Update CHANGELOG for Tantivy 0.26 release

The early return for `scorers.len() == 1` in `scorer_union` short-circuits a single TermScorer into `SpecializedScorer::Other`, bypassing the `TermUnion` path that enables block-max WAND (BMW) in `for_each_pruning`.

This was originally addressed in PR #2898 (backed out), which added a special case in `BooleanWeight::for_each_pruning`. PR #2912 (merged as d27ca164a) added a single-scorer fast path inside `block_wand` itself, but did not remove this early return — so a single SHOULD TermScorer still never reaches the BMW path.

Removing the early return lets a single TermScorer with freq reading flow through to `SpecializedScorer::TermUnion`, where `block_wand` → `block_wand_single_scorer` handles it efficiently.

.github/workflows/coverage.yml

@@ -4,6 +4,9 @@ on:
   push:
     branches: [main]
 
+permissions:
+  contents: read
+
 # Ensures that we cancel running jobs for the same PR / same workflow.
 concurrency:
   group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}
@@ -12,16 +15,20 @@ concurrency:
 jobs:
   coverage:
     runs-on: ubuntu-latest
+
+    permissions:
+      contents: read
+
     steps:
-      - uses: actions/checkout@v6
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
       - name: Install Rust
         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
+      - uses: Swatinem/rust-cache@c19371144df3bb44fab255c43d04cbc2ab54d1c4 # v2.9.1
+      - uses: taiki-e/install-action@e4b3a0453201addddc06d3a72db90326aad87084 # cargo-llvm-cov
       - name: Generate code coverage
         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@v6
+        uses: codecov/codecov-action@57e3a136b779b570ffcdbf80b3bdc90e7fab3de2 # v6.0.0
         continue-on-error: true
         with:
           token: ${{ secrets.CODECOV_TOKEN }} # not required for public repos

.github/workflows/long_running.yml

@@ -8,6 +8,9 @@ env:
   CARGO_TERM_COLOR: always
   NUM_FUNCTIONAL_TEST_ITERATIONS: 20000
 
+permissions:
+  contents: read
+
 # Ensures that we cancel running jobs for the same PR / same workflow.
 concurrency:
   group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}
@@ -18,10 +21,13 @@ jobs:
 
     runs-on: ubuntu-latest
 
+    permissions:
+      contents: read
+
     steps:
-    - uses: actions/checkout@v6
+    - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
     - name: Install stable
-      uses: actions-rs/toolchain@v1
+      uses: actions-rs/toolchain@16499b5e05bf2e26879000db0c1d13f7e13fa3af # v1.0.7
       with:
           toolchain: stable
           profile: minimal

.github/workflows/scorecard.yml

@@ -0,0 +1,49 @@
+name: OpenSSF Scorecard
+
+on:
+  schedule:
+    - cron: '0 0 * * 0'
+  push:
+    branches:
+      - main
+
+permissions:
+  contents: read
+
+jobs:
+  analysis:
+    name: Scorecards analysis
+    runs-on: ubuntu-latest
+    permissions:
+      # Needed to upload the results to code-scanning dashboard.
+      security-events: write
+      # Needed to publish results
+      id-token: write
+
+    steps:
+      - name: 'Checkout code'
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+        with:
+          persist-credentials: false
+
+      - name: 'Run analysis'
+        uses: ossf/scorecard-action@4eaacf0543bb3f2c246792bd56e8cdeffafb205a # v2.4.3
+        with:
+          results_file: results.sarif
+          results_format: sarif
+          repo_token: ${{ secrets.GITHUB_TOKEN }}
+          publish_results: true
+
+      # Upload the results as artifacts.
+      - name: 'Upload artifact'
+        uses: actions/upload-artifact@bbbca2ddaa5d8feaa63e36b76fdaad77386f024f # v7.0.0
+        with:
+          name: SARIF file
+          path: results.sarif
+          retention-days: 5
+
+      # Upload the results to GitHub's code scanning dashboard.
+      - name: 'Upload to code-scanning'
+        uses: github/codeql-action/upload-sarif@95e58e9a2cdfd71adc6e0353d5c52f41a045d225 # v4.35.2
+        with:
+          sarif_file: results.sarif

.github/workflows/test.yml

@@ -9,6 +9,9 @@ on:
 env:
   CARGO_TERM_COLOR: always
 
+permissions:
+  contents: read
+
 # Ensures that we cancel running jobs for the same PR / same workflow.
 concurrency:
   group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}
@@ -19,23 +22,27 @@ jobs:
 
     runs-on: ubuntu-latest
 
+    permissions:
+      contents: read
+      checks: write
+
     steps:
-    - uses: actions/checkout@v6
+    - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
 
     - name: Install nightly
-      uses: actions-rs/toolchain@v1
+      uses: actions-rs/toolchain@16499b5e05bf2e26879000db0c1d13f7e13fa3af # v1.0.7
       with:
             toolchain: nightly
             profile: minimal
             components: rustfmt
     - name: Install stable
-      uses: actions-rs/toolchain@v1
+      uses: actions-rs/toolchain@16499b5e05bf2e26879000db0c1d13f7e13fa3af # v1.0.7
       with:
             toolchain: stable
             profile: minimal
             components: clippy
 
-    - uses: Swatinem/rust-cache@v2
+    - uses: Swatinem/rust-cache@c19371144df3bb44fab255c43d04cbc2ab54d1c4 # v2.9.1
 
     - name: Check Formatting
       run: cargo +nightly fmt --all -- --check
@@ -47,7 +54,7 @@ jobs:
     - name: Check Bench Compilation
       run: cargo +nightly bench --no-run --profile=dev --all-features
 
-    - uses: actions-rs/clippy-check@v1
+    - uses: actions-rs/clippy-check@b5b5f21f4797c02da247df37026fcd0a5024aa4d # v1.0.7
       with:
         toolchain: stable
         token: ${{ secrets.GITHUB_TOKEN }}
@@ -57,6 +64,9 @@ jobs:
 
     runs-on: ubuntu-latest
 
+    permissions:
+      contents: read
+
     strategy:
       matrix:
         features:
@@ -67,17 +77,17 @@ jobs:
     name: test-${{ matrix.features.label}}
 
     steps:
-    - uses: actions/checkout@v6
+    - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
 
     - name: Install stable
-      uses: actions-rs/toolchain@v1
+      uses: actions-rs/toolchain@16499b5e05bf2e26879000db0c1d13f7e13fa3af # v1.0.7
       with:
             toolchain: stable
             profile: minimal
             override: true
 
-    - uses: taiki-e/install-action@nextest
-    - uses: Swatinem/rust-cache@v2
+    - uses: taiki-e/install-action@56cc9adf3a3e2c23eafb56e8acaf9d0373cb845a # nextest
+    - uses: Swatinem/rust-cache@c19371144df3bb44fab255c43d04cbc2ab54d1c4 # v2.9.1
 
     - name: Run tests
       run: |

CHANGELOG.md

@@ -4,7 +4,7 @@ Tantivy 0.26.1
 ## Performance
 - Fix quadratic runtime in nested term and composite aggregations: memory accounting scanned all parent buckets on every collect instead of just the current parent (@PSeitz @fulmicoton)
 
-Tantivy 0.26 (Unreleased)
+Tantivy 0.26
 ================================
 
 ## Bugfixes

README.md

@@ -1,6 +1,7 @@
 [![Docs](https://docs.rs/tantivy/badge.svg)](https://docs.rs/crate/tantivy/)
 [![Build Status](https://github.com/quickwit-oss/tantivy/actions/workflows/test.yml/badge.svg)](https://github.com/quickwit-oss/tantivy/actions/workflows/test.yml)
 [![codecov](https://codecov.io/gh/quickwit-oss/tantivy/branch/main/graph/badge.svg)](https://codecov.io/gh/quickwit-oss/tantivy)
+[![OpenSSF Scorecard](https://api.scorecard.dev/projects/github.com/quickwit-oss/tantivy/badge)](https://scorecard.dev/viewer/?uri=github.com/quickwit-oss/tantivy)
 [![Join the chat at https://discord.gg/MT27AG5EVE](https://shields.io/discord/908281611840282624?label=chat%20on%20discord)](https://discord.gg/MT27AG5EVE)
 [![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
 [![Crates.io](https://img.shields.io/crates/v/tantivy.svg)](https://crates.io/crates/tantivy)

benches/agg_bench.rs

@@ -79,8 +79,12 @@ fn bench_agg(mut group: InputGroup<Index>) {
     register!(group, composite_histogram_calendar);
 
     register!(group, cardinality_agg);
+    register!(group, cardinality_agg_high_card);
+    register!(group, cardinality_agg_low_card);
     register!(group, terms_status_with_cardinality_agg);
     register!(group, terms_100_buckets_with_cardinality_agg);
+    register!(group, terms_many_with_single_term_order_by_card);
+    register!(group, terms_many_with_single_term_2_order_by_card);
 
     register!(group, range_agg);
     register!(group, range_agg_with_avg_sub_agg);
@@ -168,6 +172,32 @@ fn cardinality_agg(index: &Index) {
     });
     execute_agg(index, agg_req);
 }
+// Full-scan cardinality on a near-1M-cardinality string field.
+// Hits the dense (PagedBitset) path: every doc has a unique term,
+// so the bucket promotes from FxHashSet shortly into the scan.
+fn cardinality_agg_high_card(index: &Index) {
+    let agg_req = json!({
+        "cardinality": {
+            "cardinality": {
+                "field": "text_all_unique_terms"
+            },
+        }
+    });
+    execute_agg(index, agg_req);
+}
+// Full-scan cardinality on a tiny-cardinality string field (7 distinct
+// values). Stays on the FxHashSet path — the promotion threshold is
+// never crossed. Validates no regression on the sparse path.
+fn cardinality_agg_low_card(index: &Index) {
+    let agg_req = json!({
+        "cardinality": {
+            "cardinality": {
+                "field": "text_few_terms_status"
+            },
+        }
+    });
+    execute_agg(index, agg_req);
+}
 fn terms_status_with_cardinality_agg(index: &Index) {
     let agg_req = json!({
         "my_texts": {
@@ -200,6 +230,58 @@ fn terms_100_buckets_with_cardinality_agg(index: &Index) {
     execute_agg(index, agg_req);
 }
 
