Store List of Fields in Segment

Fiels may be encoded in the columnar storage or in the inverted index
for JSON fields.
Add a new Segment file that contains the list of fields (schema +
encoded)

Cargo.toml

@@ -25,20 +25,20 @@ aho-corasick = "1.0"
 tantivy-fst = "0.5"
 memmap2 = { version = "0.9.0", optional = true }
 lz4_flex = { version = "0.11", default-features = false, optional = true }
-zstd = { version = "0.13", optional = true, default-features = false }
+zstd = { version = "0.13", default-features = false }
 tempfile = { version = "3.3.0", optional = true }
 log = "0.4.16"
 serde = { version = "1.0.136", features = ["derive"] }
 serde_json = "1.0.79"
 num_cpus = "1.13.1"
-fs4 = { version = "0.8.0", optional = true }
+fs4 = { version = "0.7.0", optional = true }
 levenshtein_automata = "0.2.1"
 uuid = { version = "1.0.0", features = ["v4", "serde"] }
 crossbeam-channel = "0.5.4"
 rust-stemmers = "1.2.0"
 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"
@@ -77,7 +77,6 @@ futures = "0.3.21"
 paste = "1.0.11"
 more-asserts = "0.3.1"
 rand_distr = "0.4.3"
-time = { version = "0.3.10", features = ["serde-well-known", "macros"] }
 
 [target.'cfg(not(windows))'.dev-dependencies]
 criterion = { version = "0.5", default-features = false }
@@ -106,7 +105,7 @@ mmap = ["fs4", "tempfile", "memmap2"]
 stopwords = []
 
 lz4-compression = ["lz4_flex"]
-zstd-compression = ["zstd"]
+zstd-compression = []
 
 failpoints = ["fail", "fail/failpoints"]
 unstable = [] # useful for benches.

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.**
-
-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
-an off-the-shelf search engine server, but rather a crate that can be used to build such a search engine.
+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
+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?
 

bitpacker/src/bitpacker.rs

@@ -125,8 +125,6 @@ impl BitUnpacker {
 
     // Decodes the range of bitpacked `u32` values with idx
     // in [start_idx, start_idx + output.len()).
-    // It is guaranteed to completely fill `output` and not read from it, so passing a vector with
-    // un-initialized values is safe.
     //
     // #Panics
     //
@@ -239,19 +237,7 @@ impl BitUnpacker {
         data: &[u8],
         positions: &mut Vec<u32>,
     ) {
-        // We use the code below instead of positions.resize(id_range.len(), 0u32) for performance
-        // reasons: on some queries, the CPU cost of memsetting the array and of using a bigger
-        // vector than necessary is noticeable (~5%).
-        // In particular, searches are a few percent faster when using reserve_exact() as below
-        // instead of reserve().
-        // The un-initialized values are safe as get_batch_u32s() completely fills `positions`
-        // and does not read from it.
-        positions.clear();
-        positions.reserve_exact(id_range.len());
-        #[allow(clippy::uninit_vec)]
-        unsafe {
-            positions.set_len(id_range.len());
-        }
+        positions.resize(id_range.len(), 0u32);
         self.get_batch_u32s(id_range.start, data, positions);
         crate::filter_vec::filter_vec_in_place(value_range, id_range.start, positions)
     }

columnar/src/column_index/merge/stacked.rs

@@ -111,7 +111,10 @@ fn stack_multivalued_indexes<'a>(
     let mut last_row_id = 0;
     let mut current_it = multivalued_indexes.next();
     Box::new(std::iter::from_fn(move || loop {
-        if let Some(row_id) = current_it.as_mut()?.next() {
+        let Some(multivalued_index) = current_it.as_mut() else {
+            return None;
+        };
+        if let Some(row_id) = multivalued_index.next() {
             last_row_id = offset + row_id;
             return Some(last_row_id);
         }

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_end = optional_index.rank(doc_id_range.end);
+                let row_start = optional_index.rank(doc_id.start);
+                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 start_docid = doc_id_range.start.min(end_docid);
+                let end_docid = doc_id.end.min(multivalued_index.num_docs() - 1) + 1;
+                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,
-        in_block_row_id: (row_id % ELEMENTS_PER_BLOCK) as u16,
+        block_id: (row_id / BLOCK_SIZE) 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
-    /// elements in the Set.
+    /// May panic if rank is greater than the number of 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));
-    test_optional_index_iter_aux(&block, 3 * ELEMENTS_PER_BLOCK);
+    block.extend((0..BLOCK_SIZE).map(|i| i + BLOCK_SIZE + 1));
+    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);

columnar/src/columnar/column_type.rs

@@ -58,7 +58,7 @@ impl ColumnType {
         self == &ColumnType::DateTime
     }
 
-    pub(crate) fn try_from_code(code: u8) -> Result<ColumnType, InvalidData> {
+    pub fn try_from_code(code: u8) -> Result<ColumnType, InvalidData> {
         COLUMN_TYPES.get(code as usize).copied().ok_or(InvalidData)
     }
 }

columnar/src/columnar/writer/mod.rs

@@ -333,7 +333,7 @@ impl ColumnarWriter {
         num_docs: RowId,
         old_to_new_row_ids: Option<&[RowId]>,
         wrt: &mut dyn io::Write,
-    ) -> io::Result<()> {
+    ) -> io::Result<Vec<(String, ColumnType)>> {
         let mut serializer = ColumnarSerializer::new(wrt);
         let mut columns: Vec<(&[u8], ColumnType, Addr)> = self
             .numerical_field_hash_map
@@ -374,7 +374,9 @@ impl ColumnarWriter {
 
         let (arena, buffers, dictionaries) = (&self.arena, &mut self.buffers, &self.dictionaries);
         let mut symbol_byte_buffer: Vec<u8> = Vec::new();
-        for (column_name, column_type, addr) in columns {
+        for (column_name, column_type, addr) in columns.iter() {
+            let column_type = *column_type;
+            let addr = *addr;
             match column_type {
                 ColumnType::Bool => {
                     let column_writer: ColumnWriter = self.bool_field_hash_map.read(addr);
@@ -485,7 +487,15 @@ impl ColumnarWriter {
             };
         }
         serializer.finalize(num_docs)?;
-        Ok(())
+        Ok(columns
+            .into_iter()
+            .map(|(column_name, column_type, _)| {
+                (
+                    String::from_utf8_lossy(column_name).to_string(),
+                    column_type,
+                )
+            })
+            .collect())
     }
 }
 

common/src/bitset.rs

@@ -6,7 +6,7 @@ use ownedbytes::OwnedBytes;
 
 use crate::ByteCount;
 
-#[derive(Clone, Copy, Eq, PartialEq)]
+#[derive(Clone, Copy, Eq, PartialEq, Hash)]
 pub struct TinySet(u64);
 
 impl fmt::Debug for TinySet {

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)
-        .map(|(left, (spaces, errors))| (left, (spaces.expect("multispace0 can't fail"), errors)))
+    opt_i(nom::character::complete::space0)(input)
+        .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

@@ -3,11 +3,11 @@ 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};
-use nom::multi::{many0, many1, separated_list0};
+use nom::multi::{many0, many1, separated_list0, separated_list1};
 use nom::sequence::{delimited, preceded, separated_pair, terminated, tuple};
 use nom::IResult;
 
@@ -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)),
-                separated_list0(multispace1, tuple((opt(occur_symbol), term_or_phrase))),
+                tuple((char('('), space0)),
+                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(multispace0, range_term_val()),
+            preceded(space0, alt((tag(">="), tag("<="), tag("<"), tag(">")))),
+            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(
-        lower_bound,
-        tuple((multispace1, tag("TO"), multispace1)),
-        upper_bound,
-    );
+    let lower_to_upper =
+        separated_pair(lower_bound, tuple((space1, tag("TO"), space1)), 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, _multispace0, lower, _multispace1, to, upper, upper_bound_kind),
-            errs,
-        )| {
+        |((lower_bound_kind, _space0, lower, _space1, 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)),
-                separated_list0(multispace1, map(simple_term, |(_, term)| term)),
+                tuple((char('['), space0)),
+                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)
 }
@@ -786,23 +780,27 @@ fn binary_operand(inp: &str) -> IResult<&str, BinaryOperand> {
 }
 
 fn aggregate_binary_expressions(
-    left: (Option<Occur>, UserInputAst),
-    others: Vec<(Option<BinaryOperand>, Option<Occur>, UserInputAst)>,
-) -> Result<UserInputAst, LenientErrorInternal> {
-    let mut leafs = Vec::with_capacity(others.len() + 1);
-    leafs.push((None, left.0, Some(left.1)));
-    leafs.extend(
-        others
-            .into_iter()
-            .map(|(operand, occur, ast)| (operand, occur, Some(ast))),
-    );
-    // the parameters we pass should statically guarantee we can't get errors
-    // (no prefix BinaryOperand is provided)
-    let (res, mut errors) = aggregate_infallible_expressions(leafs);
-    if errors.is_empty() {
-        Ok(res)
+    left: UserInputAst,
+    others: Vec<(BinaryOperand, UserInputAst)>,
+) -> UserInputAst {
+    let mut dnf: Vec<Vec<UserInputAst>> = vec![vec![left]];
+    for (operator, operand_ast) in others {
+        match operator {
+            BinaryOperand::And => {
+                if let Some(last) = dnf.last_mut() {
+                    last.push(operand_ast);
+                }
+            }
+            BinaryOperand::Or => {
+                dnf.push(vec![operand_ast]);
+            }
+        }
+    }
+    if dnf.len() == 1 {
+        UserInputAst::and(dnf.into_iter().next().unwrap()) //< safe
     } else {
-        Err(errors.swap_remove(0))
+        let conjunctions = dnf.into_iter().map(UserInputAst::and).collect();
+        UserInputAst::or(conjunctions)
     }
 }
 
@@ -818,10 +816,30 @@ fn aggregate_infallible_expressions(
         return (UserInputAst::empty_query(), err);
     }
 
+    let use_operand = leafs.iter().any(|(operand, _, _)| operand.is_some());
+    let all_operand = leafs
+        .iter()
+        .skip(1)
+        .all(|(operand, _, _)| operand.is_some());
     let early_operand = leafs
         .iter()
         .take(1)
         .all(|(operand, _, _)| operand.is_some());
+    let use_occur = leafs.iter().any(|(_, occur, _)| occur.is_some());
+
+    if use_operand && use_occur {
+        err.push(LenientErrorInternal {
+            pos: 0,
+            message: "Use of mixed occur and boolean operator".to_string(),
+        });
+    }
+
+    if use_operand && !all_operand {
+        err.push(LenientErrorInternal {
+            pos: 0,
+            message: "Missing boolean operator".to_string(),
+        });
+    }
 
     if early_operand {
         err.push(LenientErrorInternal {
@@ -848,15 +866,7 @@ fn aggregate_infallible_expressions(
                     Some(BinaryOperand::And) => Some(Occur::Must),
                     _ => Some(Occur::Should),
                 };
-                if occur == &Some(Occur::MustNot) && default_op == Some(Occur::Should) {
-                    // if occur is MustNot *and* operation is OR, we synthetize a ShouldNot
-                    clauses.push(vec![(
-                        Some(Occur::Should),
-                        ast.clone().unary(Occur::MustNot),
-                    )])
-                } else {
-                    clauses.push(vec![(occur.or(default_op), ast.clone())]);
-                }
+                clauses.push(vec![(occur.or(default_op), ast.clone())]);
             }
             None => {
                 let default_op = match next_operator {
@@ -864,15 +874,7 @@ fn aggregate_infallible_expressions(
                     Some(BinaryOperand::Or) => Some(Occur::Should),
                     None => None,
                 };
-                if occur == &Some(Occur::MustNot) && default_op == Some(Occur::Should) {
-                    // if occur is MustNot *and* operation is OR, we synthetize a ShouldNot
-                    clauses.push(vec![(
-                        Some(Occur::Should),
-                        ast.clone().unary(Occur::MustNot),
-                    )])
-                } else {
-                    clauses.push(vec![(occur.or(default_op), ast.clone())])
-                }
+                clauses.push(vec![(occur.or(default_op), ast.clone())])
             }
         }
     }
@@ -889,12 +891,7 @@ fn aggregate_infallible_expressions(
             }
         }
         Some(BinaryOperand::Or) => {
-            if last_occur == Some(Occur::MustNot) {
-                // if occur is MustNot *and* operation is OR, we synthetize a ShouldNot
-                clauses.push(vec![(Some(Occur::Should), last_ast.unary(Occur::MustNot))]);
-            } else {
-                clauses.push(vec![(last_occur.or(Some(Occur::Should)), last_ast)]);
-            }
+            clauses.push(vec![(last_occur.or(Some(Occur::Should)), last_ast)]);
         }
         None => clauses.push(vec![(last_occur, last_ast)]),
     }
@@ -920,29 +917,35 @@ fn aggregate_infallible_expressions(
     }
 }
 
-fn operand_leaf(inp: &str) -> IResult<&str, (Option<BinaryOperand>, Option<Occur>, UserInputAst)> {
-    map(
-        tuple((
-            terminated(opt(binary_operand), multispace0),
-            terminated(occur_leaf, multispace0),
-        )),
-        |(operand, (occur, ast))| (operand, occur, ast),
-    )(inp)
+fn operand_leaf(inp: &str) -> IResult<&str, (BinaryOperand, UserInputAst)> {
+    tuple((
+        terminated(binary_operand, space0),
+        terminated(boosted_leaf, space0),
+    ))(inp)
 }
 
 fn ast(inp: &str) -> IResult<&str, UserInputAst> {
-    let boolean_expr = map_res(
-        separated_pair(occur_leaf, multispace1, many1(operand_leaf)),
+    let boolean_expr = map(
+        separated_pair(boosted_leaf, space1, many1(operand_leaf)),
         |(left, right)| aggregate_binary_expressions(left, right),
     );
-    let single_leaf = map(occur_leaf, |(occur, ast)| {
-        if occur == Some(Occur::MustNot) {
-            ast.unary(Occur::MustNot)
+    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) {
+                Occur::Must | Occur::Should => ast,
+                Occur::MustNot => UserInputAst::Clause(vec![(Some(Occur::MustNot), ast)]),
+            }
         } else {
-            ast
+            UserInputAst::Clause(subqueries.into_iter().collect())
         }
     });
-    delimited(multispace0, alt((boolean_expr, single_leaf)), multispace0)(inp)
+
+    delimited(
+        space0,
+        alt((boolean_expr, whitespace_separated_leaves)),
+        space0,
+    )(inp)
 }
 
 fn ast_infallible(inp: &str) -> JResult<&str, UserInputAst> {
@@ -966,7 +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)
 }
@@ -1142,43 +1145,24 @@ 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)");
-        test_parse_query_to_ast_helper("a OR b aaa", "(?a ?b *aaa)");
-        test_parse_query_to_ast_helper("a AND b aaa", "(?(+a +b) *aaa)");
-        test_parse_query_to_ast_helper("aaa a OR b ", "(*aaa ?a ?b)");
-        test_parse_query_to_ast_helper("aaa ccc a OR b ", "(*aaa *ccc ?a ?b)");
-        test_parse_query_to_ast_helper("aaa a AND b ", "(*aaa ?(+a +b))");
-        test_parse_query_to_ast_helper("aaa ccc a AND b ", "(*aaa *ccc ?(+a +b))");
+        test_is_parse_err("a OR b aaa", "(?a ?b *aaa)");
+        test_is_parse_err("a AND b aaa", "(?(+a +b) *aaa)");
+        test_is_parse_err("aaa a OR b ", "(*aaa ?a ?b)");
+        test_is_parse_err("aaa ccc a OR b ", "(*aaa *ccc ?a ?b)");
+        test_is_parse_err("aaa a AND b ", "(*aaa ?(+a +b))");
+        test_is_parse_err("aaa ccc a AND b ", "(*aaa *ccc ?(+a +b))");
     }
 
     #[test]
     fn test_parse_mixed_bool_occur() {
-        test_parse_query_to_ast_helper("+a OR +b", "(+a +b)");
-
-        test_parse_query_to_ast_helper("a AND -b", "(+a -b)");
-        test_parse_query_to_ast_helper("-a AND b", "(-a +b)");
-        test_parse_query_to_ast_helper("a AND NOT b", "(+a +(-b))");
-        test_parse_query_to_ast_helper("NOT a AND b", "(+(-a) +b)");
-
-        test_parse_query_to_ast_helper("a AND NOT b AND c", "(+a +(-b) +c)");
-        test_parse_query_to_ast_helper("a AND -b AND c", "(+a -b +c)");
-
-        test_parse_query_to_ast_helper("a OR -b", "(?a ?(-b))");
-        test_parse_query_to_ast_helper("-a OR b", "(?(-a) ?b)");
-        test_parse_query_to_ast_helper("a OR NOT b", "(?a ?(-b))");
-        test_parse_query_to_ast_helper("NOT a OR b", "(?(-a) ?b)");
-
-        test_parse_query_to_ast_helper("a OR NOT b OR c", "(?a ?(-b) ?c)");
-        test_parse_query_to_ast_helper("a OR -b OR c", "(?a ?(-b) ?c)");
-
-        test_parse_query_to_ast_helper("a OR b +aaa", "(?a ?b +aaa)");
-        test_parse_query_to_ast_helper("a AND b -aaa", "(?(+a +b) -aaa)");
-        test_parse_query_to_ast_helper("+a OR +b aaa", "(+a +b *aaa)");
-        test_parse_query_to_ast_helper("-a AND -b aaa", "(?(-a -b) *aaa)");
-        test_parse_query_to_ast_helper("-aaa +ccc -a OR b ", "(-aaa +ccc ?(-a) ?b)");
+        test_is_parse_err("a OR b +aaa", "(?a ?b +aaa)");
+        test_is_parse_err("a AND b -aaa", "(?(+a +b) -aaa)");
+        test_is_parse_err("+a OR +b aaa", "(+a +b *aaa)");
+        test_is_parse_err("-a AND -b aaa", "(?(-a -b) *aaa)");
+        test_is_parse_err("-aaa +ccc -a OR b ", "(-aaa +ccc -a ?b)");
     }
 
     #[test]

src/aggregation/agg_bench.rs

@@ -290,41 +290,6 @@ mod bench {
         });
     }
 
-    bench_all_cardinalities!(bench_aggregation_terms_many_with_top_hits_agg);
-
-    fn bench_aggregation_terms_many_with_top_hits_agg_card(
-        b: &mut Bencher,
-        cardinality: Cardinality,
-    ) {
-        let index = get_test_index_bench(cardinality).unwrap();
-        let reader = index.reader().unwrap();
-
-        b.iter(|| {
-            let agg_req: Aggregations = serde_json::from_value(json!({
-                "my_texts": {
-                    "terms": { "field": "text_many_terms" },
-                    "aggs": {
-                        "top_hits": { "top_hits":
-                            {
-                                "sort": [
-                                    { "score": "desc" }
-                                ],
-                                "size": 2,
-                                "doc_value_fields": ["score_f64"]
-                            }
-                        }
-                    }
-                },
-            }))
-            .unwrap();
-
-            let collector = get_collector(agg_req);
-
-            let searcher = reader.searcher();
-            searcher.search(&AllQuery, &collector).unwrap()
-        });
-    }
-
     bench_all_cardinalities!(bench_aggregation_terms_many_with_sub_agg);
 
     fn bench_aggregation_terms_many_with_sub_agg_card(b: &mut Bencher, cardinality: Cardinality) {

src/aggregation/agg_req.rs

@@ -35,7 +35,7 @@ use super::bucket::{
 };
 use super::metric::{
     AverageAggregation, CountAggregation, MaxAggregation, MinAggregation,
-    PercentilesAggregationReq, StatsAggregation, SumAggregation, TopHitsAggregation,
+    PercentilesAggregationReq, StatsAggregation, SumAggregation,
 };
 
 /// The top-level aggregation request structure, which contains [`Aggregation`] and their user
@@ -93,12 +93,7 @@ impl Aggregation {
     }
 
     fn get_fast_field_names(&self, fast_field_names: &mut HashSet<String>) {
-        fast_field_names.extend(
-            self.agg
-                .get_fast_field_names()
-                .iter()
-                .map(|s| s.to_string()),
-        );
+        fast_field_names.insert(self.agg.get_fast_field_name().to_string());
         fast_field_names.extend(get_fast_field_names(&self.sub_aggregation));
     }
 }
@@ -152,27 +147,23 @@ pub enum AggregationVariants {
     /// Computes the sum of the extracted values.
     #[serde(rename = "percentiles")]
     Percentiles(PercentilesAggregationReq),
-    /// Finds the top k values matching some order
-    #[serde(rename = "top_hits")]
-    TopHits(TopHitsAggregation),
 }
 
 impl AggregationVariants {
-    /// Returns the name of the fields used by the aggregation.
-    pub fn get_fast_field_names(&self) -> Vec<&str> {
+    /// Returns the name of the field used by the aggregation.
+    pub fn get_fast_field_name(&self) -> &str {
         match self {
-            AggregationVariants::Terms(terms) => vec![terms.field.as_str()],
-            AggregationVariants::Range(range) => vec![range.field.as_str()],
-            AggregationVariants::Histogram(histogram) => vec![histogram.field.as_str()],
-            AggregationVariants::DateHistogram(histogram) => vec![histogram.field.as_str()],
-            AggregationVariants::Average(avg) => vec![avg.field_name()],
-            AggregationVariants::Count(count) => vec![count.field_name()],
-            AggregationVariants::Max(max) => vec![max.field_name()],
-            AggregationVariants::Min(min) => vec![min.field_name()],
-            AggregationVariants::Stats(stats) => vec![stats.field_name()],
-            AggregationVariants::Sum(sum) => vec![sum.field_name()],
-            AggregationVariants::Percentiles(per) => vec![per.field_name()],
-            AggregationVariants::TopHits(top_hits) => top_hits.field_names(),
+            AggregationVariants::Terms(terms) => terms.field.as_str(),
+            AggregationVariants::Range(range) => range.field.as_str(),
+            AggregationVariants::Histogram(histogram) => histogram.field.as_str(),
+            AggregationVariants::DateHistogram(histogram) => histogram.field.as_str(),
+            AggregationVariants::Average(avg) => avg.field_name(),
+            AggregationVariants::Count(count) => count.field_name(),
+            AggregationVariants::Max(max) => max.field_name(),
+            AggregationVariants::Min(min) => min.field_name(),
+            AggregationVariants::Stats(stats) => stats.field_name(),
+            AggregationVariants::Sum(sum) => sum.field_name(),
+            AggregationVariants::Percentiles(per) => per.field_name(),
         }
     }
 

src/aggregation/agg_req_with_accessor.rs

@@ -1,9 +1,6 @@
 //! This will enhance the request tree with access to the fastfield and metadata.
 
