Removing the need for column on Multivalued blabla

fix num_vals on u128 value index after merge

CHANGELOG.md

@@ -1,6 +1,7 @@
 Tantivy 0.19
 ================================
 
+- Skip score calculation, when no scoring is required [#1646](https://github.com/quickwit-oss/tantivy/pull/1646) (@PSeitz)
 - Limit fast fields to u32 (`get_val(u32)`) [#1644](https://github.com/quickwit-oss/tantivy/pull/1644) (@PSeitz)
 - Major bugfix: Fix missing fieldnorms for u64, i64, f64, bool, bytes and date [#1620](https://github.com/quickwit-oss/tantivy/pull/1620) (@PSeitz)
 - Updated [Date Field Type](https://github.com/quickwit-oss/tantivy/pull/1396)

Cargo.toml

@@ -71,10 +71,10 @@ maplit = "1.0.2"
 matches = "0.1.9"
 pretty_assertions = "1.2.1"
 proptest = "1.0.0"
-criterion = "0.3.5"
+criterion = "0.4"
 test-log = "0.2.10"
 env_logger = "0.9.0"
-pprof = { version = "0.10.0", features = ["flamegraph", "criterion"] }
+pprof = { version = "0.11.0", features = ["flamegraph", "criterion"] }
 futures = "0.3.21"
 
 [dev-dependencies.fail]

common/src/vint.rs

@@ -157,7 +157,7 @@ fn vint_len(data: &[u8]) -> usize {
 /// If the buffer does not start by a valid
 /// vint payload
 pub fn read_u32_vint(data: &mut &[u8]) -> u32 {
-    let (result, vlen) = read_u32_vint_no_advance(*data);
+    let (result, vlen) = read_u32_vint_no_advance(data);
     *data = &data[vlen..];
     result
 }

fastfield_codecs/benches/bench.rs

@@ -113,7 +113,7 @@ mod tests {
 
         b.iter(|| {
             let mut positions = Vec::new();
-            column.get_positions_for_value_range(
+            column.get_docids_for_value_range(
                 major_item..=major_item,
                 0..data.len() as u32,
                 &mut positions,
@@ -129,7 +129,7 @@ mod tests {
 
         b.iter(|| {
             let mut positions = Vec::new();
-            column.get_positions_for_value_range(
+            column.get_docids_for_value_range(
                 minor_item..=minor_item,
                 0..data.len() as u32,
                 &mut positions,
@@ -145,11 +145,7 @@ mod tests {
 
         b.iter(|| {
             let mut positions = Vec::new();
-            column.get_positions_for_value_range(
-                0..=u128::MAX,
-                0..data.len() as u32,
-                &mut positions,
-            );
+            column.get_docids_for_value_range(0..=u128::MAX, 0..data.len() as u32, &mut positions);
             positions
         });
     }

fastfield_codecs/src/column.rs

@@ -35,7 +35,7 @@ pub trait Column<T: PartialOrd = u64>: Send + Sync {
     ///
     /// Note that position == docid for single value fast fields
     #[inline]
-    fn get_positions_for_value_range(
+    fn get_docids_for_value_range(
         &self,
         value_range: RangeInclusive<T>,
         doc_id_range: Range<u32>,
@@ -70,6 +70,11 @@ pub trait Column<T: PartialOrd = u64>: Send + Sync {
     /// The number of values in the column.
     fn num_vals(&self) -> u32;
 
+    /// The number of docs in the column. For single value columns this equals num_vals.
+    fn num_docs(&self) -> u32 {
+        self.num_vals()
+    }
+
     /// Returns a iterator over the data
     fn iter<'a>(&'a self) -> Box<dyn Iterator<Item = T> + 'a> {
         Box::new((0..self.num_vals()).map(|idx| self.get_val(idx)))
@@ -222,13 +227,13 @@ where
         )
     }
 
-    fn get_positions_for_value_range(
+    fn get_docids_for_value_range(
         &self,
         range: RangeInclusive<Output>,
         doc_id_range: Range<u32>,
         positions: &mut Vec<u32>,
     ) {
-        self.from_column.get_positions_for_value_range(
+        self.from_column.get_docids_for_value_range(
             self.monotonic_mapping.inverse(range.start().clone())
                 ..=self.monotonic_mapping.inverse(range.end().clone()),
             doc_id_range,
@@ -240,6 +245,7 @@ where
     // and we do not have any specialized implementation anyway.
 }
 
+/// Wraps an iterator into a `Column`.
 pub struct IterColumn<T>(T);
 
 impl<T> From<T> for IterColumn<T>

fastfield_codecs/src/compact_space/mod.rs

@@ -306,13 +306,13 @@ impl Column<u128> for CompactSpaceDecompressor {
     }
 
     #[inline]
-    fn get_positions_for_value_range(
+    fn get_docids_for_value_range(
         &self,
         value_range: RangeInclusive<u128>,
-        doc_id_range: Range<u32>,
+        positions_range: Range<u32>,
         positions: &mut Vec<u32>,
     ) {
-        self.get_positions_for_value_range(value_range, doc_id_range, positions)
+        self.get_positions_for_value_range(value_range, positions_range, positions)
     }
 }
 
@@ -351,13 +351,13 @@ impl CompactSpaceDecompressor {
     pub fn get_positions_for_value_range(
         &self,
         value_range: RangeInclusive<u128>,
-        doc_id_range: Range<u32>,
+        position_range: Range<u32>,
         positions: &mut Vec<u32>,
     ) {
         if value_range.start() > value_range.end() {
             return;
         }
-        let doc_id_range = doc_id_range.start..doc_id_range.end.min(self.num_vals());
+        let position_range = position_range.start..position_range.end.min(self.num_vals());
         let from_value = *value_range.start();
         let to_value = *value_range.end();
         assert!(to_value >= from_value);
@@ -390,10 +390,10 @@ impl CompactSpaceDecompressor {
 
         let range = compact_from..=compact_to;
 
-        let scan_num_docs = doc_id_range.end - doc_id_range.start;
+        let scan_num_docs = position_range.end - position_range.start;
 
         let step_size = 4;
-        let cutoff = doc_id_range.start + scan_num_docs - scan_num_docs % step_size;
+        let cutoff = position_range.start + scan_num_docs - scan_num_docs % step_size;
 
         let mut push_if_in_range = |idx, val| {
             if range.contains(&val) {
@@ -402,7 +402,7 @@ impl CompactSpaceDecompressor {
         };
         let get_val = |idx| self.params.bit_unpacker.get(idx, &self.data);
         // unrolled loop
-        for idx in (doc_id_range.start..cutoff).step_by(step_size as usize) {
+        for idx in (position_range.start..cutoff).step_by(step_size as usize) {
             let idx1 = idx;
             let idx2 = idx + 1;
             let idx3 = idx + 2;
@@ -418,7 +418,7 @@ impl CompactSpaceDecompressor {
         }
 
         // handle rest
-        for idx in cutoff..doc_id_range.end {
+        for idx in cutoff..position_range.end {
             push_if_in_range(idx, get_val(idx as u32));
         }
     }
@@ -704,7 +704,7 @@ mod tests {
         doc_id_range: Range<u32>,
     ) -> Vec<u32> {
         let mut positions = Vec::new();
-        column.get_positions_for_value_range(value_range, doc_id_range, &mut positions);
+        column.get_docids_for_value_range(value_range, doc_id_range, &mut positions);
         positions
     }
 

fastfield_codecs/src/lib.rs

@@ -41,7 +41,7 @@ mod serialize;
 
 use self::bitpacked::BitpackedCodec;
 use self::blockwise_linear::BlockwiseLinearCodec;
-pub use self::column::{monotonic_map_column, Column, VecColumn};
+pub use self::column::{monotonic_map_column, Column, IterColumn, VecColumn};
 use self::linear::LinearCodec;
 pub use self::monotonic_mapping::{MonotonicallyMappableToU64, StrictlyMonotonicFn};
 pub use self::monotonic_mapping_u128::MonotonicallyMappableToU128;
@@ -218,7 +218,7 @@ mod tests {
                 .map(|(pos, _)| pos as u32)
                 .collect();
             let mut positions = Vec::new();
-            reader.get_positions_for_value_range(
+            reader.get_docids_for_value_range(
                 data[test_rand_idx]..=data[test_rand_idx],
                 0..data.len() as u32,
                 &mut positions,

fastfield_codecs/src/main.rs

@@ -119,7 +119,7 @@ fn bench_ip() {
     for value in dataset.iter().take(1110).skip(1100).cloned() {
         doc_values.clear();
         print_time!("get range");
-        decompressor.get_positions_for_value_range(
+        decompressor.get_docids_for_value_range(
             value..=value,
             0..decompressor.num_vals(),
             &mut doc_values,

src/aggregation/agg_result.rs

@@ -4,8 +4,6 @@
 //! intermediate average results, which is the sum and the number of values. The actual average is
 //! calculated on the step from intermediate to final aggregation result tree.
 
-use std::collections::HashMap;
-
 use rustc_hash::FxHashMap;
 use serde::{Deserialize, Serialize};
 
@@ -18,7 +16,7 @@ use crate::TantivyError;
 
 #[derive(Clone, Default, Debug, PartialEq, Serialize, Deserialize)]
 /// The final aggegation result.
-pub struct AggregationResults(pub HashMap<String, AggregationResult>);
+pub struct AggregationResults(pub FxHashMap<String, AggregationResult>);
 
 impl AggregationResults {
     pub(crate) fn get_value_from_aggregation(

src/aggregation/intermediate_agg_result.rs

@@ -3,7 +3,6 @@
 //! indices.
 