+fn terms_many_with_single_term_order_by_card(index: &Index) {
+    let agg_req = json!({
+        "my_texts": {
+            "terms": { "field": "text_many_terms" },
+            "aggs": {
+                "nested_terms": {
+                    "terms": {
+                        "field": "single_term",
+                        "order": { "cardinality": "desc" }
+                    },
+                    "aggs": {
+                        "cardinality": {
+                            "cardinality": { "field": "text_few_terms" }
+                        }
+                    }
+                }
+            }
+        },
+    });
+    execute_agg(index, agg_req);
+}
+
+// Two-level terms ordered by cardinality at each level: a high-card outer terms
+// (text_many_terms) ordered by a cardinality sub-agg, with a nested low-card terms
+// (text_few_terms_status) also ordered by a cardinality sub-agg, plus an avg.
+fn terms_many_with_single_term_2_order_by_card(index: &Index) {
+    let agg_req = json!({
+        "by_ip": {
+            "terms": {
+                "field": "text_many_terms",
+                "order": { "card_few_terms": "desc" }
+            },
+            "aggs": {
+                "card_few_terms": {
+                    "cardinality": { "field": "text_few_terms" }
+                },
+                "nested_terms": {
+                    "terms": {
+                        "field": " single_term",
+                        "order": { "distinct_path2": "desc" }
+                    },
+                    "aggs": {
+                        "avg_botscore": { "avg": { "field": "score" } },
+                        "distinct_path2": { "cardinality": { "field": "text_few_terms" } }
+                    }
+                }
+            }
+        }
+    });
+    execute_agg(index, agg_req);
+}
+
 fn terms_7(index: &Index) {
     let agg_req = json!({
         "my_texts": { "terms": { "field": "text_few_terms_status" } },
@@ -609,7 +691,8 @@ fn get_test_index_bench(cardinality: Cardinality) -> tantivy::Result<Index> {
             TextFieldIndexing::default().set_index_option(IndexRecordOption::WithFreqs),
         )
         .set_stored();
-    let text_field = schema_builder.add_text_field("text", text_fieldtype);
+    let text_field = schema_builder.add_text_field("text", text_fieldtype.clone());
+    let single_term = schema_builder.add_text_field("single_term", FAST);
     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);
@@ -673,6 +756,8 @@ fn get_test_index_bench(cardinality: Cardinality) -> tantivy::Result<Index> {
             index_writer.add_document(doc!(
                 json_field => json!({"mixed_type": 10.0}),
                 json_field => json!({"mixed_type": 10.0}),
+                single_term => "single_term",
+                single_term => "single_term",
                 text_field => "cool",
                 text_field => "cool",
                 text_field_all_unique_terms => "cool",
@@ -707,6 +792,7 @@ fn get_test_index_bench(cardinality: Cardinality) -> tantivy::Result<Index> {
                 json!({"mixed_type": many_terms_data.choose(&mut rng).unwrap().to_string()})
             };
             index_writer.add_document(doc!(
+                single_term => "single_term",
                 text_field => "cool",
                 json_field => json,
                 text_field_all_unique_terms => format!("unique_term_{}", rng.random::<u64>()),

src/aggregation/agg_data.rs

@@ -20,8 +20,8 @@ use crate::aggregation::metric::{
     build_segment_stats_collector, AverageAggregation, CardinalityAggReqData,
     CardinalityAggregationReq, CountAggregation, ExtendedStatsAggregation, MaxAggregation,
     MetricAggReqData, MinAggregation, SegmentCardinalityCollector, SegmentExtendedStatsCollector,
-    SegmentPercentilesCollector, StatsAggregation, StatsType, SumAggregation, TopHitsAggReqData,
-    TopHitsSegmentCollector,
+    SegmentPercentilesCollector, StatsAggregation, StatsType, SumAggregation, TermOrdSet,
+    TopHitsAggReqData, TopHitsSegmentCollector, BITSET_MAX_TERM_ORD,
 };
 use crate::aggregation::segment_agg_result::{
     GenericSegmentAggregationResultsCollector, SegmentAggregationCollector,
@@ -413,12 +413,38 @@ pub(crate) fn build_segment_agg_collector(
         }
         AggKind::Cardinality => {
             let req_data = &mut req.get_cardinality_req_data_mut(node.idx_in_req_data);
-            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,
-            )))
+            // For str columns, choose the per-bucket entries representation
+            // based on the segment's column.max_value():
+            //   * small (< BITSET_MAX_TERM_ORD): `BitSet`, pre-allocated, no promotion machinery.
+            //   * large: `TermOrdSet` (sparse FxHashSet that promotes to a paged bitset).
+            // For non-str columns the `entries` field is unused (values go
+            // straight into the HLL sketch); we still pick `TermOrdSet`
+            // because its empty Sparse(FxHashSet) costs nothing.
+            let is_str = req_data.column_type == ColumnType::Str;
+            let max_term_ord_inclusive = if is_str {
+                req_data.accessor.max_value()
+            } else {
+                0
+            };
+            let collector: Box<dyn SegmentAggregationCollector> =
+                if is_str && max_term_ord_inclusive < BITSET_MAX_TERM_ORD {
+                    Box::new(SegmentCardinalityCollector::<BitSet>::from_req(
+                        req_data.column_type,
+                        node.idx_in_req_data,
+                        req_data.accessor.clone(),
+                        req_data.missing_value_for_accessor,
+                        max_term_ord_inclusive,
+                    ))
+                } else {
+                    Box::new(SegmentCardinalityCollector::<TermOrdSet>::from_req(
+                        req_data.column_type,
+                        node.idx_in_req_data,
+                        req_data.accessor.clone(),
+                        req_data.missing_value_for_accessor,
+                        max_term_ord_inclusive,
+                    ))
+                };
+            Ok(collector)
         }
         AggKind::StatsKind(stats_type) => {
             let req_data = &mut req.per_request.stats_metric_req_data[node.idx_in_req_data];
@@ -1006,10 +1032,20 @@ fn build_terms_or_cardinality_nodes(
                 (idx_in_req_data, AggKind::Terms)
             }
             TermsOrCardinalityRequest::Cardinality(ref req) => {
+                // `str_dict_column` is computed once per field; for JSON paths
+                // with mixed types it's `Some` even on the numeric req_data.
+                // Cardinality only consults it for the str column path, so
+                // gate by column_type to avoid driving non-str collectors
+                // through the coupon-cache path.
+                let str_dict_column_for_req = if column_type == ColumnType::Str {
+                    str_dict_column.clone()
+                } else {
+                    None
+                };
                 let idx_in_req_data = data.push_cardinality_req_data(CardinalityAggReqData {
                     accessor,
                     column_type,
-                    str_dict_column: str_dict_column.clone(),
+                    str_dict_column: str_dict_column_for_req,
                     missing_value_for_accessor,
                     name: agg_name.to_string(),
                     req: req.clone(),

src/aggregation/bucket/composite/collector.rs

@@ -199,6 +199,17 @@ impl SegmentAggregationCollector for SegmentCompositeCollector {
         }
         Ok(())
     }
+
+    fn compute_metric_value(
+        &self,
+        _bucket_id: BucketId,
+        _sub_agg_name: &str,
+        _sub_agg_property: &str,
+        _agg_data: &AggregationsSegmentCtx,
+    ) -> Option<f64> {
+        // Composite is a multi-bucket agg with no single value to extract.
+        None
+    }
 }
 
 impl SegmentCompositeCollector {

src/aggregation/bucket/composite/mod.rs

@@ -559,34 +559,30 @@ mod tests {
                     page_size,
                     agg_req,
                 );
-                if page_idx + 1 < page_count {
-                    assert!(
-                        res["my_composite"].get("after_key").is_some(),
-                        "expected after_key on all but last page"
-                    );
-                    after_key = Some(res["my_composite"]["after_key"].clone());
-                } else if res["my_composite"].get("after_key").is_some() {
-                    // currently we sometime have an after_key on the last page,
-                    // check that the next "page" is empty
-                    let agg_req_json = json!({
-                        "my_composite": {
-                            "composite": {
-                                "sources": composite_agg_sources,
-                                "size": page_size,
-                                "after": res["my_composite"]["after_key"].clone(),
-                            }
-                        }
-                    });
-                    let agg_req: Aggregations = serde_json::from_value(agg_req_json).unwrap();
-                    let res = exec_request(agg_req.clone(), index).unwrap();
-                    assert_eq!(
-                        res["my_composite"]["buckets"],
-                        json!([]),
-                        "expected no buckets when using after_key from last page, query: {:?}",
-                        agg_req
-                    );
-                }
+                assert!(
+                    res["my_composite"].get("after_key").is_some(),
+                    "expected after_key on every non-empty page"
+                );
+                after_key = Some(res["my_composite"]["after_key"].clone());
             }
+            // Using the after_key from the last page must yield an empty page.
+            let agg_req_json = json!({
+                "my_composite": {
+                    "composite": {
+                        "sources": composite_agg_sources,
+                        "size": page_size,
+                        "after": after_key,
+                    }
+                }
+            });
+            let agg_req: Aggregations = serde_json::from_value(agg_req_json).unwrap();
+            let res = exec_request(agg_req.clone(), index).unwrap();
+            assert_eq!(
+                res["my_composite"]["buckets"],
+                json!([]),
+                "expected no buckets when using after_key from last page, query: {:?}",
+                agg_req
+            );
         }
     }
 
@@ -711,8 +707,28 @@ mod tests {
                 {"key": {"myterm": "terme"}, "doc_count": 1}
             ])
         );
-        assert!(res["my_composite"].get("after_key").is_none());
 
