Provide StoreReader::enumerate to simplify creation of secondary indexes

For secondary indexes, it is often necessary to read all documents, compute
some function on them and associated the result with a document ID.

Currently, this requires something like

```rust
let reader = segment.get_store_reader(1)?;

for doc_id in segment.doc_ids_alive() {
    let doc = reader.get(doc_id)?;

    // Use doc and doc_id here ...
}
```

which can be simplified to

```rust
let reader = segment.get_store_reader(1)?;

for res in reader.enumerate() {
    let (doc_id, doc) = res?;

    // Use doc and doc_id here ...
}
```

using the method proposed here.

(I added a new method instead of modifying `StoreReader::iter` to make the
change backwards compatible, i.e. possible to include in a point release.)

Cargo.toml

@@ -15,7 +15,8 @@ rust-version = "1.63"
 exclude = ["benches/*.json", "benches/*.txt"]
 
 [dependencies]
-oneshot = "0.1.7"
+# Switch back to the non-forked oneshot crate once https://github.com/faern/oneshot/pull/35 is merged
+oneshot = { git = "https://github.com/fulmicoton/oneshot.git", rev = "c10a3ba" }
 base64 = "0.22.0"
 byteorder = "1.4.3"
 crc32fast = "1.3.2"
@@ -63,7 +64,7 @@ query-grammar = { version = "0.22.0", path = "./query-grammar", package = "tanti
 tantivy-bitpacker = { version = "0.6", path = "./bitpacker" }
 common = { version = "0.7", path = "./common/", package = "tantivy-common" }
 tokenizer-api = { version = "0.3", path = "./tokenizer-api", package = "tantivy-tokenizer-api" }
-sketches-ddsketch = { version = "0.3.0", features = ["use_serde"] }
+sketches-ddsketch = { version = "0.2.1", features = ["use_serde"] }
 futures-util = { version = "0.3.28", optional = true }
 fnv = "1.0.7"
 

README.md

@@ -101,8 +101,7 @@ cargo test
 ## Companies Using Tantivy
 
 <p align="left">
-<img align="center" src="doc/assets/images/etsy.png" alt="Etsy" height="25" width="auto" /> &nbsp;
-<img align="center" src="doc/assets/images/paradedb.png" alt="ParadeDB" height="25" width="auto" /> &nbsp;
+<img align="center" src="doc/assets/images/etsy.png" alt="Etsy" height="25" width="auto" />&nbsp;
 <img align="center" src="doc/assets/images/Nuclia.png#gh-light-mode-only" alt="Nuclia" height="25" width="auto" /> &nbsp;
 <img align="center" src="doc/assets/images/humanfirst.png#gh-light-mode-only" alt="Humanfirst.ai" height="30" width="auto" />
 <img align="center" src="doc/assets/images/element.io.svg#gh-light-mode-only" alt="Element.io" height="25" width="auto" />

benches/agg_bench.rs

@@ -47,7 +47,6 @@ fn bench_agg(mut group: InputGroup<Index>) {
     register!(group, average_f64);
     register!(group, average_f64_u64);
     register!(group, stats_f64);
-    register!(group, extendedstats_f64);
     register!(group, percentiles_f64);
     register!(group, terms_few);
     register!(group, terms_many);
@@ -106,12 +105,7 @@ fn stats_f64(index: &Index) {
     });
     exec_term_with_agg(index, agg_req)
 }
-fn extendedstats_f64(index: &Index) {
-    let agg_req = json!({
-        "extendedstats_f64": { "extended_stats": { "field": "score_f64", } }
-    });
-    exec_term_with_agg(index, agg_req)
-}
+
 fn percentiles_f64(index: &Index) {
     let agg_req = json!({
       "mypercentiles": {
@@ -355,7 +349,7 @@ fn get_test_index_bench(cardinality: Cardinality) -> tantivy::Result<Index> {
     let lg_norm = rand_distr::LogNormal::new(2.996f64, 0.979f64).unwrap();
 
     let many_terms_data = (0..150_000)
-        .map(|num| format!("author{num}"))
+        .map(|num| format!("author{}", num))
         .collect::<Vec<_>>();
     {
         let mut rng = StdRng::from_seed([1u8; 32]);

benches/index-bench.rs

@@ -141,12 +141,12 @@ pub fn hdfs_index_benchmark(c: &mut Criterion) {
                 let parse_json = false;
                 // for parse_json in [false, true] {
                 let suffix = if parse_json {
-                    format!("{suffix}-with-json-parsing")
+                    format!("{}-with-json-parsing", suffix)
                 } else {
                     suffix.to_string()
                 };
 
-                let bench_name = format!("{prefix}{suffix}");
+                let bench_name = format!("{}{}", prefix, suffix);
                 group.bench_function(bench_name, |b| {
                     benchmark(b, HDFS_LOGS, schema.clone(), commit, parse_json, is_dynamic)
                 });

columnar/Cargo.toml

@@ -23,16 +23,6 @@ downcast-rs = "1.2.0"
 proptest = "1"
 more-asserts = "0.3.1"
 rand = "0.8"
-binggan = "0.8.1"
-
-[[bench]]
-name = "bench_merge"
-harness = false
-
-[[bench]]
-name = "bench_access"
-harness = false
-
 
 [features]
 unstable = []

columnar/benches/bench_access.rs

@@ -1,67 +0,0 @@
-use binggan::{black_box, InputGroup};
-use common::*;
-use tantivy_columnar::Column;
-
-pub mod common;
-
-const NUM_DOCS: u32 = 2_000_000;
-
-pub fn generate_columnar_and_open(card: Card, num_docs: u32) -> Column {
-    let reader = generate_columnar_with_name(card, num_docs, "price");
-    reader.read_columns("price").unwrap()[0]
-        .open_u64_lenient()
-        .unwrap()
-        .unwrap()
-}
-
-fn main() {
-    let mut inputs = Vec::new();
-
-    let mut add_card = |card1: Card| {
-        inputs.push((
-            format!("{card1}"),
-            generate_columnar_and_open(card1, NUM_DOCS),
-        ));
-    };
-
-    add_card(Card::MultiSparse);
-    add_card(Card::Multi);
-    add_card(Card::Sparse);
-    add_card(Card::Dense);
-    add_card(Card::Full);
-
-    bench_group(InputGroup::new_with_inputs(inputs));
-}
-
-fn bench_group(mut runner: InputGroup<Column>) {
-    runner.register("access_values_for_doc", |column| {
-        let mut sum = 0;
-        for i in 0..NUM_DOCS {
-            for value in column.values_for_doc(i) {
-                sum += value;
-            }
-        }
-        black_box(sum);
-    });
-    runner.register("access_first_vals", |column| {
-        let mut sum = 0;
-        const BLOCK_SIZE: usize = 32;
-        let mut docs = vec![0; BLOCK_SIZE];
-        let mut buffer = vec![None; BLOCK_SIZE];
-        for i in (0..NUM_DOCS).step_by(BLOCK_SIZE) {
-            // fill docs
-            for idx in 0..BLOCK_SIZE {
-                docs[idx] = idx as u32 + i;
-            }
-
-            column.first_vals(&docs, &mut buffer);
-            for val in buffer.iter() {
-                let Some(val) = val else { continue };
-                sum += *val;
-            }
-        }
-
-        black_box(sum);
-    });
-    runner.run();
-}

columnar/benches/bench_first_vals.rs

@@ -31,7 +31,7 @@ fn get_test_columns() -> Columns {
     }
     let mut buffer: Vec<u8> = Vec::new();
     dataframe_writer
-        .serialize(data.len() as u32, &mut buffer)
+        .serialize(data.len() as u32, None, &mut buffer)
         .unwrap();
     let columnar = ColumnarReader::open(buffer).unwrap();
 

columnar/benches/bench_merge.rs

@@ -1,49 +0,0 @@
-pub mod common;
-
-use binggan::{black_box, BenchRunner};
-use common::{generate_columnar_with_name, Card};
-use tantivy_columnar::*;
-
-const NUM_DOCS: u32 = 100_000;
-
-fn main() {
-    let mut inputs = Vec::new();
-
-    let mut add_combo = |card1: Card, card2: Card| {
-        inputs.push((
-            format!("merge_{card1}_and_{card2}"),
-            vec![
-                generate_columnar_with_name(card1, NUM_DOCS, "price"),
-                generate_columnar_with_name(card2, NUM_DOCS, "price"),
-            ],
-        ));
-    };
-
-    add_combo(Card::Multi, Card::Multi);
-    add_combo(Card::MultiSparse, Card::MultiSparse);
-    add_combo(Card::Dense, Card::Dense);
-    add_combo(Card::Sparse, Card::Sparse);
-    add_combo(Card::Sparse, Card::Dense);
-    add_combo(Card::MultiSparse, Card::Dense);
-    add_combo(Card::MultiSparse, Card::Sparse);
-    add_combo(Card::Multi, Card::Dense);
-    add_combo(Card::Multi, Card::Sparse);
-
-    let runner: BenchRunner = BenchRunner::new();
-    let mut group = runner.new_group();
-    for (input_name, columnar_readers) in inputs.iter() {
-        group.register_with_input(
-            input_name,
-            columnar_readers,
-            move |columnar_readers: &Vec<ColumnarReader>| {
-                let mut out = Vec::new();
-                let columnar_readers = columnar_readers.iter().collect::<Vec<_>>();
-                let merge_row_order = StackMergeOrder::stack(&columnar_readers[..]);
-
-                merge_columnar(&columnar_readers, &[], merge_row_order.into(), &mut out).unwrap();
-                black_box(out);
-            },
-        );
-    }
-    group.run();
-}

columnar/benches/common.rs

@@ -1,59 +0,0 @@
-extern crate tantivy_columnar;
-
-use core::fmt;
-use std::fmt::{Display, Formatter};
-
-use tantivy_columnar::{ColumnarReader, ColumnarWriter};
-
-pub enum Card {
-    MultiSparse,
-    Multi,
-    Sparse,
-    Dense,
-    Full,
-}
-impl Display for Card {
-    fn fmt(&self, f: &mut Formatter) -> fmt::Result {
-        match self {
-            Card::MultiSparse => write!(f, "multi sparse 1/13"),
-            Card::Multi => write!(f, "multi 2x"),
-            Card::Sparse => write!(f, "sparse 1/13"),
-            Card::Dense => write!(f, "dense 1/12"),
-            Card::Full => write!(f, "full"),
-        }
-    }
-}
-pub fn generate_columnar_with_name(card: Card, num_docs: u32, column_name: &str) -> ColumnarReader {
-    let mut columnar_writer = ColumnarWriter::default();
-
-    if let Card::MultiSparse = card {
-        columnar_writer.record_numerical(0, column_name, 10u64);
-        columnar_writer.record_numerical(0, column_name, 10u64);
-    }
-
-    for i in 0..num_docs {
-        match card {
-            Card::MultiSparse | Card::Sparse => {
-                if i % 13 == 0 {
-                    columnar_writer.record_numerical(i, column_name, i as u64);
-                }
-            }
-            Card::Dense => {
-                if i % 12 == 0 {
-                    columnar_writer.record_numerical(i, column_name, i as u64);
-                }
-            }
-            Card::Full => {
-                columnar_writer.record_numerical(i, column_name, i as u64);
-            }
-            Card::Multi => {
-                columnar_writer.record_numerical(i, column_name, i as u64);
-                columnar_writer.record_numerical(i, column_name, i as u64);
-            }
-        }
-    }
-
-    let mut wrt: Vec<u8> = Vec::new();
-    columnar_writer.serialize(num_docs, &mut wrt).unwrap();
-    ColumnarReader::open(wrt).unwrap()
-}

columnar/src/column/mod.rs

@@ -136,7 +136,7 @@ impl<T: PartialOrd + Copy + Debug + Send + Sync + 'static> Column<T> {
             .map(|value_row_id: RowId| self.values.get_val(value_row_id))
     }
 
-    /// Get the docids of values which are in the provided value and docid range.
+    /// Get the docids of values which are in the provided value range.
     #[inline]
     pub fn get_docids_for_value_range(
         &self,

columnar/src/column/serialize.rs

@@ -12,7 +12,7 @@ use crate::column_values::{
     CodecType, MonotonicallyMappableToU128, MonotonicallyMappableToU64,
 };
 use crate::iterable::Iterable;
-use crate::{StrColumn, Version};
+use crate::StrColumn;
 
 pub fn serialize_column_mappable_to_u128<T: MonotonicallyMappableToU128>(
     column_index: SerializableColumnIndex<'_>,
@@ -40,10 +40,7 @@ pub fn serialize_column_mappable_to_u64<T: MonotonicallyMappableToU64>(
     Ok(())
 }
 
-pub fn open_column_u64<T: MonotonicallyMappableToU64>(
-    bytes: OwnedBytes,
-    format_version: Version,
-) -> io::Result<Column<T>> {
+pub fn open_column_u64<T: MonotonicallyMappableToU64>(bytes: OwnedBytes) -> io::Result<Column<T>> {
     let (body, column_index_num_bytes_payload) = bytes.rsplit(4);
     let column_index_num_bytes = u32::from_le_bytes(
         column_index_num_bytes_payload
@@ -52,7 +49,7 @@ pub fn open_column_u64<T: MonotonicallyMappableToU64>(
             .unwrap(),
     );
     let (column_index_data, column_values_data) = body.split(column_index_num_bytes as usize);
-    let column_index = crate::column_index::open_column_index(column_index_data, format_version)?;
+    let column_index = crate::column_index::open_column_index(column_index_data)?;
     let column_values = load_u64_based_column_values(column_values_data)?;
     Ok(Column {
         index: column_index,
@@ -62,7 +59,6 @@ pub fn open_column_u64<T: MonotonicallyMappableToU64>(
 
 pub fn open_column_u128<T: MonotonicallyMappableToU128>(
     bytes: OwnedBytes,
-    format_version: Version,
 ) -> io::Result<Column<T>> {
     let (body, column_index_num_bytes_payload) = bytes.rsplit(4);
     let column_index_num_bytes = u32::from_le_bytes(
@@ -72,7 +68,7 @@ pub fn open_column_u128<T: MonotonicallyMappableToU128>(
             .unwrap(),
     );
     let (column_index_data, column_values_data) = body.split(column_index_num_bytes as usize);
-    let column_index = crate::column_index::open_column_index(column_index_data, format_version)?;
+    let column_index = crate::column_index::open_column_index(column_index_data)?;
     let column_values = crate::column_values::open_u128_mapped(column_values_data)?;
     Ok(Column {
         index: column_index,
@@ -83,10 +79,7 @@ pub fn open_column_u128<T: MonotonicallyMappableToU128>(
 /// Open the column as u64.
 ///
 /// See [`open_u128_as_compact_u64`] for more details.
-pub fn open_column_u128_as_compact_u64(
-    bytes: OwnedBytes,
-    format_version: Version,
-) -> io::Result<Column<u64>> {
+pub fn open_column_u128_as_compact_u64(bytes: OwnedBytes) -> io::Result<Column<u64>> {
     let (body, column_index_num_bytes_payload) = bytes.rsplit(4);
     let column_index_num_bytes = u32::from_le_bytes(
         column_index_num_bytes_payload
@@ -95,7 +88,7 @@ pub fn open_column_u128_as_compact_u64(
             .unwrap(),
     );
     let (column_index_data, column_values_data) = body.split(column_index_num_bytes as usize);
-    let column_index = crate::column_index::open_column_index(column_index_data, format_version)?;
+    let column_index = crate::column_index::open_column_index(column_index_data)?;
     let column_values = crate::column_values::open_u128_as_compact_u64(column_values_data)?;
     Ok(Column {
         index: column_index,
@@ -103,19 +96,19 @@ pub fn open_column_u128_as_compact_u64(
     })
 }
 
-pub fn open_column_bytes(data: OwnedBytes, format_version: Version) -> io::Result<BytesColumn> {
+pub fn open_column_bytes(data: OwnedBytes) -> io::Result<BytesColumn> {
     let (body, dictionary_len_bytes) = data.rsplit(4);
     let dictionary_len = u32::from_le_bytes(dictionary_len_bytes.as_slice().try_into().unwrap());
     let (dictionary_bytes, column_bytes) = body.split(dictionary_len as usize);
     let dictionary = Arc::new(Dictionary::from_bytes(dictionary_bytes)?);
-    let term_ord_column = crate::column::open_column_u64::<u64>(column_bytes, format_version)?;
+    let term_ord_column = crate::column::open_column_u64::<u64>(column_bytes)?;
     Ok(BytesColumn {
         dictionary,
         term_ord_column,
     })
 }
 
-pub fn open_column_str(data: OwnedBytes, format_version: Version) -> io::Result<StrColumn> {
-    let bytes_column = open_column_bytes(data, format_version)?;
+pub fn open_column_str(data: OwnedBytes) -> io::Result<StrColumn> {
+    let bytes_column = open_column_bytes(data)?;
     Ok(StrColumn::wrap(bytes_column))
 }

columnar/src/column_index/merge/mod.rs

@@ -95,12 +95,8 @@ pub fn merge_column_index<'a>(
 
 #[cfg(test)]
 mod tests {
-    use common::OwnedBytes;
-
     use crate::column_index::merge::detect_cardinality;
-    use crate::column_index::multivalued_index::{
-        open_multivalued_index, serialize_multivalued_index, MultiValueIndex,
-    };
+    use crate::column_index::multivalued_index::MultiValueIndex;
     use crate::column_index::{merge_column_index, OptionalIndex, SerializableColumnIndex};
     use crate::{
         Cardinality, ColumnIndex, MergeRowOrder, RowAddr, RowId, ShuffleMergeOrder, StackMergeOrder,
@@ -175,11 +171,7 @@ mod tests {
         let SerializableColumnIndex::Multivalued(start_index_iterable) = merged_column_index else {
             panic!("Excpected a multivalued index")
         };
-        let mut output = Vec::new();
-        serialize_multivalued_index(&start_index_iterable, &mut output).unwrap();
-        let multivalue =
-            open_multivalued_index(OwnedBytes::new(output), crate::Version::V2).unwrap();
-        let start_indexes: Vec<RowId> = multivalue.get_start_index_column().iter().collect();
+        let start_indexes: Vec<RowId> = start_index_iterable.boxed_iter().collect();
         assert_eq!(&start_indexes, &[0, 3, 5]);
     }
 
@@ -208,16 +200,11 @@ mod tests {
             ],
         )
         .into();
-
         let merged_column_index = merge_column_index(&column_indexes[..], &merge_row_order);
         let SerializableColumnIndex::Multivalued(start_index_iterable) = merged_column_index else {
             panic!("Excpected a multivalued index")
         };
-        let mut output = Vec::new();
-        serialize_multivalued_index(&start_index_iterable, &mut output).unwrap();
-        let multivalue =
-            open_multivalued_index(OwnedBytes::new(output), crate::Version::V2).unwrap();
-        let start_indexes: Vec<RowId> = multivalue.get_start_index_column().iter().collect();
+        let start_indexes: Vec<RowId> = start_index_iterable.boxed_iter().collect();
         assert_eq!(&start_indexes, &[0, 3, 5, 6]);
     }
 }

columnar/src/column_index/merge/shuffled.rs

@@ -1,8 +1,6 @@
 use std::iter;
 
-use crate::column_index::{
-    SerializableColumnIndex, SerializableMultivalueIndex, SerializableOptionalIndex, Set,
-};
+use crate::column_index::{SerializableColumnIndex, Set};
 use crate::iterable::Iterable;
 use crate::{Cardinality, ColumnIndex, RowId, ShuffleMergeOrder};
 
@@ -16,24 +14,15 @@ pub fn merge_column_index_shuffled<'a>(
         Cardinality::Optional => {
             let non_null_row_ids =
                 merge_column_index_shuffled_optional(column_indexes, shuffle_merge_order);
-            SerializableColumnIndex::Optional(SerializableOptionalIndex {
+            SerializableColumnIndex::Optional {
                 non_null_row_ids,
                 num_rows: shuffle_merge_order.num_rows(),
-            })
+            }
         }
         Cardinality::Multivalued => {
-            let non_null_row_ids =
-                merge_column_index_shuffled_optional(column_indexes, shuffle_merge_order);
-            SerializableColumnIndex::Multivalued(SerializableMultivalueIndex {
-                doc_ids_with_values: SerializableOptionalIndex {
-                    non_null_row_ids,
-                    num_rows: shuffle_merge_order.num_rows(),
-                },
-                start_offsets: merge_column_index_shuffled_multivalued(
-                    column_indexes,
-                    shuffle_merge_order,
-                ),
-            })
+            let multivalue_start_index =
+                merge_column_index_shuffled_multivalued(column_indexes, shuffle_merge_order);
+            SerializableColumnIndex::Multivalued(multivalue_start_index)
         }
     }
 }
@@ -113,18 +102,11 @@ fn iter_num_values<'a>(
 
 /// Transforms an iterator containing the number of vals per row (with `num_rows` elements)
 /// into a `start_offset` iterator starting at 0 and (with `num_rows + 1` element)
-///
-/// This will filter values with 0 values as these are covered by the optional index in the
-/// multivalue index.
 fn integrate_num_vals(num_vals: impl Iterator<Item = u32>) -> impl Iterator<Item = RowId> {
-    iter::once(0u32).chain(
-        num_vals
-            .filter(|num_vals| *num_vals != 0)
-            .scan(0, |state, num_vals| {
-                *state += num_vals;
-                Some(*state)
-            }),
-    )
+    iter::once(0u32).chain(num_vals.scan(0, |state, num_vals| {
+        *state += num_vals;
+        Some(*state)
+    }))
 }
 
 impl<'a> Iterable<u32> for ShuffledMultivaluedIndex<'a> {
@@ -152,7 +134,7 @@ mod tests {
 
     #[test]
     fn test_integrate_num_vals_several() {
-        assert!(integrate_num_vals([3, 0, 10, 20].into_iter()).eq([0, 3, 13, 33].into_iter()));
+        assert!(integrate_num_vals([3, 0, 10, 20].into_iter()).eq([0, 3, 3, 13, 33].into_iter()));
     }
 
     #[test]
@@ -175,10 +157,10 @@ mod tests {
             Cardinality::Optional,
             &shuffle_merge_order,
         );
-        let SerializableColumnIndex::Optional(SerializableOptionalIndex {
+        let SerializableColumnIndex::Optional {
             non_null_row_ids,
             num_rows,
-        }) = serializable_index
+        } = serializable_index
         else {
             panic!()
         };

columnar/src/column_index/merge/stacked.rs

@@ -1,8 +1,6 @@
-use std::ops::Range;
+use std::iter;
 
-use crate::column_index::multivalued_index::{MultiValueIndex, SerializableMultivalueIndex};
-use crate::column_index::serialize::SerializableOptionalIndex;
-use crate::column_index::SerializableColumnIndex;
+use crate::column_index::{SerializableColumnIndex, Set};
 use crate::iterable::Iterable;
 use crate::{Cardinality, ColumnIndex, RowId, StackMergeOrder};
 
@@ -17,146 +15,20 @@ pub fn merge_column_index_stacked<'a>(
 ) -> SerializableColumnIndex<'a> {
     match cardinality_after_merge {
         Cardinality::Full => SerializableColumnIndex::Full,
-        Cardinality::Optional => SerializableColumnIndex::Optional(SerializableOptionalIndex {
+        Cardinality::Optional => SerializableColumnIndex::Optional {
             non_null_row_ids: Box::new(StackedOptionalIndex {
                 columns,
                 stack_merge_order,
             }),
             num_rows: stack_merge_order.num_rows(),
-        }),
-        Cardinality::Multivalued => {
-            let serializable_multivalue_index =
-                make_serializable_multivalued_index(columns, stack_merge_order);
-            SerializableColumnIndex::Multivalued(serializable_multivalue_index)
-        }
-    }
-}
-
-struct StackedDocIdsWithValues<'a> {
-    column_indexes: &'a [ColumnIndex],
-    stack_merge_order: &'a StackMergeOrder,
-}
-
-impl Iterable<u32> for StackedDocIdsWithValues<'_> {
-    fn boxed_iter(&self) -> Box<dyn Iterator<Item = u32> + '_> {
-        Box::new((0..self.column_indexes.len()).flat_map(|i| {
-            let column_index = &self.column_indexes[i];
-            let doc_range = self.stack_merge_order.columnar_range(i);
-            get_doc_ids_with_values(column_index, doc_range)
-        }))
-    }
-}
-
-fn get_doc_ids_with_values<'a>(
-    column_index: &'a ColumnIndex,
-    doc_range: Range<u32>,
-) -> Box<dyn Iterator<Item = u32> + 'a> {
-    match column_index {
-        ColumnIndex::Empty { .. } => Box::new(0..0),
-        ColumnIndex::Full => Box::new(doc_range),
-        ColumnIndex::Optional(optional_index) => Box::new(
-            optional_index
-                .iter_rows()
-                .map(move |row| row + doc_range.start),
-        ),
-        ColumnIndex::Multivalued(multivalued_index) => match multivalued_index {
-            MultiValueIndex::MultiValueIndexV1(multivalued_index) => {
-                Box::new((0..multivalued_index.num_docs()).filter_map(move |docid| {
-                    let range = multivalued_index.range(docid);
-                    if range.is_empty() {
-                        None
-                    } else {
-                        Some(docid + doc_range.start)
-                    }
-                }))
-            }
-            MultiValueIndex::MultiValueIndexV2(multivalued_index) => Box::new(
-                multivalued_index
-                    .optional_index
-                    .iter_rows()
-                    .map(move |row| row + doc_range.start),
-            ),
         },
-    }
-}
-
-fn stack_doc_ids_with_values<'a>(
-    column_indexes: &'a [ColumnIndex],
-    stack_merge_order: &'a StackMergeOrder,
-) -> SerializableOptionalIndex<'a> {
-    let num_rows = stack_merge_order.num_rows();
-    SerializableOptionalIndex {
-        non_null_row_ids: Box::new(StackedDocIdsWithValues {
-            column_indexes,
-            stack_merge_order,
-        }),
-        num_rows,
-    }
-}
-
-struct StackedStartOffsets<'a> {
-    column_indexes: &'a [ColumnIndex],
-    stack_merge_order: &'a StackMergeOrder,
-}
-
-fn get_num_values_iterator<'a>(
-    column_index: &'a ColumnIndex,
-    num_docs: u32,
-) -> Box<dyn Iterator<Item = u32> + 'a> {
-    match column_index {
-        ColumnIndex::Empty { .. } => Box::new(std::iter::empty()),
-        ColumnIndex::Full => Box::new(std::iter::repeat(1u32).take(num_docs as usize)),
-        ColumnIndex::Optional(optional_index) => {
-            Box::new(std::iter::repeat(1u32).take(optional_index.num_non_nulls() as usize))
+        Cardinality::Multivalued => {
+            let stacked_multivalued_index = StackedMultivaluedIndex {
+                columns,
+                stack_merge_order,
+            };
+            SerializableColumnIndex::Multivalued(Box::new(stacked_multivalued_index))
         }
-        ColumnIndex::Multivalued(multivalued_index) => Box::new(
-            multivalued_index
-                .get_start_index_column()
-                .iter()
-                .scan(0u32, |previous_start_offset, current_start_offset| {
-                    let num_vals = current_start_offset - *previous_start_offset;
-                    *previous_start_offset = current_start_offset;
-                    Some(num_vals)
-                })
-                .skip(1),
-        ),
-    }
-}
-
-impl<'a> Iterable<u32> for StackedStartOffsets<'a> {
-    fn boxed_iter(&self) -> Box<dyn Iterator<Item = u32> + '_> {
-        let num_values_it = (0..self.column_indexes.len()).flat_map(|columnar_id| {
-            let num_docs = self.stack_merge_order.columnar_range(columnar_id).len() as u32;
-            let column_index = &self.column_indexes[columnar_id];
-            get_num_values_iterator(column_index, num_docs)
-        });
-        Box::new(std::iter::once(0u32).chain(num_values_it.into_iter().scan(
-            0u32,
-            |cumulated, el| {
-                *cumulated += el;
-                Some(*cumulated)
-            },
-        )))
-    }
-}
-
-fn stack_start_offsets<'a>(
-    column_indexes: &'a [ColumnIndex],
-    stack_merge_order: &'a StackMergeOrder,
-) -> Box<dyn Iterable<u32> + 'a> {
-    Box::new(StackedStartOffsets {
-        column_indexes,
-        stack_merge_order,
-    })
-}
-
-fn make_serializable_multivalued_index<'a>(
-    columns: &'a [ColumnIndex],
-    stack_merge_order: &'a StackMergeOrder,
-) -> SerializableMultivalueIndex<'a> {
-    SerializableMultivalueIndex {
-        doc_ids_with_values: stack_doc_ids_with_values(columns, stack_merge_order),
-        start_offsets: stack_start_offsets(columns, stack_merge_order),
     }
 }
 
@@ -190,3 +62,87 @@ impl<'a> Iterable<RowId> for StackedOptionalIndex<'a> {
         )
     }
 }
+
+#[derive(Clone, Copy)]
+struct StackedMultivaluedIndex<'a> {
+    columns: &'a [ColumnIndex],
+    stack_merge_order: &'a StackMergeOrder,
+}
+
+fn convert_column_opt_to_multivalued_index<'a>(
+    column_index_opt: &'a ColumnIndex,
+    num_rows: RowId,
+) -> Box<dyn Iterator<Item = RowId> + 'a> {
+    match column_index_opt {
+        ColumnIndex::Empty { .. } => Box::new(iter::repeat(0u32).take(num_rows as usize + 1)),
+        ColumnIndex::Full => Box::new(0..num_rows + 1),
+        ColumnIndex::Optional(optional_index) => {
+            Box::new(
+                (0..num_rows)
+                    // TODO optimize
+                    .map(|row_id| optional_index.rank(row_id))
+                    .chain(std::iter::once(optional_index.num_non_nulls())),
+            )
+        }
+        ColumnIndex::Multivalued(multivalued_index) => multivalued_index.start_index_column.iter(),
+    }
+}
+
+impl<'a> Iterable<RowId> for StackedMultivaluedIndex<'a> {
+    fn boxed_iter(&self) -> Box<dyn Iterator<Item = RowId> + '_> {
+        let multivalued_indexes =
+            self.columns
+                .iter()
+                .enumerate()
+                .map(|(columnar_id, column_opt)| {
+                    let num_rows =
+                        self.stack_merge_order.columnar_range(columnar_id).len() as RowId;
+                    convert_column_opt_to_multivalued_index(column_opt, num_rows)
+                });
+        stack_multivalued_indexes(multivalued_indexes)
+    }
+}
+
+// Refactor me
+fn stack_multivalued_indexes<'a>(
+    mut multivalued_indexes: impl Iterator<Item = Box<dyn Iterator<Item = RowId> + 'a>> + 'a,
+) -> Box<dyn Iterator<Item = RowId> + 'a> {
+    let mut offset = 0;
+    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() {
+            last_row_id = offset + row_id;
+            return Some(last_row_id);
+        }
+        offset = last_row_id;
+        loop {
+            current_it = multivalued_indexes.next();
+            if current_it.as_mut()?.next().is_some() {
+                break;
+            }
+        }
+    }))
+}
+
+#[cfg(test)]
+mod tests {
+    use crate::RowId;
+
+    fn it<'a>(row_ids: &'a [RowId]) -> Box<dyn Iterator<Item = RowId> + 'a> {
+        Box::new(row_ids.iter().copied())
+    }
+
+    #[test]
+    fn test_stack() {
+        let columns = [
+            it(&[0u32, 0u32]),
+            it(&[0u32, 1u32, 1u32, 4u32]),
+            it(&[0u32, 3u32, 5u32]),
+            it(&[0u32, 4u32]),
+        ]
+        .into_iter();
+        let start_offsets: Vec<RowId> = super::stack_multivalued_indexes(columns).collect();
+        assert_eq!(start_offsets, &[0, 0, 1, 1, 4, 7, 9, 13]);
+    }
+}

columnar/src/column_index/mod.rs

@@ -11,11 +11,8 @@ mod serialize;
 use std::ops::Range;
 
 pub use merge::merge_column_index;
-pub(crate) use multivalued_index::SerializableMultivalueIndex;
 pub use optional_index::{OptionalIndex, Set};
-pub use serialize::{
-    open_column_index, serialize_column_index, SerializableColumnIndex, SerializableOptionalIndex,
-};
+pub use serialize::{open_column_index, serialize_column_index, SerializableColumnIndex};
 
 use crate::column_index::multivalued_index::MultiValueIndex;
 use crate::{Cardinality, DocId, RowId};
@@ -134,41 +131,15 @@ impl ColumnIndex {
                 let row_end = optional_index.rank(doc_id_range.end);
                 row_start..row_end
             }
-            ColumnIndex::Multivalued(multivalued_index) => match multivalued_index {
-                MultiValueIndex::MultiValueIndexV1(index) => {
-                    let row_start = index.start_index_column.get_val(doc_id_range.start);
-                    let row_end = index.start_index_column.get_val(doc_id_range.end);
-                    row_start..row_end
-                }
-                MultiValueIndex::MultiValueIndexV2(index) => {
-                    // In this case we will use the optional_index select the next values
-                    // that are valid. There are different cases to consider:
-                    // Not exists below means does not exist in the optional
-                    // index, because it has no values.
-                    // * doc_id_range may cover a range of docids which are non existent
-                    // => rank
-                    //   will give us the next document outside the range with a value. They both
-                    //   get the same rank and therefore return a zero range
-                    //
-                    // * doc_id_range.start and doc_id_range.end may not exist, but docids in
-                    // between may have values
-                    // => rank will give us the next document outside the range with a value.
-                    //
-                    // * doc_id_range.start may be not existent but doc_id_range.end may exist
-                    // * doc_id_range.start may exist but doc_id_range.end may not exist
-                    // * doc_id_range.start and doc_id_range.end may exist
-                    // => rank on doc_id_range.end will give use the next value, which matches
-                    // how the `start_index_column` works, so we get the value start of the next
-                    // docid which we use to create the exclusive range.
-                    //
-                    let rank_start = index.optional_index.rank(doc_id_range.start);
-                    let row_start = index.start_index_column.get_val(rank_start);
-                    let rank_end = index.optional_index.rank(doc_id_range.end);
-                    let row_end = index.start_index_column.get_val(rank_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);
 
-                    row_start..row_end
-                }
-            },
+                let row_start = multivalued_index.start_index_column.get_val(start_docid);
+                let row_end = multivalued_index.start_index_column.get_val(end_docid);
+
+                row_start..row_end
+            }
         }
     }
 

columnar/src/column_index/multivalued_index.rs

@@ -3,98 +3,64 @@ use std::io::Write;
 use std::ops::Range;
 use std::sync::Arc;
 
-use common::{CountingWriter, OwnedBytes};
+use common::OwnedBytes;
 
-use super::optional_index::{open_optional_index, serialize_optional_index};
-use super::{OptionalIndex, SerializableOptionalIndex, Set};
 use crate::column_values::{
     load_u64_based_column_values, serialize_u64_based_column_values, CodecType, ColumnValues,
 };
 use crate::iterable::Iterable;
-use crate::{DocId, RowId, Version};
-
-pub struct SerializableMultivalueIndex<'a> {
-    pub doc_ids_with_values: SerializableOptionalIndex<'a>,
-    pub start_offsets: Box<dyn Iterable<u32> + 'a>,
-}
+use crate::{DocId, RowId};
 
 pub fn serialize_multivalued_index(
-    multivalued_index: &SerializableMultivalueIndex,
+    multivalued_index: &dyn Iterable<RowId>,
     output: &mut impl Write,
 ) -> io::Result<()> {
-    let SerializableMultivalueIndex {
-        doc_ids_with_values,
-        start_offsets,
-    } = multivalued_index;
-    let mut count_writer = CountingWriter::wrap(output);
-    let SerializableOptionalIndex {
-        non_null_row_ids,
-        num_rows,
-    } = doc_ids_with_values;
-    serialize_optional_index(&**non_null_row_ids, *num_rows, &mut count_writer)?;
-    let optional_len = count_writer.written_bytes() as u32;
-    let output = count_writer.finish();
     serialize_u64_based_column_values(
-        &**start_offsets,
+        multivalued_index,
         &[CodecType::Bitpacked, CodecType::Linear],
         output,
     )?;
-    output.write_all(&optional_len.to_le_bytes())?;
     Ok(())
 }
 
-pub fn open_multivalued_index(
-    bytes: OwnedBytes,
-    format_version: Version,
-) -> io::Result<MultiValueIndex> {
-    match format_version {
-        Version::V1 => {
-            let start_index_column: Arc<dyn ColumnValues<RowId>> =
-                load_u64_based_column_values(bytes)?;
-            Ok(MultiValueIndex::MultiValueIndexV1(MultiValueIndexV1 {
-                start_index_column,
-            }))
-        }
-        Version::V2 => {
-            let (body_bytes, optional_index_len) = bytes.rsplit(4);
-            let optional_index_len =
-                u32::from_le_bytes(optional_index_len.as_slice().try_into().unwrap());
-            let (optional_index_bytes, start_index_bytes) =
-                body_bytes.split(optional_index_len as usize);
-            let optional_index = open_optional_index(optional_index_bytes)?;
-            let start_index_column: Arc<dyn ColumnValues<RowId>> =
-                load_u64_based_column_values(start_index_bytes)?;
-            Ok(MultiValueIndex::MultiValueIndexV2(MultiValueIndexV2 {
-                optional_index,
-                start_index_column,
-            }))
-        }
-    }
+pub fn open_multivalued_index(bytes: OwnedBytes) -> io::Result<MultiValueIndex> {
+    let start_index_column: Arc<dyn ColumnValues<RowId>> = load_u64_based_column_values(bytes)?;
+    Ok(MultiValueIndex { start_index_column })
 }
 
 #[derive(Clone)]
 /// Index to resolve value range for given doc_id.
 /// Starts at 0.
-pub enum MultiValueIndex {
-    MultiValueIndexV1(MultiValueIndexV1),
-    MultiValueIndexV2(MultiValueIndexV2),
-}
-
-#[derive(Clone)]
-/// Index to resolve value range for given doc_id.
-/// Starts at 0.
-pub struct MultiValueIndexV1 {
+pub struct MultiValueIndex {
     pub start_index_column: Arc<dyn crate::ColumnValues<RowId>>,
 }
 
