Added field norm readers

* Assert nearly equals macro

* Renamed specialized_scorer in TermScorer

Cargo.toml

@@ -58,7 +58,6 @@ winapi = "0.3"
 rand = "0.7"
 maplit = "1"
 matches = "0.1.8"
-proptest = "0.10"
 
 [dev-dependencies.fail]
 version = "0.4"

src/collector/top_score_collector.rs

@@ -519,7 +519,7 @@ impl Collector for TopDocs {
             })?;
         }
 
-        let fruit: Vec<(Score, DocAddress)> = heap
+        let fruit = heap
             .into_sorted_vec()
             .into_iter()
             .map(|cid| (cid.feature, DocAddress(segment_ord, cid.doc)))

src/common/serialize.rs

@@ -89,19 +89,6 @@ impl FixedSize for u64 {
     const SIZE_IN_BYTES: usize = 8;
 }
 
-impl BinarySerializable for f32 {
-    fn serialize<W: Write>(&self, writer: &mut W) -> io::Result<()> {
-        writer.write_f32::<Endianness>(*self)
-    }
-    fn deserialize<R: Read>(reader: &mut R) -> io::Result<Self> {
-        reader.read_f32::<Endianness>()
-    }
-}
-
-impl FixedSize for f32 {
-    const SIZE_IN_BYTES: usize = 4;
-}
-
 impl BinarySerializable for i64 {
     fn serialize<W: Write>(&self, writer: &mut W) -> io::Result<()> {
         writer.write_i64::<Endianness>(*self)

src/common/vint.rs

@@ -5,7 +5,7 @@ use std::io::Read;
 use std::io::Write;
 
 ///   Wrapper over a `u64` that serializes as a variable int.
-#[derive(Clone, Copy, Debug, Eq, PartialEq)]
+#[derive(Debug, Eq, PartialEq)]
 pub struct VInt(pub u64);
 
 const STOP_BIT: u8 = 128;

src/core/segment_reader.rs

@@ -126,8 +126,8 @@ impl SegmentReader {
     /// They are simply stored as a fast field, serialized in
     /// the `.fieldnorm` file of the segment.
     pub fn get_fieldnorms_reader(&self, field: Field) -> FieldNormReader {
-        if let Some(fieldnorm_source) = self.fieldnorm_readers.get_field(field) {
-            fieldnorm_source
+        if let Some(fieldnorm_reader) = self.fieldnorm_readers.get_field(field) {
+            fieldnorm_reader
         } else {
             let field_name = self.schema.get_field_name(field);
             let err_msg = format!(

src/directory/footer.rs

@@ -94,24 +94,12 @@ impl Footer {
         match &self.versioned_footer {
             VersionedFooter::V1 {
                 crc32: _crc,
-                store_compression,
+                store_compression: compression,
             } => {
-                if &library_version.store_compression != store_compression {
+                if &library_version.store_compression != compression {
                     return Err(Incompatibility::CompressionMismatch {
                         library_compression_format: library_version.store_compression.to_string(),
-                        index_compression_format: store_compression.to_string(),
-                    });
-                }
-                Ok(())
-            }
-            VersionedFooter::V2 {
-                crc32: _crc,
-                store_compression,
-            } => {
-                if &library_version.store_compression != store_compression {
-                    return Err(Incompatibility::CompressionMismatch {
-                        library_compression_format: library_version.store_compression.to_string(),
-                        index_compression_format: store_compression.to_string(),
+                        index_compression_format: compression.to_string(),
                     });
                 }
                 Ok(())
@@ -132,29 +120,24 @@ pub enum VersionedFooter {
         crc32: CrcHashU32,
         store_compression: String,
     },
-    // Introduction of the Block WAND information.
-    V2 {
-        crc32: CrcHashU32,
-        store_compression: String,
-    },
 }
 
 impl BinarySerializable for VersionedFooter {
     fn serialize<W: io::Write>(&self, writer: &mut W) -> io::Result<()> {
         let mut buf = Vec::new();
         match self {
-            VersionedFooter::V2 {
+            VersionedFooter::V1 {
                 crc32,
                 store_compression: compression,
             } => {
                 // Serializes a valid `VersionedFooter` or panics if the version is unknown
                 // [   version    |   crc_hash  | compression_mode ]
                 // [    0..4      |     4..8    |     variable     ]
-                BinarySerializable::serialize(&2u32, &mut buf)?;
+                BinarySerializable::serialize(&1u32, &mut buf)?;
                 BinarySerializable::serialize(crc32, &mut buf)?;
                 BinarySerializable::serialize(compression, &mut buf)?;
             }
-            VersionedFooter::V1 { .. } | VersionedFooter::UnknownVersion => {
+            VersionedFooter::UnknownVersion => {
                 return Err(io::Error::new(
                     io::ErrorKind::InvalidInput,
                     "Cannot serialize an unknown versioned footer ",
@@ -185,17 +168,10 @@ impl BinarySerializable for VersionedFooter {
         let version = u32::deserialize(&mut cursor)?;
         if version == 1 {
             let crc32 = u32::deserialize(&mut cursor)?;
-            let store_compression = String::deserialize(&mut cursor)?;
+            let compression = String::deserialize(&mut cursor)?;
             Ok(VersionedFooter::V1 {
                 crc32,
-                store_compression,
-            })
-        } else if version == 2 {
-            let crc32 = u32::deserialize(&mut cursor)?;
-            let store_compression = String::deserialize(&mut cursor)?;
-            Ok(VersionedFooter::V2 {
-                crc32,
-                store_compression,
+                store_compression: compression,
             })
         } else {
             Ok(VersionedFooter::UnknownVersion)
@@ -206,7 +182,6 @@ impl BinarySerializable for VersionedFooter {
 impl VersionedFooter {
     pub fn crc(&self) -> Option<CrcHashU32> {
         match self {
-            VersionedFooter::V2 { crc32, .. } => Some(*crc32),
             VersionedFooter::V1 { crc32, .. } => Some(*crc32),
             VersionedFooter::UnknownVersion { .. } => None,
         }
@@ -244,7 +219,7 @@ impl<W: TerminatingWrite> Write for FooterProxy<W> {
 impl<W: TerminatingWrite> TerminatingWrite for FooterProxy<W> {
     fn terminate_ref(&mut self, _: AntiCallToken) -> io::Result<()> {
         let crc32 = self.hasher.take().unwrap().finalize();
-        let footer = Footer::new(VersionedFooter::V2 {
+        let footer = Footer::new(VersionedFooter::V1 {
             crc32,
             store_compression: crate::store::COMPRESSION.to_string(),
         });
@@ -273,11 +248,15 @@ mod tests {
         assert!(footer_proxy.terminate().is_ok());
         assert_eq!(vec.len(), 167);
         let footer = Footer::deserialize(&mut &vec[..]).unwrap();
-        assert!(matches!(
-           footer.versioned_footer,
-           VersionedFooter::V2 { store_compression, .. }
-           if store_compression == crate::store::COMPRESSION
-        ));
+        if let VersionedFooter::V1 {
+            crc32: _,
+            store_compression,
+        } = footer.versioned_footer
+        {
+            assert_eq!(store_compression, crate::store::COMPRESSION);
+        } else {
+            panic!("Versioned footer should be V1.");
+        }
         assert_eq!(&footer.version, crate::version());
     }
 
@@ -285,7 +264,7 @@ mod tests {
     fn test_serialize_deserialize_footer() {
         let mut buffer = Vec::new();
         let crc32 = 123456u32;
-        let footer: Footer = Footer::new(VersionedFooter::V2 {
+        let footer: Footer = Footer::new(VersionedFooter::V1 {
             crc32,
             store_compression: "lz4".to_string(),
         });
@@ -297,7 +276,7 @@ mod tests {
     #[test]
     fn footer_length() {
         let crc32 = 1111111u32;
-        let versioned_footer = VersionedFooter::V2 {
+        let versioned_footer = VersionedFooter::V1 {
             crc32,
             store_compression: "lz4".to_string(),
         };
@@ -318,7 +297,7 @@ mod tests {
             // versionned footer length
             12 | 128,
             // index format version
-            2,
+            1,
             0,
             0,
             0,
@@ -337,7 +316,7 @@ mod tests {
         let versioned_footer = VersionedFooter::deserialize(&mut cursor).unwrap();
         assert!(cursor.is_empty());
         let expected_crc: u32 = LittleEndian::read_u32(&v_footer_bytes[5..9]) as CrcHashU32;
-        let expected_versioned_footer: VersionedFooter = VersionedFooter::V2 {
+        let expected_versioned_footer: VersionedFooter = VersionedFooter::V1 {
             crc32: expected_crc,
             store_compression: "lz4".to_string(),
         };

src/indexer/merger.rs

@@ -8,9 +8,9 @@ use crate::fastfield::DeleteBitSet;
 use crate::fastfield::FastFieldReader;
 use crate::fastfield::FastFieldSerializer;
 use crate::fastfield::MultiValueIntFastFieldReader;
+use crate::fieldnorm::FieldNormReader;
 use crate::fieldnorm::FieldNormsSerializer;
 use crate::fieldnorm::FieldNormsWriter;
-use crate::fieldnorm::{FieldNormReader, FieldNormReaders};
 use crate::indexer::SegmentSerializer;
 use crate::postings::InvertedIndexSerializer;
 use crate::postings::Postings;
@@ -20,7 +20,7 @@ use crate::schema::{Field, Schema};
 use crate::store::StoreWriter;
 use crate::termdict::TermMerger;
 use crate::termdict::TermOrdinal;
-use crate::{DocId, SegmentComponent};
+use crate::DocId;
 use std::cmp;
 use std::collections::HashMap;
 
@@ -493,7 +493,6 @@ impl IndexMerger {
         indexed_field: Field,
         field_type: &FieldType,
         serializer: &mut InvertedIndexSerializer,
-        fieldnorm_reader: Option<FieldNormReader>,
     ) -> crate::Result<Option<TermOrdinalMapping>> {
         let mut positions_buffer: Vec<u32> = Vec::with_capacity(1_000);
         let mut delta_computer = DeltaComputer::new();
@@ -552,8 +551,7 @@ impl IndexMerger {
         // - Segment 2's doc ids become  [seg0.max_doc + seg1.max_doc,
         //                                seg0.max_doc + seg1.max_doc + seg2.max_doc]
         // ...
-        let mut field_serializer =
-            serializer.new_field(indexed_field, total_num_tokens, fieldnorm_reader)?;
+        let mut field_serializer = serializer.new_field(indexed_field, total_num_tokens)?;
 
         let field_entry = self.schema.get_field_entry(indexed_field);
 
