => map.memory_consumption(),
+            DynArrayHeapMapInner::Dim5(map) => map.memory_consumption(),
+            DynArrayHeapMapInner::Dim6(map) => map.memory_consumption(),
+            DynArrayHeapMapInner::Dim7(map) => map.memory_consumption(),
+            DynArrayHeapMapInner::Dim8(map) => map.memory_consumption(),
+            DynArrayHeapMapInner::Dim9(map) => map.memory_consumption(),
+            DynArrayHeapMapInner::Dim10(map) => map.memory_consumption(),
+            DynArrayHeapMapInner::Dim11(map) => map.memory_consumption(),
+            DynArrayHeapMapInner::Dim12(map) => map.memory_consumption(),
+            DynArrayHeapMapInner::Dim13(map) => map.memory_consumption(),
+            DynArrayHeapMapInner::Dim14(map) => map.memory_consumption(),
+            DynArrayHeapMapInner::Dim15(map) => map.memory_consumption(),
+            DynArrayHeapMapInner::Dim16(map) => map.memory_consumption(),
+        }
+    }
+}
+
+impl<K: Ord + Clone + Copy + 'static, V: 'static> DynArrayHeapMap<K, V> {
+    /// Turns this map into an iterator over key-value pairs.
+    pub fn into_iter(self) -> impl Iterator<Item = (SmallVec<[K; MAX_DYN_ARRAY_SIZE]>, V)> {
+        match self.0 {
+            DynArrayHeapMapInner::Dim1(map) => map.into_iter(),
+            DynArrayHeapMapInner::Dim2(map) => map.into_iter(),
+            DynArrayHeapMapInner::Dim3(map) => map.into_iter(),
+            DynArrayHeapMapInner::Dim4(map) => map.into_iter(),
+            DynArrayHeapMapInner::Dim5(map) => map.into_iter(),
+            DynArrayHeapMapInner::Dim6(map) => map.into_iter(),
+            DynArrayHeapMapInner::Dim7(map) => map.into_iter(),
+            DynArrayHeapMapInner::Dim8(map) => map.into_iter(),
+            DynArrayHeapMapInner::Dim9(map) => map.into_iter(),
+            DynArrayHeapMapInner::Dim10(map) => map.into_iter(),
+            DynArrayHeapMapInner::Dim11(map) => map.into_iter(),
+            DynArrayHeapMapInner::Dim12(map) => map.into_iter(),
+            DynArrayHeapMapInner::Dim13(map) => map.into_iter(),
+            DynArrayHeapMapInner::Dim14(map) => map.into_iter(),
+            DynArrayHeapMapInner::Dim15(map) => map.into_iter(),
+            DynArrayHeapMapInner::Dim16(map) => map.into_iter(),
+        }
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn test_dyn_array_heap_map() {
+        let mut map = DynArrayHeapMap::<u32, &str>::try_new(2).unwrap();
+        // insert
+        let key1 = [1u32, 2u32];
+        let key2 = [2u32, 1u32];
+        map.get_or_insert_with(&key1, || "a");
+        map.get_or_insert_with(&key2, || "b");
+        assert_eq!(map.size(), 2);
+
+        // evict highest
+        assert_eq!(map.peek_highest(), Some(&key2[..]));
+        map.evict_highest();
+        assert_eq!(map.size(), 1);
+        assert_eq!(map.peek_highest(), Some(&key1[..]));
+
+        // into_iter
+        let mut iter = map.into_iter();
+        let (k, v) = iter.next().unwrap();
+        assert_eq!(k.as_slice(), &key1);
+        assert_eq!(v, "c");
+        assert_eq!(iter.next(), None);
+    }
+}

src/aggregation/bucket/composite/numeric_types.rs

@@ -0,0 +1,460 @@
+/// This modules helps comparing numerical values of different types (i64, u64
+/// and f64).
+pub(super) mod num_cmp {
+    use std::cmp::Ordering;
+
+    use crate::TantivyError;
+
+    pub fn cmp_i64_f64(left_i: i64, right_f: f64) -> crate::Result<Ordering> {
+        if right_f.is_nan() {
+            return Err(TantivyError::InvalidArgument(
+                "NaN comparison is not supported".to_string(),
+            ));
+        }
+
+        // If right_f is < i64::MIN then left_i > right_f (i64::MIN=-2^63 can be
+        // exactly represented as f64)
+        if right_f < i64::MIN as f64 {
+            return Ok(Ordering::Greater);
+        }
+        // If right_f is >= i64::MAX then left_i < right_f (i64::MAX=2^63-1 cannot
+        // be exactly represented as f64)
+        if right_f >= i64::MAX as f64 {
+            return Ok(Ordering::Less);
+        }
+
+        // Now right_f is in (i64::MIN, i64::MAX), so `right_f as i64` is
+        // well-defined (truncation toward 0)
+        let right_as_i = right_f as i64;
+
+        let result = match left_i.cmp(&right_as_i) {
+            Ordering::Less => Ordering::Less,
+            Ordering::Greater => Ordering::Greater,
+            Ordering::Equal => {
+                // they have the same integer part, compare the fraction
+                let rem = right_f - (right_as_i as f64);
+                if rem == 0.0 {
+                    Ordering::Equal
+                } else if right_f > 0.0 {
+                    Ordering::Less
+                } else {
+                    Ordering::Greater
+                }
+            }
+        };
+        Ok(result)
+    }
+
+    pub fn cmp_u64_f64(left_u: u64, right_f: f64) -> crate::Result<Ordering> {
+        if right_f.is_nan() {
+            return Err(TantivyError::InvalidArgument(
+                "NaN comparison is not supported".to_string(),
+            ));
+        }
+
+        // Negative floats are always less than any u64 >= 0
+        if right_f < 0.0 {
+            return Ok(Ordering::Greater);
+        }
+
+        // If right_f is >= u64::MAX then left_u < right_f (u64::MAX=2^64-1 cannot be exactly)
+        let max_as_f = u64::MAX as f64;
+        if right_f > max_as_f {
+            return Ok(Ordering::Less);
+        }
+
+        // Now right_f is in (0, u64::MAX), so `right_f as u64` is well-defined
+        // (truncation toward 0)
+        let right_as_u = right_f as u64;
+
+        let result = match left_u.cmp(&right_as_u) {
+            Ordering::Less => Ordering::Less,
+            Ordering::Greater => Ordering::Greater,
+            Ordering::Equal => {
+                // they have the same integer part, compare the fraction
+                let rem = right_f - (right_as_u as f64);
+                if rem == 0.0 {
+                    Ordering::Equal
+                } else {
+                    Ordering::Less
+                }
+            }
+        };
+        Ok(result)
+    }
+
+    pub fn cmp_i64_u64(left_i: i64, right_u: u64) -> Ordering {
+        if left_i < 0 {
+            Ordering::Less
+        } else {
+            let left_as_u = left_i as u64;
+            left_as_u.cmp(&right_u)
+        }
+    }
+}
+
+/// This modules helps projecting numerical values to other numerical types.
+/// When the target value space cannot exactly represent the source value, the
+/// next representable value is returned (or AfterLast if the source value is
+/// larger than the largest representable value).
+///
+/// All functions in this module assume that f64 values are not NaN.
+pub(super) mod num_proj {
+    #[derive(Debug, PartialEq)]
+    pub enum ProjectedNumber<T> {
+        Exact(T),
+        Next(T),
+        AfterLast,
+    }
+
+    pub fn i64_to_u64(value: i64) -> ProjectedNumber<u64> {
+        if value < 0 {
+            ProjectedNumber::Next(0)
+        } else {
+            ProjectedNumber::Exact(value as u64)
+        }
+    }
+
+    pub fn u64_to_i64(value: u64) -> ProjectedNumber<i64> {
+        if value > i64::MAX as u64 {
+            ProjectedNumber::AfterLast
+        } else {
+            ProjectedNumber::Exact(value as i64)
+        }
+    }
+
+    pub fn f64_to_u64(value: f64) -> ProjectedNumber<u64> {
+        if value < 0.0 {
+            ProjectedNumber::Next(0)
+        } else if value > u64::MAX as f64 {
+            ProjectedNumber::AfterLast
+        } else if value.fract() == 0.0 {
+            ProjectedNumber::Exact(value as u64)
+        } else {
+            // casting f64 to u64 truncates toward zero
+            ProjectedNumber::Next(value as u64 + 1)
+        }
+    }
+
+    pub fn f64_to_i64(value: f64) -> ProjectedNumber<i64> {
+        if value < (i64::MIN as f64) {
+            return ProjectedNumber::Next(i64::MIN);
+        } else if value >= (i64::MAX as f64) {
+            return ProjectedNumber::AfterLast;
+        } else if value.fract() == 0.0 {
+            ProjectedNumber::Exact(value as i64)
+        } else if value > 0.0 {
+            // casting f64 to i64 truncates toward zero
+            ProjectedNumber::Next(value as i64 + 1)
+        } else {
+            ProjectedNumber::Next(value as i64)
+        }
+    }
+
+    pub fn i64_to_f64(value: i64) -> ProjectedNumber<f64> {
+        let value_f = value as f64;
+        let k_roundtrip = value_f as i64;
+        if k_roundtrip == value {
+            // between -2^53 and 2^53 all i64 are exactly represented as f64
+            ProjectedNumber::Exact(value_f)
+        } else {
+            // for very large/small i64 values, it is approximated to the closest f64
+            if k_roundtrip > value {
+                ProjectedNumber::Next(value_f)
+            } else {
+                ProjectedNumber::Next(value_f.next_up())
+            }
+        }
+    }
+
+    pub fn u64_to_f64(value: u64) -> ProjectedNumber<f64> {
+        let value_f = value as f64;
+        let k_roundtrip = value_f as u64;
+        if k_roundtrip == value {
+            // between 0 and 2^53 all u64 are exactly represented as f64
+            ProjectedNumber::Exact(value_f)
+        } else if k_roundtrip > value {
+            ProjectedNumber::Next(value_f)
+        } else {
+            ProjectedNumber::Next(value_f.next_up())
+        }
+    }
+}
+
+#[cfg(test)]
+mod num_cmp_tests {
+    use std::cmp::Ordering;
+
+    use super::num_cmp::*;
+
+    #[test]
+    fn test_cmp_u64_f64() {
+        // Basic comparisons
+        assert_eq!(cmp_u64_f64(5, 5.0).unwrap(), Ordering::Equal);
+        assert_eq!(cmp_u64_f64(5, 6.0).unwrap(), Ordering::Less);
+        assert_eq!(cmp_u64_f64(6, 5.0).unwrap(), Ordering::Greater);
+        assert_eq!(cmp_u64_f64(0, 0.0).unwrap(), Ordering::Equal);
+        assert_eq!(cmp_u64_f64(0, 0.1).unwrap(), Ordering::Less);
+
+        // Negative float values should always be less than any u64
+        assert_eq!(cmp_u64_f64(0, -0.1).unwrap(), Ordering::Greater);
+        assert_eq!(cmp_u64_f64(5, -5.0).unwrap(), Ordering::Greater);
+        assert_eq!(cmp_u64_f64(u64::MAX, -1e20).unwrap(), Ordering::Greater);
+
+        // Tests with extreme values
+        assert_eq!(cmp_u64_f64(u64::MAX, 1e20).unwrap(), Ordering::Less);
+
+        // Precision edge cases: large u64 that loses precision when converted to f64
+        // => 2^54, exactly represented as f64
+        let large_f64 = 18_014_398_509_481_984.0;
+        let large_u64 = 18_014_398_509_481_984;
+        // prove that large_u64 is exactly represented as f64
+        assert_eq!(large_u64 as f64, large_f64);
+        assert_eq!(cmp_u64_f64(large_u64, large_f64).unwrap(), Ordering::Equal);
+        // => (2^54 + 1) cannot be exactly represented in f64
+        let large_u64_plus_1 = 18_014_398_509_481_985;
+        // prove that it is represented as f64 by large_f64
+        assert_eq!(large_u64_plus_1 as f64, large_f64);
+        assert_eq!(
+            cmp_u64_f64(large_u64_plus_1, large_f64).unwrap(),
+            Ordering::Greater
+        );
+        // => (2^54 - 1) cannot be exactly represented in f64
+        let large_u64_minus_1 = 18_014_398_509_481_983;
+        // prove that it is also represented as f64 by large_f64
+        assert_eq!(large_u64_minus_1 as f64, large_f64);
+        assert_eq!(
+            cmp_u64_f64(large_u64_minus_1, large_f64).unwrap(),
+            Ordering::Less
+        );
+
+        // NaN comparison results in an error
+        assert!(cmp_u64_f64(0, f64::NAN).is_err());
+    }
+
+    #[test]
+    fn test_cmp_i64_f64() {
+        // Basic comparisons
+        assert_eq!(cmp_i64_f64(5, 5.0).unwrap(), Ordering::Equal);
+        assert_eq!(cmp_i64_f64(5, 6.0).unwrap(), Ordering::Less);
+        assert_eq!(cmp_i64_f64(6, 5.0).unwrap(), Ordering::Greater);
+        assert_eq!(cmp_i64_f64(-5, -5.0).unwrap(), Ordering::Equal);
+        assert_eq!(cmp_i64_f64(-5, -4.0).unwrap(), Ordering::Less);
+        assert_eq!(cmp_i64_f64(-4, -5.0).unwrap(), Ordering::Greater);
+        assert_eq!(cmp_i64_f64(-5, 5.0).unwrap(), Ordering::Less);
+        assert_eq!(cmp_i64_f64(5, -5.0).unwrap(), Ordering::Greater);
+        assert_eq!(cmp_i64_f64(0, -0.1).unwrap(), Ordering::Greater);
+        assert_eq!(cmp_i64_f64(0, 0.1).unwrap(), Ordering::Less);
+        assert_eq!(cmp_i64_f64(-1, -0.5).unwrap(), Ordering::Less);
+        assert_eq!(cmp_i64_f64(-1, 0.0).unwrap(), Ordering::Less);
+        assert_eq!(cmp_i64_f64(0, 0.0).unwrap(), Ordering::Equal);
+
+        // Tests with extreme values
+        assert_eq!(cmp_i64_f64(i64::MAX, 1e20).unwrap(), Ordering::Less);
+        assert_eq!(cmp_i64_f64(i64::MIN, -1e20).unwrap(), Ordering::Greater);
+
+        // Precision edge cases: large i64 that loses precision when converted to f64
+        // => 2^54, exactly represented as f64
+        let large_f64 = 18_014_398_509_481_984.0;
+        let large_i64 = 18_014_398_509_481_984;
+        // prove that large_i64 is exactly represented as f64
+        assert_eq!(large_i64 as f64, large_f64);
+        assert_eq!(cmp_i64_f64(large_i64, large_f64).unwrap(), Ordering::Equal);
+        // => (1_i64 << 54) + 1 cannot be exactly represented in f64
+        let large_i64_plus_1 = 18_014_398_509_481_985;
+        // prove that it is represented as f64 by large_f64
+        assert_eq!(large_i64_plus_1 as f64, large_f64);
+        assert_eq!(
+            cmp_i64_f64(large_i64_plus_1, large_f64).unwrap(),
+            Ordering::Greater
+        );
+        // => (1_i64 << 54) - 1 cannot be exactly represented in f64
+        let large_i64_minus_1 = 18_014_398_509_481_983;
+        // prove that it is also represented as f64 by large_f64
+        assert_eq!(large_i64_minus_1 as f64, large_f64);
+        assert_eq!(
+            cmp_i64_f64(large_i64_minus_1, large_f64).unwrap(),
+            Ordering::Less
+        );
+
+        // Same precision edge case but with negative values
+        // => -2^54, exactly represented as f64
+        let large_neg_f64 = -18_014_398_509_481_984.0;
+        let large_neg_i64 = -18_014_398_509_481_984;
+        // prove that large_neg_i64 is exactly represented as f64
+        assert_eq!(large_neg_i64 as f64, large_neg_f64);
+        assert_eq!(
+            cmp_i64_f64(large_neg_i64, large_neg_f64).unwrap(),
+            Ordering::Equal
+        );
+        // => (-2^54 + 1) cannot be exactly represented in f64
+        let large_neg_i64_plus_1 = -18_014_398_509_481_985;
+        // prove that it is represented as f64 by large_neg_f64
+        assert_eq!(large_neg_i64_plus_1 as f64, large_neg_f64);
+        assert_eq!(
+            cmp_i64_f64(large_neg_i64_plus_1, large_neg_f64).unwrap(),
+            Ordering::Less
+        );
+        // => (-2^54 - 1) cannot be exactly represented in f64
+        let large_neg_i64_minus_1 = -18_014_398_509_481_983;
+        // prove that it is also represented as f64 by large_neg_f64
+        assert_eq!(large_neg_i64_minus_1 as f64, large_neg_f64);
+        assert_eq!(
+            cmp_i64_f64(large_neg_i64_minus_1, large_neg_f64).unwrap(),
+            Ordering::Greater
+        );
+
+        // NaN comparison results in an error
+        assert!(cmp_i64_f64(0, f64::NAN).is_err());
+    }
+
+    #[test]
+    fn test_cmp_i64_u64() {
+        // Test with negative i64 values (should always be less than any u64)
+        assert_eq!(cmp_i64_u64(-1, 0), Ordering::Less);
+        assert_eq!(cmp_i64_u64(i64::MIN, 0), Ordering::Less);
+        assert_eq!(cmp_i64_u64(i64::MIN, u64::MAX), Ordering::Less);
+
+        // Test with positive i64 values
+        assert_eq!(cmp_i64_u64(0, 0), Ordering::Equal);
+        assert_eq!(cmp_i64_u64(1, 0), Ordering::Greater);
+        assert_eq!(cmp_i64_u64(1, 1), Ordering::Equal);
+        assert_eq!(cmp_i64_u64(0, 1), Ordering::Less);
+        assert_eq!(cmp_i64_u64(5, 10), Ordering::Less);
+        assert_eq!(cmp_i64_u64(10, 5), Ordering::Greater);
+
+        // Test with values near i64::MAX and u64 conversion
+        assert_eq!(cmp_i64_u64(i64::MAX, i64::MAX as u64), Ordering::Equal);
+        assert_eq!(cmp_i64_u64(i64::MAX, (i64::MAX as u64) + 1), Ordering::Less);
+        assert_eq!(cmp_i64_u64(i64::MAX, u64::MAX), Ordering::Less);
+    }
+}
+
+#[cfg(test)]
+mod num_proj_tests {
+    use super::num_proj::{self, ProjectedNumber};
+
+    #[test]
+    fn test_i64_to_u64() {
+        assert_eq!(num_proj::i64_to_u64(-1), ProjectedNumber::Next(0));
+        assert_eq!(num_proj::i64_to_u64(i64::MIN), ProjectedNumber::Next(0));
+        assert_eq!(num_proj::i64_to_u64(0), ProjectedNumber::Exact(0));
+        assert_eq!(num_proj::i64_to_u64(42), ProjectedNumber::Exact(42));
+        assert_eq!(
+            num_proj::i64_to_u64(i64::MAX),
+            ProjectedNumber::Exact(i64::MAX as u64)
+        );
+    }
+
+    #[test]
+    fn test_u64_to_i64() {
+        assert_eq!(num_proj::u64_to_i64(0), ProjectedNumber::Exact(0));
+        assert_eq!(num_proj::u64_to_i64(42), ProjectedNumber::Exact(42));
+        assert_eq!(
+            num_proj::u64_to_i64(i64::MAX as u64),
+            ProjectedNumber::Exact(i64::MAX)
+        );
+        assert_eq!(
+            num_proj::u64_to_i64((i64::MAX as u64) + 1),
+            ProjectedNumber::AfterLast
+        );
+        assert_eq!(num_proj::u64_to_i64(u64::MAX), ProjectedNumber::AfterLast);
+    }
+
+    #[test]
+    fn test_f64_to_u64() {
+        assert_eq!(num_proj::f64_to_u64(-1e25), ProjectedNumber::Next(0));
+        assert_eq!(num_proj::f64_to_u64(-0.1), ProjectedNumber::Next(0));
+        assert_eq!(num_proj::f64_to_u64(1e20), ProjectedNumber::AfterLast);
+        assert_eq!(
+            num_proj::f64_to_u64(f64::INFINITY),
+            ProjectedNumber::AfterLast
+        );
+        assert_eq!(num_proj::f64_to_u64(0.0), ProjectedNumber::Exact(0));
+        assert_eq!(num_proj::f64_to_u64(42.0), ProjectedNumber::Exact(42));
+        assert_eq!(num_proj::f64_to_u64(0.5), ProjectedNumber::Next(1));
+        assert_eq!(num_proj::f64_to_u64(42.1), ProjectedNumber::Next(43));
+    }
+
+    #[test]
+    fn test_f64_to_i64() {
+        assert_eq!(num_proj::f64_to_i64(-1e20), ProjectedNumber::Next(i64::MIN));
+        assert_eq!(
+            num_proj::f64_to_i64(f64::NEG_INFINITY),
+            ProjectedNumber::Next(i64::MIN)
+        );
+        assert_eq!(num_proj::f64_to_i64(1e20), ProjectedNumber::AfterLast);
+        assert_eq!(
+            num_proj::f64_to_i64(f64::INFINITY),
+            ProjectedNumber::AfterLast
+        );
+        assert_eq!(num_proj::f64_to_i64(0.0), ProjectedNumber::Exact(0));
+        assert_eq!(num_proj::f64_to_i64(42.0), ProjectedNumber::Exact(42));
+        assert_eq!(num_proj::f64_to_i64(-42.0), ProjectedNumber::Exact(-42));
+        assert_eq!(num_proj::f64_to_i64(0.5), ProjectedNumber::Next(1));
+        assert_eq!(num_proj::f64_to_i64(42.1), ProjectedNumber::Next(43));
+        assert_eq!(num_proj::f64_to_i64(-0.5), ProjectedNumber::Next(0));
+        assert_eq!(num_proj::f64_to_i64(-42.1), ProjectedNumber::Next(-42));
+    }
+
+    #[test]
+    fn test_i64_to_f64() {
+        assert_eq!(num_proj::i64_to_f64(0), ProjectedNumber::Exact(0.0));
+        assert_eq!(num_proj::i64_to_f64(42), ProjectedNumber::Exact(42.0));
+        assert_eq!(num_proj::i64_to_f64(-42), ProjectedNumber::Exact(-42.0));
+
+        let max_exact = 9_007_199_254_740_992; // 2^53
+        assert_eq!(
+            num_proj::i64_to_f64(max_exact),
+            ProjectedNumber::Exact(max_exact as f64)
+        );
+
+        // Test values that cannot be exactly represented as f64 (integers above 2^53)
+        let large_i64 = 9_007_199_254_740_993; // 2^53 + 1
+        let closest_f64 = 9_007_199_254_740_992.0;
+        assert_eq!(large_i64 as f64, closest_f64);
+        if let ProjectedNumber::Next(val) = num_proj::i64_to_f64(large_i64) {
+            // Verify that the returned float is different from the direct cast
+            assert!(val > closest_f64);
+            assert!(val - closest_f64 < 2. * f64::EPSILON * closest_f64);
+        } else {
+            panic!("Expected ProjectedNumber::Next for large_i64");
+        }
+
+        // Test with very large negative value
+        let large_neg_i64 = -9_007_199_254_740_993; // -(2^53 + 1)
+        let closest_neg_f64 = -9_007_199_254_740_992.0;
+        assert_eq!(large_neg_i64 as f64, closest_neg_f64);
+        if let ProjectedNumber::Next(val) = num_proj::i64_to_f64(large_neg_i64) {
+            // Verify that the returned float is the closest representable f64
+            assert_eq!(val, closest_neg_f64);
+        } else {
+            panic!("Expected ProjectedNumber::Next for large_neg_i64");
+        }
+    }
+
+    #[test]
+    fn test_u64_to_f64() {
+        assert_eq!(num_proj::u64_to_f64(0), ProjectedNumber::Exact(0.0));
+        assert_eq!(num_proj::u64_to_f64(42), ProjectedNumber::Exact(42.0));
+
+        // Test the largest u64 value that can be exactly represented as f64 (2^53)
+        let max_exact = 9_007_199_254_740_992; // 2^53
+        assert_eq!(
+            num_proj::u64_to_f64(max_exact),
+            ProjectedNumber::Exact(max_exact as f64)
+        );
+
+        // Test values that cannot be exactly represented as f64 (integers above 2^53)
+        let large_u64 = 9_007_199_254_740_993; // 2^53 + 1
+        let closest_f64 = 9_007_199_254_740_992.0;
+        assert_eq!(large_u64 as f64, closest_f64);
+        if let ProjectedNumber::Next(val) = num_proj::u64_to_f64(large_u64) {
+            // Verify that the returned float is different from the direct cast
+            assert!(val > closest_f64);
+            assert!(val - closest_f64 < 2. * f64::EPSILON * closest_f64);
+        } else {
+            panic!("Expected ProjectedNumber::Next for large_u64");
+        }
+    }
+}

src/aggregation/bucket/filter.rs

@@ -6,10 +6,14 @@ use serde::{Deserialize, Deserializer, Serialize, Serializer};
 use crate::aggregation::agg_data::{
     build_segment_agg_collectors, AggRefNode, AggregationsSegmentCtx,
 };
+use crate::aggregation::cached_sub_aggs::{
+    CachedSubAggs, HighCardSubAggCache, LowCardSubAggCache, SubAggCache,
+};
 use crate::aggregation::intermediate_agg_result::{
     IntermediateAggregationResult, IntermediateAggregationResults, IntermediateBucketResult,
 };
-use crate::aggregation::segment_agg_result::{CollectorClone, SegmentAggregationCollector};
+use crate::aggregation::segment_agg_result::{BucketIdProvider, SegmentAggregationCollector};
+use crate::aggregation::BucketId;
 use crate::docset::DocSet;
 use crate::query::{AllQuery, EnableScoring, Query, QueryParser};
 use crate::schema::Schema;
@@ -32,7 +36,7 @@ use crate::{DocId, SegmentReader, TantivyError};
 ///
 /// # Implementation Requirements
 ///
-/// Implementors must:
+/// Implementers must:
 /// 1. Derive `Debug`, `Clone`, `Serialize`, and `Deserialize`
 /// 2. Use `#[typetag::serde]` attribute on the impl block
 /// 3. Implement `build_query()` to construct the query from schema/tokenizers
@@ -404,15 +408,18 @@ pub struct FilterAggReqData {
     pub evaluator: DocumentQueryEvaluator,
     /// Reusable buffer for matching documents to minimize allocations during collection
     pub matching_docs_buffer: Vec<DocId>,
+    /// True if this filter aggregation is at the top level of the aggregation tree (not nested).
+    pub is_top_level: bool,
 }
 
 impl FilterAggReqData {
     pub(crate) fn get_memory_consumption(&self) -> usize {
         // Estimate: name + segment reader reference + bitset + buffer capacity
         self.name.len()
-            + std::mem::size_of::<SegmentReader>()
-            + self.evaluator.bitset.len() / 8 // BitSet memory (bits to bytes)
-            + self.matching_docs_buffer.capacity() * std::mem::size_of::<DocId>()
+        + std::mem::size_of::<SegmentReader>()
+        + self.evaluator.bitset.len() / 8 // BitSet memory (bits to bytes)
+        + self.matching_docs_buffer.capacity() * std::mem::size_of::<DocId>()
+        + std::mem::size_of::<bool>()
     }
 }
 
