Updated CHANGELOG

- Pass Collector into TweakedScoreTopCollector and
  CustomScoreTopCollector.
- Add std:: qualifier to f32, i32 etc. Not sure why this was not failing
  already.
- Add unit tests for TopDocs with offset including for tweaked and
  custom score collectors.

In order to convert a TopCollector<Score> to a TopCollector<TScore> I
had to add a `into_tscore` method to `TopCollector`. This is a hack but
I don't know how to avoid it.

CHANGELOG.md

@@ -3,6 +3,21 @@ Tantivy 0.13.0
 - Bugfix in `FuzzyTermQuery` not matching terms by prefix when it should (@Peachball)
 - Relaxed constraints on the custom/tweak score functions. At the segment level, they can be mut, and they are not required to be Sync + Send.
 - `MMapDirectory::open` does not return a `Result` anymore.
+- Change in the DocSet and Scorer API. (@fulmicoton). 
+A freshly created DocSet point directly to their first doc. A sentinel value called TERMINATED marks the end of a DocSet.
+`.advance()` returns the new DocId. `Scorer::skip(target)` has been replaced by `Scorer::seek(target)` and returns the resulting DocId.
+As a result, iterating through DocSet now looks as follows
+```rust
+let mut doc = docset.doc();
+while doc != TERMINATED {
+   // ...
+   doc = docset.advance();
+}
+```
+The change made it possible to greatly simplify a lot of the docset's code.
+- Misc internal optimization and introduction of the `Scorer::for_each_pruning` function. (@fulmicoton)
+- Added an offset option to the Top(.*)Collectors. (@robyoung)
+
 
 Tantivy 0.12.0
 ======================

Cargo.toml

@@ -50,7 +50,6 @@ murmurhash32 = "0.2"
 chrono = "0.4"
 smallvec = "1.0"
 rayon = "1"
-# ordered-float = "1"
 
 [target.'cfg(windows)'.dependencies]
 winapi = "0.3"
@@ -59,8 +58,6 @@ winapi = "0.3"
 rand = "0.7"
 maplit = "1"
 matches = "0.1.8"
-proptest = "0.9"
-float-cmp = "0.6"
 
 [dev-dependencies.fail]
 version = "0.4"

examples/iterating_docs_and_positions.rs

@@ -10,7 +10,7 @@
 // ---
 // Importing tantivy...
 use tantivy::schema::*;
-use tantivy::{doc, DocId, DocSet, Index, Postings};
+use tantivy::{doc, DocSet, Index, Postings, TERMINATED};
 
 fn main() -> tantivy::Result<()> {
     // We first create a schema for the sake of the
@@ -62,12 +62,11 @@ fn main() -> tantivy::Result<()> {
         {
             // this buffer will be used to request for positions
             let mut positions: Vec<u32> = Vec::with_capacity(100);
-            while segment_postings.advance() {
-                // the number of time the term appears in the document.
-                let doc_id: DocId = segment_postings.doc(); //< do not try to access this before calling advance once.
-
+            let mut doc_id = segment_postings.doc();
+            while doc_id != TERMINATED {
                 // This MAY contains deleted documents as well.
                 if segment_reader.is_deleted(doc_id) {
+                    doc_id = segment_postings.advance();
                     continue;
                 }
 
@@ -86,6 +85,7 @@ fn main() -> tantivy::Result<()> {
                 // Doc 2: TermFreq 1: [0]
                 // ```
                 println!("Doc {}: TermFreq {}: {:?}", doc_id, term_freq, positions);
+                doc_id = segment_postings.advance();
             }
         }
     }

src/collector/custom_score_top_collector.rs

@@ -11,13 +11,13 @@ impl<TCustomScorer, TScore> CustomScoreTopCollector<TCustomScorer, TScore>
 where
     TScore: Clone + PartialOrd,
 {
-    pub fn new(
+    pub(crate) fn new(
         custom_scorer: TCustomScorer,
-        limit: usize,
+        collector: TopCollector<TScore>,
     ) -> CustomScoreTopCollector<TCustomScorer, TScore> {
         CustomScoreTopCollector {
             custom_scorer,
-            collector: TopCollector::with_limit(limit),
+            collector,
         }
     }
 }

src/collector/facet_collector.rs

@@ -1,6 +1,5 @@
 use crate::collector::Collector;
 use crate::collector::SegmentCollector;
-use crate::docset::SkipResult;
 use crate::fastfield::FacetReader;
 use crate::schema::Facet;
 use crate::schema::Field;
@@ -188,6 +187,11 @@ pub struct FacetSegmentCollector {
     collapse_facet_ords: Vec<u64>,
 }
 
+enum SkipResult {
+    Found,
+    NotFound,
+}
+
 fn skip<'a, I: Iterator<Item = &'a Facet>>(
     target: &[u8],
     collapse_it: &mut Peekable<I>,
@@ -197,14 +201,14 @@ fn skip<'a, I: Iterator<Item = &'a Facet>>(
             Some(facet_bytes) => match facet_bytes.encoded_str().as_bytes().cmp(target) {
                 Ordering::Less => {}
                 Ordering::Greater => {
-                    return SkipResult::OverStep;
+                    return SkipResult::NotFound;
                 }
                 Ordering::Equal => {
-                    return SkipResult::Reached;
+                    return SkipResult::Found;
                 }
             },
             None => {
-                return SkipResult::End;
+                return SkipResult::NotFound;
             }
         }
         collapse_it.next();
@@ -281,7 +285,7 @@ impl Collector for FacetCollector {
                     // is positionned on a term that has not been processed yet.
                     let skip_result = skip(facet_streamer.key(), &mut collapse_facet_it);
                     match skip_result {
-                        SkipResult::Reached => {
+                        SkipResult::Found => {
                             // we reach a facet we decided to collapse.
                             let collapse_depth = facet_depth(facet_streamer.key());
                             let mut collapsed_id = 0;
@@ -301,7 +305,7 @@ impl Collector for FacetCollector {
                             }
                             break;
                         }
-                        SkipResult::End | SkipResult::OverStep => {
+                        SkipResult::NotFound => {
                             collapse_mapping.push(0);
                             if !facet_streamer.advance() {
                                 break;

src/collector/mod.rs

@@ -84,7 +84,7 @@ See the `custom_collector` example.
 
 */
 
-use crate::{DocId, Searcher, Executor};
+use crate::DocId;
 use crate::Score;
 use crate::SegmentLocalId;
 use crate::SegmentReader;
@@ -100,9 +100,6 @@ mod top_collector;
 
 mod top_score_collector;
 pub use self::top_score_collector::TopDocs;
-#[cfg(test)]
-pub(crate) use self::top_score_collector::TopScoreSegmentCollector;
-
 
 mod custom_score_top_collector;
 pub use self::custom_score_top_collector::{CustomScorer, CustomSegmentScorer};
@@ -112,9 +109,7 @@ pub use self::tweak_score_top_collector::{ScoreSegmentTweaker, ScoreTweaker};
 
 mod facet_collector;
 pub use self::facet_collector::FacetCollector;
-use crate::fastfield::DeleteBitSet;
-use crate::query::{Scorer, Weight};
-use std::borrow::BorrowMut;
+use crate::query::Scorer;
 
 /// `Fruit` is the type for the result of our collection.
 /// e.g. `usize` for the `Count` collector.
@@ -122,8 +117,6 @@ pub trait Fruit: Send + downcast_rs::Downcast {}
 
 impl<T> Fruit for T where T: Send + downcast_rs::Downcast {}
 
-
-
 /// Collectors are in charge of collecting and retaining relevant
 /// information from the document found and scored by the query.
 ///
@@ -163,18 +156,26 @@ pub trait Collector: Sync {
     /// into one fruit.
     fn merge_fruits(&self, segment_fruits: Vec<Self::Fruit>) -> crate::Result<Self::Fruit>;
 
-    fn collect_weight(&self, searcher: &Searcher, weight: &dyn Weight, executor: &Executor) -> crate::Result<Self::Fruit> {
-        let segment_readers = searcher.segment_readers();
-        let fruits = executor.map(
-            |(segment_ord, segment_reader)| {
-                let mut scorer = weight.scorer(segment_reader, 1.0f32)?;
-                let segment_collector =
-                    self.for_segment(segment_ord as u32, segment_reader)?;
-                Ok(segment_collector.collect_scorer(scorer.borrow_mut(), segment_reader.delete_bitset()))
-            },
-            segment_readers.iter().enumerate(),
-        )?;
-        self.merge_fruits(fruits)
+    /// Created a segment collector and
+    fn collect_segment(
+        &self,
+        scorer: &mut dyn Scorer,
+        segment_ord: u32,
+        segment_reader: &SegmentReader,
+    ) -> crate::Result<<Self::Child as SegmentCollector>::Fruit> {
+        let mut segment_collector = self.for_segment(segment_ord as u32, segment_reader)?;
+        if let Some(delete_bitset) = segment_reader.delete_bitset() {
+            scorer.for_each(&mut |doc, score| {
+                if delete_bitset.is_alive(doc) {
+                    segment_collector.collect(doc, score);
+                }
+            });
+        } else {
+            scorer.for_each(&mut |doc, score| {
+                segment_collector.collect(doc, score);
+            })
+        }
+        Ok(segment_collector.harvest())
     }
 }
 
@@ -183,7 +184,7 @@ pub trait Collector: Sync {
 ///
 /// `.collect(doc, score)` will be called for every documents
 /// matching the query.
-pub trait SegmentCollector: 'static + Sized {
+pub trait SegmentCollector: 'static {
     /// `Fruit` is the type for the result of our collection.
     /// e.g. `usize` for the `Count` collector.
     type Fruit: Fruit;
@@ -193,19 +194,6 @@ pub trait SegmentCollector: 'static + Sized {
 
     /// Extract the fruit of the collection from the `SegmentCollector`.
     fn harvest(self) -> Self::Fruit;
-
-    fn collect_scorer(mut self, scorer: &mut dyn Scorer, delete_bitset: Option<&DeleteBitSet>) -> Self::Fruit {
-        if let Some(delete_bitset) = delete_bitset {
-            scorer.for_each(&mut |doc, score| {
-                if delete_bitset.is_alive(doc) {
-                    self.collect(doc, score);
-                }
-            });
-        } else {
-            scorer.for_each(&mut |doc, score| self.collect(doc, score));
-        }
-        self.harvest()
-    }
 }
 
 // -----------------------------------------------

src/collector/top_collector.rs

@@ -18,9 +18,9 @@ use std::collections::BinaryHeap;
 /// Two elements are equal if their feature is equal, and regardless of whether `doc`
 /// is equal. This should be perfectly fine for this usage, but let's make sure this
 /// struct is never public.
-struct ComparableDoc<T, D> {
-    feature: T,
-    doc: D,
+pub(crate) struct ComparableDoc<T, D> {
+    pub feature: T,
+    pub doc: D,
 }
 
 impl<T: PartialOrd, D: PartialOrd> PartialOrd for ComparableDoc<T, D> {
@@ -57,6 +57,7 @@ impl<T: PartialOrd, D: PartialOrd> Eq for ComparableDoc<T, D> {}
 
 pub(crate) struct TopCollector<T> {
     pub limit: usize,
+    pub offset: usize,
     _marker: PhantomData<T>,
 }
 
@@ -69,15 +70,23 @@ where
     /// # Panics
     /// The method panics if limit is 0
     pub fn with_limit(limit: usize) -> TopCollector<T> {
-        assert!(limit > 0, "Limit must be strictly greater than 0.");
-        TopCollector {
+        if limit < 1 {
+            panic!("Limit must be strictly greater than 0.");
+        }
+        Self {
             limit,
+            offset: 0,
             _marker: PhantomData,
         }
     }
 
-    pub fn limit(&self) -> usize {
-        self.limit
+    /// Skip the first "offset" documents when collecting.
+    ///
+    /// This is equivalent to `OFFSET` in MySQL or PostgreSQL and `start` in
+    /// Lucene's TopDocsCollector.
+    pub fn and_offset(mut self, offset: usize) -> TopCollector<T> {
+        self.offset = offset;
+        self
     }
 
     pub fn merge_fruits(
@@ -90,7 +99,7 @@ where
         let mut top_collector = BinaryHeap::new();
         for child_fruit in children {
             for (feature, doc) in child_fruit {
-                if top_collector.len() < self.limit {
+                if top_collector.len() < (self.limit + self.offset) {
                     top_collector.push(ComparableDoc { feature, doc });
                 } else if let Some(mut head) = top_collector.peek_mut() {
                     if head.feature < feature {
@@ -102,6 +111,7 @@ where
         Ok(top_collector
             .into_sorted_vec()
             .into_iter()
+            .skip(self.offset)
             .map(|cdoc| (cdoc.feature, cdoc.doc))
             .collect())
     }
@@ -111,7 +121,23 @@ where
         segment_id: SegmentLocalId,
         _: &SegmentReader,
     ) -> crate::Result<TopSegmentCollector<F>> {
-        Ok(TopSegmentCollector::new(segment_id, self.limit))
+        Ok(TopSegmentCollector::new(
+            segment_id,
+            self.limit + self.offset,
+        ))
+    }
+
+    /// Create a new TopCollector with the same limit and offset.
+    ///
+    /// Ideally we would use Into but the blanket implementation seems to cause the Scorer traits
+    /// to fail.
+    #[doc(hidden)]
+    pub(crate) fn into_tscore<TScore: PartialOrd + Clone>(self) -> TopCollector<TScore> {
+        TopCollector {
+            limit: self.limit,
+            offset: self.offset,
+            _marker: PhantomData,
+        }
     }
 }
 
@@ -122,13 +148,13 @@ where
 /// The theorical complexity for collecting the top `K` out of `n` documents
 /// is `O(n log K)`.
 pub(crate) struct TopSegmentCollector<T> {
-    pub limit: usize,
+    limit: usize,
     heap: BinaryHeap<ComparableDoc<T, DocId>>,
     segment_id: u32,
 }
 
 impl<T: PartialOrd> TopSegmentCollector<T> {
-    pub fn new(segment_id: SegmentLocalId, limit: usize) -> TopSegmentCollector<T> {
+    fn new(segment_id: SegmentLocalId, limit: usize) -> TopSegmentCollector<T> {
         TopSegmentCollector {
             limit,
             heap: BinaryHeap::with_capacity(limit),
@@ -159,10 +185,6 @@ impl<T: PartialOrd + Clone> TopSegmentCollector<T> {
         self.heap.len() >= self.limit
     }
 
-    pub fn pruning_score(&self) -> Option<T> {
-        self.heap.peek().map(|head| head.feature.clone())
-    }
-
     /// Collects a document scored by the given feature
     ///
     /// It collects documents until it has reached the max capacity. Once it reaches capacity, it
@@ -189,7 +211,7 @@ impl<T: PartialOrd + Clone> TopSegmentCollector<T> {
 
 #[cfg(test)]
 mod tests {
-    use super::TopSegmentCollector;
+    use super::{TopCollector, TopSegmentCollector};
     use crate::DocAddress;
 
     #[test]
@@ -250,6 +272,48 @@ mod tests {
             top_collector_limit_3.harvest()[..2].to_vec(),
         );
     }
+
+    #[test]
+    fn test_top_collector_with_limit_and_offset() {
+        let collector = TopCollector::with_limit(2).and_offset(1);
+
+        let results = collector
+            .merge_fruits(vec![vec![
+                (0.9, DocAddress(0, 1)),
+                (0.8, DocAddress(0, 2)),
+                (0.7, DocAddress(0, 3)),
+                (0.6, DocAddress(0, 4)),
+                (0.5, DocAddress(0, 5)),
+            ]])
+            .unwrap();
+
+        assert_eq!(
+            results,
+            vec![(0.8, DocAddress(0, 2)), (0.7, DocAddress(0, 3)),]
+        );
+    }
+
+    #[test]
+    fn test_top_collector_with_limit_larger_than_set_and_offset() {
+        let collector = TopCollector::with_limit(2).and_offset(1);
+
+        let results = collector
+            .merge_fruits(vec![vec![(0.9, DocAddress(0, 1)), (0.8, DocAddress(0, 2))]])
+            .unwrap();
+
+        assert_eq!(results, vec![(0.8, DocAddress(0, 2)),]);
+    }
+
+    #[test]
+    fn test_top_collector_with_limit_and_offset_larger_than_set() {
+        let collector = TopCollector::with_limit(2).and_offset(20);
+
+        let results = collector
+            .merge_fruits(vec![vec![(0.9, DocAddress(0, 1)), (0.8, DocAddress(0, 2))]])
+            .unwrap();
+
+        assert_eq!(results, vec![]);
+    }
 }
 
 #[cfg(all(test, feature = "unstable"))]

src/collector/top_score_collector.rs

@@ -1,20 +1,22 @@
 use super::Collector;
 use crate::collector::custom_score_top_collector::CustomScoreTopCollector;
-use crate::collector::top_collector::TopCollector;
 use crate::collector::top_collector::TopSegmentCollector;
+use crate::collector::top_collector::{ComparableDoc, TopCollector};
 use crate::collector::tweak_score_top_collector::TweakedScoreTopCollector;
 use crate::collector::{
     CustomScorer, CustomSegmentScorer, ScoreSegmentTweaker, ScoreTweaker, SegmentCollector,
 };
+use crate::docset::TERMINATED;
 use crate::fastfield::FastFieldReader;
+use crate::query::Scorer;
 use crate::schema::Field;
-use crate::{DocAddress, Executor, Searcher};
+use crate::DocAddress;
 use crate::DocId;
 use crate::Score;
 use crate::SegmentLocalId;
 use crate::SegmentReader;
+use std::collections::BinaryHeap;
 use std::fmt;
-use crate::query::{Weight, PruningScorerIfPossible};
 
 /// The `TopDocs` collector keeps track of the top `K` documents
 /// sorted by their score.
@@ -58,7 +60,11 @@ pub struct TopDocs(TopCollector<Score>);
 
 impl fmt::Debug for TopDocs {
     fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
-        write!(f, "TopDocs({})", self.0.limit())
+        write!(
+            f,
+            "TopDocs(limit={}, offset={})",
+            self.0.limit, self.0.offset
+        )
     }
 }
 
@@ -102,6 +108,45 @@ impl TopDocs {
         TopDocs(TopCollector::with_limit(limit))
     }
 
+    /// Skip the first "offset" documents when collecting.
+    ///
+    /// This is equivalent to `OFFSET` in MySQL or PostgreSQL and `start` in
+    /// Lucene's TopDocsCollector.
+    ///
+    /// ```rust
+    /// use tantivy::collector::TopDocs;
+    /// use tantivy::query::QueryParser;
+    /// use tantivy::schema::{Schema, TEXT};
+    /// use tantivy::{doc, DocAddress, Index};
+    ///
+    /// let mut schema_builder = Schema::builder();
+    /// let title = schema_builder.add_text_field("title", TEXT);
+    /// let schema = schema_builder.build();
+    /// let index = Index::create_in_ram(schema);
+    ///
+    /// let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
+    /// index_writer.add_document(doc!(title => "The Name of the Wind"));
+    /// index_writer.add_document(doc!(title => "The Diary of Muadib"));
+    /// index_writer.add_document(doc!(title => "A Dairy Cow"));
+    /// index_writer.add_document(doc!(title => "The Diary of a Young Girl"));
+    /// index_writer.add_document(doc!(title => "The Diary of Lena Mukhina"));
+    /// assert!(index_writer.commit().is_ok());
+    ///
+    /// let reader = index.reader().unwrap();
+    /// let searcher = reader.searcher();
+    ///
+    /// let query_parser = QueryParser::for_index(&index, vec![title]);
+    /// let query = query_parser.parse_query("diary").unwrap();
+    /// let top_docs = searcher.search(&query, &TopDocs::with_limit(2).and_offset(1)).unwrap();
+    ///
+    /// assert_eq!(top_docs.len(), 2);
+    /// assert_eq!(&top_docs[0], &(0.5204813, DocAddress(0, 4)));
+    /// assert_eq!(&top_docs[1], &(0.4793185, DocAddress(0, 3)));
+    /// ```
+    pub fn and_offset(self, offset: usize) -> TopDocs {
+        TopDocs(self.0.and_offset(offset))
+    }
+
     /// Set top-K to rank documents by a given fast field.
     ///
     /// ```rust
@@ -282,7 +327,7 @@ impl TopDocs {
         TScoreSegmentTweaker: ScoreSegmentTweaker<TScore> + 'static,
         TScoreTweaker: ScoreTweaker<TScore, Child = TScoreSegmentTweaker>,
     {
-        TweakedScoreTopCollector::new(score_tweaker, self.0.limit())
+        TweakedScoreTopCollector::new(score_tweaker, self.0.into_tscore())
     }
 
