Opportunistically seed forked block caches from current one.

Cargo.toml

@@ -38,7 +38,7 @@ crossbeam-channel = "0.5.4"
 rust-stemmers = "1.2.0"
 downcast-rs = "1.2.0"
 bitpacking = { version = "0.9.2", default-features = false, features = ["bitpacker4x"] }
-census = "0.4.2"
+census = "0.4.0"
 rustc-hash = "1.1.0"
 thiserror = "1.0.30"
 htmlescape = "0.3.1"

README.md

@@ -5,18 +5,19 @@
 [![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)
 
-<img src="https://tantivy-search.github.io/logo/tantivy-logo.png" alt="Tantivy, the fastest full-text search engine library written in Rust" height="250">
+![Tantivy](https://tantivy-search.github.io/logo/tantivy-logo.png)
 
-## Fast full-text search engine library written in Rust
+**Tantivy** is a **full-text search engine library** written in Rust.
 
-**If you are looking for an alternative to Elasticsearch or Apache Solr, check out [Quickwit](https://github.com/quickwit-oss/quickwit), our distributed search engine built on top of Tantivy.**
+It is closer to [Apache Lucene](https://lucene.apache.org/) than to [Elasticsearch](https://www.elastic.co/products/elasticsearch) or [Apache Solr](https://lucene.apache.org/solr/) in the sense it is not
-
+an off-the-shelf search engine server, but rather a crate that can be used
-Tantivy is closer to [Apache Lucene](https://lucene.apache.org/) than to [Elasticsearch](https://www.elastic.co/products/elasticsearch) or [Apache Solr](https://lucene.apache.org/solr/) in the sense it is not
+to build such a search engine.
-an off-the-shelf search engine server, but rather a crate that can be used to build such a search engine.
 
 Tantivy is, in fact, strongly inspired by Lucene's design.
 
-## Benchmark
+If you are looking for an alternative to Elasticsearch or Apache Solr, check out [Quickwit](https://github.com/quickwit-oss/quickwit), our search engine built on top of Tantivy.
+
+# Benchmark
 
 The following [benchmark](https://tantivy-search.github.io/bench/) breakdowns
 performance for different types of queries/collections.
@@ -27,7 +28,7 @@ Your mileage WILL vary depending on the nature of queries and their load.
 
 Details about the benchmark can be found at this [repository](https://github.com/quickwit-oss/search-benchmark-game).
 
-## Features
+# Features
 
 - Full-text search
 - Configurable tokenizer (stemming available for 17 Latin languages) with third party support for Chinese ([tantivy-jieba](https://crates.io/crates/tantivy-jieba) and [cang-jie](https://crates.io/crates/cang-jie)), Japanese ([lindera](https://github.com/lindera-morphology/lindera-tantivy), [Vaporetto](https://crates.io/crates/vaporetto_tantivy), and [tantivy-tokenizer-tiny-segmenter](https://crates.io/crates/tantivy-tokenizer-tiny-segmenter)) and Korean ([lindera](https://github.com/lindera-morphology/lindera-tantivy) + [lindera-ko-dic-builder](https://github.com/lindera-morphology/lindera-ko-dic-builder))
@@ -53,11 +54,11 @@ Details about the benchmark can be found at this [repository](https://github.com
 - Searcher Warmer API
 - Cheesy logo with a horse
 
-### Non-features
+## Non-features
 
 Distributed search is out of the scope of Tantivy, but if you are looking for this feature, check out [Quickwit](https://github.com/quickwit-oss/quickwit/).
 
-## Getting started
+# Getting started
 
 Tantivy works on stable Rust and supports Linux, macOS, and Windows.
 
@@ -67,7 +68,7 @@ index documents, and search via the CLI or a small server with a REST API.
 It walks you through getting a Wikipedia search engine up and running in a few minutes.
 - [Reference doc for the last released version](https://docs.rs/tantivy/)
 
-## How can I support this project?
+# How can I support this project?
 
 There are many ways to support this project.
 
@@ -78,16 +79,16 @@ There are many ways to support this project.
 - Contribute code (you can join [our Discord server](https://discord.gg/MT27AG5EVE))
 - Talk about Tantivy around you
 
-## Contributing code
+# Contributing code
 
 We use the GitHub Pull Request workflow: reference a GitHub ticket and/or include a comprehensive commit message when opening a PR.
 Feel free to update CHANGELOG.md with your contribution.
 
-### Tokenizer
+## Tokenizer
 
 When implementing a tokenizer for tantivy depend on the `tantivy-tokenizer-api` crate.
 