@@ -489,17 +496,24 @@ impl Debug for DocumentQueryEvaluator {
     }
 }
 
-/// Segment collector for filter aggregation
-pub struct SegmentFilterCollector {
-    /// Document count in this bucket
+#[derive(Debug, Clone, PartialEq, Copy)]
+struct DocCount {
     doc_count: u64,
+    bucket_id: BucketId,
+}
+
+/// Segment collector for filter aggregation
+pub struct SegmentFilterCollector<C: SubAggCache> {
+    /// Document counts per parent bucket
+    parent_buckets: Vec<DocCount>,
     /// Sub-aggregation collectors
-    sub_aggregations: Option<Box<dyn SegmentAggregationCollector>>,
+    sub_aggregations: Option<CachedSubAggs<C>>,
+    bucket_id_provider: BucketIdProvider,
     /// Accessor index for this filter aggregation (to access FilterAggReqData)
     accessor_idx: usize,
 }
 
-impl SegmentFilterCollector {
+impl<C: SubAggCache> SegmentFilterCollector<C> {
     /// Create a new filter segment collector following the new agg_data pattern
     pub(crate) fn from_req_and_validate(
         req: &mut AggregationsSegmentCtx,
@@ -511,47 +525,75 @@ impl SegmentFilterCollector {
         } else {
             None
         };
+        let sub_agg_collector = sub_agg_collector.map(CachedSubAggs::new);
 
         Ok(SegmentFilterCollector {
-            doc_count: 0,
+            parent_buckets: Vec::new(),
             sub_aggregations: sub_agg_collector,
             accessor_idx: node.idx_in_req_data,
+            bucket_id_provider: BucketIdProvider::default(),
         })
     }
 }
 
-impl Debug for SegmentFilterCollector {
+pub(crate) fn build_segment_filter_collector(
+    req: &mut AggregationsSegmentCtx,
+    node: &AggRefNode,
+) -> crate::Result<Box<dyn SegmentAggregationCollector>> {
+    let is_top_level = req.per_request.filter_req_data[node.idx_in_req_data]
+        .as_ref()
+        .expect("filter_req_data slot is empty")
+        .is_top_level;
+
+    if is_top_level {
+        Ok(Box::new(
+            SegmentFilterCollector::<LowCardSubAggCache>::from_req_and_validate(req, node)?,
+        ))
+    } else {
+        Ok(Box::new(
+            SegmentFilterCollector::<HighCardSubAggCache>::from_req_and_validate(req, node)?,
+        ))
+    }
+}
+
+impl<C: SubAggCache> Debug for SegmentFilterCollector<C> {
     fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
         f.debug_struct("SegmentFilterCollector")
-            .field("doc_count", &self.doc_count)
+            .field("buckets", &self.parent_buckets)
             .field("has_sub_aggs", &self.sub_aggregations.is_some())
             .field("accessor_idx", &self.accessor_idx)
             .finish()
     }
 }
 
-impl CollectorClone for SegmentFilterCollector {
-    fn clone_box(&self) -> Box<dyn SegmentAggregationCollector> {
-        // For now, panic - this needs proper implementation with weight recreation
-        panic!("SegmentFilterCollector cloning not yet implemented - requires weight recreation")
-    }
-}
-
-impl SegmentAggregationCollector for SegmentFilterCollector {
+impl<C: SubAggCache> SegmentAggregationCollector for SegmentFilterCollector<C> {
     fn add_intermediate_aggregation_result(
-        self: Box<Self>,
+        &mut self,
         agg_data: &AggregationsSegmentCtx,
         results: &mut IntermediateAggregationResults,
+        parent_bucket_id: BucketId,
     ) -> crate::Result<()> {
         let mut sub_results = IntermediateAggregationResults::default();
+        let bucket_opt = self.parent_buckets.get(parent_bucket_id as usize);
 
-        if let Some(sub_aggs) = self.sub_aggregations {
-            sub_aggs.add_intermediate_aggregation_result(agg_data, &mut sub_results)?;
+        if let Some(sub_aggs) = &mut self.sub_aggregations {
+            sub_aggs
+                .get_sub_agg_collector()
+                .add_intermediate_aggregation_result(
+                    agg_data,
+                    &mut sub_results,
+                    // Here we create a new bucket ID for sub-aggregations if the bucket doesn't
+                    // exist, so that sub-aggregations can still produce results (e.g., zero doc
+                    // count)
+                    bucket_opt
+                        .map(|bucket| bucket.bucket_id)
+                        .unwrap_or(self.bucket_id_provider.next_bucket_id()),
+                )?;
         }
 
         // Create the filter bucket result
         let filter_bucket_result = IntermediateBucketResult::Filter {
-            doc_count: self.doc_count,
+            doc_count: bucket_opt.map(|b| b.doc_count).unwrap_or(0),
             sub_aggregations: sub_results,
         };
 
@@ -570,32 +612,17 @@ impl SegmentAggregationCollector for SegmentFilterCollector {
         Ok(())
     }
 
-    fn collect(&mut self, doc: DocId, agg_data: &mut AggregationsSegmentCtx) -> crate::Result<()> {
-        // Access the evaluator from FilterAggReqData
-        let req_data = agg_data.get_filter_req_data(self.accessor_idx);
-
-        // O(1) BitSet lookup to check if document matches filter
-        if req_data.evaluator.matches_document(doc) {
-            self.doc_count += 1;
-
-            // If we have sub-aggregations, collect on them for this filtered document
-            if let Some(sub_aggs) = &mut self.sub_aggregations {
-                sub_aggs.collect(doc, agg_data)?;
-            }
-        }
-        Ok(())
-    }
-
-    #[inline]
-    fn collect_block(
+    fn collect(
         &mut self,
-        docs: &[DocId],
+        parent_bucket_id: BucketId,
+        docs: &[crate::DocId],
         agg_data: &mut AggregationsSegmentCtx,
     ) -> crate::Result<()> {
         if docs.is_empty() {
             return Ok(());
         }
 
+        let mut bucket = self.parent_buckets[parent_bucket_id as usize];
         // Take the request data to avoid borrow checker issues with sub-aggregations
         let mut req = agg_data.take_filter_req_data(self.accessor_idx);
 
@@ -604,18 +631,24 @@ impl SegmentAggregationCollector for SegmentFilterCollector {
         req.evaluator
             .filter_batch(docs, &mut req.matching_docs_buffer);
 
-        self.doc_count += req.matching_docs_buffer.len() as u64;
+        bucket.doc_count += req.matching_docs_buffer.len() as u64;
 
         // Batch process sub-aggregations if we have matches
         if !req.matching_docs_buffer.is_empty() {
             if let Some(sub_aggs) = &mut self.sub_aggregations {
-                // Use collect_block for better sub-aggregation performance
-                sub_aggs.collect_block(&req.matching_docs_buffer, agg_data)?;
+                for &doc_id in &req.matching_docs_buffer {
+                    sub_aggs.push(bucket.bucket_id, doc_id);
+                }
             }
         }
 
         // Put the request data back
         agg_data.put_back_filter_req_data(self.accessor_idx, req);
+        if let Some(sub_aggs) = &mut self.sub_aggregations {
+            sub_aggs.check_flush_local(agg_data)?;
+        }
+        // put back bucket
+        self.parent_buckets[parent_bucket_id as usize] = bucket;
 
         Ok(())
     }
@@ -626,6 +659,21 @@ impl SegmentAggregationCollector for SegmentFilterCollector {
         }
         Ok(())
     }
+
+    fn prepare_max_bucket(
+        &mut self,
+        max_bucket: BucketId,
+        _agg_data: &AggregationsSegmentCtx,
+    ) -> crate::Result<()> {
+        while self.parent_buckets.len() <= max_bucket as usize {
+            let bucket_id = self.bucket_id_provider.next_bucket_id();
+            self.parent_buckets.push(DocCount {
+                doc_count: 0,
+                bucket_id,
+            });
+        }
+        Ok(())
+    }
 }
 
 /// Intermediate result for filter aggregation
@@ -639,16 +687,14 @@ pub struct IntermediateFilterBucketResult {
 
 #[cfg(test)]
 mod tests {
-    use std::time::Instant;
-
     use serde_json::{json, Value};
 
     use super::*;
     use crate::aggregation::agg_req::Aggregations;
     use crate::aggregation::agg_result::AggregationResults;
     use crate::aggregation::{AggContextParams, AggregationCollector};
-    use crate::query::{AllQuery, QueryParser, TermQuery};
-    use crate::schema::{IndexRecordOption, Schema, Term, FAST, INDEXED, STORED, TEXT};
+    use crate::query::{AllQuery, TermQuery};
+    use crate::schema::{IndexRecordOption, Schema, Term, FAST, INDEXED, TEXT};
     use crate::{doc, Index, IndexWriter};
 
     // Test helper functions
@@ -729,12 +775,13 @@ mod tests {
 
         let schema = schema_builder.build();
         let index = Index::create_in_ram(schema);
-        let mut writer: IndexWriter = index.writer(50_000_000)?;
+        let mut writer: IndexWriter = index.writer_for_tests()?;
 
         writer.add_document(doc!(
             category => "electronics", brand => "apple",
             price => 999u64, rating => 4.5f64, in_stock => true
         ))?;
+        writer.commit()?;
         writer.add_document(doc!(
             category => "electronics", brand => "samsung",
             price => 799u64, rating => 4.2f64, in_stock => true
@@ -938,7 +985,7 @@ mod tests {
         let index = create_standard_test_index()?;
         let reader = index.reader()?;
         let searcher = reader.searcher();
-
+        assert_eq!(searcher.segment_readers().len(), 2);
         let agg = json!({
             "premium_electronics": {
                 "filter": "category:electronics AND price:[800 TO *]",
@@ -1520,9 +1567,9 @@ mod tests {
         let searcher = reader.searcher();
 
         let agg = json!({
-            "test": {
-                "filter": deserialized,
-                "aggs": { "count": { "value_count": { "field": "brand" } } }
+                "test": {
+                    "filter": deserialized,
+                    "aggs": { "count": { "value_count": { "field": "brand" } } }
             }
         });
 

src/aggregation/bucket/histogram/date_histogram.rs

@@ -207,7 +207,7 @@ fn parse_offset_into_milliseconds(input: &str) -> Result<i64, AggregationError>
     }
 }
 
-fn parse_into_milliseconds(input: &str) -> Result<i64, AggregationError> {
+pub(crate) fn parse_into_milliseconds(input: &str) -> Result<i64, AggregationError> {
     let split_boundary = input
         .as_bytes()
         .iter()

src/aggregation/bucket/histogram/histogram.rs

@@ -1,6 +1,6 @@
 use std::cmp::Ordering;
 
-use columnar::{Column, ColumnBlockAccessor, ColumnType};
+use columnar::{Column, ColumnType};
 use rustc_hash::FxHashMap;
 use serde::{Deserialize, Serialize};
 use tantivy_bitpacker::minmax;
@@ -8,14 +8,14 @@ use tantivy_bitpacker::minmax;
 use crate::aggregation::agg_data::{
     build_segment_agg_collectors, AggRefNode, AggregationsSegmentCtx,
 };
-use crate::aggregation::agg_limits::MemoryConsumption;
 use crate::aggregation::agg_req::Aggregations;
 use crate::aggregation::agg_result::BucketEntry;
+use crate::aggregation::cached_sub_aggs::{CachedSubAggs, HighCardCachedSubAggs};
 use crate::aggregation::intermediate_agg_result::{
     IntermediateAggregationResult, IntermediateAggregationResults, IntermediateBucketResult,
     IntermediateHistogramBucketEntry,
 };
-use crate::aggregation::segment_agg_result::SegmentAggregationCollector;
+use crate::aggregation::segment_agg_result::{BucketIdProvider, SegmentAggregationCollector};
 use crate::aggregation::*;
 use crate::TantivyError;
 
@@ -26,13 +26,8 @@ pub struct HistogramAggReqData {
     pub accessor: Column<u64>,
     /// The field type of the fast field.
     pub field_type: ColumnType,
-    /// The column block accessor to access the fast field values.
-    pub column_block_accessor: ColumnBlockAccessor<u64>,
     /// The name of the aggregation.
     pub name: String,
-    /// The sub aggregation blueprint, used to create sub aggregations for each bucket.
-    /// Will be filled during initialization of the collector.
-    pub sub_aggregation_blueprint: Option<Box<dyn SegmentAggregationCollector>>,
     /// The histogram aggregation request.
     pub req: HistogramAggregation,
     /// True if this is a date_histogram aggregation.
@@ -257,18 +252,24 @@ impl HistogramBounds {
 pub(crate) struct SegmentHistogramBucketEntry {
     pub key: f64,
     pub doc_count: u64,
+    pub bucket_id: BucketId,
 }
 
 impl SegmentHistogramBucketEntry {
     pub(crate) fn into_intermediate_bucket_entry(
         self,
-        sub_aggregation: Option<Box<dyn SegmentAggregationCollector>>,
+        sub_aggregation: &mut Option<HighCardCachedSubAggs>,
         agg_data: &AggregationsSegmentCtx,
     ) -> crate::Result<IntermediateHistogramBucketEntry> {
         let mut sub_aggregation_res = IntermediateAggregationResults::default();
         if let Some(sub_aggregation) = sub_aggregation {
             sub_aggregation
-                .add_intermediate_aggregation_result(agg_data, &mut sub_aggregation_res)?;
+                .get_sub_agg_collector()
+                .add_intermediate_aggregation_result(
+                    agg_data,
+                    &mut sub_aggregation_res,
+                    self.bucket_id,
+                )?;
         }
         Ok(IntermediateHistogramBucketEntry {
             key: self.key,
@@ -278,27 +279,38 @@ impl SegmentHistogramBucketEntry {
     }
 }
 
+#[derive(Clone, Debug, Default)]
+struct HistogramBuckets {
+    pub buckets: FxHashMap<i64, SegmentHistogramBucketEntry>,
+}
+
 /// The collector puts values from the fast field into the correct buckets and does a conversion to
 /// the correct datatype.
-#[derive(Clone, Debug)]
+#[derive(Debug)]
 pub struct SegmentHistogramCollector {
     /// The buckets containing the aggregation data.
-    buckets: FxHashMap<i64, SegmentHistogramBucketEntry>,
-    sub_aggregations: FxHashMap<i64, Box<dyn SegmentAggregationCollector>>,
+    /// One Histogram bucket per parent bucket id.
+    parent_buckets: Vec<HistogramBuckets>,
+    sub_agg: Option<HighCardCachedSubAggs>,
     accessor_idx: usize,
+    bucket_id_provider: BucketIdProvider,
 }
 
 impl SegmentAggregationCollector for SegmentHistogramCollector {
     fn add_intermediate_aggregation_result(
-        self: Box<Self>,
+        &mut self,
         agg_data: &AggregationsSegmentCtx,
         results: &mut IntermediateAggregationResults,
+        parent_bucket_id: BucketId,
     ) -> crate::Result<()> {
         let name = agg_data
             .get_histogram_req_data(self.accessor_idx)
             .name
             .clone();
-        let bucket = self.into_intermediate_bucket_result(agg_data)?;
+        // TODO: avoid prepare_max_bucket here and handle empty buckets.
+        self.prepare_max_bucket(parent_bucket_id, agg_data)?;
+        let histogram = std::mem::take(&mut self.parent_buckets[parent_bucket_id as usize]);
+        let bucket = self.add_intermediate_bucket_result(agg_data, histogram)?;
         results.push(name, IntermediateAggregationResult::Bucket(bucket))?;
 
         Ok(())
@@ -307,44 +319,40 @@ impl SegmentAggregationCollector for SegmentHistogramCollector {
     #[inline]
     fn collect(
         &mut self,
-        doc: crate::DocId,
-        agg_data: &mut AggregationsSegmentCtx,
-    ) -> crate::Result<()> {
-        self.collect_block(&[doc], agg_data)
-    }
-
-    #[inline]
-    fn collect_block(
-        &mut self,
+        parent_bucket_id: BucketId,
         docs: &[crate::DocId],
         agg_data: &mut AggregationsSegmentCtx,
     ) -> crate::Result<()> {
-        let mut req = agg_data.take_histogram_req_data(self.accessor_idx);
+        let req = agg_data.take_histogram_req_data(self.accessor_idx);
         let mem_pre = self.get_memory_consumption();
+        let buckets = &mut self.parent_buckets[parent_bucket_id as usize].buckets;
 
         let bounds = req.bounds;
         let interval = req.req.interval;
         let offset = req.offset;
         let get_bucket_pos = |val| get_bucket_pos_f64(val, interval, offset) as i64;
 
-        req.column_block_accessor.fetch_block(docs, &req.accessor);
-        for (doc, val) in req
+        agg_data
+            .column_block_accessor
+            .fetch_block(docs, &req.accessor);
+        for (doc, val) in agg_data
             .column_block_accessor
             .iter_docid_vals(docs, &req.accessor)
         {
-            let val = f64_from_fastfield_u64(val, &req.field_type);
+            let val = f64_from_fastfield_u64(val, req.field_type);
             let bucket_pos = get_bucket_pos(val);
             if bounds.contains(val) {
-                let bucket = self.buckets.entry(bucket_pos).or_insert_with(|| {
+                let bucket = buckets.entry(bucket_pos).or_insert_with(|| {
                     let key = get_bucket_key_from_pos(bucket_pos as f64, interval, offset);
-                    SegmentHistogramBucketEntry { key, doc_count: 0 }
+                    SegmentHistogramBucketEntry {
+                        key,
+                        doc_count: 0,
+                        bucket_id: self.bucket_id_provider.next_bucket_id(),
+                    }
                 });
                 bucket.doc_count += 1;
-                if let Some(sub_aggregation_blueprint) = req.sub_aggregation_blueprint.as_ref() {
-                    self.sub_aggregations
-                        .entry(bucket_pos)
-                        .or_insert_with(|| sub_aggregation_blueprint.clone())
-                        .collect(doc, agg_data)?;
+                if let Some(sub_agg) = &mut self.sub_agg {
+                    sub_agg.push(bucket.bucket_id, doc);
                 }
             }
         }
@@ -358,14 +366,30 @@ impl SegmentAggregationCollector for SegmentHistogramCollector {
                 .add_memory_consumed(mem_delta as u64)?;
         }
 
+        if let Some(sub_agg) = &mut self.sub_agg {
+            sub_agg.check_flush_local(agg_data)?;
+        }
+
         Ok(())
     }
 
     fn flush(&mut self, agg_data: &mut AggregationsSegmentCtx) -> crate::Result<()> {
-        for sub_aggregation in self.sub_aggregations.values_mut() {
+        if let Some(sub_aggregation) = &mut self.sub_agg {
             sub_aggregation.flush(agg_data)?;
         }
+        Ok(())
+    }
 
+    fn prepare_max_bucket(
+        &mut self,
+        max_bucket: BucketId,
+        _agg_data: &AggregationsSegmentCtx,
+    ) -> crate::Result<()> {
+        while self.parent_buckets.len() <= max_bucket as usize {
+            self.parent_buckets.push(HistogramBuckets {
+                buckets: FxHashMap::default(),
+            });
+        }
         Ok(())
     }
 }
@@ -373,22 +397,19 @@ impl SegmentAggregationCollector for SegmentHistogramCollector {
 impl SegmentHistogramCollector {
     fn get_memory_consumption(&self) -> usize {
         let self_mem = std::mem::size_of::<Self>();
-        let sub_aggs_mem = self.sub_aggregations.memory_consumption();
-        let buckets_mem = self.buckets.memory_consumption();
-        self_mem + sub_aggs_mem + buckets_mem
+        let buckets_mem = self.parent_buckets.len() * std::mem::size_of::<HistogramBuckets>();
+        self_mem + buckets_mem
     }
     /// Converts the collector result into a intermediate bucket result.
-    pub fn into_intermediate_bucket_result(
-        self,
+    fn add_intermediate_bucket_result(
+        &mut self,
         agg_data: &AggregationsSegmentCtx,
+        histogram: HistogramBuckets,
     ) -> crate::Result<IntermediateBucketResult> {
-        let mut buckets = Vec::with_capacity(self.buckets.len());
+        let mut buckets = Vec::with_capacity(histogram.buckets.len());
 
-        for (bucket_pos, bucket) in self.buckets {
-            let bucket_res = bucket.into_intermediate_bucket_entry(
-                self.sub_aggregations.get(&bucket_pos).cloned(),
-                agg_data,
-            );
+        for bucket in histogram.buckets.into_values() {
+            let bucket_res = bucket.into_intermediate_bucket_entry(&mut self.sub_agg, agg_data);
 
             buckets.push(bucket_res?);
         }
@@ -408,7 +429,7 @@ impl SegmentHistogramCollector {
         agg_data: &mut AggregationsSegmentCtx,
         node: &AggRefNode,
     ) -> crate::Result<Self> {
-        let blueprint = if !node.children.is_empty() {
+        let sub_agg = if !node.children.is_empty() {
             Some(build_segment_agg_collectors(agg_data, &node.children)?)
         } else {
             None
@@ -423,13 +444,13 @@ impl SegmentHistogramCollector {
             max: f64::MAX,
         });
         req_data.offset = req_data.req.offset.unwrap_or(0.0);
-
-        req_data.sub_aggregation_blueprint = blueprint;
+        let sub_agg = sub_agg.map(CachedSubAggs::new);
 
         Ok(Self {
-            buckets: Default::default(),
-            sub_aggregations: Default::default(),
+            parent_buckets: Default::default(),
+            sub_agg,
             accessor_idx: node.idx_in_req_data,
+            bucket_id_provider: BucketIdProvider::default(),
         })
     }
 }

src/aggregation/bucket/mod.rs

@@ -22,6 +22,7 @@
 //! - [Range](RangeAggregation)
 //! - [Terms](TermsAggregation)
 
+mod composite;
 mod filter;
 mod histogram;
 mod range;
@@ -31,6 +32,7 @@ mod term_missing_agg;
 use std::collections::HashMap;
 use std::fmt;
 
+pub use composite::*;
 pub use filter::*;
 pub use histogram::*;
 pub use range::*;

src/aggregation/bucket/range.rs

@@ -1,18 +1,22 @@
 use std::fmt::Debug;
 use std::ops::Range;
 
-use columnar::{Column, ColumnBlockAccessor, ColumnType};
+use columnar::{Column, ColumnType};
 use rustc_hash::FxHashMap;
 use serde::{Deserialize, Serialize};
 
 use crate::aggregation::agg_data::{
     build_segment_agg_collectors, AggRefNode, AggregationsSegmentCtx,
 };
+use crate::aggregation::agg_limits::AggregationLimitsGuard;
+use crate::aggregation::cached_sub_aggs::{
+    CachedSubAggs, HighCardSubAggCache, LowCardCachedSubAggs, LowCardSubAggCache, SubAggCache,
+};
 use crate::aggregation::intermediate_agg_result::{
     IntermediateAggregationResult, IntermediateAggregationResults, IntermediateBucketResult,
     IntermediateRangeBucketEntry, IntermediateRangeBucketResult,
 };
-use crate::aggregation::segment_agg_result::SegmentAggregationCollector;
+use crate::aggregation::segment_agg_result::{BucketIdProvider, SegmentAggregationCollector};
 use crate::aggregation::*;
 use crate::TantivyError;
 
@@ -23,12 +27,12 @@ pub struct RangeAggReqData {
     pub accessor: Column<u64>,
     /// The type of the fast field.
     pub field_type: ColumnType,
-    /// The column block accessor to access the fast field values.
-    pub column_block_accessor: ColumnBlockAccessor<u64>,
     /// The range aggregation request.
     pub req: RangeAggregation,
     /// The name of the aggregation.
     pub name: String,
+    /// Whether this is a top-level aggregation.
+    pub is_top_level: bool,
 }
 
 impl RangeAggReqData {
@@ -151,19 +155,47 @@ pub(crate) struct SegmentRangeAndBucketEntry {
 