@@ -599,11 +597,7 @@ impl IndexMerger {
 
             // We know that there is at least one document containing
             // the term, so we add it.
-            let term_doc_freq = segment_postings
-                .iter()
-                .map(|(_, segment_posting)| segment_posting.doc_freq())
-                .sum();
-            let to_term_ord = field_serializer.new_term(term_bytes, term_doc_freq)?;
+            let to_term_ord = field_serializer.new_term(term_bytes)?;
 
             if let Some(ref mut term_ord_mapping) = term_ord_mapping_opt {
                 for (segment_ord, from_term_ord) in merged_terms.matching_segments() {
@@ -643,18 +637,13 @@ impl IndexMerger {
     fn write_postings(
         &self,
         serializer: &mut InvertedIndexSerializer,
-        fieldnorm_readers: FieldNormReaders,
     ) -> crate::Result<HashMap<Field, TermOrdinalMapping>> {
         let mut term_ordinal_mappings = HashMap::new();
         for (field, field_entry) in self.schema.fields() {
-            let fieldnorm_reader = fieldnorm_readers.get_field(field);
             if field_entry.is_indexed() {
-                if let Some(term_ordinal_mapping) = self.write_postings_for_field(
-                    field,
-                    field_entry.field_type(),
-                    serializer,
-                    fieldnorm_reader,
-                )? {
+                if let Some(term_ordinal_mapping) =
+                    self.write_postings_for_field(field, field_entry.field_type(), serializer)?
+                {
                     term_ordinal_mappings.insert(field, term_ordinal_mapping);
                 }
             }
@@ -683,12 +672,7 @@ impl SerializableSegment for IndexMerger {
         if let Some(fieldnorms_serializer) = serializer.extract_fieldnorms_serializer() {
             self.write_fieldnorms(fieldnorms_serializer)?;
         }
-        let fieldnorm_data = serializer
-            .segment()
-            .open_read(SegmentComponent::FIELDNORMS)?;
-        let fieldnorm_readers = FieldNormReaders::new(fieldnorm_data)?;
-        let term_ord_mappings =
-            self.write_postings(serializer.get_postings_serializer(), fieldnorm_readers)?;
+        let term_ord_mappings = self.write_postings(serializer.get_postings_serializer())?;
         self.write_fast_fields(serializer.get_fast_field_serializer(), term_ord_mappings)?;
         self.write_storable_fields(serializer.get_store_writer())?;
         serializer.close()?;
@@ -698,15 +682,15 @@ impl SerializableSegment for IndexMerger {
 
 #[cfg(test)]
 mod tests {
-    use crate::assert_nearly_equals;
     use crate::collector::tests::TEST_COLLECTOR_WITH_SCORE;
     use crate::collector::tests::{BytesFastFieldTestCollector, FastFieldTestCollector};
     use crate::collector::{Count, FacetCollector};
     use crate::core::Index;
     use crate::query::AllQuery;
     use crate::query::BooleanQuery;
-    use crate::query::Scorer;
     use crate::query::TermQuery;
+    use crate::schema;
+    use crate::schema::Cardinality;
     use crate::schema::Document;
     use crate::schema::Facet;
     use crate::schema::IndexRecordOption;
@@ -714,11 +698,9 @@ mod tests {
     use crate::schema::Term;
     use crate::schema::TextFieldIndexing;
     use crate::schema::INDEXED;
-    use crate::schema::{Cardinality, TEXT};
     use crate::DocAddress;
     use crate::IndexWriter;
     use crate::Searcher;
-    use crate::{schema, DocSet, SegmentId};
     use byteorder::{BigEndian, ReadBytesExt, WriteBytesExt};
     use futures::executor::block_on;
     use std::io::Cursor;
@@ -1541,71 +1523,4 @@ mod tests {
         assert_eq!(1, index.searchable_segments()?.len());
         Ok(())
     }
-
-    #[test]
-    fn test_merged_index_has_blockwand() -> crate::Result<()> {
-        let mut builder = schema::SchemaBuilder::new();
-        let text = builder.add_text_field("text", TEXT);
-        let index = Index::create_in_ram(builder.build());
-        let mut writer = index.writer_with_num_threads(1, 3_000_000)?;
-        let happy_term = Term::from_field_text(text, "happy");
-        let term_query = TermQuery::new(happy_term, IndexRecordOption::WithFreqs);
-        for _ in 0..62 {
-            writer.add_document(doc!(text=>"hello happy tax payer"));
-        }
-        writer.commit()?;
-        let reader = index.reader()?;
-        let searcher = reader.searcher();
-        let mut term_scorer = term_query
-            .specialized_weight(&searcher, true)
-            .specialized_scorer(searcher.segment_reader(0u32), 1.0f32)?;
-        assert_eq!(term_scorer.doc(), 0);
-        assert_nearly_equals!(term_scorer.block_max_score(), 0.0079681855);
-        assert_nearly_equals!(term_scorer.score(), 0.0079681855);
-        for _ in 0..81 {
-            writer.add_document(doc!(text=>"hello happy tax payer"));
-        }
-        writer.commit()?;
-        reader.reload()?;
-        let searcher = reader.searcher();
-
-        assert_eq!(searcher.segment_readers().len(), 2);
-        for segment_reader in searcher.segment_readers() {
-            let mut term_scorer = term_query
-                .specialized_weight(&searcher, true)
-                .specialized_scorer(segment_reader, 1.0f32)?;
-            // the difference compared to before is instrinsic to the bm25 formula. no worries there.
-            for doc in segment_reader.doc_ids_alive() {
-                assert_eq!(term_scorer.doc(), doc);
-                assert_nearly_equals!(term_scorer.block_max_score(), 0.003478312);
-                assert_nearly_equals!(term_scorer.score(), 0.003478312);
-                term_scorer.advance();
-            }
-        }
-
-        let segment_ids: Vec<SegmentId> = searcher
-            .segment_readers()
-            .iter()
-            .map(|reader| reader.segment_id())
-            .collect();
-        block_on(writer.merge(&segment_ids[..]))?;
-
-        reader.reload()?;
-        let searcher = reader.searcher();
-        assert_eq!(searcher.segment_readers().len(), 1);
-
-        let segment_reader = searcher.segment_reader(0u32);
-        let mut term_scorer = term_query
-            .specialized_weight(&searcher, true)
-            .specialized_scorer(segment_reader, 1.0f32)?;
-        // the difference compared to before is instrinsic to the bm25 formula. no worries there.
-        for doc in segment_reader.doc_ids_alive() {
-            assert_eq!(term_scorer.doc(), doc);
-            assert_nearly_equals!(term_scorer.block_max_score(), 0.003478312);
-            assert_nearly_equals!(term_scorer.score(), 0.003478312);
-            term_scorer.advance();
-        }
-
-        Ok(())
-    }
 }

src/indexer/segment_serializer.rs

@@ -36,6 +36,7 @@ impl SegmentSerializer {
         })
     }
 
+    #[allow(dead_code)]
     pub fn segment(&self) -> &Segment {
         &self.segment
     }

src/indexer/segment_updater.rs

@@ -130,15 +130,14 @@ fn merge(
     // An IndexMerger is like a "view" of our merged segments.
     let merger: IndexMerger = IndexMerger::open(index.schema(), &segments[..])?;
 
-    let merged_segment_id = merged_segment.id();
-
     // ... we just serialize this index merger in our new segment to merge the two segments.
-    let segment_serializer = SegmentSerializer::for_segment(merged_segment)?;
+    let segment_serializer = SegmentSerializer::for_segment(merged_segment.clone())?;
 
     let num_docs = merger.write(segment_serializer)?;
 
-    let segment_meta = index.new_segment_meta(merged_segment_id, num_docs);
+    let merged_segment_id = merged_segment.id();
 
+    let segment_meta = index.new_segment_meta(merged_segment_id, num_docs);
     Ok(SegmentEntry::new(segment_meta, delete_cursor, None))
 }
 

src/indexer/segment_writer.rs

@@ -2,7 +2,7 @@ use super::operation::AddOperation;
 use crate::core::Segment;
 use crate::core::SerializableSegment;
 use crate::fastfield::FastFieldsWriter;
-use crate::fieldnorm::{FieldNormReaders, FieldNormsWriter};
+use crate::fieldnorm::FieldNormsWriter;
 use crate::indexer::segment_serializer::SegmentSerializer;
 use crate::postings::compute_table_size;
 use crate::postings::MultiFieldPostingsWriter;
@@ -14,8 +14,8 @@ use crate::schema::{Field, FieldEntry};
 use crate::tokenizer::{BoxTokenStream, PreTokenizedStream};
 use crate::tokenizer::{FacetTokenizer, TextAnalyzer};
 use crate::tokenizer::{TokenStreamChain, Tokenizer};
+use crate::DocId;
 use crate::Opstamp;
-use crate::{DocId, SegmentComponent};
 use std::io;
 use std::str;
 
@@ -284,12 +284,7 @@ fn write(
     if let Some(fieldnorms_serializer) = serializer.extract_fieldnorms_serializer() {
         fieldnorms_writer.serialize(fieldnorms_serializer)?;
     }
-    let fieldnorm_data = serializer
-        .segment()
-        .open_read(SegmentComponent::FIELDNORMS)?;
-    let fieldnorm_readers = FieldNormReaders::new(fieldnorm_data)?;
-    let term_ord_map =
-        multifield_postings.serialize(serializer.get_postings_serializer(), fieldnorm_readers)?;
+    let term_ord_map = multifield_postings.serialize(serializer.get_postings_serializer())?;
     fast_field_writers.serialize(serializer.get_fast_field_serializer(), &term_ord_map)?;
     serializer.close()?;
     Ok(())

src/lib.rs

@@ -173,7 +173,7 @@ use once_cell::sync::Lazy;
 use serde::{Deserialize, Serialize};
 
 /// Index format version.
-const INDEX_FORMAT_VERSION: u32 = 2;
+const INDEX_FORMAT_VERSION: u32 = 1;
 
 /// Structure version for the index.
 #[derive(Clone, PartialEq, Eq, Serialize, Deserialize)]

src/postings/block_segment_postings.rs

@@ -1,21 +1,11 @@
 use crate::common::{BinarySerializable, VInt};
 use crate::directory::ReadOnlySource;
-use crate::fieldnorm::FieldNormReader;
 use crate::postings::compression::{
     AlignedBuffer, BlockDecoder, VIntDecoder, COMPRESSION_BLOCK_SIZE,
 };
 use crate::postings::{BlockInfo, FreqReadingOption, SkipReader};
-use crate::query::BM25Weight;
 use crate::schema::IndexRecordOption;
-use crate::{DocId, Score, TERMINATED};
-
-fn max_f32<I: Iterator<Item = f32>>(mut it: I) -> Option<f32> {
-    if let Some(first) = it.next() {
-        Some(it.fold(first, f32::max))
-    } else {
-        None
-    }
-}
+use crate::{DocId, TERMINATED};
 
 /// `BlockSegmentPostings` is a cursor iterating over blocks
 /// of documents.
@@ -29,12 +19,11 @@ pub struct BlockSegmentPostings {
     loaded_offset: usize,
     freq_decoder: BlockDecoder,
     freq_reading_option: FreqReadingOption,
-    block_max_score_cache: Option<Score>,
 
-    doc_freq: u32,
+    doc_freq: usize,
 
     data: ReadOnlySource,
-    pub(crate) skip_reader: SkipReader,
+    skip_reader: SkipReader,
 }
 
 fn decode_bitpacked_block(
@@ -58,7 +47,6 @@ fn decode_vint_block(
     doc_offset: DocId,
     num_vint_docs: usize,
 ) {
-    doc_decoder.fill(TERMINATED);
     let num_consumed_bytes = doc_decoder.uncompress_vint_sorted(data, doc_offset, num_vint_docs);
     if let Some(freq_decoder) = freq_decoder_opt {
         freq_decoder.uncompress_vint_unsorted(&data[num_consumed_bytes..], num_vint_docs);
@@ -100,12 +88,12 @@ impl BlockSegmentPostings {
             None => SkipReader::new(ReadOnlySource::empty(), doc_freq, record_option),
         };
 
+        let doc_freq = doc_freq as usize;
         let mut block_segment_postings = BlockSegmentPostings {
             doc_decoder: BlockDecoder::with_val(TERMINATED),
             loaded_offset: std::usize::MAX,
             freq_decoder: BlockDecoder::with_val(1),
             freq_reading_option,
-            block_max_score_cache: None,
             doc_freq,
             data: postings_data,
             skip_reader,
@@ -114,43 +102,6 @@ impl BlockSegmentPostings {
         block_segment_postings
     }
 