-use std::collections::HashMap;
-use std::io;
-
-use columnar::{Column, ColumnBlockAccessor, ColumnType, DynamicColumn, StrColumn};
+use columnar::{Column, ColumnBlockAccessor, ColumnType, StrColumn};
 
 use super::agg_limits::ResourceLimitGuard;
 use super::agg_req::{Aggregation, AggregationVariants, Aggregations};
@@ -17,7 +14,7 @@ use super::metric::{
 use super::segment_agg_result::AggregationLimits;
 use super::VecWithNames;
 use crate::aggregation::{f64_to_fastfield_u64, Key};
-use crate::{SegmentOrdinal, SegmentReader};
+use crate::SegmentReader;
 
 #[derive(Default)]
 pub(crate) struct AggregationsWithAccessor {
@@ -35,7 +32,6 @@ impl AggregationsWithAccessor {
 }
 
 pub struct AggregationWithAccessor {
-    pub(crate) segment_ordinal: SegmentOrdinal,
     /// In general there can be buckets without fast field access, e.g. buckets that are created
     /// based on search terms. That is not that case currently, but eventually this needs to be
     /// Option or moved.
@@ -48,16 +44,10 @@ pub struct AggregationWithAccessor {
     pub(crate) limits: ResourceLimitGuard,
     pub(crate) column_block_accessor: ColumnBlockAccessor<u64>,
     /// Used for missing term aggregation, which checks all columns for existence.
-    /// And also for `top_hits` aggregation, which may sort on multiple fields.
     /// By convention the missing aggregation is chosen, when this property is set
     /// (instead bein set in `agg`).
     /// If this needs to used by other aggregations, we need to refactor this.
-    // NOTE: we can make all other aggregations use this instead of the `accessor` and `field_type`
-    // (making them obsolete) But will it have a performance impact?
-    pub(crate) accessors: Vec<(Column<u64>, ColumnType)>,
-    /// Map field names to all associated column accessors.
-    /// This field is used for `docvalue_fields`, which is currently only supported for `top_hits`.
-    pub(crate) value_accessors: HashMap<String, Vec<DynamicColumn>>,
+    pub(crate) accessors: Vec<Column<u64>>,
     pub(crate) agg: Aggregation,
 }
 
@@ -67,55 +57,19 @@ impl AggregationWithAccessor {
         agg: &Aggregation,
         sub_aggregation: &Aggregations,
         reader: &SegmentReader,
-        segment_ordinal: SegmentOrdinal,
         limits: AggregationLimits,
     ) -> crate::Result<Vec<AggregationWithAccessor>> {
-        let mut agg = agg.clone();
-
-        let add_agg_with_accessor = |agg: &Aggregation,
-                                     accessor: Column<u64>,
+        let add_agg_with_accessor = |accessor: Column<u64>,
                                      column_type: ColumnType,
                                      aggs: &mut Vec<AggregationWithAccessor>|
          -> crate::Result<()> {
             let res = AggregationWithAccessor {
-                segment_ordinal,
                 accessor,
-                accessors: Default::default(),
-                value_accessors: Default::default(),
+                accessors: Vec::new(),
                 field_type: column_type,
                 sub_aggregation: get_aggs_with_segment_accessor_and_validate(
                     sub_aggregation,
                     reader,
-                    segment_ordinal,
-                    &limits,
-                )?,
-                agg: agg.clone(),
-                limits: limits.new_guard(),
-                missing_value_for_accessor: None,
-                str_dict_column: None,
-                column_block_accessor: Default::default(),
-            };
-            aggs.push(res);
-            Ok(())
-        };
-
-        let add_agg_with_accessors = |agg: &Aggregation,
-                                      accessors: Vec<(Column<u64>, ColumnType)>,
-                                      aggs: &mut Vec<AggregationWithAccessor>,
-                                      value_accessors: HashMap<String, Vec<DynamicColumn>>|
-         -> crate::Result<()> {
-            let (accessor, field_type) = accessors.first().expect("at least one accessor");
-            let res = AggregationWithAccessor {
-                segment_ordinal,
-                // TODO: We should do away with the `accessor` field altogether
-                accessor: accessor.clone(),
-                value_accessors,
-                field_type: *field_type,
-                accessors,
-                sub_aggregation: get_aggs_with_segment_accessor_and_validate(
-                    sub_aggregation,
-                    reader,
-                    segment_ordinal,
                     &limits,
                 )?,
                 agg: agg.clone(),
@@ -130,36 +84,32 @@ impl AggregationWithAccessor {
 
         let mut res: Vec<AggregationWithAccessor> = Vec::new();
         use AggregationVariants::*;
-
-        match agg.agg {
+        match &agg.agg {
             Range(RangeAggregation {
-                field: ref field_name,
-                ..
+                field: field_name, ..
             }) => {
                 let (accessor, column_type) =
                     get_ff_reader(reader, field_name, Some(get_numeric_or_date_column_types()))?;
-                add_agg_with_accessor(&agg, accessor, column_type, &mut res)?;
+                add_agg_with_accessor(accessor, column_type, &mut res)?;
             }
             Histogram(HistogramAggregation {
-                field: ref field_name,
-                ..
+                field: field_name, ..
             }) => {
                 let (accessor, column_type) =
                     get_ff_reader(reader, field_name, Some(get_numeric_or_date_column_types()))?;
-                add_agg_with_accessor(&agg, accessor, column_type, &mut res)?;
+                add_agg_with_accessor(accessor, column_type, &mut res)?;
             }
             DateHistogram(DateHistogramAggregationReq {
-                field: ref field_name,
-                ..
+                field: field_name, ..
             }) => {
                 let (accessor, column_type) =
                     // Only DateTime is supported for DateHistogram
                     get_ff_reader(reader, field_name, Some(&[ColumnType::DateTime]))?;
-                add_agg_with_accessor(&agg, accessor, column_type, &mut res)?;
+                add_agg_with_accessor(accessor, column_type, &mut res)?;
             }
             Terms(TermsAggregation {
-                field: ref field_name,
-                ref missing,
+                field: field_name,
+                missing,
                 ..
             }) => {
                 let str_dict_column = reader.fast_fields().str(field_name)?;
@@ -169,8 +119,8 @@ impl AggregationWithAccessor {
                     ColumnType::F64,
                     ColumnType::Str,
                     ColumnType::DateTime,
-                    ColumnType::Bool,
                     // ColumnType::Bytes Unsupported
+                    // ColumnType::Bool Unsupported
                     // ColumnType::IpAddr Unsupported
                 ];
 
@@ -212,11 +162,24 @@ impl AggregationWithAccessor {
                     let column_and_types =
                         get_all_ff_reader_or_empty(reader, field_name, None, fallback_type)?;
 
-                    let accessors = column_and_types
-                        .iter()
-                        .map(|c_t| (c_t.0.clone(), c_t.1))
-                        .collect();
-                    add_agg_with_accessors(&agg, accessors, &mut res, Default::default())?;
+                    let accessors: Vec<Column> =
+                        column_and_types.iter().map(|(a, _)| a.clone()).collect();
+                    let agg_wit_acc = AggregationWithAccessor {
+                        missing_value_for_accessor: None,
+                        accessor: accessors[0].clone(),
+                        accessors,
+                        field_type: ColumnType::U64,
+                        sub_aggregation: get_aggs_with_segment_accessor_and_validate(
+                            sub_aggregation,
+                            reader,
+                            &limits,
+                        )?,
+                        agg: agg.clone(),
+                        str_dict_column: str_dict_column.clone(),
+                        limits: limits.new_guard(),
+                        column_block_accessor: Default::default(),
+                    };
+                    res.push(agg_wit_acc);
                 }
 
                 for (accessor, column_type) in column_and_types {
@@ -226,25 +189,21 @@ impl AggregationWithAccessor {
                         missing.clone()
                     };
 
-                    let missing_value_for_accessor = if let Some(missing) =
-                        missing_value_term_agg.as_ref()
-                    {
-                        get_missing_val(column_type, missing, agg.agg.get_fast_field_names()[0])?
-                    } else {
-                        None
-                    };
+                    let missing_value_for_accessor =
+                        if let Some(missing) = missing_value_term_agg.as_ref() {
+                            get_missing_val(column_type, missing, agg.agg.get_fast_field_name())?
+                        } else {
+                            None
+                        };
 
                     let agg = AggregationWithAccessor {
-                        segment_ordinal,
                         missing_value_for_accessor,
                         accessor,
-                        accessors: Default::default(),
-                        value_accessors: Default::default(),
+                        accessors: Vec::new(),
                         field_type: column_type,
                         sub_aggregation: get_aggs_with_segment_accessor_and_validate(
                             sub_aggregation,
                             reader,
-                            segment_ordinal,
                             &limits,
                         )?,
                         agg: agg.clone(),
@@ -256,63 +215,34 @@ impl AggregationWithAccessor {
                 }
             }
             Average(AverageAggregation {
-                field: ref field_name,
-                ..
+                field: field_name, ..
             })
             | Count(CountAggregation {
-                field: ref field_name,
-                ..
+                field: field_name, ..
             })
             | Max(MaxAggregation {
-                field: ref field_name,
-                ..
+                field: field_name, ..
             })
             | Min(MinAggregation {
-                field: ref field_name,
-                ..
+                field: field_name, ..
             })
             | Stats(StatsAggregation {
-                field: ref field_name,
-                ..
+                field: field_name, ..
             })
             | Sum(SumAggregation {
-                field: ref field_name,
-                ..
+                field: field_name, ..
             }) => {
                 let (accessor, column_type) =
                     get_ff_reader(reader, field_name, Some(get_numeric_or_date_column_types()))?;
-                add_agg_with_accessor(&agg, accessor, column_type, &mut res)?;
+                add_agg_with_accessor(accessor, column_type, &mut res)?;
             }
-            Percentiles(ref percentiles) => {
+            Percentiles(percentiles) => {
                 let (accessor, column_type) = get_ff_reader(
                     reader,
                     percentiles.field_name(),
                     Some(get_numeric_or_date_column_types()),
                 )?;
-                add_agg_with_accessor(&agg, accessor, column_type, &mut res)?;
-            }
-            TopHits(ref mut top_hits) => {
-                top_hits.validate_and_resolve(reader.fast_fields().columnar())?;
-                let accessors: Vec<(Column<u64>, ColumnType)> = top_hits
-                    .field_names()
-                    .iter()
-                    .map(|field| {
-                        get_ff_reader(reader, field, Some(get_numeric_or_date_column_types()))
-                    })
-                    .collect::<crate::Result<_>>()?;
-
-                let value_accessors = top_hits
-                    .value_field_names()
-                    .iter()
-                    .map(|field_name| {
-                        Ok((
-                            field_name.to_string(),
-                            get_dynamic_columns(reader, field_name)?,
-                        ))
-                    })
-                    .collect::<crate::Result<_>>()?;
-
-                add_agg_with_accessors(&agg, accessors, &mut res, value_accessors)?;
+                add_agg_with_accessor(accessor, column_type, &mut res)?;
             }
         };
 
@@ -354,7 +284,6 @@ fn get_numeric_or_date_column_types() -> &'static [ColumnType] {
 pub(crate) fn get_aggs_with_segment_accessor_and_validate(
     aggs: &Aggregations,
     reader: &SegmentReader,
-    segment_ordinal: SegmentOrdinal,
     limits: &AggregationLimits,
 ) -> crate::Result<AggregationsWithAccessor> {
     let mut aggss = Vec::new();
@@ -363,7 +292,6 @@ pub(crate) fn get_aggs_with_segment_accessor_and_validate(
             agg,
             agg.sub_aggregation(),
             reader,
-            segment_ordinal,
             limits.clone(),
         )?;
         for agg in aggs {
@@ -393,19 +321,6 @@ fn get_ff_reader(
     Ok(ff_field_with_type)
 }
 
-fn get_dynamic_columns(
-    reader: &SegmentReader,
-    field_name: &str,
-) -> crate::Result<Vec<columnar::DynamicColumn>> {
-    let ff_fields = reader.fast_fields().dynamic_column_handles(field_name)?;
-    let cols = ff_fields
-        .iter()
-        .map(|h| h.open())
-        .collect::<io::Result<_>>()?;
-    assert!(!ff_fields.is_empty(), "field {} not found", field_name);
-    Ok(cols)
-}
-
 /// Get all fast field reader or empty as default.
 ///
 /// Is guaranteed to return at least one column.

src/aggregation/agg_result.rs

@@ -8,7 +8,7 @@ use rustc_hash::FxHashMap;
 use serde::{Deserialize, Serialize};
 
 use super::bucket::GetDocCount;
-use super::metric::{PercentilesMetricResult, SingleMetricResult, Stats, TopHitsMetricResult};
+use super::metric::{PercentilesMetricResult, SingleMetricResult, Stats};
 use super::{AggregationError, Key};
 use crate::TantivyError;
 
@@ -90,10 +90,8 @@ pub enum MetricResult {
     Stats(Stats),
     /// Sum metric result.
     Sum(SingleMetricResult),
-    /// Percentiles metric result.
+    /// Sum metric result.
     Percentiles(PercentilesMetricResult),
-    /// Top hits metric result
-    TopHits(TopHitsMetricResult),
 }
 
 impl MetricResult {
@@ -108,9 +106,6 @@ impl MetricResult {
             MetricResult::Percentiles(_) => Err(TantivyError::AggregationError(
                 AggregationError::InvalidRequest("percentiles can't be used to order".to_string()),
             )),
-            MetricResult::TopHits(_) => Err(TantivyError::AggregationError(
-                AggregationError::InvalidRequest("top_hits can't be used to order".to_string()),
-            )),
         }
     }
 }

src/aggregation/agg_tests.rs

@@ -587,9 +587,6 @@ fn test_aggregation_on_json_object() {
     let schema = schema_builder.build();
     let index = Index::create_in_ram(schema);
     let mut index_writer: IndexWriter = index.writer_for_tests().unwrap();
-    index_writer
-        .add_document(doc!(json => json!({"color": "red"})))
-        .unwrap();
     index_writer
         .add_document(doc!(json => json!({"color": "red"})))
         .unwrap();
@@ -617,8 +614,8 @@ fn test_aggregation_on_json_object() {
         &serde_json::json!({
             "jsonagg": {
                 "buckets": [
-                    {"doc_count": 2, "key": "red"},
                     {"doc_count": 1, "key": "blue"},
+                    {"doc_count": 1, "key": "red"}
                 ],
                 "doc_count_error_upper_bound": 0,
                 "sum_other_doc_count": 0
@@ -640,9 +637,6 @@ fn test_aggregation_on_nested_json_object() {
     index_writer
         .add_document(doc!(json => json!({"color.dot": "blue", "color": {"nested":"blue"} })))
         .unwrap();
-    index_writer
-        .add_document(doc!(json => json!({"color.dot": "blue", "color": {"nested":"blue"} })))
-        .unwrap();
     index_writer.commit().unwrap();
     let reader = index.reader().unwrap();
     let searcher = reader.searcher();
@@ -670,7 +664,7 @@ fn test_aggregation_on_nested_json_object() {
         &serde_json::json!({
             "jsonagg1": {
                 "buckets": [
-                    {"doc_count": 2, "key": "blue"},
+                    {"doc_count": 1, "key": "blue"},
                     {"doc_count": 1, "key": "red"}
                 ],
                 "doc_count_error_upper_bound": 0,
@@ -678,7 +672,7 @@ fn test_aggregation_on_nested_json_object() {
             },
             "jsonagg2": {
                 "buckets": [
-                    {"doc_count": 2, "key": "blue"},
+                    {"doc_count": 1, "key": "blue"},
                     {"doc_count": 1, "key": "red"}
                 ],
                 "doc_count_error_upper_bound": 0,
@@ -820,12 +814,6 @@ fn test_aggregation_on_json_object_mixed_types() {
         .unwrap();
     index_writer.commit().unwrap();
     // => Segment with all values text
-    index_writer
-        .add_document(doc!(json => json!({"mixed_type": "blue"})))
-        .unwrap();
-    index_writer
-        .add_document(doc!(json => json!({"mixed_type": "blue"})))
-        .unwrap();
     index_writer
         .add_document(doc!(json => json!({"mixed_type": "blue"})))
         .unwrap();
@@ -837,9 +825,6 @@ fn test_aggregation_on_json_object_mixed_types() {
     index_writer.commit().unwrap();
 
     // => Segment with mixed values
-    index_writer
-        .add_document(doc!(json => json!({"mixed_type": "red"})))
-        .unwrap();
     index_writer
         .add_document(doc!(json => json!({"mixed_type": "red"})))
         .unwrap();
@@ -885,8 +870,6 @@ fn test_aggregation_on_json_object_mixed_types() {
 
     let aggregation_results = searcher.search(&AllQuery, &aggregation_collector).unwrap();
     let aggregation_res_json = serde_json::to_value(aggregation_results).unwrap();
-    // pretty print as json
-    use pretty_assertions::assert_eq;
     assert_eq!(
         &aggregation_res_json,
         &serde_json::json!({
@@ -902,9 +885,9 @@ fn test_aggregation_on_json_object_mixed_types() {
             "buckets": [
               { "doc_count": 1, "key": 10.0, "min_price": { "value": 10.0 } },
               { "doc_count": 1, "key": -20.5, "min_price": { "value": -20.5 } },
-              { "doc_count": 2, "key": "red", "min_price": { "value": null } },
-              { "doc_count": 2, "key": 1.0, "key_as_string": "true", "min_price": { "value": null } },
-              { "doc_count": 3, "key": "blue", "min_price": { "value": null } },
+              // TODO bool is also not yet handled in aggregation
+              { "doc_count": 1, "key": "blue", "min_price": { "value": null } },
+              { "doc_count": 1, "key": "red", "min_price": { "value": null } },
             ],
             "sum_other_doc_count": 0
           }

src/aggregation/bucket/histogram/date_histogram.rs

@@ -1,7 +1,7 @@
 use serde::{Deserialize, Serialize};
 
 use super::{HistogramAggregation, HistogramBounds};
-use crate::aggregation::*;
+use crate::aggregation::AggregationError;
 
 /// DateHistogramAggregation is similar to `HistogramAggregation`, but it can only be used with date
 /// type.
@@ -307,7 +307,6 @@ pub mod tests {
     ) -> crate::Result<Index> {
         let mut schema_builder = Schema::builder();
         schema_builder.add_date_field("date", FAST);
-        schema_builder.add_json_field("mixed", FAST);
         schema_builder.add_text_field("text", FAST | STRING);
         schema_builder.add_text_field("text2", FAST | STRING);
         let schema = schema_builder.build();
@@ -352,10 +351,8 @@ pub mod tests {
         let docs = vec![
             vec![r#"{ "date": "2015-01-01T12:10:30Z", "text": "aaa" }"#],
             vec![r#"{ "date": "2015-01-01T11:11:30Z", "text": "bbb" }"#],
-            vec![r#"{ "date": "2015-01-01T11:11:30Z", "text": "bbb" }"#],
             vec![r#"{ "date": "2015-01-02T00:00:00Z", "text": "bbb" }"#],
             vec![r#"{ "date": "2015-01-06T00:00:00Z", "text": "ccc" }"#],
-            vec![r#"{ "date": "2015-01-06T00:00:00Z", "text": "ccc" }"#],
         ];
         let index = get_test_index_from_docs(merge_segments, &docs).unwrap();
 
@@ -384,7 +381,7 @@ pub mod tests {
                         {
                             "key_as_string" : "2015-01-01T00:00:00Z",
                             "key" : 1420070400000.0,
-                            "doc_count" : 6
+                            "doc_count" : 4
                         }
                     ]
                 }
@@ -422,15 +419,15 @@ pub mod tests {
                     {
                         "key_as_string" : "2015-01-01T00:00:00Z",
                         "key" : 1420070400000.0,
-                        "doc_count" : 6,
+                        "doc_count" : 4,
                         "texts": {
                             "buckets": [
                                 {
-                                "doc_count": 3,
+                                "doc_count": 2,
                                 "key": "bbb"
                                 },
                                 {
-                                "doc_count": 2,
+                                "doc_count": 1,
                                 "key": "ccc"
                                 },
                                 {
@@ -469,7 +466,7 @@ pub mod tests {
                 "sales_over_time": {
                     "buckets": [
                         {
-                            "doc_count": 3,
+                            "doc_count": 2,
                             "key": 1420070400000.0,
                             "key_as_string": "2015-01-01T00:00:00Z"
                         },
@@ -494,7 +491,7 @@ pub mod tests {
                             "key_as_string": "2015-01-05T00:00:00Z"
                         },
                         {
-                            "doc_count": 2,
+                            "doc_count": 1,
                             "key": 1420502400000.0,
                             "key_as_string": "2015-01-06T00:00:00Z"
                         }
@@ -535,7 +532,7 @@ pub mod tests {
                             "key_as_string": "2014-12-31T00:00:00Z"
                         },
                         {
-                            "doc_count": 3,
+                            "doc_count": 2,
                             "key": 1420070400000.0,
                             "key_as_string": "2015-01-01T00:00:00Z"
                         },
@@ -560,7 +557,7 @@ pub mod tests {
                             "key_as_string": "2015-01-05T00:00:00Z"
                         },
                         {
-                            "doc_count": 2,
+                            "doc_count": 1,
                             "key": 1420502400000.0,
                             "key_as_string": "2015-01-06T00:00:00Z"
                         },

src/aggregation/bucket/histogram/histogram.rs

@@ -20,7 +20,7 @@ use crate::aggregation::intermediate_agg_result::{
 use crate::aggregation::segment_agg_result::{
     build_segment_agg_collector, AggregationLimits, SegmentAggregationCollector,
 };
-use crate::aggregation::*;
+use crate::aggregation::{f64_from_fastfield_u64, format_date};
 use crate::TantivyError;
 
 /// Histogram is a bucket aggregation, where buckets are created dynamically for given `interval`.
@@ -73,7 +73,6 @@ pub struct HistogramAggregation {
     pub field: String,
     /// The interval to chunk your data range. Each bucket spans a value range of [0..interval).
     /// Must be a positive value.
-    #[serde(deserialize_with = "deserialize_f64")]
     pub interval: f64,
     /// Intervals implicitly defines an absolute grid of buckets `[interval * k, interval * (k +
     /// 1))`.
@@ -86,7 +85,6 @@ pub struct HistogramAggregation {
     /// fall into the buckets with the key 0 and 10.
     /// With offset 5 and interval 10, they would both fall into the bucket with they key 5 and the
     /// range [5..15)
-    #[serde(default, deserialize_with = "deserialize_option_f64")]
     pub offset: Option<f64>,
     /// The minimum number of documents in a bucket to be returned. Defaults to 0.
     pub min_doc_count: Option<u64>,
@@ -598,13 +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<()> {
@@ -1356,35 +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/aggregation/bucket/range.rs

@@ -14,7 +14,9 @@ use crate::aggregation::intermediate_agg_result::{
 use crate::aggregation::segment_agg_result::{
     build_segment_agg_collector, SegmentAggregationCollector,
 };
-use crate::aggregation::*;
+use crate::aggregation::{
+    f64_from_fastfield_u64, f64_to_fastfield_u64, format_date, Key, SerializedKey,
+};
 use crate::TantivyError;
 
 /// Provide user-defined buckets to aggregate on.
@@ -70,19 +72,11 @@ pub struct RangeAggregationRange {
     pub key: Option<String>,
     /// The from range value, which is inclusive in the range.
     /// `None` equals to an open ended interval.
-    #[serde(
-        skip_serializing_if = "Option::is_none",
-        default,
-        deserialize_with = "deserialize_option_f64"
-    )]
+    #[serde(skip_serializing_if = "Option::is_none", default)]
     pub from: Option<f64>,
     /// The to range value, which is not inclusive in the range.
     /// `None` equals to an open ended interval.
-    #[serde(
-        skip_serializing_if = "Option::is_none",
-        default,
-        deserialize_with = "deserialize_option_f64"
-    )]
+    #[serde(skip_serializing_if = "Option::is_none", default)]
     pub to: Option<f64>,
 }
 

src/aggregation/bucket/term_agg.rs

@@ -99,15 +99,24 @@ pub struct TermsAggregation {
     #[serde(skip_serializing_if = "Option::is_none", default)]
     pub size: Option<u32>,
 
-    /// To get more accurate results, we fetch more than `size` from each segment.
+    /// Unused by tantivy.
+    ///
+    /// Since tantivy doesn't know shards, this parameter is merely there to be used by consumers
+    /// of tantivy. shard_size is the number of terms returned by each shard.
+    /// The default value in elasticsearch is size * 1.5 + 10.
+    ///
+    /// Should never be smaller than size.
+    #[serde(skip_serializing_if = "Option::is_none", default)]
+    #[serde(alias = "shard_size")]
+    pub split_size: Option<u32>,
+
+    /// The get more accurate results, we fetch more than `size` from each segment.
     ///
     /// Increasing this value is will increase the cost for more accuracy.
     ///
     /// Defaults to 10 * size.
     #[serde(skip_serializing_if = "Option::is_none", default)]
-    #[serde(alias = "segment_size")]
-    #[serde(alias = "split_size")]
-    pub shard_size: Option<u32>,
+    pub segment_size: Option<u32>,
 
     /// If you set the `show_term_doc_count_error` parameter to true, the terms aggregation will
     /// include doc_count_error_upper_bound, which is an upper bound to the error on the
@@ -196,7 +205,7 @@ impl TermsAggregationInternal {
     pub(crate) fn from_req(req: &TermsAggregation) -> Self {
         let size = req.size.unwrap_or(10);
 
-        let mut segment_size = req.shard_size.unwrap_or(size * 10);
+        let mut segment_size = req.segment_size.unwrap_or(size * 10);
 
         let order = req.order.clone().unwrap_or_default();
         segment_size = segment_size.max(size);
@@ -247,7 +256,7 @@ pub struct SegmentTermCollector {
     term_buckets: TermBuckets,
     req: TermsAggregationInternal,
     blueprint: Option<Box<dyn SegmentAggregationCollector>>,
-    column_type: ColumnType,
+    field_type: ColumnType,
     accessor_idx: usize,
 }
 
@@ -346,7 +355,7 @@ impl SegmentTermCollector {
         field_type: ColumnType,
         accessor_idx: usize,
     ) -> crate::Result<Self> {
-        if field_type == ColumnType::Bytes {
+        if field_type == ColumnType::Bytes || field_type == ColumnType::Bool {
             return Err(TantivyError::InvalidArgument(format!(
                 "terms aggregation is not supported for column type {:?}",
                 field_type
@@ -380,7 +389,7 @@ impl SegmentTermCollector {
             req: TermsAggregationInternal::from_req(req),
             term_buckets,
             blueprint,
-            column_type: field_type,
+            field_type,
             accessor_idx,
         })
     }
@@ -457,7 +466,7 @@ impl SegmentTermCollector {
                 Ok(intermediate_entry)
             };
 