 /// The collector puts values from the fast field into the correct buckets and does a conversion to
 /// the correct datatype.
-#[derive(Clone, Debug)]
-pub struct SegmentRangeCollector {
+pub struct SegmentRangeCollector<C: SubAggCache> {
     /// The buckets containing the aggregation data.
-    buckets: Vec<SegmentRangeAndBucketEntry>,
+    /// One for each ParentBucketId
+    parent_buckets: Vec<Vec<SegmentRangeAndBucketEntry>>,
     column_type: ColumnType,
     pub(crate) accessor_idx: usize,
+    sub_agg: Option<CachedSubAggs<C>>,
+    /// Here things get a bit weird. We need to assign unique bucket ids across all
+    /// parent buckets. So we keep track of the next available bucket id here.
+    /// This allows a kind of flattening of the bucket ids across all parent buckets.
+    /// E.g. in nested aggregations:
+    /// Term Agg -> Range aggregation -> Stats aggregation
+    /// E.g. the Term Agg creates 3 buckets ["INFO", "ERROR", "WARN"], each of these has a Range
+    /// aggregation with 4 buckets. The Range aggregation will create buckets with ids:
+    /// - INFO: 0,1,2,3
+    /// - ERROR: 4,5,6,7
+    /// - WARN: 8,9,10,11
+    ///
+    /// This allows the Stats aggregation to have unique bucket ids to refer to.
+    bucket_id_provider: BucketIdProvider,
+    limits: AggregationLimitsGuard,
 }
 
+impl<C: SubAggCache> Debug for SegmentRangeCollector<C> {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        f.debug_struct("SegmentRangeCollector")
+            .field("parent_buckets_len", &self.parent_buckets.len())
+            .field("column_type", &self.column_type)
+            .field("accessor_idx", &self.accessor_idx)
+            .field("has_sub_agg", &self.sub_agg.is_some())
+            .finish()
+    }
+}
+
+/// TODO: Bad naming, there's also SegmentRangeAndBucketEntry
 #[derive(Clone)]
 pub(crate) struct SegmentRangeBucketEntry {
     pub key: Key,
     pub doc_count: u64,
-    pub sub_aggregation: Option<Box<dyn SegmentAggregationCollector>>,
+    // pub sub_aggregation: Option<Box<dyn SegmentAggregationCollector>>,
+    pub bucket_id: BucketId,
     /// The from range of the bucket. Equals `f64::MIN` when `None`.
     pub from: Option<f64>,
     /// The to range of the bucket. Equals `f64::MAX` when `None`. Open interval, `to` is not
@@ -184,48 +216,50 @@ impl Debug for SegmentRangeBucketEntry {
 impl SegmentRangeBucketEntry {
     pub(crate) fn into_intermediate_bucket_entry(
         self,
-        agg_data: &AggregationsSegmentCtx,
     ) -> crate::Result<IntermediateRangeBucketEntry> {
-        let mut sub_aggregation_res = IntermediateAggregationResults::default();
-        if let Some(sub_aggregation) = self.sub_aggregation {
-            sub_aggregation
-                .add_intermediate_aggregation_result(agg_data, &mut sub_aggregation_res)?
-        } else {
-            Default::default()
-        };
+        let sub_aggregation = IntermediateAggregationResults::default();
 
         Ok(IntermediateRangeBucketEntry {
             key: self.key.into(),
             doc_count: self.doc_count,
-            sub_aggregation: sub_aggregation_res,
+            sub_aggregation_res: sub_aggregation,
             from: self.from,
             to: self.to,
         })
     }
 }
 
-impl SegmentAggregationCollector for SegmentRangeCollector {
+impl<C: SubAggCache> SegmentAggregationCollector for SegmentRangeCollector<C> {
     fn add_intermediate_aggregation_result(
-        self: Box<Self>,
+        &mut self,
         agg_data: &AggregationsSegmentCtx,
         results: &mut IntermediateAggregationResults,
+        parent_bucket_id: BucketId,
     ) -> crate::Result<()> {
+        self.prepare_max_bucket(parent_bucket_id, agg_data)?;
         let field_type = self.column_type;
         let name = agg_data
             .get_range_req_data(self.accessor_idx)
             .name
             .to_string();
 
-        let buckets: FxHashMap<SerializedKey, IntermediateRangeBucketEntry> = self
-            .buckets
+        let buckets = std::mem::take(&mut self.parent_buckets[parent_bucket_id as usize]);
+
+        let buckets: FxHashMap<SerializedKey, IntermediateRangeBucketEntry> = buckets
             .into_iter()
-            .map(move |range_bucket| {
-                Ok((
-                    range_to_string(&range_bucket.range, &field_type)?,
-                    range_bucket
-                        .bucket
-                        .into_intermediate_bucket_entry(agg_data)?,
-                ))
+            .map(|range_bucket| {
+                let bucket_id = range_bucket.bucket.bucket_id;
+                let mut agg = range_bucket.bucket.into_intermediate_bucket_entry()?;
+                if let Some(sub_aggregation) = &mut self.sub_agg {
+                    sub_aggregation
+                        .get_sub_agg_collector()
+                        .add_intermediate_aggregation_result(
+                            agg_data,
+                            &mut agg.sub_aggregation_res,
+                            bucket_id,
+                        )?;
+                }
+                Ok((range_to_string(&range_bucket.range, &field_type)?, agg))
             })
             .collect::<crate::Result<_>>()?;
 
@@ -242,73 +276,114 @@ impl SegmentAggregationCollector for SegmentRangeCollector {
     #[inline]
     fn collect(
         &mut self,
-        doc: crate::DocId,
-        agg_data: &mut AggregationsSegmentCtx,
-    ) -> crate::Result<()> {
-        self.collect_block(&[doc], agg_data)
-    }
-
-    #[inline]
-    fn collect_block(
-        &mut self,
+        parent_bucket_id: BucketId,
         docs: &[crate::DocId],
         agg_data: &mut AggregationsSegmentCtx,
     ) -> crate::Result<()> {
-        // Take request data to avoid borrow conflicts during sub-aggregation
-        let mut req = agg_data.take_range_req_data(self.accessor_idx);
+        let req = agg_data.take_range_req_data(self.accessor_idx);
 
-        req.column_block_accessor.fetch_block(docs, &req.accessor);
+        agg_data
+            .column_block_accessor
+            .fetch_block(docs, &req.accessor);
 
-        for (doc, val) in req
+        let buckets = &mut self.parent_buckets[parent_bucket_id as usize];
+
+        for (doc, val) in agg_data
             .column_block_accessor
             .iter_docid_vals(docs, &req.accessor)
         {
-            let bucket_pos = self.get_bucket_pos(val);
-            let bucket = &mut self.buckets[bucket_pos];
+            let bucket_pos = get_bucket_pos(val, buckets);
+            let bucket = &mut buckets[bucket_pos];
             bucket.bucket.doc_count += 1;
-            if let Some(sub_agg) = bucket.bucket.sub_aggregation.as_mut() {
-                sub_agg.collect(doc, agg_data)?;
+            if let Some(sub_agg) = self.sub_agg.as_mut() {
+                sub_agg.push(bucket.bucket.bucket_id, doc);
             }
         }
 
         agg_data.put_back_range_req_data(self.accessor_idx, req);
+        if let Some(sub_agg) = self.sub_agg.as_mut() {
+            sub_agg.check_flush_local(agg_data)?;
+        }
 
         Ok(())
     }
 
     fn flush(&mut self, agg_data: &mut AggregationsSegmentCtx) -> crate::Result<()> {
-        for bucket in self.buckets.iter_mut() {
-            if let Some(sub_agg) = bucket.bucket.sub_aggregation.as_mut() {
-                sub_agg.flush(agg_data)?;
-            }
+        if let Some(sub_agg) = self.sub_agg.as_mut() {
+            sub_agg.flush(agg_data)?;
         }
         Ok(())
     }
+
+    fn prepare_max_bucket(
+        &mut self,
+        max_bucket: BucketId,
+        agg_data: &AggregationsSegmentCtx,
+    ) -> crate::Result<()> {
+        while self.parent_buckets.len() <= max_bucket as usize {
+            let new_buckets = self.create_new_buckets(agg_data)?;
+            self.parent_buckets.push(new_buckets);
+        }
+
+        Ok(())
+    }
+}
+/// Build a concrete `SegmentRangeCollector` with either a Vec- or HashMap-backed
+/// bucket storage, depending on the column type and aggregation level.
+pub(crate) fn build_segment_range_collector(
+    agg_data: &mut AggregationsSegmentCtx,
+    node: &AggRefNode,
+) -> crate::Result<Box<dyn SegmentAggregationCollector>> {
+    let accessor_idx = node.idx_in_req_data;
+    let req_data = agg_data.get_range_req_data(node.idx_in_req_data);
+    let field_type = req_data.field_type;
+
+    // TODO: A better metric instead of is_top_level would be the number of buckets expected.
+    // E.g. If range agg is not top level, but the parent is a bucket agg with less than 10 buckets,
+    // we can are still in low cardinality territory.
+    let is_low_card = req_data.is_top_level && req_data.req.ranges.len() <= 64;
+
+    let sub_agg = if !node.children.is_empty() {
+        Some(build_segment_agg_collectors(agg_data, &node.children)?)
+    } else {
+        None
+    };
+
+    if is_low_card {
+        Ok(Box::new(SegmentRangeCollector::<LowCardSubAggCache> {
+            sub_agg: sub_agg.map(LowCardCachedSubAggs::new),
+            column_type: field_type,
+            accessor_idx,
+            parent_buckets: Vec::new(),
+            bucket_id_provider: BucketIdProvider::default(),
+            limits: agg_data.context.limits.clone(),
+        }))
+    } else {
+        Ok(Box::new(SegmentRangeCollector::<HighCardSubAggCache> {
+            sub_agg: sub_agg.map(CachedSubAggs::new),
+            column_type: field_type,
+            accessor_idx,
+            parent_buckets: Vec::new(),
+            bucket_id_provider: BucketIdProvider::default(),
+            limits: agg_data.context.limits.clone(),
+        }))
+    }
 }
 
-impl SegmentRangeCollector {
-    pub(crate) fn from_req_and_validate(
-        req_data: &mut AggregationsSegmentCtx,
-        node: &AggRefNode,
-    ) -> crate::Result<Self> {
-        let accessor_idx = node.idx_in_req_data;
-        let (field_type, ranges) = {
-            let req_view = req_data.get_range_req_data(node.idx_in_req_data);
-            (req_view.field_type, req_view.req.ranges.clone())
-        };
-
+impl<C: SubAggCache> SegmentRangeCollector<C> {
+    pub(crate) fn create_new_buckets(
+        &mut self,
+        agg_data: &AggregationsSegmentCtx,
+    ) -> crate::Result<Vec<SegmentRangeAndBucketEntry>> {
+        let field_type = self.column_type;
+        let req_data = agg_data.get_range_req_data(self.accessor_idx);
         // The range input on the request is f64.
         // We need to convert to u64 ranges, because we read the values as u64.
         // The mapping from the conversion is monotonic so ordering is preserved.
-        let sub_agg_prototype = if !node.children.is_empty() {
-            Some(build_segment_agg_collectors(req_data, &node.children)?)
-        } else {
-            None
-        };
-
-        let buckets: Vec<_> = extend_validate_ranges(&ranges, &field_type)?
+        let buckets: Vec<_> = extend_validate_ranges(&req_data.req.ranges, &field_type)?
             .iter()
             .map(|range| {
+                let bucket_id = self.bucket_id_provider.next_bucket_id();
                 let key = range
                     .key
                     .clone()
@@ -317,20 +392,20 @@ impl SegmentRangeCollector {
                 let to = if range.range.end == u64::MAX {
                     None
                 } else {
-                    Some(f64_from_fastfield_u64(range.range.end, &field_type))
+                    Some(f64_from_fastfield_u64(range.range.end, field_type))
                 };
                 let from = if range.range.start == u64::MIN {
                     None
                 } else {
-                    Some(f64_from_fastfield_u64(range.range.start, &field_type))
+                    Some(f64_from_fastfield_u64(range.range.start, field_type))
                 };
-                let sub_aggregation = sub_agg_prototype.clone();
+                // let sub_aggregation = sub_agg_prototype.clone();
 
                 Ok(SegmentRangeAndBucketEntry {
                     range: range.range.clone(),
                     bucket: SegmentRangeBucketEntry {
                         doc_count: 0,
-                        sub_aggregation,
+                        bucket_id,
                         key,
                         from,
                         to,
@@ -339,27 +414,20 @@ impl SegmentRangeCollector {
             })
             .collect::<crate::Result<_>>()?;
 
-        req_data.context.limits.add_memory_consumed(
+        self.limits.add_memory_consumed(
             buckets.len() as u64 * std::mem::size_of::<SegmentRangeAndBucketEntry>() as u64,
         )?;
-
-        Ok(SegmentRangeCollector {
-            buckets,
-            column_type: field_type,
-            accessor_idx,
-        })
-    }
-
-    #[inline]
-    fn get_bucket_pos(&self, val: u64) -> usize {
-        let pos = self
-            .buckets
-            .binary_search_by_key(&val, |probe| probe.range.start)
-            .unwrap_or_else(|pos| pos - 1);
-        debug_assert!(self.buckets[pos].range.contains(&val));
-        pos
+        Ok(buckets)
     }
 }
+#[inline]
+fn get_bucket_pos(val: u64, buckets: &[SegmentRangeAndBucketEntry]) -> usize {
+    let pos = buckets
+        .binary_search_by_key(&val, |probe| probe.range.start)
+        .unwrap_or_else(|pos| pos - 1);
+    debug_assert!(buckets[pos].range.contains(&val));
+    pos
+}
 
 /// Converts the user provided f64 range value to fast field value space.
 ///
@@ -456,7 +524,7 @@ pub(crate) fn range_to_string(
             let val = i64::from_u64(val);
             format_date(val)
         } else {
-            Ok(f64_from_fastfield_u64(val, field_type).to_string())
+            Ok(f64_from_fastfield_u64(val, *field_type).to_string())
         }
     };
 
@@ -486,7 +554,7 @@ mod tests {
     pub fn get_collector_from_ranges(
         ranges: Vec<RangeAggregationRange>,
         field_type: ColumnType,
-    ) -> SegmentRangeCollector {
+    ) -> SegmentRangeCollector<HighCardSubAggCache> {
         let req = RangeAggregation {
             field: "dummy".to_string(),
             ranges,
@@ -506,30 +574,33 @@ mod tests {
                 let to = if range.range.end == u64::MAX {
                     None
                 } else {
-                    Some(f64_from_fastfield_u64(range.range.end, &field_type))
+                    Some(f64_from_fastfield_u64(range.range.end, field_type))
                 };
                 let from = if range.range.start == u64::MIN {
                     None
                 } else {
-                    Some(f64_from_fastfield_u64(range.range.start, &field_type))
+                    Some(f64_from_fastfield_u64(range.range.start, field_type))
                 };
                 SegmentRangeAndBucketEntry {
                     range: range.range.clone(),
                     bucket: SegmentRangeBucketEntry {
                         doc_count: 0,
-                        sub_aggregation: None,
                         key,
                         from,
                         to,
+                        bucket_id: 0,
                     },
                 }
             })
             .collect();
 
         SegmentRangeCollector {
-            buckets,
+            parent_buckets: vec![buckets],
             column_type: field_type,
             accessor_idx: 0,
+            sub_agg: None,
+            bucket_id_provider: Default::default(),
+            limits: AggregationLimitsGuard::default(),
         }
     }
 
@@ -776,7 +847,7 @@ mod tests {
         let buckets = vec![(10f64..20f64).into(), (30f64..40f64).into()];
         let collector = get_collector_from_ranges(buckets, ColumnType::F64);
 
-        let buckets = collector.buckets;
+        let buckets = collector.parent_buckets[0].clone();
         assert_eq!(buckets[0].range.start, u64::MIN);
         assert_eq!(buckets[0].range.end, 10f64.to_u64());
         assert_eq!(buckets[1].range.start, 10f64.to_u64());
@@ -799,7 +870,7 @@ mod tests {
         ];
         let collector = get_collector_from_ranges(buckets, ColumnType::F64);
 
-        let buckets = collector.buckets;
+        let buckets = collector.parent_buckets[0].clone();
         assert_eq!(buckets[0].range.start, u64::MIN);
         assert_eq!(buckets[0].range.end, 10f64.to_u64());
         assert_eq!(buckets[1].range.start, 10f64.to_u64());
@@ -814,7 +885,7 @@ mod tests {
         let buckets = vec![(-10f64..-1f64).into()];
         let collector = get_collector_from_ranges(buckets, ColumnType::F64);
 
-        let buckets = collector.buckets;
+        let buckets = collector.parent_buckets[0].clone();
         assert_eq!(&buckets[0].bucket.key.to_string(), "*--10");
         assert_eq!(&buckets[buckets.len() - 1].bucket.key.to_string(), "-1-*");
     }
@@ -823,7 +894,7 @@ mod tests {
         let buckets = vec![(0f64..10f64).into()];
         let collector = get_collector_from_ranges(buckets, ColumnType::F64);
 
-        let buckets = collector.buckets;
+        let buckets = collector.parent_buckets[0].clone();
         assert_eq!(&buckets[0].bucket.key.to_string(), "*-0");
         assert_eq!(&buckets[buckets.len() - 1].bucket.key.to_string(), "10-*");
     }
@@ -832,7 +903,7 @@ mod tests {
     fn range_binary_search_test_u64() {
         let check_ranges = |ranges: Vec<RangeAggregationRange>| {
             let collector = get_collector_from_ranges(ranges, ColumnType::U64);
-            let search = |val: u64| collector.get_bucket_pos(val);
+            let search = |val: u64| get_bucket_pos(val, &collector.parent_buckets[0]);
 
             assert_eq!(search(u64::MIN), 0);
             assert_eq!(search(9), 0);
@@ -878,7 +949,7 @@ mod tests {
         let ranges = vec![(10.0..100.0).into()];
 
         let collector = get_collector_from_ranges(ranges, ColumnType::F64);
-        let search = |val: u64| collector.get_bucket_pos(val);
+        let search = |val: u64| get_bucket_pos(val, &collector.parent_buckets[0]);
 
         assert_eq!(search(u64::MIN), 0);
         assert_eq!(search(9f64.to_u64()), 0);
@@ -890,63 +961,3 @@ mod tests {
                                              // the max value
     }
 }
-
-#[cfg(all(test, feature = "unstable"))]
-mod bench {
-
-    use itertools::Itertools;
-    use rand::seq::SliceRandom;
-    use rand::thread_rng;
-
-    use super::*;
-    use crate::aggregation::bucket::range::tests::get_collector_from_ranges;
-
-    const TOTAL_DOCS: u64 = 1_000_000u64;
-    const NUM_DOCS: u64 = 50_000u64;
-
-    fn get_collector_with_buckets(num_buckets: u64, num_docs: u64) -> SegmentRangeCollector {
-        let bucket_size = num_docs / num_buckets;
-        let mut buckets: Vec<RangeAggregationRange> = vec![];
-        for i in 0..num_buckets {
-            let bucket_start = (i * bucket_size) as f64;
-            buckets.push((bucket_start..bucket_start + bucket_size as f64).into())
-        }
-
-        get_collector_from_ranges(buckets, ColumnType::U64)
-    }
-
-    fn get_rand_docs(total_docs: u64, num_docs_returned: u64) -> Vec<u64> {
-        let mut rng = thread_rng();
-
-        let all_docs = (0..total_docs - 1).collect_vec();
-        let mut vals = all_docs
-            .as_slice()
-            .choose_multiple(&mut rng, num_docs_returned as usize)
-            .cloned()
-            .collect_vec();
-        vals.sort();
-        vals
-    }
-
-    fn bench_range_binary_search(b: &mut test::Bencher, num_buckets: u64) {
-        let collector = get_collector_with_buckets(num_buckets, TOTAL_DOCS);
-        let vals = get_rand_docs(TOTAL_DOCS, NUM_DOCS);
-        b.iter(|| {
-            let mut bucket_pos = 0;
-            for val in &vals {
-                bucket_pos = collector.get_bucket_pos(*val);
-            }
-            bucket_pos
-        })
-    }
-
-    #[bench]
-    fn bench_range_100_buckets(b: &mut test::Bencher) {
-        bench_range_binary_search(b, 100)
-    }
-
-    #[bench]
-    fn bench_range_10_buckets(b: &mut test::Bencher) {
-        bench_range_binary_search(b, 10)
-    }
-}

src/aggregation/bucket/term_agg/default_impl.rs

@@ -1,196 +0,0 @@
-use std::fmt::Debug;
-
-use columnar::ColumnType;
-use rustc_hash::FxHashMap;
-
-use super::OrderTarget;
-use crate::aggregation::agg_data::{
-    build_segment_agg_collectors, AggRefNode, AggregationsSegmentCtx,
-};
-use crate::aggregation::agg_limits::MemoryConsumption;
-use crate::aggregation::bucket::get_agg_name_and_property;
-use crate::aggregation::intermediate_agg_result::{
-    IntermediateAggregationResult, IntermediateAggregationResults,
-};
-use crate::aggregation::segment_agg_result::SegmentAggregationCollector;
-use crate::TantivyError;
-
-#[derive(Clone, Debug, Default)]
-/// Container to store term_ids/or u64 values and their buckets.
-struct TermBuckets {
-    pub(crate) entries: FxHashMap<u64, u32>,
-    pub(crate) sub_aggs: FxHashMap<u64, Box<dyn SegmentAggregationCollector>>,
-}
-
-impl TermBuckets {
-    fn get_memory_consumption(&self) -> usize {
-        let sub_aggs_mem = self.sub_aggs.memory_consumption();
-        let buckets_mem = self.entries.memory_consumption();
-        sub_aggs_mem + buckets_mem
-    }
-
-    fn force_flush(&mut self, agg_data: &mut AggregationsSegmentCtx) -> crate::Result<()> {
-        for sub_aggregations in &mut self.sub_aggs.values_mut() {
-            sub_aggregations.as_mut().flush(agg_data)?;
-        }
-        Ok(())
-    }
-}
-
-/// The collector puts values from the fast field into the correct buckets and does a conversion to
-/// the correct datatype.
-#[derive(Clone, Debug)]
-pub struct SegmentTermCollector {
-    /// The buckets containing the aggregation data.
-    term_buckets: TermBuckets,
-    accessor_idx: usize,
-}
-
-impl SegmentAggregationCollector for SegmentTermCollector {
-    fn add_intermediate_aggregation_result(
-        self: Box<Self>,
-        agg_data: &AggregationsSegmentCtx,
-        results: &mut IntermediateAggregationResults,
-    ) -> crate::Result<()> {
-        let name = agg_data.get_term_req_data(self.accessor_idx).name.clone();
-
-        let entries: Vec<(u64, u32)> = self.term_buckets.entries.into_iter().collect();
-        let bucket = super::into_intermediate_bucket_result(
-            self.accessor_idx,
-            entries,
-            self.term_buckets.sub_aggs,
-            agg_data,
-        )?;
-        results.push(name, IntermediateAggregationResult::Bucket(bucket))?;
-
-        Ok(())
-    }
-
-    #[inline]
-    fn collect(
-        &mut self,
-        doc: crate::DocId,
-        agg_data: &mut AggregationsSegmentCtx,
-    ) -> crate::Result<()> {
-        self.collect_block(&[doc], agg_data)
-    }
-
-    #[inline]
-    fn collect_block(
-        &mut self,
-        docs: &[crate::DocId],
-        agg_data: &mut AggregationsSegmentCtx,
-    ) -> crate::Result<()> {
-        let mut req_data = agg_data.take_term_req_data(self.accessor_idx);
-
-        let mem_pre = self.get_memory_consumption();
-
-        if let Some(missing) = req_data.missing_value_for_accessor {
-            req_data.column_block_accessor.fetch_block_with_missing(
-                docs,
-                &req_data.accessor,
-                missing,
-            );
-        } else {
-            req_data
-                .column_block_accessor
-                .fetch_block(docs, &req_data.accessor);
-        }
-
-        for term_id in req_data.column_block_accessor.iter_vals() {
-            if let Some(allowed_bs) = req_data.allowed_term_ids.as_ref() {
-                if !allowed_bs.contains(term_id as u32) {
-                    continue;
-                }
-            }
-            let entry = self.term_buckets.entries.entry(term_id).or_default();
-            *entry += 1;
-        }
-        // has subagg
-        if let Some(blueprint) = req_data.sub_aggregation_blueprint.as_ref() {
-            for (doc, term_id) in req_data
-                .column_block_accessor
-                .iter_docid_vals(docs, &req_data.accessor)
-            {
-                if let Some(allowed_bs) = req_data.allowed_term_ids.as_ref() {
-                    if !allowed_bs.contains(term_id as u32) {
-                        continue;
-                    }
-                }
-                let sub_aggregations = self
-                    .term_buckets
-                    .sub_aggs
-                    .entry(term_id)
-                    .or_insert_with(|| blueprint.clone());
-                sub_aggregations.collect(doc, agg_data)?;
-            }
-        }
-
-        let mem_delta = self.get_memory_consumption() - mem_pre;
-        if mem_delta > 0 {
-            agg_data
-                .context
-                .limits
-                .add_memory_consumed(mem_delta as u64)?;
-        }
-        agg_data.put_back_term_req_data(self.accessor_idx, req_data);
-
-        Ok(())
-    }
-
-    fn flush(&mut self, agg_data: &mut AggregationsSegmentCtx) -> crate::Result<()> {
-        self.term_buckets.force_flush(agg_data)?;
-        Ok(())
-    }
-}
-
-impl SegmentTermCollector {
-    pub fn from_req_and_validate(
-        req_data: &mut AggregationsSegmentCtx,
-        node: &AggRefNode,
-    ) -> crate::Result<Self> {
-        let terms_req_data = req_data.get_term_req_data(node.idx_in_req_data);
-        let column_type = terms_req_data.column_type;
-        let accessor_idx = node.idx_in_req_data;
-        if column_type == ColumnType::Bytes {
-            return Err(TantivyError::InvalidArgument(format!(
-                "terms aggregation is not supported for column type {column_type:?}"
-            )));
-        }
-        let term_buckets = TermBuckets::default();
-
-        // Validate sub aggregation exists
-        if let OrderTarget::SubAggregation(sub_agg_name) = &terms_req_data.req.order.target {
-            let (agg_name, _agg_property) = get_agg_name_and_property(sub_agg_name);
-
-            node.get_sub_agg(agg_name, &req_data.per_request)
-                .ok_or_else(|| {
-                    TantivyError::InvalidArgument(format!(
-                        "could not find aggregation with name {agg_name} in metric \
-                         sub_aggregations"
-                    ))
-                })?;
-        }
-
-        let has_sub_aggregations = !node.children.is_empty();
-        let blueprint = if has_sub_aggregations {
-            let sub_aggregation = build_segment_agg_collectors(req_data, &node.children)?;
-            Some(sub_aggregation)
-        } else {
-            None
-        };
-        let terms_req_data = req_data.get_term_req_data_mut(node.idx_in_req_data);
-        terms_req_data.sub_aggregation_blueprint = blueprint;
-
-        Ok(SegmentTermCollector {
-            term_buckets,
-            accessor_idx,
-        })
-    }
-
-    fn get_memory_consumption(&self) -> usize {
-        let self_mem = std::mem::size_of::<Self>();
-        let term_buckets_mem = self.term_buckets.get_memory_consumption();
-        self_mem + term_buckets_mem
-    }
-}