-    /// Returns the block_max_score for the current block.
-    /// It does not require the block to be loaded. For instance, it is ok to call this method
-    /// after having called `.shallow_advance(..)`.
-    ///
-    /// See `TermScorer::block_max_score(..)` for more information.
-    pub fn block_max_score(
-        &mut self,
-        fieldnorm_reader: &FieldNormReader,
-        bm25_weight: &BM25Weight,
-    ) -> Score {
-        let (block_max_score_cache, skip_reader, doc_decoder, freq_decoder) = (
-            &mut self.block_max_score_cache,
-            &self.skip_reader,
-            &self.doc_decoder,
-            &self.freq_decoder,
-        );
-        *block_max_score_cache.get_or_insert_with(|| {
-            skip_reader
-                .block_max_score(bm25_weight)
-                .or_else(|| {
-                    let docs = doc_decoder.output_array();
-                    let freqs = freq_decoder.output_array();
-                    max_f32(docs.iter().cloned().zip(freqs.iter().cloned()).map(
-                        |(doc, term_freq)| {
-                            let fieldnorm_id = fieldnorm_reader.fieldnorm_id(doc);
-                            bm25_weight.score(fieldnorm_id, term_freq)
-                        },
-                    ))
-                })
-                .unwrap_or(0f32)
-        })
-    }
-
-    pub(crate) fn freq_reading_option(&self) -> FreqReadingOption {
-        self.freq_reading_option
-    }
-
     // Resets the block segment postings on another position
     // in the postings file.
     //
@@ -170,16 +121,15 @@ impl BlockSegmentPostings {
         } else {
             self.skip_reader.reset(ReadOnlySource::empty(), doc_freq);
         }
-        self.doc_freq = doc_freq;
+        self.doc_freq = doc_freq as usize;
         self.load_block();
     }
 
-    /// Returns the overall number of documents in the block postings.
-    /// It does not take in account whether documents are deleted or not.
+    /// Returns the document frequency associated to this block postings.
     ///
     /// This `doc_freq` is simply the sum of the length of all of the blocks
     /// length, and it does not take in account deleted documents.
-    pub fn doc_freq(&self) -> u32 {
+    pub fn doc_freq(&self) -> usize {
         self.doc_freq
     }
 
@@ -189,20 +139,11 @@ impl BlockSegmentPostings {
     /// returned by `.docs()` is empty.
     #[inline]
     pub fn docs(&self) -> &[DocId] {
-        debug_assert!(self.block_is_loaded());
         self.doc_decoder.output_array()
     }
 
-    /// Returns a full block, regardless of whetehr the block is complete or incomplete (
-    /// as it happens for the last block of the posting list).
-    ///
-    /// In the latter case, the block is guaranteed to be padded with the sentinel value:
-    /// `TERMINATED`. The array is also guaranteed to be aligned on 16 bytes = 128 bits.
-    ///
-    /// This method is useful to run SSE2 linear search.
     #[inline(always)]
     pub(crate) fn docs_aligned(&self) -> &AlignedBuffer {
-        debug_assert!(self.block_is_loaded());
         self.doc_decoder.output_aligned()
     }
 
@@ -215,14 +156,12 @@ impl BlockSegmentPostings {
     /// Return the array of `term freq` in the block.
     #[inline]
     pub fn freqs(&self) -> &[u32] {
-        debug_assert!(self.block_is_loaded());
         self.freq_decoder.output_array()
     }
 
     /// Return the frequency at index `idx` of the block.
     #[inline]
     pub fn freq(&self, idx: usize) -> u32 {
-        debug_assert!(self.block_is_loaded());
         self.freq_decoder.output(idx)
     }
 
@@ -233,10 +172,13 @@ impl BlockSegmentPostings {
     /// of any number between 1 and `NUM_DOCS_PER_BLOCK - 1`
     #[inline]
     pub fn block_len(&self) -> usize {
-        debug_assert!(self.block_is_loaded());
         self.doc_decoder.output_len
     }
 
+    pub(crate) fn position_offset(&self) -> u64 {
+        self.skip_reader.position_offset()
+    }
+
     /// Position on a block that may contains `target_doc`.
     ///
     /// If all docs are smaller than target, the block loaded may be empty,
@@ -246,26 +188,7 @@ impl BlockSegmentPostings {
         self.load_block();
     }
 
-    pub(crate) fn position_offset(&self) -> u64 {
-        self.skip_reader.position_offset()
-    }
-
-    /// Dangerous API! This calls seek on the skip list,
-    /// but does not `.load_block()` afterwards.
-    ///
-    /// `.load_block()` needs to be called manually afterwards.
-    /// If all docs are smaller than target, the block loaded may be empty,
-    /// or be the last an incomplete VInt block.
-    pub(crate) fn shallow_seek(&mut self, target_doc: DocId) {
-        self.skip_reader.seek(target_doc);
-    }
-
-    pub(crate) fn block_is_loaded(&self) -> bool {
-        self.loaded_offset == self.skip_reader.byte_offset()
-    }
-
-    pub(crate) fn load_block(&mut self) {
-        self.block_max_score_cache = None;
+    fn load_block(&mut self) {
         let offset = self.skip_reader.byte_offset();
         if self.loaded_offset == offset {
             return;
@@ -290,14 +213,11 @@ impl BlockSegmentPostings {
                     tf_num_bits,
                 );
             }
-            BlockInfo::VInt { num_docs } => {
-                let data = {
-                    if num_docs == 0 {
-                        &[]
-                    } else {
-                        &self.data.as_slice()[offset..]
-                    }
-                };
+            BlockInfo::VInt(num_vint_docs) => {
+                self.doc_decoder.clear();
+                if num_vint_docs == 0 {
+                    return;
+                }
                 decode_vint_block(
                     &mut self.doc_decoder,
                     if let FreqReadingOption::ReadFreq = self.freq_reading_option {
@@ -305,9 +225,9 @@ impl BlockSegmentPostings {
                     } else {
                         None
                     },
-                    data,
+                    &self.data.as_slice()[offset..],
                     self.skip_reader.last_doc_in_previous_block,
-                    num_docs as usize,
+                    num_vint_docs as usize,
                 );
             }
         }
@@ -325,10 +245,9 @@ impl BlockSegmentPostings {
     pub fn empty() -> BlockSegmentPostings {
         BlockSegmentPostings {
             doc_decoder: BlockDecoder::with_val(TERMINATED),
-            loaded_offset: 0,
+            loaded_offset: std::usize::MAX,
             freq_decoder: BlockDecoder::with_val(1),
             freq_reading_option: FreqReadingOption::NoFreq,
-            block_max_score_cache: None,
             doc_freq: 0,
             data: ReadOnlySource::new(vec![]),
             skip_reader: SkipReader::new(ReadOnlySource::new(vec![]), 0, IndexRecordOption::Basic),
@@ -354,10 +273,8 @@ mod tests {
     #[test]
     fn test_empty_segment_postings() {
         let mut postings = SegmentPostings::empty();
-        assert_eq!(postings.doc(), TERMINATED);
         assert_eq!(postings.advance(), TERMINATED);
         assert_eq!(postings.advance(), TERMINATED);
-        assert_eq!(postings.doc_freq(), 0);
         assert_eq!(postings.len(), 0);
     }
 
@@ -377,8 +294,6 @@ mod tests {
     #[test]
     fn test_empty_block_segment_postings() {
         let mut postings = BlockSegmentPostings::empty();
-        assert!(postings.docs().is_empty());
-        assert_eq!(postings.doc_freq(), 0);
         postings.advance();
         assert!(postings.docs().is_empty());
         assert_eq!(postings.doc_freq(), 0);
@@ -460,21 +375,19 @@ mod tests {
     }
 
     #[test]
-    fn test_block_segment_postings_seek2() {
+    fn test_block_segment_postings_skip2() {
         let mut docs = vec![0];
         for i in 0..1300 {
             docs.push((i * i / 100) + i);
         }
         let mut block_postings = build_block_postings(&docs[..]);
         for i in vec![0, 424, 10000] {
-            block_postings.shallow_seek(i);
-            block_postings.load_block();
+            block_postings.seek(i);
             let docs = block_postings.docs();
             assert!(docs[0] <= i);
             assert!(docs.last().cloned().unwrap_or(0u32) >= i);
         }
-        block_postings.shallow_seek(100_000);
-        block_postings.load_block();
+        block_postings.seek(100_000);
         assert_eq!(block_postings.doc(COMPRESSION_BLOCK_SIZE - 1), TERMINATED);
     }
 

src/postings/compression/mod.rs

@@ -108,8 +108,9 @@ impl BlockDecoder {
         self.output.0[idx]
     }
 
-    pub fn fill(&mut self, val: u32) {
-        self.output.0.iter_mut().for_each(|el| *el = val);
+    pub fn clear(&mut self) {
+        self.output_len = 0;
+        self.output.0.iter_mut().for_each(|el| *el = TERMINATED);
     }
 }
 
@@ -244,6 +245,19 @@ pub mod tests {
         }
     }
 
+    #[test]
+    fn test_clearing() {
+        let mut encoder = BlockEncoder::new();
+        let vals = (0u32..128u32).map(|i| i * 3).collect::<Vec<_>>();
+        let (num_bits, compressed) = encoder.compress_block_sorted(&vals[..], 0u32);
+        let mut decoder = BlockDecoder::default();
+        decoder.uncompress_block_sorted(compressed, 0u32, num_bits);
+        assert_eq!(decoder.output_len, 128);
+        assert_eq!(decoder.output_array(), &vals[..]);
+        decoder.clear();
+        assert!(decoder.output_array().is_empty());
+    }
+
     #[test]
     fn test_encode_unsorted_block_with_junk() {
         let mut compressed: Vec<u8> = Vec::new();

src/postings/mod.rs

@@ -73,10 +73,8 @@ pub mod tests {
         let mut segment = index.new_segment();
         let mut posting_serializer = InvertedIndexSerializer::open(&mut segment).unwrap();
         {
-            let mut field_serializer = posting_serializer
-                .new_field(text_field, 120 * 4, None)
-                .unwrap();
-            field_serializer.new_term("abc".as_bytes(), 12u32).unwrap();
+            let mut field_serializer = posting_serializer.new_field(text_field, 120 * 4).unwrap();
+            field_serializer.new_term("abc".as_bytes()).unwrap();
             for doc_id in 0u32..120u32 {
                 let delta_positions = vec![1, 2, 3, 2];
                 field_serializer

src/postings/postings_writer.rs

@@ -1,6 +1,5 @@
 use super::stacker::{Addr, MemoryArena, TermHashMap};
 
-use crate::fieldnorm::FieldNormReaders;
 use crate::postings::recorder::{
     BufferLender, NothingRecorder, Recorder, TFAndPositionRecorder, TermFrequencyRecorder,
 };
@@ -129,7 +128,6 @@ impl MultiFieldPostingsWriter {
     pub fn serialize(
         &self,
         serializer: &mut InvertedIndexSerializer,
-        fieldnorm_readers: FieldNormReaders,
     ) -> crate::Result<HashMap<Field, FnvHashMap<UnorderedTermId, TermOrdinal>>> {
         let mut term_offsets: Vec<(&[u8], Addr, UnorderedTermId)> =
             self.term_index.iter().collect();
@@ -163,12 +161,8 @@ impl MultiFieldPostingsWriter {
             }
 
             let postings_writer = &self.per_field_postings_writers[field.field_id() as usize];
-            let fieldnorm_reader = fieldnorm_readers.get_field(field);
-            let mut field_serializer = serializer.new_field(
-                field,
-                postings_writer.total_num_tokens(),
-                fieldnorm_reader,
-            )?;
+            let mut field_serializer =
+                serializer.new_field(field, postings_writer.total_num_tokens())?;
             postings_writer.serialize(
                 &term_offsets[start..stop],
                 &mut field_serializer,
@@ -303,8 +297,7 @@ impl<Rec: Recorder + 'static> PostingsWriter for SpecializedPostingsWriter<Rec>
         let mut buffer_lender = BufferLender::default();
         for &(term_bytes, addr, _) in term_addrs {
             let recorder: Rec = termdict_heap.read(addr);
-            let term_doc_freq = recorder.term_doc_freq().unwrap_or(0u32);
-            serializer.new_term(&term_bytes[4..], term_doc_freq)?;
+            serializer.new_term(&term_bytes[4..])?;
             recorder.serialize(&mut buffer_lender, serializer, heap)?;
             serializer.close_term()?;
         }

src/postings/recorder.rs

@@ -75,8 +75,6 @@ pub(crate) trait Recorder: Copy + 'static {
         serializer: &mut FieldSerializer<'_>,
         heap: &MemoryArena,
     ) -> io::Result<()>;
-    /// Returns the number of document containg this term.
-    fn term_doc_freq(&self) -> Option<u32>;
 }
 