+        // paginating past last page should be empty
+        let agg_req_json = json!({
+            "my_composite": {
+                "composite": {
+                    "sources": [
+                        {"myterm": {"terms": {"field": "string_id"}}}
+                    ],
+                    "size": 3,
+                    "after":  &res["my_composite"]["after_key"]
+                }
+            }
+        });
+        let agg_req: Aggregations = serde_json::from_value(agg_req_json).unwrap();
+        let res = exec_request(agg_req.clone(), &index).unwrap();
+        assert!(res["my_composite"].get("after_key").is_none());
+        assert_eq!(
+            res["my_composite"]["buckets"],
+            json!([]),
+            "expected no buckets when using after_key from last page, query: {:?}",
+            agg_req
+        );
         Ok(())
     }
 
@@ -820,7 +836,10 @@ mod tests {
                 {"key": {"myterm": "apple"}, "doc_count": 1}
             ])
         );
-        assert!(res["fruity_aggreg"].get("after_key").is_none());
+        assert_eq!(
+            res["fruity_aggreg"]["after_key"],
+            json!({"myterm": "str:apple"})
+        );
 
         Ok(())
     }
@@ -1792,7 +1811,14 @@ mod tests {
                 {"key": {"month": ms_timestamp_from_iso_str("2021-02-01T00:00:00Z"), "category": "books"}, "doc_count": 1},
             ]),
         );
-        assert!(res["my_composite"].get("after_key").is_none());
+        let feb_2021_ns = ms_timestamp_from_iso_str("2021-02-01T00:00:00Z") * 1_000_000;
+        assert_eq!(
+            res["my_composite"]["after_key"],
+            json!({
+                "month": format!("dt:{}", feb_2021_ns),
+                "category": "str:books"
+            })
+        );
 
         Ok(())
     }

src/aggregation/bucket/filter.rs

@@ -674,6 +674,17 @@ impl<B: SubAggBuffer> SegmentAggregationCollector for SegmentFilterCollector<B>
         }
         Ok(())
     }
+
+    fn compute_metric_value(
+        &self,
+        _bucket_id: BucketId,
+        _sub_agg_name: &str,
+        _sub_agg_property: &str,
+        _agg_data: &AggregationsSegmentCtx,
+    ) -> Option<f64> {
+        // TODO: forward into the inner `sub_agg` for nested order paths (`filter.metric`).
+        None
+    }
 }
 
 /// Intermediate result for filter aggregation

src/aggregation/bucket/histogram/histogram.rs

@@ -394,6 +394,17 @@ impl SegmentAggregationCollector for SegmentHistogramCollector {
         }
         Ok(())
     }
+
+    fn compute_metric_value(
+        &self,
+        _bucket_id: BucketId,
+        _sub_agg_name: &str,
+        _sub_agg_property: &str,
+        _agg_data: &AggregationsSegmentCtx,
+    ) -> Option<f64> {
+        // Histogram is a multi-bucket agg with no single value to extract.
+        None
+    }
 }
 
 impl SegmentHistogramCollector {

src/aggregation/bucket/range.rs

@@ -328,6 +328,17 @@ impl<B: SubAggBuffer> SegmentAggregationCollector for SegmentRangeCollector<B> {
 
         Ok(())
     }
+
+    fn compute_metric_value(
+        &self,
+        _bucket_id: BucketId,
+        _sub_agg_name: &str,
+        _sub_agg_property: &str,
+        _agg_data: &AggregationsSegmentCtx,
+    ) -> Option<f64> {
+        // Range is a multi-bucket agg with no single value to extract.
+        None
+    }
 }
 /// Build a concrete `SegmentRangeCollector` with either a Vec- or HashMap-backed
 /// bucket storage, depending on the column type and aggregation level.

src/aggregation/bucket/term_agg.rs

@@ -352,19 +352,15 @@ pub(crate) fn build_segment_term_collector(
         )));
     }
 
-    // Validate sub aggregation exists when ordering by sub-aggregation.
-    {
-        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"
-                    ))
-                })?;
-        }
+    // Validate that the referenced sub-aggregation exists when ordering by one.
+    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"
+                ))
+            })?;
     }
 
     // Build sub-aggregation blueprint if there are children.
@@ -887,6 +883,17 @@ impl<TermMap: TermAggregationMap, B: SubAggBuffer> SegmentAggregationCollector
         }
         Ok(())
     }
+
+    fn compute_metric_value(
+        &self,
+        _bucket_id: BucketId,
+        _sub_agg_name: &str,
+        _sub_agg_property: &str,
+        _agg_data: &AggregationsSegmentCtx,
+    ) -> Option<f64> {
+        // Terms is a multi-bucket agg with no single value to extract.
+        None
+    }
 }
 
 /// Missing value are represented as a sentinel value in the column.
@@ -960,9 +967,6 @@ where
     ) -> crate::Result<IntermediateBucketResult> {
         let mut entries: Vec<(u64, Bucket)> = term_buckets.into_vec();
 
-        let order_by_sub_aggregation =
-            matches!(term_req.req.order.target, OrderTarget::SubAggregation(_));
-
         match &term_req.req.order.target {
             OrderTarget::Key => {
                 // We rely on the fact, that term ordinals match the order of the strings
@@ -974,10 +978,37 @@ where
                     entries.sort_unstable_by_key(|bucket| bucket.0);
                 }
             }
-            OrderTarget::SubAggregation(_name) => {
-                // don't sort and cut off since it's hard to make assumptions on the quality of the
-                // results when cutting off du to unknown nature of the sub_aggregation (possible
-                // to check).
+            OrderTarget::SubAggregation(sub_agg_path) => {
+                // Peek segment-level metric values, sort, then fall through to
+                // `cut_off_buckets`. Like Elasticsearch, we always cut off when ordering
+                // by a sub-agg: top-K results are approximate and may differ from the
+                // global ordering, especially for non-monotonic metrics like avg/min.
+                let coll = sub_agg_collector.as_deref().ok_or_else(|| {
+                    TantivyError::InvalidArgument(format!(
+                        "Could not find sub-aggregation collector for path {sub_agg_path}"
+                    ))
+                })?;
+                let (agg_name, agg_prop) = get_agg_name_and_property(sub_agg_path);
+                // Fetch values up-front; otherwise sort would re-compute per comparison
+                let mut keyed: Vec<(f64, (u64, Bucket))> = entries
+                    .into_iter()
+                    .map(|bucket| {
+                        let metric_value = coll
+                            .compute_metric_value(bucket.1.bucket_id, agg_name, agg_prop, agg_data)
+                            .unwrap_or(0.0);
+                        (metric_value, bucket)
+                    })
+                    .collect();
+                if term_req.req.order.order == Order::Desc {
+                    keyed.sort_unstable_by(|a, b| {
+                        b.0.partial_cmp(&a.0).unwrap_or(std::cmp::Ordering::Equal)
+                    });
+                } else {
+                    keyed.sort_unstable_by(|a, b| {
+                        a.0.partial_cmp(&b.0).unwrap_or(std::cmp::Ordering::Equal)
+                    });
+                }
+                entries = keyed.into_iter().map(|(_, e)| e).collect();
             }
             OrderTarget::Count => {
                 if term_req.req.order.order == Order::Desc {
@@ -988,11 +1019,8 @@ where
             }
         }
 
-        let (term_doc_count_before_cutoff, sum_other_doc_count) = if order_by_sub_aggregation {
-            (0, 0)
-        } else {
-            cut_off_buckets(&mut entries, term_req.req.segment_size as usize)
-        };
+        let (term_doc_count_before_cutoff, sum_other_doc_count) =
+            cut_off_buckets(&mut entries, term_req.req.segment_size as usize);
 
         let mut dict: FxHashMap<IntermediateKey, IntermediateTermBucketEntry> = Default::default();
         dict.reserve(entries.len());
@@ -1767,6 +1795,263 @@ mod tests {
         Ok(())
     }
 