     /// Ranks the documents using a custom score.
@@ -396,7 +441,7 @@ impl TopDocs {
         TCustomSegmentScorer: CustomSegmentScorer<TScore> + 'static,
         TCustomScorer: CustomScorer<TScore, Child = TCustomSegmentScorer>,
     {
-        CustomScoreTopCollector::new(custom_score, self.0.limit())
+        CustomScoreTopCollector::new(custom_score, self.0.into_tscore())
     }
 }
 
@@ -418,64 +463,74 @@ impl Collector for TopDocs {
         true
     }
 
-    fn collect_weight(&self, searcher: &Searcher, weight: &dyn Weight, executor: &Executor) -> crate::Result<Self::Fruit> {
-        let segment_readers = searcher.segment_readers();
-        let fruits = executor.map(
-            |(segment_ord, segment_reader)| {
-                match weight.pruning_scorer(segment_reader, 1.0f32)? {
-                    PruningScorerIfPossible::NonPruning(mut scorer) => {
-                        let segment_collector =
-                            self.for_segment(segment_ord as u32, segment_reader)?;
-                        let fruit =
-                            segment_collector.collect_scorer(scorer.as_mut(), segment_reader.delete_bitset());
-                        Ok(fruit)
-                    }
-                    PruningScorerIfPossible::Pruning(mut pruning_scorer) => {
-                        let limit = self.0.limit;
-                        let mut segment_collector =
-                            self.for_segment(segment_ord as u32, segment_reader)?;
-                        for _ in 0..limit {
-                            if !pruning_scorer.advance() {
-                                return Ok(segment_collector.harvest());
-                            }
-                            segment_collector.collect(pruning_scorer.doc(), pruning_scorer.score());
-                        }
-                        let mut pruning_score = segment_collector.0.pruning_score().unwrap_or(0.0f32);
-                        while pruning_scorer.advance_with_pruning(pruning_score) {
-                            segment_collector.0.collect(pruning_scorer.doc(), pruning_scorer.score());
-                            pruning_score = segment_collector.0.pruning_score().unwrap_or(0.0f32);
-                        }
-                        Ok(segment_collector.harvest())
-                    }
-                }
-            },
-            segment_readers.iter().enumerate(),
-        )?;
-        self.merge_fruits(fruits)
-    }
-
     fn merge_fruits(
         &self,
         child_fruits: Vec<Vec<(Score, DocAddress)>>,
     ) -> crate::Result<Self::Fruit> {
         self.0.merge_fruits(child_fruits)
     }
+
+    fn collect_segment(
+        &self,
+        scorer: &mut dyn Scorer,
+        segment_ord: u32,
+        segment_reader: &SegmentReader,
+    ) -> crate::Result<<Self::Child as SegmentCollector>::Fruit> {
+        let mut heap: BinaryHeap<ComparableDoc<Score, DocId>> =
+            BinaryHeap::with_capacity(self.0.limit + self.0.offset);
+        // first we fill the heap with the first `limit` elements.
+        let mut doc = scorer.doc();
+        while doc != TERMINATED && heap.len() < (self.0.limit + self.0.offset) {
+            if !segment_reader.is_deleted(doc) {
+                let score = scorer.score();
+                heap.push(ComparableDoc {
+                    feature: score,
+                    doc,
+                });
+            }
+            doc = scorer.advance();
+        }
+
+        let threshold = heap.peek().map(|el| el.feature).unwrap_or(std::f32::MIN);
+
+        if let Some(delete_bitset) = segment_reader.delete_bitset() {
+            scorer.for_each_pruning(threshold, &mut |doc, score| {
+                if delete_bitset.is_alive(doc) {
+                    *heap.peek_mut().unwrap() = ComparableDoc {
+                        feature: score,
+                        doc,
+                    };
+                }
+                heap.peek().map(|el| el.feature).unwrap_or(std::f32::MIN)
+            });
+        } else {
+            scorer.for_each_pruning(threshold, &mut |doc, score| {
+                *heap.peek_mut().unwrap() = ComparableDoc {
+                    feature: score,
+                    doc,
+                };
+                heap.peek().map(|el| el.feature).unwrap_or(std::f32::MIN)
+            });
+        }
+
+        let fruit = heap
+            .into_sorted_vec()
+            .into_iter()
+            .map(|cid| (cid.feature, DocAddress(segment_ord, cid.doc)))
+            .collect();
+
+        Ok(fruit)
+    }
 }
 
 /// Segment Collector associated to `TopDocs`.
 pub struct TopScoreSegmentCollector(TopSegmentCollector<Score>);
 
-impl TopScoreSegmentCollector {
-    pub fn new(segment_id: SegmentLocalId, limit: usize) -> Self {
-        TopScoreSegmentCollector(TopSegmentCollector::new(segment_id, limit))
-    }
-}
-
 impl SegmentCollector for TopScoreSegmentCollector {
     type Fruit = Vec<(Score, DocAddress)>;
 
     fn collect(&mut self, doc: DocId, score: Score) {
-        self.0.collect(doc, score)
+        self.0.collect(doc, score);
     }
 
     fn harvest(self) -> Vec<(Score, DocAddress)> {
@@ -489,10 +544,10 @@ mod tests {
     use crate::collector::Collector;
     use crate::query::{AllQuery, Query, QueryParser};
     use crate::schema::{Field, Schema, FAST, STORED, TEXT};
-    use crate::DocAddress;
     use crate::Index;
     use crate::IndexWriter;
     use crate::Score;
+    use crate::{DocAddress, DocId, SegmentReader};
 
     fn make_index() -> Index {
         let mut schema_builder = Schema::builder();
@@ -532,6 +587,21 @@ mod tests {
         );
     }
 
+    #[test]
+    fn test_top_collector_not_at_capacity_with_offset() {
+        let index = make_index();
+        let field = index.schema().get_field("text").unwrap();
+        let query_parser = QueryParser::for_index(&index, vec![field]);
+        let text_query = query_parser.parse_query("droopy tax").unwrap();
+        let score_docs: Vec<(Score, DocAddress)> = index
+            .reader()
+            .unwrap()
+            .searcher()
+            .search(&text_query, &TopDocs::with_limit(4).and_offset(2))
+            .unwrap();
+        assert_eq!(score_docs, vec![(0.48527452, DocAddress(0, 0))]);
+    }
+
     #[test]
     fn test_top_collector_at_capacity() {
         let index = make_index();
@@ -553,6 +623,27 @@ mod tests {
         );
     }
 
+    #[test]
+    fn test_top_collector_at_capacity_with_offset() {
+        let index = make_index();
+        let field = index.schema().get_field("text").unwrap();
+        let query_parser = QueryParser::for_index(&index, vec![field]);
+        let text_query = query_parser.parse_query("droopy tax").unwrap();
+        let score_docs: Vec<(Score, DocAddress)> = index
+            .reader()
+            .unwrap()
+            .searcher()
+            .search(&text_query, &TopDocs::with_limit(2).and_offset(1))
+            .unwrap();
+        assert_eq!(
+            score_docs,
+            vec![
+                (0.5376842, DocAddress(0u32, 2)),
+                (0.48527452, DocAddress(0, 0))
+            ]
+        );
+    }
+
     #[test]
     fn test_top_collector_stable_sorting() {
         let index = make_index();
@@ -666,6 +757,50 @@ mod tests {
         }
     }
 
+    #[test]
+    fn test_tweak_score_top_collector_with_offset() {
+        let index = make_index();
+        let field = index.schema().get_field("text").unwrap();
+        let query_parser = QueryParser::for_index(&index, vec![field]);
+        let text_query = query_parser.parse_query("droopy tax").unwrap();
+        let collector = TopDocs::with_limit(2).and_offset(1).tweak_score(
+            move |_segment_reader: &SegmentReader| move |doc: DocId, _original_score: Score| doc,
+        );
+        let score_docs: Vec<(u32, DocAddress)> = index
+            .reader()
+            .unwrap()
+            .searcher()
+            .search(&text_query, &collector)
+            .unwrap();
+
+        assert_eq!(
+            score_docs,
+            vec![(1, DocAddress(0, 1)), (0, DocAddress(0, 0)),]
+        );
+    }
+
+    #[test]
+    fn test_custom_score_top_collector_with_offset() {
+        let index = make_index();
+        let field = index.schema().get_field("text").unwrap();
+        let query_parser = QueryParser::for_index(&index, vec![field]);
+        let text_query = query_parser.parse_query("droopy tax").unwrap();
+        let collector = TopDocs::with_limit(2)
+            .and_offset(1)
+            .custom_score(move |_segment_reader: &SegmentReader| move |doc: DocId| doc);
+        let score_docs: Vec<(u32, DocAddress)> = index
+            .reader()
+            .unwrap()
+            .searcher()
+            .search(&text_query, &collector)
+            .unwrap();
+
+        assert_eq!(
+            score_docs,
+            vec![(1, DocAddress(0, 1)), (0, DocAddress(0, 0)),]
+        );
+    }
+
     fn index(
         query: &str,
         query_field: Field,

src/collector/tweak_score_top_collector.rs

@@ -14,11 +14,11 @@ where
 {
     pub fn new(
         score_tweaker: TScoreTweaker,
-        limit: usize,
+        collector: TopCollector<TScore>,
     ) -> TweakedScoreTopCollector<TScoreTweaker, TScore> {
         TweakedScoreTopCollector {
             score_tweaker,
-            collector: TopCollector::with_limit(limit),
+            collector,
         }
     }
 }

src/common/bitset.rs

@@ -33,6 +33,10 @@ impl TinySet {
         TinySet(0u64)
     }
 
+    pub fn clear(&mut self) {
+        self.0 = 0u64;
+    }
+
     /// Returns the complement of the set in `[0, 64[`.
     fn complement(self) -> TinySet {
         TinySet(!self.0)
@@ -43,6 +47,11 @@ impl TinySet {
         !self.intersect(TinySet::singleton(el)).is_empty()
     }
 
+    /// Returns the number of elements in the TinySet.
+    pub fn len(self) -> u32 {
+        self.0.count_ones()
+    }
+
     /// Returns the intersection of `self` and `other`
     pub fn intersect(self, other: TinySet) -> TinySet {
         TinySet(self.0 & other.0)
@@ -109,22 +118,12 @@ impl TinySet {
     pub fn range_greater_or_equal(from_included: u32) -> TinySet {
         TinySet::range_lower(from_included).complement()
     }
-
-    pub fn clear(&mut self) {
-        self.0 = 0u64;
-    }
-
-    pub fn len(self) -> u32 {
-        self.0.count_ones()
-    }
 }
 
 #[derive(Clone)]
 pub struct BitSet {
     tinysets: Box<[TinySet]>,
-    len: usize, //< Technically it should be u32, but we
-    // count multiple inserts.
-    // `usize` guards us from overflow.
+    len: usize,
     max_value: u32,
 }
 
@@ -204,7 +203,7 @@ mod tests {
 
     use super::BitSet;
     use super::TinySet;
-    use crate::docset::DocSet;
+    use crate::docset::{DocSet, TERMINATED};
     use crate::query::BitSetDocSet;
     use crate::tests;
     use crate::tests::generate_nonunique_unsorted;
@@ -278,11 +277,13 @@ mod tests {
         }
         assert_eq!(btreeset.len(), bitset.len());
         let mut bitset_docset = BitSetDocSet::from(bitset);
+        let mut remaining = true;
         for el in btreeset.into_iter() {
-            bitset_docset.advance();
+            assert!(remaining);
             assert_eq!(bitset_docset.doc(), el);
+            remaining = bitset_docset.advance() != TERMINATED;
         }
-        assert!(!bitset_docset.advance());
+        assert!(!remaining);
     }
 
     #[test]

src/core/searcher.rs

@@ -1,10 +1,8 @@
 use crate::collector::Collector;
-use crate::collector::SegmentCollector;
 use crate::core::Executor;
 use crate::core::InvertedIndexReader;
 use crate::core::SegmentReader;
 use crate::query::Query;
-use crate::query::Weight;
 use crate::schema::Document;
 use crate::schema::Schema;
 use crate::schema::{Field, Term};
@@ -16,18 +14,6 @@ use crate::Index;
 use std::fmt;
 use std::sync::Arc;
 
-fn collect_segment<C: Collector>(
-    collector: &C,
-    weight: &dyn Weight,
-    segment_ord: u32,
-    segment_reader: &SegmentReader,
-) -> crate::Result<C::Fruit> {
-    let mut scorer = weight.scorer(segment_reader, 1.0f32)?;
-    let segment_collector =
-        collector.for_segment(segment_ord as u32, segment_reader)?;
-    Ok(segment_collector.collect_scorer(&mut scorer, segment_reader.delete_bitset()))
-}
-
 /// Holds a list of `SegmentReader`s ready for search.
 ///
 /// It guarantees that the `Segment` will not be removed before
@@ -154,12 +140,8 @@ impl Searcher {
         let segment_readers = self.segment_readers();
         let fruits = executor.map(
             |(segment_ord, segment_reader)| {
-                collect_segment(
-                    collector,
-                    weight.as_ref(),
-                    segment_ord as u32,
-                    segment_reader,
-                )
+                let mut scorer = weight.scorer(segment_reader, 1.0f32)?;
+                collector.collect_segment(scorer.as_mut(), segment_ord as u32, segment_reader)
             },
             segment_readers.iter().enumerate(),
         )?;

src/docset.rs

@@ -1,58 +1,47 @@
-use crate::common::BitSet;
 use crate::fastfield::DeleteBitSet;
 use crate::DocId;
 use std::borrow::Borrow;
 use std::borrow::BorrowMut;
-use std::cmp::Ordering;
 
-/// Expresses the outcome of a call to `DocSet`'s `.skip_next(...)`.
-#[derive(PartialEq, Eq, Debug)]
-pub enum SkipResult {
-    /// target was in the docset
-    Reached,
-    /// target was not in the docset, skipping stopped as a greater element was found
-    OverStep,
-    /// the docset was entirely consumed without finding the target, nor any
-    /// element greater than the target.
-    End,
-}
+/// Sentinel value returned when a DocSet has been entirely consumed.
+///
+/// This is not u32::MAX as one would have expected, due to the lack of SSE2 instructions
+/// to compare [u32; 4].
+pub const TERMINATED: DocId = std::i32::MAX as u32;
 
 /// Represents an iterable set of sorted doc ids.
 pub trait DocSet {
     /// Goes to the next element.
-    /// `.advance(...)` needs to be called a first time to point to the correct
-    /// element.
-    fn advance(&mut self) -> bool;
+    ///
+    /// The DocId of the next element is returned.
+    /// In other words we should always have :
+    /// ```ignore
+    /// let doc = docset.advance();
+    /// assert_eq!(doc, docset.doc());
+    /// ```
+    ///
+    /// If we reached the end of the DocSet, TERMINATED should be returned.
+    ///
+    /// Calling `.advance()` on a terminated DocSet should be supported, and TERMINATED should
+    /// be returned.
+    /// TODO Test existing docsets.
+    fn advance(&mut self) -> DocId;
 
-    /// After skipping, position the iterator in such a way that `.doc()`
-    /// will return a value greater than or equal to target.
+    /// Advances the DocSet forward until reaching the target, or going to the
+    /// lowest DocId greater than the target.
     ///
-    /// SkipResult expresses whether the `target value` was reached, overstepped,
-    /// or if the `DocSet` was entirely consumed without finding any value
-    /// greater or equal to the `target`.
+    /// If the end of the DocSet is reached, TERMINATED is returned.
     ///
-    /// WARNING: Calling skip always advances the docset.
-    /// More specifically, if the docset is already positionned on the target
-    /// skipping will advance to the next position and return SkipResult::Overstep.
+    /// Calling `.seek(target)` on a terminated DocSet is legal. Implementation
+    /// of DocSet should support it.
     ///
-    /// If `.skip_next()` oversteps, then the docset must be positionned correctly
-    /// on an existing document. In other words, `.doc()` should return the first document
-    /// greater than `DocId`.
-    fn skip_next(&mut self, target: DocId) -> SkipResult {
-        if !self.advance() {
-            return SkipResult::End;
-        }
-        loop {
-            match self.doc().cmp(&target) {
-                Ordering::Less => {
-                    if !self.advance() {
-                        return SkipResult::End;
-                    }
-                }
-                Ordering::Equal => return SkipResult::Reached,
-                Ordering::Greater => return SkipResult::OverStep,
-            }
+    /// Calling `seek(TERMINATED)` is also legal and is the normal way to consume a DocSet.
+    fn seek(&mut self, target: DocId) -> DocId {
+        let mut doc = self.doc();
+        while doc < target {
+            doc = self.advance();
         }
+        doc
     }
 
     /// Fills a given mutable buffer with the next doc ids from the
@@ -71,38 +60,38 @@ pub trait DocSet {
     /// use case where batching. The normal way to
     /// go through the `DocId`'s is to call `.advance()`.
     fn fill_buffer(&mut self, buffer: &mut [DocId]) -> usize {
+        if self.doc() == TERMINATED {
+            return 0;
+        }
         for (i, buffer_val) in buffer.iter_mut().enumerate() {
-            if self.advance() {
-                *buffer_val = self.doc();
-            } else {
-                return i;
+            *buffer_val = self.doc();
+            if self.advance() == TERMINATED {
+                return i + 1;
             }
         }
         buffer.len()
     }
 
     /// Returns the current document
+    /// Right after creating a new DocSet, the docset points to the first document.
+    ///
+    /// If the DocSet is empty, .doc() should return `TERMINATED`.
     fn doc(&self) -> DocId;
 
     /// Returns a best-effort hint of the
     /// length of the docset.
     fn size_hint(&self) -> u32;
 
-    /// Appends all docs to a `bitset`.
-    fn append_to_bitset(&mut self, bitset: &mut BitSet) {
-        while self.advance() {
-            bitset.insert(self.doc());
-        }
-    }
-
     /// Returns the number documents matching.
     /// Calling this method consumes the `DocSet`.
     fn count(&mut self, delete_bitset: &DeleteBitSet) -> u32 {
         let mut count = 0u32;
-        while self.advance() {
-            if !delete_bitset.is_deleted(self.doc()) {
+        let mut doc = self.doc();
+        while doc != TERMINATED {
+            if !delete_bitset.is_deleted(doc) {
                 count += 1u32;
             }
+            doc = self.advance();
         }
         count
     }
@@ -114,22 +103,42 @@ pub trait DocSet {
     /// given by `count()`.
     fn count_including_deleted(&mut self) -> u32 {
         let mut count = 0u32;
-        while self.advance() {
+        let mut doc = self.doc();
+        while doc != TERMINATED {
             count += 1u32;
+            doc = self.advance();
         }
         count
     }
 }
 