src/aggregation/bucket/term_agg/low_cardinality_impl.rs

@@ -1,228 +0,0 @@
-use std::vec;
-
-use rustc_hash::FxHashMap;
-
-use crate::aggregation::agg_data::{
-    build_segment_agg_collectors, AggRefNode, AggregationsSegmentCtx,
-};
-use crate::aggregation::bucket::{get_agg_name_and_property, OrderTarget};
-use crate::aggregation::intermediate_agg_result::{
-    IntermediateAggregationResult, IntermediateAggregationResults,
-};
-use crate::aggregation::segment_agg_result::SegmentAggregationCollector;
-use crate::{DocId, TantivyError};
-
-const MAX_BATCH_SIZE: usize = 1_024;
-
-#[derive(Debug, Clone)]
-struct LowCardTermBuckets {
-    entries: Box<[u32]>,
-    sub_aggs: Vec<Box<dyn SegmentAggregationCollector>>,
-    doc_buffers: Box<[Vec<DocId>]>,
-}
-
-impl LowCardTermBuckets {
-    pub fn with_num_buckets(
-        num_buckets: usize,
-        sub_aggs_blueprint_opt: Option<&Box<dyn SegmentAggregationCollector>>,
-    ) -> Self {
-        let sub_aggs = sub_aggs_blueprint_opt
-            .as_ref()
-            .map(|blueprint| {
-                std::iter::repeat_with(|| blueprint.clone_box())
-                    .take(num_buckets)
-                    .collect::<Vec<_>>()
-            })
-            .unwrap_or_default();
-        Self {
-            entries: vec![0; num_buckets].into_boxed_slice(),
-            sub_aggs,
-            doc_buffers: std::iter::repeat_with(|| Vec::with_capacity(MAX_BATCH_SIZE))
-                .take(num_buckets)
-                .collect::<Vec<_>>()
-                .into_boxed_slice(),
-        }
-    }
-
-    fn get_memory_consumption(&self) -> usize {
-        std::mem::size_of::<Self>()
-            + self.entries.len() * std::mem::size_of::<u32>()
-            + self.doc_buffers.len()
-                * (std::mem::size_of::<Vec<DocId>>()
-                    + std::mem::size_of::<DocId>() * MAX_BATCH_SIZE)
-    }
-}
-
-#[derive(Debug, Clone)]
-pub struct LowCardSegmentTermCollector {
-    term_buckets: LowCardTermBuckets,
-    accessor_idx: usize,
-}
-
-impl LowCardSegmentTermCollector {
-    pub fn from_req_and_validate(
-        req_data: &mut AggregationsSegmentCtx,
-        node: &AggRefNode,
-    ) -> crate::Result<Self> {
-        let terms_req_data = req_data.get_term_req_data(node.idx_in_req_data);
-        let accessor_idx = node.idx_in_req_data;
-        let cardinality = terms_req_data
-            .accessor
-            .max_value()
-            .max(terms_req_data.missing_value_for_accessor.unwrap_or(0))
-            + 1;
-        assert!(cardinality <= super::LOW_CARDINALITY_THRESHOLD);
-
-        // Validate sub aggregation exists
-        if let OrderTarget::SubAggregation(sub_agg_name) = &terms_req_data.req.order.target {
-            let (agg_name, _agg_property) = get_agg_name_and_property(sub_agg_name);
-
-            node.get_sub_agg(agg_name, &req_data.per_request)
-                .ok_or_else(|| {
-                    TantivyError::InvalidArgument(format!(
-                        "could not find aggregation with name {agg_name} in metric \
-                         sub_aggregations"
-                    ))
-                })?;
-        }
-
-        let has_sub_aggregations = !node.children.is_empty();
-        let blueprint = if has_sub_aggregations {
-            let sub_aggregation = build_segment_agg_collectors(req_data, &node.children)?;
-            Some(sub_aggregation)
-        } else {
-            None
-        };
-        let terms_req_data = req_data.get_term_req_data_mut(node.idx_in_req_data);
-
-        let term_buckets =
-            LowCardTermBuckets::with_num_buckets(cardinality as usize, blueprint.as_ref());
-
-        terms_req_data.sub_aggregation_blueprint = blueprint;
-
-        Ok(LowCardSegmentTermCollector {
-            term_buckets,
-            accessor_idx,
-        })
-    }
-
-    fn get_memory_consumption(&self) -> usize {
-        let self_mem = std::mem::size_of::<Self>();
-        let term_buckets_mem = self.term_buckets.get_memory_consumption();
-        self_mem + term_buckets_mem
-    }
-}
-
-impl SegmentAggregationCollector for LowCardSegmentTermCollector {
-    fn add_intermediate_aggregation_result(
-        self: Box<Self>,
-        agg_data: &AggregationsSegmentCtx,
-        results: &mut IntermediateAggregationResults,
-    ) -> crate::Result<()> {
-        let name = agg_data.get_term_req_data(self.accessor_idx).name.clone();
-        let sub_aggs: FxHashMap<u64, Box<dyn SegmentAggregationCollector>> = self
-            .term_buckets
-            .sub_aggs
-            .into_iter()
-            .enumerate()
-            .filter(|(bucket_id, _sub_agg)| self.term_buckets.entries[*bucket_id] > 0)
-            .map(|(bucket_id, sub_agg)| (bucket_id as u64, sub_agg))
-            .collect();
-        let entries: Vec<(u64, u32)> = self
-            .term_buckets
-            .entries
-            .iter()
-            .enumerate()
-            .filter(|(_, count)| **count > 0)
-            .map(|(bucket_id, count)| (bucket_id as u64, *count))
-            .collect();
-
-        let bucket =
-            super::into_intermediate_bucket_result(self.accessor_idx, entries, sub_aggs, agg_data)?;
-        results.push(name, IntermediateAggregationResult::Bucket(bucket))?;
-        Ok(())
-    }
-
-    fn collect_block(
-        &mut self,
-        docs: &[crate::DocId],
-        agg_data: &mut AggregationsSegmentCtx,
-    ) -> crate::Result<()> {
-        if docs.len() > MAX_BATCH_SIZE {
-            for batch in docs.chunks(MAX_BATCH_SIZE) {
-                self.collect_block(batch, agg_data)?;
-            }
-        }
-
-        let mut req_data = agg_data.take_term_req_data(self.accessor_idx);
-
-        let mem_pre = self.get_memory_consumption();
-
-        if let Some(missing) = req_data.missing_value_for_accessor {
-            req_data.column_block_accessor.fetch_block_with_missing(
-                docs,
-                &req_data.accessor,
-                missing,
-            );
-        } else {
-            req_data
-                .column_block_accessor
-                .fetch_block(docs, &req_data.accessor);
-        }
-
-        // has subagg
-        if req_data.sub_aggregation_blueprint.is_some() {
-            for (doc, term_id) in req_data
-                .column_block_accessor
-                .iter_docid_vals(docs, &req_data.accessor)
-            {
-                if let Some(allowed_bs) = req_data.allowed_term_ids.as_ref() {
-                    if !allowed_bs.contains(term_id as u32) {
-                        continue;
-                    }
-                }
-                self.term_buckets.doc_buffers[term_id as usize].push(doc);
-            }
-            for (bucket_id, docs) in self.term_buckets.doc_buffers.iter_mut().enumerate() {
-                self.term_buckets.entries[bucket_id] += docs.len() as u32;
-                self.term_buckets.sub_aggs[bucket_id].collect_block(&docs[..], agg_data)?;
-                docs.clear();
-            }
-        } else {
-            for term_id in req_data.column_block_accessor.iter_vals() {
-                if let Some(allowed_bs) = req_data.allowed_term_ids.as_ref() {
-                    if !allowed_bs.contains(term_id as u32) {
-                        continue;
-                    }
-                }
-                self.term_buckets.entries[term_id as usize] += 1;
-            }
-        }
-
-        let mem_delta = self.get_memory_consumption() - mem_pre;
-        if mem_delta > 0 {
-            agg_data
-                .context
-                .limits
-                .add_memory_consumed(mem_delta as u64)?;
-        }
-        agg_data.put_back_term_req_data(self.accessor_idx, req_data);
-
-        Ok(())
-    }
-
-    fn collect(
-        &mut self,
-        doc: crate::DocId,
-        agg_data: &mut AggregationsSegmentCtx,
-    ) -> crate::Result<()> {
-        self.collect_block(&[doc], agg_data)
-    }
-
-    fn flush(&mut self, agg_data: &mut AggregationsSegmentCtx) -> crate::Result<()> {
-        for sub_aggregations in &mut self.term_buckets.sub_aggs.iter_mut() {
-            sub_aggregations.as_mut().flush(agg_data)?;
-        }
-        Ok(())
-    }
-}

src/aggregation/bucket/term_missing_agg.rs

@@ -5,11 +5,13 @@ use crate::aggregation::agg_data::{
     build_segment_agg_collectors, AggRefNode, AggregationsSegmentCtx,
 };
 use crate::aggregation::bucket::term_agg::TermsAggregation;
+use crate::aggregation::cached_sub_aggs::{CachedSubAggs, HighCardCachedSubAggs};
 use crate::aggregation::intermediate_agg_result::{
     IntermediateAggregationResult, IntermediateAggregationResults, IntermediateBucketResult,
     IntermediateKey, IntermediateTermBucketEntry, IntermediateTermBucketResult,
 };
-use crate::aggregation::segment_agg_result::SegmentAggregationCollector;
+use crate::aggregation::segment_agg_result::{BucketIdProvider, SegmentAggregationCollector};
+use crate::aggregation::BucketId;
 
 /// Special aggregation to handle missing values for term aggregations.
 /// This missing aggregation will check multiple columns for existence.
@@ -35,41 +37,55 @@ impl MissingTermAggReqData {
     }
 }
 
-/// The specialized missing term aggregation.
 #[derive(Default, Debug, Clone)]
-pub struct TermMissingAgg {
+struct MissingCount {
     missing_count: u32,
+    bucket_id: BucketId,
+}
+
+/// The specialized missing term aggregation.
+#[derive(Default, Debug)]
+pub struct TermMissingAgg {
     accessor_idx: usize,
-    sub_agg: Option<Box<dyn SegmentAggregationCollector>>,
+    sub_agg: Option<HighCardCachedSubAggs>,
+    /// Idx = parent bucket id, Value = missing count for that bucket
+    missing_count_per_bucket: Vec<MissingCount>,
+    bucket_id_provider: BucketIdProvider,
 }
 impl TermMissingAgg {
     pub(crate) fn new(
-        req_data: &mut AggregationsSegmentCtx,
+        agg_data: &mut AggregationsSegmentCtx,
         node: &AggRefNode,
     ) -> crate::Result<Self> {
         let has_sub_aggregations = !node.children.is_empty();
         let accessor_idx = node.idx_in_req_data;
         let sub_agg = if has_sub_aggregations {
-            let sub_aggregation = build_segment_agg_collectors(req_data, &node.children)?;
+            let sub_aggregation = build_segment_agg_collectors(agg_data, &node.children)?;
             Some(sub_aggregation)
         } else {
             None
         };
 
+        let sub_agg = sub_agg.map(CachedSubAggs::new);
+        let bucket_id_provider = BucketIdProvider::default();
+
         Ok(Self {
             accessor_idx,
             sub_agg,
-            ..Default::default()
+            missing_count_per_bucket: Vec::new(),
+            bucket_id_provider,
         })
     }
 }
 
 impl SegmentAggregationCollector for TermMissingAgg {
     fn add_intermediate_aggregation_result(
-        self: Box<Self>,
+        &mut self,
         agg_data: &AggregationsSegmentCtx,
         results: &mut IntermediateAggregationResults,
+        parent_bucket_id: BucketId,
     ) -> crate::Result<()> {
+        self.prepare_max_bucket(parent_bucket_id, agg_data)?;
         let req_data = agg_data.get_missing_term_req_data(self.accessor_idx);
         let term_agg = &req_data.req;
         let missing = term_agg
@@ -80,13 +96,16 @@ impl SegmentAggregationCollector for TermMissingAgg {
         let mut entries: FxHashMap<IntermediateKey, IntermediateTermBucketEntry> =
             Default::default();
 
+        let missing_count = &self.missing_count_per_bucket[parent_bucket_id as usize];
         let mut missing_entry = IntermediateTermBucketEntry {
-            doc_count: self.missing_count,
+            doc_count: missing_count.missing_count,
             sub_aggregation: Default::default(),
         };
-        if let Some(sub_agg) = self.sub_agg {
+        if let Some(sub_agg) = &mut self.sub_agg {
             let mut res = IntermediateAggregationResults::default();
-            sub_agg.add_intermediate_aggregation_result(agg_data, &mut res)?;
+            sub_agg
+                .get_sub_agg_collector()
+                .add_intermediate_aggregation_result(agg_data, &mut res, missing_count.bucket_id)?;
             missing_entry.sub_aggregation = res;
         }
         entries.insert(missing.into(), missing_entry);
@@ -109,30 +128,52 @@ impl SegmentAggregationCollector for TermMissingAgg {
 
     fn collect(
         &mut self,
-        doc: crate::DocId,
+        parent_bucket_id: BucketId,
+        docs: &[crate::DocId],
         agg_data: &mut AggregationsSegmentCtx,
     ) -> crate::Result<()> {
+        let bucket = &mut self.missing_count_per_bucket[parent_bucket_id as usize];
         let req_data = agg_data.get_missing_term_req_data(self.accessor_idx);
-        let has_value = req_data
-            .accessors
-            .iter()
-            .any(|(acc, _)| acc.index.has_value(doc));
-        if !has_value {
-            self.missing_count += 1;
-            if let Some(sub_agg) = self.sub_agg.as_mut() {
-                sub_agg.collect(doc, agg_data)?;
+
+        for doc in docs {
+            let doc = *doc;
+            let has_value = req_data
+                .accessors
+                .iter()
+                .any(|(acc, _)| acc.index.has_value(doc));
+            if !has_value {
+                bucket.missing_count += 1;
+
+                if let Some(sub_agg) = self.sub_agg.as_mut() {
+                    sub_agg.push(bucket.bucket_id, doc);
+                }
             }
         }
+
+        if let Some(sub_agg) = self.sub_agg.as_mut() {
+            sub_agg.check_flush_local(agg_data)?;
+        }
         Ok(())
     }
 
-    fn collect_block(
+    fn prepare_max_bucket(
         &mut self,
-        docs: &[crate::DocId],
-        agg_data: &mut AggregationsSegmentCtx,
+        max_bucket: BucketId,
+        _agg_data: &AggregationsSegmentCtx,
     ) -> crate::Result<()> {
-        for doc in docs {
-            self.collect(*doc, agg_data)?;
+        while self.missing_count_per_bucket.len() <= max_bucket as usize {
+            let bucket_id = self.bucket_id_provider.next_bucket_id();
+            self.missing_count_per_bucket.push(MissingCount {
+                missing_count: 0,
+                bucket_id,
+            });
+        }
+        Ok(())
+    }
+
+    fn flush(&mut self, agg_data: &mut AggregationsSegmentCtx) -> crate::Result<()> {
+        if let Some(sub_agg) = self.sub_agg.as_mut() {
+            sub_agg.flush(agg_data)?;
         }
         Ok(())
     }

src/aggregation/buf_collector.rs

@@ -1,83 +0,0 @@
-use super::intermediate_agg_result::IntermediateAggregationResults;
-use super::segment_agg_result::SegmentAggregationCollector;
-use crate::aggregation::agg_data::AggregationsSegmentCtx;
-use crate::DocId;
-
-pub(crate) const DOC_BLOCK_SIZE: usize = 64;
-pub(crate) type DocBlock = [DocId; DOC_BLOCK_SIZE];
-
-/// BufAggregationCollector buffers documents before calling collect_block().
-#[derive(Clone)]
-pub(crate) struct BufAggregationCollector {
-    pub(crate) collector: Box<dyn SegmentAggregationCollector>,
-    staged_docs: DocBlock,
-    num_staged_docs: usize,
-}
-
-impl std::fmt::Debug for BufAggregationCollector {
-    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
-        f.debug_struct("SegmentAggregationResultsCollector")
-            .field("staged_docs", &&self.staged_docs[..self.num_staged_docs])
-            .field("num_staged_docs", &self.num_staged_docs)
-            .finish()
-    }
-}
-
-impl BufAggregationCollector {
-    pub fn new(collector: Box<dyn SegmentAggregationCollector>) -> Self {
-        Self {
-            collector,
-            num_staged_docs: 0,
-            staged_docs: [0; DOC_BLOCK_SIZE],
-        }
-    }
-}
-
-impl SegmentAggregationCollector for BufAggregationCollector {
-    #[inline]
-    fn add_intermediate_aggregation_result(
-        self: Box<Self>,
-        agg_data: &AggregationsSegmentCtx,
-        results: &mut IntermediateAggregationResults,
-    ) -> crate::Result<()> {
-        Box::new(self.collector).add_intermediate_aggregation_result(agg_data, results)
-    }
-
-    #[inline]
-    fn collect(
-        &mut self,
-        doc: crate::DocId,
-        agg_data: &mut AggregationsSegmentCtx,
-    ) -> crate::Result<()> {
-        self.staged_docs[self.num_staged_docs] = doc;
-        self.num_staged_docs += 1;
-        if self.num_staged_docs == self.staged_docs.len() {
-            self.collector
-                .collect_block(&self.staged_docs[..self.num_staged_docs], agg_data)?;
-            self.num_staged_docs = 0;
-        }
-        Ok(())
-    }
-
-    #[inline]
-    fn collect_block(
-        &mut self,
-        docs: &[crate::DocId],
-        agg_data: &mut AggregationsSegmentCtx,
-    ) -> crate::Result<()> {
-        self.collector.collect_block(docs, agg_data)?;
-
-        Ok(())
-    }
-
-    #[inline]
-    fn flush(&mut self, agg_data: &mut AggregationsSegmentCtx) -> crate::Result<()> {
-        self.collector
-            .collect_block(&self.staged_docs[..self.num_staged_docs], agg_data)?;
-        self.num_staged_docs = 0;
-
-        self.collector.flush(agg_data)?;
-
-        Ok(())
-    }
-}

src/aggregation/cached_sub_aggs.rs

@@ -0,0 +1,245 @@
+use std::fmt::Debug;
+
+use super::segment_agg_result::SegmentAggregationCollector;
+use crate::aggregation::agg_data::AggregationsSegmentCtx;
+use crate::aggregation::bucket::MAX_NUM_TERMS_FOR_VEC;
+use crate::aggregation::BucketId;
+use crate::DocId;
+
+/// A cache for sub-aggregations, storing doc ids per bucket id.
+/// Depending on the cardinality of the parent aggregation, we use different
+/// storage strategies.
+///
+/// ## Low Cardinality
+/// Cardinality here refers to the number of unique flattened buckets that can be created
+/// by the parent aggregation.
+/// Flattened buckets are the result of combining all buckets per collector
+/// into a single list of buckets, where each bucket is identified by its BucketId.
+///
+/// ## Usage
+/// Since this is caching for sub-aggregations, it is only used by bucket
+/// aggregations.
+///
+/// TODO: consider using a more advanced data structure for high cardinality
+/// aggregations.
+/// What this datastructure does in general is to group docs by bucket id.
+#[derive(Debug)]
+pub(crate) struct CachedSubAggs<C: SubAggCache> {
+    cache: C,
+    sub_agg_collector: Box<dyn SegmentAggregationCollector>,
+    num_docs: usize,
+}
+
+pub type LowCardCachedSubAggs = CachedSubAggs<LowCardSubAggCache>;
+pub type HighCardCachedSubAggs = CachedSubAggs<HighCardSubAggCache>;
+
+const FLUSH_THRESHOLD: usize = 2048;
+
+/// A trait for caching sub-aggregation doc ids per bucket id.
+/// Different implementations can be used depending on the cardinality
+/// of the parent aggregation.
+pub trait SubAggCache: Debug {
+    fn new() -> Self;
+    fn push(&mut self, bucket_id: BucketId, doc_id: DocId);
+    fn flush_local(
+        &mut self,
+        sub_agg: &mut Box<dyn SegmentAggregationCollector>,
+        agg_data: &mut AggregationsSegmentCtx,
+        force: bool,
+    ) -> crate::Result<()>;
+}
+
+impl<Backend: SubAggCache + Debug> CachedSubAggs<Backend> {
+    pub fn new(sub_agg: Box<dyn SegmentAggregationCollector>) -> Self {
+        Self {
+            cache: Backend::new(),
+            sub_agg_collector: sub_agg,
+            num_docs: 0,
+        }
+    }
+
+    pub fn get_sub_agg_collector(&mut self) -> &mut Box<dyn SegmentAggregationCollector> {
+        &mut self.sub_agg_collector
+    }
+
+    #[inline]
+    pub fn push(&mut self, bucket_id: BucketId, doc_id: DocId) {
+        self.cache.push(bucket_id, doc_id);
+        self.num_docs += 1;
+    }
+
+    /// Check if we need to flush based on the number of documents cached.
+    /// If so, flushes the cache to the provided aggregation collector.
+    pub fn check_flush_local(
+        &mut self,
+        agg_data: &mut AggregationsSegmentCtx,
+    ) -> crate::Result<()> {
+        if self.num_docs >= FLUSH_THRESHOLD {
+            self.cache
+                .flush_local(&mut self.sub_agg_collector, agg_data, false)?;
+            self.num_docs = 0;
+        }
+        Ok(())
+    }
+
+    /// Note: this _does_ flush the sub aggregations.
+    pub fn flush(&mut self, agg_data: &mut AggregationsSegmentCtx) -> crate::Result<()> {
+        if self.num_docs != 0 {
+            self.cache
+                .flush_local(&mut self.sub_agg_collector, agg_data, true)?;
+            self.num_docs = 0;
+        }
+        self.sub_agg_collector.flush(agg_data)?;
+        Ok(())
+    }
+}
+
+/// Number of partitions for high cardinality sub-aggregation cache.
+const NUM_PARTITIONS: usize = 16;
+
+#[derive(Debug)]
+pub(crate) struct HighCardSubAggCache {
+    /// This weird partitioning is used to do some cheap grouping on the bucket ids.
+    /// bucket ids are dense, e.g. when we don't detect the cardinality as low cardinality,
+    /// but there are just 16 bucket ids, each bucket id will go to its own partition.
+    ///
+    /// We want to keep this cheap, because high cardinality aggregations can have a lot of
+    /// buckets, and there may be nothing to group.
+    partitions: Box<[PartitionEntry; NUM_PARTITIONS]>,
+}
+
+impl HighCardSubAggCache {
+    #[inline]
+    fn clear(&mut self) {
+        for partition in self.partitions.iter_mut() {
+            partition.clear();
+        }
+    }
+}
+
+#[derive(Debug, Clone, Default)]
+struct PartitionEntry {
+    bucket_ids: Vec<BucketId>,
+    docs: Vec<DocId>,
+}
+
+impl PartitionEntry {
+    #[inline]
+    fn clear(&mut self) {
+        self.bucket_ids.clear();
+        self.docs.clear();
+    }
+}
+
+impl SubAggCache for HighCardSubAggCache {
+    fn new() -> Self {
+        Self {
+            partitions: Box::new(core::array::from_fn(|_| PartitionEntry::default())),
+        }
+    }
+
+    fn push(&mut self, bucket_id: BucketId, doc_id: DocId) {
+        let idx = bucket_id % NUM_PARTITIONS as u32;
+        let slot = &mut self.partitions[idx as usize];
+        slot.bucket_ids.push(bucket_id);
+        slot.docs.push(doc_id);
+    }
+
+    fn flush_local(
+        &mut self,
+        sub_agg: &mut Box<dyn SegmentAggregationCollector>,
+        agg_data: &mut AggregationsSegmentCtx,
+        _force: bool,
+    ) -> crate::Result<()> {
+        let mut max_bucket = 0u32;
+        for partition in self.partitions.iter() {
+            if let Some(&local_max) = partition.bucket_ids.iter().max() {
+                max_bucket = max_bucket.max(local_max);
+            }
+        }
+
+        sub_agg.prepare_max_bucket(max_bucket, agg_data)?;
+
+        for slot in self.partitions.iter() {
+            if !slot.bucket_ids.is_empty() {
+                // Reduce dynamic dispatch overhead by collecting a full partition in one call.
+                sub_agg.collect_multiple(&slot.bucket_ids, &slot.docs, agg_data)?;
+            }
+        }
+
+        self.clear();
+        Ok(())
+    }
+}
+
+#[derive(Debug)]
+pub(crate) struct LowCardSubAggCache {
+    /// Cache doc ids per bucket for sub-aggregations.
+    ///
+    /// The outer Vec is indexed by BucketId.
+    per_bucket_docs: Vec<Vec<DocId>>,
+}
+
+impl LowCardSubAggCache {
+    #[inline]
+    fn clear(&mut self) {
+        for v in &mut self.per_bucket_docs {
+            v.clear();
+        }
+    }
+}
+
+impl SubAggCache for LowCardSubAggCache {
+    fn new() -> Self {
+        Self {
+            per_bucket_docs: Vec::new(),
+        }
+    }
+
+    fn push(&mut self, bucket_id: BucketId, doc_id: DocId) {
+        let idx = bucket_id as usize;
+        if self.per_bucket_docs.len() <= idx {
+            self.per_bucket_docs.resize_with(idx + 1, Vec::new);
+        }
+        self.per_bucket_docs[idx].push(doc_id);
+    }
+
+    fn flush_local(
+        &mut self,
+        sub_agg: &mut Box<dyn SegmentAggregationCollector>,
+        agg_data: &mut AggregationsSegmentCtx,
+        force: bool,
+    ) -> crate::Result<()> {
+        // Pre-aggregated: call collect per bucket.
+        let max_bucket = (self.per_bucket_docs.len() as BucketId).saturating_sub(1);
+        sub_agg.prepare_max_bucket(max_bucket, agg_data)?;
+        // The threshold above which we flush buckets individually.
+        // Note: We need to make sure that we don't lock ourselves into a situation where we hit
+        // the FLUSH_THRESHOLD, but never flush any buckets. (except the final flush)
+        let mut bucket_treshold = FLUSH_THRESHOLD / (self.per_bucket_docs.len().max(1) * 2);
+        const _: () = {
+            // MAX_NUM_TERMS_FOR_VEC threshold is used for term aggregations
+            // Note: There may be other flexible values, for other aggregations, but we can use the
+            // const value here as a upper bound. (better than nothing)
+            let bucket_treshold_limit = FLUSH_THRESHOLD / (MAX_NUM_TERMS_FOR_VEC as usize * 2);
+            assert!(
+                bucket_treshold_limit > 0,
+                "Bucket threshold must be greater than 0"
+            );
+        };
+        if force {
+            bucket_treshold = 0;
+        }
+        for (bucket_id, docs) in self
+            .per_bucket_docs
+            .iter()
+            .enumerate()
+            .filter(|(_, docs)| docs.len() > bucket_treshold)
+        {
+            sub_agg.collect(bucket_id as BucketId, docs, agg_data)?;
+        }
+
+        self.clear();
+        Ok(())
+    }
+}