+    #[test]
+    fn terms_aggregation_order_by_cardinality_desc_single_segment() -> crate::Result<()> {
+        terms_aggregation_order_by_cardinality_desc(true)
+    }
+    #[test]
+    fn terms_aggregation_order_by_cardinality_desc_multi_segment() -> crate::Result<()> {
+        terms_aggregation_order_by_cardinality_desc(false)
+    }
+    fn terms_aggregation_order_by_cardinality_desc(merge_segments: bool) -> crate::Result<()> {
+        // Distinct score values per bucket key: A→5, B→1, C→3.
+        // Order by cardinality desc must yield A, C, B.
+        let segment_and_terms = vec![vec![
+            (1.0, "A".to_string()),
+            (2.0, "A".to_string()),
+            (3.0, "A".to_string()),
+            (4.0, "A".to_string()),
+            (5.0, "A".to_string()),
+            (1.0, "B".to_string()),
+            (1.0, "B".to_string()),
+            (1.0, "B".to_string()),
+            (1.0, "C".to_string()),
+            (2.0, "C".to_string()),
+            (3.0, "C".to_string()),
+        ]];
+        let index = get_test_index_from_values_and_terms(merge_segments, &segment_and_terms)?;
+
+        let agg_req: Aggregations = serde_json::from_value(json!({
+            "my_texts": {
+                "terms": {
+                    "field": "string_id",
+                    "order": { "card": "desc" }
+                },
+                "aggs": {
+                    "card": { "cardinality": { "field": "score" } }
+                }
+            }
+        }))
+        .unwrap();
+
+        let res = exec_request(agg_req, &index)?;
+        assert_eq!(res["my_texts"]["buckets"][0]["key"], "A");
+        assert_eq!(res["my_texts"]["buckets"][0]["card"]["value"], 5.0);
+        assert_eq!(res["my_texts"]["buckets"][1]["key"], "C");
+        assert_eq!(res["my_texts"]["buckets"][1]["card"]["value"], 3.0);
+        assert_eq!(res["my_texts"]["buckets"][2]["key"], "B");
+        assert_eq!(res["my_texts"]["buckets"][2]["card"]["value"], 1.0);
+
+        // Asc engages the segment-cutoff path too (monotonic-safe: discarded buckets had
+        // local card >= cutoff, so merged card >= cutoff and they cannot be globally smallest).
+        let agg_req: Aggregations = serde_json::from_value(json!({
+            "my_texts": {
+                "terms": {
+                    "field": "string_id",
+                    "order": { "card": "asc" }
+                },
+                "aggs": {
+                    "card": { "cardinality": { "field": "score" } }
+                }
+            }
+        }))
+        .unwrap();
+        let res = exec_request(agg_req, &index)?;
+        assert_eq!(res["my_texts"]["buckets"][0]["key"], "B");
+        assert_eq!(res["my_texts"]["buckets"][1]["key"], "C");
+        assert_eq!(res["my_texts"]["buckets"][2]["key"], "A");
+
+        // size=2 with desc engages the segment cutoff: must keep top-2 by cardinality (A, C),
+        // and `sum_other_doc_count` reflects the dropped B (3 docs).
+        let agg_req: Aggregations = serde_json::from_value(json!({
+            "my_texts": {
+                "terms": {
+                    "field": "string_id",
+                    "size": 2,
+                    "order": { "card": "desc" }
+                },
+                "aggs": {
+                    "card": { "cardinality": { "field": "score" } }
+                }
+            }
+        }))
+        .unwrap();
+        let res = exec_request(agg_req, &index)?;
+        assert_eq!(res["my_texts"]["buckets"][0]["key"], "A");
+        assert_eq!(res["my_texts"]["buckets"][1]["key"], "C");
+        assert_eq!(res["my_texts"]["buckets"].as_array().unwrap().len(), 2);
+
+        // size=2 with asc engages the segment cutoff: must keep bottom-2 by cardinality (B, C).
+        let agg_req: Aggregations = serde_json::from_value(json!({
+            "my_texts": {
+                "terms": {
+                    "field": "string_id",
+                    "size": 2,
+                    "order": { "card": "asc" }
+                },
+                "aggs": {
+                    "card": { "cardinality": { "field": "score" } }
+                }
+            }
+        }))
+        .unwrap();
+        let res = exec_request(agg_req, &index)?;
+        assert_eq!(res["my_texts"]["buckets"][0]["key"], "B");
+        assert_eq!(res["my_texts"]["buckets"][1]["key"], "C");
+        assert_eq!(res["my_texts"]["buckets"].as_array().unwrap().len(), 2);
+
+        Ok(())
+    }
+
+    #[test]
+    fn terms_aggregation_order_by_sum_single_segment() -> crate::Result<()> {
+        terms_aggregation_order_by_sum(true)
+    }
+    #[test]
+    fn terms_aggregation_order_by_sum_multi_segment() -> crate::Result<()> {
+        terms_aggregation_order_by_sum(false)
+    }
+    fn terms_aggregation_order_by_sum(merge_segments: bool) -> crate::Result<()> {
+        // Per-bucket sums on the U64 `score` column (non-negative => sum is monotonic):
+        //   A → 1+2+3+4+5 = 15, B → 1+1+1 = 3, C → 1+2+3 = 6.
+        let segment_and_terms = vec![
+            vec![
+                (1.0, "A".to_string()),
+                (2.0, "A".to_string()),
+                (3.0, "A".to_string()),
+                (1.0, "B".to_string()),
+                (1.0, "C".to_string()),
+            ],
+            vec![
+                (4.0, "A".to_string()),
+                (5.0, "A".to_string()),
+                (1.0, "B".to_string()),
+                (1.0, "B".to_string()),
+                (2.0, "C".to_string()),
+                (3.0, "C".to_string()),
+            ],
+        ];
+        let index = get_test_index_from_values_and_terms(merge_segments, &segment_and_terms)?;
+
+        // Desc on a Sum metric engages the fast path (column is U64).
+        let agg_req: Aggregations = serde_json::from_value(json!({
+            "my_texts": {
+                "terms": {
+                    "field": "string_id",
+                    "order": { "total": "desc" }
+                },
+                "aggs": {
+                    "total": { "sum": { "field": "score" } }
+                }
+            }
+        }))
+        .unwrap();
+        let res = exec_request(agg_req, &index)?;
+        assert_eq!(res["my_texts"]["buckets"][0]["key"], "A");
+        assert_eq!(res["my_texts"]["buckets"][0]["total"]["value"], 15.0);
+        assert_eq!(res["my_texts"]["buckets"][1]["key"], "C");
+        assert_eq!(res["my_texts"]["buckets"][1]["total"]["value"], 6.0);
+        assert_eq!(res["my_texts"]["buckets"][2]["key"], "B");
+        assert_eq!(res["my_texts"]["buckets"][2]["total"]["value"], 3.0);
+
+        // Asc engages the fast path too — discarded buckets had local sum >= cutoff,
+        // and merged sum >= local (non-negative addends), so they cannot be globally smallest.
+        let agg_req: Aggregations = serde_json::from_value(json!({
+            "my_texts": {
+                "terms": {
+                    "field": "string_id",
+                    "order": { "total": "asc" }
+                },
+                "aggs": {
+                    "total": { "sum": { "field": "score" } }
+                }
+            }
+        }))
+        .unwrap();
+        let res = exec_request(agg_req, &index)?;
+        assert_eq!(res["my_texts"]["buckets"][0]["key"], "B");
+        assert_eq!(res["my_texts"]["buckets"][1]["key"], "C");
+        assert_eq!(res["my_texts"]["buckets"][2]["key"], "A");
+
+        // size=2 desc with cutoff: top-2 by sum (A, C).
+        let agg_req: Aggregations = serde_json::from_value(json!({
+            "my_texts": {
+                "terms": {
+                    "field": "string_id",
+                    "size": 2,
+                    "order": { "total": "desc" }
+                },
+                "aggs": {
+                    "total": { "sum": { "field": "score" } }
+                }
+            }
+        }))
+        .unwrap();
+        let res = exec_request(agg_req, &index)?;
+        assert_eq!(res["my_texts"]["buckets"][0]["key"], "A");
+        assert_eq!(res["my_texts"]["buckets"][1]["key"], "C");
+        assert_eq!(res["my_texts"]["buckets"].as_array().unwrap().len(), 2);
+
+        // Stats sub-property: ordering by `mystats.sum` on a U64 column also engages.
+        let agg_req: Aggregations = serde_json::from_value(json!({
+            "my_texts": {
+                "terms": {
+                    "field": "string_id",
+                    "order": { "mystats.sum": "desc" }
+                },
+                "aggs": {
+                    "mystats": { "stats": { "field": "score" } }
+                }
+            }
+        }))
+        .unwrap();
+        let res = exec_request(agg_req, &index)?;
+        assert_eq!(res["my_texts"]["buckets"][0]["key"], "A");
+        assert_eq!(res["my_texts"]["buckets"][1]["key"], "C");
+        assert_eq!(res["my_texts"]["buckets"][2]["key"], "B");
+
+        // Sum on a signed column (I64) takes the same cutoff path. Results may be
+        // approximate near the boundary on adversarial data, but for this dataset the
+        // top-K is unambiguous.
+        let agg_req: Aggregations = serde_json::from_value(json!({
+            "my_texts": {
+                "terms": {
+                    "field": "string_id",
+                    "order": { "total": "desc" }
+                },
+                "aggs": {
+                    "total": { "sum": { "field": "score_i64" } }
+                }
+            }
+        }))
+        .unwrap();
+        let res = exec_request(agg_req, &index)?;
+        assert_eq!(res["my_texts"]["buckets"][0]["key"], "A");
+        assert_eq!(res["my_texts"]["buckets"][1]["key"], "C");
+        assert_eq!(res["my_texts"]["buckets"][2]["key"], "B");
+
+        // Order by extended_stats sub-property exercises compute_metric_value on the
+        // ExtendedStats collector. A→max=5, B→max=1, C→max=3, so desc by max → A, C, B.
+        let agg_req: Aggregations = serde_json::from_value(json!({
+            "my_texts": {
+                "terms": {
+                    "field": "string_id",
+                    "order": { "ext.max": "desc" }
+                },
+                "aggs": {
+                    "ext": { "extended_stats": { "field": "score" } }
+                }
+            }
+        }))
+        .unwrap();
+        let res = exec_request(agg_req, &index)?;
+        assert_eq!(res["my_texts"]["buckets"][0]["key"], "A");
+        assert_eq!(res["my_texts"]["buckets"][1]["key"], "C");
+        assert_eq!(res["my_texts"]["buckets"][2]["key"], "B");
+
+        Ok(())
+    }
+
     #[test]
     fn terms_aggregation_test_order_key_single_segment() -> crate::Result<()> {
         terms_aggregation_test_order_key_merge_segment(true)

src/aggregation/bucket/term_missing_agg.rs

@@ -177,6 +177,17 @@ impl SegmentAggregationCollector for TermMissingAgg {
         }
         Ok(())
     }
+
+    fn compute_metric_value(
+        &self,
+        _bucket_id: BucketId,
+        _sub_agg_name: &str,
+        _sub_agg_property: &str,
+        _agg_data: &AggregationsSegmentCtx,
+    ) -> Option<f64> {
+        // TODO: forward to `sub_agg` for nested order paths (`missing_agg>metric`).
+        None
+    }
 }
 
 #[cfg(test)]

src/aggregation/intermediate_agg_result.rs

@@ -1004,24 +1004,20 @@ impl IntermediateCompositeBucketResult {
     ) -> 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 after_key = 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_or_default();
 
         let buckets = trimmed_entry_vec
             .into_iter()

src/aggregation/metric/cardinality.rs

@@ -4,6 +4,7 @@ use std::io;
 
 use columnar::column_values::CompactSpaceU64Accessor;
 use columnar::{Column, ColumnType, Dictionary, StrColumn};
+use common::{BitSet, TinySet};
 use datasketches::hll::{Coupon, HllSketch, HllType, HllUnion};
 use rustc_hash::{FxBuildHasher, FxHashMap, FxHashSet};
 use serde::{Deserialize, Deserializer, Serialize, Serializer};
@@ -20,6 +21,12 @@ use crate::TantivyError;
 /// 2^11 = 2048 registers, giving ~2.3% relative error and ~1KB per sketch (Hll4).
 const LG_K: u8 = 11;
 