+impl<'a> DocSet for &'a mut dyn DocSet {
+    fn advance(&mut self) -> u32 {
+        (**self).advance()
+    }
+
+    fn seek(&mut self, target: DocId) -> DocId {
+        (**self).seek(target)
+    }
+
+    fn doc(&self) -> u32 {
+        (**self).doc()
+    }
+
+    fn size_hint(&self) -> u32 {
+        (**self).size_hint()
+    }
+}
+
 impl<TDocSet: DocSet + ?Sized> DocSet for Box<TDocSet> {
-    fn advance(&mut self) -> bool {
+    fn advance(&mut self) -> DocId {
         let unboxed: &mut TDocSet = self.borrow_mut();
         unboxed.advance()
     }
 
-    fn skip_next(&mut self, target: DocId) -> SkipResult {
+    fn seek(&mut self, target: DocId) -> DocId {
         let unboxed: &mut TDocSet = self.borrow_mut();
-        unboxed.skip_next(target)
+        unboxed.seek(target)
     }
 
     fn doc(&self) -> DocId {
@@ -151,9 +160,4 @@ impl<TDocSet: DocSet + ?Sized> DocSet for Box<TDocSet> {
         let unboxed: &mut TDocSet = self.borrow_mut();
         unboxed.count_including_deleted()
     }
-
-    fn append_to_bitset(&mut self, bitset: &mut BitSet) {
-        let unboxed: &mut TDocSet = self.borrow_mut();
-        unboxed.append_to_bitset(bitset);
-    }
 }

src/indexer/index_writer.rs

@@ -10,7 +10,7 @@ use crate::core::SegmentMeta;
 use crate::core::SegmentReader;
 use crate::directory::TerminatingWrite;
 use crate::directory::{DirectoryLock, GarbageCollectionResult};
-use crate::docset::DocSet;
+use crate::docset::{DocSet, TERMINATED};
 use crate::error::TantivyError;
 use crate::fastfield::write_delete_bitset;
 use crate::indexer::delete_queue::{DeleteCursor, DeleteQueue};
@@ -112,15 +112,15 @@ fn compute_deleted_bitset(
         if let Some(mut docset) =
             inverted_index.read_postings(&delete_op.term, IndexRecordOption::Basic)
         {
-            while docset.advance() {
-                let deleted_doc = docset.doc();
+            let mut deleted_doc = docset.doc();
+            while deleted_doc != TERMINATED {
                 if deleted_doc < limit_doc {
                     delete_bitset.insert(deleted_doc);
                     might_have_changed = true;
                 }
+                deleted_doc = docset.advance();
             }
         }
-
         delete_cursor.advance();
     }
     Ok(might_have_changed)

src/indexer/merger.rs

@@ -2,7 +2,7 @@ use crate::common::MAX_DOC_LIMIT;
 use crate::core::Segment;
 use crate::core::SegmentReader;
 use crate::core::SerializableSegment;
-use crate::docset::DocSet;
+use crate::docset::{DocSet, TERMINATED};
 use crate::fastfield::BytesFastFieldReader;
 use crate::fastfield::DeleteBitSet;
 use crate::fastfield::FastFieldReader;
@@ -574,10 +574,12 @@ impl IndexMerger {
                     let inverted_index = segment_reader.inverted_index(indexed_field);
                     let mut segment_postings = inverted_index
                         .read_postings_from_terminfo(term_info, segment_postings_option);
-                    while segment_postings.advance() {
-                        if !segment_reader.is_deleted(segment_postings.doc()) {
+                    let mut doc = segment_postings.doc();
+                    while doc != TERMINATED {
+                        if !segment_reader.is_deleted(doc) {
                             return Some((segment_ord, segment_postings));
                         }
+                        doc = segment_postings.advance();
                     }
                     None
                 })
@@ -604,17 +606,9 @@ impl IndexMerger {
                 // postings serializer.
                 for (segment_ord, mut segment_postings) in segment_postings {
                     let old_to_new_doc_id = &merged_doc_id_map[segment_ord];
-                    loop {
-                        let doc = segment_postings.doc();
-
-                        // `.advance()` has been called once before the loop.
-                        //
-                        // It was required to make sure we only consider segments
-                        // that effectively contain at least one non-deleted document
-                        // and remove terms that do not have documents associated.
-                        //
-                        //  For this reason, we cannot use a `while segment_postings.advance()` loop.
 
+                    let mut doc = segment_postings.doc();
+                    while doc != TERMINATED {
                         // deleted doc are skipped as they do not have a `remapped_doc_id`.
                         if let Some(remapped_doc_id) = old_to_new_doc_id[doc as usize] {
                             // we make sure to only write the term iff
@@ -629,9 +623,8 @@ impl IndexMerger {
                                 delta_positions,
                             )?;
                         }
-                        if !segment_postings.advance() {
-                            break;
-                        }
+
+                        doc = segment_postings.advance();
                     }
                 }
 

src/lib.rs

@@ -156,7 +156,7 @@ mod snippet;
 pub use self::snippet::{Snippet, SnippetGenerator};
 
 mod docset;
-pub use self::docset::{DocSet, SkipResult};
+pub use self::docset::{DocSet, TERMINATED};
 pub use crate::common::{f64_to_u64, i64_to_u64, u64_to_f64, u64_to_i64};
 pub use crate::core::{Executor, SegmentComponent};
 pub use crate::core::{Index, IndexMeta, Searcher, Segment, SegmentId, SegmentMeta};
@@ -285,7 +285,7 @@ mod tests {
 
     use crate::collector::tests::TEST_COLLECTOR_WITH_SCORE;
     use crate::core::SegmentReader;
-    use crate::docset::DocSet;
+    use crate::docset::{DocSet, TERMINATED};
     use crate::query::BooleanQuery;
     use crate::schema::*;
     use crate::DocAddress;
@@ -381,19 +381,12 @@ mod tests {
             index_writer.commit().unwrap();
         }
         {
-            {
-                let doc = doc!(text_field=>"a");
-                index_writer.add_document(doc);
-            }
-            {
-                let doc = doc!(text_field=>"a a");
-                index_writer.add_document(doc);
-            }
+            index_writer.add_document(doc!(text_field=>"a"));
+            index_writer.add_document(doc!(text_field=>"a a"));
             index_writer.commit().unwrap();
         }
         {
-            let doc = doc!(text_field=>"c");
-            index_writer.add_document(doc);
+            index_writer.add_document(doc!(text_field=>"c"));
             index_writer.commit().unwrap();
         }
         {
@@ -472,10 +465,12 @@ mod tests {
     }
 
     fn advance_undeleted(docset: &mut dyn DocSet, reader: &SegmentReader) -> bool {
-        while docset.advance() {
-            if !reader.is_deleted(docset.doc()) {
+        let mut doc = docset.advance();
+        while doc != TERMINATED {
+            if !reader.is_deleted(doc) {
                 return true;
             }
+            doc = docset.advance();
         }
         false
     }
@@ -641,9 +636,8 @@ mod tests {
             .inverted_index(term.field())
             .read_postings(&term, IndexRecordOption::Basic)
             .unwrap();
-        assert!(postings.advance());
         assert_eq!(postings.doc(), 0);
-        assert!(!postings.advance());
+        assert_eq!(postings.advance(), TERMINATED);
     }
 
     #[test]
@@ -665,9 +659,8 @@ mod tests {
             .inverted_index(term.field())
             .read_postings(&term, IndexRecordOption::Basic)
             .unwrap();
-        assert!(postings.advance());
         assert_eq!(postings.doc(), 0);
-        assert!(!postings.advance());
+        assert_eq!(postings.advance(), TERMINATED);
     }
 
     #[test]
@@ -689,9 +682,8 @@ mod tests {
             .inverted_index(term.field())
             .read_postings(&term, IndexRecordOption::Basic)
             .unwrap();
-        assert!(postings.advance());
         assert_eq!(postings.doc(), 0);
-        assert!(!postings.advance());
+        assert_eq!(postings.advance(), TERMINATED);
     }
 
     #[test]
@@ -760,10 +752,8 @@ mod tests {
         {
             // writing the segment
             let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
-            {
-                let doc = doc!(text_field=>"af af af bc bc");
-                index_writer.add_document(doc);
-            }
+            let doc = doc!(text_field=>"af af af bc bc");
+            index_writer.add_document(doc);
             index_writer.commit().unwrap();
         }
         {
@@ -779,10 +769,9 @@ mod tests {
             let mut postings = inverted_index
                 .read_postings(&term_af, IndexRecordOption::WithFreqsAndPositions)
                 .unwrap();
-            assert!(postings.advance());
             assert_eq!(postings.doc(), 0);
             assert_eq!(postings.term_freq(), 3);
-            assert!(!postings.advance());
+            assert_eq!(postings.advance(), TERMINATED);
         }
     }
 

src/postings/block_max_postings.rs

@@ -1,16 +0,0 @@
-use crate::postings::Postings;
-use crate::DocId;
-
-/// Inverted list with additional information about the maximum term frequency
-/// within a block, as well as globally within the list.
-pub trait BlockMaxPostings: Postings {
-    /// Returns the maximum frequency in the entire list.
-    fn max_term_freq(&self) -> u32;
-    /// Returns the maximum frequency in the current block.
-    fn block_max_term_freq(&mut self) -> u32;
-    /// Returns the document with the largest frequency.
-    fn max_doc(&self) -> DocId;
-    /// Returns the document with the largest frequency within the current
-    /// block.
-    fn block_max_doc(&self) -> DocId;
-}

src/postings/block_max_segment_postings.rs

@@ -1,76 +0,0 @@
-use crate::postings::{BlockMaxPostings, Postings, SegmentPostings};
-use crate::{DocId, DocSet, SkipResult};
-
-/// A wrapper over [`SegmentPostings`](./struct.SegmentPostings.html)
-/// with max block frequencies.
-pub struct BlockMaxSegmentPostings {
-    postings: SegmentPostings,
-    max_blocks: SegmentPostings,
-    doc_with_max_term_freq: DocId,
-    max_term_freq: u32,
-}
-
-impl BlockMaxSegmentPostings {
-    /// Constructs a new segment postings with block-max information.
-    pub fn new(
-        postings: SegmentPostings,
-        max_blocks: SegmentPostings,
-        doc_with_max_term_freq: DocId,
-        max_term_freq: u32,
-    ) -> Self {
-        Self {
-            postings,
-            max_blocks,
-            doc_with_max_term_freq,
-            max_term_freq,
-        }
-    }
-}
-
-impl DocSet for BlockMaxSegmentPostings {
-    fn advance(&mut self) -> bool {
-        self.postings.advance()
-    }
-
-    fn doc(&self) -> DocId {
-        self.postings.doc()
-    }
-
-    fn size_hint(&self) -> u32 {
-        self.postings.size_hint()
-    }
-
-    fn skip_next(&mut self, target: DocId) -> SkipResult {
-        self.postings.skip_next(target)
-    }
-}
-
-impl Postings for BlockMaxSegmentPostings {
-    fn term_freq(&self) -> u32 {
-        self.postings.term_freq()
-    }
-    fn positions_with_offset(&mut self, offset: u32, output: &mut Vec<u32>) {
-        self.postings.positions_with_offset(offset, output);
-    }
-    fn positions(&mut self, output: &mut Vec<u32>) {
-        self.postings.positions(output);
-    }
-}
-
-impl BlockMaxPostings for BlockMaxSegmentPostings {
-    fn max_term_freq(&self) -> u32 {
-        self.max_term_freq
-    }
-    fn block_max_term_freq(&mut self) -> u32 {
-        if let SkipResult::End = self.max_blocks.skip_next(self.doc()) {
-            panic!("Max blocks corrupted: reached end of max block");
-        }
-        self.max_blocks.term_freq()
-    }
-    fn max_doc(&self) -> DocId {
-        self.doc_with_max_term_freq
-    }
-    fn block_max_doc(&self) -> DocId {
-        self.max_blocks.doc()
-    }
-}

src/postings/block_search.rs

@@ -129,23 +129,23 @@ impl BlockSearcher {
     ///
     /// If SSE2 instructions are available in the `(platform, running CPU)`,
     /// then we use a different implementation that does an exhaustive linear search over
-    /// the full block whenever the block is full (`len == 128`). It is surprisingly faster, most likely because of the lack
-    /// of branch.
+    /// the block regardless of whether the block is full or not.
+    ///
+    /// Indeed, if the block is not full, the remaining items are TERMINATED.
+    /// It is surprisingly faster, most likely because of the lack of branch misprediction.
     pub(crate) fn search_in_block(
         self,
         block_docs: &AlignedBuffer,
-        len: usize,
         start: usize,
         target: u32,
     ) -> usize {
         #[cfg(target_arch = "x86_64")]
         {
-            use crate::postings::compression::COMPRESSION_BLOCK_SIZE;
-            if self == BlockSearcher::SSE2 && len == COMPRESSION_BLOCK_SIZE {
+            if self == BlockSearcher::SSE2 {
                 return sse2::linear_search_sse2_128(block_docs, target);
             }
         }
-        start + galloping(&block_docs.0[start..len], target)
+        start + galloping(&block_docs.0[start..], target)
     }
 }
 
@@ -166,6 +166,7 @@ mod tests {
     use super::exponential_search;
     use super::linear_search;
     use super::BlockSearcher;
+    use crate::docset::TERMINATED;
     use crate::postings::compression::{AlignedBuffer, COMPRESSION_BLOCK_SIZE};
 
     #[test]
@@ -196,16 +197,11 @@ mod tests {
     fn util_test_search_in_block(block_searcher: BlockSearcher, block: &[u32], target: u32) {
         let cursor = search_in_block_trivial_but_slow(block, target);
         assert!(block.len() < COMPRESSION_BLOCK_SIZE);
-        let mut output_buffer = [u32::max_value(); COMPRESSION_BLOCK_SIZE];
+        let mut output_buffer = [TERMINATED; COMPRESSION_BLOCK_SIZE];
         output_buffer[..block.len()].copy_from_slice(block);
         for i in 0..cursor {
             assert_eq!(
-                block_searcher.search_in_block(
-                    &AlignedBuffer(output_buffer),
-                    block.len(),
-                    i,
-                    target
-                ),
+                block_searcher.search_in_block(&AlignedBuffer(output_buffer), i, target),
                 cursor
             );
         }