-        if self.column_type == ColumnType::Str {
+        if self.field_type == ColumnType::Str {
             let term_dict = agg_with_accessor
                 .str_dict_column
                 .as_ref()
@@ -522,34 +531,28 @@ impl SegmentTermCollector {
                         });
                 }
             }
-        } else if self.column_type == ColumnType::DateTime {
+        } else if self.field_type == ColumnType::DateTime {
             for (val, doc_count) in entries {
                 let intermediate_entry = into_intermediate_bucket_entry(val, doc_count)?;
                 let val = i64::from_u64(val);
                 let date = format_date(val)?;
                 dict.insert(IntermediateKey::Str(date), intermediate_entry);
             }
-        } else if self.column_type == ColumnType::Bool {
-            for (val, doc_count) in entries {
-                let intermediate_entry = into_intermediate_bucket_entry(val, doc_count)?;
-                let val = bool::from_u64(val);
-                dict.insert(IntermediateKey::Bool(val), intermediate_entry);
-            }
         } else {
             for (val, doc_count) in entries {
                 let intermediate_entry = into_intermediate_bucket_entry(val, doc_count)?;
-                let val = f64_from_fastfield_u64(val, &self.column_type);
+                let val = f64_from_fastfield_u64(val, &self.field_type);
                 dict.insert(IntermediateKey::F64(val), intermediate_entry);
             }
         };
 
-        Ok(IntermediateBucketResult::Terms {
-            buckets: IntermediateTermBucketResult {
+        Ok(IntermediateBucketResult::Terms(
+            IntermediateTermBucketResult {
                 entries: dict,
                 sum_other_doc_count,
                 doc_count_error_upper_bound: term_doc_count_before_cutoff,
             },
-        })
+        ))
     }
 }
 
@@ -1362,7 +1365,7 @@ mod tests {
 
     #[test]
     fn terms_aggregation_different_tokenizer_on_ff_test() -> crate::Result<()> {
-        let terms = vec!["Hello Hello", "Hallo Hallo", "Hallo Hallo"];
+        let terms = vec!["Hello Hello", "Hallo Hallo"];
 
         let index = get_test_index_from_terms(true, &[terms])?;
 
@@ -1380,7 +1383,7 @@ mod tests {
         println!("{}", serde_json::to_string_pretty(&res).unwrap());
 
         assert_eq!(res["my_texts"]["buckets"][0]["key"], "Hallo Hallo");
-        assert_eq!(res["my_texts"]["buckets"][0]["doc_count"], 2);
+        assert_eq!(res["my_texts"]["buckets"][0]["doc_count"], 1);
 
         assert_eq!(res["my_texts"]["buckets"][1]["key"], "Hello Hello");
         assert_eq!(res["my_texts"]["buckets"][1]["doc_count"], 1);
@@ -1891,40 +1894,4 @@ mod tests {
 
         Ok(())
     }
-
-    #[test]
-    fn terms_aggregation_bool() -> crate::Result<()> {
-        let mut schema_builder = Schema::builder();
-        let field = schema_builder.add_bool_field("bool_field", FAST);
-        let schema = schema_builder.build();
-        let index = Index::create_in_ram(schema);
-        {
-            let mut writer = index.writer_with_num_threads(1, 15_000_000)?;
-            writer.add_document(doc!(field=>true))?;
-            writer.add_document(doc!(field=>false))?;
-            writer.add_document(doc!(field=>true))?;
-            writer.commit()?;
-        }
-
-        let agg_req: Aggregations = serde_json::from_value(json!({
-            "my_bool": {
-                "terms": {
-                    "field": "bool_field"
-                },
-            }
-        }))
-        .unwrap();
-
-        let res = exec_request_with_query(agg_req, &index, None)?;
-
-        assert_eq!(res["my_bool"]["buckets"][0]["key"], 1.0);
-        assert_eq!(res["my_bool"]["buckets"][0]["key_as_string"], "true");
-        assert_eq!(res["my_bool"]["buckets"][0]["doc_count"], 2);
-        assert_eq!(res["my_bool"]["buckets"][1]["key"], 0.0);
-        assert_eq!(res["my_bool"]["buckets"][1]["key_as_string"], "false");
-        assert_eq!(res["my_bool"]["buckets"][1]["doc_count"], 1);
-        assert_eq!(res["my_bool"]["buckets"][2]["key"], serde_json::Value::Null);
-
-        Ok(())
-    }
 }

src/aggregation/bucket/term_missing_agg.rs

@@ -73,13 +73,11 @@ impl SegmentAggregationCollector for TermMissingAgg {
 
         entries.insert(missing.into(), missing_entry);
 
-        let bucket = IntermediateBucketResult::Terms {
-            buckets: IntermediateTermBucketResult {
-                entries,
-                sum_other_doc_count: 0,
-                doc_count_error_upper_bound: 0,
-            },
-        };
+        let bucket = IntermediateBucketResult::Terms(IntermediateTermBucketResult {
+            entries,
+            sum_other_doc_count: 0,
+            doc_count_error_upper_bound: 0,
+        });
 
         results.push(name, IntermediateAggregationResult::Bucket(bucket))?;
 
@@ -92,10 +90,7 @@ impl SegmentAggregationCollector for TermMissingAgg {
         agg_with_accessor: &mut AggregationsWithAccessor,
     ) -> crate::Result<()> {
         let agg = &mut agg_with_accessor.aggs.values[self.accessor_idx];
-        let has_value = agg
-            .accessors
-            .iter()
-            .any(|(acc, _)| acc.index.has_value(doc));
+        let has_value = agg.accessors.iter().any(|acc| acc.index.has_value(doc));
         if !has_value {
             self.missing_count += 1;
             if let Some(sub_agg) = self.sub_agg.as_mut() {

src/aggregation/collector.rs

@@ -8,7 +8,7 @@ use super::segment_agg_result::{
 };
 use crate::aggregation::agg_req_with_accessor::get_aggs_with_segment_accessor_and_validate;
 use crate::collector::{Collector, SegmentCollector};
-use crate::{DocId, SegmentOrdinal, SegmentReader, TantivyError};
+use crate::{DocId, SegmentReader, TantivyError};
 
 /// The default max bucket count, before the aggregation fails.
 pub const DEFAULT_BUCKET_LIMIT: u32 = 65000;
@@ -64,15 +64,10 @@ impl Collector for DistributedAggregationCollector {
 
     fn for_segment(
         &self,
-        segment_local_id: crate::SegmentOrdinal,
+        _segment_local_id: crate::SegmentOrdinal,
         reader: &crate::SegmentReader,
     ) -> crate::Result<Self::Child> {
-        AggregationSegmentCollector::from_agg_req_and_reader(
-            &self.agg,
-            reader,
-            segment_local_id,
-            &self.limits,
-        )
+        AggregationSegmentCollector::from_agg_req_and_reader(&self.agg, reader, &self.limits)
     }
 
     fn requires_scoring(&self) -> bool {
@@ -94,15 +89,10 @@ impl Collector for AggregationCollector {
 
     fn for_segment(
         &self,
-        segment_local_id: crate::SegmentOrdinal,
+        _segment_local_id: crate::SegmentOrdinal,
         reader: &crate::SegmentReader,
     ) -> crate::Result<Self::Child> {
-        AggregationSegmentCollector::from_agg_req_and_reader(
-            &self.agg,
-            reader,
-            segment_local_id,
-            &self.limits,
-        )
+        AggregationSegmentCollector::from_agg_req_and_reader(&self.agg, reader, &self.limits)
     }
 
     fn requires_scoring(&self) -> bool {
@@ -145,11 +135,10 @@ impl AggregationSegmentCollector {
     pub fn from_agg_req_and_reader(
         agg: &Aggregations,
         reader: &SegmentReader,
-        segment_ordinal: SegmentOrdinal,
         limits: &AggregationLimits,
     ) -> crate::Result<Self> {
         let mut aggs_with_accessor =
-            get_aggs_with_segment_accessor_and_validate(agg, reader, segment_ordinal, limits)?;
+            get_aggs_with_segment_accessor_and_validate(agg, reader, limits)?;
         let result =
             BufAggregationCollector::new(build_segment_agg_collector(&mut aggs_with_accessor)?);
         Ok(AggregationSegmentCollector {

src/aggregation/intermediate_agg_result.rs

@@ -19,7 +19,7 @@ use super::bucket::{
 };
 use super::metric::{
     IntermediateAverage, IntermediateCount, IntermediateMax, IntermediateMin, IntermediateStats,
-    IntermediateSum, PercentilesCollector, TopHitsCollector,
+    IntermediateSum, PercentilesCollector,
 };
 use super::segment_agg_result::AggregationLimits;
 use super::{format_date, AggregationError, Key, SerializedKey};
@@ -41,8 +41,6 @@ pub struct IntermediateAggregationResults {
 /// This might seem redundant with `Key`, but the point is to have a different
 /// Serialize implementation.
 pub enum IntermediateKey {
-    /// Bool key
-    Bool(bool),
     /// String key
     Str(String),
     /// `f64` key
@@ -61,7 +59,6 @@ impl From<IntermediateKey> for Key {
         match value {
             IntermediateKey::Str(s) => Self::Str(s),
             IntermediateKey::F64(f) => Self::F64(f),
-            IntermediateKey::Bool(f) => Self::F64(f as u64 as f64),
         }
     }
 }
@@ -74,7 +71,6 @@ impl std::hash::Hash for IntermediateKey {
         match self {
             IntermediateKey::Str(text) => text.hash(state),
             IntermediateKey::F64(val) => val.to_bits().hash(state),
-            IntermediateKey::Bool(val) => val.hash(state),
         }
     }
 }
@@ -170,9 +166,9 @@ impl IntermediateAggregationResults {
 pub(crate) fn empty_from_req(req: &Aggregation) -> IntermediateAggregationResult {
     use AggregationVariants::*;
     match req.agg {
-        Terms(_) => IntermediateAggregationResult::Bucket(IntermediateBucketResult::Terms {
-            buckets: Default::default(),
-        }),
+        Terms(_) => IntermediateAggregationResult::Bucket(IntermediateBucketResult::Terms(
+            Default::default(),
+        )),
         Range(_) => IntermediateAggregationResult::Bucket(IntermediateBucketResult::Range(
             Default::default(),
         )),
@@ -209,9 +205,6 @@ pub(crate) fn empty_from_req(req: &Aggregation) -> IntermediateAggregationResult
         Percentiles(_) => IntermediateAggregationResult::Metric(
             IntermediateMetricResult::Percentiles(PercentilesCollector::default()),
         ),
-        TopHits(_) => IntermediateAggregationResult::Metric(IntermediateMetricResult::TopHits(
-            TopHitsCollector::default(),
-        )),
     }
 }
 
@@ -272,8 +265,6 @@ pub enum IntermediateMetricResult {
     Stats(IntermediateStats),
     /// Intermediate sum result.
     Sum(IntermediateSum),
-    /// Intermediate top_hits result
-    TopHits(TopHitsCollector),
 }
 
 impl IntermediateMetricResult {
@@ -301,13 +292,9 @@ impl IntermediateMetricResult {
                 percentiles
                     .into_final_result(req.agg.as_percentile().expect("unexpected metric type")),
             ),
-            IntermediateMetricResult::TopHits(top_hits) => {
-                MetricResult::TopHits(top_hits.finalize())
-            }
         }
     }
 
-    // TODO: this is our top-of-the-chain fruit merge mech
     fn merge_fruits(&mut self, other: IntermediateMetricResult) -> crate::Result<()> {
         match (self, other) {
             (
@@ -343,9 +330,6 @@ impl IntermediateMetricResult {
             ) => {
                 left.merge_fruits(right)?;
             }
-            (IntermediateMetricResult::TopHits(left), IntermediateMetricResult::TopHits(right)) => {
-                left.merge_fruits(right)?;
-            }
             _ => {
                 panic!("incompatible fruit types in tree or missing merge_fruits handler");
             }
@@ -367,14 +351,11 @@ pub enum IntermediateBucketResult {
     Histogram {
         /// The column_type of the underlying `Column` is DateTime
         is_date_agg: bool,
-        /// The histogram buckets
+        /// The buckets
         buckets: Vec<IntermediateHistogramBucketEntry>,
     },
     /// Term aggregation
-    Terms {
-        /// The term buckets
-        buckets: IntermediateTermBucketResult,
-    },
+    Terms(IntermediateTermBucketResult),
 }
 
 impl IntermediateBucketResult {
@@ -451,7 +432,7 @@ impl IntermediateBucketResult {
                 };
                 Ok(BucketResult::Histogram { buckets })
             }
-            IntermediateBucketResult::Terms { buckets: terms } => terms.into_final_result(
+            IntermediateBucketResult::Terms(terms) => terms.into_final_result(
                 req.agg
                     .as_term()
                     .expect("unexpected aggregation, expected term aggregation"),
@@ -464,12 +445,8 @@ impl IntermediateBucketResult {
     fn merge_fruits(&mut self, other: IntermediateBucketResult) -> crate::Result<()> {
         match (self, other) {
             (
-                IntermediateBucketResult::Terms {
-                    buckets: term_res_left,
-                },
-                IntermediateBucketResult::Terms {
-                    buckets: term_res_right,
-                },
+                IntermediateBucketResult::Terms(term_res_left),
+                IntermediateBucketResult::Terms(term_res_right),
             ) => {
                 merge_maps(&mut term_res_left.entries, term_res_right.entries)?;
                 term_res_left.sum_other_doc_count += term_res_right.sum_other_doc_count;
@@ -553,15 +530,8 @@ impl IntermediateTermBucketResult {
             .into_iter()
             .filter(|bucket| bucket.1.doc_count as u64 >= req.min_doc_count)
             .map(|(key, entry)| {
-                let key_as_string = match key {
-                    IntermediateKey::Bool(key) => {
-                        let val = if key { "true" } else { "false" };
-                        Some(val.to_string())
-                    }
-                    _ => None,
-                };
                 Ok(BucketEntry {
-                    key_as_string,
+                    key_as_string: None,
                     key: key.into(),
                     doc_count: entry.doc_count as u64,
                     sub_aggregation: entry

src/aggregation/metric/average.rs

@@ -2,8 +2,7 @@ use std::fmt::Debug;
 
 use serde::{Deserialize, Serialize};
 
-use super::*;
-use crate::aggregation::*;
+use super::{IntermediateStats, SegmentStatsCollector};
 
 /// A single-value metric aggregation that computes the average of numeric values that are
 /// extracted from the aggregated documents.
@@ -25,7 +24,7 @@ pub struct AverageAggregation {
     /// By default they will be ignored but it is also possible to treat them as if they had a
     /// value. Examples in JSON format:
     /// { "field": "my_numbers", "missing": "10.0" }
-    #[serde(default, deserialize_with = "deserialize_option_f64")]
+    #[serde(default)]
     pub missing: Option<f64>,
 }
 
@@ -66,71 +65,3 @@ impl IntermediateAverage {
         self.stats.finalize().avg
     }
 }
-
-#[cfg(test)]
-mod tests {
-    use super::*;
-
-    #[test]
-    fn deserialization_with_missing_test1() {
-        let json = r#"{
-            "field": "score",
-            "missing": "10.0"
-        }"#;
-        let avg: AverageAggregation = serde_json::from_str(json).unwrap();
-        assert_eq!(avg.field, "score");
-        assert_eq!(avg.missing, Some(10.0));
-        // no dot
-        let json = r#"{
-            "field": "score",
-            "missing": "10"
-        }"#;
-        let avg: AverageAggregation = serde_json::from_str(json).unwrap();
-        assert_eq!(avg.field, "score");
-        assert_eq!(avg.missing, Some(10.0));
-
-        // from value
-        let avg: AverageAggregation = serde_json::from_value(json!({
-            "field": "score_f64",
-            "missing": 10u64,
-        }))
-        .unwrap();
-        assert_eq!(avg.missing, Some(10.0));
-        // from value
-        let avg: AverageAggregation = serde_json::from_value(json!({
-            "field": "score_f64",
-            "missing": 10u32,
-        }))
-        .unwrap();
-        assert_eq!(avg.missing, Some(10.0));
-        let avg: AverageAggregation = serde_json::from_value(json!({
-            "field": "score_f64",
-            "missing": 10i8,
-        }))
-        .unwrap();
-        assert_eq!(avg.missing, Some(10.0));
-    }
-
-    #[test]
-    fn deserialization_with_missing_test_fail() {
-        let json = r#"{
-            "field": "score",
-            "missing": "a"
-        }"#;
-        let avg: Result<AverageAggregation, _> = serde_json::from_str(json);
-        assert!(avg.is_err());
-        assert!(avg
-            .unwrap_err()
-            .to_string()
-            .contains("Failed to parse f64 from string: \"a\""));
-
-        // Disallow NaN
-        let json = r#"{
-            "field": "score",
-            "missing": "NaN"
-        }"#;
-        let avg: Result<AverageAggregation, _> = serde_json::from_str(json);
-        assert!(avg.is_err());
-        assert!(avg.unwrap_err().to_string().contains("NaN"));
-    }
-}

src/aggregation/metric/count.rs

@@ -2,8 +2,7 @@ use std::fmt::Debug;
 
 use serde::{Deserialize, Serialize};
 
-use super::*;
-use crate::aggregation::*;
+use super::{IntermediateStats, SegmentStatsCollector};
 
 /// A single-value metric aggregation that counts the number of values that are
 /// extracted from the aggregated documents.
@@ -25,7 +24,7 @@ pub struct CountAggregation {
     /// By default they will be ignored but it is also possible to treat them as if they had a
     /// value. Examples in JSON format:
     /// { "field": "my_numbers", "missing": "10.0" }
-    #[serde(default, deserialize_with = "deserialize_option_f64")]
+    #[serde(default)]
     pub missing: Option<f64>,
 }
 

src/aggregation/metric/max.rs

@@ -2,8 +2,7 @@ use std::fmt::Debug;
 
 use serde::{Deserialize, Serialize};
 
-use super::*;
-use crate::aggregation::*;
+use super::{IntermediateStats, SegmentStatsCollector};
 
 /// A single-value metric aggregation that computes the maximum of numeric values that are
 /// extracted from the aggregated documents.
@@ -25,7 +24,7 @@ pub struct MaxAggregation {
     /// By default they will be ignored but it is also possible to treat them as if they had a
     /// value. Examples in JSON format:
     /// { "field": "my_numbers", "missing": "10.0" }
-    #[serde(default, deserialize_with = "deserialize_option_f64")]
+    #[serde(default)]
     pub missing: Option<f64>,
 }
 

src/aggregation/metric/min.rs

@@ -2,8 +2,7 @@ use std::fmt::Debug;
 
 use serde::{Deserialize, Serialize};
 
-use super::*;
-use crate::aggregation::*;
+use super::{IntermediateStats, SegmentStatsCollector};
 
 /// A single-value metric aggregation that computes the minimum of numeric values that are
 /// extracted from the aggregated documents.
@@ -25,7 +24,7 @@ pub struct MinAggregation {
     /// By default they will be ignored but it is also possible to treat them as if they had a
     /// value. Examples in JSON format:
     /// { "field": "my_numbers", "missing": "10.0" }
-    #[serde(default, deserialize_with = "deserialize_option_f64")]
+    #[serde(default)]
     pub missing: Option<f64>,
 }
 

src/aggregation/metric/mod.rs

@@ -23,8 +23,6 @@ mod min;
 mod percentiles;
 mod stats;
 mod sum;
-mod top_hits;
-
 pub use average::*;
 pub use count::*;
 pub use max::*;
@@ -34,7 +32,6 @@ use rustc_hash::FxHashMap;
 use serde::{Deserialize, Serialize};
 pub use stats::*;
 pub use sum::*;
-pub use top_hits::*;
 
 /// Single-metric aggregations use this common result structure.
 ///
@@ -84,27 +81,6 @@ pub struct PercentilesMetricResult {
     pub values: PercentileValues,
 }
 
-/// The top_hits metric results entry
-#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
-pub struct TopHitsVecEntry {
-    /// The sort values of the document, depending on the sort criteria in the request.
-    pub sort: Vec<Option<u64>>,
-
-    /// Search results, for queries that include field retrieval requests
-    /// (`docvalue_fields`).
-    #[serde(flatten)]
-    pub search_results: FieldRetrivalResult,
-}
-
-/// The top_hits metric aggregation results a list of top hits by sort criteria.
-///
-/// The main reason for wrapping it in `hits` is to match elasticsearch output structure.
-#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
-pub struct TopHitsMetricResult {
-    /// The result of the top_hits metric.
-    pub hits: Vec<TopHitsVecEntry>,
-}
-
 #[cfg(test)]
 mod tests {
     use crate::aggregation::agg_req::Aggregations;

src/aggregation/metric/percentiles.rs

@@ -11,7 +11,7 @@ use crate::aggregation::intermediate_agg_result::{
     IntermediateAggregationResult, IntermediateAggregationResults, IntermediateMetricResult,
 };
 use crate::aggregation::segment_agg_result::SegmentAggregationCollector;
-use crate::aggregation::*;
+use crate::aggregation::{f64_from_fastfield_u64, f64_to_fastfield_u64, AggregationError};
 use crate::{DocId, TantivyError};
 
 /// # Percentiles
@@ -84,11 +84,7 @@ pub struct PercentilesAggregationReq {
     /// By default they will be ignored but it is also possible to treat them as if they had a
     /// value. Examples in JSON format:
     /// { "field": "my_numbers", "missing": "10.0" }
-    #[serde(
-        skip_serializing_if = "Option::is_none",
-        default,
-        deserialize_with = "deserialize_option_f64"
-    )]
+    #[serde(skip_serializing_if = "Option::is_none", default)]
     pub missing: Option<f64>,
 }
 fn default_percentiles() -> &'static [f64] {
@@ -137,6 +133,7 @@ pub(crate) struct SegmentPercentilesCollector {
     field_type: ColumnType,
     pub(crate) percentiles: PercentilesCollector,
     pub(crate) accessor_idx: usize,
+    val_cache: Vec<u64>,
     missing: Option<u64>,
 }
 
@@ -246,6 +243,7 @@ impl SegmentPercentilesCollector {
             field_type,
             percentiles: PercentilesCollector::new(),
             accessor_idx,
+            val_cache: Default::default(),
             missing,
         })
     }

src/aggregation/metric/stats.rs

@@ -9,7 +9,7 @@ use crate::aggregation::intermediate_agg_result::{
     IntermediateAggregationResult, IntermediateAggregationResults, IntermediateMetricResult,
 };
 use crate::aggregation::segment_agg_result::SegmentAggregationCollector;
-use crate::aggregation::*;
+use crate::aggregation::{f64_from_fastfield_u64, f64_to_fastfield_u64};
 use crate::{DocId, TantivyError};
 
 /// A multi-value metric aggregation that computes a collection of statistics on numeric values that
@@ -33,7 +33,7 @@ pub struct StatsAggregation {
     /// By default they will be ignored but it is also possible to treat them as if they had a
     /// value. Examples in JSON format:
     /// { "field": "my_numbers", "missing": "10.0" }
-    #[serde(default, deserialize_with = "deserialize_option_f64")]
+    #[serde(default)]
     pub missing: Option<f64>,
 }
 
@@ -580,30 +580,6 @@ mod tests {
             })
         );
 
-        // From string
-        let agg_req: Aggregations = serde_json::from_value(json!({
-            "my_stats": {
-                "stats": {
-                    "field": "json.partially_empty",
-                    "missing": "0.0"
-                },
-            }
-        }))
-        .unwrap();
-
-        let res = exec_request_with_query(agg_req, &index, None)?;
-
-        assert_eq!(
-            res["my_stats"],
-            json!({
-                "avg":  2.5,
-                "count": 4,
-                "max": 10.0,
-                "min": 0.0,
-                "sum": 10.0
-            })
-        );
-
         Ok(())
     }
 

src/aggregation/metric/sum.rs

@@ -2,8 +2,7 @@ use std::fmt::Debug;
 
 use serde::{Deserialize, Serialize};
 
-use super::*;
-use crate::aggregation::*;
+use super::{IntermediateStats, SegmentStatsCollector};
 