src/aggregation/collector.rs

@@ -1,9 +1,9 @@
 use super::agg_req::Aggregations;
 use super::agg_result::AggregationResults;
-use super::buf_collector::BufAggregationCollector;
+use super::cached_sub_aggs::LowCardCachedSubAggs;
 use super::intermediate_agg_result::IntermediateAggregationResults;
-use super::segment_agg_result::SegmentAggregationCollector;
 use super::AggContextParams;
+// group buffering strategy is chosen explicitly by callers; no need to hash-group on the fly.
 use crate::aggregation::agg_data::{
     build_aggregations_data_from_req, build_segment_agg_collectors_root, AggregationsSegmentCtx,
 };
@@ -136,7 +136,7 @@ fn merge_fruits(
 /// `AggregationSegmentCollector` does the aggregation collection on a segment.
 pub struct AggregationSegmentCollector {
     aggs_with_accessor: AggregationsSegmentCtx,
-    agg_collector: BufAggregationCollector,
+    agg_collector: LowCardCachedSubAggs,
     error: Option<TantivyError>,
 }
 
@@ -151,8 +151,11 @@ impl AggregationSegmentCollector {
     ) -> crate::Result<Self> {
         let mut agg_data =
             build_aggregations_data_from_req(agg, reader, segment_ordinal, context.clone())?;
-        let result =
-            BufAggregationCollector::new(build_segment_agg_collectors_root(&mut agg_data)?);
+        let mut result =
+            LowCardCachedSubAggs::new(build_segment_agg_collectors_root(&mut agg_data)?);
+        result
+            .get_sub_agg_collector()
+            .prepare_max_bucket(0, &agg_data)?; // prepare for bucket zero
 
         Ok(AggregationSegmentCollector {
             aggs_with_accessor: agg_data,
@@ -170,26 +173,31 @@ impl SegmentCollector for AggregationSegmentCollector {
         if self.error.is_some() {
             return;
         }
-        if let Err(err) = self
+        self.agg_collector.push(0, doc);
+        match self
             .agg_collector
-            .collect(doc, &mut self.aggs_with_accessor)
+            .check_flush_local(&mut self.aggs_with_accessor)
         {
-            self.error = Some(err);
+            Ok(_) => {}
+            Err(e) => {
+                self.error = Some(e);
+            }
         }
     }
-
-    /// The query pushes the documents to the collector via this method.
-    ///
-    /// Only valid for Collectors that ignore docs
     fn collect_block(&mut self, docs: &[DocId]) {
         if self.error.is_some() {
             return;
         }
-        if let Err(err) = self
-            .agg_collector
-            .collect_block(docs, &mut self.aggs_with_accessor)
-        {
-            self.error = Some(err);
+
+        match self.agg_collector.get_sub_agg_collector().collect(
+            0,
+            docs,
+            &mut self.aggs_with_accessor,
+        ) {
+            Ok(_) => {}
+            Err(e) => {
+                self.error = Some(e);
+            }
         }
     }
 
@@ -200,10 +208,13 @@ impl SegmentCollector for AggregationSegmentCollector {
         self.agg_collector.flush(&mut self.aggs_with_accessor)?;
 
         let mut sub_aggregation_res = IntermediateAggregationResults::default();
-        Box::new(self.agg_collector).add_intermediate_aggregation_result(
-            &self.aggs_with_accessor,
-            &mut sub_aggregation_res,
-        )?;
+        self.agg_collector
+            .get_sub_agg_collector()
+            .add_intermediate_aggregation_result(
+                &self.aggs_with_accessor,
+                &mut sub_aggregation_res,
+                0,
+            )?;
 
         Ok(sub_aggregation_res)
     }

src/aggregation/intermediate_agg_result.rs

@@ -15,8 +15,9 @@ use serde::{Deserialize, Serialize};
 use super::agg_req::{Aggregation, AggregationVariants, Aggregations};
 use super::agg_result::{AggregationResult, BucketResult, MetricResult, RangeBucketEntry};
 use super::bucket::{
-    cut_off_buckets, get_agg_name_and_property, intermediate_histogram_buckets_to_final_buckets,
-    GetDocCount, Order, OrderTarget, RangeAggregation, TermsAggregation,
+    composite_intermediate_key_ordering, cut_off_buckets, get_agg_name_and_property,
+    intermediate_histogram_buckets_to_final_buckets, CompositeAggregation, GetDocCount,
+    MissingOrder, Order, OrderTarget, RangeAggregation, TermsAggregation,
 };
 use super::metric::{
     IntermediateAverage, IntermediateCount, IntermediateExtendedStats, IntermediateMax,
@@ -25,7 +26,7 @@ use super::metric::{
 use super::segment_agg_result::AggregationLimitsGuard;
 use super::{format_date, AggregationError, Key, SerializedKey};
 use crate::aggregation::agg_result::{
-    AggregationResults, BucketEntries, BucketEntry, FilterBucketResult,
+    AggregationResults, BucketEntries, BucketEntry, CompositeBucketEntry, FilterBucketResult,
 };
 use crate::aggregation::bucket::TermsAggregationInternal;
 use crate::aggregation::metric::CardinalityCollector;
@@ -90,6 +91,19 @@ impl From<IntermediateKey> for Key {
 
 impl Eq for IntermediateKey {}
 
+impl std::fmt::Display for IntermediateKey {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        match self {
+            IntermediateKey::Str(val) => f.write_str(val),
+            IntermediateKey::F64(val) => f.write_str(&val.to_string()),
+            IntermediateKey::U64(val) => f.write_str(&val.to_string()),
+            IntermediateKey::I64(val) => f.write_str(&val.to_string()),
+            IntermediateKey::Bool(val) => f.write_str(&val.to_string()),
+            IntermediateKey::IpAddr(val) => f.write_str(&val.to_string()),
+        }
+    }
+}
+
 impl std::hash::Hash for IntermediateKey {
     fn hash<H: std::hash::Hasher>(&self, state: &mut H) {
         core::mem::discriminant(self).hash(state);
@@ -105,6 +119,21 @@ impl std::hash::Hash for IntermediateKey {
 }
 
 impl IntermediateAggregationResults {
+    /// Returns a reference to the intermediate aggregation result for the given key.
+    pub fn get(&self, key: &str) -> Option<&IntermediateAggregationResult> {
+        self.aggs_res.get(key)
+    }
+
+    /// Removes and returns the intermediate aggregation result for the given key.
+    pub fn remove(&mut self, key: &str) -> Option<IntermediateAggregationResult> {
+        self.aggs_res.remove(key)
+    }
+
+    /// Returns an iterator over the keys in the intermediate aggregation results.
+    pub fn keys(&self) -> impl Iterator<Item = &String> {
+        self.aggs_res.keys()
+    }
+
     /// Add a result
     pub fn push(&mut self, key: String, value: IntermediateAggregationResult) -> crate::Result<()> {
         let entry = self.aggs_res.entry(key);
@@ -252,6 +281,11 @@ pub(crate) fn empty_from_req(req: &Aggregation) -> IntermediateAggregationResult
             doc_count: 0,
             sub_aggregations: IntermediateAggregationResults::default(),
         }),
+        Composite(_) => {
+            IntermediateAggregationResult::Bucket(IntermediateBucketResult::Composite {
+                buckets: IntermediateCompositeBucketResult::default(),
+            })
+        }
     }
 }
 
@@ -445,6 +479,11 @@ pub enum IntermediateBucketResult {
         /// Sub-aggregation results
         sub_aggregations: IntermediateAggregationResults,
     },
+    /// Composite aggregation
+    Composite {
+        /// The composite buckets
+        buckets: IntermediateCompositeBucketResult,
+    },
 }
 
 impl IntermediateBucketResult {
@@ -540,6 +579,13 @@ impl IntermediateBucketResult {
                     sub_aggregations: final_sub_aggregations,
                 }))
             }
+            IntermediateBucketResult::Composite { buckets } => {
+                let composite_req = req
+                    .agg
+                    .as_composite()
+                    .expect("unexpected aggregation, expected composite aggregation");
+                buckets.into_final_result(composite_req, req.sub_aggregation(), limits)
+            }
         }
     }
 
@@ -606,6 +652,16 @@ impl IntermediateBucketResult {
                 *doc_count_left += doc_count_right;
                 sub_aggs_left.merge_fruits(sub_aggs_right)?;
             }
+            (
+                IntermediateBucketResult::Composite {
+                    buckets: composite_left,
+                },
+                IntermediateBucketResult::Composite {
+                    buckets: composite_right,
+                },
+            ) => {
+                composite_left.merge_fruits(composite_right)?;
+            }
             (IntermediateBucketResult::Range(_), _) => {
                 panic!("try merge on different types")
             }
@@ -618,6 +674,9 @@ impl IntermediateBucketResult {
             (IntermediateBucketResult::Filter { .. }, _) => {
                 panic!("try merge on different types")
             }
+            (IntermediateBucketResult::Composite { .. }, _) => {
+                panic!("try merge on different types")
+            }
         }
         Ok(())
     }
@@ -639,6 +698,21 @@ pub struct IntermediateTermBucketResult {
 }
 
 impl IntermediateTermBucketResult {
+    /// Returns a reference to the map of bucket entries keyed by [`IntermediateKey`].
+    pub fn entries(&self) -> &FxHashMap<IntermediateKey, IntermediateTermBucketEntry> {
+        &self.entries
+    }
+
+    /// Returns the count of documents not included in the returned buckets.
+    pub fn sum_other_doc_count(&self) -> u64 {
+        self.sum_other_doc_count
+    }
+
+    /// Returns the upper bound of the error on document counts in the returned buckets.
+    pub fn doc_count_error_upper_bound(&self) -> u64 {
+        self.doc_count_error_upper_bound
+    }
+
     pub(crate) fn into_final_result(
         self,
         req: &TermsAggregation,
@@ -792,7 +866,7 @@ pub struct IntermediateRangeBucketEntry {
     /// The number of documents in the bucket.
     pub doc_count: u64,
     /// The sub_aggregation in this bucket.
-    pub sub_aggregation: IntermediateAggregationResults,
+    pub sub_aggregation_res: IntermediateAggregationResults,
     /// The from range of the bucket. Equals `f64::MIN` when `None`.
     pub from: Option<f64>,
     /// The to range of the bucket. Equals `f64::MAX` when `None`.
@@ -811,7 +885,7 @@ impl IntermediateRangeBucketEntry {
             key: self.key.into(),
             doc_count: self.doc_count,
             sub_aggregation: self
-                .sub_aggregation
+                .sub_aggregation_res
                 .into_final_result_internal(req, limits)?,
             to: self.to,
             from: self.from,
@@ -820,7 +894,7 @@ impl IntermediateRangeBucketEntry {
         };
 
         // If we have a date type on the histogram buckets, we add the `key_as_string` field as
-        // rfc339
+        // rfc3339
         if column_type == Some(ColumnType::DateTime) {
             if let Some(val) = range_bucket_entry.to {
                 let key_as_string = format_date(val as i64)?;
@@ -857,7 +931,8 @@ impl MergeFruits for IntermediateTermBucketEntry {
 impl MergeFruits for IntermediateRangeBucketEntry {
     fn merge_fruits(&mut self, other: IntermediateRangeBucketEntry) -> crate::Result<()> {
         self.doc_count += other.doc_count;
-        self.sub_aggregation.merge_fruits(other.sub_aggregation)?;
+        self.sub_aggregation_res
+            .merge_fruits(other.sub_aggregation_res)?;
         Ok(())
     }
 }
@@ -870,6 +945,176 @@ impl MergeFruits for IntermediateHistogramBucketEntry {
     }
 }
 
+/// Entry for the composite bucket.
+pub type IntermediateCompositeBucketEntry = IntermediateTermBucketEntry;
+
+/// The fully typed key for composite aggregation
+#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
+pub enum CompositeIntermediateKey {
+    /// Bool key
+    Bool(bool),
+    /// String key
+    Str(String),
+    /// Float key
+    F64(f64),
+    /// Signed integer key
+    I64(i64),
+    /// Unsigned integer key
+    U64(u64),
+    /// DateTime key, nanoseconds since epoch
+    DateTime(i64),
+    /// IP Address key
+    IpAddr(Ipv6Addr),
+    /// Missing value key
+    Null,
+}
+
+impl Eq for CompositeIntermediateKey {}
+
+impl std::hash::Hash for CompositeIntermediateKey {
+    fn hash<H: std::hash::Hasher>(&self, state: &mut H) {
+        core::mem::discriminant(self).hash(state);
+        match self {
+            CompositeIntermediateKey::Bool(val) => val.hash(state),
+            CompositeIntermediateKey::Str(text) => text.hash(state),
+            CompositeIntermediateKey::F64(val) => val.to_bits().hash(state),
+            CompositeIntermediateKey::U64(val) => val.hash(state),
+            CompositeIntermediateKey::I64(val) => val.hash(state),
+            CompositeIntermediateKey::DateTime(val) => val.hash(state),
+            CompositeIntermediateKey::IpAddr(val) => val.hash(state),
+            CompositeIntermediateKey::Null => {}
+        }
+    }
+}
+
+/// Composite aggregation page.
+#[derive(Default, Clone, Debug, PartialEq, Serialize, Deserialize)]
+pub struct IntermediateCompositeBucketResult {
+    pub(crate) entries: FxHashMap<Vec<CompositeIntermediateKey>, IntermediateCompositeBucketEntry>,
+    pub(crate) target_size: u32,
+    pub(crate) orders: Vec<(Order, MissingOrder)>,
+}
+
+impl IntermediateCompositeBucketResult {
+    pub(crate) fn into_final_result(
+        self,
+        req: &CompositeAggregation,
+        sub_aggregation_req: &Aggregations,
+        limits: &mut AggregationLimitsGuard,
+    ) -> crate::Result<BucketResult> {
+        let trimmed_entry_vec =
+            trim_composite_buckets(self.entries, &self.orders, self.target_size)?;
+        let after_key = if trimmed_entry_vec.len() == req.size as usize {
+            trimmed_entry_vec
+                .last()
+                .map(|bucket| {
+                    let (intermediate_key, _entry) = bucket;
+                    intermediate_key
+                        .iter()
+                        .enumerate()
+                        .map(|(idx, intermediate_key)| {
+                            let source = &req.sources[idx];
+                            (source.name().to_string(), intermediate_key.clone().into())
+                        })
+                        .collect()
+                })
+                .unwrap()
+        } else {
+            FxHashMap::default()
+        };
+
+        let buckets = trimmed_entry_vec
+            .into_iter()
+            .map(|(intermediate_key, entry)| {
+                let key = intermediate_key
+                    .into_iter()
+                    .enumerate()
+                    .map(|(idx, intermediate_key)| {
+                        let source = &req.sources[idx];
+                        (source.name().to_string(), intermediate_key.into())
+                    })
+                    .collect();
+                Ok(CompositeBucketEntry {
+                    key,
+                    doc_count: entry.doc_count as u64,
+                    sub_aggregation: entry
+                        .sub_aggregation
+                        .into_final_result_internal(sub_aggregation_req, limits)?,
+                })
+            })
+            .collect::<crate::Result<Vec<_>>>()?;
+
+        Ok(BucketResult::Composite { after_key, buckets })
+    }
+
+    fn merge_fruits(&mut self, other: IntermediateCompositeBucketResult) -> crate::Result<()> {
+        merge_maps(&mut self.entries, other.entries)?;
+        if self.entries.len() as u32 > 2 * self.target_size {
+            self.trim()?;
+        }
+        Ok(())
+    }
+
+    /// Trim the composite buckets to the target size, according to the ordering.
+    pub(crate) fn trim(&mut self) -> crate::Result<()> {
+        if self.entries.len() as u32 <= self.target_size {
+            return Ok(());
+        }
+
+        let sorted_entries = trim_composite_buckets(
+            std::mem::take(&mut self.entries),
+            &self.orders,
+            self.target_size,
+        )?;
+
+        self.entries = sorted_entries.into_iter().collect();
+        Ok(())
+    }
+}
+
+fn trim_composite_buckets(
+    entries: FxHashMap<Vec<CompositeIntermediateKey>, IntermediateCompositeBucketEntry>,
+    orders: &[(Order, MissingOrder)],
+    target_size: u32,
+) -> crate::Result<
+    Vec<(
+        Vec<CompositeIntermediateKey>,
+        IntermediateCompositeBucketEntry,
+    )>,
+> {
+    let mut entries: Vec<_> = entries.into_iter().collect();
+    let mut sort_error: Option<TantivyError> = None;
+    entries.sort_by(|(left_key, _), (right_key, _)| {
+        if sort_error.is_some() {
+            return Ordering::Equal;
+        }
+
+        for idx in 0..orders.len() {
+            match composite_intermediate_key_ordering(
+                &left_key[idx],
+                &right_key[idx],
+                orders[idx].0,
+                orders[idx].1,
+            ) {
+                Ok(ordering) if ordering != Ordering::Equal => return ordering,
+                Ok(_) => continue,
+                Err(err) => {
+                    sort_error = Some(err);
+                    break;
+                }
+            }
+        }
+        Ordering::Equal
+    });
+
+    if let Some(err) = sort_error {
+        return Err(err);
+    }
+
+    entries.truncate(target_size as usize);
+    Ok(entries)
+}
+
 #[cfg(test)]
 mod tests {
     use std::collections::HashMap;
@@ -887,7 +1132,7 @@ mod tests {
                 IntermediateRangeBucketEntry {
                     key: IntermediateKey::Str(key.to_string()),
                     doc_count: *doc_count,
-                    sub_aggregation: Default::default(),
+                    sub_aggregation_res: Default::default(),
                     from: None,
                     to: None,
                 },
@@ -920,7 +1165,7 @@ mod tests {
                     doc_count: *doc_count,
                     from: None,
                     to: None,
-                    sub_aggregation: get_sub_test_tree(&[(
+                    sub_aggregation_res: get_sub_test_tree(&[(
                         sub_aggregation_key.to_string(),
                         *sub_aggregation_count,
                     )]),

src/aggregation/metric/average.rs

@@ -52,11 +52,15 @@ pub struct IntermediateAverage {
 
 impl IntermediateAverage {
     /// Creates a new [`IntermediateAverage`] instance from a [`SegmentStatsCollector`].
-    pub(crate) fn from_collector(collector: SegmentStatsCollector) -> Self {
-        Self {
-            stats: collector.stats,
-        }
+    pub(crate) fn from_stats(stats: IntermediateStats) -> Self {
+        Self { stats }
     }
+
+    /// Returns a reference to the underlying [`IntermediateStats`].
+    pub fn stats(&self) -> &IntermediateStats {
+        &self.stats
+    }
+
     /// Merges the other intermediate result into self.
     pub fn merge_fruits(&mut self, other: IntermediateAverage) {
         self.stats.merge_fruits(other.stats);

src/aggregation/metric/cardinality.rs

@@ -1,12 +1,11 @@
-use std::collections::hash_map::DefaultHasher;
-use std::hash::{BuildHasher, Hasher};
+use std::hash::Hash;
 
 use columnar::column_values::CompactSpaceU64Accessor;
-use columnar::{Column, ColumnBlockAccessor, ColumnType, Dictionary, StrColumn};
+use columnar::{Column, ColumnType, Dictionary, StrColumn};
 use common::f64_to_u64;
-use hyperloglogplus::{HyperLogLog, HyperLogLogPlus};
+use datasketches::hll::{HllSketch, HllType, HllUnion};
 use rustc_hash::FxHashSet;
-use serde::{Deserialize, Serialize};
+use serde::{Deserialize, Deserializer, Serialize, Serializer};
 
 use crate::aggregation::agg_data::AggregationsSegmentCtx;
 use crate::aggregation::intermediate_agg_result::{
@@ -16,29 +15,17 @@ use crate::aggregation::segment_agg_result::SegmentAggregationCollector;
 use crate::aggregation::*;
 use crate::TantivyError;
 
-#[derive(Clone, Debug, Serialize, Deserialize)]
-struct BuildSaltedHasher {
-    salt: u8,
-}
-
-impl BuildHasher for BuildSaltedHasher {
-    type Hasher = DefaultHasher;
-
-    fn build_hasher(&self) -> Self::Hasher {
-        let mut hasher = DefaultHasher::new();
-        hasher.write_u8(self.salt);
-
-        hasher
-    }
-}
+/// Log2 of the number of registers for the HLL sketch.
+/// 2^11 = 2048 registers, giving ~2.3% relative error and ~1KB per sketch (Hll4).
+const LG_K: u8 = 11;
 
 /// # Cardinality
 ///
 /// The cardinality aggregation allows for computing an estimate
 /// of the number of different values in a data set based on the
-/// HyperLogLog++ algorithm. This is particularly useful for understanding the
-/// uniqueness of values in a large dataset where counting each unique value
-/// individually would be computationally expensive.
+/// Apache DataSketches HyperLogLog algorithm. This is particularly useful for
+/// understanding the uniqueness of values in a large dataset where counting
+/// each unique value individually would be computationally expensive.
 ///
 /// For example, you might use a cardinality aggregation to estimate the number
 /// of unique visitors to a website by aggregating on a field that contains
@@ -106,8 +93,6 @@ pub struct CardinalityAggReqData {
     pub str_dict_column: Option<StrColumn>,
     /// The missing value normalized to the internal u64 representation of the field type.
     pub missing_value_for_accessor: Option<u64>,
-    /// The column block accessor to access the fast field values.
-    pub(crate) column_block_accessor: ColumnBlockAccessor<u64>,
     /// The name of the aggregation.
     pub name: String,
     /// The aggregation request.
@@ -135,45 +120,34 @@ impl CardinalityAggregationReq {
     }
 }
 