-### Clone and build locally
+## Clone and build locally
 
 Tantivy compiles on stable Rust.
 To check out and run tests, you can simply run:
@@ -98,7 +99,7 @@ cd tantivy
 cargo test
 ```
 
-## Companies Using Tantivy
+# Companies Using Tantivy
 
 <p align="left">
 <img align="center" src="doc/assets/images/etsy.png" alt="Etsy" height="25" width="auto" />&nbsp;
@@ -110,7 +111,7 @@ cargo test
 <img align="center" src="doc/assets/images/element-dark-theme.png#gh-dark-mode-only" alt="Element.io" height="25" width="auto" />
 </p>
 
-## FAQ
+# FAQ
 
 ### Can I use Tantivy in other languages?
 

columnar/src/column_index/mod.rs

@@ -126,18 +126,18 @@ impl ColumnIndex {
         }
     }
 
-    pub fn docid_range_to_rowids(&self, doc_id_range: Range<DocId>) -> Range<RowId> {
+    pub fn docid_range_to_rowids(&self, doc_id: Range<DocId>) -> Range<RowId> {
         match self {
             ColumnIndex::Empty { .. } => 0..0,
-            ColumnIndex::Full => doc_id_range,
+            ColumnIndex::Full => doc_id,
             ColumnIndex::Optional(optional_index) => {
-                let row_start = optional_index.rank(doc_id_range.start);
+                let row_start = optional_index.rank(doc_id.start);
-                let row_end = optional_index.rank(doc_id_range.end);
+                let row_end = optional_index.rank(doc_id.end);
                 row_start..row_end
             }
             ColumnIndex::Multivalued(multivalued_index) => {
-                let end_docid = doc_id_range.end.min(multivalued_index.num_docs() - 1) + 1;
+                let end_docid = doc_id.end.min(multivalued_index.num_docs() - 1) + 1;
-                let start_docid = doc_id_range.start.min(end_docid);
+                let start_docid = doc_id.start.min(end_docid);
 
                 let row_start = multivalued_index.start_index_column.get_val(start_docid);
                 let row_end = multivalued_index.start_index_column.get_val(end_docid);

columnar/src/column_index/optional_index/mod.rs

@@ -21,6 +21,8 @@ const DENSE_BLOCK_THRESHOLD: u32 =
 
 const ELEMENTS_PER_BLOCK: u32 = u16::MAX as u32 + 1;
 
+const BLOCK_SIZE: RowId = 1 << 16;
+
 #[derive(Copy, Clone, Debug)]
 struct BlockMeta {
     non_null_rows_before_block: u32,
@@ -107,8 +109,8 @@ struct RowAddr {
 #[inline(always)]
 fn row_addr_from_row_id(row_id: RowId) -> RowAddr {
     RowAddr {
-        block_id: (row_id / ELEMENTS_PER_BLOCK) as u16,
+        block_id: (row_id / BLOCK_SIZE) as u16,
-        in_block_row_id: (row_id % ELEMENTS_PER_BLOCK) as u16,
+        in_block_row_id: (row_id % BLOCK_SIZE) as u16,
     }
 }
 
@@ -183,13 +185,8 @@ impl Set<RowId> for OptionalIndex {
         }
     }
 
-    /// Any value doc_id is allowed.
-    /// In particular, doc_id = num_rows.
     #[inline]
     fn rank(&self, doc_id: DocId) -> RowId {
-        if doc_id >= self.num_docs() {
-            return self.num_non_nulls();
-        }
         let RowAddr {
             block_id,
             in_block_row_id,
@@ -203,15 +200,13 @@ impl Set<RowId> for OptionalIndex {
         block_meta.non_null_rows_before_block + block_offset_row_id
     }
 
-    /// Any value doc_id is allowed.
-    /// In particular, doc_id = num_rows.
     #[inline]
     fn rank_if_exists(&self, doc_id: DocId) -> Option<RowId> {
         let RowAddr {
             block_id,
             in_block_row_id,
         } = row_addr_from_row_id(doc_id);
-        let block_meta = *self.block_metas.get(block_id as usize)?;
+        let block_meta = self.block_metas[block_id as usize];
         let block = self.block(block_meta);
         let block_offset_row_id = match block {
             Block::Dense(dense_block) => dense_block.rank_if_exists(in_block_row_id),
@@ -496,7 +491,7 @@ fn deserialize_optional_index_block_metadatas(
         non_null_rows_before_block += num_non_null_rows;
     }
     block_metas.resize(
-        ((num_rows + ELEMENTS_PER_BLOCK - 1) / ELEMENTS_PER_BLOCK) as usize,
+        ((num_rows + BLOCK_SIZE - 1) / BLOCK_SIZE) as usize,
         BlockMeta {
             non_null_rows_before_block,
             start_byte_offset,

columnar/src/column_index/optional_index/set.rs

@@ -39,8 +39,7 @@ pub trait Set<T> {
     ///
     /// # Panics
     ///
-    /// May panic if rank is greater or equal to the number of
+    /// May panic if rank is greater than the number of elements in the Set.
-    /// elements in the Set.
     fn select(&self, rank: T) -> T;
 
     /// Creates a brand new select cursor.

columnar/src/column_index/optional_index/tests.rs

@@ -3,30 +3,6 @@ use proptest::strategy::Strategy;
 use proptest::{prop_oneof, proptest};
 
 use super::*;
-use crate::{ColumnarReader, ColumnarWriter, DynamicColumnHandle};
-
-#[test]
-fn test_optional_index_bug_2293() {
-    // tests for panic in docid_range_to_rowids for docid == num_docs
-    test_optional_index_with_num_docs(ELEMENTS_PER_BLOCK - 1);
-    test_optional_index_with_num_docs(ELEMENTS_PER_BLOCK);
-    test_optional_index_with_num_docs(ELEMENTS_PER_BLOCK + 1);
-}
-fn test_optional_index_with_num_docs(num_docs: u32) {
-    let mut dataframe_writer = ColumnarWriter::default();
-    dataframe_writer.record_numerical(100, "score", 80i64);
-    let mut buffer: Vec<u8> = Vec::new();
-    dataframe_writer
-        .serialize(num_docs, None, &mut buffer)
-        .unwrap();
-    let columnar = ColumnarReader::open(buffer).unwrap();
-    assert_eq!(columnar.num_columns(), 1);
-    let cols: Vec<DynamicColumnHandle> = columnar.read_columns("score").unwrap();
-    assert_eq!(cols.len(), 1);
-
-    let col = cols[0].open().unwrap();
-    col.column_index().docid_range_to_rowids(0..num_docs);
-}
 
 #[test]
 fn test_dense_block_threshold() {
@@ -59,7 +35,7 @@ proptest! {
 
 #[test]
 fn test_with_random_sets_simple() {
-    let vals = 10..ELEMENTS_PER_BLOCK * 2;
+    let vals = 10..BLOCK_SIZE * 2;
     let mut out: Vec<u8> = Vec::new();
     serialize_optional_index(&vals, 100, &mut out).unwrap();
     let null_index = open_optional_index(OwnedBytes::new(out)).unwrap();
@@ -195,7 +171,7 @@ fn test_optional_index_rank() {
     test_optional_index_rank_aux(&[0u32, 1u32]);
     let mut block = Vec::new();
     block.push(3u32);
-    block.extend((0..ELEMENTS_PER_BLOCK).map(|i| i + ELEMENTS_PER_BLOCK + 1));
+    block.extend((0..BLOCK_SIZE).map(|i| i + BLOCK_SIZE + 1));
     test_optional_index_rank_aux(&block);
 }
 
@@ -209,8 +185,8 @@ fn test_optional_index_iter_empty_one() {
 fn test_optional_index_iter_dense_block() {
     let mut block = Vec::new();
     block.push(3u32);
-    block.extend((0..ELEMENTS_PER_BLOCK).map(|i| i + ELEMENTS_PER_BLOCK + 1));
+    block.extend((0..BLOCK_SIZE).map(|i| i + BLOCK_SIZE + 1));
-    test_optional_index_iter_aux(&block, 3 * ELEMENTS_PER_BLOCK);
+    test_optional_index_iter_aux(&block, 3 * BLOCK_SIZE);
 }
 
 #[test]

columnar/src/column_values/mod.rs

@@ -101,7 +101,7 @@ pub trait ColumnValues<T: PartialOrd = u64>: Send + Sync {
         row_id_hits: &mut Vec<RowId>,
     ) {
         let row_id_range = row_id_range.start..row_id_range.end.min(self.num_vals());
-        for idx in row_id_range {
+        for idx in row_id_range.start..row_id_range.end {
             let val = self.get_val(idx);
             if value_range.contains(&val) {
                 row_id_hits.push(idx);

query-grammar/src/infallible.rs

@@ -81,8 +81,8 @@ where
     T: InputTakeAtPosition + Clone,
     <T as InputTakeAtPosition>::Item: AsChar + Clone,
 {
-    opt_i(nom::character::complete::multispace0)(input)
+    opt_i(nom::character::complete::space0)(input)