 /// A single-value metric aggregation that sums up numeric values that are
 /// extracted from the aggregated documents.
@@ -25,7 +24,7 @@ pub struct SumAggregation {
     /// By default they will be ignored but it is also possible to treat them as if they had a
     /// value. Examples in JSON format:
     /// { "field": "my_numbers", "missing": "10.0" }
-    #[serde(default, deserialize_with = "deserialize_option_f64")]
+    #[serde(default)]
     pub missing: Option<f64>,
 }
 

src/aggregation/metric/top_hits.rs

@@ -1,837 +0,0 @@
-use std::collections::HashMap;
-use std::fmt::Formatter;
-
-use columnar::{ColumnarReader, DynamicColumn};
-use regex::Regex;
-use serde::ser::SerializeMap;
-use serde::{Deserialize, Deserializer, Serialize, Serializer};
-
-use super::{TopHitsMetricResult, TopHitsVecEntry};
-use crate::aggregation::bucket::Order;
-use crate::aggregation::intermediate_agg_result::{
-    IntermediateAggregationResult, IntermediateMetricResult,
-};
-use crate::aggregation::segment_agg_result::SegmentAggregationCollector;
-use crate::collector::TopNComputer;
-use crate::schema::term::JSON_PATH_SEGMENT_SEP_STR;
-use crate::schema::OwnedValue;
-use crate::{DocAddress, DocId, SegmentOrdinal};
-
-/// # Top Hits
-///
-/// The top hits aggregation is a useful tool to answer questions like:
-/// - "What are the most recent posts by each author?"
-/// - "What are the most popular items in each category?"
-///
-/// It does so by keeping track of the most relevant document being aggregated,
-/// in terms of a sort criterion that can consist of multiple fields and their
-/// sort-orders (ascending or descending).
-///
-/// `top_hits` should not be used as a top-level aggregation. It is intended to be
-/// used as a sub-aggregation, inside a `terms` aggregation or a `filters` aggregation,
-/// for example.
-///
-/// Note that this aggregator does not return the actual document addresses, but
-/// rather a list of the values of the fields that were requested to be retrieved.
-/// These values can be specified in the `docvalue_fields` parameter, which can include
-/// a list of fast fields to be retrieved. At the moment, only fast fields are supported
-/// but it is possible that we support the `fields` parameter to retrieve any stored
-/// field in the future.
-///
-/// The following example demonstrates a request for the top_hits aggregation:
-/// ```JSON
-/// {
-///     "aggs": {
-///         "top_authors": {
-///             "terms": {
-///                 "field": "author",
-///                 "size": 5
-///             }
-///         },
-///         "aggs": {
-///             "top_hits": {
-///                 "size": 2,
-///                 "from": 0
-///                 "sort": [
-///                     { "date": "desc" }
-///                 ]
-///                 "docvalue_fields": ["date", "title", "iden"]
-///             }
-///         }
-/// }
-/// ```
-///
-/// This request will return an object containing the top two documents, sorted
-/// by the `date` field in descending order. You can also sort by multiple fields, which
-/// helps to resolve ties. The aggregation object for each bucket will look like:
-/// ```JSON
-/// {
-///     "hits": [
-///         {
-///           "score": [<time_u64>],
-///           "docvalue_fields": {
-///             "date": "<date_RFC3339>",
-///             "title": "<title>",
-///             "iden": "<iden>"
-///           }
-///         },
-///         {
-///           "score": [<time_u64>]
-///           "docvalue_fields": {
-///             "date": "<date_RFC3339>",
-///             "title": "<title>",
-///             "iden": "<iden>"
-///           }
-///         }
-///     ]
-/// }
-/// ```
-#[derive(Debug, Clone, PartialEq, Serialize, Deserialize, Default)]
-pub struct TopHitsAggregation {
-    sort: Vec<KeyOrder>,
-    size: usize,
-    from: Option<usize>,
-
-    #[serde(flatten)]
-    retrieval: RetrievalFields,
-}
-
-const fn default_doc_value_fields() -> Vec<String> {
-    Vec::new()
-}
-
-/// Search query spec for each matched document
-/// TODO: move this to a common module
-#[derive(Debug, Clone, PartialEq, Serialize, Deserialize, Default)]
-pub struct RetrievalFields {
-    /// The fast fields to return for each hit.
-    /// This is the only variant supported for now.
-    /// TODO: support the {field, format} variant for custom formatting.
-    #[serde(rename = "docvalue_fields")]
-    #[serde(default = "default_doc_value_fields")]
-    pub doc_value_fields: Vec<String>,
-}
-
-/// Search query result for each matched document
-/// TODO: move this to a common module
-#[derive(Debug, Clone, PartialEq, Serialize, Deserialize, Default)]
-pub struct FieldRetrivalResult {
-    /// The fast fields returned for each hit.
-    #[serde(rename = "docvalue_fields")]
-    #[serde(skip_serializing_if = "HashMap::is_empty")]
-    pub doc_value_fields: HashMap<String, OwnedValue>,
-}
-
-impl RetrievalFields {
-    fn get_field_names(&self) -> Vec<&str> {
-        self.doc_value_fields.iter().map(|s| s.as_str()).collect()
-    }
-
-    fn resolve_field_names(&mut self, reader: &ColumnarReader) -> crate::Result<()> {
-        // Tranform a glob (`pattern*`, for example) into a regex::Regex (`^pattern.*$`)
-        let globbed_string_to_regex = |glob: &str| {
-            // Replace `*` glob with `.*` regex
-            let sanitized = format!("^{}$", regex::escape(glob).replace(r"\*", ".*"));
-            Regex::new(&sanitized.replace('*', ".*")).map_err(|e| {
-                crate::TantivyError::SchemaError(format!(
-                    "Invalid regex '{}' in docvalue_fields: {}",
-                    glob, e
-                ))
-            })
-        };
-        self.doc_value_fields = self
-            .doc_value_fields
-            .iter()
-            .map(|field| {
-                if !field.contains('*')
-                    && reader
-                        .iter_columns()?
-                        .any(|(name, _)| name.as_str() == field)
-                {
-                    return Ok(vec![field.to_owned()]);
-                }
-
-                let pattern = globbed_string_to_regex(field)?;
-                let fields = reader
-                    .iter_columns()?
-                    .map(|(name, _)| {
-                        // normalize path from internal fast field repr
-                        name.replace(JSON_PATH_SEGMENT_SEP_STR, ".")
-                    })
-                    .filter(|name| pattern.is_match(name))
-                    .collect::<Vec<_>>();
-                assert!(
-                    !fields.is_empty(),
-                    "No fields matched the glob '{}' in docvalue_fields",
-                    field
-                );
-                Ok(fields)
-            })
-            .collect::<crate::Result<Vec<_>>>()?
-            .into_iter()
-            .flatten()
-            .collect();
-
-        Ok(())
-    }
-
-    fn get_document_field_data(
-        &self,
-        accessors: &HashMap<String, Vec<DynamicColumn>>,
-        doc_id: DocId,
-    ) -> FieldRetrivalResult {
-        let dvf = self
-            .doc_value_fields
-            .iter()
-            .map(|field| {
-                let accessors = accessors
-                    .get(field)
-                    .unwrap_or_else(|| panic!("field '{}' not found in accessors", field));
-
-                let values: Vec<OwnedValue> = accessors
-                    .iter()
-                    .flat_map(|accessor| match accessor {
-                        DynamicColumn::U64(accessor) => accessor
-                            .values_for_doc(doc_id)
-                            .map(OwnedValue::U64)
-                            .collect::<Vec<_>>(),
-                        DynamicColumn::I64(accessor) => accessor
-                            .values_for_doc(doc_id)
-                            .map(OwnedValue::I64)
-                            .collect::<Vec<_>>(),
-                        DynamicColumn::F64(accessor) => accessor
-                            .values_for_doc(doc_id)
-                            .map(OwnedValue::F64)
-                            .collect::<Vec<_>>(),
-                        DynamicColumn::Bytes(accessor) => accessor
-                            .term_ords(doc_id)
-                            .map(|term_ord| {
-                                let mut buffer = vec![];
-                                assert!(
-                                    accessor
-                                        .ord_to_bytes(term_ord, &mut buffer)
-                                        .expect("could not read term dictionary"),
-                                    "term corresponding to term_ord does not exist"
-                                );
-                                OwnedValue::Bytes(buffer)
-                            })
-                            .collect::<Vec<_>>(),
-                        DynamicColumn::Str(accessor) => accessor
-                            .term_ords(doc_id)
-                            .map(|term_ord| {
-                                let mut buffer = vec![];
-                                assert!(
-                                    accessor
-                                        .ord_to_bytes(term_ord, &mut buffer)
-                                        .expect("could not read term dictionary"),
-                                    "term corresponding to term_ord does not exist"
-                                );
-                                OwnedValue::Str(String::from_utf8(buffer).unwrap())
-                            })
-                            .collect::<Vec<_>>(),
-                        DynamicColumn::Bool(accessor) => accessor
-                            .values_for_doc(doc_id)
-                            .map(OwnedValue::Bool)
-                            .collect::<Vec<_>>(),
-                        DynamicColumn::IpAddr(accessor) => accessor
-                            .values_for_doc(doc_id)
-                            .map(OwnedValue::IpAddr)
-                            .collect::<Vec<_>>(),
-                        DynamicColumn::DateTime(accessor) => accessor
-                            .values_for_doc(doc_id)
-                            .map(OwnedValue::Date)
-                            .collect::<Vec<_>>(),
-                    })
-                    .collect();
-
-                (field.to_owned(), OwnedValue::Array(values))
-            })
-            .collect();
-        FieldRetrivalResult {
-            doc_value_fields: dvf,
-        }
-    }
-}
-
-#[derive(Debug, Clone, PartialEq, Default)]
-struct KeyOrder {
-    field: String,
-    order: Order,
-}
-
-impl Serialize for KeyOrder {
-    fn serialize<S: Serializer>(&self, serializer: S) -> Result<S::Ok, S::Error> {
-        let KeyOrder { field, order } = self;
-        let mut map = serializer.serialize_map(Some(1))?;
-        map.serialize_entry(field, order)?;
-        map.end()
-    }
-}
-
-impl<'de> Deserialize<'de> for KeyOrder {
-    fn deserialize<D>(deserializer: D) -> Result<Self, D::Error>
-    where D: Deserializer<'de> {
-        let mut k_o = <HashMap<String, Order>>::deserialize(deserializer)?.into_iter();
-        let (k, v) = k_o.next().ok_or(serde::de::Error::custom(
-            "Expected exactly one key-value pair in KeyOrder, found none",
-        ))?;
-        if k_o.next().is_some() {
-            return Err(serde::de::Error::custom(
-                "Expected exactly one key-value pair in KeyOrder, found more",
-            ));
-        }
-        Ok(Self { field: k, order: v })
-    }
-}
-
-impl TopHitsAggregation {
-    /// Validate and resolve field retrieval parameters
-    pub fn validate_and_resolve(&mut self, reader: &ColumnarReader) -> crate::Result<()> {
-        self.retrieval.resolve_field_names(reader)
-    }
-
-    /// Return fields accessed by the aggregator, in order.
-    pub fn field_names(&self) -> Vec<&str> {
-        self.sort
-            .iter()
-            .map(|KeyOrder { field, .. }| field.as_str())
-            .collect()
-    }
-
-    /// Return fields accessed by the aggregator's value retrieval.
-    pub fn value_field_names(&self) -> Vec<&str> {
-        self.retrieval.get_field_names()
-    }
-}
-
-/// Holds a single comparable doc feature, and the order in which it should be sorted.
-#[derive(Clone, Serialize, Deserialize, Debug)]
-struct ComparableDocFeature {
-    /// Stores any u64-mappable feature.
-    value: Option<u64>,
-    /// Sort order for the doc feature
-    order: Order,
-}
-
-impl Ord for ComparableDocFeature {
-    fn cmp(&self, other: &Self) -> std::cmp::Ordering {
-        let invert = |cmp: std::cmp::Ordering| match self.order {
-            Order::Asc => cmp,
-            Order::Desc => cmp.reverse(),
-        };
-
-        match (self.value, other.value) {
-            (Some(self_value), Some(other_value)) => invert(self_value.cmp(&other_value)),
-            (Some(_), None) => std::cmp::Ordering::Greater,
-            (None, Some(_)) => std::cmp::Ordering::Less,
-            (None, None) => std::cmp::Ordering::Equal,
-        }
-    }
-}
-
-impl PartialOrd for ComparableDocFeature {
-    fn partial_cmp(&self, other: &Self) -> Option<std::cmp::Ordering> {
-        Some(self.cmp(other))
-    }
-}
-
-impl PartialEq for ComparableDocFeature {
-    fn eq(&self, other: &Self) -> bool {
-        self.value.cmp(&other.value) == std::cmp::Ordering::Equal
-    }
-}
-
-impl Eq for ComparableDocFeature {}
-
-#[derive(Clone, Serialize, Deserialize, Debug)]
-struct ComparableDocFeatures(Vec<ComparableDocFeature>, FieldRetrivalResult);
-
-impl Ord for ComparableDocFeatures {
-    fn cmp(&self, other: &Self) -> std::cmp::Ordering {
-        for (self_feature, other_feature) in self.0.iter().zip(other.0.iter()) {
-            let cmp = self_feature.cmp(other_feature);
-            if cmp != std::cmp::Ordering::Equal {
-                return cmp;
-            }
-        }
-        std::cmp::Ordering::Equal
-    }
-}
-
-impl PartialOrd for ComparableDocFeatures {
-    fn partial_cmp(&self, other: &Self) -> Option<std::cmp::Ordering> {
-        Some(self.cmp(other))
-    }
-}
-
-impl PartialEq for ComparableDocFeatures {
-    fn eq(&self, other: &Self) -> bool {
-        self.cmp(other) == std::cmp::Ordering::Equal
-    }
-}
-
-impl Eq for ComparableDocFeatures {}
-
-/// The TopHitsCollector used for collecting over segments and merging results.
-#[derive(Clone, Serialize, Deserialize)]
-pub struct TopHitsCollector {
-    req: TopHitsAggregation,
-    top_n: TopNComputer<ComparableDocFeatures, DocAddress, false>,
-}
-
-impl Default for TopHitsCollector {
-    fn default() -> Self {
-        Self {
-            req: TopHitsAggregation::default(),
-            top_n: TopNComputer::new(1),
-        }
-    }
-}
-
-impl std::fmt::Debug for TopHitsCollector {
-    fn fmt(&self, f: &mut Formatter<'_>) -> std::fmt::Result {
-        f.debug_struct("TopHitsCollector")
-            .field("req", &self.req)
-            .field("top_n_threshold", &self.top_n.threshold)
-            .finish()
-    }
-}
-
-impl std::cmp::PartialEq for TopHitsCollector {
-    fn eq(&self, _other: &Self) -> bool {
-        false
-    }
-}
-
-impl TopHitsCollector {
-    fn collect(&mut self, features: ComparableDocFeatures, doc: DocAddress) {
-        self.top_n.push(features, doc);
-    }
-
-    pub(crate) fn merge_fruits(&mut self, other_fruit: Self) -> crate::Result<()> {
-        for doc in other_fruit.top_n.into_vec() {
-            self.collect(doc.feature, doc.doc);
-        }
-        Ok(())
-    }
-
-    /// Finalize by converting self into the final result form
-    pub fn finalize(self) -> TopHitsMetricResult {
-        let mut hits: Vec<TopHitsVecEntry> = self
-            .top_n
-            .into_sorted_vec()
-            .into_iter()
-            .map(|doc| TopHitsVecEntry {
-                sort: doc.feature.0.iter().map(|f| f.value).collect(),
-                search_results: doc.feature.1,
-            })
-            .collect();
-
-        // Remove the first `from` elements
-        // Truncating from end would be more efficient, but we need to truncate from the front
-        // because `into_sorted_vec` gives us a descending order because of the inverted
-        // `Ord` semantics of the heap elements.
-        hits.drain(..self.req.from.unwrap_or(0));
-        TopHitsMetricResult { hits }
-    }
-}
-
-#[derive(Clone)]
-pub(crate) struct SegmentTopHitsCollector {
-    segment_ordinal: SegmentOrdinal,
-    accessor_idx: usize,
-    inner_collector: TopHitsCollector,
-}
-
-impl SegmentTopHitsCollector {
-    pub fn from_req(
-        req: &TopHitsAggregation,
-        accessor_idx: usize,
-        segment_ordinal: SegmentOrdinal,
-    ) -> Self {
-        Self {
-            inner_collector: TopHitsCollector {
-                req: req.clone(),
-                top_n: TopNComputer::new(req.size + req.from.unwrap_or(0)),
-            },
-            segment_ordinal,
-            accessor_idx,
-        }
-    }
-}
-
-impl std::fmt::Debug for SegmentTopHitsCollector {
-    fn fmt(&self, f: &mut Formatter<'_>) -> std::fmt::Result {
-        f.debug_struct("SegmentTopHitsCollector")
-            .field("segment_id", &self.segment_ordinal)
-            .field("accessor_idx", &self.accessor_idx)
-            .field("inner_collector", &self.inner_collector)
-            .finish()
-    }
-}
-
-impl SegmentAggregationCollector for SegmentTopHitsCollector {
-    fn add_intermediate_aggregation_result(
-        self: Box<Self>,
-        agg_with_accessor: &crate::aggregation::agg_req_with_accessor::AggregationsWithAccessor,
-        results: &mut crate::aggregation::intermediate_agg_result::IntermediateAggregationResults,
-    ) -> crate::Result<()> {
-        let name = agg_with_accessor.aggs.keys[self.accessor_idx].to_string();
-        let intermediate_result = IntermediateMetricResult::TopHits(self.inner_collector);
-        results.push(
-            name,
-            IntermediateAggregationResult::Metric(intermediate_result),
-        )
-    }
-
-    fn collect(
-        &mut self,
-        doc_id: crate::DocId,
-        agg_with_accessor: &mut crate::aggregation::agg_req_with_accessor::AggregationsWithAccessor,
-    ) -> crate::Result<()> {
-        let accessors = &agg_with_accessor.aggs.values[self.accessor_idx].accessors;
-        let value_accessors = &agg_with_accessor.aggs.values[self.accessor_idx].value_accessors;
-        let features: Vec<ComparableDocFeature> = self
-            .inner_collector
-            .req
-            .sort
-            .iter()
-            .enumerate()
-            .map(|(idx, KeyOrder { order, .. })| {
-                let order = *order;
-                let value = accessors
-                    .get(idx)
-                    .expect("could not find field in accessors")
-                    .0
-                    .values_for_doc(doc_id)
-                    .next();
-                ComparableDocFeature { value, order }
-            })
-            .collect();
-
-        let retrieval_result = self
-            .inner_collector
-            .req
-            .retrieval
-            .get_document_field_data(value_accessors, doc_id);
-
-        self.inner_collector.collect(
-            ComparableDocFeatures(features, retrieval_result),
-            DocAddress {
-                segment_ord: self.segment_ordinal,
-                doc_id,
-            },
-        );
-        Ok(())
-    }
-
-    fn collect_block(
-        &mut self,
-        docs: &[crate::DocId],
-        agg_with_accessor: &mut crate::aggregation::agg_req_with_accessor::AggregationsWithAccessor,
-    ) -> crate::Result<()> {
-        // TODO: Consider getting fields with the column block accessor and refactor this.
-        // ---
-        // Would the additional complexity of getting fields with the column_block_accessor
-        // make sense here? Probably yes, but I want to get a first-pass review first
-        // before proceeding.
-        for doc in docs {
-            self.collect(*doc, agg_with_accessor)?;
-        }
-        Ok(())
-    }
-}
-
-#[cfg(test)]
-mod tests {
-    use common::DateTime;
-    use pretty_assertions::assert_eq;
-    use serde_json::Value;
-    use time::macros::datetime;
-
-    use super::{ComparableDocFeature, ComparableDocFeatures, Order};
-    use crate::aggregation::agg_req::Aggregations;
-    use crate::aggregation::agg_result::AggregationResults;
-    use crate::aggregation::bucket::tests::get_test_index_from_docs;
-    use crate::aggregation::tests::get_test_index_from_values;
-    use crate::aggregation::AggregationCollector;
-    use crate::collector::ComparableDoc;
-    use crate::query::AllQuery;
-    use crate::schema::OwnedValue as SchemaValue;
-
-    fn invert_order(cmp_feature: ComparableDocFeature) -> ComparableDocFeature {
-        let ComparableDocFeature { value, order } = cmp_feature;
-        let order = match order {
-            Order::Asc => Order::Desc,
-            Order::Desc => Order::Asc,
-        };
-        ComparableDocFeature { value, order }
-    }
-
-    fn collector_with_capacity(capacity: usize) -> super::TopHitsCollector {
-        super::TopHitsCollector {
-            top_n: super::TopNComputer::new(capacity),
-            ..Default::default()
-        }
-    }
-
-    fn invert_order_features(cmp_features: ComparableDocFeatures) -> ComparableDocFeatures {
-        let ComparableDocFeatures(cmp_features, search_results) = cmp_features;
-        let cmp_features = cmp_features
-            .into_iter()
-            .map(invert_order)
-            .collect::<Vec<_>>();
-        ComparableDocFeatures(cmp_features, search_results)
-    }
-
-    #[test]
-    fn test_comparable_doc_feature() -> crate::Result<()> {
-        let small = ComparableDocFeature {
-            value: Some(1),
-            order: Order::Asc,
-        };
-        let big = ComparableDocFeature {
-            value: Some(2),
-            order: Order::Asc,
-        };
-        let none = ComparableDocFeature {
-            value: None,
-            order: Order::Asc,
-        };
-
-        assert!(small < big);
-        assert!(none < small);
-        assert!(none < big);
-
-        let small = invert_order(small);
-        let big = invert_order(big);
-        let none = invert_order(none);
-
-        assert!(small > big);
-        assert!(none < small);
-        assert!(none < big);
-
-        Ok(())
-    }
-
-    #[test]
-    fn test_comparable_doc_features() -> crate::Result<()> {
-        let features_1 = ComparableDocFeatures(
-            vec![ComparableDocFeature {
-                value: Some(1),
-                order: Order::Asc,
-            }],
-            Default::default(),
-        );
-
-        let features_2 = ComparableDocFeatures(
-            vec![ComparableDocFeature {
-                value: Some(2),
-                order: Order::Asc,
-            }],
-            Default::default(),
-        );
-
-        assert!(features_1 < features_2);
-
-        assert!(invert_order_features(features_1.clone()) > invert_order_features(features_2));
-
-        Ok(())
-    }
-
-    #[test]
-    fn test_aggregation_top_hits_empty_index() -> crate::Result<()> {
-        let values = vec![];
-
-        let index = get_test_index_from_values(false, &values)?;
-
-        let d: Aggregations = serde_json::from_value(json!({
-            "top_hits_req": {
-                "top_hits": {
-                    "size": 2,
-                    "sort": [
-                        { "date": "desc" }
-                    ],
-                    "from": 0,
-                }
-        }
-        }))
-        .unwrap();
-
-        let collector = AggregationCollector::from_aggs(d, 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).expect("JSON serialization failed"),
-        )
-        .expect("JSON parsing failed");
-
-        assert_eq!(
-            res,
-            json!({
-                "top_hits_req": {
-                    "hits": []
-                }
-            })
-        );
-
-        Ok(())
-    }
-
-    #[test]
-    fn test_top_hits_collector_single_feature() -> crate::Result<()> {
-        let docs = vec![
-            ComparableDoc::<_, _, false> {
-                doc: crate::DocAddress {
-                    segment_ord: 0,
-                    doc_id: 0,
-                },
-                feature: ComparableDocFeatures(
-                    vec![ComparableDocFeature {
-                        value: Some(1),
-                        order: Order::Asc,
-                    }],
-                    Default::default(),
-                ),
-            },
-            ComparableDoc {
-                doc: crate::DocAddress {
-                    segment_ord: 0,
-                    doc_id: 2,
-                },
-                feature: ComparableDocFeatures(
-                    vec![ComparableDocFeature {
-                        value: Some(3),
-                        order: Order::Asc,
-                    }],
-                    Default::default(),
-                ),
-            },
-            ComparableDoc {
-                doc: crate::DocAddress {
-                    segment_ord: 0,
-                    doc_id: 1,
-                },
-                feature: ComparableDocFeatures(
-                    vec![ComparableDocFeature {
-                        value: Some(5),
-                        order: Order::Asc,
-                    }],
-                    Default::default(),
-                ),
-            },
-        ];
-
-        let mut collector = collector_with_capacity(3);
-        for doc in docs.clone() {
-            collector.collect(doc.feature, doc.doc);
-        }
-
-        let res = collector.finalize();
-
-        assert_eq!(
-            res,
-            super::TopHitsMetricResult {
-                hits: vec![
-                    super::TopHitsVecEntry {
-                        sort: vec![docs[0].feature.0[0].value],
-                        search_results: Default::default(),
-                    },
-                    super::TopHitsVecEntry {
-                        sort: vec![docs[1].feature.0[0].value],
-                        search_results: Default::default(),
-                    },
-                    super::TopHitsVecEntry {
-                        sort: vec![docs[2].feature.0[0].value],
-                        search_results: Default::default(),
-                    },
-                ]
-            }
-        );
-
-        Ok(())
-    }
-
-    fn test_aggregation_top_hits(merge_segments: bool) -> crate::Result<()> {
-        let docs = vec![
-            vec![
-                r#"{ "date": "2015-01-02T00:00:00Z", "text": "bbb", "text2": "bbb", "mixed": { "dyn_arr": [1, "2"] } }"#,
-                r#"{ "date": "2017-06-15T00:00:00Z", "text": "ccc", "text2": "ddd", "mixed": { "dyn_arr": [3, "4"] } }"#,
-            ],
-            vec![
-                r#"{ "text": "aaa", "text2": "bbb", "date": "2018-01-02T00:00:00Z", "mixed": { "dyn_arr": ["9", 8] } }"#,
-                r#"{ "text": "aaa", "text2": "bbb", "date": "2016-01-02T00:00:00Z", "mixed": { "dyn_arr": ["7", 6] } }"#,
-            ],
-        ];
-
-        let index = get_test_index_from_docs(merge_segments, &docs)?;
-
-        let d: Aggregations = serde_json::from_value(json!({
-            "top_hits_req": {
-                "top_hits": {
-                    "size": 2,
-                    "sort": [
-                        { "date": "desc" }
-                    ],
-                    "from": 1,
-                    "docvalue_fields": [
-                        "date",
-                        "tex*",
-                        "mixed.*",
-                    ],
-                }
-        }
-        }))?;
-
-        let collector = AggregationCollector::from_aggs(d, Default::default());
-        let reader = index.reader()?;
-        let searcher = reader.searcher();
-
-        let agg_res =
-            serde_json::to_value(searcher.search(&AllQuery, &collector).unwrap()).unwrap();
-
-        let date_2017 = datetime!(2017-06-15 00:00:00 UTC);
-        let date_2016 = datetime!(2016-01-02 00:00:00 UTC);
-
-        assert_eq!(
-            agg_res["top_hits_req"],
-            json!({
-                "hits": [
-                    {
-                        "sort": [common::i64_to_u64(date_2017.unix_timestamp_nanos() as i64)],
-                        "docvalue_fields": {
-                            "date": [ SchemaValue::Date(DateTime::from_utc(date_2017)) ],
-                            "text": [ "ccc" ],
-                            "text2": [ "ddd" ],
-                            "mixed.dyn_arr": [ 3, "4" ],
-                        }
-                    },
-                    {
-                        "sort": [common::i64_to_u64(date_2016.unix_timestamp_nanos() as i64)],
-                        "docvalue_fields": {
-                            "date": [ SchemaValue::Date(DateTime::from_utc(date_2016)) ],
-                            "text": [ "aaa" ],
-                            "text2": [ "bbb" ],
-                            "mixed.dyn_arr": [ 6, "7" ],
-                        }
-                    }
-                ]
-            }),
-        );
-
-        Ok(())
-    }
-
-    #[test]
-    fn test_aggregation_top_hits_single_segment() -> crate::Result<()> {
-        test_aggregation_top_hits(true)
-    }
-
-    #[test]
-    fn test_aggregation_top_hits_multi_segment() -> crate::Result<()> {
-        test_aggregation_top_hits(false)
-    }
-}

src/aggregation/mod.rs

@@ -145,8 +145,6 @@ mod agg_tests;
 
 mod agg_bench;
 