 /// Only records the doc ids
@@ -115,16 +113,11 @@ impl Recorder for NothingRecorder {
     ) -> io::Result<()> {
         let buffer = buffer_lender.lend_u8();
         self.stack.read_to_end(heap, buffer);
-        // TODO avoid reading twice.
         for doc in VInt32Reader::new(&buffer[..]) {
             serializer.write_doc(doc as u32, 0u32, &[][..])?;
         }
         Ok(())
     }
-
-    fn term_doc_freq(&self) -> Option<u32> {
-        None
-    }
 }
 
 /// Recorder encoding document ids, and term frequencies
@@ -133,7 +126,6 @@ pub struct TermFrequencyRecorder {
     stack: ExpUnrolledLinkedList,
     current_doc: DocId,
     current_tf: u32,
-    term_doc_freq: u32,
 }
 
 impl Recorder for TermFrequencyRecorder {
@@ -142,7 +134,6 @@ impl Recorder for TermFrequencyRecorder {
             stack: ExpUnrolledLinkedList::new(),
             current_doc: u32::max_value(),
             current_tf: 0u32,
-            term_doc_freq: 0u32,
         }
     }
 
@@ -151,7 +142,6 @@ impl Recorder for TermFrequencyRecorder {
     }
 
     fn new_doc(&mut self, doc: DocId, heap: &mut MemoryArena) {
-        self.term_doc_freq += 1;
         self.current_doc = doc;
         let _ = write_u32_vint(doc, &mut self.stack.writer(heap));
     }
@@ -182,10 +172,6 @@ impl Recorder for TermFrequencyRecorder {
 
         Ok(())
     }
-
-    fn term_doc_freq(&self) -> Option<u32> {
-        Some(self.term_doc_freq)
-    }
 }
 
 /// Recorder encoding term frequencies as well as positions.
@@ -193,14 +179,12 @@ impl Recorder for TermFrequencyRecorder {
 pub struct TFAndPositionRecorder {
     stack: ExpUnrolledLinkedList,
     current_doc: DocId,
-    term_doc_freq: u32,
 }
 impl Recorder for TFAndPositionRecorder {
     fn new() -> Self {
         TFAndPositionRecorder {
             stack: ExpUnrolledLinkedList::new(),
             current_doc: u32::max_value(),
-            term_doc_freq: 0u32,
         }
     }
 
@@ -210,7 +194,6 @@ impl Recorder for TFAndPositionRecorder {
 
     fn new_doc(&mut self, doc: DocId, heap: &mut MemoryArena) {
         self.current_doc = doc;
-        self.term_doc_freq += 1u32;
         let _ = write_u32_vint(doc, &mut self.stack.writer(heap));
     }
 
@@ -250,10 +233,6 @@ impl Recorder for TFAndPositionRecorder {
         }
         Ok(())
     }
-
-    fn term_doc_freq(&self) -> Option<u32> {
-        Some(self.term_doc_freq)
-    }
 }
 
 #[cfg(test)]

src/postings/segment_postings.rs

@@ -13,7 +13,6 @@ use crate::schema::IndexRecordOption;
 use crate::DocId;
 
 use crate::directory::ReadOnlySource;
-use crate::fieldnorm::FieldNormReader;
 use crate::postings::BlockSegmentPostings;
 
 /// `SegmentPostings` represents the inverted list or postings associated to
@@ -22,7 +21,7 @@ use crate::postings::BlockSegmentPostings;
 /// As we iterate through the `SegmentPostings`, the frequencies are optionally decoded.
 /// Positions on the other hand, are optionally entirely decoded upfront.
 pub struct SegmentPostings {
-    pub(crate) block_cursor: BlockSegmentPostings,
+    block_cursor: BlockSegmentPostings,
     cur: usize,
     position_reader: Option<PositionReader>,
     block_searcher: BlockSearcher,
@@ -39,12 +38,6 @@ impl SegmentPostings {
         }
     }
 
-    /// Returns the overall number of documents in the block postings.
-    /// It does not take in account whether documents are deleted or not.
-    pub fn doc_freq(&self) -> u32 {
-        self.block_cursor.doc_freq()
-    }
-
     /// Creates a segment postings object with the given documents
     /// and no frequency encoded.
     ///
@@ -56,8 +49,7 @@ impl SegmentPostings {
     pub fn create_from_docs(docs: &[u32]) -> SegmentPostings {
         let mut buffer = Vec::new();
         {
-            let mut postings_serializer = PostingsSerializer::new(&mut buffer, false, false, None);
-            postings_serializer.new_term(docs.len() as u32);
+            let mut postings_serializer = PostingsSerializer::new(&mut buffer, false, false);
             for &doc in docs {
                 postings_serializer.write_doc(doc, 1u32);
             }
@@ -74,29 +66,6 @@ impl SegmentPostings {
         SegmentPostings::from_block_postings(block_segment_postings, None)
     }
 
-    /// Helper functions to create `SegmentPostings` for tests.
-    pub fn create_from_docs_and_tfs(
-        doc_and_tfs: &[(u32, u32)],
-        fieldnorm_reader: Option<FieldNormReader>,
-    ) -> crate::Result<SegmentPostings> {
-        let mut buffer = Vec::new();
-        let mut postings_serializer =
-            PostingsSerializer::new(&mut buffer, true, false, fieldnorm_reader);
-        postings_serializer.new_term(doc_and_tfs.len() as u32);
-        for &(doc, tf) in doc_and_tfs {
-            postings_serializer.write_doc(doc, tf);
-        }
-        postings_serializer
-            .close_term(doc_and_tfs.len() as u32)?;
-        let block_segment_postings = BlockSegmentPostings::from_data(
-            doc_and_tfs.len() as u32,
-            ReadOnlySource::from(buffer),
-            IndexRecordOption::WithFreqs,
-            IndexRecordOption::WithFreqs,
-        );
-        Ok(SegmentPostings::from_block_postings(block_segment_postings, None))
-    }
-
     /// Reads a Segment postings from an &[u8]
     ///
     /// * `len` - number of document in the posting lists.
@@ -121,7 +90,6 @@ impl DocSet for SegmentPostings {
     // next needs to be called a first time to point to the correct element.
     #[inline]
     fn advance(&mut self) -> DocId {
-        assert!(self.block_cursor.block_is_loaded());
         if self.cur == COMPRESSION_BLOCK_SIZE - 1 {
             self.cur = 0;
             self.block_cursor.advance();
@@ -173,7 +141,7 @@ impl DocSet for SegmentPostings {
 
 impl HasLen for SegmentPostings {
     fn len(&self) -> usize {
-        self.block_cursor.doc_freq() as usize
+        self.block_cursor.doc_freq()
     }
 }
 

src/postings/serializer.rs

@@ -3,16 +3,13 @@ use crate::common::{BinarySerializable, VInt};
 use crate::common::{CompositeWrite, CountingWriter};
 use crate::core::Segment;
 use crate::directory::WritePtr;
-use crate::fieldnorm::FieldNormReader;
 use crate::positions::PositionSerializer;
 use crate::postings::compression::{BlockEncoder, VIntEncoder, COMPRESSION_BLOCK_SIZE};
 use crate::postings::skip::SkipSerializer;
-use crate::query::BM25Weight;
 use crate::schema::Schema;
 use crate::schema::{Field, FieldEntry, FieldType};
 use crate::termdict::{TermDictionaryBuilder, TermOrdinal};
 use crate::DocId;
-use std::cmp::Ordering;
 use std::io::{self, Write};
 
 /// `InvertedIndexSerializer` is in charge of serializing
@@ -92,7 +89,6 @@ impl InvertedIndexSerializer {
         &mut self,
         field: Field,
         total_num_tokens: u64,
-        fieldnorm_reader: Option<FieldNormReader>,
     ) -> io::Result<FieldSerializer<'_>> {
         let field_entry: &FieldEntry = self.schema.get_field_entry(field);
         let term_dictionary_write = self.terms_write.for_field(field);
@@ -107,7 +103,6 @@ impl InvertedIndexSerializer {
             postings_write,
             positions_write,
             positionsidx_write,
-            fieldnorm_reader,
         )
     }
 
@@ -139,7 +134,6 @@ impl<'a> FieldSerializer<'a> {
         postings_write: &'a mut CountingWriter<WritePtr>,
         positions_write: &'a mut CountingWriter<WritePtr>,
         positionsidx_write: &'a mut CountingWriter<WritePtr>,
-        fieldnorm_reader: Option<FieldNormReader>,
     ) -> io::Result<FieldSerializer<'a>> {
         let (term_freq_enabled, position_enabled): (bool, bool) = match field_type {
             FieldType::Str(ref text_options) => {
@@ -153,12 +147,8 @@ impl<'a> FieldSerializer<'a> {
             _ => (false, false),
         };
         let term_dictionary_builder = TermDictionaryBuilder::create(term_dictionary_write)?;
-        let postings_serializer = PostingsSerializer::new(
-            postings_write,
-            term_freq_enabled,
-            position_enabled,
-            fieldnorm_reader,
-        );
+        let postings_serializer =
+            PostingsSerializer::new(postings_write, term_freq_enabled, position_enabled);
         let positions_serializer_opt = if position_enabled {
             Some(PositionSerializer::new(positions_write, positionsidx_write))
         } else {
@@ -191,8 +181,8 @@ impl<'a> FieldSerializer<'a> {
     /// Starts the postings for a new term.
     /// * term - the term. It needs to come after the previous term according
     ///   to the lexicographical order.
-    /// * term_doc_freq - return the number of document containing the term.
-    pub fn new_term(&mut self, term: &[u8], term_doc_freq: u32) -> io::Result<TermOrdinal> {
+    /// * doc_freq - return the number of document containing the term.
+    pub fn new_term(&mut self, term: &[u8]) -> io::Result<TermOrdinal> {
         assert!(
             !self.term_open,
             "Called new_term, while the previous term was not closed."
@@ -203,7 +193,6 @@ impl<'a> FieldSerializer<'a> {
         self.term_dictionary_builder.insert_key(term)?;
         let term_ordinal = self.num_terms;
         self.num_terms += 1;
-        self.postings_serializer.new_term(term_doc_freq);
         Ok(term_ordinal)
     }
 