 use std::cmp::Ordering;
-use std::collections::HashMap;
 
 use itertools::Itertools;
 use rustc_hash::FxHashMap;
@@ -51,7 +50,7 @@ impl IntermediateAggregationResults {
         // Important assumption:
         // When the tree contains buckets/metric, we expect it to have all buckets/metrics from the
         // request
-        let mut results: HashMap<String, AggregationResult> = HashMap::new();
+        let mut results: FxHashMap<String, AggregationResult> = FxHashMap::default();
 
         if let Some(buckets) = self.buckets {
             convert_and_add_final_buckets_to_result(&mut results, buckets, &req.buckets)?
@@ -132,7 +131,7 @@ impl IntermediateAggregationResults {
 }
 
 fn convert_and_add_final_metrics_to_result(
-    results: &mut HashMap<String, AggregationResult>,
+    results: &mut FxHashMap<String, AggregationResult>,
     metrics: VecWithNames<IntermediateMetricResult>,
 ) {
     results.extend(
@@ -143,7 +142,7 @@ fn convert_and_add_final_metrics_to_result(
 }
 
 fn add_empty_final_metrics_to_result(
-    results: &mut HashMap<String, AggregationResult>,
+    results: &mut FxHashMap<String, AggregationResult>,
     req_metrics: &VecWithNames<MetricAggregation>,
 ) -> crate::Result<()> {
     results.extend(req_metrics.iter().map(|(key, req)| {
@@ -157,7 +156,7 @@ fn add_empty_final_metrics_to_result(
 }
 
 fn add_empty_final_buckets_to_result(
-    results: &mut HashMap<String, AggregationResult>,
+    results: &mut FxHashMap<String, AggregationResult>,
     req_buckets: &VecWithNames<BucketAggregationInternal>,
 ) -> crate::Result<()> {
     let requested_buckets = req_buckets.iter();
@@ -169,7 +168,7 @@ fn add_empty_final_buckets_to_result(
 }
 
 fn convert_and_add_final_buckets_to_result(
-    results: &mut HashMap<String, AggregationResult>,
+    results: &mut FxHashMap<String, AggregationResult>,
     buckets: VecWithNames<IntermediateBucketResult>,
     req_buckets: &VecWithNames<BucketAggregationInternal>,
 ) -> crate::Result<()> {

src/collector/facet_collector.rs

@@ -616,7 +616,7 @@ mod tests {
             .map(|mut doc| {
                 doc.add_facet(
                     facet_field,
-                    &format!("/facet/{}", thread_rng().sample(&uniform)),
+                    &format!("/facet/{}", thread_rng().sample(uniform)),
                 );
                 doc
             })

src/collector/mod.rs

@@ -172,17 +172,33 @@ pub trait Collector: Sync + Send {
     ) -> crate::Result<<Self::Child as SegmentCollector>::Fruit> {
         let mut segment_collector = self.for_segment(segment_ord as u32, reader)?;
 
-        if let Some(alive_bitset) = reader.alive_bitset() {
-            weight.for_each(reader, &mut |doc, score| {
-                if alive_bitset.is_alive(doc) {
+        match (reader.alive_bitset(), self.requires_scoring()) {
+            (Some(alive_bitset), true) => {
+                weight.for_each(reader, &mut |doc, score| {
+                    if alive_bitset.is_alive(doc) {
+                        segment_collector.collect(doc, score);
+                    }
+                })?;
+            }
+            (Some(alive_bitset), false) => {
+                weight.for_each_no_score(reader, &mut |doc| {
+                    if alive_bitset.is_alive(doc) {
+                        segment_collector.collect(doc, 0.0);
+                    }
+                })?;
+            }
+            (None, true) => {
+                weight.for_each(reader, &mut |doc, score| {
                     segment_collector.collect(doc, score);
-                }
-            })?;
-        } else {
-            weight.for_each(reader, &mut |doc, score| {
-                segment_collector.collect(doc, score);
-            })?;
+                })?;
+            }
+            (None, false) => {
+                weight.for_each_no_score(reader, &mut |doc| {
+                    segment_collector.collect(doc, 0.0);
+                })?;
+            }
         }
+
         Ok(segment_collector.harvest())
     }
 }

src/core/index_meta.rs

@@ -133,7 +133,7 @@ impl SegmentMeta {
     /// associated with a segment component.
     pub fn relative_path(&self, component: SegmentComponent) -> PathBuf {
         let mut path = self.id().uuid_string();
-        path.push_str(&*match component {
+        path.push_str(&match component {
             SegmentComponent::Postings => ".idx".to_string(),
             SegmentComponent::Positions => ".pos".to_string(),
             SegmentComponent::Terms => ".term".to_string(),

src/directory/directory.rs

@@ -55,7 +55,7 @@ impl<T: Send + Sync + 'static> From<Box<T>> for DirectoryLock {
 
 impl Drop for DirectoryLockGuard {
     fn drop(&mut self) {
-        if let Err(e) = self.directory.delete(&*self.path) {
+        if let Err(e) = self.directory.delete(&self.path) {
             error!("Failed to remove the lock file. {:?}", e);
         }
     }

src/fastfield/bytes/reader.rs

@@ -1,10 +1,9 @@
-use std::ops::Range;
 use std::sync::Arc;
 
 use fastfield_codecs::Column;
 
 use crate::directory::{FileSlice, OwnedBytes};
-use crate::fastfield::MultiValueLength;
+use crate::fastfield::MultiValueIndex;
 use crate::DocId;
 
 /// Reader for byte array fast fields
@@ -19,7 +18,7 @@ use crate::DocId;
 /// and the start index for the next document, and keeping the bytes in between.
 #[derive(Clone)]
 pub struct BytesFastFieldReader {
-    idx_reader: Arc<dyn Column<u64>>,
+    idx_reader: MultiValueIndex,
     values: OwnedBytes,
 }
 
@@ -29,24 +28,26 @@ impl BytesFastFieldReader {
         values_file: FileSlice,
     ) -> crate::Result<BytesFastFieldReader> {
         let values = values_file.read_bytes()?;
-        Ok(BytesFastFieldReader { idx_reader, values })
+        Ok(BytesFastFieldReader {
+            idx_reader: MultiValueIndex::new(idx_reader),
+            values,
+        })
     }
 
-    fn range(&self, doc: DocId) -> Range<u32> {
-        let start = self.idx_reader.get_val(doc) as u32;
-        let end = self.idx_reader.get_val(doc + 1) as u32;
-        start..end
+    /// returns the multivalue index
+    pub fn get_index_reader(&self) -> &MultiValueIndex {
+        &self.idx_reader
     }
 
     /// Returns the bytes associated with the given `doc`
     pub fn get_bytes(&self, doc: DocId) -> &[u8] {
-        let range = self.range(doc);
+        let range = self.idx_reader.range(doc);
         &self.values.as_slice()[range.start as usize..range.end as usize]
     }
 
     /// Returns the length of the bytes associated with the given `doc`
     pub fn num_bytes(&self, doc: DocId) -> u64 {
-        let range = self.range(doc);
+        let range = self.idx_reader.range(doc);
         (range.end - range.start) as u64
     }
 
@@ -55,15 +56,3 @@ impl BytesFastFieldReader {
         self.values.len() as u64
     }
 }
-
-impl MultiValueLength for BytesFastFieldReader {
-    fn get_range(&self, doc_id: DocId) -> std::ops::Range<u32> {
-        self.range(doc_id)
-    }
-    fn get_len(&self, doc_id: DocId) -> u64 {
-        self.num_bytes(doc_id)
-    }
-    fn get_total_len(&self) -> u64 {
-        self.total_num_bytes()
-    }
-}

src/fastfield/mod.rs

@@ -27,8 +27,8 @@ pub use self::error::{FastFieldNotAvailableError, Result};
 pub use self::facet_reader::FacetReader;
 pub(crate) use self::multivalued::{get_fastfield_codecs_for_multivalue, MultivalueStartIndex};
 pub use self::multivalued::{
-    MultiValueU128FastFieldWriter, MultiValuedFastFieldReader, MultiValuedFastFieldWriter,
-    MultiValuedU128FastFieldReader,
+    MultiValueIndex, MultiValueU128FastFieldWriter, MultiValuedFastFieldReader,
+    MultiValuedFastFieldWriter, MultiValuedU128FastFieldReader,
 };
 pub use self::readers::FastFieldReaders;
 pub(crate) use self::readers::{type_and_cardinality, FastType};
@@ -36,7 +36,7 @@ pub use self::serializer::{Column, CompositeFastFieldSerializer};
 use self::writer::unexpected_value;
 pub use self::writer::{FastFieldsWriter, IntFastFieldWriter};
 use crate::schema::{Type, Value};
-use crate::{DateTime, DocId};
+use crate::DateTime;
 
 mod alive_bitset;
 mod bytes;
@@ -47,17 +47,6 @@ mod readers;
 mod serializer;
 mod writer;
 
-/// Trait for `BytesFastFieldReader` and `MultiValuedFastFieldReader` to return the length of data
-/// for a doc_id
-pub trait MultiValueLength {
-    /// returns the positions for a docid
-    fn get_range(&self, doc_id: DocId) -> std::ops::Range<u32>;
-    /// returns the num of values associated with a doc_id
-    fn get_len(&self, doc_id: DocId) -> u64;
-    /// returns the sum of num values for all doc_ids
-    fn get_total_len(&self) -> u64;
-}
-
 /// Trait for types that are allowed for fast fields:
 /// (u64, i64 and f64, bool, DateTime).
 pub trait FastValue:

src/fastfield/multivalued/index.rs

@@ -0,0 +1,140 @@
+use std::ops::Range;
+use std::sync::Arc;
+
+use fastfield_codecs::Column;
+
+use crate::DocId;
+
+#[derive(Clone)]
+/// Index to resolve value range for given doc_id.
+/// Starts at 0.
+pub struct MultiValueIndex {
+    idx: Arc<dyn Column<u64>>,
+}
+
+impl MultiValueIndex {
+    pub(crate) fn new(idx: Arc<dyn Column<u64>>) -> Self {
+        Self { idx }
+    }
+
+    /// Returns `[start, end)`, such that the values associated with
+    /// the given document are `start..end`.
+    #[inline]
+    pub(crate) fn range(&self, doc: DocId) -> Range<u32> {
+        let start = self.idx.get_val(doc) as u32;
+        let end = self.idx.get_val(doc + 1) as u32;
+        start..end
+    }
+
+    /// Returns `[start, end)`, such that the values associated with
+    /// the given documents are `start..end`.
+    ///
+    /// The passed end range is allowed to be out of bounds.
+    #[inline]
+    pub(crate) fn docid_range_to_position_range(&self, range: Range<DocId>) -> Range<u32> {
+        let end_docid = range.end.min(self.num_docs() - 1) + 1;
+        let start_docid = range.start.min(end_docid);
+
+        let start = self.idx.get_val(start_docid) as u32;
+        let end = self.idx.get_val(end_docid) as u32;
+        assert!(start <= end);
+
+        start..end
+    }
+
+    /// returns the num of values associated with a doc_id
+    pub(crate) fn num_vals_for_doc(&self, doc: DocId) -> u32 {
+        let range = self.range(doc);
+        range.end - range.start
+    }
+
+    /// Returns the overall number of values in this field.
+    #[inline]
+    pub fn total_num_vals(&self) -> u64 {
+        self.idx.max_value()
+    }
+
+    /// Returns the number of documents in the index.
+    #[inline]
+    pub fn num_docs(&self) -> u32 {
+        self.idx.num_vals() - 1
+    }
+
+    /// Converts a list of positions 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 positions_to_docids(&self, doc_id_range: Range<u32>, positions: &mut Vec<u32>) {
+        if positions.is_empty() {
+            return;
+        }
+        let mut cur_doc = doc_id_range.start;
+        let mut last_doc = None;
+
+        assert!(self.idx.get_val(doc_id_range.start) as u32 <= positions[0]);
+
+        let mut write_doc_pos = 0;
+        for i in 0..positions.len() {
+            let pos = positions[i];
+            loop {
+                let end = self.idx.get_val(cur_doc + 1) as u32;
+                if end > pos {
+                    positions[write_doc_pos] = cur_doc;
+                    write_doc_pos += if last_doc == Some(cur_doc) { 0 } else { 1 };
+                    last_doc = Some(cur_doc);
+                    break;
+                }
+                cur_doc += 1;
+            }
+        }
+        positions.truncate(write_doc_pos);
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use std::ops::Range;
+    use std::sync::Arc;
+
+    use fastfield_codecs::IterColumn;
+
+    use crate::fastfield::MultiValueIndex;
+
+    fn index_to_pos_helper(
+        index: &MultiValueIndex,
+        doc_id_range: Range<u32>,
+        positions: &[u32],
+    ) -> Vec<u32> {
+        let mut positions = positions.to_vec();
+        index.positions_to_docids(doc_id_range, &mut positions);
+        positions
+    }
+
+    #[test]
+    fn test_positions_to_docid() {
+        let offsets = vec![0, 10, 12, 15, 22, 23]; // docid values are [0..10, 10..12, 12..15, etc.]
+        let column = IterColumn::from(offsets.into_iter());
+        let index = MultiValueIndex::new(Arc::new(column));
+        assert_eq!(index.num_docs(), 5);
+        {
+            let positions = vec![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]);
+            assert_eq!(index_to_pos_helper(&index, 2..5, &[12]), vec![2]);
+            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]);
+        }
+    }
+}