-        .map(|(left, (spaces, errors))| (left, (spaces.expect("multispace0 can't fail"), errors)))
+        .map(|(left, (spaces, errors))| (left, (spaces.expect("space0 can't fail"), errors)))
 }
 
 pub(crate) fn space1_infallible<T>(input: T) -> JResult<T, Option<T>>
@@ -90,7 +90,7 @@ where
     T: InputTakeAtPosition + Clone + InputLength,
     <T as InputTakeAtPosition>::Item: AsChar + Clone,
 {
-    opt_i(nom::character::complete::multispace1)(input).map(|(left, (spaces, mut errors))| {
+    opt_i(nom::character::complete::space1)(input).map(|(left, (spaces, mut errors))| {
         if spaces.is_none() {
             errors.push(LenientErrorInternal {
                 pos: left.input_len(),

query-grammar/src/query_grammar.rs

@@ -3,7 +3,7 @@ use std::iter::once;
 use nom::branch::alt;
 use nom::bytes::complete::tag;
 use nom::character::complete::{
-    anychar, char, digit1, multispace0, multispace1, none_of, one_of, satisfy, u32,
+    anychar, char, digit1, none_of, one_of, satisfy, space0, space1, u32,
 };
 use nom::combinator::{eof, map, map_res, opt, peek, recognize, value, verify};
 use nom::error::{Error, ErrorKind};
@@ -65,7 +65,7 @@ fn word_infallible(delimiter: &str) -> impl Fn(&str) -> JResult<&str, Option<&st
     |inp| {
         opt_i_err(
             preceded(
-                multispace0,
+                space0,
                 recognize(many1(satisfy(|c| {
                     !c.is_whitespace() && !delimiter.contains(c)
                 }))),
@@ -225,10 +225,10 @@ fn term_group(inp: &str) -> IResult<&str, UserInputAst> {
 
     map(
         tuple((
-            terminated(field_name, multispace0),
+            terminated(field_name, space0),
             delimited(
-                tuple((char('('), multispace0)),
+                tuple((char('('), space0)),
-                separated_list0(multispace1, tuple((opt(occur_symbol), term_or_phrase))),
+                separated_list0(space1, tuple((opt(occur_symbol), term_or_phrase))),
                 char(')'),
             ),
         )),
@@ -250,7 +250,7 @@ fn term_group_precond(inp: &str) -> IResult<&str, (), ()> {
         (),
         peek(tuple((
             field_name,
-            multispace0,
+            space0,
             char('('), // when we are here, we know it can't be anything but a term group
         ))),
     )(inp)
@@ -259,7 +259,7 @@ fn term_group_precond(inp: &str) -> IResult<&str, (), ()> {
 
 fn term_group_infallible(inp: &str) -> JResult<&str, UserInputAst> {
     let (mut inp, (field_name, _, _, _)) =
-        tuple((field_name, multispace0, char('('), multispace0))(inp).expect("precondition failed");
+        tuple((field_name, space0, char('('), space0))(inp).expect("precondition failed");
 
     let mut terms = Vec::new();
     let mut errs = Vec::new();
@@ -305,7 +305,7 @@ fn exists(inp: &str) -> IResult<&str, UserInputLeaf> {
         UserInputLeaf::Exists {
             field: String::new(),
         },
-        tuple((multispace0, char('*'))),
+        tuple((space0, char('*'))),
     )(inp)
 }
 
@@ -314,7 +314,7 @@ fn exists_precond(inp: &str) -> IResult<&str, (), ()> {
         (),
         peek(tuple((
             field_name,
-            multispace0,
+            space0,
             char('*'), // when we are here, we know it can't be anything but a exists
         ))),
     )(inp)
@@ -323,7 +323,7 @@ fn exists_precond(inp: &str) -> IResult<&str, (), ()> {
 
 fn exists_infallible(inp: &str) -> JResult<&str, UserInputAst> {
     let (inp, (field_name, _, _)) =
-        tuple((field_name, multispace0, char('*')))(inp).expect("precondition failed");
+        tuple((field_name, space0, char('*')))(inp).expect("precondition failed");
 
     let exists = UserInputLeaf::Exists { field: field_name }.into();
     Ok((inp, (exists, Vec::new())))
@@ -349,7 +349,7 @@ fn literal_no_group_infallible(inp: &str) -> JResult<&str, Option<UserInputAst>>
             alt_infallible(
                 (
                     (
-                        value((), tuple((tag("IN"), multispace0, char('[')))),
+                        value((), tuple((tag("IN"), space0, char('[')))),
                         map(set_infallible, |(set, errs)| (Some(set), errs)),
                     ),
                     (
@@ -430,8 +430,8 @@ fn range(inp: &str) -> IResult<&str, UserInputLeaf> {
     // check for unbounded range in the form of <5, <=10, >5, >=5
     let elastic_unbounded_range = map(
         tuple((
-            preceded(multispace0, alt((tag(">="), tag("<="), tag("<"), tag(">")))),
+            preceded(space0, alt((tag(">="), tag("<="), tag("<"), tag(">")))),
-            preceded(multispace0, range_term_val()),
+            preceded(space0, range_term_val()),
         )),
         |(comparison_sign, bound)| match comparison_sign {
             ">=" => (UserInputBound::Inclusive(bound), UserInputBound::Unbounded),
@@ -444,7 +444,7 @@ fn range(inp: &str) -> IResult<&str, UserInputLeaf> {
     );
 
     let lower_bound = map(
-        separated_pair(one_of("{["), multispace0, range_term_val()),
+        separated_pair(one_of("{["), space0, range_term_val()),
         |(boundary_char, lower_bound)| {
             if lower_bound == "*" {
                 UserInputBound::Unbounded
@@ -457,7 +457,7 @@ fn range(inp: &str) -> IResult<&str, UserInputLeaf> {
     );
 
     let upper_bound = map(
-        separated_pair(range_term_val(), multispace0, one_of("}]")),
+        separated_pair(range_term_val(), space0, one_of("}]")),
         |(upper_bound, boundary_char)| {
             if upper_bound == "*" {
                 UserInputBound::Unbounded
@@ -469,11 +469,8 @@ fn range(inp: &str) -> IResult<&str, UserInputLeaf> {
         },
     );
 
-    let lower_to_upper = separated_pair(
+    let lower_to_upper =
-        lower_bound,
+        separated_pair(lower_bound, tuple((space1, tag("TO"), space1)), upper_bound);
-        tuple((multispace1, tag("TO"), multispace1)),
-        upper_bound,
-    );
 
     map(
         alt((elastic_unbounded_range, lower_to_upper)),
@@ -493,16 +490,13 @@ fn range_infallible(inp: &str) -> JResult<&str, UserInputLeaf> {
             word_infallible("]}"),
             space1_infallible,
             opt_i_err(
-                terminated(tag("TO"), alt((value((), multispace1), value((), eof)))),
+                terminated(tag("TO"), alt((value((), space1), value((), eof)))),
                 "missing keyword TO",
             ),
             word_infallible("]}"),
             opt_i_err(one_of("]}"), "missing range delimiter"),
         )),
-        |(
+        |((lower_bound_kind, _space0, lower, _space1, to, upper, upper_bound_kind), errs)| {
-            (lower_bound_kind, _multispace0, lower, _multispace1, to, upper, upper_bound_kind),
-            errs,
-        )| {
             let lower_bound = match (lower_bound_kind, lower) {
                 (_, Some("*")) => UserInputBound::Unbounded,
                 (_, None) => UserInputBound::Unbounded,
@@ -602,10 +596,10 @@ fn range_infallible(inp: &str) -> JResult<&str, UserInputLeaf> {
 fn set(inp: &str) -> IResult<&str, UserInputLeaf> {
     map(
         preceded(
-            tuple((multispace0, tag("IN"), multispace1)),
+            tuple((space0, tag("IN"), space1)),
             delimited(
-                tuple((char('['), multispace0)),
+                tuple((char('['), space0)),
-                separated_list0(multispace1, map(simple_term, |(_, term)| term)),
+                separated_list0(space1, map(simple_term, |(_, term)| term)),
                 char(']'),
             ),
         ),