-impl MultiValueIndexV1 {
+impl std::fmt::Debug for MultiValueIndex {
+    fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result {
+        f.debug_struct("MultiValuedIndex")
+            .field("num_rows", &self.start_index_column.num_vals())
+            .finish_non_exhaustive()
+    }
+}
+
+impl From<Arc<dyn ColumnValues<RowId>>> for MultiValueIndex {
+    fn from(start_index_column: Arc<dyn ColumnValues<RowId>>) -> Self {
+        MultiValueIndex { start_index_column }
+    }
+}
+
+impl MultiValueIndex {
+    pub fn for_test(start_offsets: &[RowId]) -> MultiValueIndex {
+        let mut buffer = Vec::new();
+        serialize_multivalued_index(&start_offsets, &mut buffer).unwrap();
+        let bytes = OwnedBytes::new(buffer);
+        open_multivalued_index(bytes).unwrap()
+    }
+
     /// Returns `[start, end)`, such that the values associated with
     /// the given document are `start..end`.
     #[inline]
     pub(crate) fn range(&self, doc_id: DocId) -> Range<RowId> {
-        if doc_id >= self.num_docs() {
-            return 0..0;
-        }
         let start = self.start_index_column.get_val(doc_id);
         let end = self.start_index_column.get_val(doc_id + 1);
         start..end
@@ -117,6 +83,7 @@ impl MultiValueIndexV1 {
     ///
     /// TODO: Instead of a linear scan we can employ a exponential search into binary search to
     /// match a docid to its value position.
+    #[allow(clippy::bool_to_int_with_if)]
     pub(crate) fn select_batch_in_place(&self, docid_start: DocId, ranks: &mut Vec<u32>) {
         if ranks.is_empty() {
             return;
@@ -144,170 +111,11 @@ impl MultiValueIndexV1 {
     }
 }
 
-#[derive(Clone)]
-/// Index to resolve value range for given doc_id.
-/// Starts at 0.
-pub struct MultiValueIndexV2 {
-    pub optional_index: OptionalIndex,
-    pub start_index_column: Arc<dyn crate::ColumnValues<RowId>>,
-}
-
-impl std::fmt::Debug for MultiValueIndex {
-    fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result {
-        let index = match self {
-            MultiValueIndex::MultiValueIndexV1(idx) => &idx.start_index_column,
-            MultiValueIndex::MultiValueIndexV2(idx) => &idx.start_index_column,
-        };
-        f.debug_struct("MultiValuedIndex")
-            .field("num_rows", &index.num_vals())
-            .finish_non_exhaustive()
-    }
-}
-
-impl MultiValueIndex {
-    pub fn for_test(start_offsets: &[RowId]) -> MultiValueIndex {
-        assert!(!start_offsets.is_empty());
-        assert_eq!(start_offsets[0], 0);
-        let mut doc_with_values = Vec::new();
-        let mut compact_start_offsets: Vec<u32> = vec![0];
-        for doc in 0..start_offsets.len() - 1 {
-            if start_offsets[doc] < start_offsets[doc + 1] {
-                doc_with_values.push(doc as RowId);
-                compact_start_offsets.push(start_offsets[doc + 1]);
-            }
-        }
-        let serializable_multivalued_index = SerializableMultivalueIndex {
-            doc_ids_with_values: SerializableOptionalIndex {
-                non_null_row_ids: Box::new(&doc_with_values[..]),
-                num_rows: start_offsets.len() as u32 - 1,
-            },
-            start_offsets: Box::new(&compact_start_offsets[..]),
-        };
-        let mut buffer = Vec::new();
-        serialize_multivalued_index(&serializable_multivalued_index, &mut buffer).unwrap();
-        let bytes = OwnedBytes::new(buffer);
-        open_multivalued_index(bytes, Version::V2).unwrap()
-    }
-
-    pub fn get_start_index_column(&self) -> &Arc<dyn crate::ColumnValues<RowId>> {
-        match self {
-            MultiValueIndex::MultiValueIndexV1(idx) => &idx.start_index_column,
-            MultiValueIndex::MultiValueIndexV2(idx) => &idx.start_index_column,
-        }
-    }
-
-    /// Returns `[start, end)` values range, such that the values associated with
-    /// the given document are `start..end`.
-    #[inline]
-    pub(crate) fn range(&self, doc_id: DocId) -> Range<RowId> {
-        match self {
-            MultiValueIndex::MultiValueIndexV1(idx) => idx.range(doc_id),
-            MultiValueIndex::MultiValueIndexV2(idx) => idx.range(doc_id),
-        }
-    }
-
-    /// Returns the number of documents in the index.
-    #[inline]
-    pub fn num_docs(&self) -> u32 {
-        match self {
-            MultiValueIndex::MultiValueIndexV1(idx) => idx.start_index_column.num_vals() - 1,
-            MultiValueIndex::MultiValueIndexV2(idx) => idx.optional_index.num_docs(),
-        }
-    }
-
-    /// Converts a list of ranks (row ids of values) in a 1:n index to the corresponding list of
-    /// docids. Positions are converted inplace to docids.
-    ///
-    /// Since there is no index for value pos -> docid, but docid -> value pos range, we scan the
-    /// index.
-    ///
-    /// Correctness: positions needs to be sorted. idx_reader needs to contain monotonically
-    /// increasing positions.
-    ///
-    /// TODO: Instead of a linear scan we can employ a exponential search into binary search to
-    /// match a docid to its value position.
-    pub(crate) fn select_batch_in_place(&self, docid_start: DocId, ranks: &mut Vec<u32>) {
-        match self {
-            MultiValueIndex::MultiValueIndexV1(idx) => {
-                idx.select_batch_in_place(docid_start, ranks)
-            }
-            MultiValueIndex::MultiValueIndexV2(idx) => {
-                idx.select_batch_in_place(docid_start, ranks)
-            }
-        }
-    }
-}
-impl MultiValueIndexV2 {
-    /// Returns `[start, end)`, such that the values associated with
-    /// the given document are `start..end`.
-    #[inline]
-    pub(crate) fn range(&self, doc_id: DocId) -> Range<RowId> {
-        let Some(rank) = self.optional_index.rank_if_exists(doc_id) else {
-            return 0..0;
-        };
-        let start = self.start_index_column.get_val(rank);
-        let end = self.start_index_column.get_val(rank + 1);
-        start..end
-    }
-
-    /// Returns the number of documents in the index.
-    #[inline]
-    pub fn num_docs(&self) -> u32 {
-        self.optional_index.num_docs()
-    }
-
-    /// Converts a list of ranks (row ids of values) in a 1:n index to the corresponding list of
-    /// docids. Positions are converted inplace to docids.
-    ///
-    /// Since there is no index for value pos -> docid, but docid -> value pos range, we scan the
-    /// index.
-    ///
-    /// Correctness: positions needs to be sorted. idx_reader needs to contain monotonically
-    /// increasing positions.
-    ///
-    /// TODO: Instead of a linear scan we can employ a exponential search into binary search to
-    /// match a docid to its value position.
-    pub(crate) fn select_batch_in_place(&self, docid_start: DocId, ranks: &mut Vec<u32>) {
-        if ranks.is_empty() {
-            return;
-        }
-        let mut cur_pos_in_idx = self.optional_index.rank(docid_start);
-        let mut last_doc = None;
-
-        assert!(cur_pos_in_idx <= ranks[0]);
-
-        let mut write_doc_pos = 0;
-        for i in 0..ranks.len() {
-            let pos = ranks[i];
-            loop {
-                let end = self.start_index_column.get_val(cur_pos_in_idx + 1);
-                if end > pos {
-                    ranks[write_doc_pos] = cur_pos_in_idx;
-                    write_doc_pos += if last_doc == Some(cur_pos_in_idx) {
-                        0
-                    } else {
-                        1
-                    };
-                    last_doc = Some(cur_pos_in_idx);
-                    break;
-                }
-                cur_pos_in_idx += 1;
-            }
-        }
-        ranks.truncate(write_doc_pos);
-
-        for rank in ranks.iter_mut() {
-            *rank = self.optional_index.select(*rank);
-        }
-    }
-}
-
 #[cfg(test)]
 mod tests {
     use std::ops::Range;
 
     use super::MultiValueIndex;
-    use crate::{ColumnarReader, DynamicColumn};
 
     fn index_to_pos_helper(
         index: &MultiValueIndex,
@@ -326,7 +134,6 @@ mod tests {
         let positions = &[10u32, 11, 15, 20, 21, 22];
         assert_eq!(index_to_pos_helper(&index, 0..5, positions), vec![1, 3, 4]);
         assert_eq!(index_to_pos_helper(&index, 1..5, positions), vec![1, 3, 4]);
-
         assert_eq!(index_to_pos_helper(&index, 0..5, &[9]), vec![0]);
         assert_eq!(index_to_pos_helper(&index, 1..5, &[10]), vec![1]);
         assert_eq!(index_to_pos_helper(&index, 1..5, &[11]), vec![1]);
@@ -334,67 +141,4 @@ mod tests {
         assert_eq!(index_to_pos_helper(&index, 2..5, &[12, 14]), vec![2]);
         assert_eq!(index_to_pos_helper(&index, 2..5, &[12, 14, 15]), vec![2, 3]);
     }
-
-    #[test]
-    fn test_range_to_rowids() {
-        use crate::ColumnarWriter;
-
-        let mut columnar_writer = ColumnarWriter::default();
-
-        // This column gets coerced to u64
-        columnar_writer.record_numerical(1, "full", u64::MAX);
-        columnar_writer.record_numerical(1, "full", u64::MAX);
-
-        columnar_writer.record_numerical(5, "full", u64::MAX);
-        columnar_writer.record_numerical(5, "full", u64::MAX);
-
-        let mut wrt: Vec<u8> = Vec::new();
-        columnar_writer.serialize(7, &mut wrt).unwrap();
-
-        let reader = ColumnarReader::open(wrt).unwrap();
-        // Open the column as u64
-        let column = reader.read_columns("full").unwrap()[0]
-            .open()
-            .unwrap()
-            .coerce_numerical(crate::NumericalType::U64)
-            .unwrap();
-        let DynamicColumn::U64(column) = column else {
-            panic!();
-        };
-
-        let row_id_range = column.index.docid_range_to_rowids(1..2);
-        assert_eq!(row_id_range, 0..2);
-
-        let row_id_range = column.index.docid_range_to_rowids(0..2);
-        assert_eq!(row_id_range, 0..2);
-
-        let row_id_range = column.index.docid_range_to_rowids(0..4);
-        assert_eq!(row_id_range, 0..2);
-
-        let row_id_range = column.index.docid_range_to_rowids(3..4);
-        assert_eq!(row_id_range, 2..2);
-
-        let row_id_range = column.index.docid_range_to_rowids(1..6);
-        assert_eq!(row_id_range, 0..4);
-
-        let row_id_range = column.index.docid_range_to_rowids(3..6);
-        assert_eq!(row_id_range, 2..4);
-
-        let row_id_range = column.index.docid_range_to_rowids(0..6);
-        assert_eq!(row_id_range, 0..4);
-
-        let row_id_range = column.index.docid_range_to_rowids(0..6);
-        assert_eq!(row_id_range, 0..4);
-
-        let check = |range, expected| {
-            let full_range = 0..=u64::MAX;
-            let mut docids = Vec::new();
-            column.get_docids_for_value_range(full_range, range, &mut docids);
-            assert_eq!(docids, expected);
-        };
-
-        // check(0..1, vec![]);
-        // check(0..2, vec![1]);
-        check(1..2, vec![1]);
-    }
 }

columnar/src/column_index/optional_index/mod.rs

@@ -86,14 +86,8 @@ pub struct OptionalIndex {
     block_metas: Arc<[BlockMeta]>,
 }
 
-impl<'a> Iterable<u32> for &'a OptionalIndex {
-    fn boxed_iter(&self) -> Box<dyn Iterator<Item = u32> + '_> {
-        Box::new(self.iter_rows())
-    }
-}
-
 impl std::fmt::Debug for OptionalIndex {
-    fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
         f.debug_struct("OptionalIndex")
             .field("num_rows", &self.num_rows)
             .field("num_non_null_rows", &self.num_non_null_rows)
@@ -202,7 +196,6 @@ impl Set<RowId> for OptionalIndex {
         } = row_addr_from_row_id(doc_id);
         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(in_block_row_id),
             Block::Sparse(sparse_block) => sparse_block.rank(in_block_row_id),

columnar/src/column_index/optional_index/set.rs

@@ -28,11 +28,10 @@ pub trait Set<T> {
     /// Returns true if the elements is contained in the Set
     fn contains(&self, el: T) -> bool;
 
-    /// Returns the element's rank (its position in the set).
-    /// If the set does not contain the element, it will return the next existing elements rank.
+    /// Returns the number of rows in the set that are < `el`
     fn rank(&self, el: T) -> T;
 
-    /// If the set contains `el`, returns the element's rank (its position in the set).
+    /// If the set contains `el` returns the element rank.
     /// If the set does not contain the element, it returns `None`.
     fn rank_if_exists(&self, el: T) -> Option<T>;
 

columnar/src/column_index/optional_index/set_block/tests.rs

@@ -22,8 +22,8 @@ fn test_set_helper<C: SetCodec<Item = u16>>(vals: &[u16]) -> usize {
             vals.iter().cloned().take_while(|v| *v < val).count() as u16
         );
     }
-    for (rank, val) in vals.iter().enumerate() {
-        assert_eq!(tested_set.select(rank as u16), *val);
+    for rank in 0..vals.len() {
+        assert_eq!(tested_set.select(rank as u16), vals[rank]);
     }
     buffer.len()
 }
@@ -107,41 +107,3 @@ fn test_simple_translate_codec_idx_to_original_idx_dense() {
         assert_eq!(i, select_cursor.select(i));
     }
 }
-
-#[test]
-fn test_simple_translate_idx_to_value_idx_dense() {
-    let mut buffer = Vec::new();
-    DenseBlockCodec::serialize([1, 10].iter().copied(), &mut buffer).unwrap();
-    let tested_set = DenseBlockCodec::open(buffer.as_slice());
-    assert!(tested_set.contains(1));
-    assert!(!tested_set.contains(2));
-    assert_eq!(tested_set.rank(0), 0);
-    assert_eq!(tested_set.rank(1), 0);
-    for rank in 2..10 {
-        // ranks that don't exist select the next highest one
-        assert_eq!(tested_set.rank_if_exists(rank), None);
-        assert_eq!(tested_set.rank(rank), 1);
-    }
-    assert_eq!(tested_set.rank(10), 1);
-}
-
-#[test]
-fn test_simple_translate_idx_to_value_idx_sparse() {
-    let mut buffer = Vec::new();
-    SparseBlockCodec::serialize([1, 10].iter().copied(), &mut buffer).unwrap();
-    let tested_set = SparseBlockCodec::open(buffer.as_slice());
-    assert!(tested_set.contains(1));
-    assert!(!tested_set.contains(2));
-    assert_eq!(tested_set.rank(0), 0);
-    assert_eq!(tested_set.select(tested_set.rank(0)), 1);
-    assert_eq!(tested_set.rank(1), 0);
-    assert_eq!(tested_set.select(tested_set.rank(1)), 1);
-    for rank in 2..10 {
-        // ranks that don't exist select the next highest one
-        assert_eq!(tested_set.rank_if_exists(rank), None);
-        assert_eq!(tested_set.rank(rank), 1);
-        assert_eq!(tested_set.select(tested_set.rank(rank)), 10);
-    }
-    assert_eq!(tested_set.rank(10), 1);
-    assert_eq!(tested_set.select(tested_set.rank(10)), 10);
-}

columnar/src/column_index/optional_index/tests.rs

@@ -15,7 +15,9 @@ 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, &mut buffer).unwrap();
+    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();

columnar/src/column_index/serialize.rs

@@ -3,39 +3,28 @@ use std::io::Write;
 
 use common::{CountingWriter, OwnedBytes};
 
-use super::multivalued_index::SerializableMultivalueIndex;
-use super::OptionalIndex;
 use crate::column_index::multivalued_index::serialize_multivalued_index;
 use crate::column_index::optional_index::serialize_optional_index;
 use crate::column_index::ColumnIndex;
 use crate::iterable::Iterable;
-use crate::{Cardinality, RowId, Version};
-
-pub struct SerializableOptionalIndex<'a> {
-    pub non_null_row_ids: Box<dyn Iterable<RowId> + 'a>,
-    pub num_rows: RowId,
-}
-
-impl<'a> From<&'a OptionalIndex> for SerializableOptionalIndex<'a> {
-    fn from(optional_index: &'a OptionalIndex) -> Self {
-        SerializableOptionalIndex {
-            non_null_row_ids: Box::new(optional_index),
-            num_rows: optional_index.num_docs(),
-        }
-    }
-}
+use crate::{Cardinality, RowId};
 
 pub enum SerializableColumnIndex<'a> {
     Full,
-    Optional(SerializableOptionalIndex<'a>),
-    Multivalued(SerializableMultivalueIndex<'a>),
+    Optional {
+        non_null_row_ids: Box<dyn Iterable<RowId> + 'a>,
+        num_rows: RowId,
+    },
+    // TODO remove the Arc<dyn> apart from serialization this is not
+    // dynamic at all.
+    Multivalued(Box<dyn Iterable<RowId> + 'a>),
 }
 
 impl<'a> SerializableColumnIndex<'a> {
     pub fn get_cardinality(&self) -> Cardinality {
         match self {
             SerializableColumnIndex::Full => Cardinality::Full,
-            SerializableColumnIndex::Optional(_) => Cardinality::Optional,
+            SerializableColumnIndex::Optional { .. } => Cardinality::Optional,
             SerializableColumnIndex::Multivalued(_) => Cardinality::Multivalued,
         }
     }
@@ -51,12 +40,12 @@ pub fn serialize_column_index(
     output.write_all(&[cardinality])?;
     match column_index {
         SerializableColumnIndex::Full => {}
-        SerializableColumnIndex::Optional(SerializableOptionalIndex {
+        SerializableColumnIndex::Optional {
             non_null_row_ids,
             num_rows,
-        }) => serialize_optional_index(non_null_row_ids.as_ref(), num_rows, &mut output)?,
+        } => serialize_optional_index(non_null_row_ids.as_ref(), num_rows, &mut output)?,
         SerializableColumnIndex::Multivalued(multivalued_index) => {
-            serialize_multivalued_index(&multivalued_index, &mut output)?
+            serialize_multivalued_index(&*multivalued_index, &mut output)?
         }
     }
     let column_index_num_bytes = output.written_bytes() as u32;
@@ -64,10 +53,7 @@ pub fn serialize_column_index(
 }
 
 /// Open a serialized column index.
-pub fn open_column_index(
-    mut bytes: OwnedBytes,
-    format_version: Version,
-) -> io::Result<ColumnIndex> {
+pub fn open_column_index(mut bytes: OwnedBytes) -> io::Result<ColumnIndex> {
     if bytes.is_empty() {
         return Err(io::Error::new(
             io::ErrorKind::UnexpectedEof,
@@ -84,8 +70,7 @@ pub fn open_column_index(
             Ok(ColumnIndex::Optional(optional_index))
         }
         Cardinality::Multivalued => {
-            let multivalue_index =
-                super::multivalued_index::open_multivalued_index(bytes, format_version)?;
+            let multivalue_index = super::multivalued_index::open_multivalued_index(bytes)?;
             Ok(ColumnIndex::Multivalued(multivalue_index))
         }
     }

columnar/src/columnar/format_version.rs

@@ -1,6 +1,3 @@
-use core::fmt;
-use std::fmt::{Display, Formatter};
-
 use crate::InvalidData;
 
 pub const VERSION_FOOTER_NUM_BYTES: usize = MAGIC_BYTES.len() + std::mem::size_of::<u32>();
@@ -11,7 +8,7 @@ const MAGIC_BYTES: [u8; 4] = [2, 113, 119, 66];
 
 pub fn footer() -> [u8; VERSION_FOOTER_NUM_BYTES] {
     let mut footer_bytes = [0u8; VERSION_FOOTER_NUM_BYTES];
-    footer_bytes[0..4].copy_from_slice(&CURRENT_VERSION.to_bytes());
+    footer_bytes[0..4].copy_from_slice(&Version::V1.to_bytes());
     footer_bytes[4..8].copy_from_slice(&MAGIC_BYTES[..]);
     footer_bytes
 }
@@ -23,22 +20,10 @@ pub fn parse_footer(footer_bytes: [u8; VERSION_FOOTER_NUM_BYTES]) -> Result<Vers
     Version::try_from_bytes(footer_bytes[0..4].try_into().unwrap())
 }
 
-pub const CURRENT_VERSION: Version = Version::V2;
-
 #[derive(Debug, Copy, Clone, Eq, PartialEq)]
 #[repr(u32)]
 pub enum Version {
     V1 = 1u32,
-    V2 = 2u32,
-}
-
-impl Display for Version {
-    fn fmt(&self, f: &mut Formatter) -> fmt::Result {
-        match self {
-            Version::V1 => write!(f, "v1"),
-            Version::V2 => write!(f, "v2"),
-        }
-    }
 }
 
 impl Version {
@@ -50,7 +35,6 @@ impl Version {
         let code = u32::from_le_bytes(bytes);
         match code {
             1u32 => Ok(Version::V1),
-            2u32 => Ok(Version::V2),
             _ => Err(InvalidData),
         }
     }
@@ -63,9 +47,9 @@ mod tests {
     use super::*;
 
     #[test]
-    fn test_footer_deserialization() {
+    fn test_footer_dserialization() {
         let parsed_version: Version = parse_footer(footer()).unwrap();
-        assert_eq!(Version::V2, parsed_version);
+        assert_eq!(Version::V1, parsed_version);
     }
 
     #[test]
@@ -79,10 +63,11 @@ mod tests {
         for &i in &version_to_tests {
             let version_res = Version::try_from_bytes(i.to_le_bytes());
             if let Ok(version) = version_res {
+                assert_eq!(version, Version::V1);
                 assert_eq!(version.to_bytes(), i.to_le_bytes());
                 valid_versions.insert(i);
             }
         }
-        assert_eq!(valid_versions.len(), 2);
+        assert_eq!(valid_versions.len(), 1);
     }
 }

columnar/src/columnar/merge/mod.rs

@@ -7,6 +7,7 @@ use std::io;
 use std::net::Ipv6Addr;
 use std::sync::Arc;
 
+use itertools::Itertools;
 pub use merge_mapping::{MergeRowOrder, ShuffleMergeOrder, StackMergeOrder};
 
 use super::writer::ColumnarSerializer;
@@ -370,8 +371,20 @@ fn is_empty_after_merge(
                         true
                     }
                     ColumnIndex::Multivalued(multivalued_index) => {
-                        for alive_docid in alive_bitset.iter() {
-                            if !multivalued_index.range(alive_docid).is_empty() {
+                        for (doc_id, (start_index, end_index)) in multivalued_index
+                            .start_index_column
+                            .iter()
+                            .tuple_windows()
+                            .enumerate()
+                        {
+                            let doc_id = doc_id as u32;
+                            if start_index == end_index {
+                                // There are no values in this document
+                                continue;
+                            }
+                            // The document contains values and is present in the alive bitset.
+                            // The column is therefore not empty.
+                            if alive_bitset.contains(doc_id) {
                                 return false;
                             }
                         }

columnar/src/columnar/merge/tests.rs

@@ -1,5 +1,3 @@
-use itertools::Itertools;
-
 use super::*;
 use crate::{Cardinality, ColumnarWriter, HasAssociatedColumnType, RowId};
 
@@ -14,7 +12,7 @@ fn make_columnar<T: Into<NumericalValue> + HasAssociatedColumnType + Copy>(
     }
     let mut buffer: Vec<u8> = Vec::new();
     dataframe_writer
-        .serialize(vals.len() as RowId, &mut buffer)
+        .serialize(vals.len() as RowId, None, &mut buffer)
         .unwrap();
     ColumnarReader::open(buffer).unwrap()
 }
@@ -159,7 +157,9 @@ fn make_numerical_columnar_multiple_columns(
         .max()
         .unwrap_or(0u32);
     let mut buffer: Vec<u8> = Vec::new();
-    dataframe_writer.serialize(num_rows, &mut buffer).unwrap();
+    dataframe_writer
+        .serialize(num_rows, None, &mut buffer)
+        .unwrap();
     ColumnarReader::open(buffer).unwrap()
 }
 
@@ -182,7 +182,9 @@ fn make_byte_columnar_multiple_columns(
         }
     }
     let mut buffer: Vec<u8> = Vec::new();
-    dataframe_writer.serialize(num_rows, &mut buffer).unwrap();
+    dataframe_writer
+        .serialize(num_rows, None, &mut buffer)
+        .unwrap();
     ColumnarReader::open(buffer).unwrap()
 }
 
@@ -201,7 +203,9 @@ fn make_text_columnar_multiple_columns(columns: &[(&str, &[&[&str]])]) -> Column
         .max()
         .unwrap_or(0u32);
     let mut buffer: Vec<u8> = Vec::new();
-    dataframe_writer.serialize(num_rows, &mut buffer).unwrap();
+    dataframe_writer
+        .serialize(num_rows, None, &mut buffer)
+        .unwrap();
     ColumnarReader::open(buffer).unwrap()
 }
 

columnar/src/columnar/mod.rs

@@ -5,7 +5,6 @@ mod reader;
 mod writer;
 
 pub use column_type::{ColumnType, HasAssociatedColumnType};
-pub use format_version::{Version, CURRENT_VERSION};
 #[cfg(test)]
 pub(crate) use merge::ColumnTypeCategory;
 pub use merge::{merge_columnar, MergeRowOrder, ShuffleMergeOrder, StackMergeOrder};

columnar/src/columnar/reader/mod.rs

@@ -6,7 +6,7 @@ use sstable::{Dictionary, RangeSSTable};
 
 use crate::columnar::{format_version, ColumnType};
 use crate::dynamic_column::DynamicColumnHandle;
-use crate::{RowId, Version};
+use crate::RowId;
 
 fn io_invalid_data(msg: String) -> io::Error {
     io::Error::new(io::ErrorKind::InvalidData, msg)
@@ -19,7 +19,6 @@ pub struct ColumnarReader {
     column_dictionary: Dictionary<RangeSSTable>,
     column_data: FileSlice,
     num_rows: RowId,
-    format_version: Version,
 }
 
 impl fmt::Debug for ColumnarReader {
@@ -54,7 +53,6 @@ impl fmt::Debug for ColumnarReader {
 fn read_all_columns_in_stream(
     mut stream: sstable::Streamer<'_, RangeSSTable>,
     column_data: &FileSlice,
-    format_version: Version,
 ) -> io::Result<Vec<DynamicColumnHandle>> {
     let mut results = Vec::new();
     while stream.advance() {
@@ -69,7 +67,6 @@ fn read_all_columns_in_stream(
         let dynamic_column_handle = DynamicColumnHandle {
             file_slice,
             column_type,
-            format_version,
         };
         results.push(dynamic_column_handle);
     }
@@ -91,7 +88,7 @@ impl ColumnarReader {
         let num_rows = u32::deserialize(&mut &footer_bytes[8..12])?;
         let version_footer_bytes: [u8; format_version::VERSION_FOOTER_NUM_BYTES] =
             footer_bytes[12..].try_into().unwrap();
-        let format_version = format_version::parse_footer(version_footer_bytes)?;
+        let _version = format_version::parse_footer(version_footer_bytes)?;
         let (column_data, sstable) =
             file_slice_without_sstable_len.split_from_end(sstable_len as usize);
         let column_dictionary = Dictionary::open(sstable)?;
@@ -99,7 +96,6 @@ impl ColumnarReader {
             column_dictionary,
             column_data,
             num_rows,
-            format_version,
         })
     }
 
@@ -130,7 +126,6 @@ impl ColumnarReader {
                 let column_handle = DynamicColumnHandle {
                     file_slice,
                     column_type,
-                    format_version: self.format_version,
                 };
                 Some((column_name, column_handle))
             } else {
@@ -172,7 +167,7 @@ impl ColumnarReader {
             .stream_for_column_range(column_name)
             .into_stream_async()
             .await?;
-        read_all_columns_in_stream(stream, &self.column_data, self.format_version)
+        read_all_columns_in_stream(stream, &self.column_data)
     }
 
     /// Get all columns for the given column name.
@@ -181,7 +176,7 @@ impl ColumnarReader {
     /// different types.
     pub fn read_columns(&self, column_name: &str) -> io::Result<Vec<DynamicColumnHandle>> {
         let stream = self.stream_for_column_range(column_name).into_stream()?;
-        read_all_columns_in_stream(stream, &self.column_data, self.format_version)
+        read_all_columns_in_stream(stream, &self.column_data)
     }
 
     /// Return the number of columns in the columnar.
@@ -200,7 +195,7 @@ mod tests {
         columnar_writer.record_column_type("col1", ColumnType::Str, false);
         columnar_writer.record_column_type("col2", ColumnType::U64, false);
         let mut buffer = Vec::new();
-        columnar_writer.serialize(1, &mut buffer).unwrap();
+        columnar_writer.serialize(1, None, &mut buffer).unwrap();
         let columnar = ColumnarReader::open(buffer).unwrap();
         let columns = columnar.list_columns().unwrap();
         assert_eq!(columns.len(), 2);
@@ -216,7 +211,7 @@ mod tests {
         columnar_writer.record_column_type("count", ColumnType::U64, false);
         columnar_writer.record_numerical(1, "count", 1u64);
         let mut buffer = Vec::new();
-        columnar_writer.serialize(2, &mut buffer).unwrap();
+        columnar_writer.serialize(2, None, &mut buffer).unwrap();
         let columnar = ColumnarReader::open(buffer).unwrap();
         let columns = columnar.list_columns().unwrap();
         assert_eq!(columns.len(), 1);

columnar/src/columnar/writer/column_writers.rs

@@ -41,10 +41,31 @@ impl ColumnWriter {
     pub(super) fn operation_iterator<'a, V: SymbolValue>(
         &self,
         arena: &MemoryArena,
+        old_to_new_ids_opt: Option<&[RowId]>,
         buffer: &'a mut Vec<u8>,
     ) -> impl Iterator<Item = ColumnOperation<V>> + 'a {
         buffer.clear();
         self.values.read_to_end(arena, buffer);
+        if let Some(old_to_new_ids) = old_to_new_ids_opt {
+            // TODO avoid the extra deserialization / serialization.
+            let mut sorted_ops: Vec<(RowId, ColumnOperation<V>)> = Vec::new();
+            let mut new_doc = 0u32;
+            let mut cursor = &buffer[..];
+            for op in std::iter::from_fn(|| ColumnOperation::<V>::deserialize(&mut cursor)) {
+                if let ColumnOperation::NewDoc(doc) = &op {
+                    new_doc = old_to_new_ids[*doc as usize];
+                    sorted_ops.push((new_doc, ColumnOperation::NewDoc(new_doc)));
+                } else {
+                    sorted_ops.push((new_doc, op));
+                }
+            }
+            // stable sort is crucial here.
+            sorted_ops.sort_by_key(|(new_doc_id, _)| *new_doc_id);
+            buffer.clear();
+            for (_, op) in sorted_ops {
+                buffer.extend_from_slice(op.serialize().as_ref());
+            }
+        }
         let mut cursor: &[u8] = &buffer[..];
         std::iter::from_fn(move || ColumnOperation::deserialize(&mut cursor))
     }
@@ -210,9 +231,11 @@ impl NumericalColumnWriter {
     pub(super) fn operation_iterator<'a>(
         self,
         arena: &MemoryArena,
+        old_to_new_ids: Option<&[RowId]>,
         buffer: &'a mut Vec<u8>,
     ) -> impl Iterator<Item = ColumnOperation<NumericalValue>> + 'a {
-        self.column_writer.operation_iterator(arena, buffer)
+        self.column_writer
+            .operation_iterator(arena, old_to_new_ids, buffer)
     }
 }
 
@@ -254,9 +277,11 @@ impl StrOrBytesColumnWriter {
     pub(super) fn operation_iterator<'a>(
         &self,
         arena: &MemoryArena,
+        old_to_new_ids: Option<&[RowId]>,
         byte_buffer: &'a mut Vec<u8>,
     ) -> impl Iterator<Item = ColumnOperation<UnorderedId>> + 'a {
-        self.column_writer.operation_iterator(arena, byte_buffer)
+        self.column_writer
+            .operation_iterator(arena, old_to_new_ids, byte_buffer)
     }
 }
 

columnar/src/columnar/writer/mod.rs

@@ -12,7 +12,7 @@ use common::CountingWriter;
 pub(crate) use serializer::ColumnarSerializer;
 use stacker::{Addr, ArenaHashMap, MemoryArena};
 
-use crate::column_index::{SerializableColumnIndex, SerializableOptionalIndex};
+use crate::column_index::SerializableColumnIndex;
 use crate::column_values::{MonotonicallyMappableToU128, MonotonicallyMappableToU64};
 use crate::columnar::column_type::ColumnType;
 use crate::columnar::writer::column_writers::{
@@ -43,7 +43,7 @@ struct SpareBuffers {
 /// columnar_writer.record_str(1u32 /* doc id */, "product_name", "Apple");
 /// columnar_writer.record_numerical(0u32 /* doc id */, "price", 10.5f64); //< uh oh we ended up mixing integer and floats.
 /// let mut wrt: Vec<u8> =  Vec::new();
-/// columnar_writer.serialize(2u32, &mut wrt).unwrap();
+/// columnar_writer.serialize(2u32, None, &mut wrt).unwrap();
 /// ```
 #[derive(Default)]
 pub struct ColumnarWriter {
@@ -75,6 +75,63 @@ impl ColumnarWriter {
                 .sum::<usize>()
     }
 
+    /// Returns the list of doc ids from 0..num_docs sorted by the `sort_field`
+    /// column.
+    ///
+    /// If the column is multivalued, use the first value for scoring.
+    /// If no value is associated to a specific row, the document is assigned
+    /// the lowest possible score.
+    ///
+    /// The sort applied is stable.
+    pub fn sort_order(&self, sort_field: &str, num_docs: RowId, reversed: bool) -> Vec<u32> {
+        let Some(numerical_col_writer) = self
+            .numerical_field_hash_map
+            .get::<NumericalColumnWriter>(sort_field.as_bytes())
+            .or_else(|| {
+                self.datetime_field_hash_map
+                    .get::<NumericalColumnWriter>(sort_field.as_bytes())
+            })
+        else {
+            return Vec::new();
+        };
+        let mut symbols_buffer = Vec::new();
+        let mut values = Vec::new();
+        let mut start_doc_check_fill = 0;
+        let mut current_doc_opt: Option<RowId> = None;
+        // Assumption: NewDoc will never call the same doc twice and is strictly increasing between
+        // calls
+        for op in numerical_col_writer.operation_iterator(&self.arena, None, &mut symbols_buffer) {
+            match op {
+                ColumnOperation::NewDoc(doc) => {
+                    current_doc_opt = Some(doc);
+                }
+                ColumnOperation::Value(numerical_value) => {
+                    if let Some(current_doc) = current_doc_opt {
+                        // Fill up with 0.0 since last doc
+                        values.extend((start_doc_check_fill..current_doc).map(|doc| (0.0, doc)));
+                        start_doc_check_fill = current_doc + 1;
+                        // handle multi values
+                        current_doc_opt = None;
+
+                        let score: f32 = f64::coerce(numerical_value) as f32;
+                        values.push((score, current_doc));
+                    }
+                }
+            }
+        }
+        for doc in values.len() as u32..num_docs {
+            values.push((0.0f32, doc));
+        }
+        values.sort_by(|(left_score, _), (right_score, _)| {
+            if reversed {
+                right_score.total_cmp(left_score)
+            } else {
+                left_score.total_cmp(right_score)
+            }
+        });
+        values.into_iter().map(|(_score, doc)| doc).collect()
+    }
+
     /// Records a column type. This is useful to bypass the coercion process,
     /// makes sure the empty is present in the resulting columnar, or set
     /// the `sort_values_within_row`.
@@ -245,9 +302,13 @@ impl ColumnarWriter {
             },
         );
     }
-    pub fn serialize(&mut self, num_docs: RowId, wrt: &mut dyn io::Write) -> io::Result<()> {
+    pub fn serialize(
+        &mut self,
+        num_docs: RowId,
+        old_to_new_row_ids: Option<&[RowId]>,
+        wrt: &mut dyn io::Write,
+    ) -> io::Result<()> {
         let mut serializer = ColumnarSerializer::new(wrt);
-
         let mut columns: Vec<(&[u8], ColumnType, Addr)> = self
             .numerical_field_hash_map
             .iter()
@@ -261,7 +322,7 @@ impl ColumnarWriter {
         columns.extend(
             self.bytes_field_hash_map
                 .iter()
-                .map(|(column_name, addr)| (column_name, ColumnType::Bytes, addr)),
+                .map(|(term, addr)| (term, ColumnType::Bytes, addr)),
         );
         columns.extend(
             self.str_field_hash_map
@@ -283,7 +344,6 @@ impl ColumnarWriter {
                 .iter()
                 .map(|(column_name, addr)| (column_name, ColumnType::DateTime, addr)),
         );
-        // TODO: replace JSON_END_OF_PATH with b'0' in columns
         columns.sort_unstable_by_key(|(column_name, col_type, _)| (*column_name, *col_type));
 
         let (arena, buffers, dictionaries) = (&self.arena, &mut self.buffers, &self.dictionaries);
@@ -298,7 +358,11 @@ impl ColumnarWriter {
                     serialize_bool_column(
                         cardinality,
                         num_docs,
-                        column_writer.operation_iterator(arena, &mut symbol_byte_buffer),
+                        column_writer.operation_iterator(
+                            arena,
+                            old_to_new_row_ids,
+                            &mut symbol_byte_buffer,
+                        ),
                         buffers,
                         &mut column_serializer,
                     )?;
@@ -312,7 +376,11 @@ impl ColumnarWriter {
                     serialize_ip_addr_column(
                         cardinality,
                         num_docs,
-                        column_writer.operation_iterator(arena, &mut symbol_byte_buffer),
+                        column_writer.operation_iterator(
+                            arena,
+                            old_to_new_row_ids,
+                            &mut symbol_byte_buffer,
+                        ),
                         buffers,
                         &mut column_serializer,
                     )?;
@@ -337,8 +405,11 @@ impl ColumnarWriter {
                         num_docs,
                         str_or_bytes_column_writer.sort_values_within_row,
                         dictionary_builder,
-                        str_or_bytes_column_writer
-                            .operation_iterator(arena, &mut symbol_byte_buffer),
+                        str_or_bytes_column_writer.operation_iterator(
+                            arena,
+                            old_to_new_row_ids,
+                            &mut symbol_byte_buffer,
+                        ),
                         buffers,
                         &self.arena,
                         &mut column_serializer,
@@ -356,7 +427,11 @@ impl ColumnarWriter {
                         cardinality,
                         num_docs,
                         numerical_type,
-                        numerical_column_writer.operation_iterator(arena, &mut symbol_byte_buffer),
+                        numerical_column_writer.operation_iterator(
+                            arena,
+                            old_to_new_row_ids,
+                            &mut symbol_byte_buffer,
+                        ),
                         buffers,
                         &mut column_serializer,
                     )?;
@@ -371,7 +446,11 @@ impl ColumnarWriter {
                         cardinality,
                         num_docs,
                         NumericalType::I64,
-                        column_writer.operation_iterator(arena, &mut symbol_byte_buffer),
+                        column_writer.operation_iterator(
+                            arena,
+                            old_to_new_row_ids,
+                            &mut symbol_byte_buffer,
+                        ),
                         buffers,
                         &mut column_serializer,
                     )?;
@@ -556,16 +635,16 @@ fn send_to_serialize_column_mappable_to_u128<
             let optional_index_builder = value_index_builders.borrow_optional_index_builder();
             consume_operation_iterator(op_iterator, optional_index_builder, values);
             let optional_index = optional_index_builder.finish(num_rows);
-            SerializableColumnIndex::Optional(SerializableOptionalIndex {
+            SerializableColumnIndex::Optional {
                 num_rows,
                 non_null_row_ids: Box::new(optional_index),
-            })
+            }
         }
         Cardinality::Multivalued => {
             let multivalued_index_builder = value_index_builders.borrow_multivalued_index_builder();
             consume_operation_iterator(op_iterator, multivalued_index_builder, values);
-            let serializable_multivalued_index = multivalued_index_builder.finish(num_rows);
-            SerializableColumnIndex::Multivalued(serializable_multivalued_index)
+            let multivalued_index = multivalued_index_builder.finish(num_rows);
+            SerializableColumnIndex::Multivalued(Box::new(multivalued_index))
         }
     };
     crate::column::serialize_column_mappable_to_u128(
@@ -576,6 +655,15 @@ fn send_to_serialize_column_mappable_to_u128<
     Ok(())
 }
 