src/postings/block_segment_postings.rs

@@ -1,316 +0,0 @@
-use crate::DocId;
-use tantivy_fst::Streamer;
-use crate::postings::{SkipReader, FreqReadingOption, USE_SKIP_INFO_LIMIT};
-use owned_read::OwnedRead;
-use crate::postings::compression::{BlockDecoder, COMPRESSION_BLOCK_SIZE, VIntDecoder, compressed_block_size, AlignedBuffer};
-use crate::schema::IndexRecordOption;
-use crate::common::{VInt, BinarySerializable};
-
-
-fn split_into_skips_and_postings(
-    doc_freq: u32,
-    mut data: OwnedRead,
-) -> (Option<OwnedRead>, OwnedRead) {
-    if doc_freq >= USE_SKIP_INFO_LIMIT {
-        let skip_len = VInt::deserialize(&mut data).expect("Data corrupted").0 as usize;
-        let mut postings_data = data.clone();
-        postings_data.advance(skip_len);
-        data.clip(skip_len);
-        (Some(data), postings_data)
-    } else {
-        (None, data)
-    }
-}
-
-/// `BlockSegmentPostings` is a cursor iterating over blocks
-/// of documents.
-///
-/// # Warning
-///
-/// While it is useful for some very specific high-performance
-/// use cases, you should prefer using `SegmentPostings` for most usage.
-pub struct BlockSegmentPostings {
-    doc_decoder: BlockDecoder,
-    freq_decoder: BlockDecoder,
-    freq_reading_option: FreqReadingOption,
-
-    doc_freq: usize,
-    doc_offset: DocId,
-
-    num_vint_docs: usize,
-
-    remaining_data: OwnedRead,
-    skip_reader: SkipReader,
-}
-
-
-#[derive(Debug, Eq, PartialEq)]
-pub enum BlockSegmentPostingsSkipResult {
-    Terminated,
-    Success(u32), //< number of term freqs to skip
-}
-
-impl BlockSegmentPostings {
-    pub(crate) fn from_data(
-        doc_freq: u32,
-        data: OwnedRead,
-        record_option: IndexRecordOption,
-        requested_option: IndexRecordOption,
-    ) -> BlockSegmentPostings {
-        let freq_reading_option = match (record_option, requested_option) {
-            (IndexRecordOption::Basic, _) => FreqReadingOption::NoFreq,
-            (_, IndexRecordOption::Basic) => FreqReadingOption::SkipFreq,
-            (_, _) => FreqReadingOption::ReadFreq,
-        };
-
-        let (skip_data_opt, postings_data) = split_into_skips_and_postings(doc_freq, data);
-        let skip_reader = match skip_data_opt {
-            Some(skip_data) => SkipReader::new(skip_data, record_option),
-            None => SkipReader::new(OwnedRead::new(&[][..]), record_option),
-        };
-        let doc_freq = doc_freq as usize;
-        let num_vint_docs = doc_freq % COMPRESSION_BLOCK_SIZE;
-        BlockSegmentPostings {
-            num_vint_docs,
-            doc_decoder: BlockDecoder::new(),
-            freq_decoder: BlockDecoder::with_val(1),
-            freq_reading_option,
-            doc_offset: 0,
-            doc_freq,
-            remaining_data: postings_data,
-            skip_reader,
-        }
-    }
-
-    // Resets the block segment postings on another position
-    // in the postings file.
-    //
-    // This is useful for enumerating through a list of terms,
-    // and consuming the associated posting lists while avoiding
-    // reallocating a `BlockSegmentPostings`.
-    //
-    // # Warning
-    //
-    // This does not reset the positions list.
-    pub(crate) fn reset(&mut self, doc_freq: u32, postings_data: OwnedRead) {
-        let (skip_data_opt, postings_data) = split_into_skips_and_postings(doc_freq, postings_data);
-        let num_vint_docs = (doc_freq as usize) & (COMPRESSION_BLOCK_SIZE - 1);
-        self.num_vint_docs = num_vint_docs;
-        self.remaining_data = postings_data;
-        if let Some(skip_data) = skip_data_opt {
-            self.skip_reader.reset(skip_data);
-        } else {
-            self.skip_reader.reset(OwnedRead::new(&[][..]))
-        }
-        self.doc_offset = 0;
-        self.doc_freq = doc_freq as usize;
-    }
-
-    /// 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) -> usize {
-        self.doc_freq
-    }
-
-    /// Returns the array of docs in the current block.
-    ///
-    /// Before the first call to `.advance()`, the block
-    /// returned by `.docs()` is empty.
-    #[inline]
-    pub fn docs(&self) -> &[DocId] {
-        self.doc_decoder.output_array()
-    }
-
-    pub(crate) fn docs_aligned(&self) -> (&AlignedBuffer, usize) {
-        self.doc_decoder.output_aligned()
-    }
-
-    /// Return the document at index `idx` of the block.
-    #[inline]
-    pub fn doc(&self, idx: usize) -> u32 {
-        self.doc_decoder.output(idx)
-    }
-
-    /// Return the array of `term freq` in the block.
-    #[inline]
-    pub fn freqs(&self) -> &[u32] {
-        self.freq_decoder.output_array()
-    }
-
-    /// Return the frequency at index `idx` of the block.
-    #[inline]
-    pub fn freq(&self, idx: usize) -> u32 {
-        self.freq_decoder.output(idx)
-    }
-
-    /// Returns the length of the current block.
-    ///
-    /// All blocks have a length of `NUM_DOCS_PER_BLOCK`,
-    /// except the last block that may have a length
-    /// of any number between 1 and `NUM_DOCS_PER_BLOCK - 1`
-    #[inline]
-    pub(crate) fn block_len(&self) -> usize {
-        self.doc_decoder.output_len
-    }
-
-    /// position on a block that may contains `doc_id`.
-    /// Always advance the current block.
-    ///
-    /// Returns true if a block that has an element greater or equal to the target is found.
-    /// Returning true does not guarantee that the smallest element of the block is smaller
-    /// than the target. It only guarantees that the last element is greater or equal.
-    ///
-    /// Returns false iff all of the document remaining are smaller than
-    /// `doc_id`. In that case, all of these document are consumed.
-    ///
-    pub fn skip_to(&mut self, target_doc: DocId) -> BlockSegmentPostingsSkipResult {
-        let mut skip_freqs = 0u32;
-        while self.skip_reader.advance() {
-            if self.skip_reader.doc() >= target_doc {
-                // the last document of the current block is larger
-                // than the target.
-                //
-                // We found our block!
-                let num_bits = self.skip_reader.doc_num_bits();
-                let num_consumed_bytes = self.doc_decoder.uncompress_block_sorted(
-                    self.remaining_data.as_ref(),
-                    self.doc_offset,
-                    num_bits,
-                );
-                self.remaining_data.advance(num_consumed_bytes);
-                let tf_num_bits = self.skip_reader.tf_num_bits();
-                match self.freq_reading_option {
-                    FreqReadingOption::NoFreq => {}
-                    FreqReadingOption::SkipFreq => {
-                        let num_bytes_to_skip = compressed_block_size(tf_num_bits);
-                        self.remaining_data.advance(num_bytes_to_skip);
-                    }
-                    FreqReadingOption::ReadFreq => {
-                        let num_consumed_bytes = self
-                            .freq_decoder
-                            .uncompress_block_unsorted(self.remaining_data.as_ref(), tf_num_bits);
-                        self.remaining_data.advance(num_consumed_bytes);
-                    }
-                }
-                self.doc_offset = self.skip_reader.doc();
-                return BlockSegmentPostingsSkipResult::Success(skip_freqs);
-            } else {
-                skip_freqs += self.skip_reader.tf_sum();
-                let advance_len = self.skip_reader.total_block_len();
-                self.doc_offset = self.skip_reader.doc();
-                self.remaining_data.advance(advance_len);
-            }
-        }
-
-        // we are now on the last, incomplete, variable encoded block.
-        if self.num_vint_docs > 0 {
-            let num_compressed_bytes = self.doc_decoder.uncompress_vint_sorted(
-                self.remaining_data.as_ref(),
-                self.doc_offset,
-                self.num_vint_docs,
-            );
-            self.remaining_data.advance(num_compressed_bytes);
-            match self.freq_reading_option {
-                FreqReadingOption::NoFreq | FreqReadingOption::SkipFreq => {}
-                FreqReadingOption::ReadFreq => {
-                    self.freq_decoder
-                        .uncompress_vint_unsorted(self.remaining_data.as_ref(), self.num_vint_docs);
-                }
-            }
-            self.num_vint_docs = 0;
-            return self
-                .docs()
-                .last()
-                .map(|last_doc| {
-                    if *last_doc >= target_doc {
-                        BlockSegmentPostingsSkipResult::Success(skip_freqs)
-                    } else {
-                        BlockSegmentPostingsSkipResult::Terminated
-                    }
-                })
-                .unwrap_or(BlockSegmentPostingsSkipResult::Terminated);
-        }
-        BlockSegmentPostingsSkipResult::Terminated
-    }
-
-    /// Advance to the next block.
-    ///
-    /// Returns false iff there was no remaining blocks.
-    pub fn advance(&mut self) -> bool {
-        if self.skip_reader.advance() {
-            let num_bits = self.skip_reader.doc_num_bits();
-            let num_consumed_bytes = self.doc_decoder.uncompress_block_sorted(
-                self.remaining_data.as_ref(),
-                self.doc_offset,
-                num_bits,
-            );
-            self.remaining_data.advance(num_consumed_bytes);
-            let tf_num_bits = self.skip_reader.tf_num_bits();
-            match self.freq_reading_option {
-                FreqReadingOption::NoFreq => {}
-                FreqReadingOption::SkipFreq => {
-                    let num_bytes_to_skip = compressed_block_size(tf_num_bits);
-                    self.remaining_data.advance(num_bytes_to_skip);
-                }
-                FreqReadingOption::ReadFreq => {
-                    let num_consumed_bytes = self
-                        .freq_decoder
-                        .uncompress_block_unsorted(self.remaining_data.as_ref(), tf_num_bits);
-                    self.remaining_data.advance(num_consumed_bytes);
-                }
-            }
-            // it will be used as the next offset.
-            self.doc_offset = self.doc_decoder.output(COMPRESSION_BLOCK_SIZE - 1);
-            true
-        } else if self.num_vint_docs > 0 {
-            let num_compressed_bytes = self.doc_decoder.uncompress_vint_sorted(
-                self.remaining_data.as_ref(),
-                self.doc_offset,
-                self.num_vint_docs,
-            );
-            self.remaining_data.advance(num_compressed_bytes);
-            match self.freq_reading_option {
-                FreqReadingOption::NoFreq | FreqReadingOption::SkipFreq => {}
-                FreqReadingOption::ReadFreq => {
-                    self.freq_decoder
-                        .uncompress_vint_unsorted(self.remaining_data.as_ref(), self.num_vint_docs);
-                }
-            }
-            self.num_vint_docs = 0;
-            true
-        } else {
-            false
-        }
-    }
-
-    /// Returns an empty segment postings object
-    pub fn empty() -> BlockSegmentPostings {
-        BlockSegmentPostings {
-            num_vint_docs: 0,
-
-            doc_decoder: BlockDecoder::new(),
-            freq_decoder: BlockDecoder::with_val(1),
-            freq_reading_option: FreqReadingOption::NoFreq,
-
-            doc_offset: 0,
-            doc_freq: 0,
-
-            remaining_data: OwnedRead::new(vec![]),
-            skip_reader: SkipReader::new(OwnedRead::new(vec![]), IndexRecordOption::Basic),
-        }
-    }
-}
-
-impl<'a> Streamer<'a> for BlockSegmentPostings {
-    type Item = &'a [DocId];
-
-    fn next(&'a mut self) -> Option<&'a [DocId]> {
-        if self.advance() {
-            Some(self.docs())
-        } else {
-            None
-        }
-    }
-}

src/postings/compression/mod.rs

@@ -1,4 +1,5 @@
 use crate::common::FixedSize;
+use crate::docset::TERMINATED;
 use bitpacking::{BitPacker, BitPacker4x};
 
 pub const COMPRESSION_BLOCK_SIZE: usize = BitPacker4x::BLOCK_LEN;
@@ -90,14 +91,18 @@ impl BlockDecoder {
     }
 
     #[inline]
-    pub(crate) fn output_aligned(&self) -> (&AlignedBuffer, usize) {
-        (&self.output, self.output_len)
+    pub(crate) fn output_aligned(&self) -> &AlignedBuffer {
+        &self.output
     }
 
     #[inline]
     pub fn output(&self, idx: usize) -> u32 {
         self.output.0[idx]
     }
+
+    pub fn clear(&mut self) {
+        self.output.0.iter_mut().for_each(|el| *el = TERMINATED);
+    }
 }
 
 pub trait VIntEncoder {
@@ -134,9 +139,9 @@ pub trait VIntDecoder {
     /// For instance, if delta encoded are `1, 3, 9`, and the
     /// `offset` is 5, then the output will be:
     /// `5 + 1 = 6, 6 + 3= 9, 9 + 9 = 18`
-    fn uncompress_vint_sorted<'a>(
+    fn uncompress_vint_sorted(
         &mut self,
-        compressed_data: &'a [u8],
+        compressed_data: &[u8],
         offset: u32,
         num_els: usize,
     ) -> usize;
@@ -146,7 +151,7 @@ pub trait VIntDecoder {
     ///
     /// The method takes a number of int to decompress, and returns
     /// the amount of bytes that were read to decompress them.
-    fn uncompress_vint_unsorted<'a>(&mut self, compressed_data: &'a [u8], num_els: usize) -> usize;
+    fn uncompress_vint_unsorted(&mut self, compressed_data: &[u8], num_els: usize) -> usize;
 }
 
 impl VIntEncoder for BlockEncoder {
@@ -160,9 +165,9 @@ impl VIntEncoder for BlockEncoder {
 }
 
 impl VIntDecoder for BlockDecoder {
-    fn uncompress_vint_sorted<'a>(
+    fn uncompress_vint_sorted(
         &mut self,
-        compressed_data: &'a [u8],
+        compressed_data: &[u8],
         offset: u32,
         num_els: usize,
     ) -> usize {

src/postings/compression/vint.rs

@@ -42,7 +42,7 @@ pub(crate) fn compress_unsorted<'a>(input: &[u32], output: &'a mut [u8]) -> &'a
 }
 
 #[inline(always)]
-pub fn uncompress_sorted<'a>(compressed_data: &'a [u8], output: &mut [u32], offset: u32) -> usize {
+pub fn uncompress_sorted(compressed_data: &[u8], output: &mut [u32], offset: u32) -> usize {
     let mut read_byte = 0;
     let mut result = offset;
     for output_mut in output.iter_mut() {

src/postings/mod.rs

@@ -2,8 +2,6 @@
 Postings module (also called inverted index)
 */
 
-mod block_max_postings;
-mod block_max_segment_postings;
 mod block_search;
 pub(crate) mod compression;
 /// Postings module
@@ -14,7 +12,6 @@ mod postings;
 mod postings_writer;
 mod recorder;
 mod segment_postings;
-mod block_segment_postings;
 mod serializer;
 mod skip;
 mod stacker;
@@ -30,11 +27,7 @@ pub use self::postings::Postings;
 pub(crate) use self::skip::SkipReader;
 pub use self::term_info::TermInfo;
 
-pub use self::segment_postings::SegmentPostings;
-
-pub use self::block_segment_postings::BlockSegmentPostings;
-pub use self::block_max_postings::BlockMaxPostings;
-pub use self::block_max_segment_postings::BlockMaxSegmentPostings;
+pub use self::segment_postings::{BlockSegmentPostings, SegmentPostings};
 
 pub(crate) use self::stacker::compute_table_size;
 
@@ -58,7 +51,7 @@ pub mod tests {
     use crate::core::Index;
     use crate::core::SegmentComponent;
     use crate::core::SegmentReader;
-    use crate::docset::{DocSet, SkipResult};
+    use crate::docset::{DocSet, TERMINATED};
     use crate::fieldnorm::FieldNormReader;
     use crate::indexer::operation::AddOperation;
     use crate::indexer::SegmentWriter;
@@ -122,29 +115,12 @@ pub mod tests {
             let mut postings = inverted_index
                 .read_postings(&term, IndexRecordOption::WithFreqsAndPositions)
                 .unwrap();
-            postings.advance();
+            assert_eq!(postings.doc(), 0);
             postings.positions(&mut positions);
             assert_eq!(&[0, 1, 2], &positions[..]);
             postings.positions(&mut positions);
             assert_eq!(&[0, 1, 2], &positions[..]);
-            postings.advance();
-            postings.positions(&mut positions);
-            assert_eq!(&[0, 5], &positions[..]);
-        }
-        {
-            let mut postings = inverted_index
-                .read_postings(&term, IndexRecordOption::WithFreqsAndPositions)
-                .unwrap();
-            postings.advance();
-            postings.advance();
-            postings.positions(&mut positions);
-            assert_eq!(&[0, 5], &positions[..]);
-        }
-        {
-            let mut postings = inverted_index
-                .read_postings(&term, IndexRecordOption::WithFreqsAndPositions)
-                .unwrap();
-            assert_eq!(postings.skip_next(1), SkipResult::Reached);
+            assert_eq!(postings.advance(), 1);
             assert_eq!(postings.doc(), 1);
             postings.positions(&mut positions);
             assert_eq!(&[0, 5], &positions[..]);
@@ -153,7 +129,25 @@ pub mod tests {
             let mut postings = inverted_index
                 .read_postings(&term, IndexRecordOption::WithFreqsAndPositions)
                 .unwrap();
-            assert_eq!(postings.skip_next(1002), SkipResult::Reached);
+            assert_eq!(postings.doc(), 0);
+            assert_eq!(postings.advance(), 1);
+            postings.positions(&mut positions);
+            assert_eq!(&[0, 5], &positions[..]);
+        }
+        {
+            let mut postings = inverted_index
+                .read_postings(&term, IndexRecordOption::WithFreqsAndPositions)
+                .unwrap();
+            assert_eq!(postings.seek(1), 1);
+            assert_eq!(postings.doc(), 1);
+            postings.positions(&mut positions);
+            assert_eq!(&[0, 5], &positions[..]);
+        }
+        {
+            let mut postings = inverted_index
+                .read_postings(&term, IndexRecordOption::WithFreqsAndPositions)
+                .unwrap();
+            assert_eq!(postings.seek(1002), 1002);
             assert_eq!(postings.doc(), 1002);
             postings.positions(&mut positions);
             assert_eq!(&[0, 5], &positions[..]);
@@ -162,8 +156,8 @@ pub mod tests {
             let mut postings = inverted_index
                 .read_postings(&term, IndexRecordOption::WithFreqsAndPositions)
                 .unwrap();
-            assert_eq!(postings.skip_next(100), SkipResult::Reached);
-            assert_eq!(postings.skip_next(1002), SkipResult::Reached);
+            assert_eq!(postings.seek(100), 100);
+            assert_eq!(postings.seek(1002), 1002);
             assert_eq!(postings.doc(), 1002);
             postings.positions(&mut positions);
             assert_eq!(&[0, 5], &positions[..]);
@@ -288,22 +282,21 @@ pub mod tests {
                     .read_postings(&term_a, IndexRecordOption::WithFreqsAndPositions)
                     .unwrap();
                 assert_eq!(postings_a.len(), 1000);
-                assert!(postings_a.advance());
                 assert_eq!(postings_a.doc(), 0);
                 assert_eq!(postings_a.term_freq(), 6);
                 postings_a.positions(&mut positions);
                 assert_eq!(&positions[..], [0, 2, 4, 6, 7, 13]);
-                assert!(postings_a.advance());
+                assert_eq!(postings_a.advance(), 1u32);
                 assert_eq!(postings_a.doc(), 1u32);
                 assert_eq!(postings_a.term_freq(), 1);
                 for i in 2u32..1000u32 {
-                    assert!(postings_a.advance());
+                    assert_eq!(postings_a.advance(), i);
                     assert_eq!(postings_a.term_freq(), 1);
                     postings_a.positions(&mut positions);
                     assert_eq!(&positions[..], [i]);
                     assert_eq!(postings_a.doc(), i);
                 }
-                assert!(!postings_a.advance());
+                assert_eq!(postings_a.advance(), TERMINATED);
             }
             {
                 let term_e = Term::from_field_text(text_field, "e");
@@ -313,7 +306,6 @@ pub mod tests {
                     .unwrap();
                 assert_eq!(postings_e.len(), 1000 - 2);
                 for i in 2u32..1000u32 {
-                    assert!(postings_e.advance());
                     assert_eq!(postings_e.term_freq(), i);
                     postings_e.positions(&mut positions);
                     assert_eq!(positions.len(), i as usize);
@@ -321,8 +313,9 @@ pub mod tests {
                         assert_eq!(positions[j], (j as u32));
                     }
                     assert_eq!(postings_e.doc(), i);
+                    postings_e.advance();
                 }
-                assert!(!postings_e.advance());
+                assert_eq!(postings_e.doc(), TERMINATED);
             }
         }
     }
@@ -336,16 +329,8 @@ pub mod tests {
         let index = Index::create_in_ram(schema);
         {
             let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
-            {
-                let mut doc = Document::default();
-                doc.add_text(text_field, "g b b d c g c");
-                index_writer.add_document(doc);
-            }
-            {
-                let mut doc = Document::default();
-                doc.add_text(text_field, "g a b b a d c g c");
-                index_writer.add_document(doc);
-            }
+            index_writer.add_document(doc!(text_field => "g b b d c g c"));
+            index_writer.add_document(doc!(text_field => "g a b b a d c g c"));
             assert!(index_writer.commit().is_ok());
         }
         let term_a = Term::from_field_text(text_field, "a");
@@ -355,7 +340,6 @@ pub mod tests {
             .inverted_index(text_field)
             .read_postings(&term_a, IndexRecordOption::WithFreqsAndPositions)
             .unwrap();
-        assert!(postings.advance());
         assert_eq!(postings.doc(), 1u32);
         postings.positions(&mut positions);
         assert_eq!(&positions[..], &[1u32, 4]);
@@ -377,11 +361,8 @@ pub mod tests {
             let index = Index::create_in_ram(schema);
             {
                 let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
-                for i in 0..num_docs {
-                    let mut doc = Document::default();
-                    doc.add_u64(value_field, 2);
-                    doc.add_u64(value_field, (i % 2) as u64);
-
+                for i in 0u64..num_docs as u64 {
+                    let doc = doc!(value_field => 2u64, value_field => i % 2u64);
                     index_writer.add_document(doc);
                 }
                 assert!(index_writer.commit().is_ok());
@@ -398,11 +379,10 @@ pub mod tests {
                     .inverted_index(term_2.field())
                     .read_postings(&term_2, IndexRecordOption::Basic)
                     .unwrap();
-
-                assert_eq!(segment_postings.skip_next(i), SkipResult::Reached);
+                assert_eq!(segment_postings.seek(i), i);
                 assert_eq!(segment_postings.doc(), i);
 
-                assert_eq!(segment_postings.skip_next(j), SkipResult::Reached);
+                assert_eq!(segment_postings.seek(j), j);
                 assert_eq!(segment_postings.doc(), j);
             }
         }
@@ -414,17 +394,16 @@ pub mod tests {
                 .unwrap();
 
             // check that `skip_next` advances the iterator
-            assert!(segment_postings.advance());
             assert_eq!(segment_postings.doc(), 0);
 
-            assert_eq!(segment_postings.skip_next(1), SkipResult::Reached);
+            assert_eq!(segment_postings.seek(1), 1);
             assert_eq!(segment_postings.doc(), 1);
 
-            assert_eq!(segment_postings.skip_next(1), SkipResult::OverStep);
-            assert_eq!(segment_postings.doc(), 2);
+            assert_eq!(segment_postings.seek(1), 1);
+            assert_eq!(segment_postings.doc(), 1);
 
             // check that going beyond the end is handled
-            assert_eq!(segment_postings.skip_next(num_docs), SkipResult::End);
+            assert_eq!(segment_postings.seek(num_docs), TERMINATED);
         }
 
         // check that filtering works
@@ -435,7 +414,7 @@ pub mod tests {
                 .unwrap();
 
             for i in 0..num_docs / 2 {
-                assert_eq!(segment_postings.skip_next(i * 2), SkipResult::Reached);
+                assert_eq!(segment_postings.seek(i * 2), i * 2);
                 assert_eq!(segment_postings.doc(), i * 2);
             }
 