@@ -673,7 +667,7 @@ fn leaf(inp: &str) -> IResult<&str, UserInputAst> {
     alt((
         delimited(char('('), ast, char(')')),
         map(char('*'), |_| UserInputAst::from(UserInputLeaf::All)),
-        map(preceded(tuple((tag("NOT"), multispace1)), leaf), negate),
+        map(preceded(tuple((tag("NOT"), space1)), leaf), negate),
         literal,
     ))(inp)
 }
@@ -925,17 +919,17 @@ fn aggregate_infallible_expressions(
 
 fn operand_leaf(inp: &str) -> IResult<&str, (BinaryOperand, UserInputAst)> {
     tuple((
-        terminated(binary_operand, multispace0),
+        terminated(binary_operand, space0),
-        terminated(boosted_leaf, multispace0),
+        terminated(boosted_leaf, space0),
     ))(inp)
 }
 
 fn ast(inp: &str) -> IResult<&str, UserInputAst> {
     let boolean_expr = map(
-        separated_pair(boosted_leaf, multispace1, many1(operand_leaf)),
+        separated_pair(boosted_leaf, space1, many1(operand_leaf)),
         |(left, right)| aggregate_binary_expressions(left, right),
     );
-    let whitespace_separated_leaves = map(separated_list1(multispace1, occur_leaf), |subqueries| {
+    let whitespace_separated_leaves = map(separated_list1(space1, occur_leaf), |subqueries| {
         if subqueries.len() == 1 {
             let (occur_opt, ast) = subqueries.into_iter().next().unwrap();
             match occur_opt.unwrap_or(Occur::Should) {
@@ -948,9 +942,9 @@ fn ast(inp: &str) -> IResult<&str, UserInputAst> {
     });
 
     delimited(
-        multispace0,
+        space0,
         alt((boolean_expr, whitespace_separated_leaves)),
-        multispace0,
+        space0,
     )(inp)
 }
 
@@ -975,7 +969,7 @@ fn ast_infallible(inp: &str) -> JResult<&str, UserInputAst> {
 }
 
 pub fn parse_to_ast(inp: &str) -> IResult<&str, UserInputAst> {
-    map(delimited(multispace0, opt(ast), eof), |opt_ast| {
+    map(delimited(space0, opt(ast), eof), |opt_ast| {
         rewrite_ast(opt_ast.unwrap_or_else(UserInputAst::empty_query))
     })(inp)
 }
@@ -1151,7 +1145,6 @@ mod test {
     #[test]
     fn test_parse_query_to_ast_binary_op() {
         test_parse_query_to_ast_helper("a AND b", "(+a +b)");
-        test_parse_query_to_ast_helper("a\nAND b", "(+a +b)");
         test_parse_query_to_ast_helper("a OR b", "(?a ?b)");
         test_parse_query_to_ast_helper("a OR b AND c", "(?a ?(+b +c))");
         test_parse_query_to_ast_helper("a AND b         AND c", "(+a +b +c)");

src/aggregation/bucket/histogram/histogram.rs

@@ -596,13 +596,10 @@ mod tests {
 
     use super::*;
     use crate::aggregation::agg_req::Aggregations;
-    use crate::aggregation::agg_result::AggregationResults;
     use crate::aggregation::tests::{
         exec_request, exec_request_with_query, exec_request_with_query_and_memory_limit,
         get_test_index_2_segments, get_test_index_from_values, get_test_index_with_num_docs,
     };
-    use crate::aggregation::AggregationCollector;
-    use crate::query::AllQuery;
 
     #[test]
     fn histogram_test_crooked_values() -> crate::Result<()> {
@@ -1354,35 +1351,6 @@ mod tests {
             })
         );
 
-        Ok(())
-    }
-    #[test]
-    fn test_aggregation_histogram_empty_index() -> crate::Result<()> {
-        // test index without segments
-        let values = vec![];
-
-        let index = get_test_index_from_values(false, &values)?;
-
-        let agg_req_1: Aggregations = serde_json::from_value(json!({
-            "myhisto": {
-                "histogram": {
-                    "field": "score",
-                    "interval": 10.0
-                },
-            }
-        }))
-        .unwrap();
-
-        let collector = AggregationCollector::from_aggs(agg_req_1, Default::default());
-
-        let reader = index.reader()?;
-        let searcher = reader.searcher();
-        let agg_res: AggregationResults = searcher.search(&AllQuery, &collector).unwrap();
-
-        let res: Value = serde_json::from_str(&serde_json::to_string(&agg_res)?)?;
-        // Make sure the result structure is correct
-        assert_eq!(res["myhisto"]["buckets"].as_array().unwrap().len(), 0);
-
         Ok(())
     }
 }

src/collector/top_score_collector.rs

@@ -309,7 +309,7 @@ impl TopDocs {
     ///
     /// To comfortably work with `u64`s, `i64`s, `f64`s, or `date`s, please refer to
     /// the [.order_by_fast_field(...)](TopDocs::order_by_fast_field) method.
-    pub fn order_by_u64_field(
+    fn order_by_u64_field(
         self,
         field: impl ToString,
         order: Order,

src/core/searcher.rs

@@ -1,4 +1,6 @@
 use std::collections::BTreeMap;
+#[cfg(feature = "quickwit")]
+use std::future::Future;
 use std::sync::Arc;
 use std::{fmt, io};
 
@@ -112,6 +114,108 @@ impl Searcher {
         store_reader.get_async(doc_address.doc_id).await
     }
 
+    /// Fetches multiple documents in an asynchronous manner.
+    ///
+    /// This method is more efficient than calling [`doc_async`](Self::doc_async) multiple times,
+    /// as it groups overlapping requests to segments and blocks and avoids concurrent requests
+    /// trashing the caches of each other. However, it does so using intermediate data structures
+    /// and independent block caches so it will be slower if documents from very few blocks are
+    /// fetched which would have fit into the global block cache.
+    ///
+    /// The caller is expected to poll these futures concurrently (e.g. using `FuturesUnordered`)
+    /// or in parallel (e.g. using `JoinSet`) as fits best with the given use case, i.e. whether
+    /// it is predominately I/O-bound or rather CPU-bound.
+    ///
+    /// Note that any blocks brought into any of the per-segment-and-block groups will not be pulled
+    /// into the global block cache and hence not be available for subsequent calls.
+    ///
+    /// Note that there is no synchronous variant of this method as the same degree of efficiency
+    /// can be had by accessing documents in address order.
+    ///
+    /// # Example
+    ///
+    /// ```rust,no_run
+    /// # use futures::executor::block_on;
+    /// # use futures::stream::{FuturesUnordered, StreamExt};
+    /// #
+    /// # use tantivy::schema::Schema;
+    /// # use tantivy::{DocAddress, Index, TantivyDocument, TantivyError};
+    /// #
+    /// # let index = Index::create_in_ram(Schema::builder().build());
+    /// # let searcher = index.reader()?.searcher();
+    /// #
+    /// # let doc_addresses = (0..10).map(|_| DocAddress::new(0, 0));
+    /// #
+    /// let mut groups: FuturesUnordered<_> = searcher
+    ///     .docs_async::<TantivyDocument>(doc_addresses)?
+    ///     .collect();
+    ///
+    /// let mut docs = Vec::new();
+    ///
+    /// block_on(async {
+    ///     while let Some(group) = groups.next().await {
+    ///         docs.extend(group?);
+    ///     }
+    ///
+    ///     Ok::<_, TantivyError>(())
+    /// })?;
+    /// #
+    /// # Ok::<_, TantivyError>(())
+    /// ```
+    #[cfg(feature = "quickwit")]
+    pub fn docs_async<D: DocumentDeserialize>(
+        &self,
+        doc_addresses: impl IntoIterator<Item = DocAddress>,
+    ) -> crate::Result<
+        impl Iterator<Item = impl Future<Output = crate::Result<Vec<(DocAddress, D)>>>> + '_,
+    > {
+        use rustc_hash::FxHashMap;
+
+        use crate::store::CacheKey;
+        use crate::{DocId, SegmentOrdinal};
+
+        let mut groups: FxHashMap<(SegmentOrdinal, CacheKey), Vec<DocId>> = Default::default();
+
+        for doc_address in doc_addresses {
+            let store_reader = &self.inner.store_readers[doc_address.segment_ord as usize];
+            let cache_key = store_reader.cache_key(doc_address.doc_id)?;