-use core::fmt;
-
 pub use agg_limits::AggregationLimits;
 pub use collector::{
     AggregationCollector, AggregationSegmentCollector, DistributedAggregationCollector,
@@ -156,106 +154,7 @@ use columnar::{ColumnType, MonotonicallyMappableToU64};
 pub(crate) use date::format_date;
 pub use error::AggregationError;
 use itertools::Itertools;
-use serde::de::{self, Visitor};
-use serde::{Deserialize, Deserializer, Serialize};
-
-fn parse_str_into_f64<E: de::Error>(value: &str) -> Result<f64, E> {
-    let parsed = value.parse::<f64>().map_err(|_err| {
-        de::Error::custom(format!("Failed to parse f64 from string: {:?}", value))
-    })?;
-
-    // Check if the parsed value is NaN or infinity
-    if parsed.is_nan() || parsed.is_infinite() {
-        Err(de::Error::custom(format!(
-            "Value is not a valid f64 (NaN or Infinity): {:?}",
-            value
-        )))
-    } else {
-        Ok(parsed)
-    }
-}
-
-/// deserialize Option<f64> from string or float
-pub(crate) fn deserialize_option_f64<'de, D>(deserializer: D) -> Result<Option<f64>, D::Error>
-where D: Deserializer<'de> {
-    struct StringOrFloatVisitor;
-
-    impl<'de> Visitor<'de> for StringOrFloatVisitor {
-        type Value = Option<f64>;
-
-        fn expecting(&self, formatter: &mut fmt::Formatter) -> fmt::Result {
-            formatter.write_str("a string or a float")
-        }
-
-        fn visit_str<E>(self, value: &str) -> Result<Self::Value, E>
-        where E: de::Error {
-            parse_str_into_f64(value).map(Some)
-        }
-
-        fn visit_f64<E>(self, value: f64) -> Result<Self::Value, E>
-        where E: de::Error {
-            Ok(Some(value))
-        }
-
-        fn visit_i64<E>(self, value: i64) -> Result<Self::Value, E>
-        where E: de::Error {
-            Ok(Some(value as f64))
-        }
-
-        fn visit_u64<E>(self, value: u64) -> Result<Self::Value, E>
-        where E: de::Error {
-            Ok(Some(value as f64))
-        }
-
-        fn visit_none<E>(self) -> Result<Self::Value, E>
-        where E: de::Error {
-            Ok(None)
-        }
-
-        fn visit_unit<E>(self) -> Result<Self::Value, E>
-        where E: de::Error {
-            Ok(None)
-        }
-    }
-
-    deserializer.deserialize_any(StringOrFloatVisitor)
-}
-
-/// deserialize f64 from string or float
-pub(crate) fn deserialize_f64<'de, D>(deserializer: D) -> Result<f64, D::Error>
-where D: Deserializer<'de> {
-    struct StringOrFloatVisitor;
-
-    impl<'de> Visitor<'de> for StringOrFloatVisitor {
-        type Value = f64;
-
-        fn expecting(&self, formatter: &mut fmt::Formatter) -> fmt::Result {
-            formatter.write_str("a string or a float")
-        }
-
-        fn visit_str<E>(self, value: &str) -> Result<Self::Value, E>
-        where E: de::Error {
-            parse_str_into_f64(value)
-        }
-
-        fn visit_f64<E>(self, value: f64) -> Result<Self::Value, E>
-        where E: de::Error {
-            Ok(value)
-        }
-
-        fn visit_i64<E>(self, value: i64) -> Result<Self::Value, E>
-        where E: de::Error {
-            Ok(value as f64)
-        }
-
-        fn visit_u64<E>(self, value: u64) -> Result<Self::Value, E>
-        where E: de::Error {
-            Ok(value as f64)
-        }
-    }
-
-    deserializer.deserialize_any(StringOrFloatVisitor)
-}
+use serde::{Deserialize, Serialize};
 
 /// Represents an associative array `(key => values)` in a very efficient manner.
 #[derive(PartialEq, Serialize, Deserialize)]
@@ -382,7 +281,6 @@ pub(crate) fn f64_from_fastfield_u64(val: u64, field_type: &ColumnType) -> f64 {
         ColumnType::U64 => val as f64,
         ColumnType::I64 | ColumnType::DateTime => i64::from_u64(val) as f64,
         ColumnType::F64 => f64::from_u64(val),
-        ColumnType::Bool => val as f64,
         _ => {
             panic!("unexpected type {field_type:?}. This should not happen")
         }
@@ -403,7 +301,6 @@ pub(crate) fn f64_to_fastfield_u64(val: f64, field_type: &ColumnType) -> Option<
         ColumnType::U64 => Some(val as u64),
         ColumnType::I64 | ColumnType::DateTime => Some((val as i64).to_u64()),
         ColumnType::F64 => Some(val.to_u64()),
-        ColumnType::Bool => Some(val as u64),
         _ => None,
     }
 }

src/aggregation/segment_agg_result.rs

@@ -16,7 +16,6 @@ use super::metric::{
     SumAggregation,
 };
 use crate::aggregation::bucket::TermMissingAgg;
-use crate::aggregation::metric::SegmentTopHitsCollector;
 
 pub(crate) trait SegmentAggregationCollector: CollectorClone + Debug {
     fn add_intermediate_aggregation_result(
@@ -161,11 +160,6 @@ pub(crate) fn build_single_agg_segment_collector(
                 accessor_idx,
             )?,
         )),
-        TopHits(top_hits_req) => Ok(Box::new(SegmentTopHitsCollector::from_req(
-            top_hits_req,
-            accessor_idx,
-            req.segment_ordinal,
-        ))),
     }
 }
 

src/collector/facet_collector.rs

@@ -410,7 +410,6 @@ impl SegmentCollector for FacetSegmentCollector {
 
 /// Intermediary result of the `FacetCollector` that stores
 /// the facet counts for all the segments.
-#[derive(Default, Clone)]
 pub struct FacetCounts {
     facet_counts: BTreeMap<Facet, u64>,
 }
@@ -494,7 +493,7 @@ mod tests {
     use super::{FacetCollector, FacetCounts};
     use crate::collector::facet_collector::compress_mapping;
     use crate::collector::Count;
-    use crate::index::Index;
+    use crate::core::Index;
     use crate::query::{AllQuery, QueryParser, TermQuery};
     use crate::schema::{Facet, FacetOptions, IndexRecordOption, Schema, TantivyDocument};
     use crate::{IndexWriter, Term};

src/collector/mod.rs

@@ -97,7 +97,6 @@ pub use self::multi_collector::{FruitHandle, MultiCollector, MultiFruit};
 mod top_collector;
 
 mod top_score_collector;
-pub use self::top_collector::ComparableDoc;
 pub use self::top_score_collector::{TopDocs, TopNComputer};
 
 mod custom_score_top_collector;

src/collector/tests.rs

@@ -2,7 +2,7 @@ use columnar::{BytesColumn, Column};
 
 use super::*;
 use crate::collector::{Count, FilterCollector, TopDocs};
-use crate::index::SegmentReader;
+use crate::core::SegmentReader;
 use crate::query::{AllQuery, QueryParser};
 use crate::schema::{Schema, FAST, TEXT};
 use crate::time::format_description::well_known::Rfc3339;

src/collector/top_collector.rs

@@ -1,58 +1,47 @@
 use std::cmp::Ordering;
 use std::marker::PhantomData;
 
-use serde::{Deserialize, Serialize};
-
 use super::top_score_collector::TopNComputer;
 use crate::{DocAddress, DocId, SegmentOrdinal, SegmentReader};
 
 /// Contains a feature (field, score, etc.) of a document along with the document address.
 ///
-/// It guarantees stable sorting: in case of a tie on the feature, the document
-/// address is used.
+/// It has a custom implementation of `PartialOrd` that reverses the order. This is because the
+/// default Rust heap is a max heap, whereas a min heap is needed.
 ///
-/// The REVERSE_ORDER generic parameter controls whether the by-feature order
-/// should be reversed, which is useful for achieving for example largest-first
-/// semantics without having to wrap the feature in a `Reverse`.
+/// Additionally, it guarantees stable sorting: in case of a tie on the feature, the document
+/// address is used.
 ///
 /// WARNING: equality is not what you would expect here.
 /// Two elements are equal if their feature is equal, and regardless of whether `doc`
 /// is equal. This should be perfectly fine for this usage, but let's make sure this
 /// struct is never public.
-#[derive(Clone, Default, Serialize, Deserialize)]
-pub struct ComparableDoc<T, D, const REVERSE_ORDER: bool = false> {
-    /// The feature of the document. In practice, this is
-    /// is any type that implements `PartialOrd`.
+pub(crate) struct ComparableDoc<T, D> {
     pub feature: T,
-    /// The document address. In practice, this is any
-    /// type that implements `PartialOrd`, and is guaranteed
-    /// to be unique for each document.
     pub doc: D,
 }
-impl<T: std::fmt::Debug, D: std::fmt::Debug, const R: bool> std::fmt::Debug
-    for ComparableDoc<T, D, R>
-{
+impl<T: std::fmt::Debug, D: std::fmt::Debug> std::fmt::Debug for ComparableDoc<T, D> {
     fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
-        f.debug_struct(format!("ComparableDoc<_, _ {R}").as_str())
+        f.debug_struct("ComparableDoc")
             .field("feature", &self.feature)
             .field("doc", &self.doc)
             .finish()
     }
 }
 
-impl<T: PartialOrd, D: PartialOrd, const R: bool> PartialOrd for ComparableDoc<T, D, R> {
+impl<T: PartialOrd, D: PartialOrd> PartialOrd for ComparableDoc<T, D> {
     fn partial_cmp(&self, other: &Self) -> Option<Ordering> {
         Some(self.cmp(other))
     }
 }
 
-impl<T: PartialOrd, D: PartialOrd, const R: bool> Ord for ComparableDoc<T, D, R> {
+impl<T: PartialOrd, D: PartialOrd> Ord for ComparableDoc<T, D> {
     #[inline]
     fn cmp(&self, other: &Self) -> Ordering {
-        let by_feature = self
+        // Reversed to make BinaryHeap work as a min-heap
+        let by_feature = other
             .feature
-            .partial_cmp(&other.feature)
-            .map(|ord| if R { ord.reverse() } else { ord })
+            .partial_cmp(&self.feature)
             .unwrap_or(Ordering::Equal);
 
         let lazy_by_doc_address = || self.doc.partial_cmp(&other.doc).unwrap_or(Ordering::Equal);
@@ -64,13 +53,13 @@ impl<T: PartialOrd, D: PartialOrd, const R: bool> Ord for ComparableDoc<T, D, R>
     }
 }
 
-impl<T: PartialOrd, D: PartialOrd, const R: bool> PartialEq for ComparableDoc<T, D, R> {
+impl<T: PartialOrd, D: PartialOrd> PartialEq for ComparableDoc<T, D> {
     fn eq(&self, other: &Self) -> bool {
         self.cmp(other) == Ordering::Equal
     }
 }
 
-impl<T: PartialOrd, D: PartialOrd, const R: bool> Eq for ComparableDoc<T, D, R> {}
+impl<T: PartialOrd, D: PartialOrd> Eq for ComparableDoc<T, D> {}
 
 pub(crate) struct TopCollector<T> {
     pub limit: usize,
@@ -110,10 +99,10 @@ where T: PartialOrd + Clone
         if self.limit == 0 {
             return Ok(Vec::new());
         }
-        let mut top_collector: TopNComputer<_, _> = TopNComputer::new(self.limit + self.offset);
+        let mut top_collector = TopNComputer::new(self.limit + self.offset);
         for child_fruit in children {
             for (feature, doc) in child_fruit {
-                top_collector.push(feature, doc);
+                top_collector.push(ComparableDoc { feature, doc });
             }
         }
 
@@ -154,8 +143,6 @@ where T: PartialOrd + Clone
 /// The theoretical complexity for collecting the top `K` out of `n` documents
 /// is `O(n + K)`.
 pub(crate) struct TopSegmentCollector<T> {
-    /// We reverse the order of the feature in order to
-    /// have top-semantics instead of bottom semantics.
     topn_computer: TopNComputer<T, DocId>,
     segment_ord: u32,
 }
@@ -193,7 +180,7 @@ impl<T: PartialOrd + Clone> TopSegmentCollector<T> {
     /// will compare the lowest scoring item with the given one and keep whichever is greater.
     #[inline]
     pub fn collect(&mut self, doc: DocId, feature: T) {
-        self.topn_computer.push(feature, doc);
+        self.topn_computer.push(ComparableDoc { feature, doc });
     }
 }
 

src/collector/top_score_collector.rs

@@ -3,8 +3,6 @@ use std::marker::PhantomData;
 use std::sync::Arc;
 
 use columnar::ColumnValues;
-use serde::de::DeserializeOwned;
-use serde::{Deserialize, Serialize};
 
 use super::Collector;
 use crate::collector::custom_score_top_collector::CustomScoreTopCollector;
@@ -311,7 +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,
@@ -665,7 +663,7 @@ impl Collector for TopDocs {
         reader: &SegmentReader,
     ) -> crate::Result<<Self::Child as SegmentCollector>::Fruit> {
         let heap_len = self.0.limit + self.0.offset;
-        let mut top_n: TopNComputer<_, _> = TopNComputer::new(heap_len);
+        let mut top_n = TopNComputer::new(heap_len);
 
         if let Some(alive_bitset) = reader.alive_bitset() {
             let mut threshold = Score::MIN;
@@ -674,13 +672,21 @@ impl Collector for TopDocs {
                 if alive_bitset.is_deleted(doc) {
                     return threshold;
                 }
-                top_n.push(score, doc);
+                let doc = ComparableDoc {
+                    feature: score,
+                    doc,
+                };
+                top_n.push(doc);
                 threshold = top_n.threshold.unwrap_or(Score::MIN);
                 threshold
             })?;
         } else {
             weight.for_each_pruning(Score::MIN, reader, &mut |doc, score| {
-                top_n.push(score, doc);
+                let doc = ComparableDoc {
+                    feature: score,
+                    doc,
+                };
+                top_n.push(doc);
                 top_n.threshold.unwrap_or(Score::MIN)
             })?;
         }
@@ -719,65 +725,17 @@ impl SegmentCollector for TopScoreSegmentCollector {
 
 /// Fast TopN Computation
 ///
-/// Capacity of the vec is 2 * top_n.
-/// The buffer is truncated to the top_n elements when it reaches the capacity of the Vec.
-/// That means capacity has special meaning and should be carried over when cloning or serializing.
-///
 /// For TopN == 0, it will be relative expensive.
-#[derive(Serialize, Deserialize)]
-#[serde(from = "TopNComputerDeser<Score, D, REVERSE_ORDER>")]
-pub struct TopNComputer<Score, D, const REVERSE_ORDER: bool = true> {
-    /// The buffer reverses sort order to get top-semantics instead of bottom-semantics
-    buffer: Vec<ComparableDoc<Score, D, REVERSE_ORDER>>,
+pub struct TopNComputer<Score, DocId> {
+    buffer: Vec<ComparableDoc<Score, DocId>>,
     top_n: usize,
     pub(crate) threshold: Option<Score>,
 }
-// Intermediate struct for TopNComputer for deserialization, to keep vec capacity
-#[derive(Deserialize)]
-struct TopNComputerDeser<Score, D, const REVERSE_ORDER: bool> {
-    buffer: Vec<ComparableDoc<Score, D, REVERSE_ORDER>>,
-    top_n: usize,
-    threshold: Option<Score>,
-}
 
-// Custom clone to keep capacity
-impl<Score: Clone, D: Clone, const REVERSE_ORDER: bool> Clone
-    for TopNComputer<Score, D, REVERSE_ORDER>
-{
-    fn clone(&self) -> Self {
-        let mut buffer_clone = Vec::with_capacity(self.buffer.capacity());
-        buffer_clone.extend(self.buffer.iter().cloned());
-
-        TopNComputer {
-            buffer: buffer_clone,
-            top_n: self.top_n,
-            threshold: self.threshold.clone(),
-        }
-    }
-}
-
-impl<Score, D, const R: bool> From<TopNComputerDeser<Score, D, R>> for TopNComputer<Score, D, R> {
-    fn from(mut value: TopNComputerDeser<Score, D, R>) -> Self {
-        let expected_cap = value.top_n.max(1) * 2;
-        let current_cap = value.buffer.capacity();
-        if current_cap < expected_cap {
-            value.buffer.reserve_exact(expected_cap - current_cap);
-        } else {
-            value.buffer.shrink_to(expected_cap);
-        }
-
-        TopNComputer {
-            buffer: value.buffer,
-            top_n: value.top_n,
-            threshold: value.threshold,
-        }
-    }
-}
-
-impl<Score, D, const R: bool> TopNComputer<Score, D, R>
+impl<Score, DocId> TopNComputer<Score, DocId>
 where
     Score: PartialOrd + Clone,
-    D: Serialize + DeserializeOwned + Ord + Clone,
+    DocId: Ord + Clone,
 {
     /// Create a new `TopNComputer`.
     /// Internally it will allocate a buffer of size `2 * top_n`.
@@ -790,12 +748,10 @@ where
         }
     }
 
-    /// Push a new document to the top n.
-    /// If the document is below the current threshold, it will be ignored.
     #[inline]
-    pub fn push(&mut self, feature: Score, doc: D) {
+    pub(crate) fn push(&mut self, doc: ComparableDoc<Score, DocId>) {
         if let Some(last_median) = self.threshold.clone() {
-            if feature < last_median {
+            if doc.feature < last_median {
                 return;
             }
         }
@@ -810,7 +766,7 @@ where
         let uninit = self.buffer.spare_capacity_mut();
         // This cannot panic, because we truncate_median will at least remove one element, since
         // the min capacity is 2.
-        uninit[0].write(ComparableDoc { doc, feature });
+        uninit[0].write(doc);
         // This is safe because it would panic in the line above
         unsafe {
             self.buffer.set_len(self.buffer.len() + 1);
@@ -829,24 +785,13 @@ where
         median_score
     }
 
-    /// Returns the top n elements in sorted order.
-    pub fn into_sorted_vec(mut self) -> Vec<ComparableDoc<Score, D, R>> {
+    pub(crate) fn into_sorted_vec(mut self) -> Vec<ComparableDoc<Score, DocId>> {
         if self.buffer.len() > self.top_n {
             self.truncate_top_n();
         }
         self.buffer.sort_unstable();
         self.buffer
     }
-
-    /// Returns the top n elements in stored order.
-    /// Useful if you do not need the elements in sorted order,
-    /// for example when merging the results of multiple segments.
-    pub fn into_vec(mut self) -> Vec<ComparableDoc<Score, D, R>> {
-        if self.buffer.len() > self.top_n {
-            self.truncate_top_n();
-        }
-        self.buffer
-    }
 }
 
 #[cfg(test)]
@@ -880,44 +825,49 @@ mod tests {
             crate::assert_nearly_equals!(result.0, expected.0);
         }
     }
-    #[test]
-    fn test_topn_computer_serde() {
-        let computer: TopNComputer<u32, u32> = TopNComputer::new(1);
-
-        let computer_ser = serde_json::to_string(&computer).unwrap();
-        let mut computer: TopNComputer<u32, u32> = serde_json::from_str(&computer_ser).unwrap();
-
-        computer.push(1u32, 5u32);
-        computer.push(1u32, 0u32);
-        computer.push(1u32, 7u32);
-
-        assert_eq!(
-            computer.into_sorted_vec(),
-            &[ComparableDoc {
-                feature: 1u32,
-                doc: 0u32,
-            },]
-        );
-    }
 
     #[test]
     fn test_empty_topn_computer() {
         let mut computer: TopNComputer<u32, u32> = TopNComputer::new(0);
 
-        computer.push(1u32, 1u32);
-        computer.push(1u32, 2u32);
-        computer.push(1u32, 3u32);
+        computer.push(ComparableDoc {
+            feature: 1u32,
+            doc: 1u32,
+        });
+        computer.push(ComparableDoc {
+            feature: 1u32,
+            doc: 2u32,
+        });
+        computer.push(ComparableDoc {
+            feature: 1u32,
+            doc: 3u32,
+        });
         assert!(computer.into_sorted_vec().is_empty());
     }
     #[test]
     fn test_topn_computer() {
         let mut computer: TopNComputer<u32, u32> = TopNComputer::new(2);
 
-        computer.push(1u32, 1u32);
-        computer.push(2u32, 2u32);
-        computer.push(3u32, 3u32);
-        computer.push(2u32, 4u32);
-        computer.push(1u32, 5u32);
+        computer.push(ComparableDoc {
+            feature: 1u32,
+            doc: 1u32,
+        });
+        computer.push(ComparableDoc {
+            feature: 2u32,
+            doc: 2u32,
+        });
+        computer.push(ComparableDoc {
+            feature: 3u32,
+            doc: 3u32,
+        });
+        computer.push(ComparableDoc {
+            feature: 2u32,
+            doc: 4u32,
+        });
+        computer.push(ComparableDoc {
+            feature: 1u32,
+            doc: 5u32,
+        });
         assert_eq!(
             computer.into_sorted_vec(),
             &[
@@ -939,7 +889,10 @@ mod tests {
             let mut computer: TopNComputer<u32, u32> = TopNComputer::new(top_n);
 
             for _ in 0..1 + top_n * 2 {
-                computer.push(1u32, 1u32);
+                computer.push(ComparableDoc {
+                    feature: 1u32,
+                    doc: 1u32,
+                });
             }
             let _vals = computer.into_sorted_vec();
         }

src/core/index.rs

@@ -6,23 +6,24 @@ use std::path::PathBuf;
 use std::sync::Arc;
 
 use super::segment::Segment;
-use super::segment_reader::merge_field_meta_data;
-use super::{FieldMetadata, IndexSettings};
-use crate::core::{Executor, META_FILEPATH};
+use super::IndexSettings;
+use crate::core::single_segment_index_writer::SingleSegmentIndexWriter;
+use crate::core::{
+    Executor, IndexMeta, SegmentId, SegmentMeta, SegmentMetaInventory, META_FILEPATH,
+};
 use crate::directory::error::OpenReadError;
 #[cfg(feature = "mmap")]
 use crate::directory::MmapDirectory;
 use crate::directory::{Directory, ManagedDirectory, RamDirectory, INDEX_WRITER_LOCK};
 use crate::error::{DataCorruption, TantivyError};
-use crate::index::{IndexMeta, SegmentId, SegmentMeta, SegmentMetaInventory};
 use crate::indexer::index_writer::{MAX_NUM_THREAD, MEMORY_BUDGET_NUM_BYTES_MIN};
 use crate::indexer::segment_updater::save_metas;
-use crate::indexer::{IndexWriter, SingleSegmentIndexWriter};
+use crate::indexer::IndexWriter;
 use crate::reader::{IndexReader, IndexReaderBuilder};
 use crate::schema::document::Document;
 use crate::schema::{Field, FieldType, Schema};
 use crate::tokenizer::{TextAnalyzer, TokenizerManager};
-use crate::SegmentReader;
+use crate::{merge_field_meta_data, FieldMetadata, SegmentReader};
 
 fn load_metas(
     directory: &dyn Directory,
@@ -322,15 +323,6 @@ impl Index {
         Ok(())
     }
 