src/fastfield/multivalued/mod.rs

@@ -1,7 +1,9 @@
+mod index;
 mod reader;
 mod writer;
 
 use fastfield_codecs::FastFieldCodecType;
+pub use index::MultiValueIndex;
 
 pub use self::reader::{MultiValuedFastFieldReader, MultiValuedU128FastFieldReader};
 pub(crate) use self::writer::MultivalueStartIndex;

src/fastfield/multivalued/reader.rs

@@ -3,7 +3,8 @@ use std::sync::Arc;
 
 use fastfield_codecs::{Column, MonotonicallyMappableToU128};
 
-use crate::fastfield::{FastValue, MultiValueLength};
+use super::MultiValueIndex;
+use crate::fastfield::FastValue;
 use crate::DocId;
 
 /// Reader for a multivalued `u64` fast field.
@@ -13,9 +14,10 @@ use crate::DocId;
 /// The `vals_reader` will access the concatenated list of all
 /// values for all reader.
 /// The `idx_reader` associated, for each document, the index of its first value.
+/// Stores the start position for each document.
 #[derive(Clone)]
 pub struct MultiValuedFastFieldReader<Item: FastValue> {
-    idx_reader: Arc<dyn Column<u64>>,
+    idx_reader: MultiValueIndex,
     vals_reader: Arc<dyn Column<Item>>,
 }
 
@@ -25,20 +27,11 @@ impl<Item: FastValue> MultiValuedFastFieldReader<Item> {
         vals_reader: Arc<dyn Column<Item>>,
     ) -> MultiValuedFastFieldReader<Item> {
         MultiValuedFastFieldReader {
-            idx_reader,
+            idx_reader: MultiValueIndex::new(idx_reader),
             vals_reader,
         }
     }
 
-    /// Returns `[start, end)`, such that the values associated with
-    /// the given document are `start..end`.
-    #[inline]
-    fn range(&self, doc: DocId) -> Range<u32> {
-        let start = self.idx_reader.get_val(doc) as u32;
-        let end = self.idx_reader.get_val(doc + 1) as u32;
-        start..end
-    }
-
     /// Returns the array of values associated with the given `doc`.
     #[inline]
     fn get_vals_for_range(&self, range: Range<u32>, vals: &mut Vec<Item>) {
@@ -51,10 +44,15 @@ impl<Item: FastValue> MultiValuedFastFieldReader<Item> {
     /// Returns the array of values associated with the given `doc`.
     #[inline]
     pub fn get_vals(&self, doc: DocId, vals: &mut Vec<Item>) {
-        let range = self.range(doc);
+        let range = self.idx_reader.range(doc);
         self.get_vals_for_range(range, vals);
     }
 
+    /// returns the multivalue index
+    pub fn get_index_reader(&self) -> &MultiValueIndex {
+        &self.idx_reader
+    }
+
     /// Returns the minimum value for this fast field.
     ///
     /// The min value does not take in account of possible
@@ -75,28 +73,14 @@ impl<Item: FastValue> MultiValuedFastFieldReader<Item> {
 
     /// Returns the number of values associated with the document `DocId`.
     #[inline]
-    pub fn num_vals(&self, doc: DocId) -> usize {
-        let range = self.range(doc);
-        (range.end - range.start) as usize
+    pub fn num_vals(&self, doc: DocId) -> u32 {
+        self.idx_reader.num_vals_for_doc(doc)
     }
 
-    /// Returns the overall number of values in this field  .
+    /// Returns the overall number of values in this field.
     #[inline]
     pub fn total_num_vals(&self) -> u64 {
-        self.idx_reader.max_value()
-    }
-}
-
-impl<Item: FastValue> MultiValueLength for MultiValuedFastFieldReader<Item> {
-    fn get_range(&self, doc_id: DocId) -> Range<u32> {
-        self.range(doc_id)
-    }
-    fn get_len(&self, doc_id: DocId) -> u64 {
-        self.num_vals(doc_id) as u64
-    }
-
-    fn get_total_len(&self) -> u64 {
-        self.total_num_vals() as u64
+        self.idx_reader.total_num_vals()
     }
 }
 
@@ -109,7 +93,7 @@ impl<Item: FastValue> MultiValueLength for MultiValuedFastFieldReader<Item> {
 /// The `idx_reader` associated, for each document, the index of its first value.
 #[derive(Clone)]
 pub struct MultiValuedU128FastFieldReader<T: MonotonicallyMappableToU128> {
-    idx_reader: Arc<dyn Column<u64>>,
+    idx_reader: MultiValueIndex,
     vals_reader: Arc<dyn Column<T>>,
 }
 
@@ -119,24 +103,31 @@ impl<T: MonotonicallyMappableToU128> MultiValuedU128FastFieldReader<T> {
         vals_reader: Arc<dyn Column<T>>,
     ) -> MultiValuedU128FastFieldReader<T> {
         Self {
-            idx_reader,
+            idx_reader: MultiValueIndex::new(idx_reader),
             vals_reader,
         }
     }
 
-    /// Returns `[start, end)`, such that the values associated
-    /// to the given document are `start..end`.
     #[inline]
-    fn range(&self, doc: DocId) -> Range<u32> {
-        let start = self.idx_reader.get_val(doc) as u32;
-        let end = self.idx_reader.get_val(doc + 1) as u32;
-        start..end
+    fn get_docids_for_value_range(
+        &self,
+        value_range: RangeInclusive<T>,
+        doc_id_range: Range<u32>,
+        positions: &mut Vec<u32>,
+    ) {
+        let position_range = self
+            .get_index_reader()
+            .docid_range_to_position_range(doc_id_range.clone());
+        self.vals_reader
+            .get_docids_for_value_range(value_range, position_range, positions);
+
+        self.idx_reader.positions_to_docids(doc_id_range, positions);
     }
 
     /// Returns the array of values associated to the given `doc`.
     #[inline]
     pub fn get_first_val(&self, doc: DocId) -> Option<T> {
-        let range = self.range(doc);
+        let range = self.idx_reader.range(doc);
         if range.is_empty() {
             return None;
         }
@@ -152,26 +143,18 @@ impl<T: MonotonicallyMappableToU128> MultiValuedU128FastFieldReader<T> {
             .get_range(range.start as u64, &mut vals[..]);
     }
 
+    /// Returns the index reader
+    pub fn get_index_reader(&self) -> &MultiValueIndex {
+        &self.idx_reader
+    }
+
     /// Returns the array of values associated to the given `doc`.
     #[inline]
     pub fn get_vals(&self, doc: DocId, vals: &mut Vec<T>) {
-        let range = self.range(doc);
+        let range = self.idx_reader.range(doc);
         self.get_vals_for_range(range, vals);
     }
 
-    /// Returns all docids which are in the provided value range
-    pub fn get_positions_for_value_range(
-        &self,
-        value_range: RangeInclusive<T>,
-        doc_id_range: Range<u32>,
-    ) -> Vec<DocId> {
-        let mut positions = Vec::new(); // TODO replace
-        self.vals_reader
-            .get_positions_for_value_range(value_range, doc_id_range, &mut positions);
-
-        positions_to_docids(&positions, self.idx_reader.as_ref())
-    }
-
     /// Iterates over all elements in the fast field
     pub fn iter(&self) -> impl Iterator<Item = T> + '_ {
         self.vals_reader.iter()
@@ -197,85 +180,27 @@ impl<T: MonotonicallyMappableToU128> MultiValuedU128FastFieldReader<T> {
 
     /// Returns the number of values associated with the document `DocId`.
     #[inline]
-    pub fn num_vals(&self, doc: DocId) -> usize {
-        let range = self.range(doc);
-        (range.end - range.start) as usize
+    pub fn num_vals(&self, doc: DocId) -> u32 {
+        self.idx_reader.num_vals_for_doc(doc)
     }
 
-    /// Returns the overall number of values in this field.
+    /// Returns the overall number of values in this field. It does not include deletes.
     #[inline]
     pub fn total_num_vals(&self) -> u64 {
-        self.idx_reader.max_value()
+        assert_eq!(
+            self.vals_reader.num_vals() as u64,
+            self.get_index_reader().total_num_vals()
+        );
+        self.idx_reader.total_num_vals()
     }
 }
 
-impl<T: MonotonicallyMappableToU128> MultiValueLength for MultiValuedU128FastFieldReader<T> {
-    fn get_range(&self, doc_id: DocId) -> std::ops::Range<u32> {
-        self.range(doc_id)
-    }
-    fn get_len(&self, doc_id: DocId) -> u64 {
-        self.num_vals(doc_id) as u64
-    }
-    fn get_total_len(&self) -> u64 {
-        self.total_num_vals() as u64
-    }
-}
-
-/// Converts a list of positions of values in a 1:n index to the corresponding list of 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 expotential search into binary search to match a
-/// docid to its value position.
-fn positions_to_docids<C: Column + ?Sized>(positions: &[u32], idx_reader: &C) -> Vec<DocId> {
-    let mut docs = vec![];
-    let mut cur_doc = 0u32;
-    let mut last_doc = None;
-
-    for pos in positions {
-        loop {
-            let end = idx_reader.get_val(cur_doc + 1) as u32;
-            if end > *pos {
-                // avoid duplicates
-                if Some(cur_doc) == last_doc {
-                    break;
-                }
-                docs.push(cur_doc);
-                last_doc = Some(cur_doc);
-                break;
-            }
-            cur_doc += 1;
-        }
-    }
-
-    docs
-}
-
 #[cfg(test)]
 mod tests {
 