+fn sort_values_within_row_in_place(multivalued_index: &[RowId], values: &mut [u64]) {
+    let mut start_index: usize = 0;
+    for end_index in multivalued_index.iter().copied() {
+        let end_index = end_index as usize;
+        values[start_index..end_index].sort_unstable();
+        start_index = end_index;
+    }
+}
+
 fn send_to_serialize_column_mappable_to_u64(
     op_iterator: impl Iterator<Item = ColumnOperation<u64>>,
     cardinality: Cardinality,
@@ -599,22 +687,19 @@ fn send_to_serialize_column_mappable_to_u64(
             let optional_index_builder = value_index_builders.borrow_optional_index_builder();
             consume_operation_iterator(op_iterator, optional_index_builder, values);
             let optional_index = optional_index_builder.finish(num_rows);
-            SerializableColumnIndex::Optional(SerializableOptionalIndex {
+            SerializableColumnIndex::Optional {
                 non_null_row_ids: Box::new(optional_index),
                 num_rows,
-            })
+            }
         }
         Cardinality::Multivalued => {
             let multivalued_index_builder = value_index_builders.borrow_multivalued_index_builder();
             consume_operation_iterator(op_iterator, multivalued_index_builder, values);
-            let serializable_multivalued_index = multivalued_index_builder.finish(num_rows);
+            let multivalued_index = multivalued_index_builder.finish(num_rows);
             if sort_values_within_row {
-                sort_values_within_row_in_place(
-                    serializable_multivalued_index.start_offsets.boxed_iter(),
-                    values,
-                );
+                sort_values_within_row_in_place(multivalued_index, values);
             }
-            SerializableColumnIndex::Multivalued(serializable_multivalued_index)
+            SerializableColumnIndex::Multivalued(Box::new(multivalued_index))
         }
     };
     crate::column::serialize_column_mappable_to_u64(
@@ -625,18 +710,6 @@ fn send_to_serialize_column_mappable_to_u64(
     Ok(())
 }
 
-fn sort_values_within_row_in_place(
-    multivalued_index: impl Iterator<Item = RowId>,
-    values: &mut [u64],
-) {
-    let mut start_index: usize = 0;
-    for end_index in multivalued_index {
-        let end_index = end_index as usize;
-        values[start_index..end_index].sort_unstable();
-        start_index = end_index;
-    }
-}
-
 fn coerce_numerical_symbol<T>(
     operation_iterator: impl Iterator<Item = ColumnOperation<NumericalValue>>,
 ) -> impl Iterator<Item = ColumnOperation<u64>>
@@ -684,7 +757,7 @@ mod tests {
         assert_eq!(column_writer.get_cardinality(3), Cardinality::Full);
         let mut buffer = Vec::new();
         let symbols: Vec<ColumnOperation<NumericalValue>> = column_writer
-            .operation_iterator(&arena, &mut buffer)
+            .operation_iterator(&arena, None, &mut buffer)
             .collect();
         assert_eq!(symbols.len(), 6);
         assert!(matches!(symbols[0], ColumnOperation::NewDoc(0u32)));
@@ -713,7 +786,7 @@ mod tests {
         assert_eq!(column_writer.get_cardinality(3), Cardinality::Optional);
         let mut buffer = Vec::new();
         let symbols: Vec<ColumnOperation<NumericalValue>> = column_writer
-            .operation_iterator(&arena, &mut buffer)
+            .operation_iterator(&arena, None, &mut buffer)
             .collect();
         assert_eq!(symbols.len(), 4);
         assert!(matches!(symbols[0], ColumnOperation::NewDoc(1u32)));
@@ -736,7 +809,7 @@ mod tests {
         assert_eq!(column_writer.get_cardinality(2), Cardinality::Optional);
         let mut buffer = Vec::new();
         let symbols: Vec<ColumnOperation<NumericalValue>> = column_writer
-            .operation_iterator(&arena, &mut buffer)
+            .operation_iterator(&arena, None, &mut buffer)
             .collect();
         assert_eq!(symbols.len(), 2);
         assert!(matches!(symbols[0], ColumnOperation::NewDoc(0u32)));
@@ -755,7 +828,7 @@ mod tests {
         assert_eq!(column_writer.get_cardinality(1), Cardinality::Multivalued);
         let mut buffer = Vec::new();
         let symbols: Vec<ColumnOperation<NumericalValue>> = column_writer
-            .operation_iterator(&arena, &mut buffer)
+            .operation_iterator(&arena, None, &mut buffer)
             .collect();
         assert_eq!(symbols.len(), 3);
         assert!(matches!(symbols[0], ColumnOperation::NewDoc(0u32)));

columnar/src/columnar/writer/serializer.rs

@@ -1,7 +1,6 @@
 use std::io;
 use std::io::Write;
 
-use common::json_path_writer::JSON_END_OF_PATH;
 use common::{BinarySerializable, CountingWriter};
 use sstable::value::RangeValueWriter;
 use sstable::RangeSSTable;
@@ -19,8 +18,13 @@ pub struct ColumnarSerializer<W: io::Write> {
 /// code.
 fn prepare_key(key: &[u8], column_type: ColumnType, buffer: &mut Vec<u8>) {
     buffer.clear();
-    buffer.extend_from_slice(key);
-    buffer.push(JSON_END_OF_PATH);
+    // Convert 0 bytes to '0' string, as 0 bytes are reserved for the end of the path.
+    if key.contains(&0u8) {
+        buffer.extend(key.iter().map(|&b| if b == 0 { b'0' } else { b }));
+    } else {
+        buffer.extend_from_slice(key);
+    }
+    buffer.push(0u8);
     buffer.push(column_type.to_code());
 }
 

columnar/src/columnar/writer/value_index.rs

@@ -1,4 +1,3 @@
-use crate::column_index::{SerializableMultivalueIndex, SerializableOptionalIndex};
 use crate::iterable::Iterable;
 use crate::RowId;
 
@@ -60,47 +59,31 @@ impl IndexBuilder for OptionalIndexBuilder {
 
 #[derive(Default)]
 pub struct MultivaluedIndexBuilder {
-    doc_with_values: Vec<RowId>,
-    start_offsets: Vec<u32>,
+    start_offsets: Vec<RowId>,
     total_num_vals_seen: u32,
-    current_row: RowId,
-    current_row_has_value: bool,
 }
 
 impl MultivaluedIndexBuilder {
-    pub fn finish(&mut self, num_docs: RowId) -> SerializableMultivalueIndex<'_> {
-        self.start_offsets.push(self.total_num_vals_seen);
-        let non_null_row_ids: Box<dyn Iterable<RowId>> = Box::new(&self.doc_with_values[..]);
-        SerializableMultivalueIndex {
-            doc_ids_with_values: SerializableOptionalIndex {
-                non_null_row_ids,
-                num_rows: num_docs,
-            },
-            start_offsets: Box::new(&self.start_offsets[..]),
-        }
+    pub fn finish(&mut self, num_docs: RowId) -> &[u32] {
+        self.start_offsets
+            .resize(num_docs as usize + 1, self.total_num_vals_seen);
+        &self.start_offsets[..]
     }
 
     fn reset(&mut self) {
-        self.doc_with_values.clear();
         self.start_offsets.clear();
+        self.start_offsets.push(0u32);
         self.total_num_vals_seen = 0;
-        self.current_row = 0;
-        self.current_row_has_value = false;
     }
 }
 
 impl IndexBuilder for MultivaluedIndexBuilder {
     fn record_row(&mut self, row_id: RowId) {
-        self.current_row = row_id;
-        self.current_row_has_value = false;
+        self.start_offsets
+            .resize(row_id as usize + 1, self.total_num_vals_seen);
     }
 
     fn record_value(&mut self) {
-        if !self.current_row_has_value {
-            self.current_row_has_value = true;
-            self.doc_with_values.push(self.current_row);
-            self.start_offsets.push(self.total_num_vals_seen);
-        }
         self.total_num_vals_seen += 1;
     }
 }
@@ -158,32 +141,6 @@ mod tests {
         );
     }
 
-    #[test]
-    fn test_multivalued_value_index_builder_simple() {
-        let mut multivalued_value_index_builder = MultivaluedIndexBuilder::default();
-        {
-            multivalued_value_index_builder.record_row(0u32);
-            multivalued_value_index_builder.record_value();
-            multivalued_value_index_builder.record_value();
-            let serialized_multivalue_index = multivalued_value_index_builder.finish(1u32);
-            let start_offsets: Vec<u32> = serialized_multivalue_index
-                .start_offsets
-                .boxed_iter()
-                .collect();
-            assert_eq!(&start_offsets, &[0, 2]);
-        }
-        multivalued_value_index_builder.reset();
-        multivalued_value_index_builder.record_row(0u32);
-        multivalued_value_index_builder.record_value();
-        multivalued_value_index_builder.record_value();
-        let serialized_multivalue_index = multivalued_value_index_builder.finish(1u32);
-        let start_offsets: Vec<u32> = serialized_multivalue_index
-            .start_offsets
-            .boxed_iter()
-            .collect();
-        assert_eq!(&start_offsets, &[0, 2]);
-    }
-
     #[test]
     fn test_multivalued_value_index_builder() {
         let mut multivalued_value_index_builder = MultivaluedIndexBuilder::default();
@@ -192,15 +149,17 @@ mod tests {
         multivalued_value_index_builder.record_value();
         multivalued_value_index_builder.record_row(2u32);
         multivalued_value_index_builder.record_value();
-        let SerializableMultivalueIndex {
-            doc_ids_with_values,
-            start_offsets,
-        } = multivalued_value_index_builder.finish(4u32);
-        assert_eq!(doc_ids_with_values.num_rows, 4u32);
-        let doc_ids_with_values: Vec<u32> =
-            doc_ids_with_values.non_null_row_ids.boxed_iter().collect();
-        assert_eq!(&doc_ids_with_values, &[1u32, 2u32]);
-        let start_offsets: Vec<u32> = start_offsets.boxed_iter().collect();
-        assert_eq!(&start_offsets[..], &[0, 2, 3]);
+        assert_eq!(
+            multivalued_value_index_builder.finish(4u32).to_vec(),
+            vec![0, 0, 2, 3, 3]
+        );
+        multivalued_value_index_builder.reset();
+        multivalued_value_index_builder.record_row(2u32);
+        multivalued_value_index_builder.record_value();
+        multivalued_value_index_builder.record_value();
+        assert_eq!(
+            multivalued_value_index_builder.finish(4u32).to_vec(),
+            vec![0, 0, 0, 2, 2]
+        );
     }
 }

columnar/src/compat_tests.rs

@@ -1,183 +0,0 @@
-use std::path::PathBuf;
-
-use itertools::Itertools;
-
-use crate::{
-    merge_columnar, Cardinality, Column, ColumnarReader, DynamicColumn, StackMergeOrder,
-    CURRENT_VERSION,
-};
-
-const NUM_DOCS: u32 = u16::MAX as u32;
-
-fn generate_columnar(num_docs: u32, value_offset: u64) -> Vec<u8> {
-    use crate::ColumnarWriter;
-
-    let mut columnar_writer = ColumnarWriter::default();
-
-    for i in 0..num_docs {
-        if i % 100 == 0 {
-            columnar_writer.record_numerical(i, "sparse", value_offset + i as u64);
-        }
-        if i % 5 == 0 {
-            columnar_writer.record_numerical(i, "dense", value_offset + i as u64);
-        }
-        columnar_writer.record_numerical(i, "full", value_offset + i as u64);
-        columnar_writer.record_numerical(i, "multi", value_offset + i as u64);
-        columnar_writer.record_numerical(i, "multi", value_offset + i as u64);
-    }
-
-    let mut wrt: Vec<u8> = Vec::new();
-    columnar_writer.serialize(num_docs, &mut wrt).unwrap();
-
-    wrt
-}
-
-#[test]
-/// Writes a columnar for the CURRENT_VERSION to disk.
-fn create_format() {
-    let version = CURRENT_VERSION.to_string();
-    let file_path = path_for_version(&version);
-    if PathBuf::from(file_path.clone()).exists() {
-        return;
-    }
-    let columnar = generate_columnar(NUM_DOCS, 0);
-    std::fs::write(file_path, columnar).unwrap();
-}
-
-fn path_for_version(version: &str) -> String {
-    format!("./compat_tests_data/{}.columnar", version)
-}
-
-#[test]
-fn test_format_v1() {
-    let path = path_for_version("v1");
-    test_format(&path);
-}
-
-#[test]
-fn test_format_v2() {
-    let path = path_for_version("v2");
-    test_format(&path);
-}
-
-fn test_format(path: &str) {
-    let file_content = std::fs::read(path).unwrap();
-    let reader = ColumnarReader::open(file_content).unwrap();
-
-    check_columns(&reader);
-
-    // Test merge
-    let reader2 = ColumnarReader::open(generate_columnar(NUM_DOCS, NUM_DOCS as u64)).unwrap();
-    let columnar_readers = vec![&reader, &reader2];
-    let merge_row_order = StackMergeOrder::stack(&columnar_readers[..]);
-    let mut out = Vec::new();
-    merge_columnar(&columnar_readers, &[], merge_row_order.into(), &mut out).unwrap();
-    let reader = ColumnarReader::open(out).unwrap();
-    check_columns(&reader);
-}
-
-fn check_columns(reader: &ColumnarReader) {
-    let column = open_column(reader, "full");
-    check_column(&column, |doc_id| vec![(doc_id, doc_id as u64).into()]);
-    assert_eq!(column.get_cardinality(), Cardinality::Full);
-
-    let column = open_column(reader, "multi");
-    check_column(&column, |doc_id| {
-        vec![
-            (doc_id * 2, doc_id as u64).into(),
-            (doc_id * 2 + 1, doc_id as u64).into(),
-        ]
-    });
-    assert_eq!(column.get_cardinality(), Cardinality::Multivalued);
-
-    let column = open_column(reader, "sparse");
-    check_column(&column, |doc_id| {
-        if doc_id % 100 == 0 {
-            vec![(doc_id / 100, doc_id as u64).into()]
-        } else {
-            vec![]
-        }
-    });
-    assert_eq!(column.get_cardinality(), Cardinality::Optional);
-
-    let column = open_column(reader, "dense");
-    check_column(&column, |doc_id| {
-        if doc_id % 5 == 0 {
-            vec![(doc_id / 5, doc_id as u64).into()]
-        } else {
-            vec![]
-        }
-    });
-    assert_eq!(column.get_cardinality(), Cardinality::Optional);
-}
-
-struct RowIdAndValue {
-    row_id: u32,
-    value: u64,
-}
-impl From<(u32, u64)> for RowIdAndValue {
-    fn from((row_id, value): (u32, u64)) -> Self {
-        Self { row_id, value }
-    }
-}
-
-fn check_column<F: Fn(u32) -> Vec<RowIdAndValue>>(column: &Column<u64>, expected: F) {
-    let num_docs = column.num_docs();
-    let test_doc = |doc: u32| {
-        if expected(doc).is_empty() {
-            assert_eq!(column.first(doc), None);
-        } else {
-            assert_eq!(column.first(doc), Some(expected(doc)[0].value));
-        }
-        let values = column.values_for_doc(doc).collect_vec();
-        assert_eq!(values, expected(doc).iter().map(|x| x.value).collect_vec());
-        let mut row_ids = Vec::new();
-        column.row_ids_for_docs(&[doc], &mut vec![], &mut row_ids);
-        assert_eq!(
-            row_ids,
-            expected(doc).iter().map(|x| x.row_id).collect_vec()
-        );
-        let values = column.values_for_doc(doc).collect_vec();
-        assert_eq!(values, expected(doc).iter().map(|x| x.value).collect_vec());
-
-        // Docid rowid conversion
-        let mut row_ids = Vec::new();
-        let safe_next_doc = |doc: u32| (doc + 1).min(num_docs - 1);
-        column
-            .index
-            .docids_to_rowids(&[doc, safe_next_doc(doc)], &mut vec![], &mut row_ids);
-        let expected_rowids = expected(doc)
-            .iter()
-            .map(|x| x.row_id)
-            .chain(expected(safe_next_doc(doc)).iter().map(|x| x.row_id))
-            .collect_vec();
-        assert_eq!(row_ids, expected_rowids);
-        let rowid_range = column
-            .index
-            .docid_range_to_rowids(doc..safe_next_doc(doc) + 1);
-        if expected_rowids.is_empty() {
-            assert!(rowid_range.is_empty());
-        } else {
-            assert_eq!(
-                rowid_range,
-                expected_rowids[0]..expected_rowids.last().unwrap() + 1
-            );
-        }
-    };
-    test_doc(0);
-    test_doc(num_docs - 1);
-    test_doc(num_docs - 2);
-    test_doc(65000);
-}
-
-fn open_column(reader: &ColumnarReader, name: &str) -> Column<u64> {
-    let column = reader.read_columns(name).unwrap()[0]
-        .open()
-        .unwrap()
-        .coerce_numerical(crate::NumericalType::U64)
-        .unwrap();
-    let DynamicColumn::U64(column) = column else {
-        panic!();
-    };
-    column
-}

columnar/src/dynamic_column.rs

@@ -8,7 +8,7 @@ use common::{ByteCount, DateTime, HasLen, OwnedBytes};
 use crate::column::{BytesColumn, Column, StrColumn};
 use crate::column_values::{monotonic_map_column, StrictlyMonotonicFn};
 use crate::columnar::ColumnType;
-use crate::{Cardinality, ColumnIndex, ColumnValues, NumericalType, Version};
+use crate::{Cardinality, ColumnIndex, ColumnValues, NumericalType};
 
 #[derive(Clone)]
 pub enum DynamicColumn {
@@ -232,7 +232,6 @@ static_dynamic_conversions!(Column<Ipv6Addr>, IpAddr);
 pub struct DynamicColumnHandle {
     pub(crate) file_slice: FileSlice,
     pub(crate) column_type: ColumnType,
-    pub(crate) format_version: Version,
 }
 
 impl DynamicColumnHandle {
@@ -261,15 +260,11 @@ impl DynamicColumnHandle {
         let column_bytes = self.file_slice.read_bytes()?;
         match self.column_type {
             ColumnType::Str | ColumnType::Bytes => {
-                let column: BytesColumn =
-                    crate::column::open_column_bytes(column_bytes, self.format_version)?;
+                let column: BytesColumn = crate::column::open_column_bytes(column_bytes)?;
                 Ok(Some(column.term_ord_column))
             }
             ColumnType::IpAddr => {
-                let column = crate::column::open_column_u128_as_compact_u64(
-                    column_bytes,
-                    self.format_version,
-                )?;
+                let column = crate::column::open_column_u128_as_compact_u64(column_bytes)?;
                 Ok(Some(column))
             }
             ColumnType::Bool
@@ -277,8 +272,7 @@ impl DynamicColumnHandle {
             | ColumnType::U64
             | ColumnType::F64
             | ColumnType::DateTime => {
-                let column =
-                    crate::column::open_column_u64::<u64>(column_bytes, self.format_version)?;
+                let column = crate::column::open_column_u64::<u64>(column_bytes)?;
                 Ok(Some(column))
             }
         }
@@ -286,31 +280,15 @@ impl DynamicColumnHandle {
 
     fn open_internal(&self, column_bytes: OwnedBytes) -> io::Result<DynamicColumn> {
         let dynamic_column: DynamicColumn = match self.column_type {
-            ColumnType::Bytes => {
-                crate::column::open_column_bytes(column_bytes, self.format_version)?.into()
-            }
-            ColumnType::Str => {
-                crate::column::open_column_str(column_bytes, self.format_version)?.into()
-            }
-            ColumnType::I64 => {
-                crate::column::open_column_u64::<i64>(column_bytes, self.format_version)?.into()
-            }
-            ColumnType::U64 => {
-                crate::column::open_column_u64::<u64>(column_bytes, self.format_version)?.into()
-            }
-            ColumnType::F64 => {
-                crate::column::open_column_u64::<f64>(column_bytes, self.format_version)?.into()
-            }
-            ColumnType::Bool => {
-                crate::column::open_column_u64::<bool>(column_bytes, self.format_version)?.into()
-            }
-            ColumnType::IpAddr => {
-                crate::column::open_column_u128::<Ipv6Addr>(column_bytes, self.format_version)?
-                    .into()
-            }
+            ColumnType::Bytes => crate::column::open_column_bytes(column_bytes)?.into(),
+            ColumnType::Str => crate::column::open_column_str(column_bytes)?.into(),
+            ColumnType::I64 => crate::column::open_column_u64::<i64>(column_bytes)?.into(),
+            ColumnType::U64 => crate::column::open_column_u64::<u64>(column_bytes)?.into(),
+            ColumnType::F64 => crate::column::open_column_u64::<f64>(column_bytes)?.into(),
+            ColumnType::Bool => crate::column::open_column_u64::<bool>(column_bytes)?.into(),
+            ColumnType::IpAddr => crate::column::open_column_u128::<Ipv6Addr>(column_bytes)?.into(),
             ColumnType::DateTime => {
-                crate::column::open_column_u64::<DateTime>(column_bytes, self.format_version)?
-                    .into()
+                crate::column::open_column_u64::<DateTime>(column_bytes)?.into()
             }
         };
         Ok(dynamic_column)

columnar/src/iterable.rs

@@ -1,7 +1,4 @@
 use std::ops::Range;
-use std::sync::Arc;
-
-use crate::{ColumnValues, RowId};
 
 pub trait Iterable<T = u64> {
     fn boxed_iter(&self) -> Box<dyn Iterator<Item = T> + '_>;
@@ -20,9 +17,3 @@ where Range<T>: Iterator<Item = T>
         Box::new(self.clone())
     }
 }
-
-impl Iterable for Arc<dyn crate::ColumnValues<RowId>> {
-    fn boxed_iter(&self) -> Box<dyn Iterator<Item = u64> + '_> {
-        Box::new(self.iter().map(|row_id| row_id as u64))
-    }
-}

columnar/src/lib.rs

@@ -48,7 +48,7 @@ pub use column_values::{
 };
 pub use columnar::{
     merge_columnar, ColumnType, ColumnarReader, ColumnarWriter, HasAssociatedColumnType,
-    MergeRowOrder, ShuffleMergeOrder, StackMergeOrder, Version, CURRENT_VERSION,
+    MergeRowOrder, ShuffleMergeOrder, StackMergeOrder,
 };
 use sstable::VoidSSTable;
 pub use value::{NumericalType, NumericalValue};
@@ -131,6 +131,3 @@ impl Cardinality {
 
 #[cfg(test)]
 mod tests;
-
-#[cfg(test)]
-mod compat_tests;

columnar/src/tests.rs

@@ -21,7 +21,7 @@ fn test_dataframe_writer_str() {
     dataframe_writer.record_str(1u32, "my_string", "hello");
     dataframe_writer.record_str(3u32, "my_string", "helloeee");
     let mut buffer: Vec<u8> = Vec::new();
-    dataframe_writer.serialize(5, &mut buffer).unwrap();
+    dataframe_writer.serialize(5, None, &mut buffer).unwrap();
     let columnar = ColumnarReader::open(buffer).unwrap();
     assert_eq!(columnar.num_columns(), 1);
     let cols: Vec<DynamicColumnHandle> = columnar.read_columns("my_string").unwrap();
@@ -35,7 +35,7 @@ fn test_dataframe_writer_bytes() {
     dataframe_writer.record_bytes(1u32, "my_string", b"hello");
     dataframe_writer.record_bytes(3u32, "my_string", b"helloeee");
     let mut buffer: Vec<u8> = Vec::new();
-    dataframe_writer.serialize(5, &mut buffer).unwrap();
+    dataframe_writer.serialize(5, None, &mut buffer).unwrap();
     let columnar = ColumnarReader::open(buffer).unwrap();
     assert_eq!(columnar.num_columns(), 1);
     let cols: Vec<DynamicColumnHandle> = columnar.read_columns("my_string").unwrap();
@@ -49,7 +49,7 @@ fn test_dataframe_writer_bool() {
     dataframe_writer.record_bool(1u32, "bool.value", false);
     dataframe_writer.record_bool(3u32, "bool.value", true);
     let mut buffer: Vec<u8> = Vec::new();
-    dataframe_writer.serialize(5, &mut buffer).unwrap();
+    dataframe_writer.serialize(5, None, &mut buffer).unwrap();
     let columnar = ColumnarReader::open(buffer).unwrap();
     assert_eq!(columnar.num_columns(), 1);
     let cols: Vec<DynamicColumnHandle> = columnar.read_columns("bool.value").unwrap();
@@ -74,12 +74,12 @@ fn test_dataframe_writer_u64_multivalued() {
     dataframe_writer.record_numerical(6u32, "divisor", 2u64);
     dataframe_writer.record_numerical(6u32, "divisor", 3u64);
     let mut buffer: Vec<u8> = Vec::new();
-    dataframe_writer.serialize(7, &mut buffer).unwrap();
+    dataframe_writer.serialize(7, None, &mut buffer).unwrap();
     let columnar = ColumnarReader::open(buffer).unwrap();
     assert_eq!(columnar.num_columns(), 1);
     let cols: Vec<DynamicColumnHandle> = columnar.read_columns("divisor").unwrap();
     assert_eq!(cols.len(), 1);
-    assert_eq!(cols[0].num_bytes(), 50);
+    assert_eq!(cols[0].num_bytes(), 29);
     let dyn_i64_col = cols[0].open().unwrap();
     let DynamicColumn::I64(divisor_col) = dyn_i64_col else {
         panic!();
@@ -97,7 +97,7 @@ fn test_dataframe_writer_ip_addr() {
     dataframe_writer.record_ip_addr(1, "ip_addr", Ipv6Addr::from_u128(1001));
     dataframe_writer.record_ip_addr(3, "ip_addr", Ipv6Addr::from_u128(1050));
     let mut buffer: Vec<u8> = Vec::new();
-    dataframe_writer.serialize(5, &mut buffer).unwrap();
+    dataframe_writer.serialize(5, None, &mut buffer).unwrap();
     let columnar = ColumnarReader::open(buffer).unwrap();
     assert_eq!(columnar.num_columns(), 1);
     let cols: Vec<DynamicColumnHandle> = columnar.read_columns("ip_addr").unwrap();
@@ -128,7 +128,7 @@ fn test_dataframe_writer_numerical() {
     dataframe_writer.record_numerical(2u32, "srical.value", NumericalValue::U64(13u64));
     dataframe_writer.record_numerical(4u32, "srical.value", NumericalValue::U64(15u64));
     let mut buffer: Vec<u8> = Vec::new();
-    dataframe_writer.serialize(6, &mut buffer).unwrap();
+    dataframe_writer.serialize(6, None, &mut buffer).unwrap();
     let columnar = ColumnarReader::open(buffer).unwrap();
     assert_eq!(columnar.num_columns(), 1);
     let cols: Vec<DynamicColumnHandle> = columnar.read_columns("srical.value").unwrap();
@@ -153,6 +153,46 @@ fn test_dataframe_writer_numerical() {
     assert_eq!(column_i64.first(6), None); //< we can change the spec for that one.
 }
 
+#[test]
+fn test_dataframe_sort_by_full() {
+    let mut dataframe_writer = ColumnarWriter::default();
+    dataframe_writer.record_numerical(0u32, "value", NumericalValue::U64(1));
+    dataframe_writer.record_numerical(1u32, "value", NumericalValue::U64(2));
+    let data = dataframe_writer.sort_order("value", 2, false);
+    assert_eq!(data, vec![0, 1]);
+}
+
+#[test]
+fn test_dataframe_sort_by_opt() {
+    let mut dataframe_writer = ColumnarWriter::default();
+    dataframe_writer.record_numerical(1u32, "value", NumericalValue::U64(3));
+    dataframe_writer.record_numerical(3u32, "value", NumericalValue::U64(2));
+    let data = dataframe_writer.sort_order("value", 5, false);
+    // 0, 2, 4 is 0.0
+    assert_eq!(data, vec![0, 2, 4, 3, 1]);
+    let data = dataframe_writer.sort_order("value", 5, true);
+    assert_eq!(
+        data,
+        vec![4, 2, 0, 3, 1].into_iter().rev().collect::<Vec<_>>()
+    );
+}
+
+#[test]
+fn test_dataframe_sort_by_multi() {
+    let mut dataframe_writer = ColumnarWriter::default();
+    // valid for sort
+    dataframe_writer.record_numerical(1u32, "value", NumericalValue::U64(2));
+    // those are ignored for sort
+    dataframe_writer.record_numerical(1u32, "value", NumericalValue::U64(4));
+    dataframe_writer.record_numerical(1u32, "value", NumericalValue::U64(4));
+    // valid for sort
+    dataframe_writer.record_numerical(3u32, "value", NumericalValue::U64(3));
+    // ignored, would change sort order
+    dataframe_writer.record_numerical(3u32, "value", NumericalValue::U64(1));
+    let data = dataframe_writer.sort_order("value", 4, false);
+    assert_eq!(data, vec![0, 2, 1, 3]);
+}
+
 #[test]
 fn test_dictionary_encoded_str() {
     let mut buffer = Vec::new();
@@ -161,7 +201,7 @@ fn test_dictionary_encoded_str() {
     columnar_writer.record_str(3, "my.column", "c");
     columnar_writer.record_str(3, "my.column2", "different_column!");
     columnar_writer.record_str(4, "my.column", "b");
-    columnar_writer.serialize(5, &mut buffer).unwrap();
+    columnar_writer.serialize(5, None, &mut buffer).unwrap();
     let columnar_reader = ColumnarReader::open(buffer).unwrap();
     assert_eq!(columnar_reader.num_columns(), 2);
     let col_handles = columnar_reader.read_columns("my.column").unwrap();
@@ -195,7 +235,7 @@ fn test_dictionary_encoded_bytes() {
     columnar_writer.record_bytes(3, "my.column", b"c");
     columnar_writer.record_bytes(3, "my.column2", b"different_column!");
     columnar_writer.record_bytes(4, "my.column", b"b");
-    columnar_writer.serialize(5, &mut buffer).unwrap();
+    columnar_writer.serialize(5, None, &mut buffer).unwrap();
     let columnar_reader = ColumnarReader::open(buffer).unwrap();
     assert_eq!(columnar_reader.num_columns(), 2);
     let col_handles = columnar_reader.read_columns("my.column").unwrap();
@@ -304,7 +344,7 @@ fn column_value_strategy() -> impl Strategy<Value = ColumnValue> {
             ip_addr_byte
         ))),
         1 => any::<bool>().prop_map(ColumnValue::Bool),
-        1 => (679_723_993i64..1_679_723_995i64)
+        1 => (0_679_723_993i64..1_679_723_995i64)
             .prop_map(|val| { ColumnValue::DateTime(DateTime::from_timestamp_secs(val)) })
     ]
 }
@@ -329,12 +369,26 @@ fn columnar_docs_strategy() -> impl Strategy<Value = Vec<Vec<(&'static str, Colu
         .prop_flat_map(|num_docs| proptest::collection::vec(doc_strategy(), num_docs))
 }
 
+fn columnar_docs_and_mapping_strategy(
+) -> impl Strategy<Value = (Vec<Vec<(&'static str, ColumnValue)>>, Vec<RowId>)> {
+    columnar_docs_strategy().prop_flat_map(|docs| {
+        permutation_strategy(docs.len()).prop_map(move |permutation| (docs.clone(), permutation))
+    })
+}
+
+fn permutation_strategy(n: usize) -> impl Strategy<Value = Vec<RowId>> {
+    Just((0u32..n as RowId).collect()).prop_shuffle()
+}
+
 fn permutation_and_subset_strategy(n: usize) -> impl Strategy<Value = Vec<usize>> {
     let vals: Vec<usize> = (0..n).collect();
     subsequence(vals, 0..=n).prop_shuffle()
 }
 
-fn build_columnar_with_mapping(docs: &[Vec<(&'static str, ColumnValue)>]) -> ColumnarReader {
+fn build_columnar_with_mapping(
+    docs: &[Vec<(&'static str, ColumnValue)>],
+    old_to_new_row_ids_opt: Option<&[RowId]>,
+) -> ColumnarReader {
     let num_docs = docs.len() as u32;
     let mut buffer = Vec::new();
     let mut columnar_writer = ColumnarWriter::default();
@@ -362,13 +416,15 @@ fn build_columnar_with_mapping(docs: &[Vec<(&'static str, ColumnValue)>]) -> Col
             }
         }
     }
-    columnar_writer.serialize(num_docs, &mut buffer).unwrap();
+    columnar_writer
+        .serialize(num_docs, old_to_new_row_ids_opt, &mut buffer)
+        .unwrap();
 
     ColumnarReader::open(buffer).unwrap()
 }
 
 fn build_columnar(docs: &[Vec<(&'static str, ColumnValue)>]) -> ColumnarReader {
-    build_columnar_with_mapping(docs)
+    build_columnar_with_mapping(docs, None)
 }
 
 fn assert_columnar_eq_strict(left: &ColumnarReader, right: &ColumnarReader) {
@@ -392,7 +448,6 @@ fn assert_columnar_eq(
     }
 }
 
-#[track_caller]
 fn assert_column_eq<T: Copy + PartialOrd + Debug + Send + Sync + 'static>(
     left: &Column<T>,
     right: &Column<T>,
@@ -628,6 +683,54 @@ proptest! {
     }
 }
 
+// Same as `test_single_columnar_builder_proptest` but with a shuffling mapping.
+proptest! {
+    #![proptest_config(ProptestConfig::with_cases(500))]
+    #[test]
+    fn test_single_columnar_builder_with_shuffle_proptest((docs, mapping) in columnar_docs_and_mapping_strategy()) {
+        let columnar = build_columnar_with_mapping(&docs[..], Some(&mapping));
+        assert_eq!(columnar.num_rows() as usize, docs.len());
+        let mut expected_columns: HashMap<(&str, ColumnTypeCategory), HashMap<u32, Vec<&ColumnValue>> > = Default::default();
+        for (doc_id, doc_vals) in docs.iter().enumerate() {
+            for (col_name, col_val) in doc_vals {
+                expected_columns
+                    .entry((col_name, col_val.column_type_category()))
+                    .or_default()
+                    .entry(mapping[doc_id])
+                    .or_default()
+                    .push(col_val);
+            }
+        }
+        let column_list = columnar.list_columns().unwrap();
+        assert_eq!(expected_columns.len(), column_list.len());
+        for (column_name, column) in column_list {
+            let dynamic_column = column.open().unwrap();
+            let col_category: ColumnTypeCategory = dynamic_column.column_type().into();
+            let expected_col_values: &HashMap<u32, Vec<&ColumnValue>> = expected_columns.get(&(column_name.as_str(), col_category)).unwrap();
+            for _doc_id in 0..columnar.num_rows() {
+                match &dynamic_column {
+                    DynamicColumn::Bool(col) =>
+                        assert_column_values(col, expected_col_values),
+                    DynamicColumn::I64(col) =>
+                        assert_column_values(col, expected_col_values),
+                    DynamicColumn::U64(col) =>
+                        assert_column_values(col, expected_col_values),
+                    DynamicColumn::F64(col) =>
+                        assert_column_values(col, expected_col_values),
+                    DynamicColumn::IpAddr(col) =>
+                        assert_column_values(col, expected_col_values),
+                    DynamicColumn::DateTime(col) =>
+                        assert_column_values(col, expected_col_values),
+                    DynamicColumn::Bytes(col) =>
+                        assert_bytes_column_values(col, expected_col_values, false),
+                    DynamicColumn::Str(col) =>
+                        assert_bytes_column_values(col, expected_col_values, true),
+                }
+            }
+        }
+    }
+}
+
 // This tests create 2 or 3 random small columnar and attempts to merge them.
 // It compares the resulting merged dataframe with what would have been obtained by building the
 // dataframe from the concatenated rows to begin with.
@@ -741,68 +844,24 @@ fn columnar_docs_and_remap(
 proptest! {
     #![proptest_config(ProptestConfig::with_cases(1000))]
     #[test]
-    fn test_columnar_merge_and_remap_proptest((columnar_docs, shuffle_merge_order) in
-columnar_docs_and_remap()) {
-        test_columnar_merge_and_remap(columnar_docs, shuffle_merge_order);
+    fn test_columnar_merge_and_remap_proptest((columnar_docs, shuffle_merge_order) in columnar_docs_and_remap()) {
+        let shuffled_rows: Vec<Vec<(&'static str, ColumnValue)>> = shuffle_merge_order.iter()
+            .map(|row_addr| columnar_docs[row_addr.segment_ord as usize][row_addr.row_id as usize].clone())
+            .collect();
+        let expected_merged_columnar = build_columnar(&shuffled_rows[..]);
+        let columnar_readers: Vec<ColumnarReader> = columnar_docs.iter()
+            .map(|docs| build_columnar(&docs[..]))
+            .collect::<Vec<_>>();
+        let columnar_readers_arr: Vec<&ColumnarReader> = columnar_readers.iter().collect();
+        let mut output: Vec<u8> = Vec::new();
+        let segment_num_rows: Vec<RowId> = columnar_docs.iter().map(|docs| docs.len() as RowId).collect();
+        let shuffle_merge_order = ShuffleMergeOrder::for_test(&segment_num_rows, shuffle_merge_order);
+        crate::merge_columnar(&columnar_readers_arr[..], &[], shuffle_merge_order.into(), &mut output).unwrap();
+        let merged_columnar = ColumnarReader::open(output).unwrap();
+        assert_columnar_eq(&merged_columnar, &expected_merged_columnar, true);
     }
 }
 