@@ -445,7 +424,7 @@ pub mod tests {
                 .unwrap();
 
             for i in 0..num_docs / 2 - 1 {
-                assert_eq!(segment_postings.skip_next(i * 2 + 1), SkipResult::OverStep);
+                assert!(segment_postings.seek(i * 2 + 1) > (i * 1) * 2);
                 assert_eq!(segment_postings.doc(), (i + 1) * 2);
             }
         }
@@ -457,6 +436,7 @@ pub mod tests {
             assert!(index_writer.commit().is_ok());
         }
         let searcher = index.reader().unwrap().searcher();
+        assert_eq!(searcher.segment_readers().len(), 1);
         let segment_reader = searcher.segment_reader(0);
 
         // make sure seeking still works
@@ -467,11 +447,11 @@ pub mod tests {
                 .unwrap();
 
             if i % 2 == 0 {
-                assert_eq!(segment_postings.skip_next(i), SkipResult::Reached);
+                assert_eq!(segment_postings.seek(i), i);
                 assert_eq!(segment_postings.doc(), i);
                 assert!(segment_reader.is_deleted(i));
             } else {
-                assert_eq!(segment_postings.skip_next(i), SkipResult::Reached);
+                assert_eq!(segment_postings.seek(i), i);
                 assert_eq!(segment_postings.doc(), i);
             }
         }
@@ -486,12 +466,16 @@ pub mod tests {
             let mut last = 2; // start from 5 to avoid seeking to 3 twice
             let mut cur = 3;
             loop {
-                match segment_postings.skip_next(cur) {
-                    SkipResult::End => break,
-                    SkipResult::Reached => assert_eq!(segment_postings.doc(), cur),
-                    SkipResult::OverStep => assert_eq!(segment_postings.doc(), cur + 1),
+                let seek = segment_postings.seek(cur);
+                if seek == TERMINATED {
+                    break;
+                }
+                assert_eq!(seek, segment_postings.doc());
+                if seek == cur {
+                    assert_eq!(segment_postings.doc(), cur);
+                } else {
+                    assert_eq!(segment_postings.doc(), cur + 1);
                 }
-
                 let next = cur + last;
                 last = cur;
                 cur = next;
@@ -577,7 +561,7 @@ pub mod tests {
     }
 
     impl<TDocSet: DocSet> DocSet for UnoptimizedDocSet<TDocSet> {
-        fn advance(&mut self) -> bool {
+        fn advance(&mut self) -> DocId {
             self.0.advance()
         }
 
@@ -603,30 +587,22 @@ pub mod tests {
         for target in targets {
             let mut postings_opt = postings_factory();
             let mut postings_unopt = UnoptimizedDocSet::wrap(postings_factory());
-            let skip_result_opt = postings_opt.skip_next(target);
-            let skip_result_unopt = postings_unopt.skip_next(target);
+            let skip_result_opt = postings_opt.seek(target);
+            let skip_result_unopt = postings_unopt.seek(target);
             assert_eq!(
                 skip_result_unopt, skip_result_opt,
                 "Failed while skipping to {}",
                 target
             );
-            match skip_result_opt {
-                SkipResult::Reached => assert_eq!(postings_opt.doc(), target),
-                SkipResult::OverStep => assert!(postings_opt.doc() > target),
-                SkipResult::End => {
-                    return;
-                }
+            assert!(skip_result_opt >= target);
+            assert_eq!(skip_result_opt, postings_opt.doc());
+            if skip_result_opt == TERMINATED {
+                return;
             }
-            while postings_opt.advance() {
-                assert!(postings_unopt.advance());
-                assert_eq!(
-                    postings_opt.doc(),
-                    postings_unopt.doc(),
-                    "Failed while skipping to {}",
-                    target
-                );
+            while postings_opt.doc() != TERMINATED {
+                assert_eq!(postings_opt.doc(), postings_unopt.doc());
+                assert_eq!(postings_opt.advance(), postings_unopt.advance());
             }
-            assert!(!postings_unopt.advance());
         }
     }
 }
@@ -635,7 +611,7 @@ pub mod tests {
 mod bench {
 
     use super::tests::*;
-    use crate::docset::SkipResult;
+    use crate::docset::TERMINATED;
     use crate::query::Intersection;
     use crate::schema::IndexRecordOption;
     use crate::tests;
@@ -653,7 +629,7 @@ mod bench {
                 .inverted_index(TERM_A.field())
                 .read_postings(&*TERM_A, IndexRecordOption::Basic)
                 .unwrap();
-            while segment_postings.advance() {}
+            while segment_postings.advance() != TERMINATED {}
         });
     }
 
@@ -685,7 +661,7 @@ mod bench {
                 segment_postings_c,
                 segment_postings_d,
             ]);
-            while intersection.advance() {}
+            while intersection.advance() != TERMINATED {}
         });
     }
 
@@ -701,11 +677,10 @@ mod bench {
             .unwrap();
 
         let mut existing_docs = Vec::new();
-        segment_postings.advance();
         for doc in &docs {
             if *doc >= segment_postings.doc() {
                 existing_docs.push(*doc);
-                if segment_postings.skip_next(*doc) == SkipResult::End {
+                if segment_postings.seek(*doc) == TERMINATED {
                     break;
                 }
             }
@@ -717,7 +692,7 @@ mod bench {
                 .read_postings(&*TERM_A, IndexRecordOption::Basic)
                 .unwrap();
             for doc in &existing_docs {
-                if segment_postings.skip_next(*doc) == SkipResult::End {
+                if segment_postings.seek(*doc) == TERMINATED {
                     break;
                 }
             }
@@ -756,8 +731,9 @@ mod bench {
                 .read_postings(&*TERM_A, IndexRecordOption::Basic)
                 .unwrap();
             let mut s = 0u32;
-            while segment_postings.advance() {
+            while segment_postings.doc() != TERMINATED {
                 s += (segment_postings.doc() & n) % 1024;
+                segment_postings.advance()
             }
             s
         });

src/postings/segment_postings.rs

@@ -1,16 +1,19 @@
-use crate::common::BitSet;
 use crate::common::HasLen;
-use crate::docset::{DocSet, SkipResult};
+use crate::common::{BinarySerializable, VInt};
+use crate::docset::{DocSet, TERMINATED};
 use crate::positions::PositionReader;
-use crate::postings::compression::COMPRESSION_BLOCK_SIZE;
+use crate::postings::compression::{compressed_block_size, AlignedBuffer};
+use crate::postings::compression::{BlockDecoder, VIntDecoder, COMPRESSION_BLOCK_SIZE};
 use crate::postings::serializer::PostingsSerializer;
-use crate::postings::{BlockSearcher, BlockSegmentPostings};
+use crate::postings::BlockSearcher;
+use crate::postings::FreqReadingOption;
 use crate::postings::Postings;
+use crate::postings::SkipReader;
+use crate::postings::USE_SKIP_INFO_LIMIT;
 use crate::schema::IndexRecordOption;
 use crate::DocId;
 use owned_read::OwnedRead;
-use std::cmp::Ordering;
-use crate::postings::block_segment_postings::BlockSegmentPostingsSkipResult;
+use tantivy_fst::Streamer;
 
 struct PositionComputer {
     // store the amount of position int
@@ -63,12 +66,14 @@ impl SegmentPostings {
     /// Returns an empty segment postings object
     pub fn empty() -> Self {
         let empty_block_cursor = BlockSegmentPostings::empty();
-        SegmentPostings {
+        let mut segment_postings = SegmentPostings {
             block_cursor: empty_block_cursor,
             cur: COMPRESSION_BLOCK_SIZE,
             position_computer: None,
             block_searcher: BlockSearcher::default(),
-        }
+        };
+        segment_postings.advance();
+        segment_postings
     }
 
     /// Creates a segment postings object with the given documents
@@ -111,12 +116,14 @@ impl SegmentPostings {
         segment_block_postings: BlockSegmentPostings,
         positions_stream_opt: Option<PositionReader>,
     ) -> SegmentPostings {
-        SegmentPostings {
+        let mut postings = SegmentPostings {
             block_cursor: segment_block_postings,
             cur: COMPRESSION_BLOCK_SIZE, // cursor within the block
             position_computer: positions_stream_opt.map(PositionComputer::new),
             block_searcher: BlockSearcher::default(),
-        }
+        };
+        postings.advance();
+        postings
     }
 }
 
@@ -124,7 +131,7 @@ impl DocSet for SegmentPostings {
     // goes to the next element.
     // next needs to be called a first time to point to the correct element.
     #[inline]
-    fn advance(&mut self) -> bool {
+    fn advance(&mut self) -> DocId {
         if self.position_computer.is_some() && self.cur < COMPRESSION_BLOCK_SIZE {
             let term_freq = self.term_freq() as usize;
             if let Some(position_computer) = self.position_computer.as_mut() {
@@ -133,29 +140,19 @@ impl DocSet for SegmentPostings {
         }
         self.cur += 1;
         if self.cur >= self.block_cursor.block_len() {
-            self.cur = 0;
-            if !self.block_cursor.advance() {
-                self.cur = COMPRESSION_BLOCK_SIZE;
-                return false;
+            if self.block_cursor.advance() {
+                self.cur = 0;
+            } else {
+                self.cur = COMPRESSION_BLOCK_SIZE - 1;
+                return TERMINATED;
             }
         }
-        true
+        self.doc()
     }
 
-    fn skip_next(&mut self, target: DocId) -> SkipResult {
-        if !self.advance() {
-            return SkipResult::End;
-        }
-        match self.doc().cmp(&target) {
-            Ordering::Equal => {
-                return SkipResult::Reached;
-            }
-            Ordering::Greater => {
-                return SkipResult::OverStep;
-            }
-            _ => {
-                // ...
-            }
+    fn seek(&mut self, target: DocId) -> DocId {
+        if self.doc() >= target {
+            return self.doc();
         }
 
         // In the following, thanks to the call to advance above,
@@ -165,44 +162,44 @@ impl DocSet for SegmentPostings {
         // skip blocks until one that might contain the target
         // check if we need to go to the next block
         let mut sum_freqs_skipped: u32 = 0;
-        if !self
+        if self
             .block_cursor
             .docs()
             .last()
-            .map(|doc| *doc >= target)
-            .unwrap_or(false)
-        // there should always be at least a document in the block
-        // since advance returned.
+            .map(|&doc| doc < target)
+            .unwrap_or(true)
         {
-            // we are not in the right block.
-            //
-            // First compute all of the freqs skipped from the current block.
+            // We are not in the right block.
             if self.position_computer.is_some() {
-                sum_freqs_skipped = self.block_cursor.freqs()[self.cur..].iter().sum();
+                // First compute all of the freqs skipped from the current block.
+                sum_freqs_skipped = self.block_cursor.freqs()[self.cur..].iter().sum::<u32>();
                 match self.block_cursor.skip_to(target) {
                     BlockSegmentPostingsSkipResult::Success(block_skip_freqs) => {
                         sum_freqs_skipped += block_skip_freqs;
                     }
                     BlockSegmentPostingsSkipResult::Terminated => {
-                        return SkipResult::End;
+                        self.block_cursor.doc_decoder.clear();
+                        self.cur = 0;
+                        return TERMINATED;
                     }
                 }
             } else if self.block_cursor.skip_to(target)
                 == BlockSegmentPostingsSkipResult::Terminated
             {
                 // no positions needed. no need to sum freqs.
-                return SkipResult::End;
+                self.block_cursor.doc_decoder.clear();
+                self.cur = 0;
+                return TERMINATED;
             }
             self.cur = 0;
         }
 
+        // At this point we are on the block, that might contain our document.
+
         let cur = self.cur;
 
-        // we're in the right block now, start with an exponential search
-        let (output, len) = self.block_cursor.docs_aligned();
-        let new_cur = self
-            .block_searcher
-            .search_in_block(&output, len, cur, target);
+        let output = self.block_cursor.docs_aligned();
+        let new_cur = self.block_searcher.search_in_block(&output, cur, target);
         if let Some(position_computer) = self.position_computer.as_mut() {
             sum_freqs_skipped += self.block_cursor.freqs()[cur..new_cur].iter().sum::<u32>();
             position_computer.add_skip(sum_freqs_skipped as usize);
@@ -212,11 +209,7 @@ impl DocSet for SegmentPostings {
         // `doc` is now the first element >= `target`
         let doc = output.0[new_cur];
         debug_assert!(doc >= target);
-        if doc == target {
-            SkipResult::Reached
-        } else {
-            SkipResult::OverStep
-        }
+        doc
     }
 
     /// Return the current document's `DocId`.
@@ -226,18 +219,14 @@ impl DocSet for SegmentPostings {
     /// Will panics if called without having called advance before.
     #[inline]
     fn doc(&self) -> DocId {
-        let docs = self.block_cursor.docs();
-        debug_assert!(
-            self.cur < docs.len(),
-            "Have you forgotten to call `.advance()` at least once before calling `.doc()`                                      ."
-        );
-        docs[self.cur]
+        self.block_cursor.doc(self.cur)
     }
 
     fn size_hint(&self) -> u32 {
         self.len() as u32
     }
 
+    /*
     fn append_to_bitset(&mut self, bitset: &mut BitSet) {
         // finish the current block
         if self.advance() {
@@ -252,6 +241,7 @@ impl DocSet for SegmentPostings {
             }
         }
     }
+     */
 }
 
 impl HasLen for SegmentPostings {
@@ -294,6 +284,314 @@ impl Postings for SegmentPostings {
     }
 }
 
+/// `BlockSegmentPostings` is a cursor iterating over blocks
+/// of documents.
+///
+/// # Warning
+///
+/// While it is useful for some very specific high-performance
+/// use cases, you should prefer using `SegmentPostings` for most usage.
+pub struct BlockSegmentPostings {
+    doc_decoder: BlockDecoder,
+    freq_decoder: BlockDecoder,
+    freq_reading_option: FreqReadingOption,
+
+    doc_freq: usize,
+    doc_offset: DocId,
+
+    num_vint_docs: usize,
+
+    remaining_data: OwnedRead,
+    skip_reader: SkipReader,
+}
+
+fn split_into_skips_and_postings(
+    doc_freq: u32,
+    mut data: OwnedRead,
+) -> (Option<OwnedRead>, OwnedRead) {
+    if doc_freq < USE_SKIP_INFO_LIMIT {
+        return (None, data);
+    }
+    let skip_len = VInt::deserialize(&mut data).expect("Data corrupted").0 as usize;
+    let mut postings_data = data.clone();
+    postings_data.advance(skip_len);
+    data.clip(skip_len);
+    (Some(data), postings_data)
+}
+
+#[derive(Debug, Eq, PartialEq)]
+pub enum BlockSegmentPostingsSkipResult {
+    Terminated,
+    Success(u32), //< number of term freqs to skip
+}
+
+impl BlockSegmentPostings {
+    pub(crate) fn from_data(
+        doc_freq: u32,
+        data: OwnedRead,
+        record_option: IndexRecordOption,
+        requested_option: IndexRecordOption,
+    ) -> BlockSegmentPostings {
+        let freq_reading_option = match (record_option, requested_option) {
+            (IndexRecordOption::Basic, _) => FreqReadingOption::NoFreq,
+            (_, IndexRecordOption::Basic) => FreqReadingOption::SkipFreq,
+            (_, _) => FreqReadingOption::ReadFreq,
+        };
+
+        let (skip_data_opt, postings_data) = split_into_skips_and_postings(doc_freq, data);
+        let skip_reader = match skip_data_opt {
+            Some(skip_data) => SkipReader::new(skip_data, record_option),
+            None => SkipReader::new(OwnedRead::new(&[][..]), record_option),
+        };
+        let doc_freq = doc_freq as usize;
+        let num_vint_docs = doc_freq % COMPRESSION_BLOCK_SIZE;
+        BlockSegmentPostings {
+            num_vint_docs,
+            doc_decoder: BlockDecoder::new(),
+            freq_decoder: BlockDecoder::with_val(1),
+            freq_reading_option,
+            doc_offset: 0,
+            doc_freq,
+            remaining_data: postings_data,
+            skip_reader,
+        }
+    }
+
+    // Resets the block segment postings on another position
+    // in the postings file.
+    //
+    // This is useful for enumerating through a list of terms,
+    // and consuming the associated posting lists while avoiding
+    // reallocating a `BlockSegmentPostings`.
+    //
+    // # Warning
+    //
+    // This does not reset the positions list.
+    pub(crate) fn reset(&mut self, doc_freq: u32, postings_data: OwnedRead) {
+        let (skip_data_opt, postings_data) = split_into_skips_and_postings(doc_freq, postings_data);
+        let num_vint_docs = (doc_freq as usize) & (COMPRESSION_BLOCK_SIZE - 1);
+        self.num_vint_docs = num_vint_docs;
+        self.remaining_data = postings_data;
+        if let Some(skip_data) = skip_data_opt {
+            self.skip_reader.reset(skip_data);
+        } else {
+            self.skip_reader.reset(OwnedRead::new(&[][..]))
+        }
+        self.doc_offset = 0;
+        self.doc_freq = doc_freq as usize;
+    }
+
+    /// 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) -> usize {
+        self.doc_freq
+    }
+
+    /// Returns the array of docs in the current block.
+    ///
+    /// Before the first call to `.advance()`, the block
+    /// returned by `.docs()` is empty.
+    #[inline]
+    pub fn docs(&self) -> &[DocId] {
+        self.doc_decoder.output_array()
+    }
+
+    pub(crate) fn docs_aligned(&self) -> &AlignedBuffer {
+        self.doc_decoder.output_aligned()
+    }
+
+    /// Return the document at index `idx` of the block.
+    #[inline]
+    pub fn doc(&self, idx: usize) -> u32 {
+        self.doc_decoder.output(idx)
+    }
+
+    /// Return the array of `term freq` in the block.
+    #[inline]
+    pub fn freqs(&self) -> &[u32] {
+        self.freq_decoder.output_array()
+    }
+
+    /// Return the frequency at index `idx` of the block.
+    #[inline]
+    pub fn freq(&self, idx: usize) -> u32 {
+        self.freq_decoder.output(idx)
+    }
+
+    /// Returns the length of the current block.
+    ///
+    /// All blocks have a length of `NUM_DOCS_PER_BLOCK`,
+    /// except the last block that may have a length
+    /// of any number between 1 and `NUM_DOCS_PER_BLOCK - 1`
+    #[inline]
+    fn block_len(&self) -> usize {
+        self.doc_decoder.output_len
+    }
+
+    /// position on a block that may contains `doc_id`.
+    /// Always advance the current block.
+    ///
+    /// Returns true if a block that has an element greater or equal to the target is found.
+    /// Returning true does not guarantee that the smallest element of the block is smaller
+    /// than the target. It only guarantees that the last element is greater or equal.
+    ///
+    /// Returns false iff all of the document remaining are smaller than
+    /// `doc_id`. In that case, all of these document are consumed.
+    ///
+    pub fn skip_to(&mut self, target_doc: DocId) -> BlockSegmentPostingsSkipResult {
+        let mut skip_freqs = 0u32;
+        while self.skip_reader.advance() {
+            if self.skip_reader.doc() >= target_doc {
+                // the last document of the current block is larger
+                // than the target.
+                //
+                // We found our block!
+                let num_bits = self.skip_reader.doc_num_bits();
+                let num_consumed_bytes = self.doc_decoder.uncompress_block_sorted(
+                    self.remaining_data.as_ref(),
+                    self.doc_offset,
+                    num_bits,
+                );
+                self.remaining_data.advance(num_consumed_bytes);
+                let tf_num_bits = self.skip_reader.tf_num_bits();
+                match self.freq_reading_option {
+                    FreqReadingOption::NoFreq => {}
+                    FreqReadingOption::SkipFreq => {
+                        let num_bytes_to_skip = compressed_block_size(tf_num_bits);
+                        self.remaining_data.advance(num_bytes_to_skip);
+                    }
+                    FreqReadingOption::ReadFreq => {
+                        let num_consumed_bytes = self
+                            .freq_decoder
+                            .uncompress_block_unsorted(self.remaining_data.as_ref(), tf_num_bits);
+                        self.remaining_data.advance(num_consumed_bytes);
+                    }
+                }
+                self.doc_offset = self.skip_reader.doc();
+                return BlockSegmentPostingsSkipResult::Success(skip_freqs);
+            } else {
+                skip_freqs += self.skip_reader.tf_sum();
+                let advance_len = self.skip_reader.total_block_len();
+                self.doc_offset = self.skip_reader.doc();
+                self.remaining_data.advance(advance_len);
+            }
+        }
+
+        self.doc_decoder.clear();
+
+        if self.num_vint_docs == 0 {
+            return BlockSegmentPostingsSkipResult::Terminated;
+        }
+        // we are now on the last, incomplete, variable encoded block.
+        let num_compressed_bytes = self.doc_decoder.uncompress_vint_sorted(
+            self.remaining_data.as_ref(),
+            self.doc_offset,
+            self.num_vint_docs,
+        );
+        self.remaining_data.advance(num_compressed_bytes);
+        match self.freq_reading_option {
+            FreqReadingOption::NoFreq | FreqReadingOption::SkipFreq => {}
+            FreqReadingOption::ReadFreq => {
+                self.freq_decoder
+                    .uncompress_vint_unsorted(self.remaining_data.as_ref(), self.num_vint_docs);
+            }
+        }
+        self.num_vint_docs = 0;
+        self.docs()
+            .last()
+            .map(|last_doc| {
+                if *last_doc >= target_doc {
+                    BlockSegmentPostingsSkipResult::Success(skip_freqs)
+                } else {
+                    BlockSegmentPostingsSkipResult::Terminated
+                }
+            })
+            .unwrap_or(BlockSegmentPostingsSkipResult::Terminated)
+    }
+
+    /// Advance to the next block.
+    ///
+    /// Returns false iff there was no remaining blocks.
+    pub fn advance(&mut self) -> bool {
+        if self.skip_reader.advance() {
+            let num_bits = self.skip_reader.doc_num_bits();
+            let num_consumed_bytes = self.doc_decoder.uncompress_block_sorted(
+                self.remaining_data.as_ref(),
+                self.doc_offset,
+                num_bits,
+            );
+            self.remaining_data.advance(num_consumed_bytes);
+            let tf_num_bits = self.skip_reader.tf_num_bits();
+            match self.freq_reading_option {
+                FreqReadingOption::NoFreq => {}
+                FreqReadingOption::SkipFreq => {
+                    let num_bytes_to_skip = compressed_block_size(tf_num_bits);
+                    self.remaining_data.advance(num_bytes_to_skip);
+                }
+                FreqReadingOption::ReadFreq => {
+                    let num_consumed_bytes = self
+                        .freq_decoder
+                        .uncompress_block_unsorted(self.remaining_data.as_ref(), tf_num_bits);
+                    self.remaining_data.advance(num_consumed_bytes);
+                }
+            }
+            // it will be used as the next offset.
+            self.doc_offset = self.doc_decoder.output(COMPRESSION_BLOCK_SIZE - 1);
+            return true;
+        }
+        self.doc_decoder.clear();
+        if self.num_vint_docs == 0 {
+            return false;
+        }
+        let num_compressed_bytes = self.doc_decoder.uncompress_vint_sorted(
+            self.remaining_data.as_ref(),
+            self.doc_offset,
+            self.num_vint_docs,
+        );
+        self.remaining_data.advance(num_compressed_bytes);
+        match self.freq_reading_option {
+            FreqReadingOption::NoFreq | FreqReadingOption::SkipFreq => {}
+            FreqReadingOption::ReadFreq => {
+                self.freq_decoder
+                    .uncompress_vint_unsorted(self.remaining_data.as_ref(), self.num_vint_docs);
+            }
+        }
+        self.num_vint_docs = 0;
+        true
+    }
+
+    /// Returns an empty segment postings object
+    pub fn empty() -> BlockSegmentPostings {
+        BlockSegmentPostings {
+            num_vint_docs: 0,
+
+            doc_decoder: BlockDecoder::new(),
+            freq_decoder: BlockDecoder::with_val(1),
+            freq_reading_option: FreqReadingOption::NoFreq,
+
+            doc_offset: 0,
+            doc_freq: 0,
+
+            remaining_data: OwnedRead::new(vec![]),
+            skip_reader: SkipReader::new(OwnedRead::new(vec![]), IndexRecordOption::Basic),
+        }
+    }
+}
+
+impl<'b> Streamer<'b> for BlockSegmentPostings {
+    type Item = &'b [DocId];
+
+    fn next(&'b mut self) -> Option<&'b [DocId]> {
+        if self.advance() {
+            Some(self.docs())
+        } else {
+            None
+        }
+    }
+}
+
 #[cfg(test)]
 mod tests {
     use super::BlockSegmentPostings;
@@ -301,34 +599,34 @@ mod tests {
     use super::SegmentPostings;
     use crate::common::HasLen;
     use crate::core::Index;
-    use crate::docset::DocSet;
+    use crate::docset::{DocSet, TERMINATED};
     use crate::postings::postings::Postings;
     use crate::schema::IndexRecordOption;
     use crate::schema::Schema;
     use crate::schema::Term;
     use crate::schema::INDEXED;
     use crate::DocId;
-    use crate::SkipResult;
     use tantivy_fst::Streamer;
 