-    use fastfield_codecs::VecColumn;
-
     use crate::core::Index;
-    use crate::fastfield::multivalued::reader::positions_to_docids;
     use crate::schema::{Cardinality, Facet, FacetOptions, NumericOptions, Schema};
 
-    #[test]
-    fn test_positions_to_docid() {
-        let positions = vec![10u32, 11, 15, 20, 21, 22];
-
-        let offsets = vec![0, 10, 12, 15, 22, 23];
-        {
-            let column = VecColumn::from(&offsets);
-
-            let docids = positions_to_docids(&positions, &column);
-            assert_eq!(docids, vec![1, 3, 4]);
-        }
-    }
-
     #[test]
     fn test_multifastfield_reader() -> crate::Result<()> {
         let mut schema_builder = Schema::builder();

src/indexer/index_writer.rs

@@ -95,7 +95,7 @@ fn compute_deleted_bitset(
         // document that were inserted before it.
         delete_op
             .target
-            .for_each(segment_reader, &mut |doc_matching_delete_query, _| {
+            .for_each_no_score(segment_reader, &mut |doc_matching_delete_query| {
                 if doc_opstamps.is_deleted(doc_matching_delete_query, delete_op.opstamp) {
                     alive_bitset.remove(doc_matching_delete_query);
                     might_have_changed = true;
@@ -805,7 +805,7 @@ mod tests {
     use std::collections::{HashMap, HashSet};
     use std::net::Ipv6Addr;
 
-    use fastfield_codecs::MonotonicallyMappableToU128;
+    use fastfield_codecs::{Column, MonotonicallyMappableToU128};
     use proptest::prelude::*;
     use proptest::prop_oneof;
     use proptest::strategy::Strategy;
@@ -1591,6 +1591,25 @@ mod tests {
         (existing_ids, deleted_ids)
     }
 
+    fn get_id_list(ops: &[IndexingOp]) -> Vec<u64> {
+        let mut id_list = Vec::new();
+        for &op in ops {
+            match op {
+                IndexingOp::AddDoc { id } => {
+                    id_list.push(id);
+                }
+                IndexingOp::DeleteDoc { id } => {
+                    id_list.retain(|el| *el != id);
+                }
+                IndexingOp::DeleteDocQuery { id } => {
+                    id_list.retain(|el| *el != id);
+                }
+                _ => {}
+            }
+        }
+        id_list
+    }
+
     fn test_operation_strategy(
         ops: &[IndexingOp],
         sort_index: bool,
@@ -1600,7 +1619,9 @@ mod tests {
         let ip_field = schema_builder.add_ip_addr_field("ip", FAST | INDEXED | STORED);
         let ips_field = schema_builder.add_ip_addr_field(
             "ips",
-            IpAddrOptions::default().set_fast(Cardinality::MultiValues),
+            IpAddrOptions::default()
+                .set_fast(Cardinality::MultiValues)
+                .set_indexed(),
         );
         let id_field = schema_builder.add_u64_field("id", FAST | INDEXED | STORED);
         let i64_field = schema_builder.add_i64_field("i64", INDEXED);
@@ -1665,11 +1686,13 @@ mod tests {
         // rotate right
         let multi_text_field_text3 = "test3 test1 test2 test3 test1 test2";
 
+        let ip_from_id = |id| Ipv6Addr::from_u128(id as u128);
+
         for &op in ops {
             match op {
                 IndexingOp::AddDoc { id } => {
                     let facet = Facet::from(&("/cola/".to_string() + &id.to_string()));
-                    let ip_from_id = Ipv6Addr::from_u128(id as u128);
+                    let ip = ip_from_id(id);
 
                     if !ip_exists(id) {
                         // every 3rd doc has no ip field
@@ -1693,9 +1716,9 @@ mod tests {
                     } else {
                         index_writer.add_document(doc!(id_field=>id,
                                 bytes_field => id.to_le_bytes().as_slice(),
-                                ip_field => ip_from_id,
-                                ips_field => ip_from_id,
-                                ips_field => ip_from_id,
+                                ip_field => ip,
+                                ips_field => ip,
+                                ips_field => ip,
                                 multi_numbers=> id,
                                 multi_numbers => id,
                                 bool_field => (id % 2u64) != 0,
@@ -1738,6 +1761,7 @@ mod tests {
         index_writer.commit()?;
 
         let searcher = index.reader()?.searcher();
+        let num_segments_before_merge = searcher.segment_readers().len();
         if force_end_merge {
             index_writer.wait_merging_threads()?;
             let mut index_writer = index.writer_for_tests()?;
@@ -1749,6 +1773,7 @@ mod tests {
                 assert!(index_writer.wait_merging_threads().is_ok());
             }
         }
+        let num_segments_after_merge = searcher.segment_readers().len();
 
         old_reader.reload()?;
         let old_searcher = old_reader.searcher();
@@ -1776,6 +1801,22 @@ mod tests {
             .collect();
 
         let (expected_ids_and_num_occurrences, deleted_ids) = expected_ids(ops);
+
+        let id_list = get_id_list(ops);
+
+        // multivalue fast field content
+        let mut all_ips = Vec::new();
+        let mut num_ips = 0;
+        for segment_reader in searcher.segment_readers().iter() {
+            let ip_reader = segment_reader.fast_fields().ip_addrs(ips_field).unwrap();
+            for doc in segment_reader.doc_ids_alive() {
+                let mut vals = vec![];
+                ip_reader.get_vals(doc, &mut vals);
+                all_ips.extend_from_slice(&vals);
+            }
+            num_ips += ip_reader.total_num_vals();
+        }
+
         let num_docs_expected = expected_ids_and_num_occurrences
             .iter()
             .map(|(_, id_occurrences)| *id_occurrences as usize)
@@ -1797,6 +1838,30 @@ mod tests {
                 .collect::<HashSet<_>>()
         );
 
+        if force_end_merge && num_segments_before_merge > 1 && num_segments_after_merge == 1 {
+            let mut expected_multi_ips: Vec<_> = id_list
+                .iter()
+                .filter(|id| ip_exists(**id))
+                .flat_map(|id| vec![ip_from_id(*id), ip_from_id(*id)])
+                .collect();
+            assert_eq!(num_ips, expected_multi_ips.len() as u64);
+
+            expected_multi_ips.sort();
+            all_ips.sort();
+            assert_eq!(expected_multi_ips, all_ips);
+
+            // Test fastfield num_docs
+            let num_docs: usize = searcher
+                .segment_readers()
+                .iter()
+                .map(|segment_reader| {
+                    let ff_reader = segment_reader.fast_fields().ip_addrs(ips_field).unwrap();
+                    ff_reader.num_docs() as usize
+                })
+                .sum();
+            assert_eq!(num_docs, num_docs_expected);
+        }
+
         // Load all ips addr
         let ips: HashSet<Ipv6Addr> = searcher
             .segment_readers()
@@ -2000,6 +2065,51 @@ mod tests {
                 assert_eq!(do_search_ip_field(&format!("\"{}\"", ip_addr)), count);
             }
         }
+
+        // assert data is like expected
+        //
+        for (existing_id, count) in expected_ids_and_num_occurrences.iter().take(10) {
+            let (existing_id, count) = (*existing_id, *count);
+            if !ip_exists(existing_id) {
+                continue;
+            }
+            let gen_query_inclusive = |field: &str, from: Ipv6Addr, to: Ipv6Addr| {
+                format!("{}:[{} TO {}]", field, &from.to_string(), &to.to_string())
+            };
+            let ip = ip_from_id(existing_id);
+
+            let do_search_ip_field = |term: &str| do_search(term, ip_field).len() as u64;
+            // Range query on single value field
+            // let query = gen_query_inclusive("ip", ip, ip);
+            // assert_eq!(do_search_ip_field(&query), count);
+
+            // Range query on multi value field
+            let query = gen_query_inclusive("ips", ip, ip);
+            assert_eq!(do_search_ip_field(&query), count);
+        }
+
+        // ip range query on fast field
+        //
+        for (existing_id, count) in expected_ids_and_num_occurrences.iter().take(10) {
+            let (existing_id, count) = (*existing_id, *count);
+            if !ip_exists(existing_id) {
+                continue;
+            }
+            let gen_query_inclusive = |field: &str, from: Ipv6Addr, to: Ipv6Addr| {
+                format!("{}:[{} TO {}]", field, &from.to_string(), &to.to_string())
+            };
+            let ip = ip_from_id(existing_id);
+
+            let do_search_ip_field = |term: &str| do_search(term, ip_field).len() as u64;
+            // Range query on single value field
+            // let query = gen_query_inclusive("ip", ip, ip);
+            // assert_eq!(do_search_ip_field(&query), count);
+
+            // Range query on multi value field
+            let query = gen_query_inclusive("ips", ip, ip);
+            assert_eq!(do_search_ip_field(&query), count);
+        }
+
         // test facets
         for segment_reader in searcher.segment_readers().iter() {
             let mut facet_reader = segment_reader.facet_reader(facet_field).unwrap();
@@ -2021,6 +2131,40 @@ mod tests {
         Ok(())
     }
 
+    #[test]
+    fn test_ip_range_query_multivalue_bug() {
+        assert!(test_operation_strategy(
+            &[
+                IndexingOp::AddDoc { id: 2 },
+                IndexingOp::Commit,
+                IndexingOp::AddDoc { id: 1 },
+                IndexingOp::AddDoc { id: 1 },
+                IndexingOp::Commit,
+                IndexingOp::Merge
+            ],
+            true,
+            false
+        )
+        .is_ok());
+    }
+
+    #[test]
+    fn test_ff_num_ips_regression() {
+        assert!(test_operation_strategy(
+            &[
+                IndexingOp::AddDoc { id: 13 },
+                IndexingOp::AddDoc { id: 1 },
+                IndexingOp::Commit,
+                IndexingOp::DeleteDocQuery { id: 13 },
+                IndexingOp::AddDoc { id: 1 },
+                IndexingOp::Commit,
+            ],
+            false,
+            true
+        )
+        .is_ok());
+    }
+
     #[test]
     fn test_minimal() {
         assert!(test_operation_strategy(
@@ -2030,7 +2174,7 @@ mod tests {
                 IndexingOp::DeleteDoc { id: 13 }
             ],
             true,
-            false
+            true
         )
         .is_ok());
 

src/indexer/merger.rs

@@ -13,7 +13,7 @@ use crate::docset::{DocSet, TERMINATED};
 use crate::error::DataCorruption;
 use crate::fastfield::{
     get_fastfield_codecs_for_multivalue, AliveBitSet, Column, CompositeFastFieldSerializer,
-    MultiValueLength, MultiValuedFastFieldReader, MultiValuedU128FastFieldReader,
+    MultiValueIndex, MultiValuedFastFieldReader, MultiValuedU128FastFieldReader,
 };
 use crate::fieldnorm::{FieldNormReader, FieldNormReaders, FieldNormsSerializer, FieldNormsWriter};
 use crate::indexer::doc_id_mapping::{expect_field_id_for_sort_field, SegmentDocIdMapping};
@@ -348,9 +348,29 @@ impl IndexMerger {
             field,
             fast_field_serializer,
             doc_id_mapping,
-            &segment_and_ff_readers,
+            &segment_and_ff_readers
+                .iter()
+                .map(|(segment_reader, u64s_reader)| {
+                    (*segment_reader, u64s_reader.get_index_reader())
+                })
+                .collect::<Vec<_>>(),
         )?;
 