-#[derive(Clone, Debug, PartialEq)]
+#[derive(Clone, Debug)]
 pub(crate) struct SegmentCardinalityCollector {
-    cardinality: CardinalityCollector,
-    entries: FxHashSet<u64>,
+    buckets: Vec<SegmentCardinalityCollectorBucket>,
     accessor_idx: usize,
+    /// The column accessor to access the fast field values.
+    accessor: Column<u64>,
+    /// The column_type of the field.
+    column_type: ColumnType,
+    /// The missing value normalized to the internal u64 representation of the field type.
+    missing_value_for_accessor: Option<u64>,
 }
 
-impl SegmentCardinalityCollector {
-    pub fn from_req(column_type: ColumnType, accessor_idx: usize) -> Self {
+#[derive(Clone, Debug, PartialEq, Default)]
+pub(crate) struct SegmentCardinalityCollectorBucket {
+    cardinality: CardinalityCollector,
+    entries: FxHashSet<u64>,
+}
+impl SegmentCardinalityCollectorBucket {
+    pub fn new(column_type: ColumnType) -> Self {
         Self {
             cardinality: CardinalityCollector::new(column_type as u8),
-            entries: Default::default(),
-            accessor_idx,
+            entries: FxHashSet::default(),
         }
     }
-
-    fn fetch_block_with_field(
-        &mut self,
-        docs: &[crate::DocId],
-        agg_data: &mut CardinalityAggReqData,
-    ) {
-        if let Some(missing) = agg_data.missing_value_for_accessor {
-            agg_data.column_block_accessor.fetch_block_with_missing(
-                docs,
-                &agg_data.accessor,
-                missing,
-            );
-        } else {
-            agg_data
-                .column_block_accessor
-                .fetch_block(docs, &agg_data.accessor);
-        }
-    }
-
     fn into_intermediate_metric_result(
         mut self,
-        agg_data: &AggregationsSegmentCtx,
+        req_data: &CardinalityAggReqData,
     ) -> crate::Result<IntermediateMetricResult> {
-        let req_data = &agg_data.get_cardinality_req_data(self.accessor_idx);
         if req_data.column_type == ColumnType::Str {
             let fallback_dict = Dictionary::empty();
             let dict = req_data
@@ -194,9 +168,10 @@ impl SegmentCardinalityCollector {
                     term_ids.push(term_ord as u32);
                 }
             }
+
             term_ids.sort_unstable();
             dict.sorted_ords_to_term_cb(term_ids.iter().map(|term| *term as u64), |term| {
-                self.cardinality.sketch.insert_any(&term);
+                self.cardinality.insert(term);
                 Ok(())
             })?;
             if has_missing {
@@ -207,17 +182,17 @@ impl SegmentCardinalityCollector {
                     );
                 match missing_key {
                     Key::Str(missing) => {
-                        self.cardinality.sketch.insert_any(&missing);
+                        self.cardinality.insert(missing.as_str());
                     }
                     Key::F64(val) => {
                         let val = f64_to_u64(*val);
-                        self.cardinality.sketch.insert_any(&val);
+                        self.cardinality.insert(val);
                     }
                     Key::U64(val) => {
-                        self.cardinality.sketch.insert_any(&val);
+                        self.cardinality.insert(*val);
                     }
                     Key::I64(val) => {
-                        self.cardinality.sketch.insert_any(&val);
+                        self.cardinality.insert(*val);
                     }
                 }
             }
@@ -227,16 +202,49 @@ impl SegmentCardinalityCollector {
     }
 }
 
+impl SegmentCardinalityCollector {
+    pub fn from_req(
+        column_type: ColumnType,
+        accessor_idx: usize,
+        accessor: Column<u64>,
+        missing_value_for_accessor: Option<u64>,
+    ) -> Self {
+        Self {
+            buckets: vec![SegmentCardinalityCollectorBucket::new(column_type); 1],
+            column_type,
+            accessor_idx,
+            accessor,
+            missing_value_for_accessor,
+        }
+    }
+
+    fn fetch_block_with_field(
+        &mut self,
+        docs: &[crate::DocId],
+        agg_data: &mut AggregationsSegmentCtx,
+    ) {
+        agg_data.column_block_accessor.fetch_block_with_missing(
+            docs,
+            &self.accessor,
+            self.missing_value_for_accessor,
+        );
+    }
+}
+
 impl SegmentAggregationCollector for SegmentCardinalityCollector {
     fn add_intermediate_aggregation_result(
-        self: Box<Self>,
+        &mut self,
         agg_data: &AggregationsSegmentCtx,
         results: &mut IntermediateAggregationResults,
+        parent_bucket_id: BucketId,
     ) -> crate::Result<()> {
+        self.prepare_max_bucket(parent_bucket_id, agg_data)?;
         let req_data = &agg_data.get_cardinality_req_data(self.accessor_idx);
         let name = req_data.name.to_string();
+        // take the bucket in buckets and replace it with a new empty one
+        let bucket = std::mem::take(&mut self.buckets[parent_bucket_id as usize]);
 
-        let intermediate_result = self.into_intermediate_metric_result(agg_data)?;
+        let intermediate_result = bucket.into_intermediate_metric_result(req_data)?;
         results.push(
             name,
             IntermediateAggregationResult::Metric(intermediate_result),
@@ -247,27 +255,20 @@ impl SegmentAggregationCollector for SegmentCardinalityCollector {
 
     fn collect(
         &mut self,
-        doc: crate::DocId,
-        agg_data: &mut AggregationsSegmentCtx,
-    ) -> crate::Result<()> {
-        self.collect_block(&[doc], agg_data)
-    }
-
-    fn collect_block(
-        &mut self,
+        parent_bucket_id: BucketId,
         docs: &[crate::DocId],
         agg_data: &mut AggregationsSegmentCtx,
     ) -> crate::Result<()> {
-        let req_data = agg_data.get_cardinality_req_data_mut(self.accessor_idx);
-        self.fetch_block_with_field(docs, req_data);
+        self.fetch_block_with_field(docs, agg_data);
+        let bucket = &mut self.buckets[parent_bucket_id as usize];
 
-        let col_block_accessor = &req_data.column_block_accessor;
-        if req_data.column_type == ColumnType::Str {
+        let col_block_accessor = &agg_data.column_block_accessor;
+        if self.column_type == ColumnType::Str {
             for term_ord in col_block_accessor.iter_vals() {
-                self.entries.insert(term_ord);
+                bucket.entries.insert(term_ord);
             }
-        } else if req_data.column_type == ColumnType::IpAddr {
-            let compact_space_accessor = req_data
+        } else if self.column_type == ColumnType::IpAddr {
+            let compact_space_accessor = self
                 .accessor
                 .values
                 .clone()
@@ -282,23 +283,43 @@ impl SegmentAggregationCollector for SegmentCardinalityCollector {
                 })?;
             for val in col_block_accessor.iter_vals() {
                 let val: u128 = compact_space_accessor.compact_to_u128(val as u32);
-                self.cardinality.sketch.insert_any(&val);
+                bucket.cardinality.insert(val);
             }
         } else {
             for val in col_block_accessor.iter_vals() {
-                self.cardinality.sketch.insert_any(&val);
+                bucket.cardinality.insert(val);
             }
         }
 
         Ok(())
     }
+
+    fn prepare_max_bucket(
+        &mut self,
+        max_bucket: BucketId,
+        _agg_data: &AggregationsSegmentCtx,
+    ) -> crate::Result<()> {
+        if max_bucket as usize >= self.buckets.len() {
+            self.buckets.resize_with(max_bucket as usize + 1, || {
+                SegmentCardinalityCollectorBucket::new(self.column_type)
+            });
+        }
+        Ok(())
+    }
 }
 
-#[derive(Clone, Debug, Serialize, Deserialize)]
-/// The percentiles collector used during segment collection and for merging results.
+#[derive(Clone, Debug)]
+/// The cardinality collector used during segment collection and for merging results.
+/// Uses Apache DataSketches HLL (lg_k=11, Hll4) for compact binary serialization
+/// and cross-language compatibility (e.g. Java `datasketches` library).
 pub struct CardinalityCollector {
-    sketch: HyperLogLogPlus<u64, BuildSaltedHasher>,
+    sketch: HllSketch,
+    /// Salt derived from `ColumnType`, used to differentiate values of different column types
+    /// that map to the same u64 (e.g. bool `false` = 0 vs i64 `0`).
+    /// Not serialized — only needed during insertion, not after sketch registers are populated.
+    salt: u8,
 }
+
 impl Default for CardinalityCollector {
     fn default() -> Self {
         Self::new(0)
@@ -311,25 +332,52 @@ impl PartialEq for CardinalityCollector {
     }
 }
 
-impl CardinalityCollector {
-    /// Compute the final cardinality estimate.
-    pub fn finalize(self) -> Option<f64> {
-        Some(self.sketch.clone().count().trunc())
+impl Serialize for CardinalityCollector {
+    fn serialize<S: Serializer>(&self, serializer: S) -> Result<S::Ok, S::Error> {
+        let bytes = self.sketch.serialize();
+        serializer.serialize_bytes(&bytes)
     }
+}
 
+impl<'de> Deserialize<'de> for CardinalityCollector {
+    fn deserialize<D: Deserializer<'de>>(deserializer: D) -> Result<Self, D::Error> {
+        let bytes: Vec<u8> = Deserialize::deserialize(deserializer)?;
+        let sketch = HllSketch::deserialize(&bytes).map_err(serde::de::Error::custom)?;
+        Ok(Self { sketch, salt: 0 })
+    }
+}
+
+impl CardinalityCollector {
     fn new(salt: u8) -> Self {
         Self {
-            sketch: HyperLogLogPlus::new(16, BuildSaltedHasher { salt }).unwrap(),
+            sketch: HllSketch::new(LG_K, HllType::Hll4),
+            salt,
         }
     }
 
-    pub(crate) fn merge_fruits(&mut self, right: CardinalityCollector) -> crate::Result<()> {
-        self.sketch.merge(&right.sketch).map_err(|err| {
-            TantivyError::AggregationError(AggregationError::InternalError(format!(
-                "Error while merging cardinality {err:?}"
-            )))
-        })?;
+    /// Insert a value into the HLL sketch, salted by the column type.
+    /// The salt ensures that identical u64 values from different column types
+    /// (e.g. bool `false` vs i64 `0`) are counted as distinct.
+    pub(crate) fn insert<T: Hash>(&mut self, value: T) {
+        self.sketch.update((self.salt, value));
+    }
 
+    /// Compute the final cardinality estimate.
+    pub fn finalize(self) -> Option<f64> {
+        Some(self.sketch.estimate().trunc())
+    }
+
+    /// Serialize the HLL sketch to its compact binary representation.
+    /// The format is cross-language compatible with Apache DataSketches (Java, C++, Python).
+    pub fn to_sketch_bytes(&self) -> Vec<u8> {
+        self.sketch.serialize()
+    }
+
+    pub(crate) fn merge_fruits(&mut self, right: CardinalityCollector) -> crate::Result<()> {
+        let mut union = HllUnion::new(LG_K);
+        union.update(&self.sketch);
+        union.update(&right.sketch);
+        self.sketch = union.get_result(HllType::Hll4);
         Ok(())
     }
 }
@@ -491,4 +539,75 @@ mod tests {
 
         Ok(())
     }
+
+    #[test]
+    fn cardinality_collector_serde_roundtrip() {
+        use super::CardinalityCollector;
+
+        let mut collector = CardinalityCollector::default();
+        collector.insert("hello");
+        collector.insert("world");
+        collector.insert("hello"); // duplicate
+
+        let serialized = serde_json::to_vec(&collector).unwrap();
+        let deserialized: CardinalityCollector = serde_json::from_slice(&serialized).unwrap();
+
+        let original_estimate = collector.finalize().unwrap();
+        let roundtrip_estimate = deserialized.finalize().unwrap();
+        assert_eq!(original_estimate, roundtrip_estimate);
+        assert_eq!(original_estimate, 2.0);
+    }
+
+    #[test]
+    fn cardinality_collector_merge() {
+        use super::CardinalityCollector;
+
+        let mut left = CardinalityCollector::default();
+        left.insert("a");
+        left.insert("b");
+
+        let mut right = CardinalityCollector::default();
+        right.insert("b");
+        right.insert("c");
+
+        left.merge_fruits(right).unwrap();
+        let estimate = left.finalize().unwrap();
+        assert_eq!(estimate, 3.0);
+    }
+
+    #[test]
+    fn cardinality_collector_serialize_deserialize_binary() {
+        use datasketches::hll::HllSketch;
+
+        use super::CardinalityCollector;
+
+        let mut collector = CardinalityCollector::default();
+        collector.insert("apple");
+        collector.insert("banana");
+        collector.insert("cherry");
+
+        let bytes = collector.to_sketch_bytes();
+        let deserialized = HllSketch::deserialize(&bytes).unwrap();
+        assert!((deserialized.estimate() - 3.0).abs() < 0.01);
+    }
+
+    #[test]
+    fn cardinality_collector_salt_differentiates_types() {
+        use super::CardinalityCollector;
+
+        // Without salt, same u64 value from different column types would collide
+        let mut collector_bool = CardinalityCollector::new(5); // e.g. ColumnType::Bool
+        collector_bool.insert(0u64); // false
+        collector_bool.insert(1u64); // true
+
+        let mut collector_i64 = CardinalityCollector::new(2); // e.g. ColumnType::I64
+        collector_i64.insert(0u64);
+        collector_i64.insert(1u64);
+
+        // Merge them
+        collector_bool.merge_fruits(collector_i64).unwrap();
+        let estimate = collector_bool.finalize().unwrap();
+        // Should be 4 because salt makes (5, 0) != (2, 0) and (5, 1) != (2, 1)
+        assert_eq!(estimate, 4.0);
+    }
 }

src/aggregation/metric/count.rs

@@ -52,10 +52,8 @@ pub struct IntermediateCount {
 
 impl IntermediateCount {
     /// Creates a new [`IntermediateCount`] instance from a [`SegmentStatsCollector`].
-    pub(crate) fn from_collector(collector: SegmentStatsCollector) -> Self {
-        Self {
-            stats: collector.stats,
-        }
+    pub(crate) fn from_stats(stats: IntermediateStats) -> Self {
+        Self { stats }
     }
     /// Merges the other intermediate result into self.
     pub fn merge_fruits(&mut self, other: IntermediateCount) {

src/aggregation/metric/extended_stats.rs

@@ -8,10 +8,9 @@ use crate::aggregation::agg_data::AggregationsSegmentCtx;
 use crate::aggregation::intermediate_agg_result::{
     IntermediateAggregationResult, IntermediateAggregationResults, IntermediateMetricResult,
 };
-use crate::aggregation::metric::MetricAggReqData;
 use crate::aggregation::segment_agg_result::SegmentAggregationCollector;
 use crate::aggregation::*;
-use crate::{DocId, TantivyError};
+use crate::TantivyError;
 
 /// A multi-value metric aggregation that computes a collection of extended statistics
 /// on numeric values that are extracted
@@ -62,7 +61,7 @@ impl ExtendedStatsAggregation {
 
 /// Extended stats contains a collection of statistics
 /// they extends stats adding variance, standard deviation
-/// and bound informations
+/// and bound information
 #[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
 pub struct ExtendedStats {
     /// The number of documents.
@@ -318,51 +317,28 @@ impl IntermediateExtendedStats {
     }
 }
 
-#[derive(Clone, Debug, PartialEq)]
+#[derive(Clone, Debug)]
 pub(crate) struct SegmentExtendedStatsCollector {
+    name: String,
     missing: Option<u64>,
     field_type: ColumnType,
-    pub(crate) extended_stats: IntermediateExtendedStats,
-    pub(crate) accessor_idx: usize,
-    val_cache: Vec<u64>,
+    accessor: columnar::Column<u64>,
+    buckets: Vec<IntermediateExtendedStats>,
+    sigma: Option<f64>,
 }
 
 impl SegmentExtendedStatsCollector {
-    pub fn from_req(
-        field_type: ColumnType,
-        sigma: Option<f64>,
-        accessor_idx: usize,
-        missing: Option<f64>,
-    ) -> Self {
-        let missing = missing.and_then(|val| f64_to_fastfield_u64(val, &field_type));
+    pub fn from_req(req: &MetricAggReqData, sigma: Option<f64>) -> Self {
+        let missing = req
+            .missing
+            .and_then(|val| f64_to_fastfield_u64(val, &req.field_type));
         Self {
-            field_type,
-            extended_stats: IntermediateExtendedStats::with_sigma(sigma),
-            accessor_idx,
+            name: req.name.clone(),
+            field_type: req.field_type,
+            accessor: req.accessor.clone(),
             missing,
-            val_cache: Default::default(),
-        }
-    }
-    #[inline]
-    pub(crate) fn collect_block_with_field(
-        &mut self,
-        docs: &[DocId],
-        req_data: &mut MetricAggReqData,
-    ) {
-        if let Some(missing) = self.missing.as_ref() {
-            req_data.column_block_accessor.fetch_block_with_missing(
-                docs,
-                &req_data.accessor,
-                *missing,
-            );
-        } else {
-            req_data
-                .column_block_accessor
-                .fetch_block(docs, &req_data.accessor);
-        }
-        for val in req_data.column_block_accessor.iter_vals() {
-            let val1 = f64_from_fastfield_u64(val, &self.field_type);
-            self.extended_stats.collect(val1);
+            buckets: vec![IntermediateExtendedStats::with_sigma(sigma); 16],
+            sigma,
         }
     }
 }
@@ -370,15 +346,18 @@ impl SegmentExtendedStatsCollector {
 impl SegmentAggregationCollector for SegmentExtendedStatsCollector {
     #[inline]
     fn add_intermediate_aggregation_result(
-        self: Box<Self>,
+        &mut self,
         agg_data: &AggregationsSegmentCtx,
         results: &mut IntermediateAggregationResults,
+        parent_bucket_id: BucketId,
     ) -> crate::Result<()> {
-        let name = agg_data.get_metric_req_data(self.accessor_idx).name.clone();
+        let name = self.name.clone();
+        self.prepare_max_bucket(parent_bucket_id, agg_data)?;
+        let extended_stats = std::mem::take(&mut self.buckets[parent_bucket_id as usize]);
         results.push(
             name,
             IntermediateAggregationResult::Metric(IntermediateMetricResult::ExtendedStats(
-                self.extended_stats,
+                extended_stats,
             )),
         )?;
 
@@ -388,39 +367,36 @@ impl SegmentAggregationCollector for SegmentExtendedStatsCollector {
     #[inline]
     fn collect(
         &mut self,
-        doc: crate::DocId,
+        parent_bucket_id: BucketId,
+        docs: &[crate::DocId],
         agg_data: &mut AggregationsSegmentCtx,
     ) -> crate::Result<()> {
-        let req_data = agg_data.get_metric_req_data(self.accessor_idx);
-        if let Some(missing) = self.missing {
-            let mut has_val = false;
-            for val in req_data.accessor.values_for_doc(doc) {
-                let val1 = f64_from_fastfield_u64(val, &self.field_type);
-                self.extended_stats.collect(val1);
-                has_val = true;
-            }
-            if !has_val {
-                self.extended_stats
-                    .collect(f64_from_fastfield_u64(missing, &self.field_type));
-            }
-        } else {
-            for val in req_data.accessor.values_for_doc(doc) {
-                let val1 = f64_from_fastfield_u64(val, &self.field_type);
-                self.extended_stats.collect(val1);
-            }
+        let mut extended_stats = self.buckets[parent_bucket_id as usize].clone();
+
+        agg_data
+            .column_block_accessor
+            .fetch_block_with_missing(docs, &self.accessor, self.missing);
+        for val in agg_data.column_block_accessor.iter_vals() {
+            let val1 = f64_from_fastfield_u64(val, self.field_type);
+            extended_stats.collect(val1);
         }
 
+        // store back
+        self.buckets[parent_bucket_id as usize] = extended_stats;
+
         Ok(())
     }
 
-    #[inline]
-    fn collect_block(
+    fn prepare_max_bucket(
         &mut self,
-        docs: &[crate::DocId],
-        agg_data: &mut AggregationsSegmentCtx,
+        max_bucket: BucketId,
+        _agg_data: &AggregationsSegmentCtx,
     ) -> crate::Result<()> {
-        let req_data = agg_data.get_metric_req_data_mut(self.accessor_idx);
-        self.collect_block_with_field(docs, req_data);
+        if self.buckets.len() <= max_bucket as usize {
+            self.buckets.resize_with(max_bucket as usize + 1, || {
+                IntermediateExtendedStats::with_sigma(self.sigma)
+            });
+        }
         Ok(())
     }
 }

src/aggregation/metric/max.rs

@@ -52,10 +52,8 @@ pub struct IntermediateMax {
 
 impl IntermediateMax {
     /// Creates a new [`IntermediateMax`] instance from a [`SegmentStatsCollector`].
-    pub(crate) fn from_collector(collector: SegmentStatsCollector) -> Self {
-        Self {
-            stats: collector.stats,
-        }
+    pub(crate) fn from_stats(stats: IntermediateStats) -> Self {
+        Self { stats }
     }
     /// Merges the other intermediate result into self.
     pub fn merge_fruits(&mut self, other: IntermediateMax) {

src/aggregation/metric/min.rs

@@ -52,10 +52,8 @@ pub struct IntermediateMin {
 
 impl IntermediateMin {
     /// Creates a new [`IntermediateMin`] instance from a [`SegmentStatsCollector`].
-    pub(crate) fn from_collector(collector: SegmentStatsCollector) -> Self {
-        Self {
-            stats: collector.stats,
-        }
+    pub(crate) fn from_stats(stats: IntermediateStats) -> Self {
+        Self { stats }
     }
     /// Merges the other intermediate result into self.
     pub fn merge_fruits(&mut self, other: IntermediateMin) {

src/aggregation/metric/mod.rs

@@ -31,7 +31,7 @@ use std::collections::HashMap;
 
 pub use average::*;
 pub use cardinality::*;
-use columnar::{Column, ColumnBlockAccessor, ColumnType};
+use columnar::{Column, ColumnType};
 pub use count::*;
 pub use extended_stats::*;
 pub use max::*;
@@ -55,8 +55,6 @@ pub struct MetricAggReqData {
     pub field_type: ColumnType,
     /// The missing value normalized to the internal u64 representation of the field type.
     pub missing_u64: Option<u64>,
-    /// The column block accessor to access the fast field values.
-    pub column_block_accessor: ColumnBlockAccessor<u64>,
     /// The column accessor to access the fast field values.
     pub accessor: Column<u64>,
     /// Used when converting to intermediate result
@@ -109,8 +107,11 @@ pub enum PercentileValues {
 #[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
 /// The entry when requesting percentiles with keyed: false
 pub struct PercentileValuesVecEntry {
-    key: f64,
-    value: f64,
+    /// Percentile
+    pub key: f64,
+
+    /// Value at the percentile
+    pub value: f64,
 }
 
 /// Single-metric aggregations use this common result structure.

src/aggregation/metric/percentiles.rs

@@ -7,10 +7,9 @@ use crate::aggregation::agg_data::AggregationsSegmentCtx;
 use crate::aggregation::intermediate_agg_result::{
     IntermediateAggregationResult, IntermediateAggregationResults, IntermediateMetricResult,
 };
-use crate::aggregation::metric::MetricAggReqData;
 use crate::aggregation::segment_agg_result::SegmentAggregationCollector;
 use crate::aggregation::*;
-use crate::{DocId, TantivyError};
+use crate::TantivyError;
 
 /// # Percentiles
 ///
@@ -131,10 +130,16 @@ impl PercentilesAggregationReq {
     }
 }
 
-#[derive(Clone, Debug, PartialEq)]
+#[derive(Clone, Debug)]
 pub(crate) struct SegmentPercentilesCollector {
-    pub(crate) percentiles: PercentilesCollector,
+    pub(crate) buckets: Vec<PercentilesCollector>,
     pub(crate) accessor_idx: usize,
+    /// The type of the field.
+    pub field_type: ColumnType,
+    /// The missing value normalized to the internal u64 representation of the field type.
+    pub missing_u64: Option<u64>,
+    /// The column accessor to access the fast field values.
+    pub accessor: Column<u64>,
 }
 
 #[derive(Clone, Serialize, Deserialize)]
@@ -217,6 +222,12 @@ impl PercentilesCollector {
         self.sketch.add(val);
     }
 