@@ -317,21 +306,6 @@ pub struct PostingsSerializer<W: Write> {
 
     termfreq_enabled: bool,
     termfreq_sum_enabled: bool,
-    fieldnorm_reader: Option<FieldNormReader>,
-
-    bm25_weight: Option<BM25Weight>,
-
-    num_docs: u32, // Number of docs in the segment
-    avg_fieldnorm: f32, // Average number of term in the field for that segment.
-                   // this value is used to compute the block wand information.
-}
-
-fn get_avg_fieldnorm(fieldnorm_reader: &FieldNormReader) -> f32 {
-    let num_docs = fieldnorm_reader.num_docs();
-    let sum_fieldnorm: f32 = (0u32..num_docs)
-        .map(|doc| fieldnorm_reader.fieldnorm(doc) as f32)
-        .sum();
-    sum_fieldnorm / (num_docs as f32)
 }
 
 impl<W: Write> PostingsSerializer<W> {
@@ -339,16 +313,7 @@ impl<W: Write> PostingsSerializer<W> {
         write: W,
         termfreq_enabled: bool,
         termfreq_sum_enabled: bool,
-        fieldnorm_reader: Option<FieldNormReader>,
     ) -> PostingsSerializer<W> {
-        let avg_fieldnorm: f32 = fieldnorm_reader
-            .as_ref()
-            .map(get_avg_fieldnorm)
-            .unwrap_or(0f32);
-        let num_docs = fieldnorm_reader
-            .as_ref()
-            .map(|fieldnorm_reader| fieldnorm_reader.num_docs())
-            .unwrap_or(0u32);
         PostingsSerializer {
             output_write: CountingWriter::wrap(write),
 
@@ -361,23 +326,6 @@ impl<W: Write> PostingsSerializer<W> {
             last_doc_id_encoded: 0u32,
             termfreq_enabled,
             termfreq_sum_enabled,
-
-            fieldnorm_reader,
-            bm25_weight: None,
-
-            num_docs,
-            avg_fieldnorm,
-        }
-    }
-
-    pub fn new_term(&mut self, term_doc_freq: u32) {
-        if self.termfreq_enabled && self.num_docs > 0 {
-            let bm25_weight = BM25Weight::for_one_term(
-                term_doc_freq as u64,
-                self.num_docs as u64,
-                self.avg_fieldnorm,
-            );
-            self.bm25_weight = Some(bm25_weight);
         }
     }
 
@@ -394,6 +342,7 @@ impl<W: Write> PostingsSerializer<W> {
             self.postings_write.extend(block_encoded);
         }
         if self.termfreq_enabled {
+            // encode the term_freqs
             let (num_bits, block_encoded): (u8, &[u8]) = self
                 .block_encoder
                 .compress_block_unsorted(&self.block.term_freqs());
@@ -403,32 +352,6 @@ impl<W: Write> PostingsSerializer<W> {
                 let sum_freq = self.block.term_freqs().iter().cloned().sum();
                 self.skip_write.write_total_term_freq(sum_freq);
             }
-            let mut blockwand_params_opt = None;
-            if let Some(bm25_weight) = self.bm25_weight.as_ref() {
-                if let Some(fieldnorm_reader) = self.fieldnorm_reader.as_ref() {
-                    let docs = self.block.doc_ids();
-                    let term_freqs = self.block.term_freqs();
-                    blockwand_params_opt = docs
-                        .iter()
-                        .cloned()
-                        .map(|doc| fieldnorm_reader.fieldnorm_id(doc))
-                        .zip(term_freqs.iter().cloned())
-                        .max_by(
-                            |(left_fieldnorm_id, left_term_freq),
-                             (right_fieldnorm_id, right_term_freq)| {
-                                let left_score =
-                                    bm25_weight.tf_factor(*left_fieldnorm_id, *left_term_freq);
-                                let right_score =
-                                    bm25_weight.tf_factor(*right_fieldnorm_id, *right_term_freq);
-                                left_score
-                                    .partial_cmp(&right_score)
-                                    .unwrap_or(Ordering::Equal)
-                            },
-                        );
-                }
-            }
-            let (fieldnorm_id, term_freq) = blockwand_params_opt.unwrap_or((0u8, 0u32));
-            self.skip_write.write_blockwand_max(fieldnorm_id, term_freq);
         }
         self.block.clear();
     }
@@ -477,7 +400,6 @@ impl<W: Write> PostingsSerializer<W> {
         }
         self.skip_write.clear();
         self.postings_write.clear();
-        self.bm25_weight = None;
         Ok(())
     }
 

src/postings/skip.rs

@@ -1,9 +1,8 @@
-use crate::common::{BinarySerializable, VInt};
+use crate::common::BinarySerializable;
 use crate::directory::ReadOnlySource;
 use crate::postings::compression::{compressed_block_size, COMPRESSION_BLOCK_SIZE};
-use crate::query::BM25Weight;
 use crate::schema::IndexRecordOption;
-use crate::{DocId, Score, TERMINATED};
+use crate::{DocId, TERMINATED};
 use owned_read::OwnedRead;
 
 pub struct SkipSerializer {
@@ -41,11 +40,6 @@ impl SkipSerializer {
             .expect("Should never fail");
     }
 
-    pub fn write_blockwand_max(&mut self, fieldnorm_id: u8, term_freq: u32) {
-        self.buffer.push(fieldnorm_id);
-        VInt(term_freq as u64).serialize_into_vec(&mut self.buffer);
-    }
-
     pub fn data(&self) -> &[u8] {
         &self.buffer[..]
     }
@@ -69,23 +63,19 @@ pub(crate) struct SkipReader {
     position_offset: u64,
 }
 
-#[derive(Clone, Copy, Debug)]
+#[derive(Clone, Eq, PartialEq, Copy, Debug)]
 pub(crate) enum BlockInfo {
     BitPacked {
         doc_num_bits: u8,
         tf_num_bits: u8,
         tf_sum: u32,
-        block_wand_fieldnorm_id: u8,
-        block_wand_term_freq: u32,
-    },
-    VInt {
-        num_docs: u32,
     },
+    VInt(u32),
 }
 
 impl Default for BlockInfo {
     fn default() -> Self {
-        BlockInfo::VInt { num_docs: 0u32 }
+        BlockInfo::VInt(0)
     }
 }
 
@@ -100,7 +90,7 @@ impl SkipReader {
             last_doc_in_previous_block: 0u32,
             owned_read: OwnedRead::new(data),
             skip_info,
-            block_info: BlockInfo::VInt { num_docs: doc_freq },
+            block_info: BlockInfo::VInt(doc_freq),
             byte_offset: 0,
             remaining_docs: doc_freq,
             position_offset: 0u64,
@@ -119,7 +109,7 @@ impl SkipReader {
         };
         self.last_doc_in_previous_block = 0u32;
         self.owned_read = OwnedRead::new(data);
-        self.block_info = BlockInfo::VInt { num_docs: doc_freq };
+        self.block_info = BlockInfo::VInt(doc_freq);
         self.byte_offset = 0;
         self.remaining_docs = doc_freq;
         self.position_offset = 0u64;
@@ -128,17 +118,8 @@ impl SkipReader {
         }
     }
 
-    pub fn block_max_score(&self, bm25_weight: &BM25Weight) -> Option<Score> {
-        match self.block_info {
-            BlockInfo::BitPacked {
-                block_wand_fieldnorm_id,
-                block_wand_term_freq,
-                ..
-            } => Some(bm25_weight.score(block_wand_fieldnorm_id, block_wand_term_freq)),
-            BlockInfo::VInt { .. } => None,
-        }
-    }
-
+    #[cfg(test)]
+    #[inline(always)]
     pub(crate) fn last_doc_in_block(&self) -> DocId {
         self.last_doc_in_block
     }
@@ -162,38 +143,25 @@ impl SkipReader {
                     doc_num_bits,
                     tf_num_bits: 0,
                     tf_sum: 0,
-                    block_wand_fieldnorm_id: 0,
-                    block_wand_term_freq: 0,
                 };
             }
             IndexRecordOption::WithFreqs => {
                 let tf_num_bits = self.owned_read.get(1);
-                let block_wand_fieldnorm_id = self.owned_read.get(2);
-                self.owned_read.advance(3);
-                let block_wand_term_freq =
-                    VInt::deserialize_u64(&mut self.owned_read).unwrap() as u32;
                 self.block_info = BlockInfo::BitPacked {
                     doc_num_bits,
                     tf_num_bits,
                     tf_sum: 0,
-                    block_wand_fieldnorm_id,
-                    block_wand_term_freq,
                 };
+                self.owned_read.advance(2);
             }
             IndexRecordOption::WithFreqsAndPositions => {
                 let tf_num_bits = self.owned_read.get(1);
                 self.owned_read.advance(2);
                 let tf_sum = u32::deserialize(&mut self.owned_read).expect("Failed reading tf_sum");
-                let block_wand_fieldnorm_id = self.owned_read.get(0);
-                self.owned_read.advance(1);
-                let block_wand_term_freq =
-                    VInt::deserialize_u64(&mut self.owned_read).unwrap() as u32;
                 self.block_info = BlockInfo::BitPacked {
                     doc_num_bits,
                     tf_num_bits,
                     tf_sum,
-                    block_wand_fieldnorm_id,
-                    block_wand_term_freq,
                 };
             }
         }
@@ -208,7 +176,7 @@ impl SkipReader {
     /// If the target is larger than all documents, the skip_reader
     /// then advance to the last Variable In block.
     pub fn seek(&mut self, target: DocId) {
-        while self.last_doc_in_block() < target {
+        while self.last_doc_in_block < target {
             self.advance();
         }
     }
@@ -219,14 +187,13 @@ impl SkipReader {
                 doc_num_bits,
                 tf_num_bits,
                 tf_sum,
-                ..
             } => {
                 self.remaining_docs -= COMPRESSION_BLOCK_SIZE as u32;
                 self.byte_offset += compressed_block_size(doc_num_bits + tf_num_bits);
                 self.position_offset += tf_sum as u64;
             }
-            BlockInfo::VInt { num_docs} => {
-                debug_assert_eq!(num_docs, self.remaining_docs);
+            BlockInfo::VInt(num_vint_docs) => {
+                debug_assert_eq!(num_vint_docs, self.remaining_docs);
                 self.remaining_docs = 0;
                 self.byte_offset = std::usize::MAX;
             }
@@ -236,7 +203,7 @@ impl SkipReader {
             self.read_block_info();
         } else {
             self.last_doc_in_block = TERMINATED;
-            self.block_info = BlockInfo::VInt { num_docs: self.remaining_docs };
+            self.block_info = BlockInfo::VInt(self.remaining_docs);
         }
     }
 }
@@ -256,10 +223,8 @@ mod tests {
             let mut skip_serializer = SkipSerializer::new();
             skip_serializer.write_doc(1u32, 2u8);
             skip_serializer.write_term_freq(3u8);
-            skip_serializer.write_blockwand_max(13u8, 3u32);
             skip_serializer.write_doc(5u32, 5u8);
             skip_serializer.write_term_freq(2u8);
-            skip_serializer.write_blockwand_max(8u8, 2u32);
             skip_serializer.data().to_owned()
         };
         let doc_freq = 3u32 + (COMPRESSION_BLOCK_SIZE * 2) as u32;
@@ -269,34 +234,30 @@ mod tests {
             IndexRecordOption::WithFreqs,
         );
         assert_eq!(skip_reader.last_doc_in_block(), 1u32);
-        assert!(matches!(
-            skip_reader.block_info,
+        assert_eq!(
+            skip_reader.block_info(),
             BlockInfo::BitPacked {
                 doc_num_bits: 2u8,
                 tf_num_bits: 3u8,
-                tf_sum: 0,
-                block_wand_fieldnorm_id: 13,
-                block_wand_term_freq: 3
+                tf_sum: 0
             }
-       ));
+        );
         skip_reader.advance();
         assert_eq!(skip_reader.last_doc_in_block(), 5u32);
-        assert!(matches!(
+        assert_eq!(
             skip_reader.block_info(),
             BlockInfo::BitPacked {
                 doc_num_bits: 5u8,
                 tf_num_bits: 2u8,
-                tf_sum: 0,
-                block_wand_fieldnorm_id: 8,
-                block_wand_term_freq: 2
+                tf_sum: 0
             }
-       ));
+        );
         skip_reader.advance();
-        assert!(matches!(skip_reader.block_info(), BlockInfo::VInt { num_docs: 3u32 }));
+        assert_eq!(skip_reader.block_info(), BlockInfo::VInt(3u32));
         skip_reader.advance();
-        assert!(matches!(skip_reader.block_info(), BlockInfo::VInt { num_docs: 0u32 }));
+        assert_eq!(skip_reader.block_info(), BlockInfo::VInt(0u32));
         skip_reader.advance();
-        assert!(matches!(skip_reader.block_info(), BlockInfo::VInt { num_docs: 0u32 }));
+        assert_eq!(skip_reader.block_info(), BlockInfo::VInt(0u32));
     }
 