-fn test_columnar_merge_and_remap(
-    columnar_docs: Vec<Vec<Vec<(&'static str, ColumnValue)>>>,
-    shuffle_merge_order: Vec<RowAddr>,
-) {
-    let shuffled_rows: Vec<Vec<(&'static str, ColumnValue)>> = shuffle_merge_order
-        .iter()
-        .map(|row_addr| {
-            columnar_docs[row_addr.segment_ord as usize][row_addr.row_id as usize].clone()
-        })
-        .collect();
-    let expected_merged_columnar = build_columnar(&shuffled_rows[..]);
-    let columnar_readers: Vec<ColumnarReader> = columnar_docs
-        .iter()
-        .map(|docs| build_columnar(&docs[..]))
-        .collect::<Vec<_>>();
-    let columnar_readers_ref: Vec<&ColumnarReader> = columnar_readers.iter().collect();
-    let mut output: Vec<u8> = Vec::new();
-    let segment_num_rows: Vec<RowId> = columnar_docs
-        .iter()
-        .map(|docs| docs.len() as RowId)
-        .collect();
-    let shuffle_merge_order = ShuffleMergeOrder::for_test(&segment_num_rows, shuffle_merge_order);
-    crate::merge_columnar(
-        &columnar_readers_ref[..],
-        &[],
-        shuffle_merge_order.into(),
-        &mut output,
-    )
-    .unwrap();
-    let merged_columnar = ColumnarReader::open(output).unwrap();
-    assert_columnar_eq(&merged_columnar, &expected_merged_columnar, true);
-}
-
-#[test]
-fn test_columnar_merge_and_remap_bug_1() {
-    let columnar_docs = vec![vec![
-        vec![
-            ("c1", ColumnValue::Numerical(NumericalValue::U64(0))),
-            ("c1", ColumnValue::Numerical(NumericalValue::U64(0))),
-        ],
-        vec![],
-    ]];
-    let shuffle_merge_order: Vec<RowAddr> = vec![
-        RowAddr {
-            segment_ord: 0,
-            row_id: 1,
-        },
-        RowAddr {
-            segment_ord: 0,
-            row_id: 0,
-        },
-    ];
-
-    test_columnar_merge_and_remap(columnar_docs, shuffle_merge_order);
-}
-
 #[test]
 fn test_columnar_merge_empty() {
     let columnar_reader_1 = build_columnar(&[]);

common/Cargo.toml

@@ -22,6 +22,3 @@ serde = { version = "1.0.136", features = ["derive"] }
 [dev-dependencies]
 proptest = "1.0.0"
 rand = "0.8.4"
-
-[features]
-unstable = [] # useful for benches.

doc/src/index_sorting.md

@@ -7,11 +7,6 @@
     - [Other](#other)
   - [Usage](#usage)
 
-# Index Sorting has been removed!
-More infos here:
-
-https://github.com/quickwit-oss/tantivy/issues/2352
-
 # Index Sorting
 
 Tantivy allows you to sort the index according to a property.

examples/date_time_field.rs

@@ -4,7 +4,7 @@
 
 use tantivy::collector::TopDocs;
 use tantivy::query::QueryParser;
-use tantivy::schema::{DateOptions, Document, Schema, Value, INDEXED, STORED, STRING};
+use tantivy::schema::{DateOptions, Document, Schema, INDEXED, STORED, STRING};
 use tantivy::{Index, IndexWriter, TantivyDocument};
 
 fn main() -> tantivy::Result<()> {
@@ -64,7 +64,6 @@ fn main() -> tantivy::Result<()> {
             assert!(retrieved_doc
                 .get_first(occurred_at)
                 .unwrap()
-                .as_value()
                 .as_datetime()
                 .is_some(),);
             assert_eq!(

examples/index_from_multiple_threads.rs

@@ -61,7 +61,7 @@ fn main() -> tantivy::Result<()> {
                         debris of the winter’s flooding; and sycamores with mottled, white, recumbent \
                         limbs and branches that arch over the pool"
                     ))?;
-            println!("add doc {i} from thread 1 - opstamp {opstamp}");
+            println!("add doc {} from thread 1 - opstamp {}", i, opstamp);
             thread::sleep(Duration::from_millis(20));
         }
         Result::<(), TantivyError>::Ok(())
@@ -82,7 +82,7 @@ fn main() -> tantivy::Result<()> {
                     body => "Some great book description..."
                 ))?
             };
-            println!("add doc {i} from thread 2 - opstamp {opstamp}");
+            println!("add doc {} from thread 2 - opstamp {}", i, opstamp);
             thread::sleep(Duration::from_millis(10));
         }
         Result::<(), TantivyError>::Ok(())

query-grammar/src/query_grammar.rs

@@ -1,4 +1,3 @@
-use std::borrow::Cow;
 use std::iter::once;
 
 use nom::branch::alt;
@@ -20,7 +19,7 @@ use crate::Occur;
 // Note: '-' char is only forbidden at the beginning of a field name, would be clearer to add it to
 // special characters.
 const SPECIAL_CHARS: &[char] = &[
-    '+', '^', '`', ':', '{', '}', '"', '\'', '[', ']', '(', ')', '!', '\\', '*', ' ',
+    '+', '^', '`', ':', '{', '}', '"', '[', ']', '(', ')', '!', '\\', '*', ' ',
 ];
 
 /// consume a field name followed by colon. Return the field name with escape sequence
@@ -42,92 +41,36 @@ fn field_name(inp: &str) -> IResult<&str, String> {
     )(inp)
 }
 
-const ESCAPE_IN_WORD: &[char] = &['^', '`', ':', '{', '}', '"', '\'', '[', ']', '(', ')', '\\'];
-
-fn interpret_escape(source: &str) -> String {
-    let mut res = String::with_capacity(source.len());
-    let mut in_escape = false;
-    let require_escape = |c: char| c.is_whitespace() || ESCAPE_IN_WORD.contains(&c) || c == '-';
-
-    for c in source.chars() {
-        if in_escape {
-            if !require_escape(c) {
-                // we re-add the escape sequence
-                res.push('\\');
-            }
-            res.push(c);
-            in_escape = false;
-        } else if c == '\\' {
-            in_escape = true;
-        } else {
-            res.push(c);
-        }
-    }
-    res
-}
-
 /// Consume a word outside of any context.
 // TODO should support escape sequences
-fn word(inp: &str) -> IResult<&str, Cow<str>> {
+fn word(inp: &str) -> IResult<&str, &str> {
     map_res(
         recognize(tuple((
-            alt((
-                preceded(char('\\'), anychar),
-                satisfy(|c| !c.is_whitespace() && !ESCAPE_IN_WORD.contains(&c) && c != '-'),
-            )),
-            many0(alt((
-                preceded(char('\\'), anychar),
-                satisfy(|c: char| !c.is_whitespace() && !ESCAPE_IN_WORD.contains(&c)),
-            ))),
+            satisfy(|c| {
+                !c.is_whitespace()
+                    && !['-', '^', '`', ':', '{', '}', '"', '[', ']', '(', ')'].contains(&c)
+            }),
+            many0(satisfy(|c: char| {
+                !c.is_whitespace() && ![':', '^', '{', '}', '"', '[', ']', '(', ')'].contains(&c)
+            })),
         ))),
         |s| match s {
             "OR" | "AND" | "NOT" | "IN" => Err(Error::new(inp, ErrorKind::Tag)),
-            s if s.contains('\\') => Ok(Cow::Owned(interpret_escape(s))),
-            s => Ok(Cow::Borrowed(s)),
+            _ => Ok(s),
         },
     )(inp)
 }
 
-fn word_infallible(
-    delimiter: &str,
-    emit_error: bool,
-) -> impl Fn(&str) -> JResult<&str, Option<Cow<str>>> + '_ {
-    // emit error is set when receiving an unescaped `:` should emit an error
-
-    move |inp| {
-        map(
-            opt_i_err(
-                preceded(
-                    multispace0,
-                    recognize(many1(alt((
-                        preceded(char::<&str, _>('\\'), anychar),
-                        satisfy(|c| !c.is_whitespace() && !delimiter.contains(c)),
-                    )))),
-                ),
-                "expected word",
+fn word_infallible(delimiter: &str) -> impl Fn(&str) -> JResult<&str, Option<&str>> + '_ {
+    |inp| {
+        opt_i_err(
+            preceded(
+                multispace0,
+                recognize(many1(satisfy(|c| {
+                    !c.is_whitespace() && !delimiter.contains(c)
+                }))),
             ),
-            |(opt_s, mut errors)| match opt_s {
-                Some(s) => {
-                    if emit_error
-                        && (s
-                            .as_bytes()
-                            .windows(2)
-                            .any(|window| window[0] != b'\\' && window[1] == b':')
-                            || s.starts_with(':'))
-                    {
-                        errors.push(LenientErrorInternal {
-                            pos: inp.len(),
-                            message: "parsed possible invalid field as term".to_string(),
-                        });
-                    }
-                    if s.contains('\\') {
-                        (Some(Cow::Owned(interpret_escape(s))), errors)
-                    } else {
-                        (Some(Cow::Borrowed(s)), errors)
-                    }
-                }
-                None => (None, errors),
-            },
+            "expected word",
         )(inp)
     }
 }
@@ -216,7 +159,7 @@ fn simple_term_infallible(
                 (value((), char('\'')), simple_quotes),
             ),
             // numbers are parsed with words in this case, as we allow string starting with a -
-            map(word_infallible(delimiter, true), |(text, errors)| {
+            map(word_infallible(delimiter), |(text, errors)| {
                 (text.map(|text| (Delimiter::None, text.to_string())), errors)
             }),
         )(inp)
@@ -379,6 +322,15 @@ fn literal_no_group_infallible(inp: &str) -> JResult<&str, Option<UserInputAst>>
         |((field_name, _, leaf), mut errors)| {
             (
                 leaf.map(|leaf| {
+                    if matches!(&leaf, UserInputLeaf::Literal(literal)
+                            if literal.phrase.contains(':') && literal.delimiter == Delimiter::None)
+                        && field_name.is_none()
+                    {
+                        errors.push(LenientErrorInternal {
+                            pos: inp.len(),
+                            message: "parsed possible invalid field as term".to_string(),
+                        });
+                    }
                     if matches!(&leaf, UserInputLeaf::Literal(literal)
                             if literal.phrase == "NOT" && literal.delimiter == Delimiter::None)
                         && field_name.is_none()
@@ -497,20 +449,20 @@ fn range_infallible(inp: &str) -> JResult<&str, UserInputLeaf> {
         tuple_infallible((
             opt_i(anychar),
             space0_infallible,
-            word_infallible("]}", false),
+            word_infallible("]}"),
             space1_infallible,
             opt_i_err(
                 terminated(tag("TO"), alt((value((), multispace1), value((), eof)))),
                 "missing keyword TO",
             ),
-            word_infallible("]}", false),
+            word_infallible("]}"),
             opt_i_err(one_of("]}"), "missing range delimiter"),
         )),
         |(
             (lower_bound_kind, _multispace0, lower, _multispace1, to, upper, upper_bound_kind),
             errs,
         )| {
-            let lower_bound = match (lower_bound_kind, lower.as_deref()) {
+            let lower_bound = match (lower_bound_kind, lower) {
                 (_, Some("*")) => UserInputBound::Unbounded,
                 (_, None) => UserInputBound::Unbounded,
                 // if it is some, TO was actually the bound (i.e. [TO TO something])
@@ -519,7 +471,7 @@ fn range_infallible(inp: &str) -> JResult<&str, UserInputLeaf> {
                 (Some('{'), Some(bound)) => UserInputBound::Exclusive(bound.to_string()),
                 _ => unreachable!("precondition failed, range did not start with [ or {{"),
             };
-            let upper_bound = match (upper_bound_kind, upper.as_deref()) {
+            let upper_bound = match (upper_bound_kind, upper) {
                 (_, Some("*")) => UserInputBound::Unbounded,
                 (_, None) => UserInputBound::Unbounded,
                 (Some(']'), Some(bound)) => UserInputBound::Inclusive(bound.to_string()),
@@ -536,7 +488,7 @@ fn range_infallible(inp: &str) -> JResult<&str, UserInputLeaf> {
             (
                 (
                     value((), tag(">=")),
-                    map(word_infallible("", false), |(bound, err)| {
+                    map(word_infallible(""), |(bound, err)| {
                         (
                             (
                                 bound
@@ -550,7 +502,7 @@ fn range_infallible(inp: &str) -> JResult<&str, UserInputLeaf> {
                 ),
                 (
                     value((), tag("<=")),
-                    map(word_infallible("", false), |(bound, err)| {
+                    map(word_infallible(""), |(bound, err)| {
                         (
                             (
                                 UserInputBound::Unbounded,
@@ -564,7 +516,7 @@ fn range_infallible(inp: &str) -> JResult<&str, UserInputLeaf> {
                 ),
                 (
                     value((), tag(">")),
-                    map(word_infallible("", false), |(bound, err)| {
+                    map(word_infallible(""), |(bound, err)| {
                         (
                             (
                                 bound
@@ -578,7 +530,7 @@ fn range_infallible(inp: &str) -> JResult<&str, UserInputLeaf> {
                 ),
                 (
                     value((), tag("<")),
-                    map(word_infallible("", false), |(bound, err)| {
+                    map(word_infallible(""), |(bound, err)| {
                         (
                             (
                                 UserInputBound::Unbounded,
@@ -1205,12 +1157,6 @@ mod test {
         test_parse_query_to_ast_helper("weight: <= 70", "\"weight\":{\"*\" TO \"70\"]");
 
         test_parse_query_to_ast_helper("weight: <= 70.5", "\"weight\":{\"*\" TO \"70.5\"]");
-
-        test_parse_query_to_ast_helper(">a", "{\"a\" TO \"*\"}");
-        test_parse_query_to_ast_helper(">=a", "[\"a\" TO \"*\"}");
-        test_parse_query_to_ast_helper("<a", "{\"*\" TO \"a\"}");
-        test_parse_query_to_ast_helper("<=a", "{\"*\" TO \"a\"]");
-        test_parse_query_to_ast_helper("<=bsd", "{\"*\" TO \"bsd\"]");
     }
 
     #[test]
@@ -1644,21 +1590,5 @@ mod test {
             r#"myfield:'hello\"happy\'tax'"#,
             r#""myfield":'hello"happy'tax'"#,
         );
-        // we don't process escape sequence for chars which don't require it
-        test_parse_query_to_ast_helper(r#"abc\*"#, r#"abc\*"#);
-    }
-
-    #[test]
-    fn test_queries_with_colons() {
-        test_parse_query_to_ast_helper(r#""abc:def""#, r#""abc:def""#);
-        test_parse_query_to_ast_helper(r#"'abc:def'"#, r#"'abc:def'"#);
-        test_parse_query_to_ast_helper(r#"abc\:def"#, r#"abc:def"#);
-        test_parse_query_to_ast_helper(r#""abc\:def""#, r#""abc:def""#);
-        test_parse_query_to_ast_helper(r#"'abc\:def'"#, r#"'abc:def'"#);
-    }
-
-    #[test]
-    fn test_invalid_field() {
-        test_is_parse_err(r#"!bc:def"#, "!bc:def");
     }
 }

src/aggregation/agg_req.rs

@@ -34,7 +34,7 @@ use super::bucket::{
     DateHistogramAggregationReq, HistogramAggregation, RangeAggregation, TermsAggregation,
 };
 use super::metric::{
-    AverageAggregation, CountAggregation, ExtendedStatsAggregation, MaxAggregation, MinAggregation,
+    AverageAggregation, CountAggregation, MaxAggregation, MinAggregation,
     PercentilesAggregationReq, StatsAggregation, SumAggregation, TopHitsAggregation,
 };
 
@@ -146,11 +146,6 @@ pub enum AggregationVariants {
     /// extracted values.
     #[serde(rename = "stats")]
     Stats(StatsAggregation),
-    /// Computes a collection of estended statistics (`min`, `max`, `sum`, `count`, `avg`,
-    /// `sum_of_squares`, `variance`, `variance_sampling`, `std_deviation`,
-    /// `std_deviation_sampling`) over the  extracted values.
-    #[serde(rename = "extended_stats")]
-    ExtendedStats(ExtendedStatsAggregation),
     /// Computes the sum of the extracted values.
     #[serde(rename = "sum")]
     Sum(SumAggregation),
@@ -175,7 +170,6 @@ impl AggregationVariants {
             AggregationVariants::Max(max) => vec![max.field_name()],
             AggregationVariants::Min(min) => vec![min.field_name()],
             AggregationVariants::Stats(stats) => vec![stats.field_name()],
-            AggregationVariants::ExtendedStats(extended_stats) => vec![extended_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(),
@@ -203,12 +197,6 @@ impl AggregationVariants {
             _ => None,
         }
     }
-    pub(crate) fn as_top_hits(&self) -> Option<&TopHitsAggregation> {
-        match &self {
-            AggregationVariants::TopHits(top_hits) => Some(top_hits),
-            _ => None,
-        }
-    }
 
     pub(crate) fn as_percentile(&self) -> Option<&PercentilesAggregationReq> {
         match &self {

src/aggregation/agg_req_with_accessor.rs

@@ -11,8 +11,8 @@ use super::bucket::{
     DateHistogramAggregationReq, HistogramAggregation, RangeAggregation, TermsAggregation,
 };
 use super::metric::{
-    AverageAggregation, CountAggregation, ExtendedStatsAggregation, MaxAggregation, MinAggregation,
-    StatsAggregation, SumAggregation,
+    AverageAggregation, CountAggregation, MaxAggregation, MinAggregation, StatsAggregation,
+    SumAggregation,
 };
 use super::segment_agg_result::AggregationLimits;
 use super::VecWithNames;
@@ -276,10 +276,6 @@ impl AggregationWithAccessor {
                 field: ref field_name,
                 ..
             })
-            | ExtendedStats(ExtendedStatsAggregation {
-                field: ref field_name,
-                ..
-            })
             | Sum(SumAggregation {
                 field: ref field_name,
                 ..
@@ -339,8 +335,8 @@ fn get_missing_val(
         }
         _ => {
             return Err(crate::TantivyError::InvalidArgument(format!(
-                "Missing value {missing:?} for field {field_name} is not supported for column \
-                 type {column_type:?}"
+                "Missing value {:?} for field {} is not supported for column type {:?}",
+                missing, field_name, column_type
             )));
         }
     };
@@ -407,7 +403,7 @@ fn get_dynamic_columns(
         .iter()
         .map(|h| h.open())
         .collect::<io::Result<_>>()?;
-    assert!(!ff_fields.is_empty(), "field {field_name} not found");
+    assert!(!ff_fields.is_empty(), "field {} not found", field_name);
     Ok(cols)
 }
 

src/aggregation/agg_result.rs

@@ -8,9 +8,7 @@ use rustc_hash::FxHashMap;
 use serde::{Deserialize, Serialize};
 
 use super::bucket::GetDocCount;
-use super::metric::{
-    ExtendedStats, PercentilesMetricResult, SingleMetricResult, Stats, TopHitsMetricResult,
-};
+use super::metric::{PercentilesMetricResult, SingleMetricResult, Stats, TopHitsMetricResult};
 use super::{AggregationError, Key};
 use crate::TantivyError;
 
@@ -90,8 +88,6 @@ pub enum MetricResult {
     Min(SingleMetricResult),
     /// Stats metric result.
     Stats(Stats),
-    /// ExtendedStats metric result.
-    ExtendedStats(Box<ExtendedStats>),
     /// Sum metric result.
     Sum(SingleMetricResult),
     /// Percentiles metric result.
@@ -108,7 +104,6 @@ impl MetricResult {
             MetricResult::Max(max) => Ok(max.value),
             MetricResult::Min(min) => Ok(min.value),
             MetricResult::Stats(stats) => stats.get_value(agg_property),
-            MetricResult::ExtendedStats(extended_stats) => extended_stats.get_value(agg_property),
             MetricResult::Sum(sum) => Ok(sum.value),
             MetricResult::Percentiles(_) => Err(TantivyError::AggregationError(
                 AggregationError::InvalidRequest("percentiles can't be used to order".to_string()),

src/aggregation/bucket/term_agg.rs

@@ -357,7 +357,8 @@ impl SegmentTermCollector {
     ) -> crate::Result<Self> {
         if field_type == ColumnType::Bytes {
             return Err(TantivyError::InvalidArgument(format!(
-                "terms aggregation is not supported for column type {field_type:?}"
+                "terms aggregation is not supported for column type {:?}",
+                field_type
             )));
         }
         let term_buckets = TermBuckets::default();

src/aggregation/intermediate_agg_result.rs

@@ -19,8 +19,8 @@ use super::bucket::{
     GetDocCount, Order, OrderTarget, RangeAggregation, TermsAggregation,
 };
 use super::metric::{
-    IntermediateAverage, IntermediateCount, IntermediateExtendedStats, IntermediateMax,
-    IntermediateMin, IntermediateStats, IntermediateSum, PercentilesCollector, TopHitsTopNComputer,
+    IntermediateAverage, IntermediateCount, IntermediateMax, IntermediateMin, IntermediateStats,
+    IntermediateSum, PercentilesCollector, TopHitsTopNComputer,
 };
 use super::segment_agg_result::AggregationLimits;
 use super::{format_date, AggregationError, Key, SerializedKey};
@@ -215,9 +215,6 @@ pub(crate) fn empty_from_req(req: &Aggregation) -> IntermediateAggregationResult
         Stats(_) => IntermediateAggregationResult::Metric(IntermediateMetricResult::Stats(
             IntermediateStats::default(),
         )),
-        ExtendedStats(_) => IntermediateAggregationResult::Metric(
-            IntermediateMetricResult::ExtendedStats(IntermediateExtendedStats::default()),
-        ),
         Sum(_) => IntermediateAggregationResult::Metric(IntermediateMetricResult::Sum(
             IntermediateSum::default(),
         )),
@@ -225,7 +222,7 @@ pub(crate) fn empty_from_req(req: &Aggregation) -> IntermediateAggregationResult
             IntermediateMetricResult::Percentiles(PercentilesCollector::default()),
         ),
         TopHits(ref req) => IntermediateAggregationResult::Metric(
-            IntermediateMetricResult::TopHits(TopHitsTopNComputer::new(req)),
+            IntermediateMetricResult::TopHits(TopHitsTopNComputer::new(req.clone())),
         ),
     }
 }
@@ -285,8 +282,6 @@ pub enum IntermediateMetricResult {
     Min(IntermediateMin),
     /// Intermediate stats result.
     Stats(IntermediateStats),
-    /// Intermediate stats result.
-    ExtendedStats(IntermediateExtendedStats),
     /// Intermediate sum result.
     Sum(IntermediateSum),
     /// Intermediate top_hits result
@@ -311,9 +306,6 @@ impl IntermediateMetricResult {
             IntermediateMetricResult::Stats(intermediate_stats) => {
                 MetricResult::Stats(intermediate_stats.finalize())
             }
-            IntermediateMetricResult::ExtendedStats(intermediate_stats) => {
-                MetricResult::ExtendedStats(intermediate_stats.finalize())
-            }
             IntermediateMetricResult::Sum(intermediate_sum) => {
                 MetricResult::Sum(intermediate_sum.finalize().into())
             }
@@ -354,12 +346,6 @@ impl IntermediateMetricResult {
             ) => {
                 stats_left.merge_fruits(stats_right);
             }
-            (
-                IntermediateMetricResult::ExtendedStats(extended_stats_left),
-                IntermediateMetricResult::ExtendedStats(extended_stats_right),
-            ) => {
-                extended_stats_left.merge_fruits(extended_stats_right);
-            }
             (IntermediateMetricResult::Sum(sum_left), IntermediateMetricResult::Sum(sum_right)) => {
                 sum_left.merge_fruits(sum_right);
             }

src/aggregation/metric/mod.rs

@@ -18,7 +18,6 @@
 
 mod average;
 mod count;
-mod extended_stats;
 mod max;
 mod min;
 mod percentiles;
@@ -30,7 +29,6 @@ use std::collections::HashMap;
 
 pub use average::*;
 pub use count::*;
-pub use extended_stats::*;
 pub use max::*;
 pub use min::*;
 pub use percentiles::*;

src/aggregation/metric/stats.rs

@@ -1,5 +1,3 @@
-use std::fmt::Debug;
-
 use serde::{Deserialize, Serialize};
 
 use super::*;
@@ -87,15 +85,13 @@ impl Stats {
 #[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
 pub struct IntermediateStats {
     /// The number of extracted values.
-    pub(crate) count: u64,
+    count: u64,
     /// The sum of the extracted values.
-    pub(crate) sum: f64,
-    /// delta for sum needed for [Kahan algorithm for summation](https://en.wikipedia.org/wiki/Kahan_summation_algorithm)
-    pub(crate) delta: f64,
+    sum: f64,
     /// The min value.
-    pub(crate) min: f64,
+    min: f64,
     /// The max value.
-    pub(crate) max: f64,
+    max: f64,
 }
 
 impl Default for IntermediateStats {
@@ -103,7 +99,6 @@ impl Default for IntermediateStats {
         Self {
             count: 0,
             sum: 0.0,
-            delta: 0.0,
             min: f64::MAX,
             max: f64::MIN,
         }
@@ -114,13 +109,7 @@ impl IntermediateStats {
     /// Merges the other stats intermediate result into self.
     pub fn merge_fruits(&mut self, other: IntermediateStats) {
         self.count += other.count;
-
-        // kahan algorithm for sum
-        let y = other.sum - (self.delta + other.delta);
-        let t = self.sum + y;
-        self.delta = (t - self.sum) - y;
-        self.sum = t;
-
+        self.sum += other.sum;
         self.min = self.min.min(other.min);
         self.max = self.max.max(other.max);
     }
@@ -152,15 +141,9 @@ impl IntermediateStats {
     }
 
     #[inline]
-    pub(in crate::aggregation::metric) fn collect(&mut self, value: f64) {
+    fn collect(&mut self, value: f64) {
         self.count += 1;
-
-        // kahan algorithm for sum
-        let y = value - self.delta;
-        let t = self.sum + y;
-        self.delta = (t - self.sum) - y;
-        self.sum = t;
-
+        self.sum += value;
         self.min = self.min.min(value);
         self.max = self.max.max(value);
     }
@@ -305,6 +288,7 @@ impl SegmentAggregationCollector for SegmentStatsCollector {
 
 #[cfg(test)]
 mod tests {
+
     use serde_json::Value;
 
     use crate::aggregation::agg_req::{Aggregation, Aggregations};

src/aggregation/metric/top_hits.rs

@@ -1,7 +1,7 @@
 use std::collections::HashMap;
 use std::net::Ipv6Addr;
 
-use columnar::{Column, ColumnType, ColumnarReader, DynamicColumn};
+use columnar::{ColumnarReader, DynamicColumn};
 use common::json_path_writer::JSON_PATH_SEGMENT_SEP_STR;
 use common::DateTime;
 use regex::Regex;
@@ -131,8 +131,8 @@ impl<'de> Deserialize<'de> for KeyOrder {
         ))?;
         if key_order.next().is_some() {
             return Err(serde::de::Error::custom(format!(
-                "Expected exactly one key-value pair in sort parameter of top_hits, found \
-                 {key_order:?}"
+                "Expected exactly one key-value pair in sort parameter of top_hits, found {:?}",
+                key_order
             )));
         }
         Ok(Self { field, order })
@@ -144,22 +144,27 @@ fn globbed_string_to_regex(glob: &str) -> Result<Regex, crate::TantivyError> {
     // 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 '{glob}' in docvalue_fields: {e}"))
+        crate::TantivyError::SchemaError(format!(
+            "Invalid regex '{}' in docvalue_fields: {}",
+            glob, e
+        ))
     })
 }
 
 fn use_doc_value_fields_err(parameter: &str) -> crate::Result<()> {
     Err(crate::TantivyError::AggregationError(
         AggregationError::InvalidRequest(format!(
-            "The `{parameter}` parameter is not supported, only `docvalue_fields` is supported in \
-             `top_hits` aggregation"
+            "The `{}` parameter is not supported, only `docvalue_fields` is supported in \
+             `top_hits` aggregation",
+            parameter
         )),
     ))
 }
 fn unsupported_err(parameter: &str) -> crate::Result<()> {
     Err(crate::TantivyError::AggregationError(
         AggregationError::InvalidRequest(format!(
-            "The `{parameter}` parameter is not supported in the `top_hits` aggregation"
+            "The `{}` parameter is not supported in the `top_hits` aggregation",
+            parameter
         )),
     ))
 }
@@ -212,7 +217,8 @@ impl TopHitsAggregation {
                     .collect::<Vec<_>>();
                 assert!(
                     !fields.is_empty(),
-                    "No fields matched the glob '{field}' in docvalue_fields"
+                    "No fields matched the glob '{}' in docvalue_fields",
+                    field
                 );
                 Ok(fields)
             })
@@ -248,7 +254,7 @@ impl TopHitsAggregation {
             .map(|field| {
                 let accessors = accessors
                     .get(field)
-                    .unwrap_or_else(|| panic!("field '{field}' not found in accessors"));
+                    .unwrap_or_else(|| panic!("field '{}' not found in accessors", field));
 
                 let values: Vec<FastFieldValue> = accessors
                     .iter()
@@ -443,10 +449,10 @@ impl std::cmp::PartialEq for TopHitsTopNComputer {
 
 impl TopHitsTopNComputer {
     /// Create a new TopHitsCollector
-    pub fn new(req: &TopHitsAggregation) -> Self {
+    pub fn new(req: TopHitsAggregation) -> Self {
         Self {
             top_n: TopNComputer::new(req.size + req.from.unwrap_or(0)),
-            req: req.clone(),
+            req,
         }
     }
 
@@ -491,6 +497,7 @@ impl TopHitsTopNComputer {
 pub(crate) struct TopHitsSegmentCollector {
     segment_ordinal: SegmentOrdinal,
     accessor_idx: usize,
+    req: TopHitsAggregation,
     top_n: TopNComputer<Vec<DocValueAndOrder>, DocAddress, false>,
 }
 
@@ -501,6 +508,7 @@ impl TopHitsSegmentCollector {
         segment_ordinal: SegmentOrdinal,
     ) -> Self {
         Self {
+            req: req.clone(),
             top_n: TopNComputer::new(req.size + req.from.unwrap_or(0)),
             segment_ordinal,
             accessor_idx,
@@ -509,13 +517,14 @@ impl TopHitsSegmentCollector {
     fn into_top_hits_collector(
         self,
         value_accessors: &HashMap<String, Vec<DynamicColumn>>,
-        req: &TopHitsAggregation,
     ) -> TopHitsTopNComputer {
-        let mut top_hits_computer = TopHitsTopNComputer::new(req);
+        let mut top_hits_computer = TopHitsTopNComputer::new(self.req.clone());
         let top_results = self.top_n.into_vec();
 
         for res in top_results {
-            let doc_value_fields = req.get_document_field_data(value_accessors, res.doc.doc_id);
+            let doc_value_fields = self
+                .req
+                .get_document_field_data(value_accessors, res.doc.doc_id);
             top_hits_computer.collect(
                 DocSortValuesAndFields {
                     sorts: res.feature,
@@ -527,15 +536,34 @@ impl TopHitsSegmentCollector {
 
         top_hits_computer
     }
+}
 
-    /// TODO add a specialized variant for a single sort field
-    fn collect_with(
+impl SegmentAggregationCollector for TopHitsSegmentCollector {
+    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 value_accessors = &agg_with_accessor.aggs.values[self.accessor_idx].value_accessors;
+
+        let intermediate_result =
+            IntermediateMetricResult::TopHits(self.into_top_hits_collector(value_accessors));
+        results.push(
+            name,
+            IntermediateAggregationResult::Metric(intermediate_result),
+        )
+    }
+
+    fn collect(
         &mut self,
         doc_id: crate::DocId,
-        req: &TopHitsAggregation,
-        accessors: &[(Column<u64>, ColumnType)],
+        agg_with_accessor: &mut crate::aggregation::agg_req_with_accessor::AggregationsWithAccessor,
     ) -> crate::Result<()> {
-        let sorts: Vec<DocValueAndOrder> = req
+        let accessors = &agg_with_accessor.aggs.values[self.accessor_idx].accessors;
+        let sorts: Vec<DocValueAndOrder> = self
+            .req
             .sort
             .iter()
             .enumerate()
@@ -560,62 +588,15 @@ impl TopHitsSegmentCollector {
         );
         Ok(())
     }
-}
-
-impl SegmentAggregationCollector for TopHitsSegmentCollector {
-    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 value_accessors = &agg_with_accessor.aggs.values[self.accessor_idx].value_accessors;
-        let tophits_req = &agg_with_accessor.aggs.values[self.accessor_idx]
-            .agg
-            .agg
-            .as_top_hits()
-            .expect("aggregation request must be of type top hits");
-
-        let intermediate_result = IntermediateMetricResult::TopHits(
-            self.into_top_hits_collector(value_accessors, tophits_req),
-        );
-        results.push(
-            name,
-            IntermediateAggregationResult::Metric(intermediate_result),
-        )
-    }
-
-    /// TODO: Consider a caching layer to reduce the call overhead
-    fn collect(
-        &mut self,
-        doc_id: crate::DocId,
-        agg_with_accessor: &mut crate::aggregation::agg_req_with_accessor::AggregationsWithAccessor,
-    ) -> crate::Result<()> {
-        let tophits_req = &agg_with_accessor.aggs.values[self.accessor_idx]
-            .agg
-            .agg
-            .as_top_hits()
-            .expect("aggregation request must be of type top hits");
-        let accessors = &agg_with_accessor.aggs.values[self.accessor_idx].accessors;
-        self.collect_with(doc_id, tophits_req, accessors)?;
-        Ok(())
-    }
 
     fn collect_block(
         &mut self,
         docs: &[crate::DocId],
         agg_with_accessor: &mut crate::aggregation::agg_req_with_accessor::AggregationsWithAccessor,
     ) -> crate::Result<()> {
-        let tophits_req = &agg_with_accessor.aggs.values[self.accessor_idx]
-            .agg
-            .agg
-            .as_top_hits()
-            .expect("aggregation request must be of type top hits");
-        let accessors = &agg_with_accessor.aggs.values[self.accessor_idx].accessors;
         // TODO: Consider getting fields with the column block accessor.
         for doc in docs {
-            self.collect_with(*doc, tophits_req, accessors)?;
+            self.collect(*doc, agg_with_accessor)?;
         }
         Ok(())
     }

src/aggregation/mod.rs

@@ -158,14 +158,15 @@ 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:?}")))?;
+    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:?}"
+            "Value is not a valid f64 (NaN or Infinity): {:?}",
+            value
         )))
     } else {
         Ok(parsed)

src/aggregation/segment_agg_result.rs

@@ -11,12 +11,12 @@ use super::agg_req_with_accessor::{AggregationWithAccessor, AggregationsWithAcce
 use super::bucket::{SegmentHistogramCollector, SegmentRangeCollector, SegmentTermCollector};
 use super::intermediate_agg_result::IntermediateAggregationResults;
 use super::metric::{
-    AverageAggregation, CountAggregation, ExtendedStatsAggregation, MaxAggregation, MinAggregation,
+    AverageAggregation, CountAggregation, MaxAggregation, MinAggregation,
     SegmentPercentilesCollector, SegmentStatsCollector, SegmentStatsType, StatsAggregation,
     SumAggregation,
 };
 use crate::aggregation::bucket::TermMissingAgg;
-use crate::aggregation::metric::{SegmentExtendedStatsCollector, TopHitsSegmentCollector};
+use crate::aggregation::metric::TopHitsSegmentCollector;
 
 pub(crate) trait SegmentAggregationCollector: CollectorClone + Debug {
     fn add_intermediate_aggregation_result(
@@ -148,9 +148,6 @@ pub(crate) fn build_single_agg_segment_collector(
             accessor_idx,
             *missing,
         ))),
-        ExtendedStats(ExtendedStatsAggregation { missing, sigma, .. }) => Ok(Box::new(
-            SegmentExtendedStatsCollector::from_req(req.field_type, *sigma, accessor_idx, *missing),
-        )),
         Sum(SumAggregation { missing, .. }) => Ok(Box::new(SegmentStatsCollector::from_req(
             req.field_type,
             SegmentStatsType::Sum,

src/collector/facet_collector.rs

@@ -598,7 +598,7 @@ mod tests {
                 let mid = n % 4;
                 n /= 4;
                 let leaf = n % 5;
-                Facet::from(&format!("/top{top}/mid{mid}/leaf{leaf}"))
+                Facet::from(&format!("/top{}/mid{}/leaf{}", top, mid, leaf))
             })
             .collect();
         for i in 0..num_facets * 10 {
@@ -737,7 +737,7 @@ mod tests {
             vec![("a", 10), ("b", 100), ("c", 7), ("d", 12), ("e", 21)]
                 .into_iter()
                 .flat_map(|(c, count)| {
-                    let facet = Facet::from(&format!("/facet/{c}"));
+                    let facet = Facet::from(&format!("/facet/{}", c));
                     let doc = doc!(facet_field => facet);
                     iter::repeat(doc).take(count)
                 })
@@ -785,7 +785,7 @@ mod tests {
         let docs: Vec<TantivyDocument> = vec![("b", 2), ("a", 2), ("c", 4)]
             .into_iter()
             .flat_map(|(c, count)| {
-                let facet = Facet::from(&format!("/facet/{c}"));
+                let facet = Facet::from(&format!("/facet/{}", c));
                 let doc = doc!(facet_field => facet);
                 iter::repeat(doc).take(count)
             })

src/collector/top_score_collector.rs

@@ -871,10 +871,7 @@ mod tests {
     use crate::schema::{Field, Schema, FAST, STORED, TEXT};
     use crate::time::format_description::well_known::Rfc3339;
     use crate::time::OffsetDateTime;
-    use crate::{
-        assert_nearly_equals, DateTime, DocAddress, DocId, Index, IndexWriter, Order, Score,
-        SegmentReader,
-    };
+    use crate::{DateTime, DocAddress, DocId, Index, IndexWriter, Order, Score, SegmentReader};
 
     fn make_index() -> crate::Result<Index> {
         let mut schema_builder = Schema::builder();

src/core/executor.rs

@@ -195,7 +195,7 @@ mod tests {
         let (tx, rx) = crossbeam_channel::bounded::<()>(0);
         let rx = Arc::new(rx);
         let executor = Executor::multi_thread(3, "search-test").unwrap();
-        for _ in 0..1000 {
+        for i in 0..1000 {
             let counter_clone: Arc<AtomicU64> = counter.clone();
             let other_counter_clone: Arc<AtomicU64> = other_counter.clone();
 