+        let num_vals = segment_and_ff_readers
+            .iter()
+            .map(|(segment_reader, reader)| {
+                // TODO implement generic version, implement reverse scan, all - deletes
+                if let Some(alive_bitset) = segment_reader.alive_bitset() {
+                    alive_bitset
+                        .iter_alive()
+                        .map(|doc| reader.num_vals(doc))
+                        .sum()
+                } else {
+                    reader.total_num_vals() as u32
+                }
+            })
+            .sum();
+
         let fast_field_readers = segment_and_ff_readers
             .into_iter()
             .map(|(_, ff_reader)| ff_reader)
@@ -365,12 +385,7 @@ impl IndexMerger {
                 })
         };
 
-        fast_field_serializer.create_u128_fast_field_with_idx(
-            field,
-            iter_gen,
-            doc_id_mapping.len() as u32,
-            1,
-        )?;
+        fast_field_serializer.create_u128_fast_field_with_idx(field, iter_gen, num_vals, 1)?;
 
         Ok(())
     }
@@ -529,11 +544,11 @@ impl IndexMerger {
     // Creating the index file to point into the data, generic over `BytesFastFieldReader` and
     // `MultiValuedFastFieldReader`
     //
-    fn write_1_n_fast_field_idx_generic<T: MultiValueLength + Send + Sync>(
+    fn write_1_n_fast_field_idx_generic(
         field: Field,
         fast_field_serializer: &mut CompositeFastFieldSerializer,
         doc_id_mapping: &SegmentDocIdMapping,
-        segment_and_ff_readers: &[(&SegmentReader, T)],
+        segment_and_ff_readers: &[(&SegmentReader, &MultiValueIndex)],
     ) -> crate::Result<()> {
         let column =
             RemappedDocIdMultiValueIndexColumn::new(segment_and_ff_readers, doc_id_mapping);
@@ -567,7 +582,12 @@ impl IndexMerger {
             field,
             fast_field_serializer,
             doc_id_mapping,
-            &segment_and_ff_readers,
+            &segment_and_ff_readers
+                .iter()
+                .map(|(segment_reader, u64s_reader)| {
+                    (*segment_reader, u64s_reader.get_index_reader())
+                })
+                .collect::<Vec<_>>(),
         )
     }
 
@@ -697,7 +717,12 @@ impl IndexMerger {
             field,
             fast_field_serializer,
             doc_id_mapping,
-            &segment_and_ff_readers,
+            &segment_and_ff_readers
+                .iter()
+                .map(|(segment_reader, u64s_reader)| {
+                    (*segment_reader, u64s_reader.get_index_reader())
+                })
+                .collect::<Vec<_>>(),
         )?;
 
         let mut serialize_vals = fast_field_serializer.new_bytes_fast_field(field);
@@ -804,7 +829,7 @@ impl IndexMerger {
             // Let's compute the list of non-empty posting lists
             for (segment_ord, term_info) in merged_terms.current_segment_ords_and_term_infos() {
                 let segment_reader = &self.readers[segment_ord];
-                let inverted_index: &InvertedIndexReader = &*field_readers[segment_ord];
+                let inverted_index: &InvertedIndexReader = &field_readers[segment_ord];
                 let segment_postings = inverted_index
                     .read_postings_from_terminfo(&term_info, segment_postings_option)?;
                 let alive_bitset_opt = segment_reader.alive_bitset();

src/indexer/sorted_doc_id_multivalue_column.rs

@@ -3,7 +3,7 @@ use std::cmp;
 use fastfield_codecs::Column;
 
 use super::flat_map_with_buffer::FlatMapWithBufferIter;
-use crate::fastfield::{MultiValueLength, MultiValuedFastFieldReader};
+use crate::fastfield::{MultiValueIndex, MultiValuedFastFieldReader};
 use crate::indexer::doc_id_mapping::SegmentDocIdMapping;
 use crate::schema::Field;
 use crate::{DocAddress, SegmentReader};
@@ -94,17 +94,17 @@ impl<'a> Column for RemappedDocIdMultiValueColumn<'a> {
     }
 }
 
-pub(crate) struct RemappedDocIdMultiValueIndexColumn<'a, T: MultiValueLength> {
+pub(crate) struct RemappedDocIdMultiValueIndexColumn<'a> {
     doc_id_mapping: &'a SegmentDocIdMapping,
-    multi_value_length_readers: Vec<&'a T>,
+    multi_value_length_readers: Vec<&'a MultiValueIndex>,
     min_value: u64,
     max_value: u64,
     num_vals: u32,
 }
 
-impl<'a, T: MultiValueLength> RemappedDocIdMultiValueIndexColumn<'a, T> {
+impl<'a> RemappedDocIdMultiValueIndexColumn<'a> {
     pub(crate) fn new(
-        segment_and_ff_readers: &'a [(&'a SegmentReader, T)],
+        segment_and_ff_readers: &'a [(&'a SegmentReader, &'a MultiValueIndex)],
         doc_id_mapping: &'a SegmentDocIdMapping,
     ) -> Self {
         // We go through a complete first pass to compute the minimum and the
@@ -115,17 +115,19 @@ impl<'a, T: MultiValueLength> RemappedDocIdMultiValueIndexColumn<'a, T> {
         let mut multi_value_length_readers = Vec::with_capacity(segment_and_ff_readers.len());
         for segment_and_ff_reader in segment_and_ff_readers {
             let segment_reader = segment_and_ff_reader.0;
-            let multi_value_length_reader = &segment_and_ff_reader.1;
+            let multi_value_length_reader = segment_and_ff_reader.1;
             if !segment_reader.has_deletes() {
-                max_value += multi_value_length_reader.get_total_len();
+                max_value += multi_value_length_reader.total_num_vals();
             } else {
                 for doc in segment_reader.doc_ids_alive() {
-                    max_value += multi_value_length_reader.get_len(doc);
+                    max_value += multi_value_length_reader.num_vals_for_doc(doc) as u64;
                 }
             }
             num_vals += segment_reader.num_docs();
             multi_value_length_readers.push(multi_value_length_reader);
         }
+        // The value range is always get_val(doc)..get_val(doc + 1)
+        num_vals += 1;
         Self {
             doc_id_mapping,
             multi_value_length_readers,
@@ -136,7 +138,7 @@ impl<'a, T: MultiValueLength> RemappedDocIdMultiValueIndexColumn<'a, T> {
     }
 }
 
-impl<'a, T: MultiValueLength + Send + Sync> Column for RemappedDocIdMultiValueIndexColumn<'a, T> {
+impl<'a> Column for RemappedDocIdMultiValueIndexColumn<'a> {
     fn get_val(&self, _pos: u32) -> u64 {
         unimplemented!()
     }
@@ -148,8 +150,8 @@ impl<'a, T: MultiValueLength + Send + Sync> Column for RemappedDocIdMultiValueIn
                 move |old_doc_addr| {
                     let ff_reader =
                         &self.multi_value_length_readers[old_doc_addr.segment_ord as usize];
-                    offset += ff_reader.get_len(old_doc_addr.doc_id);
-                    offset
+                    offset += ff_reader.num_vals_for_doc(old_doc_addr.doc_id);
+                    offset as u64
                 },
             )),
         )

src/query/automaton_weight.rs

@@ -33,7 +33,7 @@ where
         &'a self,
         term_dict: &'a TermDictionary,
     ) -> io::Result<TermStreamer<'a, &'a A>> {
-        let automaton: &A = &*self.automaton;
+        let automaton: &A = &self.automaton;
         let term_stream_builder = term_dict.search(automaton);
         term_stream_builder.into_stream()
     }

src/query/boolean_query/boolean_weight.rs

@@ -5,7 +5,7 @@ use crate::postings::FreqReadingOption;
 use crate::query::explanation::does_not_match;
 use crate::query::score_combiner::{DoNothingCombiner, ScoreCombiner};
 use crate::query::term_query::TermScorer;
-use crate::query::weight::{for_each_pruning_scorer, for_each_scorer};
+use crate::query::weight::{for_each_docset, for_each_pruning_scorer, for_each_scorer};
 use crate::query::{
     intersect_scorers, EmptyScorer, Exclude, Explanation, Occur, RequiredOptionalScorer, Scorer,
     Union, Weight,
@@ -219,6 +219,24 @@ impl<TScoreCombiner: ScoreCombiner + Sync> Weight for BooleanWeight<TScoreCombin
         Ok(())
     }
 
+    fn for_each_no_score(
+        &self,
+        reader: &SegmentReader,
+        callback: &mut dyn FnMut(DocId),
+    ) -> crate::Result<()> {
+        let scorer = self.complex_scorer(reader, 1.0, || DoNothingCombiner)?;
+        match scorer {
+            SpecializedScorer::TermUnion(term_scorers) => {
+                let mut union_scorer = Union::build(term_scorers, &self.score_combiner_fn);
+                for_each_docset(&mut union_scorer, callback);
+            }
+            SpecializedScorer::Other(mut scorer) => {
+                for_each_docset(scorer.as_mut(), callback);
+            }
+        }
+        Ok(())
+    }
+
     /// Calls `callback` with all of the `(doc, score)` for which score
     /// is exceeding a given threshold.
     ///

src/query/range_query_ip_fastfield.rs

@@ -11,6 +11,7 @@ use fastfield_codecs::{Column, MonotonicallyMappableToU128};
 
 use super::range_query::map_bound;
 use super::{ConstScorer, Explanation, Scorer, Weight};
+use crate::fastfield::MultiValuedU128FastFieldReader;
 use crate::schema::{Cardinality, Field};
 use crate::{DocId, DocSet, Score, SegmentReader, TantivyError, TERMINATED};
 