+
+            groups
+                .entry((doc_address.segment_ord, cache_key))
+                .or_default()
+                .push(doc_address.doc_id);
+        }
+
+        let futures = groups
+            .into_iter()
+            .map(|((segment_ord, cache_key), doc_ids)| {
+                // Each group fetches documents from exactly one block and
+                // therefore gets an independent block cache of size one.
+                let store_reader =
+                    self.inner.store_readers[segment_ord as usize].fork_cache(1, &[cache_key]);
+
+                async move {
+                    let mut docs = Vec::new();
+
+                    for doc_id in doc_ids {
+                        let doc = store_reader.get_async(doc_id).await?;
+
+                        docs.push((
+                            DocAddress {
+                                segment_ord,
+                                doc_id,
+                            },
+                            doc,
+                        ));
+                    }
+
+                    Ok(docs)
+                }
+            });
+
+        Ok(futures)
+    }
+
     /// Access the schema associated with the index of this searcher.
     pub fn schema(&self) -> &Schema {
         &self.inner.schema

src/core/tests.rs

@@ -424,7 +424,7 @@ fn test_non_text_json_term_freq() {
     json_term_writer.set_fast_value(75u64);
     let postings = inv_idx
         .read_postings(
-            &json_term_writer.term(),
+            json_term_writer.term(),
             IndexRecordOption::WithFreqsAndPositions,
         )
         .unwrap()
@@ -462,7 +462,7 @@ fn test_non_text_json_term_freq_bitpacked() {
     json_term_writer.set_fast_value(75u64);
     let mut postings = inv_idx
         .read_postings(
-            &json_term_writer.term(),
+            json_term_writer.term(),
             IndexRecordOption::WithFreqsAndPositions,
         )
         .unwrap()
@@ -474,3 +474,60 @@ fn test_non_text_json_term_freq_bitpacked() {
         assert_eq!(postings.term_freq(), 1u32);
     }
 }
+
+#[cfg(feature = "quickwit")]
+#[test]
+fn test_get_many_docs() -> crate::Result<()> {
+    use futures::executor::block_on;
+    use futures::stream::{FuturesUnordered, StreamExt};
+
+    use crate::schema::{OwnedValue, STORED};
+    use crate::{DocAddress, TantivyError};
+
+    let mut schema_builder = Schema::builder();
+    let num_field = schema_builder.add_u64_field("num", STORED);
+    let schema = schema_builder.build();
+    let index = Index::create_in_ram(schema);
+    let mut index_writer: IndexWriter = index.writer_for_tests()?;
+    index_writer.set_merge_policy(Box::new(NoMergePolicy));
+    for i in 0..10u64 {
+        let doc = doc!(num_field=>i);
+        index_writer.add_document(doc)?;
+    }
+
+    index_writer.commit()?;
+    let segment_ids = index.searchable_segment_ids()?;
+    index_writer.merge(&segment_ids).wait().unwrap();
+
+    let searcher = index.reader()?.searcher();
+    assert_eq!(searcher.num_docs(), 10);
+
+    let doc_addresses = (0..10).map(|i| DocAddress::new(0, i));
+
+    let mut groups: FuturesUnordered<_> = searcher
+        .docs_async::<TantivyDocument>(doc_addresses)?
+        .collect();
+
+    let mut doc_nums = Vec::new();
+
+    block_on(async {
+        while let Some(group) = groups.next().await {
+            for (_doc_address, doc) in group? {
+                let num_value = doc.get_first(num_field).unwrap();
+
+                if let OwnedValue::U64(num) = num_value {
+                    doc_nums.push(*num);
+                } else {
+                    panic!("Expected u64 value");
+                }
+            }
+        }
+
+        Ok::<_, TantivyError>(())
+    })?;
+
+    doc_nums.sort();
+    assert_eq!(doc_nums, (0..10).collect::<Vec<u64>>());
+
+    Ok(())
+}

src/indexer/index_writer.rs

@@ -1651,7 +1651,6 @@ mod tests {
         force_end_merge: bool,
     ) -> crate::Result<Index> {
         let mut schema_builder = schema::Schema::builder();
-        let json_field = schema_builder.add_json_field("json", FAST | TEXT | STORED);
         let ip_field = schema_builder.add_ip_addr_field("ip", FAST | INDEXED | STORED);
         let ips_field = schema_builder
             .add_ip_addr_field("ips", IpAddrOptions::default().set_fast().set_indexed());
@@ -1730,9 +1729,7 @@ mod tests {
                             id_field=>id,
                         ))?;
                     } else {
-                        let json = json!({"date1": format!("2022-{id}-01T00:00:01Z"), "date2": format!("{id}-05-01T00:00:01Z"), "id": id, "ip": ip.to_string()});
                         index_writer.add_document(doc!(id_field=>id,
-                                json_field=>json,
                                 bytes_field => id.to_le_bytes().as_slice(),
                                 id_opt_field => id,
                                 ip_field => ip,

src/indexer/merger.rs

@@ -605,10 +605,6 @@ impl IndexMerger {
                             segment_postings.positions(&mut positions_buffer);
                             segment_postings.term_freq()
                         } else {
-                            // The positions_buffer may contain positions from the previous term
-                            // Existence of positions depend on the value type in JSON fields.
-                            // https://github.com/quickwit-oss/tantivy/issues/2283
-                            positions_buffer.clear();
                             0u32
                         };
 

src/indexer/segment_writer.rs

@@ -879,31 +879,6 @@ mod tests {
         assert_eq!(searcher.search(&phrase_query, &Count).unwrap(), 0);
     }
 
-    #[test]
-    fn test_json_term_with_numeric_merge_panic_regression_bug_2283() {
-        // https://github.com/quickwit-oss/tantivy/issues/2283
-        let mut schema_builder = Schema::builder();
-        let json = schema_builder.add_json_field("json", TEXT);
-        let schema = schema_builder.build();
-        let index = Index::create_in_ram(schema);
-        let mut writer = index.writer_for_tests().unwrap();
-        let doc = json!({"field": "a"});
-        writer.add_document(doc!(json=>doc)).unwrap();
-        writer.commit().unwrap();
-        let doc = json!({"field": "a", "id": 1});
-        writer.add_document(doc!(json=>doc.clone())).unwrap();
-        writer.commit().unwrap();
-
-        // Force Merge
-        writer.wait_merging_threads().unwrap();
-        let mut index_writer: IndexWriter = index.writer_for_tests().unwrap();
-        let segment_ids = index
-            .searchable_segment_ids()
-            .expect("Searchable segments failed.");
-        index_writer.merge(&segment_ids).wait().unwrap();
-        assert!(index_writer.wait_merging_threads().is_ok());
-    }
-
     #[test]
     fn test_bug_regression_1629_position_when_array_with_a_field_value_that_does_not_contain_any_token(
     ) {

src/query/exist_query.rs

@@ -1,4 +1,4 @@
-use std::borrow::Cow;
+use core::fmt::Debug;
 
 use columnar::{ColumnIndex, DynamicColumn};
 
@@ -14,7 +14,7 @@ use crate::{DocId, Score, TantivyError};
 /// All of the matched documents get the score 1.0.
 #[derive(Clone, Debug)]
 pub struct ExistsQuery {
-    field: Cow<'static, str>,
+    field_name: String,
 }
 
 impl ExistsQuery {
@@ -23,42 +23,40 @@ impl ExistsQuery {
     /// This query matches all documents with at least one non-null value in the specified field.
     /// This constructor never fails, but executing the search with this query will return an