+/// Promote FxHashSet<u64> -> PagedBitset at ~3% density (`len * 32 >
+/// dict_num_terms`). Past this point the bitset (~`dict_num_terms / 7.5`
+/// bytes) is smaller than the hashset (~10 B/entry minimum) and avoids
+/// the per-insert hash.
+const PROMOTION_RATIO: u64 = 32;
+
 /// # Cardinality
 ///
 /// The cardinality aggregation allows for computing an estimate
@@ -177,9 +184,263 @@ impl CouponCache {
     }
 }
 
-pub(crate) struct SegmentCardinalityCollector {
+// =================================================================
+// PagedBitset: a sparse bitset indexed by term_ord.
+//
+// Used as the dense alternative to FxHashSet<u64> once a string
+// cardinality bucket has accumulated enough unique term ordinals.
+// Memory is bounded to (touched pages) * (page bytes), not
+// (max_term_ord / 8).
+//
+// Page geometry mirrors `PagedTermMap` in `term_agg.rs`: 1024 ords
+// per page, lazy `Vec<Option<Box<Page>>>` directory.
+// =================================================================
+const BITSET_PAGE_SHIFT: u32 = 10;
+const BITSET_PAGE_BITS: u64 = 1u64 << BITSET_PAGE_SHIFT; // 1024
+const BITSET_PAGE_MASK: u64 = BITSET_PAGE_BITS - 1;
+const BITSET_WORDS_PER_PAGE: usize = (BITSET_PAGE_BITS / 64) as usize; // 16
+
+#[derive(Clone)]
+struct PagedBitsetPage {
+    words: [TinySet; BITSET_WORDS_PER_PAGE],
+}
+
+impl PagedBitsetPage {
+    fn new() -> Self {
+        Self {
+            words: [TinySet::empty(); BITSET_WORDS_PER_PAGE],
+        }
+    }
+}
+
+pub(crate) struct PagedBitset {
+    pages: Vec<Option<Box<PagedBitsetPage>>>,
+    /// Cached number of set bits, maintained on insert.
+    count: u64,
+}
+
+impl PagedBitset {
+    /// Allocates a directory big enough to hold ords up to and including
+    /// `max_term_ord`. Pages are allocated lazily on first set.
+    fn with_max_term_ord(max_term_ord: u64) -> Self {
+        let max_page_idx = (max_term_ord >> BITSET_PAGE_SHIFT) as usize;
+        let num_pages = max_page_idx + 1;
+        Self {
+            pages: vec![None; num_pages],
+            count: 0,
+        }
+    }
+
+    #[inline]
+    fn insert(&mut self, term_ord: u64) {
+        let page_idx = (term_ord >> BITSET_PAGE_SHIFT) as usize;
+        let intra = term_ord & BITSET_PAGE_MASK;
+        let word_idx = (intra >> 6) as usize;
+        let bit_idx = (intra & 63) as u32;
+
+        let page = match &mut self.pages[page_idx] {
+            Some(p) => p,
+            None => {
+                self.pages[page_idx] = Some(Box::new(PagedBitsetPage::new()));
+                self.pages[page_idx].as_mut().unwrap()
+            }
+        };
+        if page.words[word_idx].insert_mut(bit_idx) {
+            self.count += 1;
+        }
+    }
+
+    /// Number of set bits. O(1).
+    #[inline]
+    fn len(&self) -> u64 {
+        self.count
+    }
+
+    /// Iterate set ords in ascending order.
+    fn iter_sorted(&self) -> impl Iterator<Item = u64> + '_ {
+        self.pages
+            .iter()
+            .enumerate()
+            .filter_map(|(page_idx, page_opt)| page_opt.as_ref().map(|p| (page_idx, p)))
+            .flat_map(|(page_idx, page)| {
+                let page_base_ord = (page_idx as u64) << BITSET_PAGE_SHIFT;
+                page.words
+                    .iter()
+                    .enumerate()
+                    .flat_map(move |(word_idx, &word)| {
+                        let word_base_ord = page_base_ord + (word_idx as u64) * 64;
+                        word.into_iter()
+                            .map(move |bit| word_base_ord + u64::from(bit))
+                    })
+            })
+    }
+}
+
+/// Threshold below which we use `BitSet` instead of `TermOrdSet`.
+///
+/// Both `BitSet` and `FxHashSet<u64>` have the same 32-byte struct, so the comparison is heap only:
+///   * `BitSet` at T=256: 5 `TinySet` words covering 258 bits (with the missing-value sentinel) =
+///     40 bytes.
+///   * `FxHashSet<u64>` after one insert: 4-bucket hashbrown table ≈ 56 bytes
+pub(crate) const BITSET_MAX_TERM_ORD: u64 = 256;
+
+// =================================================================
+// TermOrdAccumulator: per-bucket abstraction over the entries set.
+//
+// Implementations:
+//   - `BitSet` (from `common`): used when `column.max_value()` is small (< BITSET_MAX_TERM_ORD).
+//     Pre-allocated, no promotion.
+//   - `TermOrdSet`: adaptive, starts as FxHashSet and promotes to a paged bitset when occupancy
+//     crosses the density threshold (only if promotion is enabled — typically gated on top-level
+//     aggregation).
+//
+// The trait lets `SegmentCardinalityCollector` be generic over the choice
+// so the hot collect() loop monomorphizes to a direct call (no enum
+// dispatch per insert).
+// =================================================================
+pub(crate) trait TermOrdAccumulator: Sized {
+    /// Construct an empty accumulator.
+    /// `max_term_ord_inclusive` is the largest term_ord that may be
+    /// inserted (used to size pre-allocated bitsets and the dense bitset
+    /// on promotion).
+    fn new(max_term_ord_inclusive: u64) -> Self;
+    fn insert(&mut self, term_ord: u64);
+    /// Bulk insert. Implementations may override to hoist any inner
+    /// dispatch outside the loop. Default loops `insert`.
+    #[inline]
+    fn extend_from_iter<I: IntoIterator<Item = u64>>(&mut self, ords: I) {
+        for ord in ords {
+            self.insert(ord);
+        }
+    }
+    /// Hook called once per ingested block. Adaptive impls use this to
+    /// decide on sparse->dense promotion.
+    fn maybe_compact(&mut self) {}
+    fn len(&self) -> usize;
+    fn iter_ords(&self) -> impl Iterator<Item = u64> + '_;
+}
+
+impl TermOrdAccumulator for BitSet {
+    #[inline]
+    fn new(max_term_ord_inclusive: u64) -> Self {
+        // `BitSet::with_max_value(M)` accepts ords in [0, M).
+        // We need ords up to and including `max_term_ord_inclusive`, plus
+        // the missing-value sentinel `column.max_value() + 1`.
+        BitSet::with_max_value((max_term_ord_inclusive + 2) as u32)
+    }
+    #[inline]
+    fn insert(&mut self, term_ord: u64) {
+        BitSet::insert(self, term_ord as u32);
+    }
+    #[inline]
+    fn len(&self) -> usize {
+        BitSet::len(self)
+    }
+    fn iter_ords(&self) -> impl Iterator<Item = u64> + '_ {
+        // `BitSet` itself doesn't expose iteration, but
+        // `BitSet::tinyset(bucket)` does. Walk per-bucket and yield each
+        // set bit. The capacity is `max_value()`; iterating to
+        // `div_ceil(64)` covers every possible ord exactly once.
+        let num_buckets = self.max_value().div_ceil(64);
+        (0..num_buckets).flat_map(move |bucket| {
+            let chunk_base = u64::from(bucket) * 64;
+            self.tinyset(bucket)
+                .into_iter()
+                .map(move |bit| chunk_base + u64::from(bit))
+        })
+    }
+}
+
+// =================================================================
+// TermOrdSet: adaptive sparse->dense accumulator.
+//
+// Starts as an FxHashSet (cheap when few ords are seen). When occupancy
+// crosses `len * PROMOTION_RATIO > max_term_ord_inclusive`, drains into
+// a `PagedBitset` and continues dense. Promotion is one-way.
+// =================================================================
+pub(crate) struct TermOrdSet {
+    inner: TermOrdSetInner,
+    /// Largest term_ord that may be inserted. Used for both sizing the
+    /// dense bitset on promotion and as the promotion-threshold reference.
+    max_term_ord_inclusive: u64,
+}
+
+enum TermOrdSetInner {
+    Sparse(FxHashSet<u64>),
+    Dense(PagedBitset),
+}
+
+impl TermOrdAccumulator for TermOrdSet {
+    fn new(max_term_ord_inclusive: u64) -> Self {
+        Self {
+            inner: TermOrdSetInner::Sparse(FxHashSet::default()),
+            max_term_ord_inclusive,
+        }
+    }
+
+    #[inline]
+    fn insert(&mut self, term_ord: u64) {
+        match &mut self.inner {
+            TermOrdSetInner::Sparse(set) => {
+                set.insert(term_ord);
+            }
+            TermOrdSetInner::Dense(bitset) => bitset.insert(term_ord),
+        }
+    }
+
+    /// Hoist the Sparse/Dense match outside the per-ord loop so that a
+    /// block of inserts dispatches once.
+    fn extend_from_iter<I: IntoIterator<Item = u64>>(&mut self, ords: I) {
+        match &mut self.inner {
+            TermOrdSetInner::Sparse(set) => {
+                for ord in ords {
+                    set.insert(ord);
+                }
+            }
+            TermOrdSetInner::Dense(bitset) => {
+                for ord in ords {
+                    bitset.insert(ord);
+                }
+            }
+        }
+    }
+
+    fn maybe_compact(&mut self) {
+        let TermOrdSetInner::Sparse(set) = &mut self.inner else {
+            return;
+        };
+        if set.len() as u64 * PROMOTION_RATIO <= self.max_term_ord_inclusive {
+            return;
+        }
+        // Size for ord <= max_term_ord_inclusive plus the missing sentinel
+        // (column.max_value() + 1, which may equal max_term_ord_inclusive
+        // when the column references every dictionary term).
+        let mut bitset = PagedBitset::with_max_term_ord(self.max_term_ord_inclusive + 1);
+        let set = std::mem::take(set);
+        for ord in set {
+            bitset.insert(ord);
+        }
+        self.inner = TermOrdSetInner::Dense(bitset);
+    }
+
+    fn len(&self) -> usize {
+        match &self.inner {
+            TermOrdSetInner::Sparse(set) => set.len(),
+            TermOrdSetInner::Dense(bitset) => bitset.len() as usize,
+        }
+    }
+
+    fn iter_ords(&self) -> impl Iterator<Item = u64> + '_ {
+        match &self.inner {
+            TermOrdSetInner::Sparse(set) => itertools::Either::Left(set.iter().copied()),
+            TermOrdSetInner::Dense(bitset) => itertools::Either::Right(bitset.iter_sorted()),
+        }
+    }
+}
+
+pub(crate) struct SegmentCardinalityCollector<S: TermOrdAccumulator> {
     /// Buckets are Some(_) until they get consumed by into_intermediate_results().
-    buckets: Vec<Option<SegmentCardinalityCollectorBucket>>,
+    buckets: Vec<Option<SegmentCardinalityCollectorBucket<S>>>,
     accessor_idx: usize,
     /// The column accessor to access the fast field values.
     accessor: Column<u64>,
@@ -188,9 +449,13 @@ pub(crate) struct SegmentCardinalityCollector {
     /// The missing value normalized to the internal u64 representation of the field type.
     missing_value_for_accessor: Option<u64>,
     coupon_cache: Option<CouponCache>,
+    /// Largest term_ord that may be inserted into a bucket. For str columns
+    /// this is `accessor.max_value()`; for non-str columns this is unused
+    /// (no inserts go into `entries`) and set to 0.
+    max_term_ord_inclusive: u64,
 }
 