@@ -43,16 +44,29 @@ impl Weight for IPFastFieldRangeWeight {
         let field_type = reader.schema().get_field_entry(self.field).field_type();
         match field_type.fastfield_cardinality().unwrap() {
             Cardinality::SingleValue => {
-                let ip_addr_fast_field = reader.fast_fields().ip_addr(self.field)?;
+                let ip_addr_fast_field: Arc<dyn Column<Ipv6Addr>> =
+                    reader.fast_fields().ip_addr(self.field)?;
                 let value_range = bound_to_value_range(
                     &self.left_bound,
                     &self.right_bound,
-                    ip_addr_fast_field.as_ref(),
+                    ip_addr_fast_field.min_value(),
+                    ip_addr_fast_field.max_value(),
                 );
                 let docset = IpRangeDocSet::new(value_range, ip_addr_fast_field);
                 Ok(Box::new(ConstScorer::new(docset, boost)))
             }
-            Cardinality::MultiValues => unimplemented!(),
+            Cardinality::MultiValues => {
+                let ip_addr_fast_field: MultiValuedU128FastFieldReader<Ipv6Addr> =
+                    reader.fast_fields().ip_addrs(self.field)?;
+                let value_range = bound_to_value_range(
+                    &self.left_bound,
+                    &self.right_bound,
+                    ip_addr_fast_field.min_value(),
+                    ip_addr_fast_field.max_value(),
+                );
+                let docset = IpRangeDocSet::new(value_range, Arc::new(ip_addr_fast_field));
+                Ok(Box::new(ConstScorer::new(docset, boost)))
+            }
         }
     }
 
@@ -73,18 +87,19 @@ impl Weight for IPFastFieldRangeWeight {
 fn bound_to_value_range(
     left_bound: &Bound<Ipv6Addr>,
     right_bound: &Bound<Ipv6Addr>,
-    column: &dyn Column<Ipv6Addr>,
+    min_value: Ipv6Addr,
+    max_value: Ipv6Addr,
 ) -> RangeInclusive<Ipv6Addr> {
     let start_value = match left_bound {
         Bound::Included(ip_addr) => *ip_addr,
         Bound::Excluded(ip_addr) => Ipv6Addr::from(ip_addr.to_u128() + 1),
-        Bound::Unbounded => column.min_value(),
+        Bound::Unbounded => min_value,
     };
 
     let end_value = match right_bound {
         Bound::Included(ip_addr) => *ip_addr,
         Bound::Excluded(ip_addr) => Ipv6Addr::from(ip_addr.to_u128() - 1),
-        Bound::Unbounded => column.max_value(),
+        Bound::Unbounded => max_value,
     };
     start_value..=end_value
 }
@@ -109,22 +124,23 @@ impl VecCursor {
     fn current(&self) -> Option<u32> {
         self.docs.get(self.current_pos).map(|el| *el as u32)
     }
-
     fn get_cleared_data(&mut self) -> &mut Vec<u32> {
         self.docs.clear();
         self.current_pos = 0;
         &mut self.docs
     }
-
+    fn last_value(&self) -> Option<u32> {
+        self.docs.iter().last().cloned()
+    }
     fn is_empty(&self) -> bool {
         self.current_pos >= self.docs.len()
     }
 }
 
-struct IpRangeDocSet {
+struct IpRangeDocSet<T> {
     /// The range filter on the values.
     value_range: RangeInclusive<Ipv6Addr>,
-    ip_addr_fast_field: Arc<dyn Column<Ipv6Addr>>,
+    ip_addrs: T,
     /// The next docid start range to fetch (inclusive).
     next_fetch_start: u32,
     /// Number of docs range checked in a batch.
@@ -141,18 +157,17 @@ struct IpRangeDocSet {
     last_seek_pos_opt: Option<u32>,
 }
 
-const DEFALT_FETCH_HORIZON: u32 = 128;
-impl IpRangeDocSet {
-    fn new(
-        value_range: RangeInclusive<Ipv6Addr>,
-        ip_addr_fast_field: Arc<dyn Column<Ipv6Addr>>,
-    ) -> Self {
+const DEFAULT_FETCH_HORIZON: u32 = 128;
+impl<T> IpRangeDocSet<T>
+where Self: SingleOrMultivalued
+{
+    fn new(value_range: RangeInclusive<Ipv6Addr>, ip_addrs: T) -> Self {
         let mut ip_range_docset = Self {
             value_range,
-            ip_addr_fast_field,
+            ip_addrs,
             loaded_docs: VecCursor::new(),
             next_fetch_start: 0,
-            fetch_horizon: DEFALT_FETCH_HORIZON,
+            fetch_horizon: DEFAULT_FETCH_HORIZON,
             last_seek_pos_opt: None,
         };
         ip_range_docset.reset_fetch_range();
@@ -161,7 +176,7 @@ impl IpRangeDocSet {
     }
 
     fn reset_fetch_range(&mut self) {
-        self.fetch_horizon = DEFALT_FETCH_HORIZON;
+        self.fetch_horizon = DEFAULT_FETCH_HORIZON;
     }
 
     /// Returns true if more data could be fetched
@@ -185,36 +200,72 @@ impl IpRangeDocSet {
             true
         }
     }
+}
 
-    /// Fetches a block for docid range [next_fetch_start .. next_fetch_start + HORIZON]
+trait SingleOrMultivalued {
+    fn num_docs(&self) -> u32;
     fn fetch_horizon(&mut self, horizon: u32) -> bool {
-        let mut finished_to_end = false;
+        // Have different implem for single value and multivalue
+        todo!();
+        // let mut finished_to_end = false;
 
-        let limit = self.ip_addr_fast_field.num_vals();
-        let mut end = self.next_fetch_start + horizon;
-        if end >= limit {
-            end = limit;
-            finished_to_end = true;
-        }
+        // let limit = self.num_docs();
+        // let mut end = self.next_fetch_start + horizon;
+        // if end >= limit {
+        //     end = limit;
+        //     finished_to_end = true;
+        // }
 
-        let data = self.loaded_docs.get_cleared_data();
-        self.ip_addr_fast_field.get_positions_for_value_range(
-            self.value_range.clone(),
-            self.next_fetch_start..end,
-            data,
-        );
-        self.next_fetch_start = end;
-        finished_to_end
+        // let last_loaded_docs_val = self
+        //     .is_multivalue
+        //     .then(|| self.loaded_docs.last_value())
+        //     .flatten();
+
+        // let last_loaded_docs_val =
+        //     if self.is_multivalue {
+        //         self.loaded_docs.last_value()
+        //     } else {
+        //         None
+        //     };
+
+        // let loaded_docs_data = self.loaded_docs.get_cleared_data();
+        // self.ip_addr_fast_field.get_docids_for_value_range(
+        //     self.value_range.clone(),
+        //     self.next_fetch_start..end,
+        //     loaded_docs_data,
+        // );
+        // // In case of multivalues, we may have an overlap of the same docid between fetching
+        // blocks if let Some(last_value) = last_loaded_docs_val {
+        //     while self.loaded_docs.current() == Some(last_value) {
+        //         self.loaded_docs.next();
+        //     }
+        // }
+        // self.next_fetch_start = end;
+        // finished_to_end
     }
 }
 
-impl DocSet for IpRangeDocSet {
+impl SingleOrMultivalued for IpRangeDocSet<Arc<dyn Column<Ipv6Addr>>> {
+    fn num_docs(&self) -> u32 {
+        self.ip_addrs.num_docs()
+    }
+}
+
+impl SingleOrMultivalued for IpRangeDocSet<Arc<MultiValuedU128FastFieldReader<Ipv6Addr>>> {
+    fn num_docs(&self) -> u32 {
+        self.ip_addrs.get_index_reader().num_docs()
+    }
+}
+
+impl<T: Send> DocSet for IpRangeDocSet<T>
+where Self: SingleOrMultivalued
+{
     #[inline]
     fn advance(&mut self) -> DocId {
         if let Some(docid) = self.loaded_docs.next() {
             docid as u32
         } else {
-            if self.next_fetch_start >= self.ip_addr_fast_field.num_vals() as u32 {
+            if self.next_fetch_start >= self.num_docs() as u32 {
                 return TERMINATED;
             }
             self.fetch_block();
@@ -269,7 +320,7 @@ mod tests {
     use super::*;
     use crate::collector::Count;
     use crate::query::QueryParser;
-    use crate::schema::{Schema, FAST, INDEXED, STORED, STRING};
+    use crate::schema::{IpAddrOptions, Schema, FAST, INDEXED, STORED, STRING};
     use crate::Index;
 
     #[derive(Clone, Debug)]
@@ -280,12 +331,13 @@ mod tests {
 
     fn operation_strategy() -> impl Strategy<Value = Doc> {
         prop_oneof![
-            (0u64..100u64).prop_map(doc_from_id_1),
-            (1u64..100u64).prop_map(doc_from_id_2),
+            (0u64..10_000u64).prop_map(doc_from_id_1),
+            (1u64..10_000u64).prop_map(doc_from_id_2),
         ]
     }
 
     pub fn doc_from_id_1(id: u64) -> Doc {
+        let id = id * 1000;
         Doc {
             // ip != id
             id: id.to_string(),
@@ -293,6 +345,7 @@ mod tests {
         }
     }
     fn doc_from_id_2(id: u64) -> Doc {
+        let id = id * 1000;
         Doc {
             // ip != id
             id: (id - 1).to_string(),
@@ -310,6 +363,12 @@ mod tests {
 
     #[test]
     fn ip_range_regression1_test() {
+        let ops = vec![doc_from_id_1(0)];
+        assert!(test_ip_range_for_docs(ops).is_ok());
+    }
+
+    #[test]
+    fn ip_range_regression2_test() {
         let ops = vec![
             doc_from_id_1(52),
             doc_from_id_1(63),
@@ -321,14 +380,20 @@ mod tests {
     }
 
     #[test]
-    fn ip_range_regression2_test() {
-        let ops = vec![doc_from_id_1(0)];
+    fn ip_range_regression3_test() {
+        let ops = vec![doc_from_id_1(1), doc_from_id_1(2), doc_from_id_1(3)];
         assert!(test_ip_range_for_docs(ops).is_ok());
     }
 
     pub fn create_index_from_docs(docs: &[Doc]) -> Index {
         let mut schema_builder = Schema::builder();
         let ip_field = schema_builder.add_ip_addr_field("ip", INDEXED | STORED | FAST);
+        let ips_field = schema_builder.add_ip_addr_field(
+            "ips",
+            IpAddrOptions::default()
+                .set_fast(Cardinality::MultiValues)
+                .set_indexed(),
+        );
         let text_field = schema_builder.add_text_field("id", STRING | STORED);
         let schema = schema_builder.build();
         let index = Index::create_in_ram(schema);
@@ -338,6 +403,8 @@ mod tests {
             for doc in docs.iter() {
                 index_writer
                     .add_document(doc!(
+                        ips_field => doc.ip,
+                        ips_field => doc.ip,
                         ip_field => doc.ip,
                         text_field => doc.id.to_string(),
                     ))
@@ -361,8 +428,8 @@ mod tests {
                 .unwrap()
         };
 