@@ -203,18 +203,18 @@ mod tests {
             let rx_clone2 = rx.clone();
             let fut = executor.spawn_blocking(move || {
                 counter_clone.fetch_add(1, Ordering::SeqCst);
-                let _ = rx_clone.recv();
+                let () = rx_clone.recv().unwrap();
             });
             futures.push(fut);
             let other_fut = executor.spawn_blocking(move || {
                 other_counter_clone.fetch_add(1, Ordering::SeqCst);
-                let _ = rx_clone2.recv();
+                let () = rx_clone2.recv().unwrap();
             });
             other_futures.push(other_fut);
         }
 
         // We execute 100 futures.
-        for _ in 0..100 {
+        for i in 0..100 {
             tx.send(()).unwrap();
         }
 
@@ -226,7 +226,7 @@ mod tests {
         drop(other_futures);
 
         // We execute 100 futures.
-        for _ in 0..100 {
+        for i in 0..100 {
             tx.send(()).unwrap();
         }
 

src/core/json_utils.rs

@@ -338,14 +338,14 @@ mod tests {
         let mut term = Term::from_field_json_path(field, "attributes.color", false);
         term.append_type_and_str("red");
         assert_eq!(
-            format!("{term:?}"),
+            format!("{:?}", term),
             "Term(field=1, type=Json, path=attributes.color, type=Str, \"red\")"
         );
 
         let mut term = Term::from_field_json_path(field, "attributes.dimensions.width", false);
         term.append_type_and_fast_value(400i64);
         assert_eq!(
-            format!("{term:?}"),
+            format!("{:?}", term),
             "Term(field=1, type=Json, path=attributes.dimensions.width, type=I64, 400)"
         );
     }

src/directory/mmap_directory.rs

@@ -566,7 +566,7 @@ mod tests {
         let mmap_directory = MmapDirectory::create_from_tempdir().unwrap();
         let num_paths = 10;
         let paths: Vec<PathBuf> = (0..num_paths)
-            .map(|i| PathBuf::from(&*format!("file_{i}")))
+            .map(|i| PathBuf::from(&*format!("file_{}", i)))
             .collect();
         {
             for path in &paths {

src/fastfield/facet_reader.rs

@@ -62,7 +62,7 @@ impl FacetReader {
 
 #[cfg(test)]
 mod tests {
-    use crate::schema::{Facet, FacetOptions, SchemaBuilder, Value, STORED};
+    use crate::schema::{Facet, FacetOptions, SchemaBuilder, STORED};
     use crate::{DocAddress, Index, IndexWriter, TantivyDocument};
 
     #[test]
@@ -88,9 +88,7 @@ mod tests {
         let doc = searcher
             .doc::<TantivyDocument>(DocAddress::new(0u32, 0u32))
             .unwrap();
-        let value = doc
-            .get_first(facet_field)
-            .and_then(|v| v.as_value().as_facet());
+        let value = doc.get_first(facet_field).and_then(|v| v.as_facet());
         assert_eq!(value, None);
     }
 

src/fastfield/mod.rs

@@ -127,7 +127,7 @@ mod tests {
             fast_field_writers
                 .add_document(&doc!(*FIELD=>2u64))
                 .unwrap();
-            fast_field_writers.serialize(&mut write).unwrap();
+            fast_field_writers.serialize(&mut write, None).unwrap();
             write.terminate().unwrap();
         }
         let file = directory.open_read(path).unwrap();
@@ -178,7 +178,7 @@ mod tests {
             fast_field_writers
                 .add_document(&doc!(*FIELD=>215u64))
                 .unwrap();
-            fast_field_writers.serialize(&mut write).unwrap();
+            fast_field_writers.serialize(&mut write, None).unwrap();
             write.terminate().unwrap();
         }
         let file = directory.open_read(path).unwrap();
@@ -211,7 +211,7 @@ mod tests {
                     .add_document(&doc!(*FIELD=>100_000u64))
                     .unwrap();
             }
-            fast_field_writers.serialize(&mut write).unwrap();
+            fast_field_writers.serialize(&mut write, None).unwrap();
             write.terminate().unwrap();
         }
         let file = directory.open_read(path).unwrap();
@@ -243,7 +243,7 @@ mod tests {
                     .add_document(&doc!(*FIELD=>5_000_000_000_000_000_000u64 + doc_id))
                     .unwrap();
             }
-            fast_field_writers.serialize(&mut write).unwrap();
+            fast_field_writers.serialize(&mut write, None).unwrap();
             write.terminate().unwrap();
         }
         let file = directory.open_read(path).unwrap();
@@ -276,7 +276,7 @@ mod tests {
                 doc.add_i64(i64_field, i);
                 fast_field_writers.add_document(&doc).unwrap();
             }
-            fast_field_writers.serialize(&mut write).unwrap();
+            fast_field_writers.serialize(&mut write, None).unwrap();
             write.terminate().unwrap();
         }
         let file = directory.open_read(path).unwrap();
@@ -315,7 +315,7 @@ mod tests {
             let mut fast_field_writers = FastFieldsWriter::from_schema(&schema).unwrap();
             let doc = TantivyDocument::default();
             fast_field_writers.add_document(&doc).unwrap();
-            fast_field_writers.serialize(&mut write).unwrap();
+            fast_field_writers.serialize(&mut write, None).unwrap();
             write.terminate().unwrap();
         }
 
@@ -348,7 +348,7 @@ mod tests {
             let mut fast_field_writers = FastFieldsWriter::from_schema(&schema).unwrap();
             let doc = TantivyDocument::default();
             fast_field_writers.add_document(&doc).unwrap();
-            fast_field_writers.serialize(&mut write).unwrap();
+            fast_field_writers.serialize(&mut write, None).unwrap();
             write.terminate().unwrap();
         }
 
@@ -385,7 +385,7 @@ mod tests {
             for &x in &permutation {
                 fast_field_writers.add_document(&doc!(*FIELD=>x)).unwrap();
             }
-            fast_field_writers.serialize(&mut write).unwrap();
+            fast_field_writers.serialize(&mut write, None).unwrap();
             write.terminate().unwrap();
         }
         let file = directory.open_read(path).unwrap();
@@ -770,7 +770,7 @@ mod tests {
             fast_field_writers
                 .add_document(&doc!(field=>false))
                 .unwrap();
-            fast_field_writers.serialize(&mut write).unwrap();
+            fast_field_writers.serialize(&mut write, None).unwrap();
             write.terminate().unwrap();
         }
         let file = directory.open_read(path).unwrap();
@@ -802,7 +802,7 @@ mod tests {
                     .add_document(&doc!(field=>false))
                     .unwrap();
             }
-            fast_field_writers.serialize(&mut write).unwrap();
+            fast_field_writers.serialize(&mut write, None).unwrap();
             write.terminate().unwrap();
         }
         let file = directory.open_read(path).unwrap();
@@ -827,7 +827,7 @@ mod tests {
             let mut fast_field_writers = FastFieldsWriter::from_schema(&schema).unwrap();
             let doc = TantivyDocument::default();
             fast_field_writers.add_document(&doc).unwrap();
-            fast_field_writers.serialize(&mut write).unwrap();
+            fast_field_writers.serialize(&mut write, None).unwrap();
             write.terminate().unwrap();
         }
         let file = directory.open_read(path).unwrap();
@@ -855,7 +855,7 @@ mod tests {
             for doc in docs {
                 fast_field_writers.add_document(doc).unwrap();
             }
-            fast_field_writers.serialize(&mut write).unwrap();
+            fast_field_writers.serialize(&mut write, None).unwrap();
             write.terminate().unwrap();
         }
         Ok(directory)

src/fastfield/writer.rs

@@ -4,6 +4,7 @@ use columnar::{ColumnarWriter, NumericalValue};
 use common::{DateTimePrecision, JsonPathWriter};
 use tokenizer_api::Token;
 
+use crate::indexer::doc_id_mapping::DocIdMapping;
 use crate::schema::document::{Document, ReferenceValue, ReferenceValueLeaf, Value};
 use crate::schema::{value_type_to_column_type, Field, FieldType, Schema, Type};
 use crate::tokenizer::{TextAnalyzer, TokenizerManager};
@@ -105,6 +106,16 @@ impl FastFieldsWriter {
         self.columnar_writer.mem_usage()
     }
 
+    pub(crate) fn sort_order(
+        &self,
+        sort_field: &str,
+        num_docs: DocId,
+        reversed: bool,
+    ) -> Vec<DocId> {
+        self.columnar_writer
+            .sort_order(sort_field, num_docs, reversed)
+    }
+
     /// Indexes all of the fastfields of a new document.
     pub fn add_document<D: Document>(&mut self, doc: &D) -> crate::Result<()> {
         let doc_id = self.num_docs;
@@ -222,9 +233,16 @@ impl FastFieldsWriter {
 
     /// Serializes all of the `FastFieldWriter`s by pushing them in
     /// order to the fast field serializer.
-    pub fn serialize(mut self, wrt: &mut dyn io::Write) -> io::Result<()> {
+    pub fn serialize(
+        mut self,
+        wrt: &mut dyn io::Write,
+        doc_id_map_opt: Option<&DocIdMapping>,
+    ) -> io::Result<()> {
         let num_docs = self.num_docs;
-        self.columnar_writer.serialize(num_docs, wrt)?;
+        let old_to_new_row_ids =
+            doc_id_map_opt.map(|doc_id_mapping| doc_id_mapping.old_to_new_ids());
+        self.columnar_writer
+            .serialize(num_docs, old_to_new_row_ids, wrt)?;
         Ok(())
     }
 }
@@ -374,7 +392,7 @@ mod tests {
         }
         let mut buffer = Vec::new();
         columnar_writer
-            .serialize(json_docs.len() as DocId, &mut buffer)
+            .serialize(json_docs.len() as DocId, None, &mut buffer)
             .unwrap();
         ColumnarReader::open(buffer).unwrap()
     }

src/fieldnorm/mod.rs

@@ -77,7 +77,7 @@ mod tests {
             let mut fieldnorm_writers = FieldNormsWriter::for_schema(&SCHEMA);
             fieldnorm_writers.record(2u32, *TXT_FIELD, 5);
             fieldnorm_writers.record(3u32, *TXT_FIELD, 3);
-            fieldnorm_writers.serialize(serializer)?;
+            fieldnorm_writers.serialize(serializer, None)?;
         }
         let file = directory.open_read(path)?;
         {

src/fieldnorm/writer.rs

@@ -2,6 +2,7 @@ use std::cmp::Ordering;
 use std::{io, iter};
 
 use super::{fieldnorm_to_id, FieldNormsSerializer};
+use crate::indexer::doc_id_mapping::DocIdMapping;
 use crate::schema::{Field, Schema};
 use crate::DocId;
 
@@ -91,7 +92,11 @@ impl FieldNormsWriter {
     }
 
     /// Serialize the seen fieldnorm values to the serializer for all fields.
-    pub fn serialize(&self, mut fieldnorms_serializer: FieldNormsSerializer) -> io::Result<()> {
+    pub fn serialize(
+        &self,
+        mut fieldnorms_serializer: FieldNormsSerializer,
+        doc_id_map: Option<&DocIdMapping>,
+    ) -> io::Result<()> {
         for (field, fieldnorms_buffer) in self.fieldnorms_buffers.iter().enumerate().filter_map(
             |(field_id, fieldnorms_buffer_opt)| {
                 fieldnorms_buffer_opt.as_ref().map(|fieldnorms_buffer| {
@@ -99,7 +104,12 @@ impl FieldNormsWriter {
                 })
             },
         ) {
-            fieldnorms_serializer.serialize_field(field, fieldnorms_buffer)?;
+            if let Some(doc_id_map) = doc_id_map {
+                let remapped_fieldnorm_buffer = doc_id_map.remap(fieldnorms_buffer);
+                fieldnorms_serializer.serialize_field(field, &remapped_fieldnorm_buffer)?;
+            } else {
+                fieldnorms_serializer.serialize_field(field, fieldnorms_buffer)?;
+            }
         }
         fieldnorms_serializer.close()?;
         Ok(())

src/functional_test.rs

@@ -7,7 +7,7 @@ use rand::{thread_rng, Rng};
 use crate::indexer::index_writer::MEMORY_BUDGET_NUM_BYTES_MIN;
 use crate::schema::*;
 #[allow(deprecated)]
-use crate::{doc, schema, Index, IndexWriter, Searcher};
+use crate::{doc, schema, Index, IndexSettings, IndexSortByField, IndexWriter, Order, Searcher};
 
 fn check_index_content(searcher: &Searcher, vals: &[u64]) -> crate::Result<()> {
     assert!(searcher.segment_readers().len() < 20);
@@ -65,6 +65,71 @@ fn get_num_iterations() -> usize {
         .map(|str| str.parse().unwrap())
         .unwrap_or(2000)
 }
+#[test]
+#[ignore]
+fn test_functional_indexing_sorted() -> crate::Result<()> {
+    let mut schema_builder = Schema::builder();
+
+    let id_field = schema_builder.add_u64_field("id", INDEXED | FAST);
+    let multiples_field = schema_builder.add_u64_field("multiples", INDEXED);
+    let text_field_options = TextOptions::default()
+        .set_indexing_options(
+            TextFieldIndexing::default()
+                .set_index_option(schema::IndexRecordOption::WithFreqsAndPositions),
+        )
+        .set_stored();
+    let text_field = schema_builder.add_text_field("text_field", text_field_options);
+    let schema = schema_builder.build();
+
+    let mut index_builder = Index::builder().schema(schema);
+    index_builder = index_builder.settings(IndexSettings {
+        sort_by_field: Some(IndexSortByField {
+            field: "id".to_string(),
+            order: Order::Desc,
+        }),
+        ..Default::default()
+    });
+    let index = index_builder.create_from_tempdir().unwrap();
+
+    let reader = index.reader()?;
+
+    let mut rng = thread_rng();
+
+    let mut index_writer: IndexWriter =
+        index.writer_with_num_threads(3, 3 * MEMORY_BUDGET_NUM_BYTES_MIN)?;
+
+    let mut committed_docs: HashSet<u64> = HashSet::new();
+    let mut uncommitted_docs: HashSet<u64> = HashSet::new();
+
+    for _ in 0..get_num_iterations() {
+        let random_val = rng.gen_range(0..20);
+        if random_val == 0 {
+            index_writer.commit()?;
+            committed_docs.extend(&uncommitted_docs);
+            uncommitted_docs.clear();
+            reader.reload()?;
+            let searcher = reader.searcher();
+            // check that everything is correct.
+            check_index_content(
+                &searcher,
+                &committed_docs.iter().cloned().collect::<Vec<u64>>(),
+            )?;
+        } else if committed_docs.remove(&random_val) || uncommitted_docs.remove(&random_val) {
+            let doc_id_term = Term::from_field_u64(id_field, random_val);
+            index_writer.delete_term(doc_id_term);
+        } else {
+            uncommitted_docs.insert(random_val);
+            let mut doc = TantivyDocument::new();
+            doc.add_u64(id_field, random_val);
+            for i in 1u64..10u64 {
+                doc.add_u64(multiples_field, random_val * i);
+            }
+            doc.add_text(text_field, get_text());
+            index_writer.add_document(doc)?;
+        }
+    }
+    Ok(())
+}
 
 const LOREM: &str = "Doc Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod \
                      tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, \

src/index/index.rs

@@ -20,7 +20,7 @@ use crate::indexer::segment_updater::save_metas;
 use crate::indexer::{IndexWriter, SingleSegmentIndexWriter};
 use crate::reader::{IndexReader, IndexReaderBuilder};
 use crate::schema::document::Document;
-use crate::schema::{Field, FieldType, Schema};
+use crate::schema::{Field, FieldType, Schema, Type};
 use crate::tokenizer::{TextAnalyzer, TokenizerManager};
 use crate::SegmentReader;
 
@@ -232,7 +232,32 @@ impl IndexBuilder {
     }
 
     fn validate(&self) -> crate::Result<()> {
-        if let Some(_schema) = self.schema.as_ref() {
+        if let Some(schema) = self.schema.as_ref() {
+            if let Some(sort_by_field) = self.index_settings.sort_by_field.as_ref() {
+                let schema_field = schema.get_field(&sort_by_field.field).map_err(|_| {
+                    TantivyError::InvalidArgument(format!(
+                        "Field to sort index {} not found in schema",
+                        sort_by_field.field
+                    ))
+                })?;
+                let entry = schema.get_field_entry(schema_field);
+                if !entry.is_fast() {
+                    return Err(TantivyError::InvalidArgument(format!(
+                        "Field {} is no fast field. Field needs to be a single value fast field \
+                         to be used to sort an index",
+                        sort_by_field.field
+                    )));
+                }
+                let supported_field_types = [Type::I64, Type::U64, Type::F64, Type::Date];
+                let field_type = entry.field_type().value_type();
+                if !supported_field_types.contains(&field_type) {
+                    return Err(TantivyError::InvalidArgument(format!(
+                        "Unsupported field type in sort_by_field: {:?}. Supported field types: \
+                         {:?} ",
+                        field_type, supported_field_types,
+                    )));
+                }
+            }
             Ok(())
         } else {
             Err(TantivyError::InvalidArgument(

src/index/index_meta.rs

@@ -249,6 +249,10 @@ fn is_true(val: &bool) -> bool {
 /// index, like presort documents.
 #[derive(Clone, Debug, Serialize, Deserialize, Eq, PartialEq)]
 pub struct IndexSettings {
+    /// Sorts the documents by information
+    /// provided in `IndexSortByField`
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub sort_by_field: Option<IndexSortByField>,
     /// The `Compressor` used to compress the doc store.
     #[serde(default)]
     pub docstore_compression: Compressor,
@@ -271,6 +275,7 @@ fn default_docstore_blocksize() -> usize {
 impl Default for IndexSettings {
     fn default() -> Self {
         Self {
+            sort_by_field: None,
             docstore_compression: Compressor::default(),
             docstore_blocksize: default_docstore_blocksize(),
             docstore_compress_dedicated_thread: true,
@@ -278,6 +283,22 @@ impl Default for IndexSettings {
     }
 }
 
+/// Settings to presort the documents in an index
+///
+/// Presorting documents can greatly improve performance
+/// in some scenarios, by applying top n
+/// optimizations.
+#[deprecated(
+    since = "0.22.0",
+    note = "We plan to remove index sorting in `0.23`. If you need index sorting, please comment on the related issue https://github.com/quickwit-oss/tantivy/issues/2352 and explain your use case."
+)]
+#[derive(Clone, Debug, Serialize, Deserialize, Eq, PartialEq)]
+pub struct IndexSortByField {
+    /// The field to sort the documents by
+    pub field: String,
+    /// The order to sort the documents by
+    pub order: Order,
+}
 /// The order to sort by
 #[derive(Clone, Debug, Serialize, Deserialize, Eq, PartialEq)]
 pub enum Order {
@@ -396,7 +417,7 @@ mod tests {
     use crate::store::Compressor;
     #[cfg(feature = "zstd-compression")]
     use crate::store::ZstdCompressor;
-    use crate::IndexSettings;
+    use crate::{IndexSettings, IndexSortByField, Order};
 
     #[test]
     fn test_serialize_metas() {
@@ -406,7 +427,13 @@ mod tests {
             schema_builder.build()
         };
         let index_metas = IndexMeta {
-            index_settings: IndexSettings::default(),
+            index_settings: IndexSettings {
+                sort_by_field: Some(IndexSortByField {
+                    field: "text".to_string(),
+                    order: Order::Asc,
+                }),
+                ..Default::default()
+            },
             segments: Vec::new(),
             schema,
             opstamp: 0u64,
@@ -415,7 +442,7 @@ mod tests {
         let json = serde_json::ser::to_string(&index_metas).expect("serialization failed");
         assert_eq!(
             json,
-            r#"{"index_settings":{"docstore_compression":"lz4","docstore_blocksize":16384},"segments":[],"schema":[{"name":"text","type":"text","options":{"indexing":{"record":"position","fieldnorms":true,"tokenizer":"default"},"stored":false,"fast":false}}],"opstamp":0}"#
+            r#"{"index_settings":{"sort_by_field":{"field":"text","order":"Asc"},"docstore_compression":"lz4","docstore_blocksize":16384},"segments":[],"schema":[{"name":"text","type":"text","options":{"indexing":{"record":"position","fieldnorms":true,"tokenizer":"default"},"stored":false,"fast":false}}],"opstamp":0}"#
         );
 
         let deser_meta: UntrackedIndexMeta = serde_json::from_str(&json).unwrap();
@@ -434,6 +461,10 @@ mod tests {
         };
         let index_metas = IndexMeta {
             index_settings: IndexSettings {
+                sort_by_field: Some(IndexSortByField {
+                    field: "text".to_string(),
+                    order: Order::Asc,
+                }),
                 docstore_compression: crate::store::Compressor::Zstd(ZstdCompressor {
                     compression_level: Some(4),
                 }),
@@ -448,7 +479,7 @@ mod tests {
         let json = serde_json::ser::to_string(&index_metas).expect("serialization failed");
         assert_eq!(
             json,
-            r#"{"index_settings":{"docstore_compression":"zstd(compression_level=4)","docstore_blocksize":1000000},"segments":[],"schema":[{"name":"text","type":"text","options":{"indexing":{"record":"position","fieldnorms":true,"tokenizer":"default"},"stored":false,"fast":false}}],"opstamp":0}"#
+            r#"{"index_settings":{"sort_by_field":{"field":"text","order":"Asc"},"docstore_compression":"zstd(compression_level=4)","docstore_blocksize":1000000},"segments":[],"schema":[{"name":"text","type":"text","options":{"indexing":{"record":"position","fieldnorms":true,"tokenizer":"default"},"stored":false,"fast":false}}],"opstamp":0}"#
         );
 
         let deser_meta: UntrackedIndexMeta = serde_json::from_str(&json).unwrap();
@@ -460,35 +491,35 @@ mod tests {
     #[test]
     #[cfg(all(feature = "lz4-compression", feature = "zstd-compression"))]
     fn test_serialize_metas_invalid_comp() {
-        let json = r#"{"index_settings":{"docstore_compression":"zsstd","docstore_blocksize":1000000},"segments":[],"schema":[{"name":"text","type":"text","options":{"indexing":{"record":"position","fieldnorms":true,"tokenizer":"default"},"stored":false,"fast":false}}],"opstamp":0}"#;
+        let json = r#"{"index_settings":{"sort_by_field":{"field":"text","order":"Asc"},"docstore_compression":"zsstd","docstore_blocksize":1000000},"segments":[],"schema":[{"name":"text","type":"text","options":{"indexing":{"record":"position","fieldnorms":true,"tokenizer":"default"},"stored":false,"fast":false}}],"opstamp":0}"#;
 
         let err = serde_json::from_str::<UntrackedIndexMeta>(json).unwrap_err();
         assert_eq!(
             err.to_string(),
             "unknown variant `zsstd`, expected one of `none`, `lz4`, `zstd`, \
-             `zstd(compression_level=5)` at line 1 column 49"
+             `zstd(compression_level=5)` at line 1 column 96"
                 .to_string()
         );
 
-        let json = r#"{"index_settings":{"docstore_compression":"zstd(bla=10)","docstore_blocksize":1000000},"segments":[],"schema":[{"name":"text","type":"text","options":{"indexing":{"record":"position","fieldnorms":true,"tokenizer":"default"},"stored":false,"fast":false}}],"opstamp":0}"#;
+        let json = r#"{"index_settings":{"sort_by_field":{"field":"text","order":"Asc"},"docstore_compression":"zstd(bla=10)","docstore_blocksize":1000000},"segments":[],"schema":[{"name":"text","type":"text","options":{"indexing":{"record":"position","fieldnorms":true,"tokenizer":"default"},"stored":false,"fast":false}}],"opstamp":0}"#;
 
         let err = serde_json::from_str::<UntrackedIndexMeta>(json).unwrap_err();
         assert_eq!(
             err.to_string(),
-            "unknown zstd option \"bla\" at line 1 column 56".to_string()
+            "unknown zstd option \"bla\" at line 1 column 103".to_string()
         );
     }
 
     #[test]
     #[cfg(not(feature = "zstd-compression"))]
     fn test_serialize_metas_unsupported_comp() {
-        let json = r#"{"index_settings":{"docstore_compression":"zstd","docstore_blocksize":1000000},"segments":[],"schema":[{"name":"text","type":"text","options":{"indexing":{"record":"position","fieldnorms":true,"tokenizer":"default"},"stored":false,"fast":false}}],"opstamp":0}"#;
+        let json = r#"{"index_settings":{"sort_by_field":{"field":"text","order":"Asc"},"docstore_compression":"zstd","docstore_blocksize":1000000},"segments":[],"schema":[{"name":"text","type":"text","options":{"indexing":{"record":"position","fieldnorms":true,"tokenizer":"default"},"stored":false,"fast":false}}],"opstamp":0}"#;
 
         let err = serde_json::from_str::<UntrackedIndexMeta>(json).unwrap_err();
         assert_eq!(
             err.to_string(),
             "unsupported variant `zstd`, please enable Tantivy's `zstd-compression` feature at \
-             line 1 column 48"
+             line 1 column 95"
                 .to_string()
         );
     }
@@ -500,6 +531,7 @@ mod tests {
         assert_eq!(
             index_settings,
             IndexSettings {
+                sort_by_field: None,
                 docstore_compression: Compressor::default(),
                 docstore_compress_dedicated_thread: true,
                 docstore_blocksize: 16_384

src/index/mod.rs

@@ -12,7 +12,7 @@ mod segment_reader;
 
 pub use self::index::{Index, IndexBuilder};
 pub(crate) use self::index_meta::SegmentMetaInventory;
-pub use self::index_meta::{IndexMeta, IndexSettings, Order, SegmentMeta};
+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;

src/index/segment_reader.rs

@@ -318,14 +318,14 @@ impl SegmentReader {
                         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 = 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}")
+                            format!("{}.{}", field_name, json_path)
                         }
                     };
                     indexed_fields.extend(

src/indexer/doc_id_mapping.rs

@@ -3,12 +3,15 @@
 
 use common::ReadOnlyBitSet;
 
-use crate::DocAddress;
+use super::SegmentWriter;
+use crate::schema::{Field, Schema};
+use crate::{DocAddress, DocId, IndexSortByField, TantivyError};
 
 #[derive(Copy, Clone, Eq, PartialEq)]
 pub enum MappingType {
     Stacked,
     StackedWithDeletes,
+    Shuffled,
 }
 
 /// Struct to provide mapping from new doc_id to old doc_id and segment.
@@ -43,4 +46,537 @@ impl SegmentDocIdMapping {
     pub(crate) fn iter_old_doc_addrs(&self) -> impl Iterator<Item = DocAddress> + '_ {
         self.new_doc_id_to_old_doc_addr.iter().copied()
     }
+
+    /// This flags means the segments are simply stacked in the order of their ordinal.
+    /// e.g. [(0, 1), .. (n, 1), (0, 2)..., (m, 2)]
+    ///
+    /// The different segment may present some deletes, in which case it is expressed by skipping a
+    /// `DocId`. [(0, 1), (0, 3)] <--- here doc_id=0 and doc_id=1 have been deleted
+    ///
+    /// Being trivial is equivalent to having the `new_doc_id_to_old_doc_addr` array sorted.
+    ///
+    /// This allows for some optimization.
+    pub(crate) fn is_trivial(&self) -> bool {
+        match self.mapping_type {
+            MappingType::Stacked | MappingType::StackedWithDeletes => true,
+            MappingType::Shuffled => false,
+        }
+    }
+}
+
+/// Struct to provide mapping from old doc_id to new doc_id and vice versa within a segment.
+pub struct DocIdMapping {
+    new_doc_id_to_old: Vec<DocId>,
+    old_doc_id_to_new: Vec<DocId>,
+}
+
+impl DocIdMapping {
+    pub fn from_new_id_to_old_id(new_doc_id_to_old: Vec<DocId>) -> Self {
+        let max_doc = new_doc_id_to_old.len();
+        let old_max_doc = new_doc_id_to_old
+            .iter()
+            .cloned()
+            .max()
+            .map(|n| n + 1)
+            .unwrap_or(0);
+        let mut old_doc_id_to_new = vec![0; old_max_doc as usize];
+        for i in 0..max_doc {
+            old_doc_id_to_new[new_doc_id_to_old[i] as usize] = i as DocId;
+        }
+        DocIdMapping {
+            new_doc_id_to_old,
+            old_doc_id_to_new,
+        }
+    }
+
+    /// returns the new doc_id for the old doc_id
+    pub fn get_new_doc_id(&self, doc_id: DocId) -> DocId {
+        self.old_doc_id_to_new[doc_id as usize]
+    }
+    /// returns the old doc_id for the new doc_id
+    pub fn get_old_doc_id(&self, doc_id: DocId) -> DocId {
+        self.new_doc_id_to_old[doc_id as usize]
+    }
+    /// iterate over old doc_ids in order of the new doc_ids
+    pub fn iter_old_doc_ids(&self) -> impl Iterator<Item = DocId> + Clone + '_ {
+        self.new_doc_id_to_old.iter().cloned()
+    }
+
+    pub fn old_to_new_ids(&self) -> &[DocId] {
+        &self.old_doc_id_to_new[..]
+    }
+
+    /// Remaps a given array to the new doc ids.
+    pub fn remap<T: Copy>(&self, els: &[T]) -> Vec<T> {
+        self.new_doc_id_to_old
+            .iter()
+            .map(|old_doc| els[*old_doc as usize])
+            .collect()
+    }
+    pub fn num_new_doc_ids(&self) -> usize {
+        self.new_doc_id_to_old.len()
+    }
+    pub fn num_old_doc_ids(&self) -> usize {
+        self.old_doc_id_to_new.len()
+    }
+}
+
+pub(crate) fn expect_field_id_for_sort_field(
+    schema: &Schema,
+    sort_by_field: &IndexSortByField,
+) -> crate::Result<Field> {
+    schema.get_field(&sort_by_field.field).map_err(|_| {
+        TantivyError::InvalidArgument(format!(
+            "field to sort index by not found: {:?}",
+            sort_by_field.field
+        ))
+    })
+}
+
+// Generates a document mapping in the form of [index new doc_id] -> old doc_id
+// TODO detect if field is already sorted and discard mapping
+pub(crate) fn get_doc_id_mapping_from_field(
+    sort_by_field: IndexSortByField,
+    segment_writer: &SegmentWriter,
+) -> crate::Result<DocIdMapping> {
+    let schema = segment_writer.segment_serializer.segment().schema();
+    expect_field_id_for_sort_field(&schema, &sort_by_field)?; // for now expect
+    let new_doc_id_to_old = segment_writer.fast_field_writers.sort_order(
+        sort_by_field.field.as_str(),
+        segment_writer.max_doc(),
+        sort_by_field.order.is_desc(),
+    );
+    // create new doc_id to old doc_id index (used in fast_field_writers)
+    Ok(DocIdMapping::from_new_id_to_old_id(new_doc_id_to_old))
+}
+
+#[cfg(test)]
+mod tests_indexsorting {
+    use common::DateTime;
+
+    use crate::collector::TopDocs;
+    use crate::indexer::doc_id_mapping::DocIdMapping;
+    use crate::indexer::NoMergePolicy;
+    use crate::query::QueryParser;
+    use crate::schema::*;
+    use crate::{DocAddress, Index, IndexBuilder, IndexSettings, IndexSortByField, Order};
+
+    fn create_test_index(
+        index_settings: Option<IndexSettings>,
+        text_field_options: TextOptions,
+    ) -> crate::Result<Index> {
+        let mut schema_builder = Schema::builder();
+
+        let my_text_field = schema_builder.add_text_field("text_field", text_field_options);
+        let my_string_field = schema_builder.add_text_field("string_field", STRING | STORED);
+        let my_number =
+            schema_builder.add_u64_field("my_number", NumericOptions::default().set_fast());
+
+        let multi_numbers =
+            schema_builder.add_u64_field("multi_numbers", NumericOptions::default().set_fast());
+
+        let schema = schema_builder.build();
+        let mut index_builder = Index::builder().schema(schema);
+        if let Some(settings) = index_settings {
+            index_builder = index_builder.settings(settings);
+        }
+        let index = index_builder.create_in_ram()?;
+
+        let mut index_writer = index.writer_for_tests()?;
+        index_writer.add_document(doc!(my_number=>40_u64))?;
+        index_writer.add_document(
+            doc!(my_number=>20_u64, multi_numbers => 5_u64, multi_numbers => 6_u64),
+        )?;
+        index_writer.add_document(doc!(my_number=>100_u64))?;
+        index_writer.add_document(
+            doc!(my_number=>10_u64, my_string_field=> "blublub", my_text_field => "some text"),
+        )?;
+        index_writer.add_document(doc!(my_number=>30_u64, multi_numbers => 3_u64 ))?;
+        index_writer.commit()?;
+        Ok(index)
+    }
+    fn get_text_options() -> TextOptions {
+        TextOptions::default().set_indexing_options(
+            TextFieldIndexing::default().set_index_option(IndexRecordOption::Basic),
+        )
+    }
+    #[test]
+    fn test_sort_index_test_text_field() -> crate::Result<()> {
+        // there are different serializers for different settings in postings/recorder.rs
+        // test remapping for all of them
+        let options = vec![
+            get_text_options(),
+            get_text_options().set_indexing_options(
+                TextFieldIndexing::default().set_index_option(IndexRecordOption::WithFreqs),
+            ),
+            get_text_options().set_indexing_options(
+                TextFieldIndexing::default()
+                    .set_index_option(IndexRecordOption::WithFreqsAndPositions),
+            ),
+        ];
+
+        for option in options {
+            // let options = get_text_options();
+            // no index_sort
+            let index = create_test_index(None, option.clone())?;
+            let my_text_field = index.schema().get_field("text_field").unwrap();
+            let searcher = index.reader()?.searcher();
+
+            let query = QueryParser::for_index(&index, vec![my_text_field]).parse_query("text")?;
+            let top_docs: Vec<(f32, DocAddress)> =
+                searcher.search(&query, &TopDocs::with_limit(3))?;
+            assert_eq!(
+                top_docs.iter().map(|el| el.1.doc_id).collect::<Vec<_>>(),
+                vec![3]
+            );
+
+            // sort by field asc
+            let index = create_test_index(
+                Some(IndexSettings {
+                    sort_by_field: Some(IndexSortByField {
+                        field: "my_number".to_string(),
+                        order: Order::Asc,
+                    }),
+                    ..Default::default()
+                }),
+                option.clone(),
+            )?;
+            let my_text_field = index.schema().get_field("text_field").unwrap();
+            let reader = index.reader()?;
+            let searcher = reader.searcher();
+
+            let query = QueryParser::for_index(&index, vec![my_text_field]).parse_query("text")?;
+            let top_docs: Vec<(f32, DocAddress)> =
+                searcher.search(&query, &TopDocs::with_limit(3))?;
+            assert_eq!(
+                top_docs.iter().map(|el| el.1.doc_id).collect::<Vec<_>>(),
+                vec![0]
+            );
+
+            // test new field norm mapping
+            {
+                let my_text_field = index.schema().get_field("text_field").unwrap();
+                let fieldnorm_reader = searcher
+                    .segment_reader(0)
+                    .get_fieldnorms_reader(my_text_field)?;
+                assert_eq!(fieldnorm_reader.fieldnorm(0), 2); // some text
+                assert_eq!(fieldnorm_reader.fieldnorm(1), 0);
+            }
+            // sort by field desc
+            let index = create_test_index(
+                Some(IndexSettings {
+                    sort_by_field: Some(IndexSortByField {
+                        field: "my_number".to_string(),
+                        order: Order::Desc,
+                    }),
+                    ..Default::default()
+                }),
+                option.clone(),
+            )?;
+            let my_string_field = index.schema().get_field("text_field").unwrap();
+            let searcher = index.reader()?.searcher();
+
+            let query =
+                QueryParser::for_index(&index, vec![my_string_field]).parse_query("text")?;
+            let top_docs: Vec<(f32, DocAddress)> =
+                searcher.search(&query, &TopDocs::with_limit(3))?;
+            assert_eq!(
+                top_docs.iter().map(|el| el.1.doc_id).collect::<Vec<_>>(),
+                vec![4]
+            );
+            // test new field norm mapping
+            {
+                let my_text_field = index.schema().get_field("text_field").unwrap();
+                let fieldnorm_reader = searcher
+                    .segment_reader(0)
+                    .get_fieldnorms_reader(my_text_field)?;
+                assert_eq!(fieldnorm_reader.fieldnorm(0), 0);
+                assert_eq!(fieldnorm_reader.fieldnorm(1), 0);
+                assert_eq!(fieldnorm_reader.fieldnorm(2), 0);
+                assert_eq!(fieldnorm_reader.fieldnorm(3), 0);
+                assert_eq!(fieldnorm_reader.fieldnorm(4), 2); // some text
+            }
+        }
+        Ok(())
+    }
+    #[test]
+    fn test_sort_index_get_documents() -> crate::Result<()> {
+        // default baseline
+        let index = create_test_index(None, get_text_options())?;
+        let my_string_field = index.schema().get_field("string_field").unwrap();
+        let searcher = index.reader()?.searcher();
+        {
+            assert!(searcher
+                .doc::<TantivyDocument>(DocAddress::new(0, 0))?
+                .get_first(my_string_field)
+                .is_none());
+            assert_eq!(
+                searcher
+                    .doc::<TantivyDocument>(DocAddress::new(0, 3))?
+                    .get_first(my_string_field)
+                    .unwrap()
+                    .as_str(),
+                Some("blublub")
+            );
+        }
+        // sort by field asc
+        let index = create_test_index(
+            Some(IndexSettings {
+                sort_by_field: Some(IndexSortByField {
+                    field: "my_number".to_string(),
+                    order: Order::Asc,
+                }),
+                ..Default::default()
+            }),
+            get_text_options(),
+        )?;
+        let my_string_field = index.schema().get_field("string_field").unwrap();
+        let searcher = index.reader()?.searcher();
+        {
+            assert_eq!(
+                searcher
+                    .doc::<TantivyDocument>(DocAddress::new(0, 0))?
+                    .get_first(my_string_field)
+                    .unwrap()
+                    .as_str(),
+                Some("blublub")
+            );
+            let doc = searcher.doc::<TantivyDocument>(DocAddress::new(0, 4))?;
+            assert!(doc.get_first(my_string_field).is_none());
+        }
+        // sort by field desc
+        let index = create_test_index(
+            Some(IndexSettings {
+                sort_by_field: Some(IndexSortByField {
+                    field: "my_number".to_string(),
+                    order: Order::Desc,
+                }),
+                ..Default::default()
+            }),
+            get_text_options(),
+        )?;
+        let my_string_field = index.schema().get_field("string_field").unwrap();
+        let searcher = index.reader()?.searcher();
+        {
+            let doc = searcher.doc::<TantivyDocument>(DocAddress::new(0, 4))?;
+            assert_eq!(
+                doc.get_first(my_string_field).unwrap().as_str(),
+                Some("blublub")
+            );
+        }
+        Ok(())
+    }
+
+    #[test]
+    fn test_sort_index_test_string_field() -> crate::Result<()> {
+        let index = create_test_index(None, get_text_options())?;
+        let my_string_field = index.schema().get_field("string_field").unwrap();
+        let searcher = index.reader()?.searcher();
+
+        let query = QueryParser::for_index(&index, vec![my_string_field]).parse_query("blublub")?;
+        let top_docs: Vec<(f32, DocAddress)> = searcher.search(&query, &TopDocs::with_limit(3))?;
+        assert_eq!(
+            top_docs.iter().map(|el| el.1.doc_id).collect::<Vec<_>>(),
+            vec![3]
+        );
+
+        let index = create_test_index(
+            Some(IndexSettings {
+                sort_by_field: Some(IndexSortByField {
+                    field: "my_number".to_string(),
+                    order: Order::Asc,
+                }),
+                ..Default::default()
+            }),
+            get_text_options(),
+        )?;
+        let my_string_field = index.schema().get_field("string_field").unwrap();
+        let reader = index.reader()?;
+        let searcher = reader.searcher();
+
+        let query = QueryParser::for_index(&index, vec![my_string_field]).parse_query("blublub")?;
+        let top_docs: Vec<(f32, DocAddress)> = searcher.search(&query, &TopDocs::with_limit(3))?;
+        assert_eq!(
+            top_docs.iter().map(|el| el.1.doc_id).collect::<Vec<_>>(),
+            vec![0]
+        );
+
+        // test new field norm mapping
+        {
+            let my_text_field = index.schema().get_field("text_field").unwrap();
+            let fieldnorm_reader = searcher
+                .segment_reader(0)
+                .get_fieldnorms_reader(my_text_field)?;
+            assert_eq!(fieldnorm_reader.fieldnorm(0), 2); // some text
+            assert_eq!(fieldnorm_reader.fieldnorm(1), 0);
+        }
+        // sort by field desc
+        let index = create_test_index(
+            Some(IndexSettings {
+                sort_by_field: Some(IndexSortByField {
+                    field: "my_number".to_string(),
+                    order: Order::Desc,
+                }),
+                ..Default::default()
+            }),
+            get_text_options(),
+        )?;
+        let my_string_field = index.schema().get_field("string_field").unwrap();
+        let searcher = index.reader()?.searcher();
+
+        let query = QueryParser::for_index(&index, vec![my_string_field]).parse_query("blublub")?;
+        let top_docs: Vec<(f32, DocAddress)> = searcher.search(&query, &TopDocs::with_limit(3))?;
+        assert_eq!(
+            top_docs.iter().map(|el| el.1.doc_id).collect::<Vec<_>>(),
+            vec![4]
+        );
+        // test new field norm mapping
+        {
+            let my_text_field = index.schema().get_field("text_field").unwrap();
+            let fieldnorm_reader = searcher
+                .segment_reader(0)
+                .get_fieldnorms_reader(my_text_field)?;
+            assert_eq!(fieldnorm_reader.fieldnorm(0), 0);
+            assert_eq!(fieldnorm_reader.fieldnorm(1), 0);
+            assert_eq!(fieldnorm_reader.fieldnorm(2), 0);
+            assert_eq!(fieldnorm_reader.fieldnorm(3), 0);
+            assert_eq!(fieldnorm_reader.fieldnorm(4), 2); // some text
+        }
+        Ok(())
+    }
+
+    #[test]
+    fn test_sort_index_fast_field() -> crate::Result<()> {
+        let index = create_test_index(
+            Some(IndexSettings {
+                sort_by_field: Some(IndexSortByField {
+                    field: "my_number".to_string(),
+                    order: Order::Asc,
+                }),
+                ..Default::default()
+            }),
+            get_text_options(),
+        )?;
+        assert_eq!(
+            index.settings().sort_by_field.as_ref().unwrap().field,
+            "my_number".to_string()
+        );
+
+        let searcher = index.reader()?.searcher();
+        assert_eq!(searcher.segment_readers().len(), 1);
+        let segment_reader = searcher.segment_reader(0);
+        let fast_fields = segment_reader.fast_fields();
+
+        let fast_field = fast_fields
+            .u64("my_number")
+            .unwrap()
+            .first_or_default_col(999);
+        assert_eq!(fast_field.get_val(0), 10u64);
+        assert_eq!(fast_field.get_val(1), 20u64);
+        assert_eq!(fast_field.get_val(2), 30u64);
+
+        let multifield = fast_fields.u64("multi_numbers").unwrap();
+        let vals: Vec<u64> = multifield.values_for_doc(0u32).collect();
+        assert_eq!(vals, &[] as &[u64]);
+        let vals: Vec<_> = multifield.values_for_doc(1u32).collect();
+        assert_eq!(vals, &[5, 6]);
+
+        let vals: Vec<_> = multifield.values_for_doc(2u32).collect();
+        assert_eq!(vals, &[3]);
+        Ok(())
+    }
+
+    #[test]
+    fn test_with_sort_by_date_field() -> crate::Result<()> {
+        let mut schema_builder = Schema::builder();
+        let date_field = schema_builder.add_date_field("date", INDEXED | STORED | FAST);
+        let schema = schema_builder.build();
+
+        let settings = IndexSettings {
+            sort_by_field: Some(IndexSortByField {
+                field: "date".to_string(),
+                order: Order::Desc,
+            }),
+            ..Default::default()
+        };
+
+        let index = Index::builder()
+            .schema(schema)
+            .settings(settings)
+            .create_in_ram()?;
+        let mut index_writer = index.writer_for_tests()?;
+        index_writer.set_merge_policy(Box::new(NoMergePolicy));
+
+        index_writer.add_document(doc!(
+            date_field => DateTime::from_timestamp_secs(1000),
+        ))?;
+        index_writer.add_document(doc!(
+            date_field => DateTime::from_timestamp_secs(999),
+        ))?;
+        index_writer.add_document(doc!(
+            date_field => DateTime::from_timestamp_secs(1001),
+        ))?;
+        index_writer.commit()?;
+
+        let searcher = index.reader()?.searcher();
+        assert_eq!(searcher.segment_readers().len(), 1);
+        let segment_reader = searcher.segment_reader(0);
+        let fast_fields = segment_reader.fast_fields();
+
+        let fast_field = fast_fields
+            .date("date")
+            .unwrap()
+            .first_or_default_col(DateTime::from_timestamp_secs(0));
+        assert_eq!(fast_field.get_val(0), DateTime::from_timestamp_secs(1001));
+        assert_eq!(fast_field.get_val(1), DateTime::from_timestamp_secs(1000));
+        assert_eq!(fast_field.get_val(2), DateTime::from_timestamp_secs(999));
+        Ok(())
+    }
+
+    #[test]
+    fn test_doc_mapping() {
+        let doc_mapping = DocIdMapping::from_new_id_to_old_id(vec![3, 2, 5]);
+        assert_eq!(doc_mapping.get_old_doc_id(0), 3);
+        assert_eq!(doc_mapping.get_old_doc_id(1), 2);
+        assert_eq!(doc_mapping.get_old_doc_id(2), 5);
+        assert_eq!(doc_mapping.get_new_doc_id(0), 0);
+        assert_eq!(doc_mapping.get_new_doc_id(1), 0);
+        assert_eq!(doc_mapping.get_new_doc_id(2), 1);
+        assert_eq!(doc_mapping.get_new_doc_id(3), 0);
+        assert_eq!(doc_mapping.get_new_doc_id(4), 0);
+        assert_eq!(doc_mapping.get_new_doc_id(5), 2);
+    }
+
+    #[test]
+    fn test_doc_mapping_remap() {
+        let doc_mapping = DocIdMapping::from_new_id_to_old_id(vec![2, 8, 3]);
+        assert_eq!(
+            &doc_mapping.remap(&[0, 1000, 2000, 3000, 4000, 5000, 6000, 7000, 8000]),
+            &[2000, 8000, 3000]
+        );
+    }
+
+    #[test]
+    fn test_text_sort() -> crate::Result<()> {
+        let mut schema_builder = SchemaBuilder::new();
+        schema_builder.add_text_field("id", STRING | FAST | STORED);
+        schema_builder.add_text_field("name", TEXT | STORED);
+
+        let resp = IndexBuilder::new()
+            .schema(schema_builder.build())
+            .settings(IndexSettings {
+                sort_by_field: Some(IndexSortByField {
+                    field: "id".to_string(),
+                    order: Order::Asc,
+                }),
+                ..Default::default()
+            })
+            .create_in_ram();
+        assert!(resp
+            .unwrap_err()
+            .to_string()
+            .contains("Unsupported field type"));
+
+        Ok(())
+    }
 }