-impl Debug for SegmentCardinalityCollector {
+impl<S: TermOrdAccumulator> Debug for SegmentCardinalityCollector<S> {
     fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result {
         f.debug_struct("SegmentCardinalityCollector")
             .field("column_type", &self.column_type)
@@ -202,16 +467,21 @@ impl Debug for SegmentCardinalityCollector {
     }
 }
 
-pub(crate) struct SegmentCardinalityCollectorBucket {
-    cardinality: CardinalityCollector,
-    entries: FxHashSet<u64>,
+/// Per-bucket state. Shape depends on column kind: str columns dedup
+/// term ords and only build the HLL sketch at finalization (saves the
+/// ~96 B `CardinalityCollector` per bucket during collect); numeric/IpAddr
+/// columns feed the sketch directly during collect.
+pub(crate) enum SegmentCardinalityCollectorBucket<S: TermOrdAccumulator> {
+    Str(S),
+    Numeric(CardinalityCollector),
 }
-impl SegmentCardinalityCollectorBucket {
+impl<S: TermOrdAccumulator> SegmentCardinalityCollectorBucket<S> {
     #[inline(always)]
-    pub fn new(column_type: ColumnType) -> Self {
-        Self {
-            cardinality: CardinalityCollector::new(column_type as u8),
-            entries: FxHashSet::default(),
+    pub fn new(column_type: ColumnType, max_term_ord_inclusive: u64) -> Self {
+        if column_type == ColumnType::Str {
+            Self::Str(S::new(max_term_ord_inclusive))
+        } else {
+            Self::Numeric(CardinalityCollector::new(column_type as u8))
         }
     }
 
@@ -222,37 +492,57 @@ impl SegmentCardinalityCollectorBucket {
     //
     // If the column is str, then the values are dictionary encoded
     // and have not been added to the sketch yet.
-    // We need to resolves the term ords accumulated in self.entries
-    // with the coupon cache, and append the results to the sketch.
+    // We need to resolves the term ords accumulated in the str entries
+    // with the coupon cache, and append the results to a fresh sketch.
     fn into_intermediate_metric_result(
-        mut self,
+        self,
         coupon_cache_opt: Option<&CouponCache>,
     ) -> crate::Result<IntermediateMetricResult> {
-        if let Some(coupon_cache) = coupon_cache_opt {
-            assert!(self.cardinality.sketch.is_empty());
-            append_to_sketch(&self.entries, coupon_cache, &mut self.cardinality);
-        }
-        Ok(IntermediateMetricResult::Cardinality(self.cardinality))
+        let cardinality = match self {
+            Self::Str(entries) => {
+                let mut cardinality = CardinalityCollector::new(ColumnType::Str as u8);
+                if let Some(coupon_cache) = coupon_cache_opt {
+                    // Sketch must be empty for str columns: coupons are appended here
+                    // from the term_ord set (and not directly during collection).
+                    assert!(cardinality.sketch.is_empty());
+                    append_to_sketch(&entries, coupon_cache, &mut cardinality);
+                }
+                cardinality
+            }
+            Self::Numeric(cardinality) => cardinality,
+        };
+        Ok(IntermediateMetricResult::Cardinality(cardinality))
     }
 }
 
 /// Builds a coupon cache from the given buckets, dictionary, and optional missing value.
 /// Returns a mapping from term_ord to the hash (coupon) of the associated term.
-fn build_coupon_cache(
-    buckets: &[Option<SegmentCardinalityCollectorBucket>],
+fn build_coupon_cache<S: TermOrdAccumulator>(
+    buckets: &[Option<SegmentCardinalityCollectorBucket<S>>],
     dictionary: &Dictionary,
     missing_value_opt: Option<&Key>,
 ) -> io::Result<CouponCache> {
-    let term_ords_capacity: usize = buckets
-        .iter()
-        .flatten()
-        .map(|bucket| bucket.entries.len())
-        .max()
-        .unwrap_or(0)
-        * 2;
-    let mut term_ords_set = FxHashSet::with_capacity_and_hasher(term_ords_capacity, FxBuildHasher);
+    // Caller restricts this to str cardinality collectors, so every
+    // present bucket must be the `Str` variant. Pass 1 validates and
+    // computes the capacity hint; pass 2 inserts.
+    let mut max_bucket_len = 0usize;
     for bucket in buckets.iter().flatten() {
-        term_ords_set.extend(bucket.entries.iter().copied());
+        match bucket {
+            SegmentCardinalityCollectorBucket::Str(entries) => {
+                max_bucket_len = max_bucket_len.max(entries.len());
+            }
+            SegmentCardinalityCollectorBucket::Numeric(_) => {
+                return Err(io::Error::other(
+                    "build_coupon_cache invoked with a non-str bucket",
+                ));
+            }
+        }
+    }
+    let mut term_ords_set = FxHashSet::with_capacity_and_hasher(max_bucket_len * 2, FxBuildHasher);
+    for bucket in buckets.iter().flatten() {
+        if let SegmentCardinalityCollectorBucket::Str(entries) = bucket {
+            term_ords_set.extend(entries.iter_ords());
+        }
     }
     let mut term_ords: Vec<u64> = term_ords_set.into_iter().collect();
     term_ords.sort_unstable();
@@ -284,8 +574,8 @@ fn build_coupon_cache(
     Ok(CouponCache::new(term_ords, coupons, missing_coupon_opt))
 }
 
-fn append_to_sketch(
-    term_ords: &FxHashSet<u64>,
+fn append_to_sketch<S: TermOrdAccumulator>(
+    term_ords: &S,
     coupon_cache: &CouponCache,
     sketch: &mut CardinalityCollector,
 ) {
@@ -294,7 +584,7 @@ fn append_to_sketch(
             coupon_map,
             missing_coupon_opt,
         } => {
-            for &term_ord in term_ords {
+            for term_ord in term_ords.iter_ords() {
                 if let Some(coupon) = coupon_map
                     .get(term_ord as usize)
                     .copied()
@@ -308,8 +598,8 @@ fn append_to_sketch(
             coupon_map,
             missing_coupon_opt,
         } => {
-            for term_ord in term_ords {
-                if let Some(coupon) = coupon_map.get(term_ord).copied().or(*missing_coupon_opt) {
+            for term_ord in term_ords.iter_ords() {
+                if let Some(coupon) = coupon_map.get(&term_ord).copied().or(*missing_coupon_opt) {
                     sketch.insert_coupon(coupon);
                 }
             }
@@ -317,12 +607,13 @@ fn append_to_sketch(
     }
 }
 
-impl SegmentCardinalityCollector {
+impl<S: TermOrdAccumulator> SegmentCardinalityCollector<S> {
     pub fn from_req(
         column_type: ColumnType,
         accessor_idx: usize,
         accessor: Column<u64>,
         missing_value_for_accessor: Option<u64>,
+        max_term_ord_inclusive: u64,
     ) -> Self {
         Self {
             buckets: Vec::new(),
@@ -331,6 +622,7 @@ impl SegmentCardinalityCollector {
             accessor,
             missing_value_for_accessor,
             coupon_cache: None,
+            max_term_ord_inclusive,
         }
     }
 