-        let gen_query_inclusive = |from: Ipv6Addr, to: Ipv6Addr| {
-            format!("ip:[{} TO {}]", &from.to_string(), &to.to_string())
+        let gen_query_inclusive = |field: &str, from: Ipv6Addr, to: Ipv6Addr| {
+            format!("{}:[{} TO {}]", field, &from.to_string(), &to.to_string())
         };
 
         let test_sample = |sample_docs: Vec<Doc>| {
@@ -373,7 +440,10 @@ mod tests {
                 .filter(|doc| (ips[0]..=ips[1]).contains(&doc.ip))
                 .count();
 
-            let query = gen_query_inclusive(ips[0], ips[1]);
+            let query = gen_query_inclusive("ip", ips[0], ips[1]);
+            assert_eq!(get_num_hits(query_from_text(&query)), expected_num_hits);
+
+            let query = gen_query_inclusive("ips", ips[0], ips[1]);
             assert_eq!(get_num_hits(query_from_text(&query)), expected_num_hits);
 
             // Intersection search
@@ -382,7 +452,20 @@ mod tests {
                 .iter()
                 .filter(|doc| (ips[0]..=ips[1]).contains(&doc.ip) && doc.id == id_filter)
                 .count();
-            let query = format!("{} AND id:{}", query, &id_filter);
+            let query = format!(
+                "{} AND id:{}",
+                gen_query_inclusive("ip", ips[0], ips[1]),
+                &id_filter
+            );
+            assert_eq!(get_num_hits(query_from_text(&query)), expected_num_hits);
+
+            // Intersection search on multivalue ip field
+            let id_filter = sample_docs[0].id.to_string();
+            let query = format!(
+                "{} AND id:{}",
+                gen_query_inclusive("ips", ips[0], ips[1]),
+                &id_filter
+            );
             assert_eq!(get_num_hits(query_from_text(&query)), expected_num_hits);
         };
 
@@ -402,7 +485,8 @@ mod tests {
 #[cfg(all(test, feature = "unstable"))]
 mod bench {
 
-    use rand::{thread_rng, Rng};
+    use rand::rngs::StdRng;
+    use rand::{Rng, SeedableRng};
     use test::Bencher;
 
     use super::tests::*;
@@ -412,7 +496,7 @@ mod bench {
     use crate::Index;
 
     fn get_index_0_to_100() -> Index {
-        let mut rng = thread_rng();
+        let mut rng = StdRng::from_seed([1u8; 32]);
         let num_vals = 100_000;
         let docs: Vec<_> = (0..num_vals)
             .map(|_i| {
@@ -424,8 +508,10 @@ mod bench {
                     "many".to_string() // 90%
                 };
                 Doc {
-                    id: id,
+                    id,
                     // Multiply by 1000, so that we create many buckets in the compact space
+                    // The benches depend on this range to select n-percent of elements with the
+                    // methods below.
                     ip: Ipv6Addr::from_u128(rng.gen_range(0..100) * 1000),
                 }
             })
@@ -434,22 +520,42 @@ mod bench {
         let index = create_index_from_docs(&docs);
         index
     }
+
+    fn get_90_percent() -> RangeInclusive<Ipv6Addr> {
+        let start = Ipv6Addr::from_u128(0);
+        let end = Ipv6Addr::from_u128(90 * 1000);
+        start..=end
+    }
+
+    fn get_10_percent() -> RangeInclusive<Ipv6Addr> {
+        let start = Ipv6Addr::from_u128(0);
+        let end = Ipv6Addr::from_u128(10 * 1000);
+        start..=end
+    }
+
+    fn get_1_percent() -> RangeInclusive<Ipv6Addr> {
+        let start = Ipv6Addr::from_u128(10 * 1000);
+        let end = Ipv6Addr::from_u128(10 * 1000);
+        start..=end
+    }
+
     fn excute_query(
-        start_inclusive: Ipv6Addr,
-        end_inclusive: Ipv6Addr,
+        field: &str,
+        ip_range: RangeInclusive<Ipv6Addr>,
         suffix: &str,
         index: &Index,
     ) -> usize {
-        let gen_query_inclusive = |from: Ipv6Addr, to: Ipv6Addr| {
+        let gen_query_inclusive = |from: &Ipv6Addr, to: &Ipv6Addr| {
             format!(
-                "ip:[{} TO {}] {}",
+                "{}:[{} TO {}] {}",
+                field,
                 &from.to_string(),
                 &to.to_string(),
                 suffix
             )
         };
 
-        let query = gen_query_inclusive(start_inclusive, end_inclusive);
+        let query = gen_query_inclusive(ip_range.start(), ip_range.end());
         let query_from_text = |text: &str| {
             QueryParser::for_index(&index, vec![])
                 .parse_query(text)
@@ -465,131 +571,153 @@ mod bench {
     fn bench_ip_range_hit_90_percent(bench: &mut Bencher) {
         let index = get_index_0_to_100();
 
-        bench.iter(|| {
-            let start = Ipv6Addr::from_u128(0);
-            let end = Ipv6Addr::from_u128(90 * 1000);
-
-            excute_query(start, end, "", &index)
-        });
+        bench.iter(|| excute_query("ip", get_90_percent(), "", &index));
     }
 
     #[bench]
     fn bench_ip_range_hit_10_percent(bench: &mut Bencher) {
         let index = get_index_0_to_100();
 
-        bench.iter(|| {
-            let start = Ipv6Addr::from_u128(0);
-            let end = Ipv6Addr::from_u128(10 * 1000);
-
-            excute_query(start, end, "", &index)
-        });
+        bench.iter(|| excute_query("ip", get_10_percent(), "", &index));
     }
 
     #[bench]
     fn bench_ip_range_hit_1_percent(bench: &mut Bencher) {
         let index = get_index_0_to_100();
 
-        bench.iter(|| {
-            let start = Ipv6Addr::from_u128(10 * 1000);
-            let end = Ipv6Addr::from_u128(10 * 1000);
-
-            excute_query(start, end, "", &index)
-        });
+        bench.iter(|| excute_query("ip", get_1_percent(), "", &index));
     }
 
     #[bench]
     fn bench_ip_range_hit_10_percent_intersect_with_10_percent(bench: &mut Bencher) {
         let index = get_index_0_to_100();
 
-        bench.iter(|| {
-            let start = Ipv6Addr::from_u128(0);
-            let end = Ipv6Addr::from_u128(10 * 1000);
-
-            excute_query(start, end, "AND id:few", &index)
-        });
+        bench.iter(|| excute_query("ip", get_10_percent(), "AND id:few", &index));
     }
 
     #[bench]
     fn bench_ip_range_hit_1_percent_intersect_with_10_percent(bench: &mut Bencher) {
         let index = get_index_0_to_100();
 
-        bench.iter(|| {
-            let start = Ipv6Addr::from_u128(10 * 1000);
-            let end = Ipv6Addr::from_u128(10 * 1000);
-
-            excute_query(start, end, "AND id:few", &index)
-        });
+        bench.iter(|| excute_query("ip", get_1_percent(), "AND id:few", &index));
     }
 
     #[bench]
     fn bench_ip_range_hit_1_percent_intersect_with_90_percent(bench: &mut Bencher) {
         let index = get_index_0_to_100();
 
-        bench.iter(|| {
-            let start = Ipv6Addr::from_u128(10 * 1000);
-            let end = Ipv6Addr::from_u128(10 * 1000);
-
-            excute_query(start, end, "AND id:many", &index)
-        });
+        bench.iter(|| excute_query("ip", get_1_percent(), "AND id:many", &index));
     }
 
     #[bench]
     fn bench_ip_range_hit_1_percent_intersect_with_1_percent(bench: &mut Bencher) {
         let index = get_index_0_to_100();
 
-        bench.iter(|| {
-            let start = Ipv6Addr::from_u128(10 * 1000);
-            let end = Ipv6Addr::from_u128(10 * 1000);
-
-            excute_query(start, end, "AND id:veryfew", &index)
-        });
+        bench.iter(|| excute_query("ip", get_1_percent(), "AND id:veryfew", &index));
     }
 
     #[bench]
     fn bench_ip_range_hit_10_percent_intersect_with_90_percent(bench: &mut Bencher) {
         let index = get_index_0_to_100();
 
-        bench.iter(|| {
-            let start = Ipv6Addr::from_u128(0);
-            let end = Ipv6Addr::from_u128(10 * 1000);
-
-            excute_query(start, end, "AND id:many", &index)
-        });
+        bench.iter(|| excute_query("ip", get_10_percent(), "AND id:many", &index));
     }
 
     #[bench]
     fn bench_ip_range_hit_90_percent_intersect_with_90_percent(bench: &mut Bencher) {
         let index = get_index_0_to_100();
 
-        bench.iter(|| {
-            let start = Ipv6Addr::from_u128(0);
-            let end = Ipv6Addr::from_u128(90 * 1000);
-
-            excute_query(start, end, "AND id:many", &index)
-        });
+        bench.iter(|| excute_query("ip", get_90_percent(), "AND id:many", &index));
     }
 
     #[bench]
     fn bench_ip_range_hit_90_percent_intersect_with_10_percent(bench: &mut Bencher) {
         let index = get_index_0_to_100();
 
-        bench.iter(|| {
-            let start = Ipv6Addr::from_u128(0);
-            let end = Ipv6Addr::from_u128(90 * 1000);
-
-            excute_query(start, end, "AND id:few", &index)
-        });
+        bench.iter(|| excute_query("ip", get_90_percent(), "AND id:few", &index));
     }
 
     #[bench]
     fn bench_ip_range_hit_90_percent_intersect_with_1_percent(bench: &mut Bencher) {
         let index = get_index_0_to_100();
 
-        bench.iter(|| {
-            let start = Ipv6Addr::from_u128(0);
-            let end = Ipv6Addr::from_u128(90 * 1000);
+        bench.iter(|| excute_query("ip", get_90_percent(), "AND id:veryfew", &index));
+    }
 