+    /// Encode the underlying DDSketch to Java-compatible binary format
+    /// for cross-language serialization with Java consumers.
+    pub fn to_sketch_bytes(&self) -> Vec<u8> {
+        self.sketch.to_java_bytes()
+    }
+
     pub(crate) fn merge_fruits(&mut self, right: PercentilesCollector) -> crate::Result<()> {
         self.sketch.merge(&right.sketch).map_err(|err| {
             TantivyError::AggregationError(AggregationError::InternalError(format!(
@@ -229,33 +240,18 @@ impl PercentilesCollector {
 }
 
 impl SegmentPercentilesCollector {
-    pub fn from_req_and_validate(accessor_idx: usize) -> crate::Result<Self> {
-        Ok(Self {
-            percentiles: PercentilesCollector::new(),
+    pub fn from_req_and_validate(
+        field_type: ColumnType,
+        missing_u64: Option<u64>,
+        accessor: Column<u64>,
+        accessor_idx: usize,
+    ) -> Self {
+        Self {
+            buckets: Vec::with_capacity(64),
+            field_type,
+            missing_u64,
+            accessor,
             accessor_idx,
-        })
-    }
-    #[inline]
-    pub(crate) fn collect_block_with_field(
-        &mut self,
-        docs: &[DocId],
-        req_data: &mut MetricAggReqData,
-    ) {
-        if let Some(missing) = req_data.missing_u64.as_ref() {
-            req_data.column_block_accessor.fetch_block_with_missing(
-                docs,
-                &req_data.accessor,
-                *missing,
-            );
-        } else {
-            req_data
-                .column_block_accessor
-                .fetch_block(docs, &req_data.accessor);
-        }
-
-        for val in req_data.column_block_accessor.iter_vals() {
-            let val1 = f64_from_fastfield_u64(val, &req_data.field_type);
-            self.percentiles.collect(val1);
         }
     }
 }
@@ -263,12 +259,18 @@ impl SegmentPercentilesCollector {
 impl SegmentAggregationCollector for SegmentPercentilesCollector {
     #[inline]
     fn add_intermediate_aggregation_result(
-        self: Box<Self>,
+        &mut self,
         agg_data: &AggregationsSegmentCtx,
         results: &mut IntermediateAggregationResults,
+        parent_bucket_id: BucketId,
     ) -> crate::Result<()> {
         let name = agg_data.get_metric_req_data(self.accessor_idx).name.clone();
-        let intermediate_metric_result = IntermediateMetricResult::Percentiles(self.percentiles);
+        self.prepare_max_bucket(parent_bucket_id, agg_data)?;
+        // Swap collector with an empty one to avoid cloning
+        let percentiles_collector = std::mem::take(&mut self.buckets[parent_bucket_id as usize]);
+
+        let intermediate_metric_result =
+            IntermediateMetricResult::Percentiles(percentiles_collector);
 
         results.push(
             name,
@@ -281,40 +283,33 @@ impl SegmentAggregationCollector for SegmentPercentilesCollector {
     #[inline]
     fn collect(
         &mut self,
-        doc: crate::DocId,
+        parent_bucket_id: BucketId,
+        docs: &[crate::DocId],
         agg_data: &mut AggregationsSegmentCtx,
     ) -> crate::Result<()> {
-        let req_data = agg_data.get_metric_req_data(self.accessor_idx);
+        let percentiles = &mut self.buckets[parent_bucket_id as usize];
+        agg_data.column_block_accessor.fetch_block_with_missing(
+            docs,
+            &self.accessor,
+            self.missing_u64,
+        );
 
-        if let Some(missing) = req_data.missing_u64 {
-            let mut has_val = false;
-            for val in req_data.accessor.values_for_doc(doc) {
-                let val1 = f64_from_fastfield_u64(val, &req_data.field_type);
-                self.percentiles.collect(val1);
-                has_val = true;
-            }
-            if !has_val {
-                self.percentiles
-                    .collect(f64_from_fastfield_u64(missing, &req_data.field_type));
-            }
-        } else {
-            for val in req_data.accessor.values_for_doc(doc) {
-                let val1 = f64_from_fastfield_u64(val, &req_data.field_type);
-                self.percentiles.collect(val1);
-            }
+        for val in agg_data.column_block_accessor.iter_vals() {
+            let val1 = f64_from_fastfield_u64(val, self.field_type);
+            percentiles.collect(val1);
         }
 
         Ok(())
     }
 
-    #[inline]
-    fn collect_block(
+    fn prepare_max_bucket(
         &mut self,
-        docs: &[crate::DocId],
-        agg_data: &mut AggregationsSegmentCtx,
+        max_bucket: BucketId,
+        _agg_data: &AggregationsSegmentCtx,
     ) -> crate::Result<()> {
-        let req_data = agg_data.get_metric_req_data_mut(self.accessor_idx);
-        self.collect_block_with_field(docs, req_data);
+        while self.buckets.len() <= max_bucket as usize {
+            self.buckets.push(PercentilesCollector::new());
+        }
         Ok(())
     }
 }
@@ -336,7 +331,7 @@ mod tests {
     use crate::aggregation::AggregationCollector;
     use crate::query::AllQuery;
     use crate::schema::{Schema, FAST};
-    use crate::Index;
+    use crate::{assert_nearly_equals, Index};
 
     #[test]
     fn test_aggregation_percentiles_empty_index() -> crate::Result<()> {
@@ -619,12 +614,16 @@ mod tests {
         let res = exec_request_with_query(agg_req, &index, None)?;
         assert_eq!(res["range_with_stats"]["buckets"][0]["doc_count"], 3);
 
-        assert_eq!(
-            res["range_with_stats"]["buckets"][0]["percentiles"]["values"]["1.0"],
+        assert_nearly_equals!(
+            res["range_with_stats"]["buckets"][0]["percentiles"]["values"]["1.0"]
+                .as_f64()
+                .unwrap(),
             5.0028295751107414
         );
-        assert_eq!(
-            res["range_with_stats"]["buckets"][0]["percentiles"]["values"]["99.0"],
+        assert_nearly_equals!(
+            res["range_with_stats"]["buckets"][0]["percentiles"]["values"]["99.0"]
+                .as_f64()
+                .unwrap(),
             10.07469668951144
         );
 
@@ -670,8 +669,14 @@ mod tests {
 
         let res = exec_request_with_query(agg_req, &index, None)?;
 
-        assert_eq!(res["percentiles"]["values"]["1.0"], 5.0028295751107414);
-        assert_eq!(res["percentiles"]["values"]["99.0"], 10.07469668951144);
+        assert_nearly_equals!(
+            res["percentiles"]["values"]["1.0"].as_f64().unwrap(),
+            5.0028295751107414
+        );
+        assert_nearly_equals!(
+            res["percentiles"]["values"]["99.0"].as_f64().unwrap(),
+            10.07469668951144
+        );
 
         Ok(())
     }

src/aggregation/metric/stats.rs

@@ -1,5 +1,6 @@
 use std::fmt::Debug;
 
+use columnar::{Column, ColumnType};
 use serde::{Deserialize, Serialize};
 
 use super::*;
@@ -7,10 +8,9 @@ use crate::aggregation::agg_data::AggregationsSegmentCtx;
 use crate::aggregation::intermediate_agg_result::{
     IntermediateAggregationResult, IntermediateAggregationResults, IntermediateMetricResult,
 };
-use crate::aggregation::metric::MetricAggReqData;
 use crate::aggregation::segment_agg_result::SegmentAggregationCollector;
 use crate::aggregation::*;
-use crate::{DocId, TantivyError};
+use crate::TantivyError;
 
 /// A multi-value metric aggregation that computes a collection of statistics on numeric values that
 /// are extracted from the aggregated documents.
@@ -83,7 +83,7 @@ impl Stats {
 
 /// Intermediate result of the stats aggregation that can be combined with other intermediate
 /// results.
-#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
+#[derive(Clone, Copy, Debug, PartialEq, Serialize, Deserialize)]
 pub struct IntermediateStats {
     /// The number of extracted values.
     pub(crate) count: u64,
@@ -110,6 +110,16 @@ impl Default for IntermediateStats {
 }
 
 impl IntermediateStats {
+    /// Returns the number of values collected.
+    pub fn count(&self) -> u64 {
+        self.count
+    }
+
+    /// Returns the sum of all values collected.
+    pub fn sum(&self) -> f64 {
+        self.sum
+    }
+
     /// Merges the other stats intermediate result into self.
     pub fn merge_fruits(&mut self, other: IntermediateStats) {
         self.count += other.count;
@@ -187,75 +197,75 @@ pub enum StatsType {
     Percentiles,
 }
 
+fn create_collector<const TYPE_ID: u8>(
+    req: &MetricAggReqData,
+) -> Box<dyn SegmentAggregationCollector> {
+    Box::new(SegmentStatsCollector::<TYPE_ID> {
+        name: req.name.clone(),
+        collecting_for: req.collecting_for,
+        is_number_or_date_type: req.is_number_or_date_type,
+        missing_u64: req.missing_u64,
+        accessor: req.accessor.clone(),
+        buckets: vec![IntermediateStats::default()],
+    })
+}
+
+/// Build a concrete `SegmentStatsCollector` depending on the column type.
+pub(crate) fn build_segment_stats_collector(
+    req: &MetricAggReqData,
+) -> crate::Result<Box<dyn SegmentAggregationCollector>> {
+    match req.field_type {
+        ColumnType::I64 => Ok(create_collector::<{ ColumnType::I64 as u8 }>(req)),
+        ColumnType::U64 => Ok(create_collector::<{ ColumnType::U64 as u8 }>(req)),
+        ColumnType::F64 => Ok(create_collector::<{ ColumnType::F64 as u8 }>(req)),
+        ColumnType::Bool => Ok(create_collector::<{ ColumnType::Bool as u8 }>(req)),
+        ColumnType::DateTime => Ok(create_collector::<{ ColumnType::DateTime as u8 }>(req)),
+        ColumnType::Bytes => Ok(create_collector::<{ ColumnType::Bytes as u8 }>(req)),
+        ColumnType::Str => Ok(create_collector::<{ ColumnType::Str as u8 }>(req)),
+        ColumnType::IpAddr => Ok(create_collector::<{ ColumnType::IpAddr as u8 }>(req)),
+    }
+}
+
+#[repr(C)]
 #[derive(Clone, Debug)]
-pub(crate) struct SegmentStatsCollector {
-    pub(crate) stats: IntermediateStats,
-    pub(crate) accessor_idx: usize,
+pub(crate) struct SegmentStatsCollector<const COLUMN_TYPE_ID: u8> {
+    pub(crate) missing_u64: Option<u64>,
+    pub(crate) accessor: Column<u64>,
+    pub(crate) is_number_or_date_type: bool,
+    pub(crate) buckets: Vec<IntermediateStats>,
+    pub(crate) name: String,
+    pub(crate) collecting_for: StatsType,
 }
 
-impl SegmentStatsCollector {
-    pub fn from_req(accessor_idx: usize) -> Self {
-        Self {
-            stats: IntermediateStats::default(),
-            accessor_idx,
-        }
-    }
-    #[inline]
-    pub(crate) fn collect_block_with_field(
-        &mut self,
-        docs: &[DocId],
-        req_data: &mut MetricAggReqData,
-    ) {
-        if let Some(missing) = req_data.missing_u64.as_ref() {
-            req_data.column_block_accessor.fetch_block_with_missing(
-                docs,
-                &req_data.accessor,
-                *missing,
-            );
-        } else {
-            req_data
-                .column_block_accessor
-                .fetch_block(docs, &req_data.accessor);
-        }
-        if req_data.is_number_or_date_type {
-            for val in req_data.column_block_accessor.iter_vals() {
-                let val1 = f64_from_fastfield_u64(val, &req_data.field_type);
-                self.stats.collect(val1);
-            }
-        } else {
-            for _val in req_data.column_block_accessor.iter_vals() {
-                // we ignore the value and simply record that we got something
-                self.stats.collect(0.0);
-            }
-        }
-    }
-}
-
-impl SegmentAggregationCollector for SegmentStatsCollector {
+impl<const COLUMN_TYPE_ID: u8> SegmentAggregationCollector
+    for SegmentStatsCollector<COLUMN_TYPE_ID>
+{
     #[inline]
     fn add_intermediate_aggregation_result(
-        self: Box<Self>,
+        &mut self,
         agg_data: &AggregationsSegmentCtx,
         results: &mut IntermediateAggregationResults,
+        parent_bucket_id: BucketId,
     ) -> crate::Result<()> {
-        let req = agg_data.get_metric_req_data(self.accessor_idx);
-        let name = req.name.clone();
+        let name = self.name.clone();
 
-        let intermediate_metric_result = match req.collecting_for {
+        self.prepare_max_bucket(parent_bucket_id, agg_data)?;
+        let stats = self.buckets[parent_bucket_id as usize];
+        let intermediate_metric_result = match self.collecting_for {
             StatsType::Average => {
-                IntermediateMetricResult::Average(IntermediateAverage::from_collector(*self))
+                IntermediateMetricResult::Average(IntermediateAverage::from_stats(stats))
             }
             StatsType::Count => {
-                IntermediateMetricResult::Count(IntermediateCount::from_collector(*self))
+                IntermediateMetricResult::Count(IntermediateCount::from_stats(stats))
             }
-            StatsType::Max => IntermediateMetricResult::Max(IntermediateMax::from_collector(*self)),
-            StatsType::Min => IntermediateMetricResult::Min(IntermediateMin::from_collector(*self)),
-            StatsType::Stats => IntermediateMetricResult::Stats(self.stats),
-            StatsType::Sum => IntermediateMetricResult::Sum(IntermediateSum::from_collector(*self)),
+            StatsType::Max => IntermediateMetricResult::Max(IntermediateMax::from_stats(stats)),
+            StatsType::Min => IntermediateMetricResult::Min(IntermediateMin::from_stats(stats)),
+            StatsType::Stats => IntermediateMetricResult::Stats(stats),
+            StatsType::Sum => IntermediateMetricResult::Sum(IntermediateSum::from_stats(stats)),
             _ => {
                 return Err(TantivyError::InvalidArgument(format!(
                     "Unsupported stats type for stats aggregation: {:?}",
-                    req.collecting_for
+                    self.collecting_for
                 )))
             }
         };
@@ -271,41 +281,67 @@ impl SegmentAggregationCollector for SegmentStatsCollector {
     #[inline]
     fn collect(
         &mut self,
-        doc: crate::DocId,
-        agg_data: &mut AggregationsSegmentCtx,
-    ) -> crate::Result<()> {
-        let req_data = agg_data.get_metric_req_data(self.accessor_idx);
-        if let Some(missing) = req_data.missing_u64 {
-            let mut has_val = false;
-            for val in req_data.accessor.values_for_doc(doc) {
-                let val1 = f64_from_fastfield_u64(val, &req_data.field_type);
-                self.stats.collect(val1);
-                has_val = true;
-            }
-            if !has_val {
-                self.stats
-                    .collect(f64_from_fastfield_u64(missing, &req_data.field_type));
-            }
-        } else {
-            for val in req_data.accessor.values_for_doc(doc) {
-                let val1 = f64_from_fastfield_u64(val, &req_data.field_type);
-                self.stats.collect(val1);
-            }
-        }
-
-        Ok(())
-    }
-
-    #[inline]
-    fn collect_block(
-        &mut self,
+        parent_bucket_id: BucketId,
         docs: &[crate::DocId],
         agg_data: &mut AggregationsSegmentCtx,
     ) -> crate::Result<()> {
-        let req_data = agg_data.get_metric_req_data_mut(self.accessor_idx);
-        self.collect_block_with_field(docs, req_data);
+        // TODO: remove once we fetch all values for all bucket ids in one go
+        if docs.len() == 1 && self.missing_u64.is_none() {
+            collect_stats::<COLUMN_TYPE_ID>(
+                &mut self.buckets[parent_bucket_id as usize],
+                self.accessor.values_for_doc(docs[0]),
+                self.is_number_or_date_type,
+            )?;
+
+            return Ok(());
+        }
+        agg_data.column_block_accessor.fetch_block_with_missing(
+            docs,
+            &self.accessor,
+            self.missing_u64,
+        );
+        collect_stats::<COLUMN_TYPE_ID>(
+            &mut self.buckets[parent_bucket_id as usize],
+            agg_data.column_block_accessor.iter_vals(),
+            self.is_number_or_date_type,
+        )?;
+
         Ok(())
     }
+
+    fn prepare_max_bucket(
+        &mut self,
+        max_bucket: BucketId,
+        _agg_data: &AggregationsSegmentCtx,
+    ) -> crate::Result<()> {
+        let required_buckets = (max_bucket as usize) + 1;
+        if self.buckets.len() < required_buckets {
+            self.buckets
+                .resize_with(required_buckets, IntermediateStats::default);
+        }
+        Ok(())
+    }
+}
+
+#[inline]
+fn collect_stats<const COLUMN_TYPE_ID: u8>(
+    stats: &mut IntermediateStats,
+    vals: impl Iterator<Item = u64>,
+    is_number_or_date_type: bool,
+) -> crate::Result<()> {
+    if is_number_or_date_type {
+        for val in vals {
+            let val1 = convert_to_f64::<COLUMN_TYPE_ID>(val);
+            stats.collect(val1);
+        }
+    } else {
+        for _val in vals {
+            // we ignore the value and simply record that we got something
+            stats.collect(0.0);
+        }
+    }
+
+    Ok(())
 }
 
 #[cfg(test)]

src/aggregation/metric/sum.rs

@@ -52,10 +52,8 @@ pub struct IntermediateSum {
 
 impl IntermediateSum {
     /// Creates a new [`IntermediateSum`] instance from a [`SegmentStatsCollector`].
-    pub(crate) fn from_collector(collector: SegmentStatsCollector) -> Self {
-        Self {
-            stats: collector.stats,
-        }
+    pub(crate) fn from_stats(stats: IntermediateStats) -> Self {
+        Self { stats }
     }
     /// Merges the other intermediate result into self.
     pub fn merge_fruits(&mut self, other: IntermediateSum) {

src/aggregation/metric/top_hits.rs

@@ -15,11 +15,11 @@ use crate::aggregation::intermediate_agg_result::{
     IntermediateAggregationResult, IntermediateMetricResult,
 };
 use crate::aggregation::segment_agg_result::SegmentAggregationCollector;
-use crate::aggregation::AggregationError;
+use crate::aggregation::{AggregationError, BucketId};
+use crate::collector::sort_key::ReverseComparator;
 use crate::collector::TopNComputer;
 use crate::schema::OwnedValue;
 use crate::{DocAddress, DocId, SegmentOrdinal};
-// duplicate import removed; already imported above
 
 /// Contains all information required by the TopHitsSegmentCollector to perform the
 /// top_hits aggregation on a segment.
@@ -458,7 +458,7 @@ impl Eq for DocSortValuesAndFields {}
 #[derive(Clone, Serialize, Deserialize, Debug)]
 pub struct TopHitsTopNComputer {
     req: TopHitsAggregationReq,
-    top_n: TopNComputer<DocSortValuesAndFields, DocAddress, false>,
+    top_n: TopNComputer<DocSortValuesAndFields, DocAddress, ReverseComparator>,
 }
 
 impl std::cmp::PartialEq for TopHitsTopNComputer {
@@ -471,7 +471,10 @@ impl TopHitsTopNComputer {
     /// Create a new TopHitsCollector
     pub fn new(req: &TopHitsAggregationReq) -> Self {
         Self {
-            top_n: TopNComputer::new(req.size + req.from.unwrap_or(0)),
+            top_n: TopNComputer::new_with_comparator(
+                req.size + req.from.unwrap_or(0),
+                ReverseComparator,
+            ),
             req: req.clone(),
         }
     }
@@ -482,7 +485,7 @@ impl TopHitsTopNComputer {
 
     pub(crate) fn merge_fruits(&mut self, other_fruit: Self) -> crate::Result<()> {
         for doc in other_fruit.top_n.into_vec() {
-            self.collect(doc.feature, doc.doc);
+            self.collect(doc.sort_key, doc.doc);
         }
         Ok(())
     }
@@ -494,9 +497,9 @@ impl TopHitsTopNComputer {
             .into_sorted_vec()
             .into_iter()
             .map(|doc| TopHitsVecEntry {
-                sort: doc.feature.sorts.iter().map(|f| f.value).collect(),
+                sort: doc.sort_key.sorts.iter().map(|f| f.value).collect(),
                 doc_value_fields: doc
-                    .feature
+                    .sort_key
                     .doc_value_fields
                     .into_iter()
                     .map(|(k, v)| (k, v.into()))
@@ -517,7 +520,8 @@ impl TopHitsTopNComputer {
 pub(crate) struct TopHitsSegmentCollector {
     segment_ordinal: SegmentOrdinal,
     accessor_idx: usize,
-    top_n: TopNComputer<Vec<DocValueAndOrder>, DocAddress, false>,
+    buckets: Vec<TopNComputer<Vec<DocValueAndOrder>, DocAddress, ReverseComparator>>,
+    num_hits: usize,
 }
 
 impl TopHitsSegmentCollector {
@@ -526,25 +530,35 @@ impl TopHitsSegmentCollector {
         accessor_idx: usize,
         segment_ordinal: SegmentOrdinal,
     ) -> Self {
+        let num_hits = req.size + req.from.unwrap_or(0);
         Self {
-            top_n: TopNComputer::new(req.size + req.from.unwrap_or(0)),
+            num_hits,
             segment_ordinal,
             accessor_idx,
+            buckets: vec![TopNComputer::new_with_comparator(num_hits, ReverseComparator); 1],
         }
     }
-    fn into_top_hits_collector(
-        self,
+    fn get_top_hits_computer(
+        &mut self,
+        parent_bucket_id: BucketId,
         value_accessors: &HashMap<String, Vec<DynamicColumn>>,
         req: &TopHitsAggregationReq,
     ) -> TopHitsTopNComputer {
+        if parent_bucket_id as usize >= self.buckets.len() {
+            return TopHitsTopNComputer::new(req);
+        }
+        let top_n = std::mem::replace(
+            &mut self.buckets[parent_bucket_id as usize],
+            TopNComputer::new(0),
+        );
         let mut top_hits_computer = TopHitsTopNComputer::new(req);
-        let top_results = self.top_n.into_vec();
+        let top_results = top_n.into_vec();
 
         for res in top_results {
             let doc_value_fields = req.get_document_field_data(value_accessors, res.doc.doc_id);
             top_hits_computer.collect(
                 DocSortValuesAndFields {
-                    sorts: res.feature,
+                    sorts: res.sort_key,
                     doc_value_fields,
                 },
                 res.doc,
@@ -553,54 +567,24 @@ impl TopHitsSegmentCollector {
 
         top_hits_computer
     }
-
-    /// TODO add a specialized variant for a single sort field
-    fn collect_with(
-        &mut self,
-        doc_id: crate::DocId,
-        req: &TopHitsAggregationReq,
-        accessors: &[(Column<u64>, ColumnType)],
-    ) -> crate::Result<()> {
-        let sorts: Vec<DocValueAndOrder> = req
-            .sort
-            .iter()
-            .enumerate()
-            .map(|(idx, KeyOrder { order, .. })| {
-                let order = *order;
-                let value = accessors
-                    .get(idx)
-                    .expect("could not find field in accessors")
-                    .0
-                    .values_for_doc(doc_id)
-                    .next();
-                DocValueAndOrder { value, order }
-            })
-            .collect();
-
-        self.top_n.push(
-            sorts,
-            DocAddress {
-                segment_ord: self.segment_ordinal,
-                doc_id,
-            },
-        );
-        Ok(())
-    }
 }
 
 impl SegmentAggregationCollector for TopHitsSegmentCollector {
     fn add_intermediate_aggregation_result(
-        self: Box<Self>,
+        &mut self,
         agg_data: &AggregationsSegmentCtx,
         results: &mut crate::aggregation::intermediate_agg_result::IntermediateAggregationResults,
+        parent_bucket_id: BucketId,
     ) -> crate::Result<()> {
         let req_data = agg_data.get_top_hits_req_data(self.accessor_idx);
 
         let value_accessors = &req_data.value_accessors;
 
-        let intermediate_result = IntermediateMetricResult::TopHits(
-            self.into_top_hits_collector(value_accessors, &req_data.req),
-        );
+        let intermediate_result = IntermediateMetricResult::TopHits(self.get_top_hits_computer(
+            parent_bucket_id,
+            value_accessors,
+            &req_data.req,
+        ));
         results.push(
             req_data.name.to_string(),
             IntermediateAggregationResult::Metric(intermediate_result),
@@ -610,26 +594,56 @@ impl SegmentAggregationCollector for TopHitsSegmentCollector {
     /// TODO: Consider a caching layer to reduce the call overhead
     fn collect(
         &mut self,
-        doc_id: crate::DocId,
-        agg_data: &mut AggregationsSegmentCtx,
-    ) -> crate::Result<()> {
-        let req_data = agg_data.get_top_hits_req_data(self.accessor_idx);
-        self.collect_with(doc_id, &req_data.req, &req_data.accessors)?;
-        Ok(())
-    }
-
-    fn collect_block(
-        &mut self,
+        parent_bucket_id: BucketId,
         docs: &[crate::DocId],
         agg_data: &mut AggregationsSegmentCtx,
     ) -> crate::Result<()> {
+        let top_n = &mut self.buckets[parent_bucket_id as usize];
         let req_data = agg_data.get_top_hits_req_data(self.accessor_idx);
-        // TODO: Consider getting fields with the column block accessor.
-        for doc in docs {
-            self.collect_with(*doc, &req_data.req, &req_data.accessors)?;
+        let req = &req_data.req;
+        let accessors = &req_data.accessors;
+        for &doc_id in docs {
+            // TODO: this is terrible, a new vec is allocated for every doc
+            // We can fetch blocks instead
+            // We don't need to store the order for every value
+            let sorts: Vec<DocValueAndOrder> = req
+                .sort
+                .iter()
+                .enumerate()
+                .map(|(idx, KeyOrder { order, .. })| {
+                    let order = *order;
+                    let value = accessors
+                        .get(idx)
+                        .expect("could not find field in accessors")
+                        .0
+                        .values_for_doc(doc_id)
+                        .next();
+                    DocValueAndOrder { value, order }
+                })
+                .collect();
+
+            top_n.push(
+                sorts,
+                DocAddress {
+                    segment_ord: self.segment_ordinal,
+                    doc_id,
+                },
+            );
         }
         Ok(())
     }
+
+    fn prepare_max_bucket(
+        &mut self,
+        max_bucket: BucketId,
+        _agg_data: &AggregationsSegmentCtx,
+    ) -> crate::Result<()> {
+        self.buckets.resize(
+            (max_bucket as usize) + 1,
+            TopNComputer::new_with_comparator(self.num_hits, ReverseComparator),
+        );
+        Ok(())
+    }
 }
 