     #[test]
     fn test_empty_segment_postings() {
         let mut postings = SegmentPostings::empty();
-        assert!(!postings.advance());
-        assert!(!postings.advance());
+        assert_eq!(postings.advance(), TERMINATED);
+        assert_eq!(postings.advance(), TERMINATED);
         assert_eq!(postings.len(), 0);
     }
 
     #[test]
-    #[should_panic(expected = "Have you forgotten to call `.advance()`")]
-    fn test_panic_if_doc_called_before_advance() {
-        SegmentPostings::empty().doc();
+    fn test_empty_postings_doc_returns_terminated() {
+        let mut postings = SegmentPostings::empty();
+        assert_eq!(postings.doc(), TERMINATED);
+        assert_eq!(postings.advance(), TERMINATED);
     }
 
     #[test]
-    #[should_panic(expected = "Have you forgotten to call `.advance()`")]
-    fn test_panic_if_freq_called_before_advance() {
-        SegmentPostings::empty().term_freq();
+    fn test_empty_postings_doc_term_freq_returns_0() {
+        let postings = SegmentPostings::empty();
+        assert_eq!(postings.term_freq(), 1);
     }
 
     #[test]
@@ -362,25 +660,27 @@ mod tests {
         {
             let block_segments = build_block_postings(&doc_ids);
             let mut docset = SegmentPostings::from_block_postings(block_segments, None);
-            assert_eq!(docset.skip_next(128), SkipResult::OverStep);
+            assert_eq!(docset.seek(128), 129);
             assert_eq!(docset.doc(), 129);
-            assert!(docset.advance());
+            assert_eq!(docset.advance(), 130);
             assert_eq!(docset.doc(), 130);
-            assert!(!docset.advance());
+            assert_eq!(docset.advance(), TERMINATED);
         }
         {
             let block_segments = build_block_postings(&doc_ids);
             let mut docset = SegmentPostings::from_block_postings(block_segments, None);
-            assert_eq!(docset.skip_next(129), SkipResult::Reached);
+            assert_eq!(docset.seek(129), 129);
             assert_eq!(docset.doc(), 129);
-            assert!(docset.advance());
+            assert_eq!(docset.advance(), 130);
             assert_eq!(docset.doc(), 130);
-            assert!(!docset.advance());
+            assert_eq!(docset.advance(), TERMINATED);
         }
         {
             let block_segments = build_block_postings(&doc_ids);
             let mut docset = SegmentPostings::from_block_postings(block_segments, None);
-            assert_eq!(docset.skip_next(131), SkipResult::End);
+            assert_eq!(docset.doc(), 0);
+            assert_eq!(docset.seek(131), TERMINATED);
+            assert_eq!(docset.doc(), TERMINATED);
         }
     }
 

src/query/all_query.rs

@@ -1,6 +1,6 @@
 use crate::core::Searcher;
 use crate::core::SegmentReader;
-use crate::docset::DocSet;
+use crate::docset::{DocSet, TERMINATED};
 use crate::query::boost_query::BoostScorer;
 use crate::query::explanation::does_not_match;
 use crate::query::{Explanation, Query, Scorer, Weight};
@@ -25,7 +25,6 @@ pub struct AllWeight;
 impl Weight for AllWeight {
     fn scorer(&self, reader: &SegmentReader, boost: f32) -> crate::Result<Box<dyn Scorer>> {
         let all_scorer = AllScorer {
-            state: State::NotStarted,
             doc: 0u32,
             max_doc: reader.max_doc(),
         };
@@ -40,39 +39,20 @@ impl Weight for AllWeight {
     }
 }
 
-enum State {
-    NotStarted,
-    Started,
-    Finished,
-}
-
 /// Scorer associated to the `AllQuery` query.
 pub struct AllScorer {
-    state: State,
     doc: DocId,
     max_doc: DocId,
 }
 
 impl DocSet for AllScorer {
-    fn advance(&mut self) -> bool {
-        match self.state {
-            State::NotStarted => {
-                self.state = State::Started;
-                self.doc = 0;
-            }
-            State::Started => {
-                self.doc += 1u32;
-            }
-            State::Finished => {
-                return false;
-            }
-        }
-        if self.doc < self.max_doc {
-            true
-        } else {
-            self.state = State::Finished;
-            false
+    fn advance(&mut self) -> DocId {
+        if self.doc + 1 >= self.max_doc {
+            self.doc = TERMINATED;
+            return TERMINATED;
         }
+        self.doc += 1;
+        self.doc
     }
 
     fn doc(&self) -> DocId {
@@ -93,6 +73,7 @@ impl Scorer for AllScorer {
 #[cfg(test)]
 mod tests {
     use super::AllQuery;
+    use crate::docset::TERMINATED;
     use crate::query::Query;
     use crate::schema::{Schema, TEXT};
     use crate::Index;
@@ -120,18 +101,16 @@ mod tests {
         {
             let reader = searcher.segment_reader(0);
             let mut scorer = weight.scorer(reader, 1.0f32).unwrap();
-            assert!(scorer.advance());
             assert_eq!(scorer.doc(), 0u32);
-            assert!(scorer.advance());
+            assert_eq!(scorer.advance(), 1u32);
             assert_eq!(scorer.doc(), 1u32);
-            assert!(!scorer.advance());
+            assert_eq!(scorer.advance(), TERMINATED);
         }
         {
             let reader = searcher.segment_reader(1);
             let mut scorer = weight.scorer(reader, 1.0f32).unwrap();
-            assert!(scorer.advance());
             assert_eq!(scorer.doc(), 0u32);
-            assert!(!scorer.advance());
+            assert_eq!(scorer.advance(), TERMINATED);
         }
     }
 
@@ -144,13 +123,11 @@ mod tests {
         let reader = searcher.segment_reader(0);
         {
             let mut scorer = weight.scorer(reader, 2.0f32).unwrap();
-            assert!(scorer.advance());
             assert_eq!(scorer.doc(), 0u32);
             assert_eq!(scorer.score(), 2.0f32);
         }
         {
             let mut scorer = weight.scorer(reader, 1.5f32).unwrap();
-            assert!(scorer.advance());
             assert_eq!(scorer.doc(), 0u32);
             assert_eq!(scorer.score(), 1.5f32);
         }

src/query/automaton_weight.rs

@@ -6,8 +6,8 @@ use crate::query::{Scorer, Weight};
 use crate::schema::{Field, IndexRecordOption};
 use crate::termdict::{TermDictionary, TermStreamer};
 use crate::DocId;
+use crate::Result;
 use crate::TantivyError;
-use crate::{Result, SkipResult};
 use std::sync::Arc;
 use tantivy_fst::Automaton;
 
@@ -64,7 +64,7 @@ where
 
     fn explain(&self, reader: &SegmentReader, doc: DocId) -> Result<Explanation> {
         let mut scorer = self.scorer(reader, 1.0f32)?;
-        if scorer.skip_next(doc) == SkipResult::Reached {
+        if scorer.seek(doc) == doc {
             Ok(Explanation::new("AutomatonScorer", 1.0f32))
         } else {
             Err(TantivyError::InvalidArgument(
@@ -77,6 +77,7 @@ where
 #[cfg(test)]
 mod tests {
     use super::AutomatonWeight;
+    use crate::docset::TERMINATED;
     use crate::query::Weight;
     use crate::schema::{Schema, STRING};
     use crate::Index;
@@ -141,13 +142,12 @@ mod tests {
         let mut scorer = automaton_weight
             .scorer(searcher.segment_reader(0u32), 1.0f32)
             .unwrap();
-        assert!(scorer.advance());
         assert_eq!(scorer.doc(), 0u32);
         assert_eq!(scorer.score(), 1.0f32);
-        assert!(scorer.advance());
+        assert_eq!(scorer.advance(), 2u32);
         assert_eq!(scorer.doc(), 2u32);
         assert_eq!(scorer.score(), 1.0f32);
-        assert!(!scorer.advance());
+        assert_eq!(scorer.advance(), TERMINATED);
     }
 
     #[test]
@@ -160,7 +160,6 @@ mod tests {
         let mut scorer = automaton_weight
             .scorer(searcher.segment_reader(0u32), 1.32f32)
             .unwrap();
-        assert!(scorer.advance());
         assert_eq!(scorer.doc(), 0u32);
         assert_eq!(scorer.score(), 1.32f32);
     }

src/query/bitset/mod.rs

@@ -1,7 +1,6 @@
 use crate::common::{BitSet, TinySet};
-use crate::docset::{DocSet, SkipResult};
+use crate::docset::{DocSet, TERMINATED};
 use crate::DocId;
-use std::cmp::Ordering;
 
 /// A `BitSetDocSet` makes it possible to iterate through a bitset as if it was a `DocSet`.
 ///
@@ -33,75 +32,50 @@ impl From<BitSet> for BitSetDocSet {
         } else {
             docs.tinyset(0)
         };
-        BitSetDocSet {
+        let mut docset = BitSetDocSet {
             docs,
             cursor_bucket: 0,
             cursor_tinybitset: first_tiny_bitset,
             doc: 0u32,
-        }
+        };
+        docset.advance();
+        docset
     }
 }
 
 impl DocSet for BitSetDocSet {
-    fn advance(&mut self) -> bool {
+    fn advance(&mut self) -> DocId {
         if let Some(lower) = self.cursor_tinybitset.pop_lowest() {
             self.doc = (self.cursor_bucket as u32 * 64u32) | lower;
-            return true;
+            return self.doc;
         }
         if let Some(cursor_bucket) = self.docs.first_non_empty_bucket(self.cursor_bucket + 1) {
             self.go_to_bucket(cursor_bucket);
             let lower = self.cursor_tinybitset.pop_lowest().unwrap();
             self.doc = (cursor_bucket * 64u32) | lower;
-            true
+            self.doc
         } else {
-            false
+            self.doc = TERMINATED;
+            TERMINATED
         }
     }
 
-    fn skip_next(&mut self, target: DocId) -> SkipResult {
-        // skip is required to advance.
-        if !self.advance() {
-            return SkipResult::End;
-        }
+    fn seek(&mut self, target: DocId) -> DocId {
         let target_bucket = target / 64u32;
 
         // Mask for all of the bits greater or equal
         // to our target document.
-        match target_bucket.cmp(&self.cursor_bucket) {
-            Ordering::Greater => {
-                self.go_to_bucket(target_bucket);
-                let greater_filter: TinySet = TinySet::range_greater_or_equal(target);
-                self.cursor_tinybitset = self.cursor_tinybitset.intersect(greater_filter);
-                if !self.advance() {
-                    SkipResult::End
-                } else if self.doc() == target {
-                    SkipResult::Reached
-                } else {
-                    debug_assert!(self.doc() > target);
-                    SkipResult::OverStep
-                }
-            }
-            Ordering::Equal => loop {
-                match self.doc().cmp(&target) {
-                    Ordering::Less => {
-                        if !self.advance() {
-                            return SkipResult::End;
-                        }
-                    }
-                    Ordering::Equal => {
-                        return SkipResult::Reached;
-                    }
-                    Ordering::Greater => {
-                        debug_assert!(self.doc() > target);
-                        return SkipResult::OverStep;
-                    }
-                }
-            },
-            Ordering::Less => {
-                debug_assert!(self.doc() > target);
-                SkipResult::OverStep
-            }
+        if target_bucket > self.cursor_bucket {
+            self.go_to_bucket(target_bucket);
+            let greater_filter: TinySet = TinySet::range_greater_or_equal(target);
+            self.cursor_tinybitset = self.cursor_tinybitset.intersect(greater_filter);
+            self.advance();
         }
+        let mut doc = self.doc();
+        while doc < target {
+            doc = self.advance();
+        }
+        doc
     }
 
     /// Returns the current document
@@ -122,7 +96,7 @@ impl DocSet for BitSetDocSet {
 mod tests {
     use super::BitSetDocSet;
     use crate::common::BitSet;
-    use crate::docset::{DocSet, SkipResult};
+    use crate::docset::{DocSet, TERMINATED};
     use crate::DocId;
 
     fn create_docbitset(docs: &[DocId], max_doc: DocId) -> BitSetDocSet {
@@ -133,19 +107,24 @@ mod tests {
         BitSetDocSet::from(docset)
     }
 
+    #[test]
+    fn test_empty() {
+        let bitset = BitSet::with_max_value(1000);
+        let mut empty = BitSetDocSet::from(bitset);
+        assert_eq!(empty.advance(), TERMINATED)
+    }
+
     fn test_go_through_sequential(docs: &[DocId]) {
         let mut docset = create_docbitset(docs, 1_000u32);
         for &doc in docs {
-            assert!(docset.advance());
             assert_eq!(doc, docset.doc());
+            docset.advance();
         }
-        assert!(!docset.advance());
-        assert!(!docset.advance());
+        assert_eq!(docset.advance(), TERMINATED);
     }
 