     /// error if the specified field doesn't exists or is not a fast field.
-    pub fn new_exists_query<F: Into<Cow<'static, str>>>(field: F) -> ExistsQuery {
+    pub fn new_exists_query(field: String) -> ExistsQuery {
-        ExistsQuery {
+        ExistsQuery { field_name: field }
-            field: field.into(),
-        }
     }
 }
 
 impl Query for ExistsQuery {
     fn weight(&self, enable_scoring: EnableScoring) -> crate::Result<Box<dyn Weight>> {
         let schema = enable_scoring.schema();
-        let Some((field, _path)) = schema.find_field(&self.field) else {
+        let Some((field, _path)) = schema.find_field(&self.field_name) else {
-            return Err(TantivyError::FieldNotFound(self.field.to_string()));
+            return Err(TantivyError::FieldNotFound(self.field_name.clone()));
         };
         let field_type = schema.get_field_entry(field).field_type();
         if !field_type.is_fast() {
             return Err(TantivyError::SchemaError(format!(
                 "Field {} is not a fast field.",
-                self.field
+                self.field_name
             )));
         }
         Ok(Box::new(ExistsWeight {
-            field: self.field.clone(),
+            field_name: self.field_name.clone(),
         }))
     }
 }
 
 /// Weight associated with the `ExistsQuery` query.
 pub struct ExistsWeight {
-    field: Cow<'static, str>,
+    field_name: String,
 }
 
 impl Weight for ExistsWeight {
     fn scorer(&self, reader: &SegmentReader, boost: Score) -> crate::Result<Box<dyn Scorer>> {
         let fast_field_reader = reader.fast_fields();
         let dynamic_columns: crate::Result<Vec<DynamicColumn>> = fast_field_reader
-            .dynamic_column_handles(&self.field)?
+            .dynamic_column_handles(&self.field_name)?
             .into_iter()
             .map(|handle| handle.open().map_err(|io_error| io_error.into()))
             .collect();

src/query/range_query/range_query.rs

@@ -1,4 +1,3 @@
-use std::borrow::Cow;
 use std::io;
 use std::net::Ipv6Addr;
 use std::ops::{Bound, Range};
@@ -69,7 +68,7 @@ use crate::{DateTime, DocId, Score};
 /// ```
 #[derive(Clone, Debug)]
 pub struct RangeQuery {
-    field: Cow<'static, str>,
+    field: String,
     value_type: Type,
     lower_bound: Bound<Vec<u8>>,
     upper_bound: Bound<Vec<u8>>,
@@ -81,15 +80,15 @@ impl RangeQuery {
     ///
     /// If the value type is not correct, something may go terribly wrong when
     /// the `Weight` object is created.
-    pub fn new_term_bounds<F: Into<Cow<'static, str>>>(
+    pub fn new_term_bounds(
-        field: F,
+        field: String,
         value_type: Type,
         lower_bound: &Bound<Term>,
         upper_bound: &Bound<Term>,
     ) -> RangeQuery {
         let verify_and_unwrap_term = |val: &Term| val.serialized_value_bytes().to_owned();
         RangeQuery {
-            field: field.into(),
+            field,
             value_type,
             lower_bound: map_bound(lower_bound, verify_and_unwrap_term),
             upper_bound: map_bound(upper_bound, verify_and_unwrap_term),
@@ -101,7 +100,7 @@ impl RangeQuery {
     ///
     /// If the field is not of the type `i64`, tantivy
     /// will panic when the `Weight` object is created.
-    pub fn new_i64<F: Into<Cow<'static, str>>>(field: F, range: Range<i64>) -> RangeQuery {
+    pub fn new_i64(field: String, range: Range<i64>) -> RangeQuery {
         RangeQuery::new_i64_bounds(
             field,
             Bound::Included(range.start),
@@ -116,8 +115,8 @@ impl RangeQuery {
     ///
     /// If the field is not of the type `i64`, tantivy
     /// will panic when the `Weight` object is created.
-    pub fn new_i64_bounds<F: Into<Cow<'static, str>>>(
+    pub fn new_i64_bounds(
-        field: F,
+        field: String,
         lower_bound: Bound<i64>,
         upper_bound: Bound<i64>,
     ) -> RangeQuery {
@@ -127,7 +126,7 @@ impl RangeQuery {
                 .to_owned()
         };
         RangeQuery {
-            field: field.into(),
+            field,
             value_type: Type::I64,
             lower_bound: map_bound(&lower_bound, make_term_val),
             upper_bound: map_bound(&upper_bound, make_term_val),
@@ -139,7 +138,7 @@ impl RangeQuery {
     ///
     /// If the field is not of the type `f64`, tantivy
     /// will panic when the `Weight` object is created.
-    pub fn new_f64<F: Into<Cow<'static, str>>>(field: F, range: Range<f64>) -> RangeQuery {
+    pub fn new_f64(field: String, range: Range<f64>) -> RangeQuery {
         RangeQuery::new_f64_bounds(
             field,
             Bound::Included(range.start),
@@ -154,8 +153,8 @@ impl RangeQuery {
     ///
     /// If the field is not of the type `f64`, tantivy
     /// will panic when the `Weight` object is created.
-    pub fn new_f64_bounds<F: Into<Cow<'static, str>>>(
+    pub fn new_f64_bounds(
-        field: F,
+        field: String,
         lower_bound: Bound<f64>,
         upper_bound: Bound<f64>,
     ) -> RangeQuery {
@@ -165,7 +164,7 @@ impl RangeQuery {
                 .to_owned()
         };
         RangeQuery {
-            field: field.into(),
+            field,
             value_type: Type::F64,
             lower_bound: map_bound(&lower_bound, make_term_val),
             upper_bound: map_bound(&upper_bound, make_term_val),
@@ -180,8 +179,8 @@ impl RangeQuery {
     ///
     /// If the field is not of the type `u64`, tantivy
     /// will panic when the `Weight` object is created.
-    pub fn new_u64_bounds<F: Into<Cow<'static, str>>>(
+    pub fn new_u64_bounds(
-        field: F,
+        field: String,
         lower_bound: Bound<u64>,
         upper_bound: Bound<u64>,
     ) -> RangeQuery {
@@ -191,7 +190,7 @@ impl RangeQuery {
                 .to_owned()
         };
         RangeQuery {
-            field: field.into(),
+            field,
             value_type: Type::U64,
             lower_bound: map_bound(&lower_bound, make_term_val),
             upper_bound: map_bound(&upper_bound, make_term_val),
@@ -203,8 +202,8 @@ impl RangeQuery {
     ///
     /// If the field is not of the type `ip`, tantivy
     /// will panic when the `Weight` object is created.
-    pub fn new_ip_bounds<F: Into<Cow<'static, str>>>(
+    pub fn new_ip_bounds(
-        field: F,
+        field: String,
         lower_bound: Bound<Ipv6Addr>,
         upper_bound: Bound<Ipv6Addr>,
     ) -> RangeQuery {
@@ -214,7 +213,7 @@ impl RangeQuery {
                 .to_owned()
         };
         RangeQuery {
-            field: field.into(),
+            field,
             value_type: Type::IpAddr,
             lower_bound: map_bound(&lower_bound, make_term_val),
             upper_bound: map_bound(&upper_bound, make_term_val),
@@ -226,7 +225,7 @@ impl RangeQuery {
     ///
     /// If the field is not of the type `u64`, tantivy
     /// will panic when the `Weight` object is created.
-    pub fn new_u64<F: Into<Cow<'static, str>>>(field: F, range: Range<u64>) -> RangeQuery {
+    pub fn new_u64(field: String, range: Range<u64>) -> RangeQuery {
         RangeQuery::new_u64_bounds(
             field,
             Bound::Included(range.start),
@@ -241,8 +240,8 @@ impl RangeQuery {
     ///
     /// If the field is not of the type `date`, tantivy
     /// will panic when the `Weight` object is created.
-    pub fn new_date_bounds<F: Into<Cow<'static, str>>>(
+    pub fn new_date_bounds(
-        field: F,
+        field: String,
         lower_bound: Bound<DateTime>,
         upper_bound: Bound<DateTime>,
     ) -> RangeQuery {
@@ -252,7 +251,7 @@ impl RangeQuery {