     #[test]
@@ -314,34 +275,30 @@ mod tests {
             IndexRecordOption::Basic,
         );
         assert_eq!(skip_reader.last_doc_in_block(), 1u32);
-        assert!(matches!(
+        assert_eq!(
             skip_reader.block_info(),
             BlockInfo::BitPacked {
                 doc_num_bits: 2u8,
                 tf_num_bits: 0,
-                tf_sum: 0u32,
-                block_wand_fieldnorm_id: 0,
-                block_wand_term_freq: 0
+                tf_sum: 0u32
             }
-        ));
+        );
         skip_reader.advance();
         assert_eq!(skip_reader.last_doc_in_block(), 5u32);
-        assert!(matches!(
+        assert_eq!(
             skip_reader.block_info(),
             BlockInfo::BitPacked {
                 doc_num_bits: 5u8,
                 tf_num_bits: 0,
-                tf_sum: 0u32,
-                block_wand_fieldnorm_id: 0,
-                block_wand_term_freq: 0
+                tf_sum: 0u32
             }
-        ));
+        );
         skip_reader.advance();
-        assert!(matches!(skip_reader.block_info(), BlockInfo::VInt { num_docs: 3u32 }));
+        assert_eq!(skip_reader.block_info(), BlockInfo::VInt(3u32));
         skip_reader.advance();
-        assert!(matches!(skip_reader.block_info(), BlockInfo::VInt { num_docs: 0u32 }));
+        assert_eq!(skip_reader.block_info(), BlockInfo::VInt(0u32));
         skip_reader.advance();
-        assert!(matches!(skip_reader.block_info(), BlockInfo::VInt { num_docs: 0u32 }));
+        assert_eq!(skip_reader.block_info(), BlockInfo::VInt(0u32));
     }
 
     #[test]
@@ -358,17 +315,15 @@ mod tests {
             IndexRecordOption::Basic,
         );
         assert_eq!(skip_reader.last_doc_in_block(), 1u32);
-        assert!(matches!(
+        assert_eq!(
             skip_reader.block_info(),
             BlockInfo::BitPacked {
                 doc_num_bits: 2u8,
                 tf_num_bits: 0,
-                tf_sum: 0u32,
-                block_wand_fieldnorm_id: 0,
-                block_wand_term_freq: 0
+                tf_sum: 0u32
             }
-        ));
+        );
         skip_reader.advance();
-        assert!(matches!(skip_reader.block_info(), BlockInfo::VInt { num_docs: 0u32 }));
+        assert_eq!(skip_reader.block_info(), BlockInfo::VInt(0u32));
     }
 }

src/query/bm25.rs

@@ -3,14 +3,11 @@ use crate::query::Explanation;
 use crate::Score;
 use crate::Searcher;
 use crate::Term;
-use serde::Deserialize;
-use serde::Serialize;
 
 const K1: f32 = 1.2;
 const B: f32 = 0.75;
 
 fn idf(doc_freq: u64, doc_count: u64) -> f32 {
-    assert!(doc_count >= doc_freq, "{} >= {}", doc_count, doc_freq);
     let x = ((doc_count - doc_freq) as f32 + 0.5) / (doc_freq as f32 + 0.5);
     (1f32 + x).ln()
 }
@@ -28,12 +25,6 @@ fn compute_tf_cache(average_fieldnorm: f32) -> [f32; 256] {
     cache
 }
 
-#[derive(Clone, PartialEq, Debug, Serialize, Deserialize)]
-pub struct BM25Params {
-    pub idf: f32,
-    pub avg_fieldnorm: f32,
-}
-
 pub struct BM25Weight {
     idf_explain: Explanation,
     weight: f32,
@@ -71,9 +62,17 @@ impl BM25Weight {
         }
         let average_fieldnorm = total_num_tokens as f32 / total_num_docs as f32;
 
+        let mut idf_explain: Explanation;
         if terms.len() == 1 {
             let term_doc_freq = searcher.doc_freq(&terms[0]);
-            BM25Weight::for_one_term(term_doc_freq, total_num_docs, average_fieldnorm)
+            let idf = idf(term_doc_freq, total_num_docs);
+            idf_explain =
+                Explanation::new("idf, computed as log(1 + (N - n + 0.5) / (n + 0.5))", idf);
+            idf_explain.add_const(
+                "n, number of docs containing this term",
+                term_doc_freq as f32,
+            );
+            idf_explain.add_const("N, total number of docs", total_num_docs as f32);
         } else {
             let idf = terms
                 .iter()
@@ -82,21 +81,9 @@ impl BM25Weight {
                     idf(term_doc_freq, total_num_docs)
                 })
                 .sum::<f32>();
-            let idf_explain = Explanation::new("idf", idf);
-            BM25Weight::new(idf_explain, average_fieldnorm)
+            idf_explain = Explanation::new("idf", idf);
         }
-    }
-
-    pub fn for_one_term(term_doc_freq: u64, total_num_docs: u64, avg_fieldnorm: f32) -> BM25Weight {
-        let idf = idf(term_doc_freq, total_num_docs);
-        let mut idf_explain =
-            Explanation::new("idf, computed as log(1 + (N - n + 0.5) / (n + 0.5))", idf);
-        idf_explain.add_const(
-            "n, number of docs containing this term",
-            term_doc_freq as f32,
-        );
-        idf_explain.add_const("N, total number of docs", total_num_docs as f32);
-        BM25Weight::new(idf_explain, avg_fieldnorm)
+        BM25Weight::new(idf_explain, average_fieldnorm)
     }
 
     fn new(idf_explain: Explanation, average_fieldnorm: f32) -> BM25Weight {
@@ -111,23 +98,15 @@ impl BM25Weight {
 
     #[inline(always)]
     pub fn score(&self, fieldnorm_id: u8, term_freq: u32) -> Score {
-        self.weight * self.tf_factor(fieldnorm_id, term_freq)
-    }
-
-    pub fn max_score(&self) -> Score {
-        self.score(255u8, 2_013_265_944)
-    }
-
-    #[inline(always)]
-    pub(crate) fn tf_factor(&self, fieldnorm_id: u8, term_freq: u32) -> f32 {
-        let term_freq = term_freq as f32;
         let norm = self.cache[fieldnorm_id as usize];
-        term_freq / (term_freq + norm)
+        let term_freq = term_freq as f32;
+        self.weight * term_freq / (term_freq + norm)
     }
 
     pub fn explain(&self, fieldnorm_id: u8, term_freq: u32) -> Explanation {
         // The explain format is directly copied from Lucene's.
         // (So, Kudos to Lucene)
+
         let score = self.score(fieldnorm_id, term_freq);
 
         let norm = self.cache[fieldnorm_id as usize];
@@ -164,6 +143,6 @@ mod tests {
 
     #[test]
     fn test_idf() {
-        assert_nearly_equals!(idf(1, 2), std::f32::consts::LN_2);
+        assert_nearly_equals!(idf(1, 2), 0.6931472);
     }
 }