src/indexer/merge_index_test.rs

@@ -1,147 +0,0 @@
-#[cfg(test)]
-mod tests {
-    use crate::collector::TopDocs;
-    use crate::fastfield::AliveBitSet;
-    use crate::index::Index;
-    use crate::postings::Postings;
-    use crate::query::QueryParser;
-    use crate::schema::{
-        self, BytesOptions, Facet, FacetOptions, IndexRecordOption, NumericOptions,
-        TextFieldIndexing, TextOptions,
-    };
-    use crate::{DocAddress, DocSet, IndexSettings, IndexWriter, Term};
-
-    fn create_test_index(index_settings: Option<IndexSettings>) -> crate::Result<Index> {
-        let mut schema_builder = schema::Schema::builder();
-        let int_options = NumericOptions::default()
-            .set_fast()
-            .set_stored()
-            .set_indexed();
-        let int_field = schema_builder.add_u64_field("intval", int_options);
-
-        let bytes_options = BytesOptions::default().set_fast().set_indexed();
-        let bytes_field = schema_builder.add_bytes_field("bytes", bytes_options);
-        let facet_field = schema_builder.add_facet_field("facet", FacetOptions::default());
-
-        let multi_numbers =
-            schema_builder.add_u64_field("multi_numbers", NumericOptions::default().set_fast());
-        let text_field_options = TextOptions::default()
-            .set_indexing_options(
-                TextFieldIndexing::default()
-                    .set_index_option(schema::IndexRecordOption::WithFreqsAndPositions),
-            )
-            .set_stored();
-        let text_field = schema_builder.add_text_field("text_field", text_field_options);
-        let schema = schema_builder.build();
-
-        let mut index_builder = Index::builder().schema(schema);
-        if let Some(settings) = index_settings {
-            index_builder = index_builder.settings(settings);
-        }
-        let index = index_builder.create_in_ram()?;
-
-        {
-            let mut index_writer = index.writer_for_tests()?;
-
-            // segment 1 - range 1-3
-            index_writer.add_document(doc!(int_field=>1_u64))?;
-            index_writer.add_document(
-                doc!(int_field=>3_u64, multi_numbers => 3_u64, multi_numbers => 4_u64, bytes_field => vec![1, 2, 3], text_field => "some text", facet_field=> Facet::from("/book/crime")),
-            )?;
-            index_writer.add_document(
-                doc!(int_field=>1_u64, text_field=> "deleteme",  text_field => "ok text more text"),
-            )?;
-            index_writer.add_document(
-                doc!(int_field=>2_u64, multi_numbers => 2_u64, multi_numbers => 3_u64, text_field => "ok text more text"),
-            )?;
-
-            index_writer.commit()?;
-            index_writer.add_document(doc!(int_field=>20_u64, multi_numbers => 20_u64))?;
-
-            let in_val = 1u64;
-            index_writer.add_document(doc!(int_field=>in_val, text_field=> "deleteme" , text_field => "ok text more text", facet_field=> Facet::from("/book/crime")))?;
-            index_writer.commit()?;
-            let int_vals = [10u64, 5];
-            index_writer.add_document( // position of this doc after delete in desc sorting = [2], in disjunct case [1]
-                doc!(int_field=>int_vals[0], multi_numbers => 10_u64, multi_numbers => 11_u64, text_field=> "blubber", facet_field=> Facet::from("/book/fantasy")),
-            )?;
-            index_writer.add_document(doc!(int_field=>int_vals[1], text_field=> "deleteme"))?;
-            index_writer.add_document(
-                doc!(int_field=>1_000u64, multi_numbers => 1001_u64, multi_numbers => 1002_u64, bytes_field => vec![5, 5],text_field => "the biggest num")
-            )?;
-
-            index_writer.delete_term(Term::from_field_text(text_field, "deleteme"));
-            index_writer.commit()?;
-        }
-
-        // Merging the segments
-        {
-            let segment_ids = index.searchable_segment_ids()?;
-            let mut index_writer: IndexWriter = index.writer_for_tests()?;
-            index_writer.merge(&segment_ids).wait()?;
-            index_writer.wait_merging_threads()?;
-        }
-        Ok(index)
-    }
-
-    #[test]
-    fn test_merge_index() {
-        let index = create_test_index(Some(IndexSettings {
-            ..Default::default()
-        }))
-        .unwrap();
-
-        let reader = index.reader().unwrap();
-        let searcher = reader.searcher();
-        assert_eq!(searcher.segment_readers().len(), 1);
-        let segment_reader = searcher.segment_readers().last().unwrap();
-
-        let searcher = index.reader().unwrap().searcher();
-        {
-            let my_text_field = index.schema().get_field("text_field").unwrap();
-
-            let do_search = |term: &str| {
-                let query = QueryParser::for_index(&index, vec![my_text_field])
-                    .parse_query(term)
-                    .unwrap();
-                let top_docs: Vec<(f32, DocAddress)> =
-                    searcher.search(&query, &TopDocs::with_limit(3)).unwrap();
-
-                top_docs.iter().map(|el| el.1.doc_id).collect::<Vec<_>>()
-            };
-
-            assert_eq!(do_search("some"), vec![1]);
-            assert_eq!(do_search("blubber"), vec![3]);
-            assert_eq!(do_search("biggest"), vec![4]);
-        }
-
-        // postings file
-        {
-            let my_text_field = index.schema().get_field("text_field").unwrap();
-            let term_a = Term::from_field_text(my_text_field, "text");
-            let inverted_index = segment_reader.inverted_index(my_text_field).unwrap();
-            let mut postings = inverted_index
-                .read_postings(&term_a, IndexRecordOption::WithFreqsAndPositions)
-                .unwrap()
-                .unwrap();
-            assert_eq!(postings.doc_freq(), 2);
-            let fallback_bitset = AliveBitSet::for_test_from_deleted_docs(&[0], 100);
-            assert_eq!(
-                postings.doc_freq_given_deletes(
-                    segment_reader.alive_bitset().unwrap_or(&fallback_bitset)
-                ),
-                2
-            );
-
-            assert_eq!(postings.term_freq(), 1);
-            let mut output = vec![];
-            postings.positions(&mut output);
-            assert_eq!(output, vec![1]);
-            postings.advance();
-
-            assert_eq!(postings.term_freq(), 2);
-            postings.positions(&mut output);
-            assert_eq!(output, vec![1, 3]);
-        }
-    }
-}

src/indexer/merger.rs

@@ -1,7 +1,8 @@
 use std::sync::Arc;
 
 use columnar::{
-    ColumnType, ColumnarReader, MergeRowOrder, RowAddr, ShuffleMergeOrder, StackMergeOrder,
+    ColumnType, ColumnValues, ColumnarReader, MergeRowOrder, RowAddr, ShuffleMergeOrder,
+    StackMergeOrder,
 };
 use common::ReadOnlyBitSet;
 use itertools::Itertools;
@@ -10,7 +11,7 @@ use measure_time::debug_time;
 use crate::directory::WritePtr;
 use crate::docset::{DocSet, TERMINATED};
 use crate::error::DataCorruption;
-use crate::fastfield::AliveBitSet;
+use crate::fastfield::{AliveBitSet, FastFieldNotAvailableError};
 use crate::fieldnorm::{FieldNormReader, FieldNormReaders, FieldNormsSerializer, FieldNormsWriter};
 use crate::index::{Segment, SegmentComponent, SegmentReader};
 use crate::indexer::doc_id_mapping::{MappingType, SegmentDocIdMapping};
@@ -19,7 +20,9 @@ use crate::postings::{InvertedIndexSerializer, Postings, SegmentPostings};
 use crate::schema::{value_type_to_column_type, Field, FieldType, Schema};
 use crate::store::StoreWriter;
 use crate::termdict::{TermMerger, TermOrdinal};
-use crate::{DocAddress, DocId, InvertedIndexReader};
+use crate::{
+    DocAddress, DocId, IndexSettings, IndexSortByField, InvertedIndexReader, Order, SegmentOrdinal,
+};
 
 /// Segment's max doc must be `< MAX_DOC_LIMIT`.
 ///
@@ -77,6 +80,7 @@ fn estimate_total_num_tokens(readers: &[SegmentReader], field: Field) -> crate::
 }
 
 pub struct IndexMerger {
+    index_settings: IndexSettings,
     schema: Schema,
     pub(crate) readers: Vec<SegmentReader>,
     max_doc: u32,
@@ -112,7 +116,7 @@ fn convert_to_merge_order(
 ) -> MergeRowOrder {
     match doc_id_mapping.mapping_type() {
         MappingType::Stacked => MergeRowOrder::Stack(StackMergeOrder::stack(columnars)),
-        MappingType::StackedWithDeletes => {
+        MappingType::StackedWithDeletes | MappingType::Shuffled => {
             // RUST/LLVM is amazing. The following conversion is actually a no-op:
             // no allocation, no copy.
             let new_row_id_to_old_row_id: Vec<RowAddr> = doc_id_mapping
@@ -145,9 +149,13 @@ fn extract_fast_field_required_columns(schema: &Schema) -> Vec<(String, ColumnTy
 }
 
 impl IndexMerger {
-    pub fn open(schema: Schema, segments: &[Segment]) -> crate::Result<IndexMerger> {
+    pub fn open(
+        schema: Schema,
+        index_settings: IndexSettings,
+        segments: &[Segment],
+    ) -> crate::Result<IndexMerger> {
         let alive_bitset = segments.iter().map(|_| None).collect_vec();
-        Self::open_with_custom_alive_set(schema, segments, alive_bitset)
+        Self::open_with_custom_alive_set(schema, index_settings, segments, alive_bitset)
     }
 
     // Create merge with a custom delete set.
@@ -164,6 +172,7 @@ impl IndexMerger {
     // segments and partitions them e.g. by a value in a field.
     pub fn open_with_custom_alive_set(
         schema: Schema,
+        index_settings: IndexSettings,
         segments: &[Segment],
         alive_bitset_opt: Vec<Option<AliveBitSet>>,
     ) -> crate::Result<IndexMerger> {
@@ -177,6 +186,9 @@ impl IndexMerger {
         }
 
         let max_doc = readers.iter().map(|reader| reader.num_docs()).sum();
+        if let Some(sort_by_field) = index_settings.sort_by_field.as_ref() {
+            readers = Self::sort_readers_by_min_sort_field(readers, sort_by_field)?;
+        }
         // sort segments by their natural sort setting
         if max_doc >= MAX_DOC_LIMIT {
             let err_msg = format!(
@@ -186,12 +198,37 @@ impl IndexMerger {
             return Err(crate::TantivyError::InvalidArgument(err_msg));
         }
         Ok(IndexMerger {
+            index_settings,
             schema,
             readers,
             max_doc,
         })
     }
 
+    fn sort_readers_by_min_sort_field(
+        readers: Vec<SegmentReader>,
+        sort_by_field: &IndexSortByField,
+    ) -> crate::Result<Vec<SegmentReader>> {
+        // presort the readers by their min_values, so that when they are disjunct, we can use
+        // the regular merge logic (implicitly sorted)
+        let mut readers_with_min_sort_values = readers
+            .into_iter()
+            .map(|reader| {
+                let accessor = Self::get_sort_field_accessor(&reader, sort_by_field)?;
+                Ok((reader, accessor.min_value()))
+            })
+            .collect::<crate::Result<Vec<_>>>()?;
+        if sort_by_field.order.is_asc() {
+            readers_with_min_sort_values.sort_by_key(|(_, min_val)| *min_val);
+        } else {
+            readers_with_min_sort_values.sort_by_key(|(_, min_val)| std::cmp::Reverse(*min_val));
+        }
+        Ok(readers_with_min_sort_values
+            .into_iter()
+            .map(|(reader, _)| reader)
+            .collect())
+    }
+
     fn write_fieldnorms(
         &self,
         mut fieldnorms_serializer: FieldNormsSerializer,
@@ -239,6 +276,128 @@ impl IndexMerger {
         Ok(())
     }
 
+    /// Checks if the readers are disjunct for their sort property and in the correct order to be
+    /// able to just stack them.
+    pub(crate) fn is_disjunct_and_sorted_on_sort_property(
+        &self,
+        sort_by_field: &IndexSortByField,
+    ) -> crate::Result<bool> {
+        let reader_ordinal_and_field_accessors =
+            self.get_reader_with_sort_field_accessor(sort_by_field)?;
+
+        let everything_is_in_order = reader_ordinal_and_field_accessors
+            .into_iter()
+            .map(|(_, col)| Arc::new(col))
+            .tuple_windows()
+            .all(|(field_accessor1, field_accessor2)| {
+                if sort_by_field.order.is_asc() {
+                    field_accessor1.max_value() <= field_accessor2.min_value()
+                } else {
+                    field_accessor1.min_value() >= field_accessor2.max_value()
+                }
+            });
+        Ok(everything_is_in_order)
+    }
+
+    pub(crate) fn get_sort_field_accessor(
+        reader: &SegmentReader,
+        sort_by_field: &IndexSortByField,
+    ) -> crate::Result<Arc<dyn ColumnValues>> {
+        reader.schema().get_field(&sort_by_field.field)?;
+        let (value_accessor, _column_type) = reader
+            .fast_fields()
+            .u64_lenient(&sort_by_field.field)?
+            .ok_or_else(|| FastFieldNotAvailableError {
+                field_name: sort_by_field.field.to_string(),
+            })?;
+        Ok(value_accessor.first_or_default_col(0u64))
+    }
+    /// Collecting value_accessors into a vec to bind the lifetime.
+    pub(crate) fn get_reader_with_sort_field_accessor(
+        &self,
+        sort_by_field: &IndexSortByField,
+    ) -> crate::Result<Vec<(SegmentOrdinal, Arc<dyn ColumnValues>)>> {
+        let reader_ordinal_and_field_accessors = self
+            .readers
+            .iter()
+            .enumerate()
+            .map(|(reader_ordinal, _)| reader_ordinal as SegmentOrdinal)
+            .map(|reader_ordinal: SegmentOrdinal| {
+                let value_accessor = Self::get_sort_field_accessor(
+                    &self.readers[reader_ordinal as usize],
+                    sort_by_field,
+                )?;
+                Ok((reader_ordinal, value_accessor))
+            })
+            .collect::<crate::Result<Vec<_>>>()?;
+        Ok(reader_ordinal_and_field_accessors)
+    }
+
+    /// Generates the doc_id mapping where position in the vec=new
+    /// doc_id.
+    /// ReaderWithOrdinal will include the ordinal position of the
+    /// reader in self.readers.
+    pub(crate) fn generate_doc_id_mapping_with_sort_by_field(
+        &self,
+        sort_by_field: &IndexSortByField,
+    ) -> crate::Result<SegmentDocIdMapping> {
+        let reader_ordinal_and_field_accessors =
+            self.get_reader_with_sort_field_accessor(sort_by_field)?;
+        // Loading the field accessor on demand causes a 15x regression
+
+        // create iterators over segment/sort_accessor/doc_id  tuple
+        let doc_id_reader_pair =
+            reader_ordinal_and_field_accessors
+                .iter()
+                .map(|(reader_ord, ff_reader)| {
+                    let reader = &self.readers[*reader_ord as usize];
+                    reader
+                        .doc_ids_alive()
+                        .map(move |doc_id| (doc_id, reader_ord, ff_reader))
+                });
+
+        let total_num_new_docs = self
+            .readers
+            .iter()
+            .map(|reader| reader.num_docs() as usize)
+            .sum();
+
+        let mut sorted_doc_ids: Vec<DocAddress> = Vec::with_capacity(total_num_new_docs);
+
+        // create iterator tuple of (old doc_id, reader) in order of the new doc_ids
+        sorted_doc_ids.extend(
+            doc_id_reader_pair
+                .into_iter()
+                .kmerge_by(|a, b| {
+                    let val1 = a.2.get_val(a.0);
+                    let val2 = b.2.get_val(b.0);
+                    if sort_by_field.order == Order::Asc {
+                        val1 < val2
+                    } else {
+                        val1 > val2
+                    }
+                })
+                .map(|(doc_id, &segment_ord, _)| DocAddress {
+                    doc_id,
+                    segment_ord,
+                }),
+        );
+
+        let alive_bitsets: Vec<Option<ReadOnlyBitSet>> = self
+            .readers
+            .iter()
+            .map(|segment_reader| {
+                let alive_bitset = segment_reader.alive_bitset()?;
+                Some(alive_bitset.bitset().clone())
+            })
+            .collect();
+        Ok(SegmentDocIdMapping::new(
+            sorted_doc_ids,
+            MappingType::Shuffled,
+            alive_bitsets,
+        ))
+    }
+
     /// Creates a mapping if the segments are stacked. this is helpful to merge codelines between
     /// index sorting and the others
     pub(crate) fn get_doc_id_from_concatenated_data(&self) -> crate::Result<SegmentDocIdMapping> {
@@ -356,6 +515,7 @@ impl IndexMerger {
         );
 
         let mut segment_postings_containing_the_term: Vec<(usize, SegmentPostings)> = vec![];
+        let mut doc_id_and_positions = vec![];
 
         while merged_terms.advance() {
             segment_postings_containing_the_term.clear();
@@ -451,13 +611,37 @@ impl IndexMerger {
                             0u32
                         };
 
-                        let delta_positions = delta_computer.compute_delta(&positions_buffer);
-                        field_serializer.write_doc(remapped_doc_id, term_freq, delta_positions);
+                        // if doc_id_mapping exists, the doc_ids are reordered, they are
+                        // not just stacked. The field serializer expects monotonically increasing
+                        // doc_ids, so we collect and sort them first, before writing.
+                        //
+                        // I think this is not strictly necessary, it would be possible to
+                        // avoid the loading into a vec via some form of kmerge, but then the merge
+                        // logic would deviate much more from the stacking case (unsorted index)
+                        if !doc_id_mapping.is_trivial() {
+                            doc_id_and_positions.push((
+                                remapped_doc_id,
+                                term_freq,
+                                positions_buffer.to_vec(),
+                            ));
+                        } else {
+                            let delta_positions = delta_computer.compute_delta(&positions_buffer);
+                            field_serializer.write_doc(remapped_doc_id, term_freq, delta_positions);
+                        }
                     }
 
                     doc = segment_postings.advance();
                 }
             }
+            if !doc_id_mapping.is_trivial() {
+                doc_id_and_positions.sort_unstable_by_key(|&(doc_id, _, _)| doc_id);
+
+                for (doc_id, term_freq, positions) in &doc_id_and_positions {
+                    let delta_positions = delta_computer.compute_delta(positions);
+                    field_serializer.write_doc(*doc_id, *term_freq, delta_positions);
+                }
+                doc_id_and_positions.clear();
+            }
             // closing the term.
             field_serializer.close_term()?;
         }
@@ -486,13 +670,47 @@ impl IndexMerger {
         Ok(())
     }
 
-    fn write_storable_fields(&self, store_writer: &mut StoreWriter) -> crate::Result<()> {
+    fn write_storable_fields(
+        &self,
+        store_writer: &mut StoreWriter,
+        doc_id_mapping: &SegmentDocIdMapping,
+    ) -> crate::Result<()> {
         debug_time!("write-storable-fields");
         debug!("write-storable-field");
 
-        for reader in &self.readers {
-            let store_reader = reader.get_store_reader(1)?;
-            if reader.has_deletes()
+        if !doc_id_mapping.is_trivial() {
+            debug!("non-trivial-doc-id-mapping");
+
+            let store_readers: Vec<_> = self
+                .readers
+                .iter()
+                .map(|reader| reader.get_store_reader(50))
+                .collect::<Result<_, _>>()?;
+
+            let mut document_iterators: Vec<_> = store_readers
+                .iter()
+                .enumerate()
+                .map(|(i, store)| store.iter_raw(self.readers[i].alive_bitset()))
+                .collect();
+
+            for old_doc_addr in doc_id_mapping.iter_old_doc_addrs() {
+                let doc_bytes_it = &mut document_iterators[old_doc_addr.segment_ord as usize];
+                if let Some(doc_bytes_res) = doc_bytes_it.next() {
+                    let (_, doc_bytes) = doc_bytes_res?;
+                    store_writer.store_bytes(&doc_bytes)?;
+                } else {
+                    return Err(DataCorruption::comment_only(format!(
+                        "unexpected missing document in docstore on merge, doc address \
+                         {old_doc_addr:?}",
+                    ))
+                    .into());
+                }
+            }
+        } else {
+            debug!("trivial-doc-id-mapping");
+            for reader in &self.readers {
+                let store_reader = reader.get_store_reader(1)?;
+                if reader.has_deletes()
                     // If there is not enough data in the store, we avoid stacking in order to
                     // avoid creating many small blocks in the doc store. Once we have 5 full blocks,
                     // we start stacking. In the worst case 2/7 of the blocks would be very small.
@@ -508,13 +726,14 @@ impl IndexMerger {
                     // take 7 in order to not walk over all checkpoints.
                     || store_reader.block_checkpoints().take(7).count() < 6
                     || store_reader.decompressor() != store_writer.compressor().into()
-            {
-                for doc_bytes_res in store_reader.iter_raw(reader.alive_bitset()) {
-                    let doc_bytes = doc_bytes_res?;
-                    store_writer.store_bytes(&doc_bytes)?;
+                {
+                    for doc_bytes_res in store_reader.iter_raw(reader.alive_bitset()) {
+                        let (_, doc_bytes) = doc_bytes_res?;
+                        store_writer.store_bytes(&doc_bytes)?;
+                    }
+                } else {
+                    store_writer.stack(store_reader)?;
                 }
-            } else {
-                store_writer.stack(store_reader)?;
             }
         }
         Ok(())
@@ -526,7 +745,18 @@ impl IndexMerger {
     /// # Returns
     /// The number of documents in the resulting segment.
     pub fn write(&self, mut serializer: SegmentSerializer) -> crate::Result<u32> {
-        let doc_id_mapping = self.get_doc_id_from_concatenated_data()?;
+        let doc_id_mapping = if let Some(sort_by_field) = self.index_settings.sort_by_field.as_ref()
+        {
+            // If the documents are already sorted and stackable, we ignore the mapping and execute
+            // it as if there was no sorting
+            if self.is_disjunct_and_sorted_on_sort_property(sort_by_field)? {
+                self.get_doc_id_from_concatenated_data()?
+            } else {
+                self.generate_doc_id_mapping_with_sort_by_field(sort_by_field)?
+            }
+        } else {
+            self.get_doc_id_from_concatenated_data()?
+        };
         debug!("write-fieldnorms");
         if let Some(fieldnorms_serializer) = serializer.extract_fieldnorms_serializer() {
             self.write_fieldnorms(fieldnorms_serializer, &doc_id_mapping)?;
@@ -543,7 +773,7 @@ impl IndexMerger {
         )?;
 
         debug!("write-storagefields");
-        self.write_storable_fields(serializer.get_store_writer())?;
+        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)?;
 
@@ -557,8 +787,6 @@ impl IndexMerger {
 mod tests {
 
     use columnar::Column;
-    use proptest::prop_oneof;
-    use proptest::strategy::Strategy;
     use schema::FAST;
 
     use crate::collector::tests::{
@@ -566,16 +794,15 @@ mod tests {
     };
     use crate::collector::{Count, FacetCollector};
     use crate::index::{Index, SegmentId};
-    use crate::indexer::NoMergePolicy;
     use crate::query::{AllQuery, BooleanQuery, EnableScoring, Scorer, TermQuery};
     use crate::schema::{
         Facet, FacetOptions, IndexRecordOption, NumericOptions, TantivyDocument, Term,
-        TextFieldIndexing, Value, INDEXED, TEXT,
+        TextFieldIndexing, INDEXED, TEXT,
     };
     use crate::time::OffsetDateTime;
     use crate::{
         assert_nearly_equals, schema, DateTime, DocAddress, DocId, DocSet, IndexSettings,
-        IndexWriter, Searcher,
+        IndexSortByField, IndexWriter, Order, Searcher,
     };
 
     #[test]
@@ -682,24 +909,15 @@ mod tests {
             }
             {
                 let doc = searcher.doc::<TantivyDocument>(DocAddress::new(0, 0))?;
-                assert_eq!(
-                    doc.get_first(text_field).unwrap().as_value().as_str(),
-                    Some("af b")
-                );
+                assert_eq!(doc.get_first(text_field).unwrap().as_str(), Some("af b"));
             }
             {
                 let doc = searcher.doc::<TantivyDocument>(DocAddress::new(0, 1))?;
-                assert_eq!(
-                    doc.get_first(text_field).unwrap().as_value().as_str(),
-                    Some("a b c")
-                );
+                assert_eq!(doc.get_first(text_field).unwrap().as_str(), Some("a b c"));
             }
             {
                 let doc = searcher.doc::<TantivyDocument>(DocAddress::new(0, 2))?;
-                assert_eq!(
-                    doc.get_first(text_field).unwrap().as_value().as_str(),
-                    Some("a b c d")
-                );
+                assert_eq!(doc.get_first(text_field).unwrap().as_str(), Some("a b c d"));
             }
             {
                 let doc = searcher.doc::<TantivyDocument>(DocAddress::new(0, 3))?;
@@ -1048,6 +1266,60 @@ mod tests {
         test_merge_facets(None, true)
     }
 
+    #[test]
+    fn test_merge_facets_sort_asc() {
+        // In the merge case this will go through the doc_id mapping code
+        test_merge_facets(
+            Some(IndexSettings {
+                sort_by_field: Some(IndexSortByField {
+                    field: "intval".to_string(),
+                    order: Order::Desc,
+                }),
+                ..Default::default()
+            }),
+            true,
+        );
+        // In the merge case this will not go through the doc_id mapping code, because the data
+        // sorted and disjunct
+        test_merge_facets(
+            Some(IndexSettings {
+                sort_by_field: Some(IndexSortByField {
+                    field: "intval".to_string(),
+                    order: Order::Desc,
+                }),
+                ..Default::default()
+            }),
+            false,
+        );
+    }
+
+    #[test]
+    fn test_merge_facets_sort_desc() {
+        // In the merge case this will go through the doc_id mapping code
+        test_merge_facets(
+            Some(IndexSettings {
+                sort_by_field: Some(IndexSortByField {
+                    field: "intval".to_string(),
+                    order: Order::Desc,
+                }),
+                ..Default::default()
+            }),
+            true,
+        );
+        // In the merge case this will not go through the doc_id mapping code, because the data
+        // sorted and disjunct
+        test_merge_facets(
+            Some(IndexSettings {
+                sort_by_field: Some(IndexSortByField {
+                    field: "intval".to_string(),
+                    order: Order::Desc,
+                }),
+                ..Default::default()
+            }),
+            false,
+        );
+    }
+
     // force_segment_value_overlap forces the int value for sorting to have overlapping min and max
     // ranges between segments so that merge algorithm can't apply certain optimizations
     fn test_merge_facets(index_settings: Option<IndexSettings>, force_segment_value_overlap: bool) {
@@ -1250,112 +1522,6 @@ mod tests {
         Ok(())
     }
 