                 .to_owned()
         };
         RangeQuery {
-            field: field.into(),
+            field,
             value_type: Type::Date,
             lower_bound: map_bound(&lower_bound, make_term_val),
             upper_bound: map_bound(&upper_bound, make_term_val),
@@ -264,7 +263,7 @@ impl RangeQuery {
     ///
     /// If the field is not of the type `date`, tantivy
     /// will panic when the `Weight` object is created.
-    pub fn new_date<F: Into<Cow<'static, str>>>(field: F, range: Range<DateTime>) -> RangeQuery {
+    pub fn new_date(field: String, range: Range<DateTime>) -> RangeQuery {
         RangeQuery::new_date_bounds(
             field,
             Bound::Included(range.start),
@@ -279,14 +278,14 @@ impl RangeQuery {
     ///
     /// If the field is not of the type `Str`, tantivy
     /// will panic when the `Weight` object is created.
-    pub fn new_str_bounds<F: Into<Cow<'static, str>>>(
+    pub fn new_str_bounds(
-        field: F,
+        field: String,
         lower_bound: Bound<&str>,
         upper_bound: Bound<&str>,
     ) -> RangeQuery {
         let make_term_val = |val: &&str| val.as_bytes().to_vec();
         RangeQuery {
-            field: field.into(),
+            field,
             value_type: Type::Str,
             lower_bound: map_bound(&lower_bound, make_term_val),
             upper_bound: map_bound(&upper_bound, make_term_val),
@@ -298,7 +297,7 @@ impl RangeQuery {
     ///
     /// If the field is not of the type `Str`, tantivy
     /// will panic when the `Weight` object is created.
-    pub fn new_str<F: Into<Cow<'static, str>>>(field: F, range: Range<&str>) -> RangeQuery {
+    pub fn new_str(field: String, range: Range<&str>) -> RangeQuery {
         RangeQuery::new_str_bounds(
             field,
             Bound::Included(range.start),
@@ -359,7 +358,7 @@ impl Query for RangeQuery {
                 let lower_bound = map_bound_res(&self.lower_bound, parse_ip_from_bytes)?;
                 let upper_bound = map_bound_res(&self.upper_bound, parse_ip_from_bytes)?;
                 Ok(Box::new(IPFastFieldRangeWeight::new(
-                    self.field.clone(),
+                    self.field.to_string(),
                     lower_bound,
                     upper_bound,
                 )))
@@ -374,14 +373,14 @@ impl Query for RangeQuery {
                 let lower_bound = map_bound(&self.lower_bound, parse_from_bytes);
                 let upper_bound = map_bound(&self.upper_bound, parse_from_bytes);
                 Ok(Box::new(FastFieldRangeWeight::new_u64_lenient(
-                    self.field.clone(),
+                    self.field.to_string(),
                     lower_bound,
                     upper_bound,
                 )))
             }
         } else {
             Ok(Box::new(RangeWeight {
-                field: self.field.clone(),
+                field: self.field.to_string(),
                 lower_bound: self.lower_bound.clone(),
                 upper_bound: self.upper_bound.clone(),
                 limit: self.limit,
@@ -391,7 +390,7 @@ impl Query for RangeQuery {
 }
 
 pub struct RangeWeight {
-    field: Cow<'static, str>,
+    field: String,
     lower_bound: Bound<Vec<u8>>,
     upper_bound: Bound<Vec<u8>>,
     limit: Option<u64>,

src/query/range_query/range_query_ip_fastfield.rs

@@ -2,7 +2,6 @@
 //! We use this variant only if the fastfield exists, otherwise the default in `range_query` is
 //! used, which uses the term dictionary + postings.
 
-use std::borrow::Cow;
 use std::net::Ipv6Addr;
 use std::ops::{Bound, RangeInclusive};
 
@@ -14,18 +13,14 @@ use crate::{DocId, DocSet, Score, SegmentReader, TantivyError};
 
 /// `IPFastFieldRangeWeight` uses the ip address fast field to execute range queries.
 pub struct IPFastFieldRangeWeight {
-    field: Cow<'static, str>,
+    field: String,
     lower_bound: Bound<Ipv6Addr>,
     upper_bound: Bound<Ipv6Addr>,
 }
 
 impl IPFastFieldRangeWeight {
     /// Creates a new IPFastFieldRangeWeight.
-    pub fn new(
+    pub fn new(field: String, lower_bound: Bound<Ipv6Addr>, upper_bound: Bound<Ipv6Addr>) -> Self {
-        field: Cow<'static, str>,
-        lower_bound: Bound<Ipv6Addr>,
-        upper_bound: Bound<Ipv6Addr>,
-    ) -> Self {
         Self {
             field,
             lower_bound,
@@ -176,7 +171,7 @@ pub mod tests {
         writer.commit().unwrap();
         let searcher = index.reader().unwrap().searcher();
         let range_weight = IPFastFieldRangeWeight {
-            field: Cow::Borrowed("ips"),
+            field: "ips".to_string(),
             lower_bound: Bound::Included(ip_addrs[1]),
             upper_bound: Bound::Included(ip_addrs[2]),
         };

src/query/range_query/range_query_u64_fastfield.rs

@@ -2,7 +2,6 @@
 //! We use this variant only if the fastfield exists, otherwise the default in `range_query` is
 //! used, which uses the term dictionary + postings.
 
-use std::borrow::Cow;
 use std::ops::{Bound, RangeInclusive};
 
 use columnar::{ColumnType, HasAssociatedColumnType, MonotonicallyMappableToU64};
@@ -15,7 +14,7 @@ use crate::{DocId, DocSet, Score, SegmentReader, TantivyError};
 /// `FastFieldRangeWeight` uses the fast field to execute range queries.
 #[derive(Clone, Debug)]
 pub struct FastFieldRangeWeight {
-    field: Cow<'static, str>,
+    field: String,
     lower_bound: Bound<u64>,
     upper_bound: Bound<u64>,
     column_type_opt: Option<ColumnType>,
@@ -24,7 +23,7 @@ pub struct FastFieldRangeWeight {
 impl FastFieldRangeWeight {
     /// Create a new FastFieldRangeWeight, using the u64 representation of any fast field.
     pub(crate) fn new_u64_lenient(
-        field: Cow<'static, str>,
+        field: String,
         lower_bound: Bound<u64>,
         upper_bound: Bound<u64>,
     ) -> Self {
@@ -40,7 +39,7 @@ impl FastFieldRangeWeight {
 
     /// Create a new `FastFieldRangeWeight` for a range of a u64-mappable type .
     pub fn new<T: HasAssociatedColumnType + MonotonicallyMappableToU64>(
-        field: Cow<'static, str>,
+        field: String,
         lower_bound: Bound<T>,
         upper_bound: Bound<T>,
     ) -> Self {
@@ -131,7 +130,6 @@ fn bound_to_value_range<T: MonotonicallyMappableToU64>(
 
 #[cfg(test)]
 pub mod tests {
-    use std::borrow::Cow;
     use std::ops::{Bound, RangeInclusive};
 
     use proptest::prelude::*;
@@ -216,7 +214,7 @@ pub mod tests {
         writer.commit().unwrap();
         let searcher = index.reader().unwrap().searcher();
         let range_query = FastFieldRangeWeight::new_u64_lenient(
-            Cow::Borrowed("test_field"),
+            "test_field".to_string(),
             Bound::Included(50_000),
             Bound::Included(50_002),
         );

src/query/regex_query.rs

@@ -63,7 +63,7 @@ impl RegexQuery {
     /// Creates a new RegexQuery from a given pattern
     pub fn from_pattern(regex_pattern: &str, field: Field) -> crate::Result<Self> {
         let regex = Regex::new(regex_pattern)
-            .map_err(|err| TantivyError::InvalidArgument(format!("RegexQueryError: {err}")))?;
+            .map_err(|_| TantivyError::InvalidArgument(regex_pattern.to_string()))?;
         Ok(RegexQuery::from_regex(regex, field))
     }
 
@@ -176,16 +176,4 @@ mod test {
         verify_regex_query(matching_one, matching_zero, reader);
         Ok(())
     }
-
-    #[test]
-    pub fn test_pattern_error() {
-        let (_reader, field) = build_test_index().unwrap();
-
-        match RegexQuery::from_pattern(r"(foo", field) {
-            Err(crate::TantivyError::InvalidArgument(msg)) => {