-    /// Custom thread pool by a outer thread pool.
-    pub fn set_shared_multithread_executor(
-        &mut self,
-        shared_thread_pool: Arc<Executor>,
-    ) -> crate::Result<()> {
-        self.executor = shared_thread_pool.clone();
-        Ok(())
-    }
-
     /// Replace the default single thread search executor pool
     /// by a thread pool with as many threads as there are CPUs on the system.
     pub fn set_default_multithread_executor(&mut self) -> crate::Result<()> {

src/core/index_meta.rs

@@ -7,7 +7,7 @@ use std::sync::Arc;
 use serde::{Deserialize, Serialize};
 
 use super::SegmentComponent;
-use crate::index::SegmentId;
+use crate::core::SegmentId;
 use crate::schema::Schema;
 use crate::store::Compressor;
 use crate::{Inventory, Opstamp, TrackedObject};
@@ -19,7 +19,7 @@ struct DeleteMeta {
 }
 
 #[derive(Clone, Default)]
-pub(crate) struct SegmentMetaInventory {
+pub struct SegmentMetaInventory {
     inventory: Inventory<InnerSegmentMeta>,
 }
 
@@ -142,6 +142,7 @@ impl SegmentMeta {
             SegmentComponent::FastFields => ".fast".to_string(),
             SegmentComponent::FieldNorms => ".fieldnorm".to_string(),
             SegmentComponent::Delete => format!(".{}.del", self.delete_opstamp().unwrap_or(0)),
+            SegmentComponent::FieldList => ".fieldlist".to_string(),
         });
         PathBuf::from(path)
     }
@@ -408,7 +409,7 @@ impl fmt::Debug for IndexMeta {
 mod tests {
 
     use super::IndexMeta;
-    use crate::index::index_meta::UntrackedIndexMeta;
+    use crate::core::index_meta::UntrackedIndexMeta;
     use crate::schema::{Schema, TEXT};
     use crate::store::Compressor;
     #[cfg(feature = "zstd-compression")]

src/core/inverted_index_reader.rs

@@ -70,7 +70,7 @@ impl InvertedIndexReader {
         &self.termdict
     }
 
-    /// Return the fields and types encoded in the dictionary in lexicographic oder.
+    /// Return the fields and types encoded in the dictionary in lexicographic order.
     /// Only valid on JSON fields.
     ///
     /// Notice: This requires a full scan and therefore **very expensive**.
@@ -266,9 +266,7 @@ impl InvertedIndexReader {
 
     /// Warmup a block postings given a `Term`.
     /// This method is for an advanced usage only.
-    ///
-    /// returns a boolean, whether the term was found in the dictionary
-    pub async fn warm_postings(&self, term: &Term, with_positions: bool) -> io::Result<bool> {
+    pub async fn warm_postings(&self, term: &Term, with_positions: bool) -> io::Result<()> {
         let term_info_opt: Option<TermInfo> = self.get_term_info_async(term).await?;
         if let Some(term_info) = term_info_opt {
             let postings = self
@@ -282,27 +280,23 @@ impl InvertedIndexReader {
             } else {
                 postings.await?;
             }
-            Ok(true)
-        } else {
-            Ok(false)
         }
+        Ok(())
     }
 
     /// Warmup a block postings given a range of `Term`s.
     /// This method is for an advanced usage only.
-    ///
-    /// returns a boolean, whether a term matching the range was found in the dictionary
     pub async fn warm_postings_range(
         &self,
         terms: impl std::ops::RangeBounds<Term>,
         limit: Option<u64>,
         with_positions: bool,
-    ) -> io::Result<bool> {
+    ) -> io::Result<()> {
         let mut term_info = self.get_term_range_async(terms, limit).await?;
 
         let Some(first_terminfo) = term_info.next() else {
             // no key matches, nothing more to load
-            return Ok(false);
+            return Ok(());
         };
 
         let last_terminfo = term_info.last().unwrap_or_else(|| first_terminfo.clone());
@@ -322,7 +316,7 @@ impl InvertedIndexReader {
         } else {
             postings.await?;
         }
-        Ok(true)
+        Ok(())
     }
 
     /// Warmup the block postings for all terms.

src/core/json_utils.rs

@@ -1,4 +1,4 @@
-use columnar::MonotonicallyMappableToU64;
+use columnar::{ColumnType, MonotonicallyMappableToU64};
 use common::{replace_in_place, JsonPathWriter};
 use rustc_hash::FxHashMap;
 
@@ -153,7 +153,7 @@ fn index_json_value<'a, V: Value<'a>>(
                 let mut token_stream = text_analyzer.token_stream(val);
                 let unordered_id = ctx
                     .path_to_unordered_id
-                    .get_or_allocate_unordered_id(json_path_writer.as_str());
+                    .get_or_allocate_unordered_id(json_path_writer.as_str(), ColumnType::Str);
 
                 // TODO: make sure the chain position works out.
                 set_path_id(term_buffer, unordered_id);
@@ -171,7 +171,7 @@ fn index_json_value<'a, V: Value<'a>>(
                 set_path_id(
                     term_buffer,
                     ctx.path_to_unordered_id
-                        .get_or_allocate_unordered_id(json_path_writer.as_str()),
+                        .get_or_allocate_unordered_id(json_path_writer.as_str(), ColumnType::U64),
                 );
                 term_buffer.append_type_and_fast_value(val);
                 postings_writer.subscribe(doc, 0u32, term_buffer, ctx);
@@ -180,7 +180,7 @@ fn index_json_value<'a, V: Value<'a>>(
                 set_path_id(
                     term_buffer,
                     ctx.path_to_unordered_id
-                        .get_or_allocate_unordered_id(json_path_writer.as_str()),
+                        .get_or_allocate_unordered_id(json_path_writer.as_str(), ColumnType::I64),
                 );
                 term_buffer.append_type_and_fast_value(val);
                 postings_writer.subscribe(doc, 0u32, term_buffer, ctx);
@@ -189,7 +189,7 @@ fn index_json_value<'a, V: Value<'a>>(
                 set_path_id(
                     term_buffer,
                     ctx.path_to_unordered_id
-                        .get_or_allocate_unordered_id(json_path_writer.as_str()),
+                        .get_or_allocate_unordered_id(json_path_writer.as_str(), ColumnType::F64),
                 );
                 term_buffer.append_type_and_fast_value(val);
                 postings_writer.subscribe(doc, 0u32, term_buffer, ctx);
@@ -198,7 +198,7 @@ fn index_json_value<'a, V: Value<'a>>(
                 set_path_id(
                     term_buffer,
                     ctx.path_to_unordered_id
-                        .get_or_allocate_unordered_id(json_path_writer.as_str()),
+                        .get_or_allocate_unordered_id(json_path_writer.as_str(), ColumnType::Bool),
                 );
                 term_buffer.append_type_and_fast_value(val);
                 postings_writer.subscribe(doc, 0u32, term_buffer, ctx);
@@ -206,8 +206,10 @@ fn index_json_value<'a, V: Value<'a>>(
             ReferenceValueLeaf::Date(val) => {
                 set_path_id(
                     term_buffer,
-                    ctx.path_to_unordered_id
-                        .get_or_allocate_unordered_id(json_path_writer.as_str()),
+                    ctx.path_to_unordered_id.get_or_allocate_unordered_id(
+                        json_path_writer.as_str(),
+                        ColumnType::DateTime,
+                    ),
                 );
                 term_buffer.append_type_and_fast_value(val);
                 postings_writer.subscribe(doc, 0u32, term_buffer, ctx);

src/core/mod.rs

@@ -1,14 +1,32 @@
 mod executor;
+pub mod index;
+mod index_meta;
+mod inverted_index_reader;
 #[doc(hidden)]
 pub mod json_utils;
 pub mod searcher;
+mod segment;
+mod segment_component;
+mod segment_id;
+mod segment_reader;
+mod single_segment_index_writer;
 
 use std::path::Path;
 
 use once_cell::sync::Lazy;
 
 pub use self::executor::Executor;
+pub use self::index::{Index, IndexBuilder};
+pub use self::index_meta::{
+    IndexMeta, IndexSettings, IndexSortByField, Order, SegmentMeta, SegmentMetaInventory,
+};
+pub use self::inverted_index_reader::InvertedIndexReader;
 pub use self::searcher::{Searcher, SearcherGeneration};
+pub use self::segment::Segment;
+pub use self::segment_component::SegmentComponent;
+pub use self::segment_id::SegmentId;
+pub use self::segment_reader::{merge_field_meta_data, FieldMetadata, SegmentReader};
+pub use self::single_segment_index_writer::SingleSegmentIndexWriter;
 
 /// The meta file contains all the information about the list of segments and the schema
 /// of the index.

src/core/searcher.rs

@@ -3,8 +3,7 @@ use std::sync::Arc;
 use std::{fmt, io};
 
 use crate::collector::Collector;
-use crate::core::Executor;
-use crate::index::SegmentReader;
+use crate::core::{Executor, SegmentReader};
 use crate::query::{Bm25StatisticsProvider, EnableScoring, Query};
 use crate::schema::document::DocumentDeserialize;
 use crate::schema::{Schema, Term};

src/core/segment.rs

@@ -2,9 +2,9 @@ use std::fmt;
 use std::path::PathBuf;
 
 use super::SegmentComponent;
+use crate::core::{Index, SegmentId, SegmentMeta};
 use crate::directory::error::{OpenReadError, OpenWriteError};
 use crate::directory::{Directory, FileSlice, WritePtr};
-use crate::index::{Index, SegmentId, SegmentMeta};
 use crate::schema::Schema;
 use crate::Opstamp;
 

src/core/segment_component.rs

@@ -27,12 +27,14 @@ pub enum SegmentComponent {
     /// Bitset describing which document of the segment is alive.
     /// (It was representing deleted docs but changed to represent alive docs from v0.17)
     Delete,
+    /// Field list describing the fields in the segment.
+    FieldList,
 }
 
 impl SegmentComponent {
     /// Iterates through the components.
     pub fn iterator() -> slice::Iter<'static, SegmentComponent> {
-        static SEGMENT_COMPONENTS: [SegmentComponent; 8] = [
+        static SEGMENT_COMPONENTS: [SegmentComponent; 9] = [
             SegmentComponent::Postings,
             SegmentComponent::Positions,
             SegmentComponent::FastFields,
@@ -41,6 +43,7 @@ impl SegmentComponent {
             SegmentComponent::Store,
             SegmentComponent::TempStore,
             SegmentComponent::Delete,
+            SegmentComponent::FieldList,
         ];
         SEGMENT_COMPONENTS.iter()
     }

src/core/segment_reader.rs

@@ -3,15 +3,14 @@ use std::ops::BitOrAssign;
 use std::sync::{Arc, RwLock};
 use std::{fmt, io};
 
-use fnv::FnvHashMap;
 use itertools::Itertools;
 
+use crate::core::{InvertedIndexReader, Segment, SegmentComponent, SegmentId};
 use crate::directory::{CompositeFile, FileSlice};
 use crate::error::DataCorruption;
 use crate::fastfield::{intersect_alive_bitsets, AliveBitSet, FacetReader, FastFieldReaders};
+use crate::field_list::read_split_fields;
 use crate::fieldnorm::{FieldNormReader, FieldNormReaders};
-use crate::index::{InvertedIndexReader, Segment, SegmentComponent, SegmentId};
-use crate::json_utils::json_path_sep_to_dot;
 use crate::schema::{Field, IndexRecordOption, Schema, Type};
 use crate::space_usage::SegmentSpaceUsage;
 use crate::store::StoreReader;
@@ -44,6 +43,7 @@ pub struct SegmentReader {
     fast_fields_readers: FastFieldReaders,
     fieldnorm_readers: FieldNormReaders,
 
+    list_fields_file: Option<FileSlice>, // Optional field list file for backwards compatibility
     store_file: FileSlice,
     alive_bitset_opt: Option<AliveBitSet>,
     schema: Schema,
@@ -153,6 +153,7 @@ impl SegmentReader {
         let termdict_composite = CompositeFile::open(&termdict_file)?;
 
         let store_file = segment.open_read(SegmentComponent::Store)?;
+        let list_fields_file = segment.open_read(SegmentComponent::FieldList).ok();
 
         crate::fail_point!("SegmentReader::open#middle");
 
@@ -201,6 +202,7 @@ impl SegmentReader {
             segment_id: segment.id(),
             delete_opstamp: segment.meta().delete_opstamp(),
             store_file,
+            list_fields_file,
             alive_bitset_opt,
             positions_composite,
             schema,
@@ -299,87 +301,25 @@ impl SegmentReader {
     /// field that is not indexed nor a fast field but is stored, it is possible for the field
     /// to not be listed.
     pub fn fields_metadata(&self) -> crate::Result<Vec<FieldMetadata>> {
-        let mut indexed_fields: Vec<FieldMetadata> = Vec::new();
-        let mut map_to_canonical = FnvHashMap::default();
-        for (field, field_entry) in self.schema().fields() {
-            let field_name = field_entry.name().to_string();
-            let is_indexed = field_entry.is_indexed();
-
-            if is_indexed {
-                let is_json = field_entry.field_type().value_type() == Type::Json;
-                if is_json {
-                    let inv_index = self.inverted_index(field)?;
-                    let encoded_fields_in_index = inv_index.list_encoded_fields()?;
-                    let mut build_path = |field_name: &str, mut json_path: String| {
-                        // In this case we need to map the potential fast field to the field name
-                        // accepted by the query parser.
-                        let create_canonical =
-                            !field_entry.is_expand_dots_enabled() && json_path.contains('.');
-                        if create_canonical {
-                            // Without expand dots enabled dots need to be escaped.
-                            let escaped_json_path = json_path.replace('.', "\\.");
-                            let full_path = format!("{}.{}", field_name, escaped_json_path);
-                            let full_path_unescaped = format!("{}.{}", field_name, &json_path);
-                            map_to_canonical.insert(full_path_unescaped, full_path.to_string());
-                            full_path
-                        } else {
-                            // With expand dots enabled, we can use '.' instead of '\u{1}'.
-                            json_path_sep_to_dot(&mut json_path);
-                            format!("{}.{}", field_name, json_path)
-                        }
-                    };
-                    indexed_fields.extend(
-                        encoded_fields_in_index
-                            .into_iter()
-                            .map(|(name, typ)| (build_path(&field_name, name), typ))
-                            .map(|(field_name, typ)| FieldMetadata {
-                                indexed: true,
-                                stored: false,
-                                field_name,
-                                fast: false,
-                                typ,
-                            }),
-                    );
-                } else {
-                    indexed_fields.push(FieldMetadata {
-                        indexed: true,
-                        stored: false,
-                        field_name: field_name.to_string(),
-                        fast: false,
-                        typ: field_entry.field_type().value_type(),
-                    });
-                }
-            }
+        if let Some(list_fields_file) = self.list_fields_file.as_ref() {
+            let file = list_fields_file.read_bytes()?;
+            let fields_metadata =
+                read_split_fields(file)?.collect::<io::Result<Vec<FieldMetadata>>>();
+            fields_metadata.map_err(|e| e.into())
+        } else {
+            // Schema fallback
+            Ok(self
+                .schema()
+                .fields()
+                .map(|(_field, entry)| FieldMetadata {
+                    field_name: entry.name().to_string(),
+                    typ: entry.field_type().value_type(),
+                    indexed: entry.is_indexed(),
+                    stored: entry.is_stored(),
+                    fast: entry.is_fast(),
+                })
+                .collect())
         }
-        let mut fast_fields: Vec<FieldMetadata> = self
-            .fast_fields()
-            .columnar()
-            .iter_columns()?
-            .map(|(mut field_name, handle)| {
-                json_path_sep_to_dot(&mut field_name);
-                // map to canonical path, to avoid similar but different entries.
-                // Eventually we should just accept '.' seperated for all cases.
-                let field_name = map_to_canonical
-                    .get(&field_name)
-                    .unwrap_or(&field_name)
-                    .to_string();
-                FieldMetadata {
-                    indexed: false,
-                    stored: false,
-                    field_name,
-                    fast: true,
-                    typ: Type::from(handle.column_type()),
-                }
-            })
-            .collect();
-        // Since the type is encoded differently in the fast field and in the inverted index,
-        // the order of the fields is not guaranteed to be the same. Therefore, we sort the fields.
-        // If we are sure that the order is the same, we can remove this sort.
-        indexed_fields.sort_unstable();
-        fast_fields.sort_unstable();
-        let merged = merge_field_meta_data(vec![indexed_fields, fast_fields], &self.schema);
-
-        Ok(merged)
     }
 
     /// Returns the segment id
@@ -515,9 +455,9 @@ impl fmt::Debug for SegmentReader {
 #[cfg(test)]
 mod test {
     use super::*;
-    use crate::index::Index;
+    use crate::core::Index;
     use crate::schema::{Schema, SchemaBuilder, Term, STORED, TEXT};
-    use crate::{DocId, IndexWriter};
+    use crate::{DocId, FieldMetadata, IndexWriter};
 
     #[test]
     fn test_merge_field_meta_data_same() {

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()

src/directory/mmap_directory.rs

@@ -479,7 +479,6 @@ impl Directory for MmapDirectory {
         let file: File = OpenOptions::new()
             .write(true)
             .create(true) //< if the file does not exist yet, create it.
-            .truncate(false)
             .open(full_path)
             .map_err(LockError::wrap_io_error)?;
         if lock.is_blocking {
@@ -674,7 +673,7 @@ mod tests {
             let num_segments = reader.searcher().segment_readers().len();
             assert!(num_segments <= 4);
             let num_components_except_deletes_and_tempstore =
-                crate::index::SegmentComponent::iterator().len() - 2;
+                crate::core::SegmentComponent::iterator().len() - 2;
             let max_num_mmapped = num_components_except_deletes_and_tempstore * num_segments;
             assert_eventually(|| {
                 let num_mmapped = mmap_directory.get_cache_info().mmapped.len();

src/directory/ram_directory.rs

@@ -85,7 +85,7 @@ impl InnerDirectory {
         self.fs
             .get(path)
             .ok_or_else(|| OpenReadError::FileDoesNotExist(PathBuf::from(path)))
-            .cloned()
+            .map(Clone::clone)
     }
 
     fn delete(&mut self, path: &Path) -> result::Result<(), DeleteError> {

src/fastfield/writer.rs

@@ -238,13 +238,17 @@ impl FastFieldsWriter {
         mut self,
         wrt: &mut dyn io::Write,
         doc_id_map_opt: Option<&DocIdMapping>,
-    ) -> io::Result<()> {
+    ) -> io::Result<Vec<(String, Type)>> {
         let num_docs = self.num_docs;
         let old_to_new_row_ids =
             doc_id_map_opt.map(|doc_id_mapping| doc_id_mapping.old_to_new_ids());
-        self.columnar_writer
+        let columns = self
+            .columnar_writer
             .serialize(num_docs, old_to_new_row_ids, wrt)?;
-        Ok(())
+        Ok(columns
+            .into_iter()
+            .map(|(field_name, column)| (field_name.to_string(), column.into()))
+            .collect())
     }
 }
 