@@ -347,7 +639,9 @@ impl SegmentCardinalityCollector {
     }
 }
 
-impl SegmentAggregationCollector for SegmentCardinalityCollector {
+impl<S: TermOrdAccumulator + 'static> SegmentAggregationCollector
+    for SegmentCardinalityCollector<S>
+{
     fn add_intermediate_aggregation_result(
         &mut self,
         agg_data: &AggregationsSegmentCtx,
@@ -402,31 +696,41 @@ impl SegmentAggregationCollector for SegmentCardinalityCollector {
             ));
         };
         let col_block_accessor = &agg_data.column_block_accessor;
-        if self.column_type == ColumnType::Str {
-            for term_ord in col_block_accessor.iter_vals() {
-                bucket.entries.insert(term_ord);
+        match bucket {
+            SegmentCardinalityCollectorBucket::Str(entries) => {
+                // Promotion check runs on the pre-block state: the first call
+                // sees an empty set (no-op), and the last block of inserts
+                // doesn't trigger a promotion of a set we won't grow further.
+                // The trait dispatches once per block (via `extend_from_iter`)
+                // for adaptive variants and inlines to a tight loop for the
+                // BitSet path.
+                entries.maybe_compact();
+                entries.extend_from_iter(col_block_accessor.iter_vals());
             }
-        } else if self.column_type == ColumnType::IpAddr {
-            let compact_space_accessor = self
-                .accessor
-                .values
-                .clone()
-                .downcast_arc::<CompactSpaceU64Accessor>()
-                .map_err(|_| {
-                    TantivyError::AggregationError(
-                        crate::aggregation::AggregationError::InternalError(
-                            "Type mismatch: Could not downcast to CompactSpaceU64Accessor"
-                                .to_string(),
-                        ),
-                    )
-                })?;
-            for val in col_block_accessor.iter_vals() {
-                let val: u128 = compact_space_accessor.compact_to_u128(val as u32);
-                bucket.cardinality.insert(val);
-            }
-        } else {
-            for val in col_block_accessor.iter_vals() {
-                bucket.cardinality.insert(val);
+            SegmentCardinalityCollectorBucket::Numeric(cardinality) => {
+                if self.column_type == ColumnType::IpAddr {
+                    let compact_space_accessor = self
+                        .accessor
+                        .values
+                        .clone()
+                        .downcast_arc::<CompactSpaceU64Accessor>()
+                        .map_err(|_| {
+                            TantivyError::AggregationError(
+                                crate::aggregation::AggregationError::InternalError(
+                                    "Type mismatch: Could not downcast to CompactSpaceU64Accessor"
+                                        .to_string(),
+                                ),
+                            )
+                        })?;
+                    for val in col_block_accessor.iter_vals() {
+                        let val: u128 = compact_space_accessor.compact_to_u128(val as u32);
+                        cardinality.insert(val);
+                    }
+                } else {
+                    for val in col_block_accessor.iter_vals() {
+                        cardinality.insert(val);
+                    }
+                }
             }
         }
 
@@ -439,12 +743,40 @@ impl SegmentAggregationCollector for SegmentCardinalityCollector {
         _agg_data: &AggregationsSegmentCtx,
     ) -> crate::Result<()> {
         if max_bucket as usize >= self.buckets.len() {
+            let column_type = self.column_type;
+            let max_term_ord_inclusive = self.max_term_ord_inclusive;
             self.buckets.resize_with(max_bucket as usize + 1, || {
-                Some(SegmentCardinalityCollectorBucket::new(self.column_type))
+                Some(SegmentCardinalityCollectorBucket::<S>::new(
+                    column_type,
+                    max_term_ord_inclusive,
+                ))
             });
         }
         Ok(())
     }
+
+    fn compute_metric_value(
+        &self,
+        bucket_id: BucketId,
+        sub_agg_name: &str,
+        sub_agg_property: &str,
+        agg_data: &AggregationsSegmentCtx,
+    ) -> Option<f64> {
+        let req_data = &agg_data.get_cardinality_req_data(self.accessor_idx);
+        if req_data.name != sub_agg_name || !sub_agg_property.is_empty() {
+            return None;
+        }
+        let bucket = self.buckets.get(bucket_id as usize)?.as_ref()?;
+        // For string columns the sketch isn't built until finalization; the
+        // term_ord set's len is the exact distinct count. For numeric columns
+        // the sketch is populated during collect.
+        match bucket {
+            SegmentCardinalityCollectorBucket::Str(entries) => Some(entries.len() as f64),
+            SegmentCardinalityCollectorBucket::Numeric(cardinality) => {
+                Some(cardinality.sketch.estimate().trunc())
+            }
+        }
+    }
 }
 
 #[derive(Clone, Debug)]
@@ -592,6 +924,134 @@ mod tests {
         Ok(())
     }
 
+    /// Build a single-segment string-cardinality index with 32 unique terms.
+    /// `column.max_value() = 31` is well below `BITSET_MAX_TERM_ORD`,
+    /// so the bucket exercises the `BitSet` path end to end.
+    #[test]
+    fn cardinality_aggregation_test_str_bitset() -> crate::Result<()> {
+        let terms: Vec<String> = (0..32).map(|i| format!("term_{i}")).collect();
+        let term_refs: Vec<Vec<&str>> = terms.iter().map(|t| vec![t.as_str()]).collect::<Vec<_>>();
+        // single segment so we have a single dictionary of 32 terms.
+        let index = get_test_index_from_terms(true, &term_refs)?;
+
+        let agg_req: Aggregations = serde_json::from_value(json!({
+            "cardinality": {
+                "cardinality": { "field": "string_id" }
+            },
+        }))
+        .unwrap();
+
+        let res = exec_request(agg_req, &index)?;
+        assert_eq!(res["cardinality"]["value"], 32.0);
+        Ok(())
+    }
+
+    /// `BitSet` path with a `missing` parameter: the column-level missing
+    /// sentinel (`column.max_value() + 1`) flows into the bitset, the
+    /// dict lookup filter at finalization drops it, and the missing
+    /// coupon is applied separately.
+    #[test]
+    fn cardinality_aggregation_test_str_bitset_with_missing() {
+        let mut schema_builder = Schema::builder();
+        let name_field = schema_builder.add_text_field("name", STRING | FAST);
+        let index = Index::create_in_ram(schema_builder.build());
+        let mut writer = index.writer_for_tests().unwrap();
+        for i in 0..16 {
+            let term = format!("t{i:02}");
+            writer.add_document(doc!(name_field => term)).unwrap();
+        }
+        // One empty doc, exercising the missing sentinel.
+        writer.add_document(doc!()).unwrap();
+        writer.commit().unwrap();
+
+        let agg_req: Aggregations = serde_json::from_value(json!({
+            "cardinality": {
+                "cardinality": {
+                    "field": "name",
+                    "missing": "MISSING_SENTINEL_KEY",
+                }
+            },
+        }))
+        .unwrap();
+
+        let res = exec_request(agg_req, &index).unwrap();
+        // 16 distinct real terms + 1 distinct "missing" value = 17.
+        assert_eq!(res["cardinality"]["value"], 17.0);
+    }
+
+    /// Unit-test the PagedBitset itself: cross-page inserts produce sorted
+    /// iteration, len() matches the inserted set, and duplicates are
+    /// idempotent.
+    #[test]
+    fn paged_bitset_basic() {
+        use super::PagedBitset;
+        // Span several pages: BITSET_PAGE_BITS = 1024, so ords > 1024 land
+        // on the second page, > 2048 on the third, etc.
+        let ords = [0u64, 1, 63, 64, 1023, 1024, 1025, 4096, 4097, 9999, 10_000];
+        let max_ord = *ords.iter().max().unwrap();
+        let mut bitset = PagedBitset::with_max_term_ord(max_ord);
+        for &ord in &ords {
+            bitset.insert(ord);
+            // Idempotent: inserting again must not increase count.
+            bitset.insert(ord);
+        }
+        assert_eq!(bitset.len(), ords.len() as u64);
+        let collected: Vec<u64> = bitset.iter_sorted().collect();
+        let mut expected: Vec<u64> = ords.to_vec();
+        expected.sort_unstable();
+        assert_eq!(collected, expected);
+    }
+
+    /// Unit-test `TermOrdSet`: starts Sparse, promotes to Dense on
+    /// `maybe_compact` once the density threshold is crossed, and
+    /// `iter_ords()` yields the same set in either state. Ords spanning
+    /// multiple paged-bitset pages exercise the Dense iter ordering.
+    #[test]
+    fn term_ord_set_promotes_on_maybe_compact() {
+        use super::{TermOrdAccumulator, TermOrdSet, PROMOTION_RATIO};
+        // Pick max so promotion needs few inserts: len * RATIO > max with
+        // RATIO=32 and max=64 trips at len=3 (3*32=96 > 64).
+        let max_term_ord = 64u64;
+        let mut set = <TermOrdSet as TermOrdAccumulator>::new(max_term_ord);
+        // Two inserts: should stay Sparse after maybe_compact (2 * RATIO = 64, not > 64).
+        set.insert(0);
+        set.insert(7);
+        set.maybe_compact();
+        assert_eq!(set.len(), 2);
+
+        // Third insert promotes on next maybe_compact.
+        set.insert(20);
+        assert_eq!(set.len(), 3);
+        // Sanity check: at len=3, 3 * PROMOTION_RATIO = 96 > 64.
+        assert!(3u64 * PROMOTION_RATIO > max_term_ord);
+        set.maybe_compact();
+
+        // Post-promotion: extending continues to work.
+        set.insert(15);
+        set.insert(15); // dup
+        assert_eq!(set.len(), 4);
+
+        let mut collected: Vec<u64> = set.iter_ords().collect();
+        collected.sort_unstable();
+        assert_eq!(collected, vec![0, 7, 15, 20]);
+    }
+
+    /// Unit-test the `BitSet` impl of `TermOrdAccumulator`: insert,
+    /// dedup, and iter_ords order.
+    #[test]
+    fn bitset_accumulator_basic() {
+        use common::BitSet;
+
+        use super::TermOrdAccumulator;
+        let mut set = <BitSet as TermOrdAccumulator>::new(255);
+        for ord in [0u64, 1, 63, 64, 65, 128, 200, 200, 0] {
+            <BitSet as TermOrdAccumulator>::insert(&mut set, ord);
+        }
+        assert_eq!(<BitSet as TermOrdAccumulator>::len(&set), 7);
+        let collected: Vec<u64> = set.iter_ords().collect();
+        assert_eq!(collected, vec![0, 1, 63, 64, 65, 128, 200]);
+    }
+
     #[test]
     fn cardinality_aggregation_u64() -> crate::Result<()> {
         let mut schema_builder = Schema::builder();
@@ -683,6 +1143,42 @@ mod tests {
         Ok(())
     }
 