     #[test]
     fn test_docbitset_sequential() {
-        test_go_through_sequential(&[]);
         test_go_through_sequential(&[1, 2, 3]);
         test_go_through_sequential(&[1, 2, 3, 4, 5, 63, 64, 65]);
         test_go_through_sequential(&[63, 64, 65]);
@@ -156,64 +135,64 @@ mod tests {
     fn test_docbitset_skip() {
         {
             let mut docset = create_docbitset(&[1, 5, 6, 7, 5112], 10_000);
-            assert_eq!(docset.skip_next(7), SkipResult::Reached);
+            assert_eq!(docset.seek(7), 7);
             assert_eq!(docset.doc(), 7);
-            assert!(docset.advance(), 7);
+            assert_eq!(docset.advance(), 5112);
             assert_eq!(docset.doc(), 5112);
-            assert!(!docset.advance());
+            assert_eq!(docset.advance(), TERMINATED);
         }
         {
             let mut docset = create_docbitset(&[1, 5, 6, 7, 5112], 10_000);
-            assert_eq!(docset.skip_next(3), SkipResult::OverStep);
+            assert_eq!(docset.seek(3), 5);
             assert_eq!(docset.doc(), 5);
-            assert!(docset.advance());
+            assert_eq!(docset.advance(), 6);
         }
         {
             let mut docset = create_docbitset(&[5112], 10_000);
-            assert_eq!(docset.skip_next(5112), SkipResult::Reached);
+            assert_eq!(docset.seek(5112), 5112);
             assert_eq!(docset.doc(), 5112);
-            assert!(!docset.advance());
+            assert_eq!(docset.advance(), TERMINATED);
         }
         {
             let mut docset = create_docbitset(&[5112], 10_000);
-            assert_eq!(docset.skip_next(5113), SkipResult::End);
-            assert!(!docset.advance());
+            assert_eq!(docset.seek(5113), TERMINATED);
+            assert_eq!(docset.advance(), TERMINATED);
         }
         {
             let mut docset = create_docbitset(&[5112], 10_000);
-            assert_eq!(docset.skip_next(5111), SkipResult::OverStep);
+            assert_eq!(docset.seek(5111), 5112);
             assert_eq!(docset.doc(), 5112);
-            assert!(!docset.advance());
+            assert_eq!(docset.advance(), TERMINATED);
         }
         {
             let mut docset = create_docbitset(&[1, 5, 6, 7, 5112, 5500, 6666], 10_000);
-            assert_eq!(docset.skip_next(5112), SkipResult::Reached);
+            assert_eq!(docset.seek(5112), 5112);
             assert_eq!(docset.doc(), 5112);
-            assert!(docset.advance());
+            assert_eq!(docset.advance(), 5500);
             assert_eq!(docset.doc(), 5500);
-            assert!(docset.advance());
+            assert_eq!(docset.advance(), 6666);
             assert_eq!(docset.doc(), 6666);
-            assert!(!docset.advance());
+            assert_eq!(docset.advance(), TERMINATED);
         }
         {
             let mut docset = create_docbitset(&[1, 5, 6, 7, 5112, 5500, 6666], 10_000);
-            assert_eq!(docset.skip_next(5111), SkipResult::OverStep);
+            assert_eq!(docset.seek(5111), 5112);
             assert_eq!(docset.doc(), 5112);
-            assert!(docset.advance());
+            assert_eq!(docset.advance(), 5500);
             assert_eq!(docset.doc(), 5500);
-            assert!(docset.advance());
+            assert_eq!(docset.advance(), 6666);
             assert_eq!(docset.doc(), 6666);
-            assert!(!docset.advance());
+            assert_eq!(docset.advance(), TERMINATED);
         }
         {
             let mut docset = create_docbitset(&[1, 5, 6, 7, 5112, 5513, 6666], 10_000);
-            assert_eq!(docset.skip_next(5111), SkipResult::OverStep);
+            assert_eq!(docset.seek(5111), 5112);
             assert_eq!(docset.doc(), 5112);
-            assert!(docset.advance());
+            assert_eq!(docset.advance(), 5513);
             assert_eq!(docset.doc(), 5513);
-            assert!(docset.advance());
+            assert_eq!(docset.advance(), 6666);
             assert_eq!(docset.doc(), 6666);
-            assert!(!docset.advance());
+            assert_eq!(docset.advance(), TERMINATED);
         }
     }
 }
@@ -223,6 +202,7 @@ mod bench {
 
     use super::BitSet;
     use super::BitSetDocSet;
+    use crate::docset::TERMINATED;
     use crate::test;
     use crate::tests;
     use crate::DocSet;
@@ -257,7 +237,7 @@ mod bench {
         }
         b.iter(|| {
             let mut docset = BitSetDocSet::from(bitset.clone());
-            while docset.advance() {}
+            while docset.advance() != TERMINATED {}
         });
     }
 }

src/query/block_max_scorer.rs

@@ -1,31 +0,0 @@
-use crate::docset::DocSet;
-use crate::DocId;
-use crate::Score;
-use downcast_rs::impl_downcast;
-use crate::query::Scorer;
-
-/// A set of documents matching a query within a specific segment
-/// and having a maximum score within certain blocks.
-///
-/// See [`Query`](./trait.Query.html) and [`Scorer`](./trait.Scorer.html).
-pub trait BlockMaxScorer: downcast_rs::Downcast + DocSet + Scorer + 'static {
-    /// Returns the maximum score within the current block.
-    ///
-    /// The blocks are defined when indexing. For example, blocks can be
-    /// have a specific number postings each, or can be optimized for
-    /// retrieval speed. Read more in
-    /// [Faster BlockMax WAND with Variable-sized Blocks][vbmw]
-    ///
-    /// This method will perform a bit of computation and is not cached.
-    ///
-    /// [vbmw]: https://dl.acm.org/doi/abs/10.1145/3077136.3080780
-    fn block_max_score(&mut self) -> Score;
-
-    /// Returns the last document in the current block.
-    fn block_max_doc(&mut self) -> DocId;
-
-    /// Returns the maximum possible score within the entire document set.
-    fn max_score(&self) -> Score;
-}
-
-impl_downcast!(BlockMaxScorer);

src/query/block_max_wand.rs

@@ -1,613 +0,0 @@
-use crate::docset::{DocSet, SkipResult};
-use crate::query::score_combiner::ScoreCombiner;
-use crate::query::{BlockMaxScorer, Scorer};
-use crate::DocId;
-use crate::Score;
-use crate::query::weight::PruningScorer;
-
-#[derive(Debug, Copy, Clone, PartialEq, Eq)]
-struct Pivot {
-    position: usize,
-    first_occurrence: usize,
-    doc: DocId,
-}
-
-
-/// Find the position in the sorted list of posting lists of the **pivot**.
-///
-/// docsets need to be advanced, and are required to be sorted by the doc they point to.
-///
-/// The pivot is then defined as the lowest DocId that has a chance of matching our condition.
-fn find_pivot_position<'a, TScorer>(
-    mut docsets: impl Iterator<Item = &'a TScorer>,
-    lower_bound_score: Score,
-) -> Option<Pivot>
-    where TScorer: BlockMaxScorer
-{
-    let mut position = 0;
-    let mut upper_bound = Score::default();
-    while let Some(docset) = docsets.next() {
-        upper_bound += docset.max_score();
-        if lower_bound_score < upper_bound {
-            let pivot_doc = docset.doc();
-            let first_occurrence = position;
-            while let Some(docset) = docsets.next() {
-                if docset.doc() != pivot_doc {
-                    break;
-                } else {
-                    position += 1;
-                }
-            }
-            return Some(Pivot {
-                position,
-                doc: pivot_doc,
-                first_occurrence,
-            });
-        }
-        position += 1;
-    }
-    None
-}
-
-/// Given an iterator over all ordered lists up to the pivot (inclusive) and the following list (if
-/// exists), it returns the next document ID that can be possibly relevant, based on the block max
-/// scores.
-fn find_next_relevant_doc<T, TScorer>(
-    docsets_up_to_pivot: &mut [T],
-    pivot_docset: &mut T,
-    docset_after_pivot: Option<&mut T>,
-) -> DocId
-    where
-        T: AsMut<TScorer>,
-        TScorer: BlockMaxScorer + Scorer,
-{
-    let pivot_docset = pivot_docset.as_mut();
-    let mut next_doc = 1 + docsets_up_to_pivot
-        .iter_mut()
-        .map(|docset| docset.as_mut().block_max_doc())
-        .chain(std::iter::once(pivot_docset.block_max_doc()))
-        .min()
-        .unwrap();
-    if let Some(docset) = docset_after_pivot {
-        let doc = docset.as_mut().doc();
-        if doc < next_doc {
-            next_doc = doc;
-        }
-    }
-    if next_doc <= pivot_docset.doc() {
-        pivot_docset.doc() + 1
-    } else {
-        next_doc
-    }
-}
-
-/// Sifts down the first element of the slice.
-///
-/// `docsets[1..]` are assumed sorted.
-/// This function swaps `docsets[0]` with its right
-/// neighbor successively -bubble sort style- until it reaches the first
-/// position such that `docsets` is sorted.
-fn sift_down<T, TScorer>(docsets: &mut [T])
-    where
-        T: AsRef<TScorer>,
-        TScorer: BlockMaxScorer + Scorer,
-{
-    for idx in 1..docsets.len() {
-        if docsets[idx].as_ref().doc() >= docsets[idx - 1].as_ref().doc() {
-            return;
-        }
-        docsets.swap(idx, idx - 1);
-    }
-}
-
-/// Creates a `DocSet` that iterates through the union of two or more `DocSet`s,
-/// applying [BlockMaxWand] dynamic pruning.
-///
-/// [BlockMaxWand]: https://dl.acm.org/doi/10.1145/2009916.2010048
-pub struct BlockMaxWand<TScorer, TScoreCombiner> {
-    docsets: Vec<Box<TScorer>>,
-    doc: DocId,
-    score: Score,
-    combiner: TScoreCombiner,
-}
-
-impl<TScorer, TScoreCombiner> BlockMaxWand<TScorer, TScoreCombiner>
-    where
-        TScoreCombiner: ScoreCombiner,
-        TScorer: BlockMaxScorer + Scorer,
-{
-    fn new(
-        docsets: Vec<TScorer>,
-        combiner: TScoreCombiner,
-    ) -> BlockMaxWand<TScorer, TScoreCombiner> {
-        let mut non_empty_docsets: Vec<_> = docsets
-            .into_iter()
-            .flat_map(|mut docset| {
-                if docset.advance() {
-                    Some(Box::new(docset))
-                } else {
-                    None
-                }
-            })
-            .collect();
-        non_empty_docsets.sort_by_key(Box::<TScorer>::doc);
-        BlockMaxWand {
-            docsets: non_empty_docsets,
-            combiner,
-            doc: 0u32,
-            score: 0f32
-        }
-    }
-
-    /// Find the position in the sorted list of posting lists of the **pivot**.
-    fn find_pivot_position(&self, lower_bound_score: Score) -> Option<Pivot> {
-        find_pivot_position(
-            self.docsets.iter().map(|docset| docset.as_ref()),
-            lower_bound_score)
-    }
-
-    fn advance_with_pivot(&mut self, pivot: Pivot, lower_bound_score: Score) -> SkipResult {
-        let block_upper_bound: Score = self.docsets[..=pivot.position]
-            .iter_mut()
-            .map(|docset| docset.block_max_score())
-            .sum();
-        if block_upper_bound > lower_bound_score {
-            if pivot.doc == self.docsets[0].doc() {
-                // Since self.docsets is sorted by their current doc, in this branch, all
-                // docsets in [0..=pivot] are positioned on pivot.doc.
-                //
-                // Lets compute the actual score for this doc.
-                //
-                // NOTE(elshize): One additional check needs to be done to improve performance:
-                // update block-wise bound while accumulating score with the actual score,
-                // and check each time if still above threshold.
-                self.combiner.clear();
-                for idx in (0..=pivot.position).rev() {
-                    self.combiner.update(self.docsets[idx].as_mut());
-                    if !self.docsets[idx].advance() {
-                        self.docsets.swap_remove(idx);
-                    }
-                }
-                self.score = self.combiner.score();
-                self.doc = pivot.doc;
-                self.docsets.sort_by_key(Box::<TScorer>::doc);
-                SkipResult::Reached
-            } else {
-                // The substraction does not underflow because otherwise we would go to the other
-                // branch.
-                //
-                // `advanced_idx` is the last idx that is not positionned on the pivot yet.
-                let advanced_idx = pivot.first_occurrence - 1;
-                if !self.docsets[advanced_idx].advance() {
-                    self.docsets.swap_remove(advanced_idx);
-                }
-                if self.docsets.is_empty() {
-                    return SkipResult::End;
-                }
-                sift_down(&mut self.docsets[advanced_idx..]);
-                SkipResult::OverStep
-            }
-        } else {
-            let (up_to_pivot, pivot_and_rest) = self.docsets.split_at_mut(pivot.position as usize);
-            let (pivot, after_pivot) = pivot_and_rest.split_first_mut().unwrap();
-            let next_doc = find_next_relevant_doc(up_to_pivot, pivot, after_pivot.first_mut());
-            // NOTE(elshize): It might be more efficient to advance the list with the higher
-            // max score, but let's advance the first one for now for simplicity.
-            if self.docsets[0].skip_next(next_doc) == SkipResult::End {
-                self.docsets.swap_remove(0);
-            }
-            if self.docsets.is_empty() {
-                return SkipResult::End;
-            }
-            sift_down(&mut self.docsets[..]);
-            SkipResult::OverStep
-        }
-    }
-}
-
-impl<TScorer, TScoreCombiner> PruningScorer
-for BlockMaxWand<TScorer, TScoreCombiner>
-    where
-        TScoreCombiner: ScoreCombiner,
-        TScorer: Scorer + BlockMaxScorer,
-{
-    fn doc(&self) -> DocId {
-        self.doc
-    }
-
-    fn score(&self) -> Score {
-        self.score
-    }
-
-    fn advance_with_pruning(&mut self, lower_bound_score: f32) -> bool {
-        while let Some(pivot) = self.find_pivot_position(lower_bound_score) {
-            match self.advance_with_pivot(pivot, lower_bound_score) {
-                SkipResult::End => { return false },
-                SkipResult::Reached=> { return true; }
-                SkipResult::OverStep => {}
-            }
-        }
-        false
-    }
-}
-
-
-#[cfg(test)]
-mod tests {
-
-    use super::*;
-    use crate::common::HasLen;
-    use crate::docset::DocSet;
-    use crate::query::score_combiner::SumCombiner;
-    use crate::query::Union;
-    use crate::query::{BlockMaxScorer, Scorer};
-    use crate::{DocId, Score};
-    use float_cmp::approx_eq;
-    use proptest::strategy::Strategy;
-    use std::cmp::Ordering;
-    use std::num::Wrapping;
-    use crate::collector::{SegmentCollector, TopScoreSegmentCollector};
-
-    /*
-    #[derive(Debug, Clone)]
-    pub struct VecDocSet {
-        postings: Vec<(DocId, Score)>,
-        cursor: Wrapping<usize>,
-        block_max_scores: Vec<(DocId, Score)>,
-        max_score: Score,
-        block_size: usize,
-    }
-
-    impl VecDocSet {
-        fn new(postings: Vec<(DocId, Score)>, block_size: usize) -> VecDocSet {
-            let block_max_scores: Vec<(DocId, f32)> = postings
-                .chunks(block_size)
-                .into_iter()
-                .map(|block| {
-                    (
-                        block.iter().last().unwrap().0,
-                        block
-                            .iter()
-                            .map(|(_, s)| *s)
-                            .fold(-f32::INFINITY, |left, right| left.max(right))
-                    )
-                })
-                .collect();
-            let max_score = block_max_scores
-                .iter()
-                .copied()
-                .map(|(_, s)| s)
-                .fold(-f32::INFINITY, |left, right| left.max(right));
-            VecDocSet {
-                postings,
-                cursor: Wrapping(0_usize) - Wrapping(1_usize),
-                block_max_scores,
-                max_score,
-                block_size,
-            }
-        }
-        /// Constructs a new set and advances it.
-        fn started(postings: Vec<(DocId, Score)>, block_size: usize) -> VecDocSet {
-            let mut docset = VecDocSet::new(postings, block_size);
-            docset.advance();
-            docset
-        }
-    }
-
-    impl DocSet for VecDocSet {
-        fn advance(&mut self) -> bool {
-            self.cursor += Wrapping(1);
-            self.postings.len() > self.cursor.0
-        }
-
-        fn doc(&self) -> DocId {
-            self.postings[self.cursor.0].0
-        }
-
-        fn size_hint(&self) -> u32 {
-            self.len() as u32
-        }
-    }
-
-    impl HasLen for VecDocSet {
-        fn len(&self) -> usize {
-            self.postings.len()
-        }
-    }
-
-    impl BlockMaxScorer for VecDocSet {
-        fn max_score(&self) -> Score {
-            self.max_score
-        }
-        fn block_max_score(&mut self) -> Score {
-            self.block_max_scores[self.cursor.0 / self.block_size].1
-        }
-        fn block_max_doc(&mut self) -> DocId {
-            self.block_max_scores[self.cursor.0 / self.block_size].0
-        }
-    }
-
-    impl Scorer for VecDocSet {
-        fn score(&mut self) -> Score {
-            self.postings[self.cursor.0].1
-        }
-    }
-
-    #[derive(Debug)]
-    struct ComparableDoc<T, D> {
-        feature: T,
-        doc: D,
-    }
-
-    impl<T: PartialOrd, D: PartialOrd> PartialOrd for ComparableDoc<T, D> {
-        fn partial_cmp(&self, other: &Self) -> Option<Ordering> {
-            Some(self.cmp(other))
-        }
-    }
-
-    impl<T: PartialOrd, D: PartialOrd> Ord for ComparableDoc<T, D> {
-        #[inline]
-        fn cmp(&self, other: &Self) -> Ordering {
-            // Reversed to make BinaryHeap work as a min-heap
-            let by_feature = other
-                .feature
-                .partial_cmp(&self.feature)
-                .unwrap_or(Ordering::Equal);
-
-            let lazy_by_doc_address =
-                || self.doc.partial_cmp(&other.doc).unwrap_or(Ordering::Equal);
-
-            // In case of a tie on the feature, we sort by ascending
-            // `DocAddress` in order to ensure a stable sorting of the
-            // documents.
-            by_feature.then_with(lazy_by_doc_address)
-        }
-    }
-
-    impl<T: PartialOrd, D: PartialOrd> PartialEq for ComparableDoc<T, D> {
-        fn eq(&self, other: &Self) -> bool {
-            self.cmp(other) == Ordering::Equal
-        }
-    }
-
-    impl<T: PartialOrd, D: PartialOrd> Eq for ComparableDoc<T, D> {}
-
-    fn union_vs_bmw(posting_lists: Vec<VecDocSet>) {
-        let mut union = Union::<VecDocSet, SumCombiner>::from(posting_lists.clone());
-        let mut top_union = TopScoreSegmentCollector::new(0, 10);
-        while union.advance() {
-            top_union.collect(union.doc(), union.score());
-        }
-        let top_bmw = TopScoreSegmentCollector::new(0, 10 );
-        let mut bmw = BlockMaxWand::new(posting_lists, SumCombiner::default());
-        let top_docs_bnw = top_bmw.collect_scorer(&mut bmw, None);
-        for ((expected_score, expected_doc), (actual_score, actual_doc)) in
-        top_union.harvest().into_iter().zip( top_docs_bnw )
-        {
-            assert!(approx_eq!(
-                f32,
-                expected_score,
-                actual_score,
-                epsilon = 0.0001
-            ));
-            assert_eq!(expected_doc, actual_doc);
-        }
-    }
-
-    #[test]
-    fn test_bmw_0() {
-        union_vs_bmw(vec![
-            VecDocSet {
-                postings: vec![
-                    (0, 1.0),
-                    (23, 1.0),
-                    (28, 1.0),
-                    (56, 1.0),
-                    (59, 1.0),
-                    (66, 1.0),
-                    (93, 1.0),
-                ],
-                cursor: Wrapping(0_usize) - Wrapping(1_usize),
-                block_max_scores: vec![(93, 1.0)],
-                max_score: 1.0,
-                block_size: 16,
-            },
-            VecDocSet {
-                postings: vec![
-                    (2, 1.6549665),
-                    (43, 2.6958032),
-                    (53, 3.5309567),
-                    (71, 2.7688136),
-                    (87, 3.4279852),
-                    (96, 3.9028034),
-                ],
-                cursor: Wrapping(0_usize) - Wrapping(1_usize),
-                block_max_scores: vec![(96, 3.9028034)],
-                max_score: 3.9028034,
-                block_size: 16,
-            },
-        ])
-    }
-
-    #[test]
-    fn test_bmw_1() {
-        union_vs_bmw(vec![
-            VecDocSet {
-                postings: vec![(73, 1.0), (82, 1.0)],
-                cursor: Wrapping(0_usize) - Wrapping(1_usize),
-                block_max_scores: vec![(82, 1.0)],
-                max_score: 1.0,
-                block_size: 16,
-            },
-            VecDocSet {
-                postings: vec![
-                    (21, 3.582513),
-                    (23, 1.6928024),
-                    (27, 3.887647),
-                    (42, 1.5469292),
-                    (61, 1.7317574),
-                    (62, 1.2968783),
-                    (82, 2.4040694),
-                    (85, 3.1487892),
-                ],
-                cursor: Wrapping(0_usize) - Wrapping(1_usize),
-                block_max_scores: vec![(85, 3.887647)],
-                max_score: 3.887647,
-                block_size: 16,
-            },
-        ])
-    }
-
-    proptest::proptest! {
-        #[test]
-        fn test_union_vs_bmw(postings in proptest::collection::vec(
-            proptest::collection::vec(0_u32..100, 1..10)
-                .prop_flat_map(|v| {
-                    let scores = proptest::collection::vec(1_f32..4_f32, v.len()..=v.len());
-                    scores.prop_map(move |s| {
-                        let mut postings: Vec<_> = v.iter().copied().zip(s.iter().copied()).collect();
-                        postings.sort_by_key(|p| p.0);
-                        postings.dedup_by_key(|p| p.0);
-                        VecDocSet::new(postings, 16)
-                    })
-                }),
-            2..5)
-        ) {
-            union_vs_bmw(postings);
-        }
-    }
-
-    #[test]
-    fn test_find_pivot_position() {
-        let postings = vec![
-            VecDocSet::started(vec![(0, 2.0)], 1),
-            VecDocSet::started(vec![(1, 3.0)], 1),
-            VecDocSet::started(vec![(2, 4.0)], 1),
-            VecDocSet::started(vec![(3, 5.0)], 1),
-            VecDocSet::started(vec![(3, 6.0)], 1),
-        ];
-        assert_eq!(
-            find_pivot_position(postings.iter(), 2.0f32),
-            Some(Pivot {
-                position: 1,
-                doc: 1,
-                first_occurrence: 1,
-            })
-        );
-        assert_eq!(
-            find_pivot_position(postings.iter(), 5.0f32),
-            Some(Pivot {
-                position: 2,
-                doc: 2,
-                first_occurrence: 2,
-            })
-        );
-        assert_eq!(
-            find_pivot_position(postings.iter(), 9.0f32),
-            Some(Pivot {
-                position: 4,
-                doc: 3,
-                first_occurrence: 3,
-            })
-        );
-        assert_eq!(
-            find_pivot_position(postings.iter(), 20.0f32),
-            None
-        );
-    }
-
-    #[test]
-    fn test_find_next_relevant_doc_before_pivot() {
-        let mut postings = vec![
-            Box::new(VecDocSet::started(vec![(0, 0.0), (3, 0.0)], 2)),
-            Box::new(VecDocSet::started(vec![(1, 0.0), (4, 0.0)], 2)),
-            Box::new(VecDocSet::started(vec![(2, 0.0), (6, 0.0)], 2)), // pivot
-            Box::new(VecDocSet::started(vec![(6, 0.0), (8, 0.0)], 2)),
-            Box::new(VecDocSet::started(vec![(6, 0.0), (8, 0.0)], 2)),
-        ];
-        let (up_to_pivot, rest) = postings.split_at_mut(2);
-        let (pivot, after_pivot) = rest.split_first_mut().unwrap();
-        let next_doc = find_next_relevant_doc(up_to_pivot, pivot, Some(&mut after_pivot[0]));
-        assert_eq!(next_doc, 4);
-    }
-
-    #[test]
-    fn test_find_next_relevant_doc_prefix_smaller_than_pivot() {
-        let mut postings = vec![
-            Box::new(VecDocSet::started(vec![(0, 0.0), (3, 0.0)], 2)),
-            Box::new(VecDocSet::started(vec![(1, 0.0), (4, 0.0)], 2)),
-            Box::new(VecDocSet::started(vec![(5, 0.0), (8, 0.0)], 2)), // pivot
-            Box::new(VecDocSet::started(vec![(6, 0.0), (8, 0.0)], 2)),
-            Box::new(VecDocSet::started(vec![(6, 0.0), (8, 0.0)], 2)),
-        ];
-        let (up_to_pivot, rest) = postings.split_at_mut(2);
-        let (pivot, after_pivot) = rest.split_first_mut().unwrap();
-        let next_doc = find_next_relevant_doc(up_to_pivot, pivot, Some(&mut after_pivot[0]));
-        assert_eq!(next_doc, 6);
-    }
-
-    #[test]
-    fn test_find_next_relevant_doc_after_pivot() {
-        let mut postings = vec![
-            Box::new(VecDocSet::started(vec![(0, 0.0), (8, 0.0)], 2)),
-            Box::new(VecDocSet::started(vec![(1, 0.0), (8, 0.0)], 2)),
-            Box::new(VecDocSet::started(vec![(2, 0.0), (8, 0.0)], 2)), // pivot
-            Box::new(VecDocSet::started(vec![(5, 0.0), (7, 0.0)], 2)),
-            Box::new(VecDocSet::started(vec![(6, 0.0), (7, 0.0)], 2)),
-        ];
-        let (up_to_pivot, rest) = postings.split_at_mut(2);
-        let (pivot, after_pivot) = rest.split_first_mut().unwrap();
-        let next_doc = find_next_relevant_doc(up_to_pivot, pivot, Some(&mut after_pivot[0]));
-        assert_eq!(next_doc, 5);
-    }
-
-    #[test]
-    fn test_sift_down_already_sifted() {
-        let mut postings = vec![
-            Box::new(VecDocSet::started(vec![(0, 0.0), (8, 0.0)], 2)),
-            Box::new(VecDocSet::started(vec![(1, 0.0), (8, 0.0)], 2)),
-            Box::new(VecDocSet::started(vec![(2, 0.0), (8, 0.0)], 2)), // pivot
-            Box::new(VecDocSet::started(vec![(5, 0.0), (7, 0.0)], 2)),
-            Box::new(VecDocSet::started(vec![(6, 0.0), (7, 0.0)], 2)),
-        ];
-        sift_down(&mut postings[2..]);
-        assert_eq!(
-            postings.into_iter().map(|p| p.doc()).collect::<Vec<_>>(),
-            vec![0, 1, 2, 5, 6]
-        );
-    }
-
-    #[test]
-    fn test_sift_down_sift_one_down() {
-        let mut postings = vec![
-            Box::new(VecDocSet::started(vec![(0, 0.0), (8, 0.0)], 2)),
-            Box::new(VecDocSet::started(vec![(1, 0.0), (8, 0.0)], 2)),
-            Box::new(VecDocSet::started(vec![(6, 0.0), (8, 0.0)], 2)), // pivot
-            Box::new(VecDocSet::started(vec![(5, 0.0), (7, 0.0)], 2)),
-            Box::new(VecDocSet::started(vec![(7, 0.0), (7, 0.0)], 2)),
-        ];
-        sift_down(&mut postings[2..]);
-        assert_eq!(
-            postings.into_iter().map(|p| p.doc()).collect::<Vec<_>>(),
-            vec![0, 1, 5, 6, 7]
-        );
-    }
-
-    #[test]
-    fn test_sift_down_to_bottom() {
-        let mut postings = vec![
-            Box::new(VecDocSet::started(vec![(0, 0.0), (8, 0.0)], 2)),
-            Box::new(VecDocSet::started(vec![(1, 0.0), (8, 0.0)], 2)),
-            Box::new(VecDocSet::started(vec![(7, 0.0), (8, 0.0)], 2)), // pivot
-            Box::new(VecDocSet::started(vec![(5, 0.0), (7, 0.0)], 2)),
-            Box::new(VecDocSet::started(vec![(6, 0.0), (7, 0.0)], 2)),
-        ];
-        sift_down(&mut postings[2..]);
-        assert_eq!(
-            postings.into_iter().map(|p| p.doc()).collect::<Vec<_>>(),
-            vec![0, 1, 5, 6, 7]
-        );
-    }
-    
-     */
-}