-            excute_query(start, end, "AND id:veryfew", &index)
-        });
+    #[bench]
+    fn bench_ip_range_hit_90_percent_multi(bench: &mut Bencher) {
+        let index = get_index_0_to_100();
+
+        bench.iter(|| excute_query("ips", get_90_percent(), "", &index));
+    }
+
+    #[bench]
+    fn bench_ip_range_hit_10_percent_multi(bench: &mut Bencher) {
+        let index = get_index_0_to_100();
+
+        bench.iter(|| excute_query("ips", get_10_percent(), "", &index));
+    }
+
+    #[bench]
+    fn bench_ip_range_hit_1_percent_multi(bench: &mut Bencher) {
+        let index = get_index_0_to_100();
+
+        bench.iter(|| excute_query("ips", get_1_percent(), "", &index));
+    }
+
+    #[bench]
+    fn bench_ip_range_hit_10_percent_intersect_with_10_percent_multi(bench: &mut Bencher) {
+        let index = get_index_0_to_100();
+
+        bench.iter(|| excute_query("ips", get_10_percent(), "AND id:few", &index));
+    }
+
+    #[bench]
+    fn bench_ip_range_hit_1_percent_intersect_with_10_percent_multi(bench: &mut Bencher) {
+        let index = get_index_0_to_100();
+
+        bench.iter(|| excute_query("ips", get_1_percent(), "AND id:few", &index));
+    }
+
+    #[bench]
+    fn bench_ip_range_hit_1_percent_intersect_with_90_percent_multi(bench: &mut Bencher) {
+        let index = get_index_0_to_100();
+
+        bench.iter(|| excute_query("ips", get_1_percent(), "AND id:many", &index));
+    }
+
+    #[bench]
+    fn bench_ip_range_hit_1_percent_intersect_with_1_percent_multi(bench: &mut Bencher) {
+        let index = get_index_0_to_100();
+
+        bench.iter(|| excute_query("ips", get_1_percent(), "AND id:veryfew", &index));
+    }
+
+    #[bench]
+    fn bench_ip_range_hit_10_percent_intersect_with_90_percent_multi(bench: &mut Bencher) {
+        let index = get_index_0_to_100();
+
+        bench.iter(|| excute_query("ips", get_10_percent(), "AND id:many", &index));
+    }
+
+    #[bench]
+    fn bench_ip_range_hit_90_percent_intersect_with_90_percent_multi(bench: &mut Bencher) {
+        let index = get_index_0_to_100();
+
+        bench.iter(|| excute_query("ips", get_90_percent(), "AND id:many", &index));
+    }
+
+    #[bench]
+    fn bench_ip_range_hit_90_percent_intersect_with_10_percent_multi(bench: &mut Bencher) {
+        let index = get_index_0_to_100();
+
+        bench.iter(|| excute_query("ips", get_90_percent(), "AND id:few", &index));
+    }
+
+    #[bench]
+    fn bench_ip_range_hit_90_percent_intersect_with_1_percent_multi(bench: &mut Bencher) {
+        let index = get_index_0_to_100();
+
+        bench.iter(|| excute_query("ips", get_90_percent(), "AND id:veryfew", &index));
     }
 }

src/query/term_query/term_weight.rs

@@ -5,7 +5,7 @@ use crate::fieldnorm::FieldNormReader;
 use crate::postings::SegmentPostings;
 use crate::query::bm25::Bm25Weight;
 use crate::query::explanation::does_not_match;
-use crate::query::weight::for_each_scorer;
+use crate::query::weight::{for_each_docset, for_each_scorer};
 use crate::query::{Explanation, Scorer, Weight};
 use crate::schema::IndexRecordOption;
 use crate::{DocId, Score, Term};
@@ -56,6 +56,18 @@ impl Weight for TermWeight {
         Ok(())
     }
 
+    /// Iterates through all of the document matched by the DocSet
+    /// `DocSet` and push the scored documents to the collector.
+    fn for_each_no_score(
+        &self,
+        reader: &SegmentReader,
+        callback: &mut dyn FnMut(DocId),
+    ) -> crate::Result<()> {
+        let mut scorer = self.specialized_scorer(reader, 1.0)?;
+        for_each_docset(&mut scorer, callback);
+        Ok(())
+    }
+
     /// Calls `callback` with all of the `(doc, score)` for which score
     /// is exceeding a given threshold.
     ///

src/query/union.rs

@@ -94,8 +94,8 @@ impl<TScorer: Scorer, TScoreCombiner: ScoreCombiner> Union<TScorer, TScoreCombin
             self.doc = min_doc;
             refill(
                 &mut self.docsets,
-                &mut *self.bitsets,
-                &mut *self.scores,
+                &mut self.bitsets,
+                &mut self.scores,
                 min_doc,
             );
             true

src/query/weight.rs

@@ -1,10 +1,10 @@
 use super::Scorer;
 use crate::core::SegmentReader;
 use crate::query::Explanation;
-use crate::{DocId, Score, TERMINATED};
+use crate::{DocId, DocSet, Score, TERMINATED};
 
-/// Iterates through all of the document matched by the DocSet
-/// `DocSet` and push the scored documents to the collector.
+/// Iterates through all of the documents and scores matched by the DocSet
+/// `DocSet`.
 pub(crate) fn for_each_scorer<TScorer: Scorer + ?Sized>(
     scorer: &mut TScorer,
     callback: &mut dyn FnMut(DocId, Score),
@@ -16,6 +16,16 @@ pub(crate) fn for_each_scorer<TScorer: Scorer + ?Sized>(
     }
 }
 
+/// Iterates through all of the documents matched by the DocSet
+/// `DocSet`.
+pub(crate) fn for_each_docset<T: DocSet + ?Sized>(docset: &mut T, callback: &mut dyn FnMut(DocId)) {
+    let mut doc = docset.doc();
+    while doc != TERMINATED {
+        callback(doc);
+        doc = docset.advance();
+    }
+}
+
 /// Calls `callback` with all of the `(doc, score)` for which score
 /// is exceeding a given threshold.
 ///
@@ -78,6 +88,18 @@ pub trait Weight: Send + Sync + 'static {
         Ok(())
     }
 
+    /// Iterates through all of the document matched by the DocSet
+    /// `DocSet` and push the scored documents to the collector.
+    fn for_each_no_score(
+        &self,
+        reader: &SegmentReader,
+        callback: &mut dyn FnMut(DocId),
+    ) -> crate::Result<()> {
+        let mut docset = self.scorer(reader, 1.0)?;
+        for_each_docset(docset.as_mut(), callback);
+        Ok(())
+    }
+
     /// Calls `callback` with all of the `(doc, score)` for which score
     /// is exceeding a given threshold.
     ///

src/store/footer.rs

@@ -2,7 +2,7 @@ use std::io;
 
 use common::{BinarySerializable, FixedSize, HasLen};
 
-use super::Decompressor;
+use super::{Decompressor, DOC_STORE_VERSION};
 use crate::directory::FileSlice;
 
 #[derive(Debug, Clone, PartialEq)]
@@ -17,6 +17,7 @@ pub struct DocStoreFooter {
 /// - reserved for future use: 15 bytes
 impl BinarySerializable for DocStoreFooter {
     fn serialize<W: io::Write>(&self, writer: &mut W) -> io::Result<()> {
+        BinarySerializable::serialize(&DOC_STORE_VERSION, writer)?;
         BinarySerializable::serialize(&self.offset, writer)?;
         BinarySerializable::serialize(&self.decompressor.get_id(), writer)?;
         writer.write_all(&[0; 15])?;
@@ -24,6 +25,13 @@ impl BinarySerializable for DocStoreFooter {
     }
 
     fn deserialize<R: io::Read>(reader: &mut R) -> io::Result<Self> {
+        let doc_store_version = u32::deserialize(reader)?;
+        if doc_store_version != DOC_STORE_VERSION {
+            panic!(
+                "actual doc store version: {}, expected: {}",
+                doc_store_version, DOC_STORE_VERSION
+            );
+        }
         let offset = u64::deserialize(reader)?;
         let compressor_id = u8::deserialize(reader)?;
         let mut skip_buf = [0; 15];
@@ -36,7 +44,7 @@ impl BinarySerializable for DocStoreFooter {
 }
 
 impl FixedSize for DocStoreFooter {
-    const SIZE_IN_BYTES: usize = 24;
+    const SIZE_IN_BYTES: usize = 28;
 }
 
 impl DocStoreFooter {

src/store/mod.rs

@@ -44,6 +44,9 @@ pub use self::reader::{CacheStats, StoreReader};
 pub use self::writer::StoreWriter;
 mod store_compressor;
 
+/// Doc store version in footer to handle format changes.
+pub(crate) const DOC_STORE_VERSION: u32 = 1;
+
 #[cfg(feature = "lz4-compression")]
 mod compression_lz4_block;
 

src/termdict/tests.rs

@@ -229,10 +229,10 @@ fn test_empty_string() -> crate::Result<()> {
     let buffer: Vec<u8> = {
         let mut term_dictionary_builder = TermDictionaryBuilder::create(vec![]).unwrap();
         term_dictionary_builder
-            .insert(&[], &make_term_info(1_u64))
+            .insert([], &make_term_info(1_u64))
             .unwrap();
         term_dictionary_builder
-            .insert(&[1u8], &make_term_info(2_u64))
+            .insert([1u8], &make_term_info(2_u64))
             .unwrap();
         term_dictionary_builder.finish()?
     };
@@ -252,7 +252,7 @@ fn stream_range_test_dict() -> crate::Result<TermDictionary> {
         let mut term_dictionary_builder = TermDictionaryBuilder::create(Vec::new())?;
         for i in 0u8..10u8 {
             let number_arr = [i; 1];
-            term_dictionary_builder.insert(&number_arr, &make_term_info(i as u64))?;
+            term_dictionary_builder.insert(number_arr, &make_term_info(i as u64))?;
         }
         term_dictionary_builder.finish()?
     };

src/tokenizer/stop_word_filter.rs

@@ -10,6 +10,8 @@
 //! assert_eq!(stream.next().unwrap().text, "crafty");
 //! assert!(stream.next().is_none());
 //! ```
+use std::sync::Arc;
+
 use rustc_hash::FxHashSet;
 
 use super::{Token, TokenFilter, TokenStream};
@@ -18,19 +20,15 @@ use crate::tokenizer::BoxTokenStream;
 /// `TokenFilter` that removes stop words from a token stream
 #[derive(Clone)]
 pub struct StopWordFilter {
-    words: FxHashSet<String>,
+    words: Arc<FxHashSet<String>>,
 }
 
 impl StopWordFilter {
     /// Creates a `StopWordFilter` given a list of words to remove
-    pub fn remove(words: Vec<String>) -> StopWordFilter {
-        let mut set = FxHashSet::default();
-
-        for word in words {
-            set.insert(word);
+    pub fn remove<W: IntoIterator<Item = String>>(words: W) -> StopWordFilter {
+        StopWordFilter {
+            words: Arc::new(words.into_iter().collect()),
         }
-
-        StopWordFilter { words: set }
     }
 
     fn english() -> StopWordFilter {
@@ -40,12 +38,12 @@ impl StopWordFilter {
             "there", "these", "they", "this", "to", "was", "will", "with",
         ];
 
-        StopWordFilter::remove(words.iter().map(|&s| s.to_string()).collect())
+        StopWordFilter::remove(words.iter().map(|&s| s.to_string()))
     }
 }
 
 pub struct StopWordFilterStream<'a> {
-    words: FxHashSet<String>,
+    words: Arc<FxHashSet<String>>,
     tail: BoxTokenStream<'a>,
 }