src/query/boolean_query/block_wand.rs

@@ -1,206 +0,0 @@
-use crate::query::term_query::TermScorer;
-use crate::query::Scorer;
-use crate::{DocId, DocSet, Score, TERMINATED};
-use std::ops::DerefMut;
-use std::ops::Deref;
-
-/// Takes a term_scorers sorted by their current doc() and a threshold and returns
-/// Returns (pivot_len, pivot_ord) defined as follows:
-/// - `pivot_doc` lowest document that has a chance of exceeding (>) the threshold score.
-/// - `before_pivot_len` number of term_scorers such that term_scorer.doc() < pivot.
-/// - `pivot_len` number of term_scorers such that term_scorer.doc() <= pivot.
-///
-/// We always have `before_pivot_len` < `pivot_len`.
-///
-/// None is returned if we establish that no document can exceed the threshold.
-fn find_pivot_doc(term_scorers: &[TermScorerWithMaxScore], threshold: f32) -> Option<(usize, usize, DocId)> {
-    let mut max_score = 0.0f32;
-    let mut before_pivot_len = 0;
-    let mut pivot_doc = TERMINATED;
-    while before_pivot_len < term_scorers.len() {
-        let term_scorer = &term_scorers[before_pivot_len];
-        max_score += term_scorer.max_score;
-        if max_score > threshold {
-            pivot_doc = term_scorer.doc();
-            break;
-        }
-        before_pivot_len += 1;
-    }
-    if pivot_doc == TERMINATED {
-        return None;
-    }
-    // Right now i is an ordinal, we want a len.
-    let mut pivot_len = before_pivot_len + 1;
-    // Some other term_scorer may be positioned on the same document.
-    pivot_len += term_scorers[pivot_len..].iter()
-        .take_while(|term_scorer| term_scorer.doc() == pivot_doc)
-        .count();
-    Some((before_pivot_len, pivot_len, pivot_doc))
-}
-
-struct TermScorerWithMaxScore<'a> {
-    scorer: &'a mut TermScorer,
-    max_score: f32,
-}
-
-impl<'a> From<&'a mut TermScorer> for TermScorerWithMaxScore<'a> {
-    fn from(scorer: &'a mut TermScorer) -> Self {
-        let max_score = scorer.max_score();
-        TermScorerWithMaxScore {
-            scorer,
-            max_score
-        }
-    }
-}
-
-impl<'a> Deref for TermScorerWithMaxScore<'a> {
-    type Target = TermScorer;
-
-    fn deref(&self) -> &Self::Target {
-        self.scorer
-    }
-}
-
-
-impl<'a> DerefMut for TermScorerWithMaxScore<'a> {
-    fn deref_mut(&mut self) -> &mut Self::Target {
-        self.scorer
-    }
-}
-
-// Before and after calling this method, scorers need to be sorted by their `.doc()`.
-fn block_max_was_too_low_advance_one_scorer(scorers: &mut Vec<TermScorerWithMaxScore>, pivot_len: usize) {
-    let mut scorer_to_seek = pivot_len - 1;
-    let mut doc_to_seek_after = scorers[scorer_to_seek].doc();
-    for scorer_ord in (0..pivot_len - 1).rev() {
-        let scorer = &scorers[scorer_ord];
-        if scorer.last_doc_in_block() <= doc_to_seek_after {
-            doc_to_seek_after = scorer.last_doc_in_block();
-            scorer_to_seek = scorer_ord;
-        }
-    }
-    for scorer in &scorers[pivot_len..] {
-        if scorer.doc() <= doc_to_seek_after {
-            doc_to_seek_after = scorer.doc();
-        }
-    }
-    scorers[scorer_to_seek].seek(doc_to_seek_after + 1);
-    restore_ordering(scorers, scorer_to_seek);
-}
-
-
-// Given a list of term_scorers and a `ord` and assuming that `term_scorers[ord]` is sorted
-// except term_scorers[ord] that might be in advance compared to its ranks,
-// bubble up term_scorers[ord] in order to restore the ordering.
-fn restore_ordering(term_scorers: &mut Vec<TermScorerWithMaxScore>, ord: usize) {
-    let doc = term_scorers[ord].doc();
-    for i in ord + 1..term_scorers.len() {
-        if term_scorers[i].doc() >= doc {
-            break;
-        }
-        term_scorers.swap(i, i - 1);
-    }
-}
-
-
-// Attempts to advance all term_scorers between `&term_scorers[0..before_len]` to the pivot.
-// If this works, return true.
-// If this fails (ie: one of the term_scorer does not contain `pivot_doc` and seek goes past the
-// pivot), reorder the term_scorers to ensure the list is still sorted and returns `false`.
-// If a term_scorer reach TERMINATED in the process return false remove the term_scorer and return.
-fn align_scorers(term_scorers: &mut Vec<TermScorerWithMaxScore>, pivot_doc: DocId, before_pivot_len: usize) -> bool {
-    debug_assert_ne!(pivot_doc, TERMINATED);
-    for i in (0..before_pivot_len).rev() {
-        let new_doc = term_scorers[i].seek(pivot_doc);
-        if new_doc != pivot_doc {
-            if new_doc == TERMINATED {
-                term_scorers.swap_remove(i);
-            }
-            // We went past the pivot.
-            // We just go through the outer loop mechanic (Note that pivot is
-            // still a possible candidate).
-            //
-            // Termination is still guaranteed since we can only consider the same
-            // pivot at most term_scorers.len() - 1 times.
-            restore_ordering(term_scorers, i);
-            return false;
-        }
-    }
-    return true;
-}
-
-// Assumes terms_scorers[..pivot_len] are positioned on the same doc (pivot_doc).
-// Advance term_scorers[..pivot_len] and out of these removes the terminated scores.
-// Restores the ordering of term_scorers.
-fn advance_all_scorers_on_pivot(term_scorers: &mut Vec<TermScorerWithMaxScore>, pivot_len: usize) {
-    let mut i = 0;
-    for _ in 0..pivot_len {
-        if term_scorers[i].advance() == TERMINATED {
-            term_scorers.swap_remove(i);
-        } else {
-            i += 1;
-        }
-    }
-    term_scorers.sort_by_key(|scorer| scorer.doc());
-}
-
-pub fn block_wand(
-    mut scorers: Vec<TermScorer>,
-    mut threshold: f32,
-    callback: &mut dyn FnMut(u32, Score) -> Score,
-) {
-    let mut scorers: Vec<TermScorerWithMaxScore> = scorers.iter_mut().map(TermScorerWithMaxScore::from).collect();
-    scorers.sort_by_key(|scorer| scorer.doc());
-    loop {
-        // At this point we need to ensure that the scorers are sorted!
-        if let Some((before_pivot_len, pivot_len, pivot_doc)) = find_pivot_doc(&scorers[..], threshold) {
-            debug_assert_ne!(pivot_doc, TERMINATED);
-            debug_assert!(before_pivot_len < pivot_len);
-            let block_max_score_upperbound: Score =  scorers[..pivot_len].iter_mut()
-                .map(|scorer| {
-                    scorer.shallow_seek(pivot_doc);
-                    scorer.block_max_score()
-                })
-                .sum();
-
-            // Beware after shallow advance, skip readers can be in advance compared to
-            // the segment posting lists.
-            //
-            // `block_segment_postings.load_block()` need to be called separately.
-            if block_max_score_upperbound <= threshold {
-                // Block max condition was not reached.
-                // We could get away by simply advancing the scorers to DocId + 1 but it would
-                // be inefficient. The optimization requires proper explanation and was
-                // isolated in a different function.
-                block_max_was_too_low_advance_one_scorer(&mut scorers, pivot_len);
-                continue;
-            }
-
-            // Block max condition is observed.
-            //
-            // Let's try and advance all scorers before the pivot to the pivot.
-            if !align_scorers(&mut scorers, pivot_doc, before_pivot_len) {
-                // At least of the scorer does not contain the pivot.
-                //
-                // Let's stop scoring this pivot and go through the pivot selection again.
-                // Note that the current pivot is not necessarily a bad candidate and it
-                // may be picked again.
-                continue;
-            }
-
-            // At this point, all scorers are positioned on the doc.
-            let score = scorers[..pivot_len]
-                .iter_mut()
-                .map(|scorer| scorer.score())
-                .sum();
-            if score > threshold {
-                threshold = callback(pivot_doc, score);
-            }
-            // let's advance all of the scorers that are currently positioned on the pivot.
-            advance_all_scorers_on_pivot(&mut scorers, pivot_len);
-
-        } else {
-            return;
-        }
-    }
-}

src/query/boolean_query/boolean_weight.rs

@@ -1,5 +1,4 @@
 use crate::core::SegmentReader;
-use crate::postings::FreqReadingOption;
 use crate::query::explanation::does_not_match;
 use crate::query::score_combiner::{DoNothingCombiner, ScoreCombiner, SumWithCoordsCombiner};
 use crate::query::term_query::TermScorer;
@@ -15,12 +14,12 @@ use crate::query::{intersect_scorers, Explanation};
 use crate::{DocId, Score};
 use std::collections::HashMap;
 
-enum SpecializedScorer {
-    TermUnion(Vec<TermScorer>),
+enum SpecializedScorer<TScoreCombiner: ScoreCombiner> {
+    TermUnion(Union<TermScorer, TScoreCombiner>),
     Other(Box<dyn Scorer>),
 }
 
-fn scorer_union<TScoreCombiner>(scorers: Vec<Box<dyn Scorer>>) -> SpecializedScorer
+fn scorer_union<TScoreCombiner>(scorers: Vec<Box<dyn Scorer>>) -> SpecializedScorer<TScoreCombiner>
 where
     TScoreCombiner: ScoreCombiner,
 {
@@ -36,29 +35,20 @@ where
                 .into_iter()
                 .map(|scorer| *(scorer.downcast::<TermScorer>().map_err(|_| ()).unwrap()))
                 .collect();
-            if scorers
-                .iter()
-                .all(|scorer| scorer.freq_reading_option() == FreqReadingOption::ReadFreq)
-            {
-                // Block wand is only available iff we read frequencies.
-                return SpecializedScorer::TermUnion(scorers);
-            } else {
-                return SpecializedScorer::Other(Box::new(Union::<_, TScoreCombiner>::from(
-                    scorers,
-                )));
-            }
+            return SpecializedScorer::TermUnion(Union::<TermScorer, TScoreCombiner>::from(
+                scorers,
+            ));
         }
     }
     SpecializedScorer::Other(Box::new(Union::<_, TScoreCombiner>::from(scorers)))
 }
 
-fn into_box_scorer<TScoreCombiner: ScoreCombiner>(scorer: SpecializedScorer) -> Box<dyn Scorer> {
-    match scorer {
-        SpecializedScorer::TermUnion(term_scorers) => {
-            let union_scorer = Union::<TermScorer, TScoreCombiner>::from(term_scorers);
-            Box::new(union_scorer)
+impl<TScoreCombiner: ScoreCombiner> Into<Box<dyn Scorer>> for SpecializedScorer<TScoreCombiner> {
+    fn into(self) -> Box<dyn Scorer> {
+        match self {
+            Self::TermUnion(union) => Box::new(union),
+            Self::Other(scorer) => scorer,
         }
-        SpecializedScorer::Other(scorer) => scorer,
     }
 }
 
@@ -95,47 +85,46 @@ impl BooleanWeight {
         &self,
         reader: &SegmentReader,
         boost: f32,
-    ) -> crate::Result<SpecializedScorer> {
+    ) -> crate::Result<SpecializedScorer<TScoreCombiner>> {
         let mut per_occur_scorers = self.per_occur_scorers(reader, boost)?;
 
-        let should_scorer_opt: Option<SpecializedScorer> = per_occur_scorers
+        let should_scorer_opt: Option<SpecializedScorer<TScoreCombiner>> = per_occur_scorers
             .remove(&Occur::Should)
             .map(scorer_union::<TScoreCombiner>);
 
         let exclude_scorer_opt: Option<Box<dyn Scorer>> = per_occur_scorers
             .remove(&Occur::MustNot)
             .map(scorer_union::<DoNothingCombiner>)
-            .map(into_box_scorer::<DoNothingCombiner>);
+            .map(Into::into);
 
         let must_scorer_opt: Option<Box<dyn Scorer>> = per_occur_scorers
             .remove(&Occur::Must)
             .map(intersect_scorers);
 
-        let positive_scorer: SpecializedScorer = match (should_scorer_opt, must_scorer_opt) {
-            (Some(should_scorer), Some(must_scorer)) => {
-                if self.scoring_enabled {
-                    SpecializedScorer::Other(Box::new(RequiredOptionalScorer::<
-                        Box<dyn Scorer>,
-                        Box<dyn Scorer>,
-                        TScoreCombiner,
-                    >::new(
-                        must_scorer,
-                        into_box_scorer::<TScoreCombiner>(should_scorer),
-                    )))
-                } else {
-                    SpecializedScorer::Other(must_scorer)
+        let positive_scorer: SpecializedScorer<TScoreCombiner> =
+            match (should_scorer_opt, must_scorer_opt) {
+                (Some(should_scorer), Some(must_scorer)) => {
+                    if self.scoring_enabled {
+                        SpecializedScorer::Other(Box::new(RequiredOptionalScorer::<
+                            Box<dyn Scorer>,
+                            Box<dyn Scorer>,
+                            TScoreCombiner,
+                        >::new(
+                            must_scorer, should_scorer.into()
+                        )))
+                    } else {
+                        SpecializedScorer::Other(must_scorer)
+                    }
                 }
-            }
-            (None, Some(must_scorer)) => SpecializedScorer::Other(must_scorer),
-            (Some(should_scorer), None) => should_scorer,
-            (None, None) => {
-                return Ok(SpecializedScorer::Other(Box::new(EmptyScorer)));
-            }
-        };
+                (None, Some(must_scorer)) => SpecializedScorer::Other(must_scorer),
+                (Some(should_scorer), None) => should_scorer,
+                (None, None) => {
+                    return Ok(SpecializedScorer::Other(Box::new(EmptyScorer)));
+                }
+            };
 
         if let Some(exclude_scorer) = exclude_scorer_opt {
-            let positive_scorer_boxed: Box<dyn Scorer> =
-                into_box_scorer::<TScoreCombiner>(positive_scorer);
+            let positive_scorer_boxed: Box<dyn Scorer> = positive_scorer.into();
             Ok(SpecializedScorer::Other(Box::new(Exclude::new(
                 positive_scorer_boxed,
                 exclude_scorer,
@@ -159,12 +148,10 @@ impl Weight for BooleanWeight {
             }
         } else if self.scoring_enabled {
             self.complex_scorer::<SumWithCoordsCombiner>(reader, boost)
-                .map(|specialized_scorer| {
-                    into_box_scorer::<SumWithCoordsCombiner>(specialized_scorer)
-                })
+                .map(Into::into)
         } else {
             self.complex_scorer::<DoNothingCombiner>(reader, boost)
-                .map(into_box_scorer::<DoNothingCombiner>)
+                .map(Into::into)
         }
     }
 
@@ -195,9 +182,7 @@ impl Weight for BooleanWeight {
     ) -> crate::Result<()> {
         let scorer = self.complex_scorer::<SumWithCoordsCombiner>(reader, 1.0f32)?;
         match scorer {
-            SpecializedScorer::TermUnion(term_scorers) => {
-                let mut union_scorer =
-                    Union::<TermScorer, SumWithCoordsCombiner>::from(term_scorers);
+            SpecializedScorer::TermUnion(mut union_scorer) => {
                 for_each_scorer(&mut union_scorer, callback);
             }
             SpecializedScorer::Other(mut scorer) => {
@@ -225,8 +210,8 @@ impl Weight for BooleanWeight {
     ) -> crate::Result<()> {
         let scorer = self.complex_scorer::<SumWithCoordsCombiner>(reader, 1.0f32)?;
         match scorer {
-            SpecializedScorer::TermUnion(term_scorers) => {
-                super::block_wand(term_scorers, threshold, callback);
+            SpecializedScorer::TermUnion(mut union_scorer) => {
+                for_each_pruning_scorer(&mut union_scorer, threshold, callback);
             }
             SpecializedScorer::Other(mut scorer) => {
                 for_each_pruning_scorer(scorer.as_mut(), threshold, callback);

src/query/boolean_query/mod.rs

@@ -1,8 +1,6 @@
-mod block_wand;
 mod boolean_query;
 mod boolean_weight;
 