src/field_list/mod.rs

@@ -0,0 +1,369 @@
+//! The list of fields that are stored in a `tantivy` `Index`.
+
+use std::collections::HashSet;
+use std::io::{self, ErrorKind, Read};
+
+use columnar::ColumnType;
+use common::TinySet;
+use fnv::FnvHashMap;
+
+use crate::indexer::path_to_unordered_id::OrderedPathId;
+use crate::json_utils::json_path_sep_to_dot;
+use crate::postings::IndexingContext;
+use crate::schema::{Field, Schema, Type};
+use crate::{merge_field_meta_data, FieldMetadata, Term};
+
+#[derive(Debug, PartialEq, Eq, Clone, Copy, Hash)]
+pub(crate) struct FieldConfig {
+    pub typ: Type,
+    pub indexed: bool,
+    pub stored: bool,
+    pub fast: bool,
+}
+
+impl FieldConfig {
+    fn serialize(&self) -> [u8; 2] {
+        let typ = self.typ.to_code();
+        let flags = (self.indexed as u8) << 2 | (self.stored as u8) << 1 | (self.fast as u8);
+        [typ, flags]
+    }
+    fn deserialize_from(data: [u8; 2]) -> io::Result<FieldConfig> {
+        let typ = Type::from_code(data[0]).ok_or_else(|| {
+            io::Error::new(
+                ErrorKind::InvalidData,
+                format!("could not deserialize type {}", data[0]),
+            )
+        })?;
+
+        let data = data[1];
+        let indexed = (data & 0b100) != 0;
+        let stored = (data & 0b010) != 0;
+        let fast = (data & 0b001) != 0;
+
+        Ok(FieldConfig {
+            typ,
+            indexed,
+            stored,
+            fast,
+        })
+    }
+}
+/// Serializes the split fields.
+pub(crate) fn serialize_segment_fields(
+    ctx: IndexingContext,
+    wrt: &mut dyn io::Write,
+    schema: &Schema,
+    unordered_id_to_ordered_id: &[(OrderedPathId, TinySet)],
+    mut columns: Vec<(String, Type)>,
+) -> crate::Result<()> {
+    let mut field_list_set: HashSet<(Field, OrderedPathId, TinySet)> = HashSet::default();
+    let mut encoded_fields = Vec::new();
+    let mut map_to_canonical = FnvHashMap::default();
+
+    // Replace unordered ids by ordered ids to be able to sort
+    let ordered_id_to_path = ctx.path_to_unordered_id.ordered_id_to_path();
+
+    for (key, _addr) in ctx.term_index.iter() {
+        let field = Term::wrap(key).field();
+        let field_entry = schema.get_field_entry(field);
+        if field_entry.field_type().value_type() == Type::Json {
+            let byte_range_unordered_id = 5..5 + 4;
+            let unordered_id =
+                u32::from_be_bytes(key[byte_range_unordered_id.clone()].try_into().unwrap());
+            let (path_id, typ_code_bitvec) = unordered_id_to_ordered_id[unordered_id as usize];
+            if !field_list_set.contains(&(field, path_id, typ_code_bitvec)) {
+                field_list_set.insert((field, path_id, typ_code_bitvec));
+                let mut build_path = |field_name: &str, mut json_path: String| {
+                    // In this case we need to map the potential fast field to the field name
+                    // accepted by the query parser.
+                    let create_canonical =
+                        !field_entry.is_expand_dots_enabled() && json_path.contains('.');
+                    if create_canonical {
+                        // Without expand dots enabled dots need to be escaped.
+                        let escaped_json_path = json_path.replace('.', "\\.");
+                        let full_path = format!("{}.{}", field_name, escaped_json_path);
+                        let full_path_unescaped = format!("{}.{}", field_name, &json_path);
+                        map_to_canonical.insert(full_path_unescaped, full_path.to_string());
+                        full_path
+                    } else {
+                        // With expand dots enabled, we can use '.' instead of '\u{1}'.
+                        json_path_sep_to_dot(&mut json_path);
+                        format!("{}.{}", field_name, json_path)
+                    }
+                };
+
+                let path = build_path(
+                    field_entry.name(),
+                    ordered_id_to_path[path_id.path_id() as usize].to_string(), /* String::from_utf8(key[5..].to_vec()).unwrap(), */
+                );
+                encoded_fields.push((path, typ_code_bitvec));
+            }
+        }
+    }
+
+    let mut indexed_fields: Vec<FieldMetadata> = Vec::new();
+    for (_field, field_entry) in schema.fields() {
+        let field_name = field_entry.name().to_string();
+        let is_indexed = field_entry.is_indexed();
+
+        let is_json = field_entry.field_type().value_type() == Type::Json;
+        if is_indexed && !is_json {
+            indexed_fields.push(FieldMetadata {
+                indexed: true,
+                stored: false,
+                field_name: field_name.to_string(),
+                fast: false,
+                typ: field_entry.field_type().value_type(),
+            });
+        }
+    }
+    for (field_name, field_type_set) in encoded_fields {
+        for field_type in field_type_set {
+            let column_type = ColumnType::try_from_code(field_type as u8).unwrap();
+            indexed_fields.push(FieldMetadata {
+                indexed: true,
+                stored: false,
+                field_name: field_name.to_string(),
+                fast: false,
+                typ: Type::from(column_type),
+            });
+        }
+    }
+    let mut fast_fields: Vec<FieldMetadata> = columns
+        .iter_mut()
+        .map(|(field_name, typ)| {
+            json_path_sep_to_dot(field_name);
+            // map to canonical path, to avoid similar but different entries.
+            // Eventually we should just accept '.' seperated for all cases.
+            let field_name = map_to_canonical
+                .get(field_name)
+                .unwrap_or(field_name)
+                .to_string();
+            FieldMetadata {
+                indexed: false,
+                stored: false,
+                field_name,
+                fast: true,
+                typ: *typ,
+            }
+        })
+        .collect();
+    // Since the type is encoded differently in the fast field and in the inverted index,
+    // the order of the fields is not guaranteed to be the same. Therefore, we sort the fields.
+    // If we are sure that the order is the same, we can remove this sort.
+    indexed_fields.sort_unstable();
+    fast_fields.sort_unstable();
+    let merged = merge_field_meta_data(vec![indexed_fields, fast_fields], schema);
+    let out = serialize_split_fields(&merged);
+    wrt.write_all(&out)?;
+
+    Ok(())
+}
+
+/// Serializes the Split fields.
+///
+/// `fields_metadata` has to be sorted.
+pub fn serialize_split_fields(fields_metadata: &[FieldMetadata]) -> Vec<u8> {
+    // ensure that fields_metadata is strictly sorted.
+    debug_assert!(fields_metadata.windows(2).all(|w| w[0] < w[1]));
+    let mut payload = Vec::new();
+    // Write Num Fields
+    let length = fields_metadata.len() as u32;
+    payload.extend_from_slice(&length.to_le_bytes());
+
+    for field_metadata in fields_metadata {
+        write_field(field_metadata, &mut payload);
+    }
+    let compression_level = 3;
+    let payload_compressed = zstd::stream::encode_all(&mut &payload[..], compression_level)
+        .expect("zstd encoding failed");
+    let mut out = Vec::new();
+    // Write Header -- Format Version
+    let format_version = 1u8;
+    out.push(format_version);
+    // Write Payload
+    out.extend_from_slice(&payload_compressed);
+    out
+}
+
+fn write_field(field_metadata: &FieldMetadata, out: &mut Vec<u8>) {
+    let field_config = FieldConfig {
+        typ: field_metadata.typ,
+        indexed: field_metadata.indexed,
+        stored: field_metadata.stored,
+        fast: field_metadata.fast,
+    };
+
+    // Write Config 2 bytes
+    out.extend_from_slice(&field_config.serialize());
+    let str_length = field_metadata.field_name.len() as u16;
+    // Write String length 2 bytes
+    out.extend_from_slice(&str_length.to_le_bytes());
+    out.extend_from_slice(field_metadata.field_name.as_bytes());
+}
+
+/// Reads a fixed number of bytes into an array and returns the array.
+fn read_exact_array<R: Read, const N: usize>(reader: &mut R) -> io::Result<[u8; N]> {
+    let mut buffer = [0u8; N];
+    reader.read_exact(&mut buffer)?;
+    Ok(buffer)
+}
+
+/// Reads the Split fields from a zstd compressed stream of bytes
+pub fn read_split_fields<R: Read>(
+    mut reader: R,
+) -> io::Result<impl Iterator<Item = io::Result<FieldMetadata>>> {
+    let format_version = read_exact_array::<_, 1>(&mut reader)?[0];
+    assert_eq!(format_version, 1);
+    let reader = zstd::Decoder::new(reader)?;
+    read_split_fields_from_zstd(reader)
+}
+
+fn read_field<R: Read>(reader: &mut R) -> io::Result<FieldMetadata> {
+    // Read FieldConfig (2 bytes)
+    let config_bytes = read_exact_array::<_, 2>(reader)?;
+    let field_config = FieldConfig::deserialize_from(config_bytes)?; // Assuming this returns a Result
+
+    // Read field name length and the field name
+    let name_len = u16::from_le_bytes(read_exact_array::<_, 2>(reader)?) as usize;
+
+    let mut data = vec![0; name_len];
+    reader.read_exact(&mut data)?;
+
+    let field_name = String::from_utf8(data).map_err(|err| {
+        io::Error::new(
+            ErrorKind::InvalidData,
+            format!(
+                "Encountered invalid utf8 when deserializing field name: {}",
+                err
+            ),
+        )
+    })?;
+    Ok(FieldMetadata {
+        field_name,
+        typ: field_config.typ,
+        indexed: field_config.indexed,
+        stored: field_config.stored,
+        fast: field_config.fast,
+    })
+}
+
+/// Reads the Split fields from a stream of bytes
+fn read_split_fields_from_zstd<R: Read>(
+    mut reader: R,
+) -> io::Result<impl Iterator<Item = io::Result<FieldMetadata>>> {
+    let mut num_fields = u32::from_le_bytes(read_exact_array::<_, 4>(&mut reader)?);
+
+    Ok(std::iter::from_fn(move || {
+        if num_fields == 0 {
+            return None;
+        }
+        num_fields -= 1;
+
+        Some(read_field(&mut reader))
+    }))
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn field_config_deser_test() {
+        let field_config = FieldConfig {
+            typ: Type::Str,
+            indexed: true,
+            stored: false,
+            fast: true,
+        };
+        let serialized = field_config.serialize();
+        let deserialized = FieldConfig::deserialize_from(serialized).unwrap();
+        assert_eq!(field_config, deserialized);
+    }
+    #[test]
+    fn write_read_field_test() {
+        for typ in Type::iter_values() {
+            let field_metadata = FieldMetadata {
+                field_name: "test".to_string(),
+                typ,
+                indexed: true,
+                stored: true,
+                fast: true,
+            };
+            let mut out = Vec::new();
+            write_field(&field_metadata, &mut out);
+            let deserialized = read_field(&mut &out[..]).unwrap();
+            assert_eq!(field_metadata, deserialized);
+        }
+        let field_metadata = FieldMetadata {
+            field_name: "test".to_string(),
+            typ: Type::Str,
+            indexed: false,
+            stored: true,
+            fast: true,
+        };
+        let mut out = Vec::new();
+        write_field(&field_metadata, &mut out);
+        let deserialized = read_field(&mut &out[..]).unwrap();
+        assert_eq!(field_metadata, deserialized);
+
+        let field_metadata = FieldMetadata {
+            field_name: "test".to_string(),
+            typ: Type::Str,
+            indexed: false,
+            stored: false,
+            fast: true,
+        };
+        let mut out = Vec::new();
+        write_field(&field_metadata, &mut out);
+        let deserialized = read_field(&mut &out[..]).unwrap();
+        assert_eq!(field_metadata, deserialized);
+
+        let field_metadata = FieldMetadata {
+            field_name: "test".to_string(),
+            typ: Type::Str,
+            indexed: true,
+            stored: false,
+            fast: false,
+        };
+        let mut out = Vec::new();
+        write_field(&field_metadata, &mut out);
+        let deserialized = read_field(&mut &out[..]).unwrap();
+        assert_eq!(field_metadata, deserialized);
+    }
+    #[test]
+    fn write_split_fields_test() {
+        let fields_metadata = vec![
+            FieldMetadata {
+                field_name: "test".to_string(),
+                typ: Type::Str,
+                indexed: true,
+                stored: true,
+                fast: true,
+            },
+            FieldMetadata {
+                field_name: "test2".to_string(),
+                typ: Type::Str,
+                indexed: true,
+                stored: false,
+                fast: false,
+            },
+            FieldMetadata {
+                field_name: "test3".to_string(),
+                typ: Type::U64,
+                indexed: true,
+                stored: false,
+                fast: true,
+            },
+        ];
+
+        let out = serialize_split_fields(&fields_metadata);
+
+        let deserialized: Vec<FieldMetadata> = read_split_fields(&mut &out[..])
+            .unwrap()
+            .map(|el| el.unwrap())
+            .collect();
+
+        assert_eq!(fields_metadata, deserialized);
+    }
+}

src/index/mod.rs

@@ -1,22 +0,0 @@
-//! # Index Module
-//!
-//! The `index` module in Tantivy contains core components to read and write indexes.
-//!
-//! It contains `Index` and `Segment`, where a `Index` consists of one or more `Segment`s.
-
-mod index;
-mod index_meta;
-mod inverted_index_reader;
-mod segment;
-mod segment_component;
-mod segment_id;
-mod segment_reader;
-
-pub use self::index::{Index, IndexBuilder};
-pub(crate) use self::index_meta::SegmentMetaInventory;
-pub use self::index_meta::{IndexMeta, IndexSettings, IndexSortByField, Order, SegmentMeta};
-pub use self::inverted_index_reader::InvertedIndexReader;
-pub use self::segment::Segment;
-pub use self::segment_component::SegmentComponent;
-pub use self::segment_id::SegmentId;
-pub use self::segment_reader::{FieldMetadata, SegmentReader};

src/indexer/index_writer.rs

@@ -9,10 +9,10 @@ use smallvec::smallvec;
 use super::operation::{AddOperation, UserOperation};
 use super::segment_updater::SegmentUpdater;
 use super::{AddBatch, AddBatchReceiver, AddBatchSender, PreparedCommit};
+use crate::core::{Index, Segment, SegmentComponent, SegmentId, SegmentMeta, SegmentReader};
 use crate::directory::{DirectoryLock, GarbageCollectionResult, TerminatingWrite};
 use crate::error::TantivyError;
 use crate::fastfield::write_alive_bitset;
-use crate::index::{Index, Segment, SegmentComponent, SegmentId, SegmentMeta, SegmentReader};
 use crate::indexer::delete_queue::{DeleteCursor, DeleteQueue};
 use crate::indexer::doc_opstamp_mapping::DocToOpstampMapping;
 use crate::indexer::index_writer_status::IndexWriterStatus;
@@ -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/log_merge_policy.rs

@@ -3,7 +3,7 @@ use std::cmp;
 use itertools::Itertools;
 
 use super::merge_policy::{MergeCandidate, MergePolicy};
-use crate::index::SegmentMeta;
+use crate::core::SegmentMeta;
 
 const DEFAULT_LEVEL_LOG_SIZE: f64 = 0.75;
 const DEFAULT_MIN_LAYER_SIZE: u32 = 10_000;
@@ -144,7 +144,7 @@ mod tests {
     use once_cell::sync::Lazy;
 
     use super::*;
-    use crate::index::{SegmentId, SegmentMeta, SegmentMetaInventory};
+    use crate::core::{SegmentId, SegmentMeta, SegmentMetaInventory};
     use crate::indexer::merge_policy::MergePolicy;
     use crate::schema;
     use crate::schema::INDEXED;

src/indexer/merge_policy.rs

@@ -1,7 +1,7 @@
 use std::fmt::Debug;
 use std::marker;
 
-use crate::index::{SegmentId, SegmentMeta};
+use crate::core::{SegmentId, SegmentMeta};
 
 /// Set of segment suggested for a merge.
 #[derive(Debug, Clone)]
@@ -39,7 +39,7 @@ impl MergePolicy for NoMergePolicy {
 pub mod tests {
 
     use super::*;
-    use crate::index::{SegmentId, SegmentMeta};
+    use crate::core::{SegmentId, SegmentMeta};
 
     /// `MergePolicy` useful for test purposes.
     ///

src/indexer/merger.rs

@@ -1,3 +1,4 @@
+use std::io::Write;
 use std::sync::Arc;
 
 use columnar::{
@@ -8,12 +9,13 @@ use common::ReadOnlyBitSet;
 use itertools::Itertools;
 use measure_time::debug_time;
 
+use crate::core::{Segment, SegmentReader};
 use crate::directory::WritePtr;
 use crate::docset::{DocSet, TERMINATED};
 use crate::error::DataCorruption;
 use crate::fastfield::{AliveBitSet, FastFieldNotAvailableError};
+use crate::field_list::serialize_split_fields;
 use crate::fieldnorm::{FieldNormReader, FieldNormReaders, FieldNormsSerializer, FieldNormsWriter};
-use crate::index::{Segment, SegmentReader};
 use crate::indexer::doc_id_mapping::{MappingType, SegmentDocIdMapping};
 use crate::indexer::SegmentSerializer;
 use crate::postings::{InvertedIndexSerializer, Postings, SegmentPostings};
@@ -21,8 +23,8 @@ use crate::schema::{value_type_to_column_type, Field, FieldType, Schema};
 use crate::store::StoreWriter;
 use crate::termdict::{TermMerger, TermOrdinal};
 use crate::{
-    DocAddress, DocId, IndexSettings, IndexSortByField, InvertedIndexReader, Order,
-    SegmentComponent, SegmentOrdinal,
+    merge_field_meta_data, DocAddress, DocId, FieldMetadata, IndexSettings, IndexSortByField,
+    InvertedIndexReader, Order, SegmentComponent, SegmentOrdinal,
 };
 
 /// Segment's max doc must be `< MAX_DOC_LIMIT`.
@@ -255,6 +257,19 @@ impl IndexMerger {
         Ok(())
     }
 
+    fn write_field_list(&self, list_field_wrt: &mut WritePtr) -> crate::Result<()> {
+        let field_metadatas: Vec<Vec<FieldMetadata>> = self
+            .readers
+            .iter()
+            .map(|reader| reader.fields_metadata())
+            .collect::<crate::Result<Vec<_>>>()?;
+        let merged = merge_field_meta_data(field_metadatas, &self.schema);
+        let out = serialize_split_fields(&merged);
+        list_field_wrt.write_all(&out)?;
+
+        Ok(())
+    }
+
     fn write_fast_fields(
         &self,
         fast_field_wrt: &mut WritePtr,
@@ -605,10 +620,6 @@ 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
                         };
 
@@ -777,6 +788,7 @@ impl IndexMerger {
         self.write_storable_fields(serializer.get_store_writer(), &doc_id_mapping)?;
         debug!("write-fastfields");
         self.write_fast_fields(serializer.get_fast_field_write(), doc_id_mapping)?;
+        self.write_field_list(serializer.get_field_list_write())?;
 
         debug!("close-serializer");
         serializer.close()?;
@@ -794,7 +806,7 @@ mod tests {
         BytesFastFieldTestCollector, FastFieldTestCollector, TEST_COLLECTOR_WITH_SCORE,
     };
     use crate::collector::{Count, FacetCollector};
-    use crate::index::Index;
+    use crate::core::Index;
     use crate::query::{AllQuery, BooleanQuery, EnableScoring, Scorer, TermQuery};
     use crate::schema::document::Value;
     use crate::schema::{

src/indexer/merger_sorted_index_test.rs

@@ -1,8 +1,8 @@
 #[cfg(test)]
 mod tests {
     use crate::collector::TopDocs;
+    use crate::core::Index;
     use crate::fastfield::AliveBitSet;
-    use crate::index::Index;
     use crate::query::QueryParser;
     use crate::schema::document::Value;
     use crate::schema::{
@@ -485,7 +485,7 @@ mod bench_sorted_index_merge {
 
     use test::{self, Bencher};
 
-    use crate::index::Index;
+    use crate::core::Index;
     use crate::indexer::merger::IndexMerger;
     use crate::schema::{NumericOptions, Schema};
     use crate::{IndexSettings, IndexSortByField, IndexWriter, Order};

src/indexer/mod.rs

@@ -25,7 +25,6 @@ mod segment_register;
 pub(crate) mod segment_serializer;
 pub(crate) mod segment_updater;
 pub(crate) mod segment_writer;
-pub(crate) mod single_segment_index_writer;
 mod stamper;
 
 use crossbeam_channel as channel;
@@ -35,14 +34,13 @@ pub use self::index_writer::IndexWriter;
 pub use self::log_merge_policy::LogMergePolicy;
 pub use self::merge_operation::MergeOperation;
 pub use self::merge_policy::{MergeCandidate, MergePolicy, NoMergePolicy};
-use self::operation::AddOperation;
 pub use self::operation::UserOperation;
 pub use self::prepared_commit::PreparedCommit;
 pub use self::segment_entry::SegmentEntry;
 pub(crate) use self::segment_serializer::SegmentSerializer;
 pub use self::segment_updater::{merge_filtered_segments, merge_indices};
 pub use self::segment_writer::SegmentWriter;
-pub use self::single_segment_index_writer::SingleSegmentIndexWriter;
+use crate::indexer::operation::AddOperation;
 
 /// Alias for the default merge policy, which is the `LogMergePolicy`.
 pub type DefaultMergePolicy = LogMergePolicy;
@@ -65,10 +63,9 @@ mod tests_mmap {
     use crate::aggregation::agg_result::AggregationResults;
     use crate::aggregation::AggregationCollector;
     use crate::collector::{Count, TopDocs};
-    use crate::index::FieldMetadata;
     use crate::query::{AllQuery, QueryParser};
     use crate::schema::{JsonObjectOptions, Schema, Type, FAST, INDEXED, STORED, TEXT};
-    use crate::{Index, IndexWriter, Term};
+    use crate::{FieldMetadata, Index, IndexWriter, Term};
 
     #[test]
     fn test_advance_delete_bug() -> crate::Result<()> {
@@ -406,10 +403,11 @@ mod tests_mmap {
 
         let searcher = reader.searcher();
 
-        let fields_and_vals = [
+        let fields_and_vals = vec![
+            // Only way to address or it gets shadowed by `json.shadow` field
             ("json.shadow\u{1}val".to_string(), "a"), // Succeeds
             //("json.shadow.val".to_string(), "a"),   // Fails
-            ("json.shadow.val".to_string(), "b"),
+            ("json.shadow.val".to_string(), "b"), // Succeeds
         ];
 
         let query_parser = QueryParser::for_index(&index, vec![]);

src/indexer/path_to_unordered_id.rs

@@ -1,3 +1,5 @@
+use columnar::ColumnType;
+use common::TinySet;
 use fnv::FnvHashMap;
 
 /// `Field` is represented by an unsigned 32-bit integer type.
@@ -24,34 +26,44 @@ impl From<u32> for OrderedPathId {
 
 #[derive(Default)]
 pub(crate) struct PathToUnorderedId {
-    map: FnvHashMap<String, u32>,
+    /// TinySet contains the type codes of the columns in the path.
+    map: FnvHashMap<String, (u32, TinySet)>,
 }
 
 impl PathToUnorderedId {
     #[inline]
-    pub(crate) fn get_or_allocate_unordered_id(&mut self, path: &str) -> u32 {
-        if let Some(id) = self.map.get(path) {
+    pub(crate) fn get_or_allocate_unordered_id(&mut self, path: &str, typ: ColumnType) -> u32 {
+        let code_bit = typ.to_code();
+        if let Some((id, all_codes)) = self.map.get_mut(path) {
+            *all_codes = all_codes.insert(code_bit as u32);
             return *id;
         }
-        self.insert_new_path(path)
+        self.insert_new_path(path, code_bit)
     }
     #[cold]
-    fn insert_new_path(&mut self, path: &str) -> u32 {
+    fn insert_new_path(&mut self, path: &str, typ_code: u8) -> u32 {
         let next_id = self.map.len() as u32;
-        self.map.insert(path.to_string(), next_id);
+        self.map.insert(
+            path.to_string(),
+            (next_id, TinySet::singleton(typ_code as u32)),
+        );
         next_id
     }
 
     /// Retuns ids which reflect the lexical order of the paths.
     ///
     /// The returned vec can be indexed with the unordered id to get the ordered id.
-    pub(crate) fn unordered_id_to_ordered_id(&self) -> Vec<OrderedPathId> {
-        let mut sorted_ids: Vec<(&str, &u32)> =
-            self.map.iter().map(|(k, v)| (k.as_str(), v)).collect();
+    pub(crate) fn unordered_id_to_ordered_id(&self) -> Vec<(OrderedPathId, TinySet)> {
+        let mut sorted_ids: Vec<(&str, (u32, TinySet))> = self
+            .map
+            .iter()
+            .map(|(k, (id, typ_code))| (k.as_str(), (*id, *typ_code)))
+            .collect();
         sorted_ids.sort_unstable_by_key(|(path, _)| *path);
-        let mut result = vec![OrderedPathId::default(); sorted_ids.len()];
-        for (ordered, unordered) in sorted_ids.iter().map(|(_k, v)| v).enumerate() {
-            result[**unordered as usize] = OrderedPathId::from_ordered_id(ordered as u32);
+        let mut result = vec![(OrderedPathId::default(), TinySet::empty()); sorted_ids.len()];
+        for (ordered, (unordered, typ_code)) in sorted_ids.iter().map(|(_k, v)| v).enumerate() {
+            result[*unordered as usize] =
+                (OrderedPathId::from_ordered_id(ordered as u32), *typ_code);
         }
         result
     }
@@ -74,12 +86,12 @@ mod tests {
         let terms = vec!["b", "a", "b", "c"];
         let ids = terms
             .iter()
-            .map(|term| path_to_id.get_or_allocate_unordered_id(term))
+            .map(|term| path_to_id.get_or_allocate_unordered_id(term, ColumnType::Str))
             .collect::<Vec<u32>>();
         assert_eq!(ids, vec![0, 1, 0, 2]);
         let ordered_ids = ids
             .iter()
-            .map(|id| path_to_id.unordered_id_to_ordered_id()[*id as usize])
+            .map(|id| path_to_id.unordered_id_to_ordered_id()[*id as usize].0)
             .collect::<Vec<OrderedPathId>>();
         assert_eq!(ordered_ids, vec![1.into(), 0.into(), 1.into(), 2.into()]);
         // Fetch terms

src/indexer/segment_entry.rs

@@ -2,7 +2,7 @@ use std::fmt;
 
 use common::BitSet;
 
-use crate::index::{SegmentId, SegmentMeta};
+use crate::core::{SegmentId, SegmentMeta};
 use crate::indexer::delete_queue::DeleteCursor;
 
 /// A segment entry describes the state of

src/indexer/segment_manager.rs

@@ -3,8 +3,8 @@ use std::fmt::{self, Debug, Formatter};
 use std::sync::{RwLock, RwLockReadGuard, RwLockWriteGuard};
 
 use super::segment_register::SegmentRegister;
+use crate::core::{SegmentId, SegmentMeta};
 use crate::error::TantivyError;
-use crate::index::{SegmentId, SegmentMeta};
 use crate::indexer::delete_queue::DeleteCursor;
 use crate::indexer::SegmentEntry;
 

src/indexer/segment_register.rs

@@ -1,7 +1,7 @@
 use std::collections::{HashMap, HashSet};
 use std::fmt::{self, Debug, Display, Formatter};
 
-use crate::index::{SegmentId, SegmentMeta};
+use crate::core::{SegmentId, SegmentMeta};
 use crate::indexer::delete_queue::DeleteCursor;
 use crate::indexer::segment_entry::SegmentEntry;
 
@@ -103,7 +103,7 @@ impl SegmentRegister {
 #[cfg(test)]
 mod tests {
     use super::*;
-    use crate::index::{SegmentId, SegmentMetaInventory};
+    use crate::core::{SegmentId, SegmentMetaInventory};
     use crate::indexer::delete_queue::*;
 
     fn segment_ids(segment_register: &SegmentRegister) -> Vec<SegmentId> {

src/indexer/segment_serializer.rs

@@ -1,8 +1,8 @@
 use common::TerminatingWrite;
 
+use crate::core::{Segment, SegmentComponent};
 use crate::directory::WritePtr;
 use crate::fieldnorm::FieldNormsSerializer;
-use crate::index::{Segment, SegmentComponent};
 use crate::postings::InvertedIndexSerializer;
 use crate::store::StoreWriter;
 
@@ -12,6 +12,7 @@ pub struct SegmentSerializer {
     segment: Segment,
     pub(crate) store_writer: StoreWriter,
     fast_field_write: WritePtr,
+    field_list_write: WritePtr,
     fieldnorms_serializer: Option<FieldNormsSerializer>,
     postings_serializer: InvertedIndexSerializer,
 }
@@ -49,6 +50,7 @@ impl SegmentSerializer {
         };
 
         let fast_field_write = segment.open_write(SegmentComponent::FastFields)?;
+        let field_list_write = segment.open_write(SegmentComponent::FieldList)?;
 
         let fieldnorms_write = segment.open_write(SegmentComponent::FieldNorms)?;
         let fieldnorms_serializer = FieldNormsSerializer::from_write(fieldnorms_write)?;
@@ -58,6 +60,7 @@ impl SegmentSerializer {
             segment,
             store_writer,
             fast_field_write,
+            field_list_write,
             fieldnorms_serializer: Some(fieldnorms_serializer),
             postings_serializer,
         })
@@ -81,6 +84,11 @@ impl SegmentSerializer {
         &mut self.postings_serializer
     }
 