-    #[derive(Debug, Clone, Copy, Eq, PartialEq)]
-    enum IndexingOp {
-        ZeroVal,
-        OneVal { val: u64 },
-        TwoVal { val: u64 },
-        Commit,
-    }
-
-    fn balanced_operation_strategy() -> impl Strategy<Value = IndexingOp> {
-        prop_oneof![
-            (0u64..1u64).prop_map(|_| IndexingOp::ZeroVal),
-            (0u64..1u64).prop_map(|val| IndexingOp::OneVal { val }),
-            (0u64..1u64).prop_map(|val| IndexingOp::TwoVal { val }),
-            (0u64..1u64).prop_map(|_| IndexingOp::Commit),
-        ]
-    }
-
-    use proptest::prelude::*;
-    proptest! {
-        #[test]
-        fn test_merge_columnar_int_proptest(ops in proptest::collection::vec(balanced_operation_strategy(), 1..20)) {
-            assert!(test_merge_int_fields(&ops[..]).is_ok());
-        }
-    }
-    fn test_merge_int_fields(ops: &[IndexingOp]) -> crate::Result<()> {
-        if ops.iter().all(|op| *op == IndexingOp::Commit) {
-            return Ok(());
-        }
-        let expected_doc_and_vals: Vec<(u32, Vec<u64>)> = ops
-            .iter()
-            .filter(|op| *op != &IndexingOp::Commit)
-            .map(|op| match op {
-                IndexingOp::ZeroVal => vec![],
-                IndexingOp::OneVal { val } => vec![*val],
-                IndexingOp::TwoVal { val } => vec![*val, *val],
-                IndexingOp::Commit => unreachable!(),
-            })
-            .enumerate()
-            .map(|(id, val)| (id as u32, val))
-            .collect();
-
-        let mut schema_builder = schema::Schema::builder();
-        let int_options = NumericOptions::default().set_fast().set_indexed();
-        let int_field = schema_builder.add_u64_field("intvals", int_options);
-        let index = Index::create_in_ram(schema_builder.build());
-        {
-            let mut index_writer = index.writer_for_tests()?;
-            index_writer.set_merge_policy(Box::new(NoMergePolicy));
-            let index_doc = |index_writer: &mut IndexWriter, int_vals: &[u64]| {
-                let mut doc = TantivyDocument::default();
-                for &val in int_vals {
-                    doc.add_u64(int_field, val);
-                }
-                index_writer.add_document(doc).unwrap();
-            };
-
-            for op in ops {
-                match op {
-                    IndexingOp::ZeroVal => index_doc(&mut index_writer, &[]),
-                    IndexingOp::OneVal { val } => index_doc(&mut index_writer, &[*val]),
-                    IndexingOp::TwoVal { val } => index_doc(&mut index_writer, &[*val, *val]),
-                    IndexingOp::Commit => {
-                        index_writer.commit().expect("commit failed");
-                    }
-                }
-            }
-            index_writer.commit().expect("commit failed");
-        }
-        {
-            let mut segment_ids = index.searchable_segment_ids()?;
-            segment_ids.sort();
-            let mut index_writer: IndexWriter = index.writer_for_tests()?;
-            index_writer.merge(&segment_ids).wait()?;
-            index_writer.wait_merging_threads()?;
-        }
-        let reader = index.reader()?;
-        reader.reload()?;
-
-        let mut vals: Vec<u64> = Vec::new();
-        let mut test_vals = move |col: &Column<u64>, doc: DocId, expected: &[u64]| {
-            vals.clear();
-            vals.extend(col.values_for_doc(doc));
-            assert_eq!(&vals[..], expected);
-        };
-
-        let mut test_col = move |col: &Column<u64>, column_expected: &[(u32, Vec<u64>)]| {
-            for (doc_id, vals) in column_expected.iter() {
-                test_vals(col, *doc_id, vals);
-            }
-        };
-
-        {
-            let searcher = reader.searcher();
-            let segment = searcher.segment_reader(0u32);
-            let col = segment
-                .fast_fields()
-                .column_opt::<u64>("intvals")
-                .unwrap()
-                .unwrap();
-
-            test_col(&col, &expected_doc_and_vals);
-        }
-
-        Ok(())
-    }
-
     #[test]
     fn test_merge_multivalued_int_fields_simple() -> crate::Result<()> {
         let mut schema_builder = schema::Schema::builder();

src/indexer/merger_sorted_index_test.rs

@@ -0,0 +1,576 @@
+#[cfg(test)]
+mod tests {
+    use crate::collector::TopDocs;
+    use crate::fastfield::AliveBitSet;
+    use crate::index::Index;
+    use crate::postings::Postings;
+    use crate::query::QueryParser;
+    use crate::schema::{
+        self, BytesOptions, Facet, FacetOptions, IndexRecordOption, NumericOptions,
+        TextFieldIndexing, TextOptions,
+    };
+    use crate::{
+        DocAddress, DocSet, IndexSettings, IndexSortByField, IndexWriter, Order, TantivyDocument,
+        Term,
+    };
+
+    fn create_test_index_posting_list_issue(index_settings: Option<IndexSettings>) -> Index {
+        let mut schema_builder = schema::Schema::builder();
+        let int_options = NumericOptions::default().set_fast().set_indexed();
+        let int_field = schema_builder.add_u64_field("intval", int_options);
+
+        let facet_field = schema_builder.add_facet_field("facet", FacetOptions::default());
+
+        let schema = schema_builder.build();
+
+        let mut index_builder = Index::builder().schema(schema);
+        if let Some(settings) = index_settings {
+            index_builder = index_builder.settings(settings);
+        }
+        let index = index_builder.create_in_ram().unwrap();
+
+        {
+            let mut index_writer: IndexWriter = index.writer_for_tests().unwrap();
+            index_writer
+                .add_document(doc!(int_field=>3_u64, facet_field=> Facet::from("/crime")))
+                .unwrap();
+            index_writer
+                .add_document(doc!(int_field=>6_u64, facet_field=> Facet::from("/crime")))
+                .unwrap();
+            index_writer.commit().unwrap();
+            index_writer
+                .add_document(doc!(int_field=>5_u64, facet_field=> Facet::from("/fanta")))
+                .unwrap();
+            index_writer.commit().unwrap();
+        }
+
+        // Merging the segments
+        {
+            let segment_ids = index
+                .searchable_segment_ids()
+                .expect("Searchable segments failed.");
+            let mut index_writer: IndexWriter = index.writer_for_tests().unwrap();
+            assert!(index_writer.merge(&segment_ids).wait().is_ok());
+            assert!(index_writer.wait_merging_threads().is_ok());
+        }
+        index
+    }
+
+    // force_disjunct_segment_sort_values forces the field, by which the index is sorted have
+    // disjunct ranges between segments, e.g. values in segment [1-3] [10 - 20] [50 - 500]
+    fn create_test_index(
+        index_settings: Option<IndexSettings>,
+        force_disjunct_segment_sort_values: bool,
+    ) -> crate::Result<Index> {
+        let mut schema_builder = schema::Schema::builder();
+        let int_options = NumericOptions::default()
+            .set_fast()
+            .set_stored()
+            .set_indexed();
+        let int_field = schema_builder.add_u64_field("intval", int_options);
+
+        let bytes_options = BytesOptions::default().set_fast().set_indexed();
+        let bytes_field = schema_builder.add_bytes_field("bytes", bytes_options);
+        let facet_field = schema_builder.add_facet_field("facet", FacetOptions::default());
+
+        let multi_numbers =
+            schema_builder.add_u64_field("multi_numbers", NumericOptions::default().set_fast());
+        let text_field_options = TextOptions::default()
+            .set_indexing_options(
+                TextFieldIndexing::default()
+                    .set_index_option(schema::IndexRecordOption::WithFreqsAndPositions),
+            )
+            .set_stored();
+        let text_field = schema_builder.add_text_field("text_field", text_field_options);
+        let schema = schema_builder.build();
+
+        let mut index_builder = Index::builder().schema(schema);
+        if let Some(settings) = index_settings {
+            index_builder = index_builder.settings(settings);
+        }
+        let index = index_builder.create_in_ram()?;
+
+        {
+            let mut index_writer = index.writer_for_tests()?;
+
+            // segment 1 - range 1-3
+            index_writer.add_document(doc!(int_field=>1_u64))?;
+            index_writer.add_document(
+                doc!(int_field=>3_u64, multi_numbers => 3_u64, multi_numbers => 4_u64, bytes_field => vec![1, 2, 3], text_field => "some text", facet_field=> Facet::from("/book/crime")),
+            )?;
+            index_writer.add_document(
+                doc!(int_field=>1_u64, text_field=> "deleteme",  text_field => "ok text more text"),
+            )?;
+            index_writer.add_document(
+                doc!(int_field=>2_u64, multi_numbers => 2_u64, multi_numbers => 3_u64, text_field => "ok text more text"),
+            )?;
+
+            index_writer.commit()?;
+            // segment 2 - range 1-20 , with force_disjunct_segment_sort_values 10-20
+            index_writer.add_document(doc!(int_field=>20_u64, multi_numbers => 20_u64))?;
+
+            let in_val = if force_disjunct_segment_sort_values {
+                10_u64
+            } else {
+                1
+            };
+            index_writer.add_document(doc!(int_field=>in_val, text_field=> "deleteme" , text_field => "ok text more text", facet_field=> Facet::from("/book/crime")))?;
+            index_writer.commit()?;
+            // segment 3 - range 5-1000, with force_disjunct_segment_sort_values 50-1000
+            let int_vals = if force_disjunct_segment_sort_values {
+                [100_u64, 50]
+            } else {
+                [10, 5]
+            };
+            index_writer.add_document( // position of this doc after delete in desc sorting = [2], in disjunct case [1]
+                doc!(int_field=>int_vals[0], multi_numbers => 10_u64, multi_numbers => 11_u64, text_field=> "blubber", facet_field=> Facet::from("/book/fantasy")),
+            )?;
+            index_writer.add_document(doc!(int_field=>int_vals[1], text_field=> "deleteme"))?;
+            index_writer.add_document(
+                doc!(int_field=>1_000u64, multi_numbers => 1001_u64, multi_numbers => 1002_u64, bytes_field => vec![5, 5],text_field => "the biggest num")
+            )?;
+
+            index_writer.delete_term(Term::from_field_text(text_field, "deleteme"));
+            index_writer.commit()?;
+        }
+
+        // Merging the segments
+        {
+            let segment_ids = index.searchable_segment_ids()?;
+            let mut index_writer: IndexWriter = index.writer_for_tests()?;
+            index_writer.merge(&segment_ids).wait()?;
+            index_writer.wait_merging_threads()?;
+        }
+        Ok(index)
+    }
+
+    #[test]
+    fn test_merge_sorted_postinglist_sort_issue() {
+        create_test_index_posting_list_issue(Some(IndexSettings {
+            sort_by_field: Some(IndexSortByField {
+                field: "intval".to_string(),
+                order: Order::Desc,
+            }),
+            ..Default::default()
+        }));
+    }
+
+    #[test]
+    fn test_merge_sorted_index_desc_not_disjunct() {
+        test_merge_sorted_index_desc_(false);
+    }
+
+    #[test]
+    fn test_merge_sorted_index_desc_disjunct() {
+        test_merge_sorted_index_desc_(true);
+    }
+
+    fn test_merge_sorted_index_desc_(force_disjunct_segment_sort_values: bool) {
+        let index = create_test_index(
+            Some(IndexSettings {
+                sort_by_field: Some(IndexSortByField {
+                    field: "intval".to_string(),
+                    order: Order::Desc,
+                }),
+                ..Default::default()
+            }),
+            force_disjunct_segment_sort_values,
+        )
+        .unwrap();
+
+        let int_field = index.schema().get_field("intval").unwrap();
+        let reader = index.reader().unwrap();
+
+        let searcher = reader.searcher();
+        assert_eq!(searcher.segment_readers().len(), 1);
+        let segment_reader = searcher.segment_readers().last().unwrap();
+
+        let fast_fields = segment_reader.fast_fields();
+        let fast_field = fast_fields.u64("intval").unwrap();
+        assert_eq!(fast_field.first(5), Some(1u64));
+        assert_eq!(fast_field.first(4), Some(2u64));
+        assert_eq!(fast_field.first(3), Some(3u64));
+        if force_disjunct_segment_sort_values {
+            assert_eq!(fast_field.first(2), Some(20u64));
+            assert_eq!(fast_field.first(1), Some(100u64));
+        } else {
+            assert_eq!(fast_field.first(2), Some(10u64));
+            assert_eq!(fast_field.first(1), Some(20u64));
+        }
+        assert_eq!(fast_field.first(0), Some(1_000u64));
+
+        // test new field norm mapping
+        {
+            let my_text_field = index.schema().get_field("text_field").unwrap();
+            let fieldnorm_reader = segment_reader.get_fieldnorms_reader(my_text_field).unwrap();
+            assert_eq!(fieldnorm_reader.fieldnorm(0), 3); // the biggest num
+            if force_disjunct_segment_sort_values {
+                assert_eq!(fieldnorm_reader.fieldnorm(1), 1); // blubber
+                assert_eq!(fieldnorm_reader.fieldnorm(2), 0);
+            } else {
+                assert_eq!(fieldnorm_reader.fieldnorm(1), 0);
+                assert_eq!(fieldnorm_reader.fieldnorm(2), 1); // blubber
+            }
+            assert_eq!(fieldnorm_reader.fieldnorm(3), 2); // some text
+            assert_eq!(fieldnorm_reader.fieldnorm(5), 0);
+        }
+
+        let my_text_field = index.schema().get_field("text_field").unwrap();
+        let searcher = index.reader().unwrap().searcher();
+        {
+            let my_text_field = index.schema().get_field("text_field").unwrap();
+
+            let do_search = |term: &str| {
+                let query = QueryParser::for_index(&index, vec![my_text_field])
+                    .parse_query(term)
+                    .unwrap();
+                let top_docs: Vec<(f32, DocAddress)> =
+                    searcher.search(&query, &TopDocs::with_limit(3)).unwrap();
+
+                top_docs.iter().map(|el| el.1.doc_id).collect::<Vec<_>>()
+            };
+
+            assert_eq!(do_search("some"), vec![3]);
+            if force_disjunct_segment_sort_values {
+                assert_eq!(do_search("blubber"), vec![1]);
+            } else {
+                assert_eq!(do_search("blubber"), vec![2]);
+            }
+            assert_eq!(do_search("biggest"), vec![0]);
+        }
+
+        // postings file
+        {
+            let my_text_field = index.schema().get_field("text_field").unwrap();
+            let term_a = Term::from_field_text(my_text_field, "text");
+            let inverted_index = segment_reader.inverted_index(my_text_field).unwrap();
+            let mut postings = inverted_index
+                .read_postings(&term_a, IndexRecordOption::WithFreqsAndPositions)
+                .unwrap()
+                .unwrap();
+
+            assert_eq!(postings.doc_freq(), 2);
+            let fallback_bitset = AliveBitSet::for_test_from_deleted_docs(&[0], 100);
+            assert_eq!(
+                postings.doc_freq_given_deletes(
+                    segment_reader.alive_bitset().unwrap_or(&fallback_bitset)
+                ),
+                2
+            );
+
+            assert_eq!(postings.term_freq(), 1);
+            let mut output = vec![];
+            postings.positions(&mut output);
+            assert_eq!(output, vec![1]);
+            postings.advance();
+
+            assert_eq!(postings.term_freq(), 2);
+            postings.positions(&mut output);
+            assert_eq!(output, vec![1, 3]);
+        }
+
+        // access doc store
+        {
+            let blubber_pos = if force_disjunct_segment_sort_values {
+                1
+            } else {
+                2
+            };
+            let doc = searcher
+                .doc::<TantivyDocument>(DocAddress::new(0, blubber_pos))
+                .unwrap();
+            assert_eq!(
+                doc.get_first(my_text_field).unwrap().as_str(),
+                Some("blubber")
+            );
+            let doc = searcher
+                .doc::<TantivyDocument>(DocAddress::new(0, 0))
+                .unwrap();
+            assert_eq!(doc.get_first(int_field).unwrap().as_u64(), Some(1000));
+        }
+    }
+
+    #[test]
+    fn test_merge_unsorted_index() {
+        let index = create_test_index(
+            Some(IndexSettings {
+                ..Default::default()
+            }),
+            false,
+        )
+        .unwrap();
+
+        let reader = index.reader().unwrap();
+        let searcher = reader.searcher();
+        assert_eq!(searcher.segment_readers().len(), 1);
+        let segment_reader = searcher.segment_readers().last().unwrap();
+
+        let searcher = index.reader().unwrap().searcher();
+        {
+            let my_text_field = index.schema().get_field("text_field").unwrap();
+
+            let do_search = |term: &str| {
+                let query = QueryParser::for_index(&index, vec![my_text_field])
+                    .parse_query(term)
+                    .unwrap();
+                let top_docs: Vec<(f32, DocAddress)> =
+                    searcher.search(&query, &TopDocs::with_limit(3)).unwrap();
+
+                top_docs.iter().map(|el| el.1.doc_id).collect::<Vec<_>>()
+            };
+
+            assert_eq!(do_search("some"), vec![1]);
+            assert_eq!(do_search("blubber"), vec![3]);
+            assert_eq!(do_search("biggest"), vec![4]);
+        }
+
+        // postings file
+        {
+            let my_text_field = index.schema().get_field("text_field").unwrap();
+            let term_a = Term::from_field_text(my_text_field, "text");
+            let inverted_index = segment_reader.inverted_index(my_text_field).unwrap();
+            let mut postings = inverted_index
+                .read_postings(&term_a, IndexRecordOption::WithFreqsAndPositions)
+                .unwrap()
+                .unwrap();
+            assert_eq!(postings.doc_freq(), 2);
+            let fallback_bitset = AliveBitSet::for_test_from_deleted_docs(&[0], 100);
+            assert_eq!(
+                postings.doc_freq_given_deletes(
+                    segment_reader.alive_bitset().unwrap_or(&fallback_bitset)
+                ),
+                2
+            );
+
+            assert_eq!(postings.term_freq(), 1);
+            let mut output = vec![];
+            postings.positions(&mut output);
+            assert_eq!(output, vec![1]);
+            postings.advance();
+
+            assert_eq!(postings.term_freq(), 2);
+            postings.positions(&mut output);
+            assert_eq!(output, vec![1, 3]);
+        }
+    }
+
+    // #[test]
+    // fn test_merge_sorted_index_asc() {
+    //     let index = create_test_index(
+    //         Some(IndexSettings {
+    //             sort_by_field: Some(IndexSortByField {
+    //                 field: "intval".to_string(),
+    //                 order: Order::Asc,
+    //             }),
+    //             ..Default::default()
+    //         }),
+    //         false,
+    //     )
+    //     .unwrap();
+
+    //     let int_field = index.schema().get_field("intval").unwrap();
+    //     let multi_numbers = index.schema().get_field("multi_numbers").unwrap();
+    //     let bytes_field = index.schema().get_field("bytes").unwrap();
+    //     let reader = index.reader().unwrap();
+    //     let searcher = reader.searcher();
+    //     assert_eq!(searcher.segment_readers().len(), 1);
+    //     let segment_reader = searcher.segment_readers().last().unwrap();
+
+    //     let fast_fields = segment_reader.fast_fields();
+    //     let fast_field = fast_fields.u64(int_field).unwrap();
+    //     assert_eq!(fast_field.get_val(0), 1u64);
+    //     assert_eq!(fast_field.get_val(1), 2u64);
+    //     assert_eq!(fast_field.get_val(2), 3u64);
+    //     assert_eq!(fast_field.get_val(3), 10u64);
+    //     assert_eq!(fast_field.get_val(4), 20u64);
+    //     assert_eq!(fast_field.get_val(5), 1_000u64);
+
+    //     let get_vals = |fast_field: &MultiValuedFastFieldReader<u64>, doc_id: u32| -> Vec<u64> {
+    //         let mut vals = vec![];
+    //         fast_field.get_vals(doc_id, &mut vals);
+    //         vals
+    //     };
+    //     let fast_fields = segment_reader.fast_fields();
+    //     let fast_field = fast_fields.u64s(multi_numbers).unwrap();
+    //     assert_eq!(&get_vals(&fast_field, 0), &[] as &[u64]);
+    //     assert_eq!(&get_vals(&fast_field, 1), &[2, 3]);
+    //     assert_eq!(&get_vals(&fast_field, 2), &[3, 4]);
+    //     assert_eq!(&get_vals(&fast_field, 3), &[10, 11]);
+    //     assert_eq!(&get_vals(&fast_field, 4), &[20]);
+    //     assert_eq!(&get_vals(&fast_field, 5), &[1001, 1002]);
+
+    //     let fast_field = fast_fields.bytes(bytes_field).unwrap();
+    //     assert_eq!(fast_field.get_bytes(0), &[] as &[u8]);
+    //     assert_eq!(fast_field.get_bytes(2), &[1, 2, 3]);
+    //     assert_eq!(fast_field.get_bytes(5), &[5, 5]);
+
+    //     // test new field norm mapping
+    //     {
+    //         let my_text_field = index.schema().get_field("text_field").unwrap();
+    //         let fieldnorm_reader = segment_reader.get_fieldnorms_reader(my_text_field).unwrap();
+    //         assert_eq!(fieldnorm_reader.fieldnorm(0), 0);
+    //         assert_eq!(fieldnorm_reader.fieldnorm(1), 4);
+    //         assert_eq!(fieldnorm_reader.fieldnorm(2), 2); // some text
+    //         assert_eq!(fieldnorm_reader.fieldnorm(3), 1);
+    //         assert_eq!(fieldnorm_reader.fieldnorm(5), 3); // the biggest num
+    //     }
+
+    //     let searcher = index.reader().unwrap().searcher();
+    //     {
+    //         let my_text_field = index.schema().get_field("text_field").unwrap();
+
+    //         let do_search = |term: &str| {
+    //             let query = QueryParser::for_index(&index, vec![my_text_field])
+    //                 .parse_query(term)
+    //                 .unwrap();
+    //             let top_docs: Vec<(f32, DocAddress)> =
+    //                 searcher.search(&query, &TopDocs::with_limit(3)).unwrap();
+
+    //             top_docs.iter().map(|el| el.1.doc_id).collect::<Vec<_>>()
+    //         };
+
+    //         assert_eq!(do_search("some"), vec![2]);
+    //         assert_eq!(do_search("blubber"), vec![3]);
+    //         assert_eq!(do_search("biggest"), vec![5]);
+    //     }
+
+    //     // postings file
+    //     {
+    //         let my_text_field = index.schema().get_field("text_field").unwrap();
+    //         let term_a = Term::from_field_text(my_text_field, "text");
+    //         let inverted_index = segment_reader.inverted_index(my_text_field).unwrap();
+    //         let mut postings = inverted_index
+    //             .read_postings(&term_a, IndexRecordOption::WithFreqsAndPositions)
+    //             .unwrap()
+    //             .unwrap();
+
+    //         assert_eq!(postings.doc_freq(), 2);
+    //         let fallback_bitset = AliveBitSet::for_test_from_deleted_docs(&[0], 100);
+    //         assert_eq!(
+    //             postings.doc_freq_given_deletes(
+    //                 segment_reader.alive_bitset().unwrap_or(&fallback_bitset)
+    //             ),
+    //             2
+    //         );
+
+    //         let mut output = vec![];
+    //         postings.positions(&mut output);
+    //         assert_eq!(output, vec![1, 3]);
+    //         postings.advance();
+
+    //         postings.positions(&mut output);
+    //         assert_eq!(output, vec![1]);
+    //     }
+
+    //     // access doc store
+    //     {
+    //         let doc = searcher.doc(DocAddress::new(0, 0)).unwrap();
+    //         assert_eq!(doc.get_first(int_field).unwrap().as_u64(), Some(1));
+    //         let doc = searcher.doc(DocAddress::new(0, 1)).unwrap();
+    //         assert_eq!(doc.get_first(int_field).unwrap().as_u64(), Some(2));
+    //         let doc = searcher.doc(DocAddress::new(0, 2)).unwrap();
+    //         assert_eq!(doc.get_first(int_field).unwrap().as_u64(), Some(3));
+    //         let doc = searcher.doc(DocAddress::new(0, 3)).unwrap();
+    //         assert_eq!(doc.get_first(int_field).unwrap().as_u64(), Some(10));
+    //         let doc = searcher.doc(DocAddress::new(0, 4)).unwrap();
+    //         assert_eq!(doc.get_first(int_field).unwrap().as_u64(), Some(20));
+    //         let doc = searcher.doc(DocAddress::new(0, 5)).unwrap();
+    //         assert_eq!(doc.get_first(int_field).unwrap().as_u64(), Some(1_000));
+    //     }
+    // }
+}
+
+#[cfg(all(test, feature = "unstable"))]
+mod bench_sorted_index_merge {
+
+    use test::{self, Bencher};
+
+    use crate::index::Index;
+    use crate::indexer::merger::IndexMerger;
+    use crate::schema::{NumericOptions, Schema};
+    use crate::{IndexSettings, IndexSortByField, IndexWriter, Order};
+    fn create_index(sort_by_field: Option<IndexSortByField>) -> Index {
+        let mut schema_builder = Schema::builder();
+        let int_options = NumericOptions::default().set_fast().set_indexed();
+        let int_field = schema_builder.add_u64_field("intval", int_options);
+        let schema = schema_builder.build();
+
+        let index_builder = Index::builder().schema(schema).settings(IndexSettings {
+            sort_by_field,
+            ..Default::default()
+        });
+        let index = index_builder.create_in_ram().unwrap();
+
+        {
+            let mut index_writer: IndexWriter = index.writer_for_tests().unwrap();
+            let index_doc = |index_writer: &mut IndexWriter, val: u64| {
+                index_writer.add_document(doc!(int_field=>val)).unwrap();
+            };
+            // 3 segments with 10_000 values in the fast fields
+            for _ in 0..3 {
+                index_doc(&mut index_writer, 5000); // fix to make it unordered
+                for i in 0..10_000 {
+                    index_doc(&mut index_writer, i);
+                }
+                index_writer.commit().unwrap();
+            }
+        }
+        index
+    }
+
+    //#[bench]
+    // fn create_sorted_index_walk_overkmerge_on_merge_fastfield(
+    // b: &mut Bencher,
+    //) -> crate::Result<()> {
+    // let sort_by_field = IndexSortByField {
+    // field: "intval".to_string(),
+    // order: Order::Desc,
+    //};
+    // let index = create_index(Some(sort_by_field.clone()));
+    // let segments = index.searchable_segments().unwrap();
+    // let merger: IndexMerger =
+    // IndexMerger::open(index.schema(), index.settings().clone(), &segments[..])?;
+    // let doc_id_mapping = merger.generate_doc_id_mapping(&sort_by_field).unwrap();
+    // b.iter(|| {
+    // let sorted_doc_ids = doc_id_mapping.iter_old_doc_addrs().map(|doc_addr| {
+    // let reader = &merger.readers[doc_addr.segment_ord as usize];
+    // let u64_reader: Arc<dyn Column<u64>> = reader
+    //.fast_fields()
+    //.typed_fast_field_reader("intval")
+    //.expect(
+    //"Failed to find a reader for single fast field. This is a tantivy bug and \
+    // it should never happen.",
+    //);
+    //(doc_addr.doc_id, reader, u64_reader)
+    //});
+    /// add values in order of the new doc_ids
+    // let mut val = 0;
+    // for (doc_id, _reader, field_reader) in sorted_doc_ids {
+    // val = field_reader.get_val(doc_id);
+    //}
+
+    // val
+    //});
+
+    // Ok(())
+    //}
+    #[bench]
+    fn create_sorted_index_create_doc_id_mapping(b: &mut Bencher) -> crate::Result<()> {
+        let sort_by_field = IndexSortByField {
+            field: "intval".to_string(),
+            order: Order::Desc,
+        };
+        let index = create_index(Some(sort_by_field.clone()));
+        // let field = index.schema().get_field("intval").unwrap();
+        let segments = index.searchable_segments().unwrap();
+        let merger: IndexMerger =
+            IndexMerger::open(index.schema(), index.settings().clone(), &segments[..])?;
+        b.iter(|| {
+            merger
+                .generate_doc_id_mapping_with_sort_by_field(&sort_by_field)
+                .unwrap();
+        });
+
+        Ok(())
+    }
+}

src/indexer/mod.rs

@@ -13,10 +13,10 @@ mod flat_map_with_buffer;
 pub(crate) mod index_writer;
 pub(crate) mod index_writer_status;
 mod log_merge_policy;
-mod merge_index_test;
 mod merge_operation;
 pub(crate) mod merge_policy;
 pub(crate) mod merger;
+mod merger_sorted_index_test;
 pub(crate) mod operation;
 pub(crate) mod prepared_commit;
 mod segment_entry;
@@ -216,7 +216,7 @@ mod tests_mmap {
         let test_query = |query_str: &str| {
             let query = parse_query.parse_query(query_str).unwrap();
             let num_docs = searcher.search(&query, &Count).unwrap();
-            assert_eq!(num_docs, 1, "{query_str}");
+            assert_eq!(num_docs, 1, "{}", query_str);
         };
         test_query(format!("json.{field_name_out}:test1").as_str());
         test_query(format!("json.a{field_name_out}:test2").as_str());
@@ -590,10 +590,10 @@ mod tests_mmap {
         let query_parser = QueryParser::for_index(&index, vec![]);
         // Test if field name can be queried
         for (indexed_field, val) in fields_and_vals.iter() {
-            let query_str = &format!("{indexed_field}:{val}");
+            let query_str = &format!("{}:{}", indexed_field, val);
             let query = query_parser.parse_query(query_str).unwrap();
             let count_docs = searcher.search(&*query, &TopDocs::with_limit(2)).unwrap();
-            assert!(!count_docs.is_empty(), "{indexed_field}:{val}");
+            assert!(!count_docs.is_empty(), "{}:{}", indexed_field, val);
         }
         // Test if field name can be used for aggregation
         for (field_name, val) in fields_and_vals.iter() {

src/indexer/path_to_unordered_id.rs

@@ -1,5 +1,3 @@
-use common::json_path_writer::JSON_END_OF_PATH;
-use common::replace_in_place;
 use fnv::FnvHashMap;
 
 /// `Field` is represented by an unsigned 32-bit integer type.
@@ -40,14 +38,7 @@ impl PathToUnorderedId {
     #[cold]
     fn insert_new_path(&mut self, path: &str) -> u32 {
         let next_id = self.map.len() as u32;
-        let mut new_path = path.to_string();
-
-        // The unsafe below is safe as long as b'.' and JSON_PATH_SEGMENT_SEP are
-        // valid single byte ut8 strings.
-        // By utf-8 design, they cannot be part of another codepoint.
-        unsafe { replace_in_place(JSON_END_OF_PATH, b'0', new_path.as_bytes_mut()) };
-
-        self.map.insert(new_path, next_id);
+        self.map.insert(path.to_string(), next_id);
         next_id
     }
 

src/indexer/segment_serializer.rs

@@ -18,9 +18,27 @@ pub struct SegmentSerializer {
 
 impl SegmentSerializer {
     /// Creates a new `SegmentSerializer`.
-    pub fn for_segment(mut segment: Segment) -> crate::Result<SegmentSerializer> {
+    pub fn for_segment(
+        mut segment: Segment,
+        is_in_merge: bool,
+    ) -> crate::Result<SegmentSerializer> {
+        // If the segment is going to be sorted, we stream the docs first to a temporary file.
+        // In the merge case this is not necessary because we can kmerge the already sorted
+        // segments
+        let remapping_required = segment.index().settings().sort_by_field.is_some() && !is_in_merge;
         let settings = segment.index().settings().clone();
-        let store_writer = {
+        let store_writer = if remapping_required {
+            let store_write = segment.open_write(SegmentComponent::TempStore)?;
+            StoreWriter::new(
+                store_write,
+                crate::store::Compressor::None,
+                // We want fast random access on the docs, so we choose a small block size.
+                // If this is zero, the skip index will contain too many checkpoints and
+                // therefore will be relatively slow.
+                16000,
+                settings.docstore_compress_dedicated_thread,
+            )?
+        } else {
             let store_write = segment.open_write(SegmentComponent::Store)?;
             StoreWriter::new(
                 store_write,
@@ -54,6 +72,10 @@ impl SegmentSerializer {
         &self.segment
     }
 
+    pub fn segment_mut(&mut self) -> &mut Segment {
+        &mut self.segment
+    }
+
     /// Accessor to the `PostingsSerializer`.
     pub fn get_postings_serializer(&mut self) -> &mut InvertedIndexSerializer {
         &mut self.postings_serializer

src/indexer/segment_updater.rs

@@ -115,10 +115,11 @@ fn merge(
         .collect();
 
     // An IndexMerger is like a "view" of our merged segments.
-    let merger: IndexMerger = IndexMerger::open(index.schema(), &segments[..])?;
+    let merger: IndexMerger =
+        IndexMerger::open(index.schema(), index.settings().clone(), &segments[..])?;
 
     // ... we just serialize this index merger in our new segment to merge the segments.
-    let segment_serializer = SegmentSerializer::for_segment(merged_segment.clone())?;
+    let segment_serializer = SegmentSerializer::for_segment(merged_segment.clone(), true)?;
 
     let num_docs = merger.write(segment_serializer)?;
 
@@ -219,9 +220,13 @@ pub fn merge_filtered_segments<T: Into<Box<dyn Directory>>>(
     )?;
     let merged_segment = merged_index.new_segment();
     let merged_segment_id = merged_segment.id();
-    let merger: IndexMerger =
-        IndexMerger::open_with_custom_alive_set(merged_index.schema(), segments, filter_doc_ids)?;
-    let segment_serializer = SegmentSerializer::for_segment(merged_segment)?;
+    let merger: IndexMerger = IndexMerger::open_with_custom_alive_set(
+        merged_index.schema(),
+        merged_index.settings().clone(),
+        segments,
+        filter_doc_ids,
+    )?;
+    let segment_serializer = SegmentSerializer::for_segment(merged_segment, true)?;
     let num_docs = merger.write(segment_serializer)?;
 
     let segment_meta = merged_index.new_segment_meta(merged_segment_id, num_docs);
@@ -1062,6 +1067,7 @@ mod tests {
             )?;
             let merger: IndexMerger = IndexMerger::open_with_custom_alive_set(
                 merged_index.schema(),
+                merged_index.settings().clone(),
                 &segments[..],
                 filter_segments,
             )?;
@@ -1077,6 +1083,7 @@ mod tests {
                 Index::create(RamDirectory::default(), target_schema, target_settings)?;
             let merger: IndexMerger = IndexMerger::open_with_custom_alive_set(
                 merged_index.schema(),
+                merged_index.settings().clone(),
                 &segments[..],
                 filter_segments,
             )?;

src/indexer/segment_writer.rs

@@ -3,6 +3,7 @@ use common::JsonPathWriter;
 use itertools::Itertools;
 use tokenizer_api::BoxTokenStream;
 
+use super::doc_id_mapping::{get_doc_id_mapping_from_field, DocIdMapping};
 use super::operation::AddOperation;
 use crate::fastfield::FastFieldsWriter;
 use crate::fieldnorm::{FieldNormReaders, FieldNormsWriter};
@@ -15,6 +16,7 @@ use crate::postings::{
 };
 use crate::schema::document::{Document, Value};
 use crate::schema::{FieldEntry, FieldType, Schema, Term, DATE_TIME_PRECISION_INDEXED};
+use crate::store::{StoreReader, StoreWriter};
 use crate::tokenizer::{FacetTokenizer, PreTokenizedStream, TextAnalyzer, Tokenizer};
 use crate::{DocId, Opstamp, TantivyError};
 
@@ -39,6 +41,20 @@ fn compute_initial_table_size(per_thread_memory_budget: usize) -> crate::Result<
         })
 }
 
+fn remap_doc_opstamps(
+    opstamps: Vec<Opstamp>,
+    doc_id_mapping_opt: Option<&DocIdMapping>,
+) -> Vec<Opstamp> {
+    if let Some(doc_id_mapping_opt) = doc_id_mapping_opt {
+        doc_id_mapping_opt
+            .iter_old_doc_ids()
+            .map(|doc| opstamps[doc as usize])
+            .collect()
+    } else {
+        opstamps
+    }
+}
+
 /// A `SegmentWriter` is in charge of creating segment index from a
 /// set of documents.
 ///
@@ -74,7 +90,7 @@ impl SegmentWriter {
         let tokenizer_manager = segment.index().tokenizers().clone();
         let tokenizer_manager_fast_field = segment.index().fast_field_tokenizer().clone();
         let table_size = compute_initial_table_size(memory_budget_in_bytes)?;
-        let segment_serializer = SegmentSerializer::for_segment(segment)?;
+        let segment_serializer = SegmentSerializer::for_segment(segment, false)?;
         let per_field_postings_writers = PerFieldPostingsWriter::for_schema(&schema);
         let per_field_text_analyzers = schema
             .fields()
@@ -123,6 +139,15 @@ impl SegmentWriter {
     /// be used afterwards.
     pub fn finalize(mut self) -> crate::Result<Vec<u64>> {
         self.fieldnorms_writer.fill_up_to_max_doc(self.max_doc);
+        let mapping: Option<DocIdMapping> = self
+            .segment_serializer
+            .segment()
+            .index()
+            .settings()
+            .sort_by_field
+            .clone()
+            .map(|sort_by_field| get_doc_id_mapping_from_field(sort_by_field, &self))
+            .transpose()?;
         remap_and_write(
             self.schema,
             &self.per_field_postings_writers,
@@ -130,8 +155,10 @@ impl SegmentWriter {
             self.fast_field_writers,
             &self.fieldnorms_writer,
             self.segment_serializer,
+            mapping.as_ref(),
         )?;
-        Ok(self.doc_opstamps)
+        let doc_opstamps = remap_doc_opstamps(self.doc_opstamps, mapping.as_ref());
+        Ok(doc_opstamps)
     }
 
     /// Returns an estimation of the current memory usage of the segment writer.
@@ -175,8 +202,9 @@ impl SegmentWriter {
             match field_entry.field_type() {
                 FieldType::Facet(_) => {
                     let mut facet_tokenizer = FacetTokenizer::default(); // this can be global
-                    for value in values {
-                        let value = value.as_value();
+                    for value_access in values {
+                        // Used to help with linting and type checking.
+                        let value = value_access as D::Value<'_>;
 
                         let facet_str = value.as_facet().ok_or_else(make_schema_error)?;
                         let mut facet_tokenizer = facet_tokenizer.token_stream(facet_str);
@@ -192,14 +220,15 @@ impl SegmentWriter {
                 }
                 FieldType::Str(_) => {
                     let mut indexing_position = IndexingPosition::default();
-                    for value in values {
-                        let value = value.as_value();
+                    for value_access in values {
+                        // Used to help with linting and type checking.
+                        let value = value_access as D::Value<'_>;
 
                         let mut token_stream = if let Some(text) = value.as_str() {
                             let text_analyzer =
                                 &mut self.per_field_text_analyzers[field.field_id() as usize];
                             text_analyzer.token_stream(text)
-                        } else if let Some(tok_str) = value.into_pre_tokenized_text() {
+                        } else if let Some(tok_str) = value.as_pre_tokenized_text() {
                             BoxTokenStream::new(PreTokenizedStream::from(*tok_str.clone()))
                         } else {
                             continue;
@@ -221,8 +250,9 @@ impl SegmentWriter {
                 }
                 FieldType::U64(_) => {
                     let mut num_vals = 0;
-                    for value in values {
-                        let value = value.as_value();
+                    for value_access in values {
+                        // Used to help with linting and type checking.
+                        let value = value_access as D::Value<'_>;
 
                         num_vals += 1;
                         let u64_val = value.as_u64().ok_or_else(make_schema_error)?;
@@ -235,8 +265,10 @@ impl SegmentWriter {
                 }
                 FieldType::Date(_) => {
                     let mut num_vals = 0;
-                    for value in values {
-                        let value = value.as_value();
+                    for value_access in values {
+                        // Used to help with linting and type checking.
+                        let value_access = value_access as D::Value<'_>;
+                        let value = value_access.as_value();
 
                         num_vals += 1;
                         let date_val = value.as_datetime().ok_or_else(make_schema_error)?;
@@ -250,8 +282,9 @@ impl SegmentWriter {
                 }
                 FieldType::I64(_) => {
                     let mut num_vals = 0;
-                    for value in values {
-                        let value = value.as_value();
+                    for value_access in values {
+                        // Used to help with linting and type checking.
+                        let value = value_access as D::Value<'_>;
 