-                assert!(msg.contains("error: unclosed group"))
-            }
-            res => panic!("unexpected result: {:?}", res),
-        }
-    }
 }

src/store/mod.rs

@@ -37,6 +37,8 @@ mod reader;
 mod writer;
 pub use self::compressors::{Compressor, ZstdCompressor};
 pub use self::decompressors::Decompressor;
+#[cfg(feature = "quickwit")]
+pub(crate) use self::reader::CacheKey;
 pub(crate) use self::reader::DOCSTORE_CACHE_CAPACITY;
 pub use self::reader::{CacheStats, StoreReader};
 pub use self::writer::StoreWriter;

src/store/reader.rs

@@ -40,6 +40,15 @@ struct BlockCache {
 }
 
 impl BlockCache {
+    fn new(cache_num_blocks: usize) -> Self {
+        Self {
+            cache: NonZeroUsize::new(cache_num_blocks)
+                .map(|cache_num_blocks| Mutex::new(LruCache::new(cache_num_blocks))),
+            cache_hits: Default::default(),
+            cache_misses: Default::default(),
+        }
+    }
+
     fn get_from_cache(&self, pos: usize) -> Option<Block> {
         if let Some(block) = self
             .cache
@@ -81,6 +90,10 @@ impl BlockCache {
     }
 }
 
+/// Opaque cache key which indicates which documents are cached together.
+#[derive(Debug, Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Hash)]
+pub(crate) struct CacheKey(usize);
+
 #[derive(Debug, Default)]
 /// CacheStats for the `StoreReader`.
 pub struct CacheStats {
@@ -128,17 +141,35 @@ impl StoreReader {
         Ok(StoreReader {
             decompressor: footer.decompressor,
             data: data_file,
-            cache: BlockCache {
+            cache: BlockCache::new(cache_num_blocks),
-                cache: NonZeroUsize::new(cache_num_blocks)
-                    .map(|cache_num_blocks| Mutex::new(LruCache::new(cache_num_blocks))),
-                cache_hits: Default::default(),
-                cache_misses: Default::default(),
-            },
             skip_index: Arc::new(skip_index),
             space_usage,
         })
     }
 
+    /// Clones the given store reader with an independent block cache of the given size.
+    ///
+    /// `cache_keys` is used to seed the forked cache from the current cache
+    /// if some blocks are already available.
+    #[cfg(feature = "quickwit")]
+    pub(crate) fn fork_cache(&self, cache_num_blocks: usize, cache_keys: &[CacheKey]) -> Self {
+        let forked = Self {
+            decompressor: self.decompressor,
+            data: self.data.clone(),
+            cache: BlockCache::new(cache_num_blocks),
+            skip_index: Arc::clone(&self.skip_index),
+            space_usage: self.space_usage.clone(),
+        };
+
+        for &CacheKey(pos) in cache_keys {
+            if let Some(block) = self.cache.get_from_cache(pos) {
+                forked.cache.put_into_cache(pos, block);
+            }
+        }
+
+        forked
+    }
+
     pub(crate) fn block_checkpoints(&self) -> impl Iterator<Item = Checkpoint> + '_ {
         self.skip_index.checkpoints()
     }
@@ -152,6 +183,21 @@ impl StoreReader {
         self.cache.stats()
     }
 
+    /// Returns the cache key for a given document
+    ///
+    /// These keys are opaque and are not used with the public API,
+    /// but having the same cache key means that the documents
+    /// will only require one I/O and decompression operation
+    /// when retrieve from the same store reader consecutively.
+    ///
+    /// Note that looking up the cache key of a document
+    /// will not yet pull anything into the block cache.
+    #[cfg(feature = "quickwit")]
+    pub(crate) fn cache_key(&self, doc_id: DocId) -> crate::Result<CacheKey> {
+        let checkpoint = self.block_checkpoint(doc_id)?;
+        Ok(CacheKey(checkpoint.byte_range.start))
+    }
+
     /// Get checkpoint for `DocId`. The checkpoint can be used to load a block containing the
     /// document.
     ///

stacker/src/memory_arena.rs

@@ -189,11 +189,6 @@ struct Page {
 
 impl Page {
     fn new(page_id: usize) -> Page {
-        // We use 32-bits addresses.
-        // - 20 bits for the in-page addressing
-        // - 12 bits for the page id.
-        // This limits us to 2^12 - 1=4095 for the page id.
-        assert!(page_id < 4096);
         Page {
             page_id,
             len: 0,
@@ -243,7 +238,6 @@ impl Page {
 mod tests {
 
     use super::MemoryArena;
-    use crate::memory_arena::PAGE_SIZE;
 
     #[test]
     fn test_arena_allocate_slice() {
@@ -261,31 +255,6 @@ mod tests {
         assert_eq!(arena.slice(addr_b, b.len()), b);
     }
 
-    #[test]
-    fn test_arena_allocate_end_of_page() {
-        let mut arena = MemoryArena::default();
-
-        // A big block
-        let len_a = PAGE_SIZE - 2;
-        let addr_a = arena.allocate_space(len_a);
-        *arena.slice_mut(addr_a, len_a).last_mut().unwrap() = 1;
-
-        // Single bytes
-        let addr_b = arena.allocate_space(1);
-        arena.slice_mut(addr_b, 1)[0] = 2;
-
-        let addr_c = arena.allocate_space(1);
-        arena.slice_mut(addr_c, 1)[0] = 3;
-
-        let addr_d = arena.allocate_space(1);
-        arena.slice_mut(addr_d, 1)[0] = 4;
-
-        assert_eq!(arena.slice(addr_a, len_a)[len_a - 1], 1);
-        assert_eq!(arena.slice(addr_b, 1)[0], 2);
-        assert_eq!(arena.slice(addr_c, 1)[0], 3);
-        assert_eq!(arena.slice(addr_d, 1)[0], 4);
-    }
-
     #[derive(Clone, Copy, Debug, Eq, PartialEq)]
     struct MyTest {
         pub a: usize,

stacker/src/shared_arena_hashmap.rs

@@ -295,8 +295,6 @@ impl SharedArenaHashMap {
     /// will be in charge of returning a default value.
     /// If the key already as an associated value, then it will be passed
     /// `Some(previous_value)`.
-    ///
-    /// The key will be truncated to u16::MAX bytes.
     #[inline]
     pub fn mutate_or_create<V>(
         &mut self,
@@ -310,8 +308,6 @@ impl SharedArenaHashMap {
         if self.is_saturated() {
             self.resize();
         }
-        // Limit the key size to u16::MAX
-        let key = &key[..std::cmp::min(key.len(), u16::MAX as usize)];
         let hash = self.get_hash(key);
         let mut probe = self.probe(hash);
         let mut bucket = probe.next_probe();
@@ -383,36 +379,6 @@ mod tests {
         }
         assert_eq!(vanilla_hash_map.len(), 2);
     }
-
-    #[test]
-    fn test_long_key_truncation() {
-        // Keys longer than u16::MAX are truncated.
-        let mut memory_arena = MemoryArena::default();
-        let mut hash_map: SharedArenaHashMap = SharedArenaHashMap::default();
-        let key1 = (0..u16::MAX as usize).map(|i| i as u8).collect::<Vec<_>>();
-        hash_map.mutate_or_create(&key1, &mut memory_arena, |opt_val: Option<u32>| {
-            assert_eq!(opt_val, None);
-            4u32
-        });
-        // Due to truncation, this key is the same as key1
-        let key2 = (0..u16::MAX as usize + 1)
-            .map(|i| i as u8)
-            .collect::<Vec<_>>();
-        hash_map.mutate_or_create(&key2, &mut memory_arena, |opt_val: Option<u32>| {
-            assert_eq!(opt_val, Some(4));
-            3u32
-        });
-        let mut vanilla_hash_map = HashMap::new();
-        let iter_values = hash_map.iter(&memory_arena);
-        for (key, addr) in iter_values {
-            let val: u32 = memory_arena.read(addr);
-            vanilla_hash_map.insert(key.to_owned(), val);
-            assert_eq!(key.len(), key1[..].len());
-            assert_eq!(key, &key1[..])
-        }
-        assert_eq!(vanilla_hash_map.len(), 1); // Both map to the same key
-    }
-
     #[test]
     fn test_empty_hashmap() {
         let memory_arena = MemoryArena::default();