+    /// A JSON path that resolves to both a Str column and a numeric column
+    /// produces two collector instances per segment — one with `Str` buckets
+    /// and one with `Numeric` buckets. Their `IntermediateMetricResult`s must
+    /// merge into the union cardinality.
+    #[test]
+    fn cardinality_aggregation_json_str_and_numeric() -> crate::Result<()> {
+        let mut schema_builder = Schema::builder();
+        let field = schema_builder.add_json_field("json", FAST);
+        let index = Index::create_in_ram(schema_builder.build());
+        {
+            let mut writer = index.writer_for_tests()?;
+            writer.add_document(doc!(field => json!({"value": "hello"})))?;
+            writer.add_document(doc!(field => json!({"value": "world"})))?;
+            writer.add_document(doc!(field => json!({"value": "hello"})))?; // dup str
+            writer.add_document(doc!(field => json!({"value": i64::from_u64(7u64)})))?;
+            writer.add_document(doc!(field => json!({"value": i64::from_u64(42u64)})))?;
+            writer.add_document(doc!(field => json!({"value": i64::from_u64(7u64)})))?; // dup num
+            writer.commit()?;
+        }
+
+        let agg_req: Aggregations = serde_json::from_value(json!({
+            "cardinality": {
+                "cardinality": {
+                    "field": "json.value"
+                },
+            }
+        }))
+        .unwrap();
+
+        let res = exec_request(agg_req, &index)?;
+        // 4 distinct values: "hello", "world", 7, 42.
+        assert_eq!(res["cardinality"]["value"], 4.0);
+
+        Ok(())
+    }
+
     #[test]
     fn cardinality_collector_serde_roundtrip() {
         use super::CardinalityCollector;

src/aggregation/metric/extended_stats.rs

@@ -399,6 +399,26 @@ impl SegmentAggregationCollector for SegmentExtendedStatsCollector {
         }
         Ok(())
     }
+
+    fn compute_metric_value(
+        &self,
+        bucket_id: BucketId,
+        sub_agg_name: &str,
+        sub_agg_property: &str,
+        _agg_data: &AggregationsSegmentCtx,
+    ) -> Option<f64> {
+        if self.name != sub_agg_name {
+            return None;
+        }
+        let extended = self.buckets.get(bucket_id as usize)?;
+        // Finalize is a pure read of accumulators — calling it here for the cutoff sort
+        // doesn't disturb the eventual intermediate result.
+        extended
+            .finalize()
+            .get_value(sub_agg_property)
+            .ok()
+            .flatten()
+    }
 }
 
 #[cfg(test)]

src/aggregation/metric/percentiles.rs

@@ -312,6 +312,26 @@ impl SegmentAggregationCollector for SegmentPercentilesCollector {
         }
         Ok(())
     }
+
+    fn compute_metric_value(
+        &self,
+        bucket_id: BucketId,
+        sub_agg_name: &str,
+        sub_agg_property: &str,
+        agg_data: &AggregationsSegmentCtx,
+    ) -> Option<f64> {
+        if agg_data.get_metric_req_data(self.accessor_idx).name != sub_agg_name {
+            return None;
+        }
+        let percentile: f64 = sub_agg_property.parse().ok()?;
+        if !(0.0..=100.0).contains(&percentile) {
+            return None;
+        }
+        let bucket = self.buckets.get(bucket_id as usize)?;
+        // DDSketch.quantile is a pure read; calling it here for the cutoff sort does
+        // not affect the intermediate state used for the final result.
+        bucket.sketch.quantile(percentile / 100.0).ok().flatten()
+    }
 }
 
 #[cfg(test)]

src/aggregation/metric/stats.rs

@@ -321,6 +321,40 @@ impl<const COLUMN_TYPE_ID: u8> SegmentAggregationCollector
         }
         Ok(())
     }
+
+    fn compute_metric_value(
+        &self,
+        bucket_id: BucketId,
+        sub_agg_name: &str,
+        sub_agg_property: &str,
+        _agg_data: &AggregationsSegmentCtx,
+    ) -> Option<f64> {
+        if self.name != sub_agg_name {
+            return None;
+        }
+        let stats = self.buckets.get(bucket_id as usize)?;
+        // The property depends on what we're collecting:
+        //   - StatsType::Stats exposes count/sum/min/max/avg via dotted property.
+        //   - Single-value kinds (Sum/Count/Min/Max/Average) expect an empty property and return
+        //     the value they were configured to collect.
+        let prop = match self.collecting_for {
+            StatsType::Stats if !sub_agg_property.is_empty() => sub_agg_property,
+            StatsType::Sum if sub_agg_property.is_empty() => "sum",
+            StatsType::Count if sub_agg_property.is_empty() => "count",
+            StatsType::Max if sub_agg_property.is_empty() => "max",
+            StatsType::Min if sub_agg_property.is_empty() => "min",
+            StatsType::Average if sub_agg_property.is_empty() => "avg",
+            _ => return None,
+        };
+        match prop {
+            "count" => Some(stats.count as f64),
+            "sum" => Some(stats.sum),
+            "min" if stats.count > 0 => Some(stats.min),
+            "max" if stats.count > 0 => Some(stats.max),
+            "avg" if stats.count > 0 => Some(stats.sum / stats.count as f64),
+            _ => None,
+        }
+    }
 }
 
 #[inline]

src/aggregation/metric/top_hits.rs

@@ -644,6 +644,17 @@ impl SegmentAggregationCollector for TopHitsSegmentCollector {
         );
         Ok(())
     }
+
+    fn compute_metric_value(
+        &self,
+        _bucket_id: BucketId,
+        _sub_agg_name: &str,
+        _sub_agg_property: &str,
+        _agg_data: &AggregationsSegmentCtx,
+    ) -> Option<f64> {
+        // top_hits is not a numeric metric and cannot be used as an order target.
+        None
+    }
 }
 
 #[cfg(test)]

src/aggregation/segment_agg_result.rs

@@ -76,6 +76,31 @@ pub trait SegmentAggregationCollector: Debug {
     fn flush(&mut self, _agg_data: &mut AggregationsSegmentCtx) -> crate::Result<()> {
         Ok(())
     }
+
+    /// Compute the segment-level metric value of the named direct-child metric for `bucket_id`.
+    ///
+    /// Used by parent term aggs that order by a sub-aggregation: the parent sorts on
+    /// this value and cuts off at segment time, matching the approximation tradeoff
+    /// Elasticsearch makes for any sub-agg ordering.
+    ///
+    /// `sub_agg_property` is the dotted suffix (e.g. `"sum"` in `mystats.sum`); empty when
+    /// the metric is a single-value kind such as cardinality.
+    ///
+    /// Returns `None` only on name mismatch, unknown property, or empty bucket. Implementations
+    /// may finalize their per-bucket state (e.g. compute a percentile from a sketch); calls
+    /// must be idempotent so the final intermediate result is unaffected.
+    ///
+    /// No default impl on purpose: every collector must decide explicitly whether it
+    /// produces a metric value, forwards into children (single-bucket aggs), or rejects
+    /// the lookup. A silent `None` default would let a parent term agg's cutoff sort all
+    /// buckets to the same key and drop arbitrary winners.
+    fn compute_metric_value(
+        &self,
+        bucket_id: BucketId,
+        sub_agg_name: &str,
+        sub_agg_property: &str,
+        agg_data: &AggregationsSegmentCtx,
+    ) -> Option<f64>;
 }
 
 #[derive(Default)]
@@ -137,4 +162,21 @@ impl SegmentAggregationCollector for GenericSegmentAggregationResultsCollector {
         }
         Ok(())
     }
+
+    fn compute_metric_value(
+        &self,
+        bucket_id: BucketId,
+        sub_agg_name: &str,
+        sub_agg_property: &str,
+        agg_data: &AggregationsSegmentCtx,
+    ) -> Option<f64> {
+        for agg in &self.aggs {
+            if let Some(value) =
+                agg.compute_metric_value(bucket_id, sub_agg_name, sub_agg_property, agg_data)
+            {
+                return Some(value);
+            }
+        }
+        None
+    }
 }

src/query/boolean_query/boolean_weight.rs

@@ -9,8 +9,8 @@ use crate::query::score_combiner::{DoNothingCombiner, ScoreCombiner};
 use crate::query::term_query::TermScorer;
 use crate::query::weight::{for_each_docset_buffered, for_each_pruning_scorer, for_each_scorer};
 use crate::query::{
-    intersect_scorers, AllScorer, BufferedUnionScorer, EmptyScorer, Exclude, Explanation,
-    Intersection, Occur, RequiredOptionalScorer, Scorer, Weight,
+    intersect_scorers, AllScorer, BufferedUnionScorer, EmptyScorer, Exclude, Explanation, Occur,
+    RequiredOptionalScorer, Scorer, Weight,
 };
 use crate::{DocId, Score};
 
@@ -50,10 +50,9 @@ where
     TScoreCombiner: ScoreCombiner,
 {
     assert!(!scorers.is_empty());
-    if scorers.len() == 1 {
+    if scorers.len() == 1 && !scorers[0].is::<TermScorer>() {
         return SpecializedScorer::Other(scorers.into_iter().next().unwrap()); //< we checked the size beforehand
     }
-
     {
         let is_all_term_queries = scorers.iter().all(|scorer| scorer.is::<TermScorer>());
         if is_all_term_queries {
@@ -67,6 +66,9 @@ where
             {
                 // Block wand is only available if we read frequencies.
                 return SpecializedScorer::TermUnion(scorers);
+            } else if scorers.len() == 1 {
+                // Single TermScorer without freq reading — unwrap directly.
+                return SpecializedScorer::Other(Box::new(scorers.into_iter().next().unwrap()));
             } else {
                 return SpecializedScorer::Other(Box::new(BufferedUnionScorer::build(
                     scorers,