                         num_vals += 1;
                         let i64_val = value.as_i64().ok_or_else(make_schema_error)?;
@@ -264,8 +297,10 @@ impl SegmentWriter {
                 }
                 FieldType::F64(_) => {
                     let mut num_vals = 0;
-                    for value in values {
-                        let value = value.as_value();
+                    for value_access in values {
+                        // Used to help with linting and type checking.
+                        let value = value_access as D::Value<'_>;
+
                         num_vals += 1;
                         let f64_val = value.as_f64().ok_or_else(make_schema_error)?;
                         term_buffer.set_f64(f64_val);
@@ -277,8 +312,10 @@ impl SegmentWriter {
                 }
                 FieldType::Bool(_) => {
                     let mut num_vals = 0;
-                    for value in values {
-                        let value = value.as_value();
+                    for value_access in values {
+                        // Used to help with linting and type checking.
+                        let value = value_access as D::Value<'_>;
+
                         num_vals += 1;
                         let bool_val = value.as_bool().ok_or_else(make_schema_error)?;
                         term_buffer.set_bool(bool_val);
@@ -290,8 +327,10 @@ impl SegmentWriter {
                 }
                 FieldType::Bytes(_) => {
                     let mut num_vals = 0;
-                    for value in values {
-                        let value = value.as_value();
+                    for value_access in values {
+                        // Used to help with linting and type checking.
+                        let value = value_access as D::Value<'_>;
+
                         num_vals += 1;
                         let bytes = value.as_bytes().ok_or_else(make_schema_error)?;
                         term_buffer.set_bytes(bytes);
@@ -325,8 +364,9 @@ impl SegmentWriter {
                 }
                 FieldType::IpAddr(_) => {
                     let mut num_vals = 0;
-                    for value in values {
-                        let value = value.as_value();
+                    for value_access in values {
+                        // Used to help with linting and type checking.
+                        let value = value_access as D::Value<'_>;
 
                         num_vals += 1;
                         let ip_addr = value.as_ip_addr().ok_or_else(make_schema_error)?;
@@ -392,10 +432,11 @@ fn remap_and_write(
     fast_field_writers: FastFieldsWriter,
     fieldnorms_writer: &FieldNormsWriter,
     mut serializer: SegmentSerializer,
+    doc_id_map: Option<&DocIdMapping>,
 ) -> crate::Result<()> {
     debug!("remap-and-write");
     if let Some(fieldnorms_serializer) = serializer.extract_fieldnorms_serializer() {
-        fieldnorms_writer.serialize(fieldnorms_serializer)?;
+        fieldnorms_writer.serialize(fieldnorms_serializer, doc_id_map)?;
     }
     let fieldnorm_data = serializer
         .segment()
@@ -406,10 +447,39 @@ fn remap_and_write(
         schema,
         per_field_postings_writers,
         fieldnorm_readers,
+        doc_id_map,
         serializer.get_postings_serializer(),
     )?;
     debug!("fastfield-serialize");
-    fast_field_writers.serialize(serializer.get_fast_field_write())?;
+    fast_field_writers.serialize(serializer.get_fast_field_write(), doc_id_map)?;
+
+    // finalize temp docstore and create version, which reflects the doc_id_map
+    if let Some(doc_id_map) = doc_id_map {
+        debug!("resort-docstore");
+        let store_write = serializer
+            .segment_mut()
+            .open_write(SegmentComponent::Store)?;
+        let settings = serializer.segment().index().settings();
+        let store_writer = StoreWriter::new(
+            store_write,
+            settings.docstore_compression,
+            settings.docstore_blocksize,
+            settings.docstore_compress_dedicated_thread,
+        )?;
+        let old_store_writer = std::mem::replace(&mut serializer.store_writer, store_writer);
+        old_store_writer.close()?;
+        let store_read = StoreReader::open(
+            serializer
+                .segment()
+                .open_read(SegmentComponent::TempStore)?,
+            1, /* The docstore is configured to have one doc per block, and each doc is accessed
+                * only once: we don't need caching. */
+        )?;
+        for old_doc_id in doc_id_map.iter_old_doc_ids() {
+            let doc_bytes = store_read.get_document_bytes(old_doc_id)?;
+            serializer.get_store_writer().store_bytes(&doc_bytes)?;
+        }
+    }
 
     debug!("serializer-close");
     serializer.close()?;
@@ -430,8 +500,8 @@ mod tests {
     use crate::postings::{Postings, TermInfo};
     use crate::query::{PhraseQuery, QueryParser};
     use crate::schema::{
-        Document, IndexRecordOption, OwnedValue, Schema, TextFieldIndexing, TextOptions, Value,
-        STORED, STRING, TEXT,
+        Document, IndexRecordOption, OwnedValue, Schema, TextFieldIndexing, TextOptions, STORED,
+        STRING, TEXT,
     };
     use crate::store::{Compressor, StoreReader, StoreWriter};
     use crate::time::format_description::well_known::Rfc3339;
@@ -485,12 +555,9 @@ mod tests {
         let doc = reader.get::<TantivyDocument>(0).unwrap();
 
         assert_eq!(doc.field_values().count(), 2);
+        assert_eq!(doc.get_all(text_field).next().unwrap().as_str(), Some("A"));
         assert_eq!(
-            doc.get_all(text_field).next().unwrap().as_value().as_str(),
-            Some("A")
-        );
-        assert_eq!(
-            doc.get_all(text_field).nth(1).unwrap().as_value().as_str(),
+            doc.get_all(text_field).nth(1).unwrap().as_str(),
             Some("title")
         );
     }

src/lib.rs

@@ -222,8 +222,8 @@ pub use crate::core::{Executor, Searcher, SearcherGeneration};
 pub use crate::directory::Directory;
 #[allow(deprecated)] // Remove with index sorting
 pub use crate::index::{
-    Index, IndexBuilder, IndexMeta, IndexSettings, InvertedIndexReader, Order, Segment,
-    SegmentMeta, SegmentReader,
+    Index, IndexBuilder, IndexMeta, IndexSettings, IndexSortByField, InvertedIndexReader, Order,
+    Segment, SegmentMeta, SegmentReader,
 };
 pub use crate::indexer::{IndexWriter, SingleSegmentIndexWriter};
 pub use crate::schema::{Document, TantivyDocument, Term};
@@ -397,20 +397,16 @@ pub mod tests {
     #[macro_export]
     macro_rules! assert_nearly_equals {
         ($left:expr, $right:expr) => {{
-            assert_nearly_equals!($left, $right, 0.0005);
-        }};
-        ($left:expr, $right:expr, $epsilon:expr) => {{
-            match (&$left, &$right, &$epsilon) {
-                (left_val, right_val, epsilon_val) => {
+            match (&$left, &$right) {
+                (left_val, right_val) => {
                     let diff = (left_val - right_val).abs();
-
-                    if diff > *epsilon_val {
+                    let add = left_val.abs() + right_val.abs();
+                    if diff > 0.0005 * add {
                         panic!(
-                            r#"assertion failed: `abs(left-right)>epsilon`
-    left: `{:?}`,
-    right: `{:?}`,
-    epsilon: `{:?}`"#,
-                            &*left_val, &*right_val, &*epsilon_val
+                            r#"assertion failed: `(left ~= right)`
+  left: `{:?}`,
+ right: `{:?}`"#,
+                            &*left_val, &*right_val
                         )
                     }
                 }

src/postings/json_postings_writer.rs

@@ -3,6 +3,7 @@ use std::io;
 use common::json_path_writer::JSON_END_OF_PATH;
 use stacker::Addr;
 
+use crate::indexer::doc_id_mapping::DocIdMapping;
 use crate::indexer::path_to_unordered_id::OrderedPathId;
 use crate::postings::postings_writer::SpecializedPostingsWriter;
 use crate::postings::recorder::{BufferLender, DocIdRecorder, Recorder};
@@ -61,6 +62,7 @@ impl<Rec: Recorder> PostingsWriter for JsonPostingsWriter<Rec> {
         &self,
         term_addrs: &[(Field, OrderedPathId, &[u8], Addr)],
         ordered_id_to_path: &[&str],
+        doc_id_map: Option<&DocIdMapping>,
         ctx: &IndexingContext,
         serializer: &mut FieldSerializer,
     ) -> io::Result<()> {
@@ -85,6 +87,7 @@ impl<Rec: Recorder> PostingsWriter for JsonPostingsWriter<Rec> {
                     SpecializedPostingsWriter::<Rec>::serialize_one_term(
                         term_buffer.serialized_value_bytes(),
                         *addr,
+                        doc_id_map,
                         &mut buffer_lender,
                         ctx,
                         serializer,
@@ -93,6 +96,7 @@ impl<Rec: Recorder> PostingsWriter for JsonPostingsWriter<Rec> {
                     SpecializedPostingsWriter::<DocIdRecorder>::serialize_one_term(
                         term_buffer.serialized_value_bytes(),
                         *addr,
+                        doc_id_map,
                         &mut buffer_lender,
                         ctx,
                         serializer,

src/postings/postings_writer.rs

@@ -5,6 +5,7 @@ use std::ops::Range;
 use stacker::Addr;
 
 use crate::fieldnorm::FieldNormReaders;
+use crate::indexer::doc_id_mapping::DocIdMapping;
 use crate::indexer::path_to_unordered_id::OrderedPathId;
 use crate::postings::recorder::{BufferLender, Recorder};
 use crate::postings::{
@@ -49,6 +50,7 @@ pub(crate) fn serialize_postings(
     schema: Schema,
     per_field_postings_writers: &PerFieldPostingsWriter,
     fieldnorm_readers: FieldNormReaders,
+    doc_id_map: Option<&DocIdMapping>,
     serializer: &mut InvertedIndexSerializer,
 ) -> crate::Result<()> {
     // Replace unordered ids by ordered ids to be able to sort
@@ -84,6 +86,7 @@ pub(crate) fn serialize_postings(
         postings_writer.serialize(
             &term_offsets[byte_offsets],
             &ordered_id_to_path,
+            doc_id_map,
             &ctx,
             &mut field_serializer,
         )?;
@@ -119,6 +122,7 @@ pub(crate) trait PostingsWriter: Send + Sync {
         &self,
         term_addrs: &[(Field, OrderedPathId, &[u8], Addr)],
         ordered_id_to_path: &[&str],
+        doc_id_map: Option<&DocIdMapping>,
         ctx: &IndexingContext,
         serializer: &mut FieldSerializer,
     ) -> io::Result<()>;
@@ -183,6 +187,7 @@ impl<Rec: Recorder> SpecializedPostingsWriter<Rec> {
     pub(crate) fn serialize_one_term(
         term: &[u8],
         addr: Addr,
+        doc_id_map: Option<&DocIdMapping>,
         buffer_lender: &mut BufferLender,
         ctx: &IndexingContext,
         serializer: &mut FieldSerializer,
@@ -190,7 +195,7 @@ impl<Rec: Recorder> SpecializedPostingsWriter<Rec> {
         let recorder: Rec = ctx.term_index.read(addr);
         let term_doc_freq = recorder.term_doc_freq().unwrap_or(0u32);
         serializer.new_term(term, term_doc_freq, recorder.has_term_freq())?;
-        recorder.serialize(&ctx.arena, serializer, buffer_lender);
+        recorder.serialize(&ctx.arena, doc_id_map, serializer, buffer_lender);
         serializer.close_term()?;
         Ok(())
     }
@@ -224,12 +229,13 @@ impl<Rec: Recorder> PostingsWriter for SpecializedPostingsWriter<Rec> {
         &self,
         term_addrs: &[(Field, OrderedPathId, &[u8], Addr)],
         _ordered_id_to_path: &[&str],
+        doc_id_map: Option<&DocIdMapping>,
         ctx: &IndexingContext,
         serializer: &mut FieldSerializer,
     ) -> io::Result<()> {
         let mut buffer_lender = BufferLender::default();
         for (_field, _path_id, term, addr) in term_addrs {
-            Self::serialize_one_term(term, *addr, &mut buffer_lender, ctx, serializer)?;
+            Self::serialize_one_term(term, *addr, doc_id_map, &mut buffer_lender, ctx, serializer)?;
         }
         Ok(())
     }

src/postings/recorder.rs

@@ -1,6 +1,7 @@
 use common::read_u32_vint;
 use stacker::{ExpUnrolledLinkedList, MemoryArena};
 
+use crate::indexer::doc_id_mapping::DocIdMapping;
 use crate::postings::FieldSerializer;
 use crate::DocId;
 
@@ -70,6 +71,7 @@ pub(crate) trait Recorder: Copy + Default + Send + Sync + 'static {
     fn serialize(
         &self,
         arena: &MemoryArena,
+        doc_id_map: Option<&DocIdMapping>,
         serializer: &mut FieldSerializer<'_>,
         buffer_lender: &mut BufferLender,
     );
@@ -113,15 +115,26 @@ impl Recorder for DocIdRecorder {
     fn serialize(
         &self,
         arena: &MemoryArena,
+        doc_id_map: Option<&DocIdMapping>,
         serializer: &mut FieldSerializer<'_>,
         buffer_lender: &mut BufferLender,
     ) {
-        let buffer = buffer_lender.lend_u8();
+        let (buffer, doc_ids) = buffer_lender.lend_all();
         // TODO avoid reading twice.
         self.stack.read_to_end(arena, buffer);
-        let iter = get_sum_reader(VInt32Reader::new(&buffer[..]));
-        for doc_id in iter {
-            serializer.write_doc(doc_id, 0u32, &[][..]);
+        if let Some(doc_id_map) = doc_id_map {
+            let iter = get_sum_reader(VInt32Reader::new(&buffer[..]));
+            doc_ids.extend(iter.map(|old_doc_id| doc_id_map.get_new_doc_id(old_doc_id)));
+            doc_ids.sort_unstable();
+
+            for doc in doc_ids {
+                serializer.write_doc(*doc, 0u32, &[][..]);
+            }
+        } else {
+            let iter = get_sum_reader(VInt32Reader::new(&buffer[..]));
+            for doc_id in iter {
+                serializer.write_doc(doc_id, 0u32, &[][..]);
+            }
         }
     }
 
@@ -181,18 +194,35 @@ impl Recorder for TermFrequencyRecorder {
     fn serialize(
         &self,
         arena: &MemoryArena,
+        doc_id_map: Option<&DocIdMapping>,
         serializer: &mut FieldSerializer<'_>,
         buffer_lender: &mut BufferLender,
     ) {
         let buffer = buffer_lender.lend_u8();
         self.stack.read_to_end(arena, buffer);
         let mut u32_it = VInt32Reader::new(&buffer[..]);
-        let mut prev_doc = 0;
-        while let Some(delta_doc_id) = u32_it.next() {
-            let doc_id = prev_doc + delta_doc_id;
-            prev_doc = doc_id;
-            let term_freq = u32_it.next().unwrap_or(self.current_tf);
-            serializer.write_doc(doc_id, term_freq, &[][..]);
+        if let Some(doc_id_map) = doc_id_map {
+            let mut doc_id_and_tf = vec![];
+            let mut prev_doc = 0;
+            while let Some(delta_doc_id) = u32_it.next() {
+                let doc_id = prev_doc + delta_doc_id;
+                prev_doc = doc_id;
+                let term_freq = u32_it.next().unwrap_or(self.current_tf);
+                doc_id_and_tf.push((doc_id_map.get_new_doc_id(doc_id), term_freq));
+            }
+            doc_id_and_tf.sort_unstable_by_key(|&(doc_id, _)| doc_id);
+
+            for (doc_id, tf) in doc_id_and_tf {
+                serializer.write_doc(doc_id, tf, &[][..]);
+            }
+        } else {
+            let mut prev_doc = 0;
+            while let Some(delta_doc_id) = u32_it.next() {
+                let doc_id = prev_doc + delta_doc_id;
+                prev_doc = doc_id;
+                let term_freq = u32_it.next().unwrap_or(self.current_tf);
+                serializer.write_doc(doc_id, term_freq, &[][..]);
+            }
         }
     }
 
@@ -238,12 +268,14 @@ impl Recorder for TfAndPositionRecorder {
     fn serialize(
         &self,
         arena: &MemoryArena,
+        doc_id_map: Option<&DocIdMapping>,
         serializer: &mut FieldSerializer<'_>,
         buffer_lender: &mut BufferLender,
     ) {
         let (buffer_u8, buffer_positions) = buffer_lender.lend_all();
         self.stack.read_to_end(arena, buffer_u8);
         let mut u32_it = VInt32Reader::new(&buffer_u8[..]);
+        let mut doc_id_and_positions = vec![];
         let mut prev_doc = 0;
         while let Some(delta_doc_id) = u32_it.next() {
             let doc_id = prev_doc + delta_doc_id;
@@ -262,7 +294,19 @@ impl Recorder for TfAndPositionRecorder {
                     }
                 }
             }
-            serializer.write_doc(doc_id, buffer_positions.len() as u32, buffer_positions);
+            if let Some(doc_id_map) = doc_id_map {
+                // this simple variant to remap may consume to much memory
+                doc_id_and_positions
+                    .push((doc_id_map.get_new_doc_id(doc_id), buffer_positions.to_vec()));
+            } else {
+                serializer.write_doc(doc_id, buffer_positions.len() as u32, buffer_positions);
+            }
+        }
+        if doc_id_map.is_some() {
+            doc_id_and_positions.sort_unstable_by_key(|&(doc_id, _)| doc_id);
+            for (doc_id, positions) in doc_id_and_positions {
+                serializer.write_doc(doc_id, positions.len() as u32, &positions);
+            }
         }
     }
 

src/query/fuzzy_query.rs

@@ -138,7 +138,8 @@ impl FuzzyTermQuery {
                 if json_path_type != Type::Str {
                     return Err(InvalidArgument(format!(
                         "The fuzzy term query requires a string path type for a json term. Found \
-                         {json_path_type:?}"
+                         {:?}",
+                        json_path_type
                     )));
                 }
             }

src/query/query_parser/logical_ast.rs

@@ -2,7 +2,7 @@ use std::fmt;
 use std::ops::Bound;
 
 use crate::query::Occur;
-use crate::schema::{Term, Type};
+use crate::schema::{Field, Term, Type};
 use crate::Score;
 
 #[derive(Clone)]
@@ -20,6 +20,8 @@ pub enum LogicalLiteral {
         upper: Bound<Term>,
     },
     Set {
+        field: Field,
+        value_type: Type,
         elements: Vec<Term>,
     },
     All,

src/query/query_parser/query_parser.rs

@@ -832,11 +832,17 @@ impl QueryParser {
                 let (field, json_path) = try_tuple!(self
                     .split_full_path(&full_path)
                     .ok_or_else(|| QueryParserError::FieldDoesNotExist(full_path.clone())));
+                let field_entry = self.schema.get_field_entry(field);
+                let value_type = field_entry.field_type().value_type();
                 let (elements, errors) = elements
                     .into_iter()
                     .map(|element| self.compute_boundary_term(field, json_path, &element))
                     .partition_result();
-                let logical_ast = LogicalAst::Leaf(Box::new(LogicalLiteral::Set { elements }));
+                let logical_ast = LogicalAst::Leaf(Box::new(LogicalLiteral::Set {
+                    elements,
+                    field,
+                    value_type,
+                }));
                 (Some(logical_ast), errors)
             }
             UserInputLeaf::Exists { .. } => (

src/query/regex_query.rs

@@ -185,7 +185,7 @@ mod test {
             Err(crate::TantivyError::InvalidArgument(msg)) => {
                 assert!(msg.contains("error: unclosed group"))
             }
-            res => panic!("unexpected result: {res:?}"),
+            res => panic!("unexpected result: {:?}", res),
         }
     }
 }

src/schema/document/default_document.rs

@@ -157,24 +157,29 @@ impl CompactDoc {
     }
 
     /// field_values accessor
-    pub fn field_values(&self) -> impl Iterator<Item = (Field, CompactDocValue<'_>)> {
+    pub fn field_values(
+        &self,
+    ) -> impl Iterator<Item = (Field, ReferenceValue<'_, CompactDocValue<'_>>)> {
         self.field_values.iter().map(|field_val| {
             let field = Field::from_field_id(field_val.field as u32);
-            let val = self.get_compact_doc_value(field_val.value_addr);
+            let val = self.extract_value(field_val.value_addr).unwrap();
             (field, val)
         })
     }
 
     /// Returns all of the `ReferenceValue`s associated the given field
-    pub fn get_all(&self, field: Field) -> impl Iterator<Item = CompactDocValue<'_>> + '_ {
+    pub fn get_all(
+        &self,
+        field: Field,
+    ) -> impl Iterator<Item = ReferenceValue<'_, CompactDocValue<'_>>> + '_ {
         self.field_values
             .iter()
             .filter(move |field_value| Field::from_field_id(field_value.field as u32) == field)
-            .map(|val| self.get_compact_doc_value(val.value_addr))
+            .map(|val| self.extract_value(val.value_addr).unwrap())
     }
 
     /// Returns the first `ReferenceValue` associated the given field
-    pub fn get_first(&self, field: Field) -> Option<CompactDocValue<'_>> {
+    pub fn get_first(&self, field: Field) -> Option<ReferenceValue<'_, CompactDocValue<'_>>> {
         self.get_all(field).next()
     }
 
@@ -294,11 +299,58 @@ impl CompactDoc {
         }
     }
 
-    /// Get CompactDocValue for address
-    fn get_compact_doc_value(&self, value_addr: ValueAddr) -> CompactDocValue<'_> {
-        CompactDocValue {
-            container: self,
-            value_addr,
+    fn extract_value(
+        &self,
+        ref_value: ValueAddr,
+    ) -> io::Result<ReferenceValue<'_, CompactDocValue<'_>>> {
+        match ref_value.type_id {
+            ValueType::Null => Ok(ReferenceValueLeaf::Null.into()),
+            ValueType::Str => {
+                let str_ref = self.extract_str(ref_value.val_addr);
+                Ok(ReferenceValueLeaf::Str(str_ref).into())
+            }
+            ValueType::Facet => {
+                let str_ref = self.extract_str(ref_value.val_addr);
+                Ok(ReferenceValueLeaf::Facet(str_ref).into())
+            }
+            ValueType::Bytes => {
+                let data = self.extract_bytes(ref_value.val_addr);
+                Ok(ReferenceValueLeaf::Bytes(data).into())
+            }
+            ValueType::U64 => self
+                .read_from::<u64>(ref_value.val_addr)
+                .map(ReferenceValueLeaf::U64)
+                .map(Into::into),
+            ValueType::I64 => self
+                .read_from::<i64>(ref_value.val_addr)
+                .map(ReferenceValueLeaf::I64)
+                .map(Into::into),
+            ValueType::F64 => self
+                .read_from::<f64>(ref_value.val_addr)
+                .map(ReferenceValueLeaf::F64)
+                .map(Into::into),
+            ValueType::Bool => Ok(ReferenceValueLeaf::Bool(ref_value.val_addr != 0).into()),
+            ValueType::Date => self
+                .read_from::<i64>(ref_value.val_addr)
+                .map(|ts| ReferenceValueLeaf::Date(DateTime::from_timestamp_nanos(ts)))
+                .map(Into::into),
+            ValueType::IpAddr => self
+                .read_from::<u128>(ref_value.val_addr)
+                .map(|num| ReferenceValueLeaf::IpAddr(Ipv6Addr::from_u128(num)))
+                .map(Into::into),
+            ValueType::PreTokStr => self
+                .read_from::<PreTokenizedString>(ref_value.val_addr)
+                .map(Into::into)
+                .map(ReferenceValueLeaf::PreTokStr)
+                .map(Into::into),
+            ValueType::Object => Ok(ReferenceValue::Object(CompactDocObjectIter::new(
+                self,
+                ref_value.val_addr,
+            )?)),
+            ValueType::Array => Ok(ReferenceValue::Array(CompactDocArrayIter::new(
+                self,
+                ref_value.val_addr,
+            )?)),
         }
     }
 
@@ -358,7 +410,7 @@ impl PartialEq for CompactDoc {
         let convert_to_comparable_map = |doc: &CompactDoc| {
             let mut field_value_set: HashMap<Field, HashSet<String>> = Default::default();
             for field_value in doc.field_values.iter() {
-                let value: OwnedValue = doc.get_compact_doc_value(field_value.value_addr).into();
+                let value: OwnedValue = doc.extract_value(field_value.value_addr).unwrap().into();
                 let value = serde_json::to_string(&value).unwrap();
                 field_value_set
                     .entry(Field::from_field_id(field_value.field as u32))
@@ -392,19 +444,7 @@ impl DocumentDeserialize for CompactDoc {
 #[derive(Debug, Clone, Copy)]
 pub struct CompactDocValue<'a> {
     container: &'a CompactDoc,
-    value_addr: ValueAddr,
-}
-impl PartialEq for CompactDocValue<'_> {
-    fn eq(&self, other: &Self) -> bool {
-        let value1: OwnedValue = (*self).into();
-        let value2: OwnedValue = (*other).into();
-        value1 == value2
-    }
-}
-impl<'a> From<CompactDocValue<'a>> for OwnedValue {
-    fn from(value: CompactDocValue) -> Self {
-        value.as_value().into()
-    }
+    value: ValueAddr,
 }
 impl<'a> Value<'a> for CompactDocValue<'a> {
     type ArrayIter = CompactDocArrayIter<'a>;
@@ -412,67 +452,7 @@ impl<'a> Value<'a> for CompactDocValue<'a> {
     type ObjectIter = CompactDocObjectIter<'a>;
 
     fn as_value(&self) -> ReferenceValue<'a, Self> {
-        self.get_ref_value().unwrap()
-    }
-}
-impl<'a> CompactDocValue<'a> {
-    fn get_ref_value(&self) -> io::Result<ReferenceValue<'a, CompactDocValue<'a>>> {
-        let addr = self.value_addr.val_addr;
-        match self.value_addr.type_id {
-            ValueType::Null => Ok(ReferenceValueLeaf::Null.into()),
-            ValueType::Str => {
-                let str_ref = self.container.extract_str(addr);
-                Ok(ReferenceValueLeaf::Str(str_ref).into())
-            }
-            ValueType::Facet => {
-                let str_ref = self.container.extract_str(addr);
-                Ok(ReferenceValueLeaf::Facet(str_ref).into())
-            }
-            ValueType::Bytes => {
-                let data = self.container.extract_bytes(addr);
-                Ok(ReferenceValueLeaf::Bytes(data).into())
-            }
-            ValueType::U64 => self
-                .container
-                .read_from::<u64>(addr)
-                .map(ReferenceValueLeaf::U64)
-                .map(Into::into),
-            ValueType::I64 => self
-                .container
-                .read_from::<i64>(addr)
-                .map(ReferenceValueLeaf::I64)
-                .map(Into::into),
-            ValueType::F64 => self
-                .container
-                .read_from::<f64>(addr)
-                .map(ReferenceValueLeaf::F64)
-                .map(Into::into),
-            ValueType::Bool => Ok(ReferenceValueLeaf::Bool(addr != 0).into()),
-            ValueType::Date => self
-                .container
-                .read_from::<i64>(addr)
-                .map(|ts| ReferenceValueLeaf::Date(DateTime::from_timestamp_nanos(ts)))
-                .map(Into::into),
-            ValueType::IpAddr => self
-                .container
-                .read_from::<u128>(addr)
-                .map(|num| ReferenceValueLeaf::IpAddr(Ipv6Addr::from_u128(num)))
-                .map(Into::into),
-            ValueType::PreTokStr => self
-                .container
-                .read_from::<PreTokenizedString>(addr)
-                .map(Into::into)
-                .map(ReferenceValueLeaf::PreTokStr)
-                .map(Into::into),
-            ValueType::Object => Ok(ReferenceValue::Object(CompactDocObjectIter::new(
-                self.container,
-                addr,
-            )?)),
-            ValueType::Array => Ok(ReferenceValue::Array(CompactDocArrayIter::new(
-                self.container,
-                addr,
-            )?)),
-        }
+        self.container.extract_value(self.value).unwrap()
     }
 }
 
@@ -557,7 +537,7 @@ impl BinarySerializable for ValueType {
         } else {
             return Err(io::Error::new(
                 io::ErrorKind::InvalidData,
-                format!("Invalid value type id: {num}"),
+                format!("Invalid value type id: {}", num),
             ));
         };
         Ok(type_id)
@@ -621,9 +601,9 @@ impl<'a> Iterator for CompactDocObjectIter<'a> {
         let value = ValueAddr::deserialize(&mut self.node_addresses_slice).ok()?;
         let value = CompactDocValue {
             container: self.container,
-            value_addr: value,
+            value,
         };
-        Some((key, value))
+        return Some((key, value));
     }
 }
 
@@ -655,9 +635,9 @@ impl<'a> Iterator for CompactDocArrayIter<'a> {
         let value = ValueAddr::deserialize(&mut self.node_addresses_slice).ok()?;
         let value = CompactDocValue {
             container: self.container,
-            value_addr: value,
+            value,
         };
-        Some(value)
+        return Some(value);
     }
 }
 
@@ -688,7 +668,7 @@ impl<'a> Iterator for FieldValueIterRef<'a> {
                 Field::from_field_id(field_value.field as u32),
                 CompactDocValue::<'a> {
                     container: self.container,
-                    value_addr: field_value.value_addr,
+                    value: field_value.value_addr,
                 },
             )
         })

src/schema/document/se.rs

@@ -58,8 +58,9 @@ where W: Write
             return Err(io::Error::new(
                 io::ErrorKind::Other,
                 format!(
-                    "Unexpected number of entries written to serializer, expected \
-                     {num_field_values} entries, got {actual_length} entries",
+                    "Unexpected number of entries written to serializer, expected {} entries, got \
+                     {} entries",
+                    num_field_values, actual_length,
                 ),
             ));
         }

src/schema/document/value.rs

@@ -17,6 +17,15 @@ pub trait Value<'a>: Send + Sync + Debug {
     /// Returns the field value represented by an enum which borrows it's data.
     fn as_value(&self) -> ReferenceValue<'a, Self>;
 
+    #[inline]
+    /// Returns if the value is `null` or not.
+    fn is_null(&self) -> bool {
+        matches!(
+            self.as_value(),
+            ReferenceValue::Leaf(ReferenceValueLeaf::Null)
+        )
+    }
+
     #[inline]
     /// If the Value is a leaf, returns the associated leaf. Returns None otherwise.
     fn as_leaf(&self) -> Option<ReferenceValueLeaf<'a>> {
@@ -108,6 +117,18 @@ pub trait Value<'a>: Send + Sync + Debug {
             None
         }
     }
+
+    #[inline]
+    /// Returns true if the Value is an array.
+    fn is_array(&self) -> bool {
+        matches!(self.as_value(), ReferenceValue::Object(_))
+    }
+
+    #[inline]
+    /// Returns true if the Value is an object.
+    fn is_object(&self) -> bool {
+        matches!(self.as_value(), ReferenceValue::Object(_))
+    }
 }
 
 /// A enum representing a leaf value for tantivy to index.

src/schema/field_type.rs

@@ -659,9 +659,9 @@ mod tests {
         let schema = schema_builder.build();
         let doc_json = r#"{"date": "2019-10-12T07:20:50.52+02:00"}"#;
         let doc = TantivyDocument::parse_json(&schema, doc_json).unwrap();
-        let date = OwnedValue::from(doc.get_first(date_field).unwrap());
+        let date = doc.get_first(date_field).unwrap();
         // Time zone is converted to UTC
-        assert_eq!("Date(2019-10-12T05:20:50.52Z)", format!("{date:?}"));
+        assert_eq!("Leaf(Date(2019-10-12T05:20:50.52Z))", format!("{date:?}"));
     }
 
     #[test]

src/schema/term.rs

@@ -249,12 +249,9 @@ impl Term {
     #[inline]
     pub fn append_path(&mut self, bytes: &[u8]) -> &mut [u8] {
         let len_before = self.0.len();
-        if bytes.contains(&JSON_END_OF_PATH) {
-            self.0.extend(
-                bytes
-                    .iter()
-                    .map(|&b| if b == JSON_END_OF_PATH { b'0' } else { b }),
-            );
+        if bytes.contains(&0u8) {
+            self.0
+                .extend(bytes.iter().map(|&b| if b == 0 { b'0' } else { b }));
         } else {
             self.0.extend_from_slice(bytes);
         }

src/store/mod.rs

@@ -60,7 +60,7 @@ pub mod tests {
     use crate::directory::{Directory, RamDirectory, WritePtr};
     use crate::fastfield::AliveBitSet;
     use crate::schema::{
-        self, Schema, TantivyDocument, TextFieldIndexing, TextOptions, Value, STORED, TEXT,
+        self, Schema, TantivyDocument, TextFieldIndexing, TextOptions, STORED, TEXT,
     };
     use crate::{Index, IndexWriter, Term};
 
@@ -122,7 +122,6 @@ pub mod tests {
                     .get::<TantivyDocument>(i)?
                     .get_first(field_title)
                     .unwrap()
-                    .as_value()
                     .as_str()
                     .unwrap(),
                 format!("Doc {i}")
@@ -134,7 +133,6 @@ pub mod tests {
             let title_content = doc
                 .get_first(field_title)
                 .unwrap()
-                .as_value()
                 .as_str()
                 .unwrap()
                 .to_string();

src/store/reader.rs

@@ -241,12 +241,23 @@ impl StoreReader {
         &'b self,
         alive_bitset: Option<&'a AliveBitSet>,
     ) -> impl Iterator<Item = crate::Result<D>> + 'b {
+        self.enumerate(alive_bitset)
+            .map(|res| res.map(|(_, doc)| doc))
+    }
+
+    /// A variant of [`iter`][Self::iter] which also yields document ID.
+    pub fn enumerate<'a: 'b, 'b, D: DocumentDeserialize>(
+        &'b self,
+        alive_bitset: Option<&'a AliveBitSet>,
+    ) -> impl Iterator<Item = crate::Result<(DocId, D)>> + 'b {
         self.iter_raw(alive_bitset).map(|doc_bytes_res| {
-            let mut doc_bytes = doc_bytes_res?;
+            let (doc_id, mut doc_bytes) = doc_bytes_res?;
 
             let deserializer = BinaryDocumentDeserializer::from_reader(&mut doc_bytes)
                 .map_err(crate::TantivyError::from)?;
-            D::deserialize(deserializer).map_err(crate::TantivyError::from)
+            let doc = D::deserialize(deserializer).map_err(crate::TantivyError::from)?;
+
+            Ok((doc_id, doc))
         })
     }
 
@@ -256,7 +267,7 @@ impl StoreReader {
     pub(crate) fn iter_raw<'a: 'b, 'b>(
         &'b self,
         alive_bitset: Option<&'a AliveBitSet>,
-    ) -> impl Iterator<Item = crate::Result<OwnedBytes>> + 'b {
+    ) -> impl Iterator<Item = crate::Result<(DocId, OwnedBytes)>> + 'b {
         let last_doc_id = self
             .block_checkpoints()
             .last()
@@ -284,14 +295,14 @@ impl StoreReader {
 
                 let alive = alive_bitset.map_or(true, |bitset| bitset.is_alive(doc_id));
                 let res = if alive {
-                    Some((curr_block.clone(), doc_pos))
+                    Some((doc_id, curr_block.clone(), doc_pos))
                 } else {
                     None
                 };
                 doc_pos += 1;
                 res
             })
-            .map(move |(block, doc_pos)| {
+            .map(move |(doc_id, block, doc_pos)| {
                 let block = block
                     .ok_or_else(|| {
                         DataCorruption::comment_only(
@@ -304,7 +315,7 @@ impl StoreReader {
                     })?;
 
                 let range = block_read_index(&block, doc_pos)?;
-                Ok(block.slice(range))
+                Ok((doc_id, block.slice(range)))
             })
     }
 
@@ -403,7 +414,7 @@ mod tests {
 
     use super::*;
     use crate::directory::RamDirectory;
-    use crate::schema::{Field, TantivyDocument, Value};
+    use crate::schema::{Field, TantivyDocument};
     use crate::store::tests::write_lorem_ipsum_store;
     use crate::store::Compressor;
     use crate::Directory;
@@ -411,7 +422,7 @@ mod tests {
     const BLOCK_SIZE: usize = 16_384;
 
     fn get_text_field<'a>(doc: &'a TantivyDocument, field: &'a Field) -> Option<&'a str> {
-        doc.get_first(*field).and_then(|f| f.as_value().as_str())
+        doc.get_first(*field).and_then(|f| f.as_str())
     }
 
     #[test]

src/termdict/fst_termdict/termdict.rs

@@ -93,7 +93,7 @@ fn open_fst_index(fst_file: FileSlice) -> io::Result<tantivy_fst::Map<OwnedBytes
     let fst = Fst::new(bytes).map_err(|err| {
         io::Error::new(
             io::ErrorKind::InvalidData,
-            format!("Fst data is corrupted: {err:?}"),
+            format!("Fst data is corrupted: {:?}", err),
         )
     })?;
     Ok(tantivy_fst::Map::from(fst))

src/termdict/tests.rs

@@ -95,7 +95,7 @@ fn test_term_dictionary_simple() -> crate::Result<()> {
 #[test]
 fn test_term_dictionary_stream() -> crate::Result<()> {
     let ids: Vec<_> = (0u32..10_000u32)
-        .map(|i| (format!("doc{i:0>6}"), i))
+        .map(|i| (format!("doc{:0>6}", i), i))
         .collect();
     let buffer: Vec<u8> = {
         let mut term_dictionary_builder = TermDictionaryBuilder::create(vec![]).unwrap();
@@ -156,7 +156,7 @@ fn test_stream_high_range_prefix_suffix() -> crate::Result<()> {
 #[test]
 fn test_stream_range() -> crate::Result<()> {
     let ids: Vec<_> = (0u32..10_000u32)
-        .map(|i| (format!("doc{i:0>6}"), i))
+        .map(|i| (format!("doc{:0>6}", i), i))
         .collect();
     let buffer: Vec<u8> = {
         let mut term_dictionary_builder = TermDictionaryBuilder::create(vec![]).unwrap();

src/tokenizer/facet_tokenizer.rs

@@ -96,7 +96,7 @@ mod tests {
         {
             let mut add_token = |token: &Token| {
                 let facet = Facet::from_encoded(token.text.as_bytes().to_owned()).unwrap();
-                tokens.push(format!("{facet}"));
+                tokens.push(format!("{}", facet));
             };
             FacetTokenizer::default()
                 .token_stream(facet.encoded_str())
@@ -116,7 +116,7 @@ mod tests {
         {
             let mut add_token = |token: &Token| {
                 let facet = Facet::from_encoded(token.text.as_bytes().to_owned()).unwrap(); // ok test
-                tokens.push(format!("{facet}"));
+                tokens.push(format!("{}", facet));
             };
             FacetTokenizer::default()
                 .token_stream(facet.encoded_str()) // ok test