-pub(crate) use self::block_wand::block_wand;
 pub use self::boolean_query::BooleanQuery;
 
 #[cfg(test)]

src/query/explanation.rs

@@ -1,6 +1,5 @@
 use crate::{DocId, TantivyError};
 use serde::Serialize;
-use std::fmt;
 
 pub(crate) fn does_not_match(doc: DocId) -> TantivyError {
     TantivyError::InvalidArgument(format!("Document #({}) does not match", doc))
@@ -19,12 +18,6 @@ pub struct Explanation {
     details: Vec<Explanation>,
 }
 
-impl fmt::Debug for Explanation {
-    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
-        write!(f, "Explanation({})", self.to_pretty_json())
-    }
-}
-
 impl Explanation {
     /// Creates a new explanation object.
     pub fn new<T: ToString>(description: T, value: f32) -> Explanation {

src/query/intersection.rs

@@ -119,6 +119,7 @@ impl<TDocSet: DocSet, TOtherDocSet: DocSet> DocSet for Intersection<TDocSet, TOt
                     continue 'outer;
                 }
             }
+
             debug_assert_eq!(candidate, self.left.doc());
             debug_assert_eq!(candidate, self.right.doc());
             debug_assert!(self.others.iter().all(|docset| docset.doc() == candidate));

src/query/mod.rs

@@ -26,7 +26,6 @@ mod weight;
 mod vec_docset;
 
 pub(crate) mod score_combiner;
-pub(crate) use self::bm25::BM25Weight;
 pub use self::intersection::Intersection;
 pub use self::union::Union;
 

src/query/term_query/mod.rs

@@ -15,7 +15,8 @@ mod tests {
     use crate::postings::compression::COMPRESSION_BLOCK_SIZE;
     use crate::query::{Query, QueryParser, Scorer, TermQuery};
     use crate::schema::{Field, IndexRecordOption, Schema, STRING, TEXT};
-    use crate::{Term, Index, TERMINATED};
+    use crate::Term;
+    use crate::{Index, TERMINATED};
 
     #[test]
     pub fn test_term_query_no_freq() {

src/query/term_query/term_scorer.rs

@@ -4,8 +4,8 @@ use crate::DocId;
 use crate::Score;
 
 use crate::fieldnorm::FieldNormReader;
+use crate::postings::Postings;
 use crate::postings::SegmentPostings;
-use crate::postings::{FreqReadingOption, Postings};
 use crate::query::bm25::BM25Weight;
 
 pub struct TermScorer {
@@ -26,70 +26,13 @@ impl TermScorer {
             similarity_weight,
         }
     }
+}
 
-    pub(crate) fn shallow_seek(&mut self, target_doc: DocId) {
-        self.postings.block_cursor.shallow_seek(target_doc)
-    }
-
-
-    #[cfg(test)]
-    pub fn create_for_test(
-        doc_and_tfs: &[(DocId, u32)],
-        fieldnorm_vals: &[u32],
-        similarity_weight: BM25Weight,
-    ) -> crate::Result<TermScorer> {
-        assert!(!doc_and_tfs.is_empty());
-        assert!(doc_and_tfs.len() <= fieldnorm_vals.len());
-
-        let doc_freq = doc_and_tfs.len();
-        let max_doc = doc_and_tfs.last().unwrap().0 + 1;
-        let mut fieldnorms: Vec<u32> = std::iter::repeat(1).take(max_doc as usize).collect();
-
-        for i in 0..doc_freq {
-            let doc = doc_and_tfs[i].0;
-            let fieldnorm = fieldnorm_vals[i];
-            fieldnorms[doc as usize] = fieldnorm;
-        }
-        let fieldnorm_reader = FieldNormReader::from(&fieldnorms[..]);
-
-        let segment_postings =
-            SegmentPostings::create_from_docs_and_tfs(doc_and_tfs, Some(fieldnorm_reader.clone()))?;
-
-        Ok(TermScorer::new(segment_postings, fieldnorm_reader, similarity_weight))
-    }
-
-    /// See `FreqReadingOption`.
-    pub(crate) fn freq_reading_option(&self) -> FreqReadingOption {
-        self.postings.block_cursor.freq_reading_option()
-    }
-
-    /// Returns the maximum score for the current block.
-    ///
-    /// In some rare case, the result may not be exact. In this case a lower value is returned,
-    /// (and may lead us to return a lesser document).
-    ///
-    /// At index time, we store the (fieldnorm_id, term frequency) pair that maximizes the
-    /// score assuming the average fieldnorm computed on this segment.
-    ///
-    /// Though extremely rare, it is theoretically possible that the actual average fieldnorm
-    /// is different enough from the current segment average fieldnorm that the maximum over a
-    /// specific is achieved on a different document.
-    ///
-    /// (The result is on the other hand guaranteed to be correct if there is only one segment).
-    pub fn block_max_score(&mut self) -> Score {
-        self.postings
-            .block_cursor
-            .block_max_score(&self.fieldnorm_reader, &self.similarity_weight)
-    }
-
+impl TermScorer {
     pub fn term_freq(&self) -> u32 {
         self.postings.term_freq()
     }
 
-    pub fn doc_freq(&self) -> usize {
-        self.postings.doc_freq() as usize
-    }
-
     pub fn fieldnorm_id(&self) -> u8 {
         self.fieldnorm_reader.fieldnorm_id(self.doc())
     }
@@ -99,14 +42,6 @@ impl TermScorer {
         let term_freq = self.term_freq();
         self.similarity_weight.explain(fieldnorm_id, term_freq)
     }
-
-    pub fn max_score(&self) -> f32 {
-        self.similarity_weight.max_score()
-    }
-
-    pub fn last_doc_in_block(&self) -> DocId {
-        self.postings.block_cursor.skip_reader.last_doc_in_block()
-    }
 }
 
 impl DocSet for TermScorer {
@@ -134,99 +69,3 @@ impl Scorer for TermScorer {
         self.similarity_weight.score(fieldnorm_id, term_freq)
     }
 }
-
-#[cfg(test)]
-mod tests {
-    use crate::assert_nearly_equals;
-    use crate::postings::compression::COMPRESSION_BLOCK_SIZE;
-    use crate::query::term_query::TermScorer;
-    use crate::query::{BM25Weight, Scorer};
-    use crate::{DocId, DocSet, TERMINATED};
-    use proptest::prelude::*;
-
-    #[test]
-    fn test_term_scorer_max_score() -> crate::Result<()> {
-        let bm25_weight = BM25Weight::for_one_term(3, 6, 10f32);
-        let mut term_scorer =
-            TermScorer::create_for_test(&[(2, 3), (3, 12), (7, 8)], &[10, 12, 100], bm25_weight)?;
-        let max_scorer = term_scorer.max_score();
-        assert_eq!(max_scorer, 1.3990127f32);
-        assert_eq!(term_scorer.doc(), 2);
-        assert_eq!(term_scorer.term_freq(), 3);
-        assert_nearly_equals!(term_scorer.block_max_score(), 1.3676447f32);
-        assert_nearly_equals!(term_scorer.score(), 1.0892314f32);
-        assert_eq!(term_scorer.advance(), 3);
-        assert_eq!(term_scorer.doc(), 3);
-        assert_eq!(term_scorer.term_freq(), 12);
-        assert_nearly_equals!(term_scorer.score(), 1.3676447f32);
-        assert_eq!(term_scorer.advance(), 7);
-        assert_eq!(term_scorer.doc(), 7);
-        assert_eq!(term_scorer.term_freq(), 8);
-        assert_nearly_equals!(term_scorer.score(), 0.72015285f32);
-        assert_eq!(term_scorer.advance(), TERMINATED);
-        Ok(())
-    }
-
-
-    #[test]
-    fn test_term_scorer_shallow_advance() -> crate::Result<()> {
-        let bm25_weight = BM25Weight::for_one_term(300, 1024, 10f32);
-        let mut doc_and_tfs = vec![];
-        for i in 0u32..300u32 {
-            let doc = i * 10;
-            doc_and_tfs.push((doc, 1u32 + doc % 3u32));
-        }
-        let fieldnorms: Vec<u32> = std::iter::repeat(10u32).take(1024).collect();
-        let mut term_scorer =
-            TermScorer::create_for_test(&doc_and_tfs, &fieldnorms, bm25_weight)?;
-        assert_eq!(term_scorer.doc(), 0u32);
-        term_scorer.shallow_seek(1289);
-        assert_eq!(term_scorer.doc(), 0u32);
-        term_scorer.seek(1289);
-        assert_eq!(term_scorer.doc(), 1290);
-        Ok(())
-    }
-
-    proptest! {
-        #[test]
-        fn test_term_scorer_block_max_score(term_freqs_fieldnorms in proptest::collection::vec((1u32..10u32, 0u32..100u32), 80..300)) {
-        let term_doc_freq = term_freqs_fieldnorms.len();
-         let doc_tfs: Vec<(u32, u32)> = term_freqs_fieldnorms.iter()
-                   .cloned()
-                  .enumerate()
-                  .map(|(doc, (tf, _))| (doc as u32, tf))
-                  .collect();
-
-         let mut fieldnorms: Vec<u32> = vec![];
-         for i in 0..term_doc_freq {
-             let (tf, num_extra_terms) = term_freqs_fieldnorms[i];
-             fieldnorms.push(tf + num_extra_terms);
-         }
-         let average_fieldnorm = fieldnorms
-             .iter()
-             .cloned()
-             .sum::<u32>() as f32 / term_doc_freq as f32;
-             // Average fieldnorm is over the entire index,
-             // not necessarily the docs that are in the posting list.
-             // For this reason we multiply by 1.1 to make a realistic value.
-         let bm25_weight = BM25Weight::for_one_term(term_doc_freq as u64,
-            term_doc_freq as u64 * 10u64,
-            average_fieldnorm);
-
-         let mut term_scorer =
-              TermScorer::create_for_test(&doc_tfs[..], &fieldnorms[..], bm25_weight).unwrap();
-
-         let docs: Vec<DocId> = (0..term_doc_freq).map(|doc| doc as DocId).collect();
-         for block in docs.chunks(COMPRESSION_BLOCK_SIZE) {
-             let block_max_score = term_scorer.block_max_score();
-             let mut block_max_score_computed = 0.0f32;
-             for &doc in block {
-                assert_eq!(term_scorer.doc(), doc);
-                block_max_score_computed = block_max_score_computed.max(term_scorer.score());
-                term_scorer.advance();
-             }
-             assert_nearly_equals!(block_max_score_computed, block_max_score);
-         }
-        }
-    }
-}

src/query/term_query/term_weight.rs

@@ -92,7 +92,7 @@ impl TermWeight {
         }
     }
 
-    pub fn specialized_scorer(&self, reader: &SegmentReader, boost: f32) -> Result<TermScorer> {
+    fn specialized_scorer(&self, reader: &SegmentReader, boost: f32) -> Result<TermScorer> {
         let field = self.term.field();
         let inverted_index = reader.inverted_index(field);
         let fieldnorm_reader = reader.get_fieldnorms_reader(field);