+    /// Accessor to the ``.
+    pub fn get_field_list_write(&mut self) -> &mut WritePtr {
+        &mut self.field_list_write
+    }
+
     /// Accessor to the `FastFieldSerializer`.
     pub fn get_fast_field_write(&mut self) -> &mut WritePtr {
         &mut self.fast_field_write
@@ -104,6 +112,7 @@ impl SegmentSerializer {
             fieldnorms_serializer.close()?;
         }
         self.fast_field_write.terminate()?;
+        self.field_list_write.terminate()?;
         self.postings_serializer.close()?;
         self.store_writer.close()?;
         Ok(())

src/indexer/segment_updater.rs

@@ -9,10 +9,11 @@ use std::sync::{Arc, RwLock};
 use rayon::{ThreadPool, ThreadPoolBuilder};
 
 use super::segment_manager::SegmentManager;
-use crate::core::META_FILEPATH;
+use crate::core::{
+    Index, IndexMeta, IndexSettings, Segment, SegmentId, SegmentMeta, META_FILEPATH,
+};
 use crate::directory::{Directory, DirectoryClone, GarbageCollectionResult};
 use crate::fastfield::AliveBitSet;
-use crate::index::{Index, IndexMeta, IndexSettings, Segment, SegmentId, SegmentMeta};
 use crate::indexer::delete_queue::DeleteCursor;
 use crate::indexer::index_writer::advance_deletes;
 use crate::indexer::merge_operation::MergeOperationInventory;

src/indexer/segment_writer.rs

@@ -6,9 +6,10 @@ use tokenizer_api::BoxTokenStream;
 use super::doc_id_mapping::{get_doc_id_mapping_from_field, DocIdMapping};
 use super::operation::AddOperation;
 use crate::core::json_utils::index_json_values;
+use crate::core::Segment;
 use crate::fastfield::FastFieldsWriter;
+use crate::field_list::serialize_segment_fields;
 use crate::fieldnorm::{FieldNormReaders, FieldNormsWriter};
-use crate::index::Segment;
 use crate::indexer::segment_serializer::SegmentSerializer;
 use crate::postings::{
     compute_table_memory_size, serialize_postings, IndexingContext, IndexingPosition,
@@ -443,16 +444,29 @@ fn remap_and_write(
         .segment()
         .open_read(SegmentComponent::FieldNorms)?;
     let fieldnorm_readers = FieldNormReaders::open(fieldnorm_data)?;
+
+    let unordered_id_to_ordered_id = ctx.path_to_unordered_id.unordered_id_to_ordered_id();
+
     serialize_postings(
-        ctx,
-        schema,
+        &ctx,
+        schema.clone(),
         per_field_postings_writers,
         fieldnorm_readers,
         doc_id_map,
+        &unordered_id_to_ordered_id,
         serializer.get_postings_serializer(),
     )?;
     debug!("fastfield-serialize");
-    fast_field_writers.serialize(serializer.get_fast_field_write(), doc_id_map)?;
+    let columns = fast_field_writers.serialize(serializer.get_fast_field_write(), doc_id_map)?;
+
+    let field_list_serializer = serializer.get_field_list_write();
+    serialize_segment_fields(
+        ctx,
+        field_list_serializer,
+        &schema,
+        &unordered_id_to_ordered_id,
+        columns,
+    )?;
 
     // finalize temp docstore and create version, which reflects the doc_id_map
     if let Some(doc_id_map) = doc_id_map {
@@ -879,31 +893,6 @@ 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/lib.rs

@@ -188,8 +188,8 @@ pub mod aggregation;
 pub mod collector;
 pub mod directory;
 pub mod fastfield;
+pub mod field_list;
 pub mod fieldnorm;
-pub mod index;
 pub mod positions;
 pub mod postings;
 
@@ -221,25 +221,27 @@ pub use self::docset::{DocSet, TERMINATED};
 pub use self::snippet::{Snippet, SnippetGenerator};
 #[doc(hidden)]
 pub use crate::core::json_utils;
-pub use crate::core::{Executor, Searcher, SearcherGeneration};
-pub use crate::directory::Directory;
-pub use crate::index::{
-    Index, IndexBuilder, IndexMeta, IndexSettings, IndexSortByField, InvertedIndexReader, Order,
-    Segment, SegmentComponent, SegmentId, SegmentMeta, SegmentReader,
+pub use crate::core::{
+    merge_field_meta_data, Executor, FieldMetadata, Index, IndexBuilder, IndexMeta, IndexSettings,
+    IndexSortByField, InvertedIndexReader, Order, Searcher, SearcherGeneration, Segment,
+    SegmentComponent, SegmentId, SegmentMeta, SegmentReader, SingleSegmentIndexWriter,
 };
+pub use crate::directory::Directory;
+pub use crate::indexer::IndexWriter;
 #[deprecated(
     since = "0.22.0",
     note = "Will be removed in tantivy 0.23. Use export from indexer module instead"
 )]
-pub use crate::indexer::PreparedCommit;
-pub use crate::indexer::{IndexWriter, SingleSegmentIndexWriter};
+pub use crate::indexer::{merge_filtered_segments, merge_indices, PreparedCommit};
 pub use crate::postings::Postings;
 #[allow(deprecated)]
 pub use crate::schema::DatePrecision;
 pub use crate::schema::{DateOptions, DateTimePrecision, Document, TantivyDocument, Term};
 
 /// Index format version.
-const INDEX_FORMAT_VERSION: u32 = 6;
+///
+/// Version 7: Add `.fieldlist` file containing the list of fields in a segment.
+const INDEX_FORMAT_VERSION: u32 = 7;
 /// Oldest index format version this tantivy version can read.
 const INDEX_FORMAT_OLDEST_SUPPORTED_VERSION: u32 = 4;
 
@@ -339,7 +341,7 @@ impl DocAddress {
 ///
 /// The id used for the segment is actually an ordinal
 /// in the list of `Segment`s held by a `Searcher`.
-#[derive(Debug, Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Hash, Serialize, Deserialize)]
+#[derive(Debug, Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Hash)]
 pub struct DocAddress {
     /// The segment ordinal id that identifies the segment
     /// hosting the document in the `Searcher` it is called from.
@@ -387,8 +389,8 @@ pub mod tests {
     use time::OffsetDateTime;
 
     use crate::collector::tests::TEST_COLLECTOR_WITH_SCORE;
+    use crate::core::SegmentReader;
     use crate::docset::{DocSet, TERMINATED};
-    use crate::index::SegmentReader;
     use crate::merge_policy::NoMergePolicy;
     use crate::query::BooleanQuery;
     use crate::schema::document::Value;

src/postings/block_segment_postings.rs

@@ -383,8 +383,8 @@ mod tests {
     use common::HasLen;
 
     use super::BlockSegmentPostings;
+    use crate::core::Index;
     use crate::docset::{DocSet, TERMINATED};
-    use crate::index::Index;
     use crate::postings::compression::COMPRESSION_BLOCK_SIZE;
     use crate::postings::postings::Postings;
     use crate::postings::SegmentPostings;

src/postings/mod.rs

@@ -42,9 +42,9 @@ pub mod tests {
     use std::mem;
 
     use super::{InvertedIndexSerializer, Postings};
+    use crate::core::{Index, SegmentComponent, SegmentReader};
     use crate::docset::{DocSet, TERMINATED};
     use crate::fieldnorm::FieldNormReader;
-    use crate::index::{Index, SegmentComponent, SegmentReader};
     use crate::indexer::operation::AddOperation;
     use crate::indexer::SegmentWriter;
     use crate::query::Scorer;

src/postings/postings_writer.rs

@@ -2,6 +2,7 @@ use std::io;
 use std::marker::PhantomData;
 use std::ops::Range;
 
+use common::TinySet;
 use stacker::Addr;
 
 use crate::fieldnorm::FieldNormReaders;
@@ -46,37 +47,38 @@ fn make_field_partition(
 /// It pushes all term, one field at a time, towards the
 /// postings serializer.
 pub(crate) fn serialize_postings(
-    ctx: IndexingContext,
+    ctx: &IndexingContext,
     schema: Schema,
     per_field_postings_writers: &PerFieldPostingsWriter,
     fieldnorm_readers: FieldNormReaders,
     doc_id_map: Option<&DocIdMapping>,
+    unordered_id_to_ordered_id: &[(OrderedPathId, TinySet)],
     serializer: &mut InvertedIndexSerializer,
 ) -> crate::Result<()> {
     // Replace unordered ids by ordered ids to be able to sort
-    let unordered_id_to_ordered_id: Vec<OrderedPathId> =
-        ctx.path_to_unordered_id.unordered_id_to_ordered_id();
+    let ordered_id_to_path = ctx.path_to_unordered_id.ordered_id_to_path();
 
     let mut term_offsets: Vec<(Field, OrderedPathId, &[u8], Addr)> =
         Vec::with_capacity(ctx.term_index.len());
-    term_offsets.extend(ctx.term_index.iter().map(|(key, addr)| {
+    for (key, addr) in ctx.term_index.iter() {
         let field = Term::wrap(key).field();
-        if schema.get_field_entry(field).field_type().value_type() == Type::Json {
-            let byte_range_path = 5..5 + 4;
-            let unordered_id = u32::from_be_bytes(key[byte_range_path.clone()].try_into().unwrap());
-            let path_id = unordered_id_to_ordered_id[unordered_id as usize];
-            (field, path_id, &key[byte_range_path.end..], addr)
+        let field_entry = schema.get_field_entry(field);
+        if field_entry.field_type().value_type() == Type::Json {
+            let byte_range_unordered_id = 5..5 + 4;
+            let unordered_id =
+                u32::from_be_bytes(key[byte_range_unordered_id.clone()].try_into().unwrap());
+            let (path_id, _typ_code_bitvec) = unordered_id_to_ordered_id[unordered_id as usize];
+            term_offsets.push((field, path_id, &key[byte_range_unordered_id.end..], addr));
         } else {
-            (field, 0.into(), &key[5..], addr)
+            term_offsets.push((field, 0.into(), &key[5..], addr));
         }
-    }));
+    }
     // Sort by field, path, and term
     term_offsets.sort_unstable_by(
         |(field1, path_id1, bytes1, _), (field2, path_id2, bytes2, _)| {
             (field1, path_id1, bytes1).cmp(&(field2, path_id2, bytes2))
         },
     );
-    let ordered_id_to_path = ctx.path_to_unordered_id.ordered_id_to_path();
     let field_offsets = make_field_partition(&term_offsets);
     for (field, byte_offsets) in field_offsets {
         let postings_writer = per_field_postings_writers.get_for_field(field);
@@ -87,7 +89,7 @@ pub(crate) fn serialize_postings(
             &term_offsets[byte_offsets],
             &ordered_id_to_path,
             doc_id_map,
-            &ctx,
+            ctx,
             &mut field_serializer,
         )?;
         field_serializer.close()?;

src/postings/serializer.rs

@@ -4,9 +4,9 @@ use std::io::{self, Write};
 use common::{BinarySerializable, CountingWriter, VInt};
 
 use super::TermInfo;
+use crate::core::Segment;
 use crate::directory::{CompositeWrite, WritePtr};
 use crate::fieldnorm::FieldNormReader;
-use crate::index::Segment;
 use crate::positions::PositionSerializer;
 use crate::postings::compression::{BlockEncoder, VIntEncoder, COMPRESSION_BLOCK_SIZE};
 use crate::postings::skip::SkipSerializer;

src/query/all_query.rs

@@ -1,5 +1,5 @@
+use crate::core::SegmentReader;
 use crate::docset::{DocSet, BUFFER_LEN, TERMINATED};
-use crate::index::SegmentReader;
 use crate::query::boost_query::BoostScorer;
 use crate::query::explanation::does_not_match;
 use crate::query::{EnableScoring, Explanation, Query, Scorer, Weight};

src/query/automaton_weight.rs

@@ -5,7 +5,7 @@ use common::BitSet;
 use tantivy_fst::Automaton;
 
 use super::phrase_prefix_query::prefix_end;
-use crate::index::SegmentReader;
+use crate::core::SegmentReader;
 use crate::query::{BitSetDocSet, ConstScorer, Explanation, Scorer, Weight};
 use crate::schema::{Field, IndexRecordOption};
 use crate::termdict::{TermDictionary, TermStreamer};

src/query/boolean_query/boolean_weight.rs

@@ -1,7 +1,7 @@
 use std::collections::HashMap;
 
+use crate::core::SegmentReader;
 use crate::docset::BUFFER_LEN;
-use crate::index::SegmentReader;
 use crate::postings::FreqReadingOption;
 use crate::query::explanation::does_not_match;
 use crate::query::score_combiner::{DoNothingCombiner, ScoreCombiner};

src/query/exist_query.rs

@@ -3,8 +3,8 @@ use core::fmt::Debug;
 use columnar::{ColumnIndex, DynamicColumn};
 
 use super::{ConstScorer, EmptyScorer};
+use crate::core::SegmentReader;
 use crate::docset::{DocSet, TERMINATED};
-use crate::index::SegmentReader;
 use crate::query::explanation::does_not_match;
 use crate::query::{EnableScoring, Explanation, Query, Scorer, Weight};
 use crate::{DocId, Score, TantivyError};

src/query/phrase_prefix_query/phrase_prefix_weight.rs

@@ -1,6 +1,6 @@
 use super::{prefix_end, PhrasePrefixScorer};
+use crate::core::SegmentReader;
 use crate::fieldnorm::FieldNormReader;
-use crate::index::SegmentReader;
 use crate::postings::SegmentPostings;
 use crate::query::bm25::Bm25Weight;
 use crate::query::explanation::does_not_match;
@@ -157,8 +157,8 @@ impl Weight for PhrasePrefixWeight {
 
 #[cfg(test)]
 mod tests {
+    use crate::core::Index;
     use crate::docset::TERMINATED;
-    use crate::index::Index;
     use crate::query::{EnableScoring, PhrasePrefixQuery, Query};
     use crate::schema::{Schema, TEXT};
     use crate::{DocSet, IndexWriter, Term};

src/query/phrase_query/mod.rs

@@ -14,7 +14,7 @@ pub mod tests {
 
     use super::*;
     use crate::collector::tests::{TEST_COLLECTOR_WITHOUT_SCORE, TEST_COLLECTOR_WITH_SCORE};
-    use crate::index::Index;
+    use crate::core::Index;
     use crate::query::{EnableScoring, QueryParser, Weight};
     use crate::schema::{Schema, Term, TEXT};
     use crate::{assert_nearly_equals, DocAddress, DocId, IndexWriter, TERMINATED};

src/query/phrase_query/phrase_weight.rs

@@ -1,6 +1,6 @@
 use super::PhraseScorer;
+use crate::core::SegmentReader;
 use crate::fieldnorm::FieldNormReader;
-use crate::index::SegmentReader;
 use crate::postings::SegmentPostings;
 use crate::query::bm25::Bm25Weight;
 use crate::query::explanation::does_not_match;

src/query/query_parser/query_parser.rs

@@ -13,7 +13,7 @@ use super::logical_ast::*;
 use crate::core::json_utils::{
     convert_to_fast_value_and_get_term, set_string_and_get_terms, JsonTermWriter,
 };
-use crate::index::Index;
+use crate::core::Index;
 use crate::query::range_query::{is_type_valid_for_fastfield_range_query, RangeQuery};
 use crate::query::{
     AllQuery, BooleanQuery, BoostQuery, EmptyQuery, FuzzyTermQuery, Occur, PhrasePrefixQuery,

src/query/range_query/range_query.rs

@@ -7,8 +7,8 @@ use common::{BinarySerializable, BitSet};
 
 use super::map_bound;
 use super::range_query_u64_fastfield::FastFieldRangeWeight;
+use crate::core::SegmentReader;
 use crate::error::TantivyError;
-use crate::index::SegmentReader;
 use crate::query::explanation::does_not_match;
 use crate::query::range_query::range_query_ip_fastfield::IPFastFieldRangeWeight;
 use crate::query::range_query::{is_type_valid_for_fastfield_range_query, map_bound_res};

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/query/term_query/term_weight.rs

@@ -1,7 +1,7 @@
 use super::term_scorer::TermScorer;
+use crate::core::SegmentReader;
 use crate::docset::{DocSet, BUFFER_LEN};
 use crate::fieldnorm::FieldNormReader;
-use crate::index::SegmentReader;
 use crate::postings::SegmentPostings;
 use crate::query::bm25::Bm25Weight;
 use crate::query::explanation::does_not_match;

src/query/weight.rs

@@ -1,6 +1,6 @@
 use super::Scorer;
+use crate::core::SegmentReader;
 use crate::docset::BUFFER_LEN;
-use crate::index::SegmentReader;
 use crate::query::Explanation;
 use crate::{DocId, DocSet, Score, TERMINATED};
 

src/schema/document/de.rs

@@ -889,7 +889,7 @@ mod tests {
 
     #[test]
     fn test_array_serialize() {
-        let elements = [serde_json::Value::Null, serde_json::Value::Null];
+        let elements = vec![serde_json::Value::Null, serde_json::Value::Null];
         let result = serialize_value(ReferenceValue::Array(elements.iter()));
         let value = deserialize_value(result);
         assert_eq!(
@@ -900,7 +900,7 @@ mod tests {
             ]),
         );
 
-        let elements = [
+        let elements = vec![
             serde_json::Value::String("Hello, world".into()),
             serde_json::Value::String("Some demo".into()),
         ];
@@ -914,12 +914,12 @@ mod tests {
             ]),
         );
 
-        let elements = [];
+        let elements = vec![];
         let result = serialize_value(ReferenceValue::Array(elements.iter()));
         let value = deserialize_value(result);
         assert_eq!(value, crate::schema::OwnedValue::Array(vec![]));
 
-        let elements = [
+        let elements = vec![
             serde_json::Value::Null,
             serde_json::Value::String("Hello, world".into()),
             serde_json::Value::Number(12345.into()),

src/schema/document/se.rs

@@ -453,7 +453,7 @@ mod tests {
 
     #[test]
     fn test_array_serialize() {
-        let elements = [serde_json::Value::Null, serde_json::Value::Null];
+        let elements = vec![serde_json::Value::Null, serde_json::Value::Null];
         let result = serialize_value(ReferenceValue::Array(elements.iter()));
         let expected = binary_repr!(
             collection type_codes::ARRAY_CODE,
@@ -466,7 +466,7 @@ mod tests {
             "Expected serialized value to match the binary representation"
         );
 
-        let elements = [
+        let elements = vec![
             serde_json::Value::String("Hello, world".into()),
             serde_json::Value::String("Some demo".into()),
         ];
@@ -482,7 +482,7 @@ mod tests {
             "Expected serialized value to match the binary representation"
         );
 
-        let elements = [];
+        let elements = vec![];
         let result = serialize_value(ReferenceValue::Array(elements.iter()));
         let expected = binary_repr!(
             collection type_codes::ARRAY_CODE,
@@ -493,7 +493,7 @@ mod tests {
             "Expected serialized value to match the binary representation"
         );
 
-        let elements = [
+        let elements = vec![
             serde_json::Value::Null,
             serde_json::Value::String("Hello, world".into()),
             serde_json::Value::Number(12345.into()),

src/snippet/mod.rs

@@ -743,12 +743,11 @@ Survey in 2016, 2017, and 2018."#;
 
     #[test]
     fn test_collapse_overlapped_ranges() {
-        #![allow(clippy::single_range_in_vec_init)]
-        assert_eq!(&collapse_overlapped_ranges(&[0..1, 2..3]), &[0..1, 2..3]);
-        assert_eq!(&collapse_overlapped_ranges(&[0..1, 1..2]), &[0..1, 1..2]);
-        assert_eq!(&collapse_overlapped_ranges(&[0..2, 1..2]), &[0..2]);
-        assert_eq!(&collapse_overlapped_ranges(&[0..2, 1..3]), &[0..3]);
-        assert_eq!(&collapse_overlapped_ranges(&[0..3, 1..2]), &[0..3]);
+        assert_eq!(&collapse_overlapped_ranges(&[0..1, 2..3,]), &[0..1, 2..3]);
+        assert_eq!(&collapse_overlapped_ranges(&[0..1, 1..2,]), &[0..1, 1..2]);
+        assert_eq!(&collapse_overlapped_ranges(&[0..2, 1..2,]), &[0..2]);
+        assert_eq!(&collapse_overlapped_ranges(&[0..2, 1..3,]), &[0..3]);
+        assert_eq!(&collapse_overlapped_ranges(&[0..3, 1..2,]), &[0..3]);
     }
 
     #[test]

src/space_usage/mod.rs

@@ -117,6 +117,7 @@ impl SegmentSpaceUsage {
         use self::ComponentSpaceUsage::*;
         use crate::SegmentComponent::*;
         match component {
+            FieldList => ComponentSpaceUsage::Basic(ByteCount::from(0u64)),
             Postings => PerField(self.postings().clone()),
             Positions => PerField(self.positions().clone()),
             FastFields => PerField(self.fast_fields().clone()),
@@ -290,7 +291,7 @@ impl FieldUsage {
 
 #[cfg(test)]
 mod test {
-    use crate::index::Index;
+    use crate::core::Index;
     use crate::schema::{Field, Schema, FAST, INDEXED, STORED, TEXT};
     use crate::space_usage::PerFieldSpaceUsage;
     use crate::{IndexWriter, Term};

src/store/mod.rs

@@ -129,7 +129,10 @@ pub mod tests {
             );
         }
 
-        for doc in store.iter::<TantivyDocument>(Some(&alive_bitset)) {
+        for (_, doc) in store
+            .iter::<TantivyDocument>(Some(&alive_bitset))
+            .enumerate()
+        {
             let doc = doc?;
             let title_content = doc.get_first(field_title).unwrap().as_str().unwrap();
             if !title_content.starts_with("Doc ") {

stacker/Cargo.toml

@@ -11,7 +11,6 @@ description = "term hashmap used for indexing"
 murmurhash32 = "0.3"
 common = { version = "0.6", path = "../common/", package = "tantivy-common" }
 ahash = { version = "0.8.3", default-features = false, optional = true }
-rand_distr = "0.4.3"
 
 [[bench]]
 harness = false

stacker/fuzz_test/Cargo.toml

@@ -1,15 +0,0 @@
-[package]
-name = "fuzz_test"
-version = "0.1.0"
-edition = "2021"
-
-# See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
-
-[dependencies]
-ahash = "0.8.7"
-rand = "0.8.5"
-rand_distr = "0.4.3"
-tantivy-stacker = { version = "0.2.0", path = ".." }
-
-[workspace]
-

stacker/fuzz_test/src/main.rs

@@ -1,45 +0,0 @@
-use ahash::AHashMap;
-use rand::{rngs::StdRng, Rng, SeedableRng};
-use rand_distr::Exp;
-use tantivy_stacker::ArenaHashMap;
-
-fn main() {
-    for _ in 0..1_000_000 {
-        let seed: u64 = rand::random();
-        test_with_seed(seed);
-    }
-}
-
-fn test_with_seed(seed: u64) {
-    let mut hash_map = AHashMap::new();
-    let mut arena_hashmap = ArenaHashMap::default();
-    let mut rng = StdRng::seed_from_u64(seed);
-    let key_count = rng.gen_range(1_000..=1_000_000);
-    let exp = Exp::new(0.05).unwrap();
-
-    for _ in 0..key_count {
-        let key_length = rng.sample::<f32, _>(exp).min(u16::MAX as f32).max(1.0) as usize;
-
-        let key: Vec<u8> = (0..key_length).map(|_| rng.gen()).collect();
-
-        arena_hashmap.mutate_or_create(&key, |current_count| {
-            let count: u64 = current_count.unwrap_or(0);
-            count + 1
-        });
-        hash_map.entry(key).and_modify(|e| *e += 1).or_insert(1);
-    }
-
-    println!(
-        "Seed: {} \t {:.2}MB",
-        seed,
-        arena_hashmap.memory_arena.len() as f32 / 1024.0 / 1024.0
-    );
-    // Check the contents of the ArenaHashMap
-    for (key, addr) in arena_hashmap.iter() {
-        let count: u64 = arena_hashmap.read(addr);
-        let count_expected = hash_map
-            .get(key)
-            .unwrap_or_else(|| panic!("NOT FOUND: Key: {:?}, Count: {}", key, count));
-        assert_eq!(count, *count_expected);
-    }
-}

stacker/src/memory_arena.rs

@@ -113,15 +113,6 @@ impl MemoryArena {
         self.pages.len() * PAGE_SIZE
     }
 
-    /// Returns the number of bytes allocated in the arena.
-    pub fn len(&self) -> usize {
-        self.pages.len().saturating_sub(1) * PAGE_SIZE + self.pages.last().unwrap().len
-    }
-
-    pub fn is_empty(&self) -> bool {
-        self.len() == 0
-    }
-
     #[inline]
     pub fn write_at<Item: Copy + 'static>(&mut self, addr: Addr, val: Item) {
         let dest = self.slice_mut(addr, std::mem::size_of::<Item>());
@@ -198,11 +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,
@@ -252,7 +238,6 @@ impl Page {
 mod tests {
 
     use super::MemoryArena;
-    use crate::memory_arena::PAGE_SIZE;
 
     #[test]
     fn test_arena_allocate_slice() {
@@ -270,31 +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();