 #[cfg(test)]
@@ -645,6 +659,7 @@ mod tests {
     use crate::aggregation::bucket::tests::get_test_index_from_docs;
     use crate::aggregation::tests::get_test_index_from_values;
     use crate::aggregation::AggregationCollector;
+    use crate::collector::sort_key::ReverseComparator;
     use crate::collector::ComparableDoc;
     use crate::query::AllQuery;
     use crate::schema::OwnedValue;
@@ -660,7 +675,7 @@ mod tests {
 
     fn collector_with_capacity(capacity: usize) -> super::TopHitsTopNComputer {
         super::TopHitsTopNComputer {
-            top_n: super::TopNComputer::new(capacity),
+            top_n: super::TopNComputer::new_with_comparator(capacity, ReverseComparator),
             req: Default::default(),
         }
     }
@@ -744,7 +759,7 @@ mod tests {
                     ],
                     "from": 0,
                 }
-        }
+            }
         }))
         .unwrap();
 
@@ -774,12 +789,12 @@ mod tests {
     #[test]
     fn test_top_hits_collector_single_feature() -> crate::Result<()> {
         let docs = vec![
-            ComparableDoc::<_, _, false> {
+            ComparableDoc::<_, _> {
                 doc: crate::DocAddress {
                     segment_ord: 0,
                     doc_id: 0,
                 },
-                feature: DocSortValuesAndFields {
+                sort_key: DocSortValuesAndFields {
                     sorts: vec![DocValueAndOrder {
                         value: Some(1),
                         order: Order::Asc,
@@ -792,7 +807,7 @@ mod tests {
                     segment_ord: 0,
                     doc_id: 2,
                 },
-                feature: DocSortValuesAndFields {
+                sort_key: DocSortValuesAndFields {
                     sorts: vec![DocValueAndOrder {
                         value: Some(3),
                         order: Order::Asc,
@@ -805,7 +820,7 @@ mod tests {
                     segment_ord: 0,
                     doc_id: 1,
                 },
-                feature: DocSortValuesAndFields {
+                sort_key: DocSortValuesAndFields {
                     sorts: vec![DocValueAndOrder {
                         value: Some(5),
                         order: Order::Asc,
@@ -817,7 +832,7 @@ mod tests {
 
         let mut collector = collector_with_capacity(3);
         for doc in docs.clone() {
-            collector.collect(doc.feature, doc.doc);
+            collector.collect(doc.sort_key, doc.doc);
         }
 
         let res = collector.into_final_result();
@@ -827,15 +842,15 @@ mod tests {
             super::TopHitsMetricResult {
                 hits: vec![
                     super::TopHitsVecEntry {
-                        sort: vec![docs[0].feature.sorts[0].value],
+                        sort: vec![docs[0].sort_key.sorts[0].value],
                         doc_value_fields: Default::default(),
                     },
                     super::TopHitsVecEntry {
-                        sort: vec![docs[1].feature.sorts[0].value],
+                        sort: vec![docs[1].sort_key.sorts[0].value],
                         doc_value_fields: Default::default(),
                     },
                     super::TopHitsVecEntry {
-                        sort: vec![docs[2].feature.sorts[0].value],
+                        sort: vec![docs[2].sort_key.sorts[0].value],
                         doc_value_fields: Default::default(),
                     },
                 ]
@@ -873,7 +888,7 @@ mod tests {
                         "mixed.*",
                     ],
                 }
-        }
+            }
         }))?;
 
         let collector = AggregationCollector::from_aggs(d, Default::default());

src/aggregation/mod.rs

@@ -133,7 +133,7 @@ mod agg_limits;
 pub mod agg_req;
 pub mod agg_result;
 pub mod bucket;
-mod buf_collector;
+pub(crate) mod cached_sub_aggs;
 mod collector;
 mod date;
 mod error;
@@ -162,6 +162,19 @@ use serde::{Deserialize, Deserializer, Serialize};
 
 use crate::tokenizer::TokenizerManager;
 
+/// A bucket id is a dense identifier for a bucket within an aggregation.
+/// It is used to index into a Vec that hold per-bucket data.
+///
+/// For example, in a terms aggregation, each unique term will be assigned a incremental BucketId.
+/// This BucketId will be forwarded to sub-aggregations to identify the parent bucket.
+///
+/// This allows to have a single AggregationCollector instance per aggregation,
+/// that can handle multiple buckets efficiently.
+///
+/// The API to call sub-aggregations is therefore a &[(BucketId, &[DocId])].
+/// For that we'll need a buffer. One Vec per bucket aggregation is needed.
+pub type BucketId = u32;
+
 /// Context parameters for aggregation execution
 ///
 /// This struct holds shared resources needed during aggregation execution:
@@ -335,19 +348,37 @@ impl Display for Key {
     }
 }
 
+pub(crate) fn convert_to_f64<const COLUMN_TYPE_ID: u8>(val: u64) -> f64 {
+    if COLUMN_TYPE_ID == ColumnType::U64 as u8 {
+        val as f64
+    } else if COLUMN_TYPE_ID == ColumnType::I64 as u8
+        || COLUMN_TYPE_ID == ColumnType::DateTime as u8
+    {
+        i64::from_u64(val) as f64
+    } else if COLUMN_TYPE_ID == ColumnType::F64 as u8 {
+        f64::from_u64(val)
+    } else if COLUMN_TYPE_ID == ColumnType::Bool as u8 {
+        val as f64
+    } else {
+        panic!(
+            "ColumnType ID {} cannot be converted to f64 metric",
+            COLUMN_TYPE_ID
+        )
+    }
+}
+
 /// Inverse of `to_fastfield_u64`. Used to convert to `f64` for metrics.
 ///
 /// # Panics
 /// Only `u64`, `f64`, `date`, and `i64` are supported.
-pub(crate) fn f64_from_fastfield_u64(val: u64, field_type: &ColumnType) -> f64 {
+pub(crate) fn f64_from_fastfield_u64(val: u64, field_type: ColumnType) -> f64 {
     match field_type {
-        ColumnType::U64 => val as f64,
-        ColumnType::I64 | ColumnType::DateTime => i64::from_u64(val) as f64,
-        ColumnType::F64 => f64::from_u64(val),
-        ColumnType::Bool => val as f64,
-        _ => {
-            panic!("unexpected type {field_type:?}. This should not happen")
-        }
+        ColumnType::U64 => convert_to_f64::<{ ColumnType::U64 as u8 }>(val),
+        ColumnType::I64 => convert_to_f64::<{ ColumnType::I64 as u8 }>(val),
+        ColumnType::F64 => convert_to_f64::<{ ColumnType::F64 as u8 }>(val),
+        ColumnType::Bool => convert_to_f64::<{ ColumnType::Bool as u8 }>(val),
+        ColumnType::DateTime => convert_to_f64::<{ ColumnType::DateTime as u8 }>(val),
+        _ => panic!("unexpected type {field_type:?}. This should not happen"),
     }
 }
 

src/aggregation/segment_agg_result.rs

@@ -8,30 +8,69 @@ use std::fmt::Debug;
 pub(crate) use super::agg_limits::AggregationLimitsGuard;
 use super::intermediate_agg_result::IntermediateAggregationResults;
 use crate::aggregation::agg_data::AggregationsSegmentCtx;
+use crate::aggregation::BucketId;
+
+/// Monotonically increasing provider of BucketIds.
+#[derive(Debug, Clone, Default)]
+pub struct BucketIdProvider(u32);
+impl BucketIdProvider {
+    /// Get the next BucketId.
+    pub fn next_bucket_id(&mut self) -> BucketId {
+        let bucket_id = self.0;
+        self.0 += 1;
+        bucket_id
+    }
+}
 
 /// A SegmentAggregationCollector is used to collect aggregation results.
-pub trait SegmentAggregationCollector: CollectorClone + Debug {
+pub trait SegmentAggregationCollector: Debug {
     fn add_intermediate_aggregation_result(
-        self: Box<Self>,
+        &mut self,
         agg_data: &AggregationsSegmentCtx,
         results: &mut IntermediateAggregationResults,
+        parent_bucket_id: BucketId,
     ) -> crate::Result<()>;
 
-    #[inline]
+    /// Note: The caller needs to call `prepare_max_bucket` before calling `collect`.
     fn collect(
         &mut self,
-        doc: crate::DocId,
-        agg_data: &mut AggregationsSegmentCtx,
-    ) -> crate::Result<()> {
-        self.collect_block(&[doc], agg_data)
-    }
-
-    fn collect_block(
-        &mut self,
+        parent_bucket_id: BucketId,
         docs: &[crate::DocId],
         agg_data: &mut AggregationsSegmentCtx,
     ) -> crate::Result<()>;
 
+    /// Collect docs for multiple buckets in one call.
+    /// Minimizes dynamic dispatch overhead when collecting many buckets.
+    ///
+    /// Note: The caller needs to call `prepare_max_bucket` before calling `collect`.
+    fn collect_multiple(
+        &mut self,
+        bucket_ids: &[BucketId],
+        docs: &[crate::DocId],
+        agg_data: &mut AggregationsSegmentCtx,
+    ) -> crate::Result<()> {
+        debug_assert_eq!(bucket_ids.len(), docs.len());
+        let mut start = 0;
+        while start < bucket_ids.len() {
+            let bucket_id = bucket_ids[start];
+            let mut end = start + 1;
+            while end < bucket_ids.len() && bucket_ids[end] == bucket_id {
+                end += 1;
+            }
+            self.collect(bucket_id, &docs[start..end], agg_data)?;
+            start = end;
+        }
+        Ok(())
+    }
+
+    /// Prepare the collector for collecting up to BucketId `max_bucket`.
+    /// This is useful so we can split allocation ahead of time of collecting.
+    fn prepare_max_bucket(
+        &mut self,
+        max_bucket: BucketId,
+        agg_data: &AggregationsSegmentCtx,
+    ) -> crate::Result<()>;
+
     /// Finalize method. Some Aggregator collect blocks of docs before calling `collect_block`.
     /// This method ensures those staged docs will be collected.
     fn flush(&mut self, _agg_data: &mut AggregationsSegmentCtx) -> crate::Result<()> {
@@ -39,26 +78,7 @@ pub trait SegmentAggregationCollector: CollectorClone + Debug {
     }
 }
 
-/// A helper trait to enable cloning of Box<dyn SegmentAggregationCollector>
-pub trait CollectorClone {
-    fn clone_box(&self) -> Box<dyn SegmentAggregationCollector>;
-}
-
-impl<T> CollectorClone for T
-where T: 'static + SegmentAggregationCollector + Clone
-{
-    fn clone_box(&self) -> Box<dyn SegmentAggregationCollector> {
-        Box::new(self.clone())
-    }
-}
-
-impl Clone for Box<dyn SegmentAggregationCollector> {
-    fn clone(&self) -> Box<dyn SegmentAggregationCollector> {
-        self.clone_box()
-    }
-}
-
-#[derive(Clone, Default)]
+#[derive(Default)]
 /// The GenericSegmentAggregationResultsCollector is the generic version of the collector, which
 /// can handle arbitrary complexity of  sub-aggregations. Ideally we never have to pick this one
 /// and can provide specialized versions instead, that remove some of its overhead.
@@ -76,12 +96,13 @@ impl Debug for GenericSegmentAggregationResultsCollector {
 
 impl SegmentAggregationCollector for GenericSegmentAggregationResultsCollector {
     fn add_intermediate_aggregation_result(
-        self: Box<Self>,
+        &mut self,
         agg_data: &AggregationsSegmentCtx,
         results: &mut IntermediateAggregationResults,
+        parent_bucket_id: BucketId,
     ) -> crate::Result<()> {
-        for agg in self.aggs {
-            agg.add_intermediate_aggregation_result(agg_data, results)?;
+        for agg in &mut self.aggs {
+            agg.add_intermediate_aggregation_result(agg_data, results, parent_bucket_id)?;
         }
 
         Ok(())
@@ -89,23 +110,13 @@ impl SegmentAggregationCollector for GenericSegmentAggregationResultsCollector {
 
     fn collect(
         &mut self,
-        doc: crate::DocId,
-        agg_data: &mut AggregationsSegmentCtx,
-    ) -> crate::Result<()> {
-        self.collect_block(&[doc], agg_data)?;
-
-        Ok(())
-    }
-
-    fn collect_block(
-        &mut self,
+        parent_bucket_id: BucketId,
         docs: &[crate::DocId],
         agg_data: &mut AggregationsSegmentCtx,
     ) -> crate::Result<()> {
         for collector in &mut self.aggs {
-            collector.collect_block(docs, agg_data)?;
+            collector.collect(parent_bucket_id, docs, agg_data)?;
         }
-
         Ok(())
     }
 
@@ -115,4 +126,15 @@ impl SegmentAggregationCollector for GenericSegmentAggregationResultsCollector {
         }
         Ok(())
     }
+
+    fn prepare_max_bucket(
+        &mut self,
+        max_bucket: BucketId,
+        agg_data: &AggregationsSegmentCtx,
+    ) -> crate::Result<()> {
+        for collector in &mut self.aggs {
+            collector.prepare_max_bucket(max_bucket, agg_data)?;
+        }
+        Ok(())
+    }
 }

src/collector/custom_score_top_collector.rs

@@ -1,121 +0,0 @@
-use crate::collector::top_collector::{TopCollector, TopSegmentCollector};
-use crate::collector::{Collector, SegmentCollector};
-use crate::{DocAddress, DocId, Score, SegmentReader};
-
-pub(crate) struct CustomScoreTopCollector<TCustomScorer, TScore = Score> {
-    custom_scorer: TCustomScorer,
-    collector: TopCollector<TScore>,
-}
-
-impl<TCustomScorer, TScore> CustomScoreTopCollector<TCustomScorer, TScore>
-where TScore: Clone + PartialOrd
-{
-    pub(crate) fn new(
-        custom_scorer: TCustomScorer,
-        collector: TopCollector<TScore>,
-    ) -> CustomScoreTopCollector<TCustomScorer, TScore> {
-        CustomScoreTopCollector {
-            custom_scorer,
-            collector,
-        }
-    }
-}
-
-/// A custom segment scorer makes it possible to define any kind of score
-/// for a given document belonging to a specific segment.
-///
-/// It is the segment local version of the [`CustomScorer`].
-pub trait CustomSegmentScorer<TScore>: 'static {
-    /// Computes the score of a specific `doc`.
-    fn score(&mut self, doc: DocId) -> TScore;
-}
-
-/// `CustomScorer` makes it possible to define any kind of score.
-///
-/// The `CustomerScorer` itself does not make much of the computation itself.
-/// Instead, it helps constructing `Self::Child` instances that will compute
-/// the score at a segment scale.
-pub trait CustomScorer<TScore>: Sync {
-    /// Type of the associated [`CustomSegmentScorer`].
-    type Child: CustomSegmentScorer<TScore>;
-    /// Builds a child scorer for a specific segment. The child scorer is associated with
-    /// a specific segment.
-    fn segment_scorer(&self, segment_reader: &SegmentReader) -> crate::Result<Self::Child>;
-}
-
-impl<TCustomScorer, TScore> Collector for CustomScoreTopCollector<TCustomScorer, TScore>
-where
-    TCustomScorer: CustomScorer<TScore> + Send + Sync,
-    TScore: 'static + PartialOrd + Clone + Send + Sync,
-{
-    type Fruit = Vec<(TScore, DocAddress)>;
-
-    type Child = CustomScoreTopSegmentCollector<TCustomScorer::Child, TScore>;
-
-    fn for_segment(
-        &self,
-        segment_local_id: u32,
-        segment_reader: &SegmentReader,
-    ) -> crate::Result<Self::Child> {
-        let segment_collector = self.collector.for_segment(segment_local_id, segment_reader);
-        let segment_scorer = self.custom_scorer.segment_scorer(segment_reader)?;
-        Ok(CustomScoreTopSegmentCollector {
-            segment_collector,
-            segment_scorer,
-        })
-    }
-
-    fn requires_scoring(&self) -> bool {
-        false
-    }
-
-    fn merge_fruits(&self, segment_fruits: Vec<Self::Fruit>) -> crate::Result<Self::Fruit> {
-        self.collector.merge_fruits(segment_fruits)
-    }
-}
-
-pub struct CustomScoreTopSegmentCollector<T, TScore>
-where
-    TScore: 'static + PartialOrd + Clone + Send + Sync + Sized,
-    T: CustomSegmentScorer<TScore>,
-{
-    segment_collector: TopSegmentCollector<TScore>,
-    segment_scorer: T,
-}
-
-impl<T, TScore> SegmentCollector for CustomScoreTopSegmentCollector<T, TScore>
-where
-    TScore: 'static + PartialOrd + Clone + Send + Sync,
-    T: 'static + CustomSegmentScorer<TScore>,
-{
-    type Fruit = Vec<(TScore, DocAddress)>;
-
-    fn collect(&mut self, doc: DocId, _score: Score) {
-        let score = self.segment_scorer.score(doc);
-        self.segment_collector.collect(doc, score);
-    }
-
-    fn harvest(self) -> Vec<(TScore, DocAddress)> {
-        self.segment_collector.harvest()
-    }
-}
-
-impl<F, TScore, T> CustomScorer<TScore> for F
-where
-    F: 'static + Send + Sync + Fn(&SegmentReader) -> T,
-    T: CustomSegmentScorer<TScore>,
-{
-    type Child = T;
-
-    fn segment_scorer(&self, segment_reader: &SegmentReader) -> crate::Result<Self::Child> {
-        Ok((self)(segment_reader))
-    }
-}
-
-impl<F, TScore> CustomSegmentScorer<TScore> for F
-where F: 'static + FnMut(DocId) -> TScore
-{
-    fn score(&mut self, doc: DocId) -> TScore {
-        (self)(doc)
-    }
-}

src/collector/facet_collector.rs

@@ -486,9 +486,9 @@ mod tests {
     use std::collections::BTreeSet;
 
     use columnar::Dictionary;
-    use rand::distributions::Uniform;
+    use rand::distr::Uniform;
     use rand::prelude::SliceRandom;
-    use rand::{thread_rng, Rng};
+    use rand::{rng, Rng};
 
     use super::{FacetCollector, FacetCounts};
     use crate::collector::facet_collector::compress_mapping;
@@ -731,7 +731,7 @@ mod tests {
         let schema = schema_builder.build();
         let index = Index::create_in_ram(schema);
 
-        let uniform = Uniform::new_inclusive(1, 100_000);
+        let uniform = Uniform::new_inclusive(1, 100_000).unwrap();
         let mut docs: Vec<TantivyDocument> =
             vec![("a", 10), ("b", 100), ("c", 7), ("d", 12), ("e", 21)]
                 .into_iter()
@@ -741,14 +741,11 @@ mod tests {
                     std::iter::repeat_n(doc, count)
                 })
                 .map(|mut doc| {
-                    doc.add_facet(
-                        facet_field,
-                        &format!("/facet/{}", thread_rng().sample(uniform)),
-                    );
+                    doc.add_facet(facet_field, &format!("/facet/{}", rng().sample(uniform)));
                     doc
                 })
                 .collect();
-        docs[..].shuffle(&mut thread_rng());
+        docs[..].shuffle(&mut rng());
 
         let mut index_writer: IndexWriter = index.writer_for_tests().unwrap();
         for doc in docs {
@@ -822,8 +819,8 @@ mod tests {
 #[cfg(all(test, feature = "unstable"))]
 mod bench {
 
+    use rand::rng;
     use rand::seq::SliceRandom;
-    use rand::thread_rng;
     use test::Bencher;
 
     use crate::collector::FacetCollector;
@@ -846,7 +843,7 @@ mod bench {
             }
         }
         // 40425 docs
-        docs[..].shuffle(&mut thread_rng());
+        docs[..].shuffle(&mut rng());
 
         let mut index_writer: IndexWriter = index.writer_for_tests().unwrap();
         for doc in docs {

src/collector/filter_collector_wrapper.rs

@@ -12,6 +12,7 @@ use std::marker::PhantomData;
 use columnar::{BytesColumn, Column, DynamicColumn, HasAssociatedColumnType};
 
 use crate::collector::{Collector, SegmentCollector};
+use crate::schema::Schema;
 use crate::{DocId, Score, SegmentReader};
 
 /// The `FilterCollector` filters docs using a fast field value and a predicate.
@@ -49,13 +50,13 @@ use crate::{DocId, Score, SegmentReader};
 ///
 /// let query_parser = QueryParser::for_index(&index, vec![title]);
 /// let query = query_parser.parse_query("diary")?;
-/// let no_filter_collector = FilterCollector::new("price".to_string(), |value: u64| value > 20_120u64, TopDocs::with_limit(2));
+/// let no_filter_collector = FilterCollector::new("price".to_string(), |value: u64| value > 20_120u64, TopDocs::with_limit(2).order_by_score());
 /// let top_docs = searcher.search(&query, &no_filter_collector)?;
 ///
 /// assert_eq!(top_docs.len(), 1);
 /// assert_eq!(top_docs[0].1, DocAddress::new(0, 1));
 ///
-/// let filter_all_collector: FilterCollector<_, _, u64> = FilterCollector::new("price".to_string(), |value| value < 5u64, TopDocs::with_limit(2));
+/// let filter_all_collector: FilterCollector<_, _, u64> = FilterCollector::new("price".to_string(), |value| value < 5u64, TopDocs::with_limit(2).order_by_score());
 /// let filtered_top_docs = searcher.search(&query, &filter_all_collector)?;
 ///
 /// assert_eq!(filtered_top_docs.len(), 0);
@@ -104,6 +105,11 @@ where
 
     type Child = FilterSegmentCollector<TCollector::Child, TPredicate, TPredicateValue>;
 
+    fn check_schema(&self, schema: &Schema) -> crate::Result<()> {
+        self.collector.check_schema(schema)?;
+        Ok(())
+    }
+
     fn for_segment(
         &self,
         segment_local_id: u32,
@@ -120,6 +126,7 @@ where
             segment_collector,
             predicate: self.predicate.clone(),
             t_predicate_value: PhantomData,
+            filtered_docs: Vec::with_capacity(crate::COLLECT_BLOCK_BUFFER_LEN),
         })
     }
 
@@ -140,6 +147,7 @@ pub struct FilterSegmentCollector<TSegmentCollector, TPredicate, TPredicateValue
     segment_collector: TSegmentCollector,
     predicate: TPredicate,
     t_predicate_value: PhantomData<TPredicateValue>,
+    filtered_docs: Vec<DocId>,
 }
 
 impl<TSegmentCollector, TPredicate, TPredicateValue>
@@ -176,6 +184,20 @@ where
         }
     }
 
+    fn collect_block(&mut self, docs: &[DocId]) {
+        self.filtered_docs.clear();
+        for &doc in docs {
+            // TODO: `accept_document` could be further optimized to do batch lookups of column
+            // values for single-valued columns.
+            if self.accept_document(doc) {
+                self.filtered_docs.push(doc);
+            }
+        }
+        if !self.filtered_docs.is_empty() {
+            self.segment_collector.collect_block(&self.filtered_docs);
+        }
+    }
+
     fn harvest(self) -> TSegmentCollector::Fruit {
         self.segment_collector.harvest()
     }
@@ -218,7 +240,7 @@ where
 ///
 /// let query_parser = QueryParser::for_index(&index, vec![title]);
 /// let query = query_parser.parse_query("diary")?;
-/// let filter_collector = BytesFilterCollector::new("barcode".to_string(), |bytes: &[u8]| bytes.starts_with(b"01"), TopDocs::with_limit(2));
+/// let filter_collector = BytesFilterCollector::new("barcode".to_string(), |bytes: &[u8]| bytes.starts_with(b"01"), TopDocs::with_limit(2).order_by_score());
 /// let top_docs = searcher.search(&query, &filter_collector)?;
 ///
 /// assert_eq!(top_docs.len(), 1);
@@ -258,6 +280,10 @@ where
 
     type Child = BytesFilterSegmentCollector<TCollector::Child, TPredicate>;
 
+    fn check_schema(&self, schema: &Schema) -> crate::Result<()> {
+        self.collector.check_schema(schema)
+    }
+
     fn for_segment(
         &self,
         segment_local_id: u32,
@@ -274,6 +300,7 @@ where
             segment_collector,
             predicate: self.predicate.clone(),
             buffer: Vec::new(),
+            filtered_docs: Vec::with_capacity(crate::COLLECT_BLOCK_BUFFER_LEN),
         })
     }
 
@@ -296,6 +323,7 @@ where TPredicate: 'static
     segment_collector: TSegmentCollector,
     predicate: TPredicate,
     buffer: Vec<u8>,
+    filtered_docs: Vec<DocId>,
 }
 
 impl<TSegmentCollector, TPredicate> BytesFilterSegmentCollector<TSegmentCollector, TPredicate>
@@ -334,6 +362,20 @@ where
         }
     }
 
+    fn collect_block(&mut self, docs: &[DocId]) {
+        self.filtered_docs.clear();
+        for &doc in docs {
+            // TODO: `accept_document` could be further optimized to do batch lookups of column
+            // values for single-valued columns.
+            if self.accept_document(doc) {
+                self.filtered_docs.push(doc);
+            }
+        }
+        if !self.filtered_docs.is_empty() {
+            self.segment_collector.collect_block(&self.filtered_docs);
+        }
+    }
+
     fn harvest(self) -> TSegmentCollector::Fruit {
         self.segment_collector.harvest()
     }