src/query/boolean_query/boolean_weight.rs

@@ -10,7 +10,7 @@ use crate::query::Scorer;
 use crate::query::Union;
 use crate::query::Weight;
 use crate::query::{intersect_scorers, Explanation};
-use crate::{DocId, SkipResult};
+use crate::DocId;
 use std::collections::HashMap;
 
 fn scorer_union<TScoreCombiner>(scorers: Vec<Box<dyn Scorer>>) -> Box<dyn Scorer>
@@ -133,7 +133,7 @@ impl Weight for BooleanWeight {
 
     fn explain(&self, reader: &SegmentReader, doc: DocId) -> crate::Result<Explanation> {
         let mut scorer = self.scorer(reader, 1.0f32)?;
-        if scorer.skip_next(doc) != SkipResult::Reached {
+        if scorer.seek(doc) != doc {
             return Err(does_not_match(doc));
         }
         if !self.scoring_enabled {

src/query/boolean_query/mod.rs

@@ -207,7 +207,6 @@ mod tests {
             let mut boolean_scorer = boolean_weight
                 .scorer(searcher.segment_reader(0u32), 1.0f32)
                 .unwrap();
-            assert!(boolean_scorer.advance());
             assert_eq!(boolean_scorer.doc(), 0u32);
             assert_nearly_equals(boolean_scorer.score(), 0.84163445f32);
         }
@@ -215,7 +214,6 @@ mod tests {
             let mut boolean_scorer = boolean_weight
                 .scorer(searcher.segment_reader(0u32), 2.0f32)
                 .unwrap();
-            assert!(boolean_scorer.advance());
             assert_eq!(boolean_scorer.doc(), 0u32);
             assert_nearly_equals(boolean_scorer.score(), 1.6832689f32);
         }

src/query/boost_query.rs

@@ -1,8 +1,7 @@
-use crate::common::BitSet;
 use crate::fastfield::DeleteBitSet;
 use crate::query::explanation::does_not_match;
 use crate::query::{Explanation, Query, Scorer, Weight};
-use crate::{DocId, DocSet, Searcher, SegmentReader, SkipResult, Term};
+use crate::{DocId, DocSet, Searcher, SegmentReader, Term};
 use std::collections::BTreeSet;
 use std::fmt;
 
@@ -72,7 +71,7 @@ impl Weight for BoostWeight {
 
     fn explain(&self, reader: &SegmentReader, doc: u32) -> crate::Result<Explanation> {
         let mut scorer = self.scorer(reader, 1.0f32)?;
-        if scorer.skip_next(doc) != SkipResult::Reached {
+        if scorer.seek(doc) != doc {
             return Err(does_not_match(doc));
         }
         let mut explanation =
@@ -99,12 +98,12 @@ impl<S: Scorer> BoostScorer<S> {
 }
 
 impl<S: Scorer> DocSet for BoostScorer<S> {
-    fn advance(&mut self) -> bool {
+    fn advance(&mut self) -> DocId {
         self.underlying.advance()
     }
 
-    fn skip_next(&mut self, target: DocId) -> SkipResult {
-        self.underlying.skip_next(target)
+    fn seek(&mut self, target: DocId) -> DocId {
+        self.underlying.seek(target)
     }
 
     fn fill_buffer(&mut self, buffer: &mut [DocId]) -> usize {
@@ -119,10 +118,6 @@ impl<S: Scorer> DocSet for BoostScorer<S> {
         self.underlying.size_hint()
     }
 
-    fn append_to_bitset(&mut self, bitset: &mut BitSet) {
-        self.underlying.append_to_bitset(bitset)
-    }
-
     fn count(&mut self, delete_bitset: &DeleteBitSet) -> u32 {
         self.underlying.count(delete_bitset)
     }

src/query/empty_query.rs

@@ -1,4 +1,5 @@
 use super::Scorer;
+use crate::docset::TERMINATED;
 use crate::query::explanation::does_not_match;
 use crate::query::Weight;
 use crate::query::{Explanation, Query};
@@ -48,15 +49,12 @@ impl Weight for EmptyWeight {
 pub struct EmptyScorer;
 
 impl DocSet for EmptyScorer {
-    fn advance(&mut self) -> bool {
-        false
+    fn advance(&mut self) -> DocId {
+        TERMINATED
     }
 
     fn doc(&self) -> DocId {
-        panic!(
-            "You may not call .doc() on a scorer \
-             where the last call to advance() did not return true."
-        );
+        TERMINATED
     }
 
     fn size_hint(&self) -> u32 {
@@ -72,18 +70,15 @@ impl Scorer for EmptyScorer {
 
 #[cfg(test)]
 mod tests {
+    use crate::docset::TERMINATED;
     use crate::query::EmptyScorer;
     use crate::DocSet;
 
     #[test]
     fn test_empty_scorer() {
         let mut empty_scorer = EmptyScorer;
-        assert!(!empty_scorer.advance());
-    }
-
-    #[test]
-    #[should_panic]
-    fn test_empty_scorer_panic_on_doc_call() {
-        EmptyScorer.doc();
+        assert_eq!(empty_scorer.doc(), TERMINATED);
+        assert_eq!(empty_scorer.advance(), TERMINATED);
+        assert_eq!(empty_scorer.doc(), TERMINATED);
     }
 }

src/query/exclude.rs

@@ -1,41 +1,37 @@
-use crate::docset::{DocSet, SkipResult};
+use crate::docset::{DocSet, TERMINATED};
 use crate::query::Scorer;
 use crate::DocId;
 use crate::Score;
 
-#[derive(Clone, Copy, Debug)]
-enum State {
-    ExcludeOne(DocId),
-    Finished,
-}
-
 /// Filters a given `DocSet` by removing the docs from a given `DocSet`.
 ///
 /// The excluding docset has no impact on scoring.
 pub struct Exclude<TDocSet, TDocSetExclude> {
     underlying_docset: TDocSet,
     excluding_docset: TDocSetExclude,
-    excluding_state: State,
 }
 
 impl<TDocSet, TDocSetExclude> Exclude<TDocSet, TDocSetExclude>
 where
+    TDocSet: DocSet,
     TDocSetExclude: DocSet,
 {
     /// Creates a new `ExcludeScorer`
     pub fn new(
-        underlying_docset: TDocSet,
+        mut underlying_docset: TDocSet,
         mut excluding_docset: TDocSetExclude,
     ) -> Exclude<TDocSet, TDocSetExclude> {
-        let state = if excluding_docset.advance() {
-            State::ExcludeOne(excluding_docset.doc())
-        } else {
-            State::Finished
-        };
+        while underlying_docset.doc() != TERMINATED {
+            let target = underlying_docset.doc();
+            if excluding_docset.seek(target) != target {
+                // this document is not excluded.
+                break;
+            }
+            underlying_docset.advance();
+        }
         Exclude {
             underlying_docset,
             excluding_docset,
-            excluding_state: state,
         }
     }
 }
@@ -51,28 +47,7 @@ where
     /// increasing `doc`.
     fn accept(&mut self) -> bool {
         let doc = self.underlying_docset.doc();
-        match self.excluding_state {
-            State::ExcludeOne(excluded_doc) => {
-                if doc == excluded_doc {
-                    return false;
-                }
-                if excluded_doc > doc {
-                    return true;
-                }
-                match self.excluding_docset.skip_next(doc) {
-                    SkipResult::OverStep => {
-                        self.excluding_state = State::ExcludeOne(self.excluding_docset.doc());
-                        true
-                    }
-                    SkipResult::End => {
-                        self.excluding_state = State::Finished;
-                        true
-                    }
-                    SkipResult::Reached => false,
-                }
-            }
-            State::Finished => true,
-        }
+        self.excluding_docset.seek(doc) != doc
     }
 }
 
@@ -81,27 +56,24 @@ where
     TDocSet: DocSet,
     TDocSetExclude: DocSet,
 {
-    fn advance(&mut self) -> bool {
-        while self.underlying_docset.advance() {
+    fn advance(&mut self) -> DocId {
+        while self.underlying_docset.advance() != TERMINATED {
             if self.accept() {
-                return true;
+                return self.doc();
             }
         }
-        false
+        TERMINATED
     }
 
-    fn skip_next(&mut self, target: DocId) -> SkipResult {
-        let underlying_skip_result = self.underlying_docset.skip_next(target);
-        if underlying_skip_result == SkipResult::End {
-            return SkipResult::End;
+    fn seek(&mut self, target: DocId) -> DocId {
+        let underlying_seek_result = self.underlying_docset.seek(target);
+        if underlying_seek_result == TERMINATED {
+            return TERMINATED;
         }
         if self.accept() {
-            underlying_skip_result
-        } else if self.advance() {
-            SkipResult::OverStep
-        } else {
-            SkipResult::End
+            return underlying_seek_result;
         }
+        self.advance()
     }
 
     fn doc(&self) -> DocId {
@@ -141,8 +113,9 @@ mod tests {
             VecDocSet::from(vec![1, 2, 3, 10, 16, 24]),
         );
         let mut els = vec![];
-        while exclude_scorer.advance() {
+        while exclude_scorer.doc() != TERMINATED {
             els.push(exclude_scorer.doc());
+            exclude_scorer.advance();
         }
         assert_eq!(els, vec![5, 8, 15]);
     }

src/query/fuzzy_query.rs

@@ -117,7 +117,7 @@ impl FuzzyTermQuery {
         }
     }
 
-    /// Creates a new Fuzzy Query that treats transpositions as cost one rather than two
+    /// Creates a new Fuzzy Query of the Term prefix
     pub fn new_prefix(term: Term, distance: u8, transposition_cost_one: bool) -> FuzzyTermQuery {
         FuzzyTermQuery {
             term,
@@ -188,6 +188,8 @@ mod test {
         }
         let reader = index.reader().unwrap();
         let searcher = reader.searcher();
+
+        // passes because Levenshtein distance is 1 (substitute 'o' with 'a')
         {
             let term = Term::from_field_text(country_field, "japon");
 
@@ -200,6 +202,18 @@ mod test {
             assert_nearly_equals(1f32, score);
         }
 
+        // fails because non-prefix Levenshtein distance is more than 1 (add 'a' and 'n')
+        {
+            let term = Term::from_field_text(country_field, "jap");
+
+            let fuzzy_query = FuzzyTermQuery::new(term, 1, true);
+            let top_docs = searcher
+                .search(&fuzzy_query, &TopDocs::with_limit(2))
+                .unwrap();
+            assert_eq!(top_docs.len(), 0, "Expected no document");
+        }
+
+        // passes because prefix Levenshtein distance is 0
         {
             let term = Term::from_field_text(country_field, "jap");
 

src/query/intersection.rs

@@ -1,4 +1,4 @@
-use crate::docset::{DocSet, SkipResult};
+use crate::docset::{DocSet, TERMINATED};
 use crate::query::term_query::TermScorer;
 use crate::query::EmptyScorer;
 use crate::query::Scorer;
@@ -20,12 +20,14 @@ pub fn intersect_scorers(mut scorers: Vec<Box<dyn Scorer>>) -> Box<dyn Scorer> {
     if scorers.len() == 1 {
         return scorers.pop().unwrap();
     }
+    scorers.sort_by_key(|scorer| scorer.size_hint());
+    let doc = go_to_first_doc(&mut scorers[..]);
+    if doc == TERMINATED {
+        return Box::new(EmptyScorer);
+    }
     // We know that we have at least 2 elements.
-    let num_docsets = scorers.len();
-    scorers.sort_by(|left, right| right.size_hint().cmp(&left.size_hint()));
-    let left = scorers.pop().unwrap();
-    let right = scorers.pop().unwrap();
-    scorers.reverse();
+    let left = scorers.remove(0);
+    let right = scorers.remove(0);
     let all_term_scorers = [&left, &right]
         .iter()
         .all(|&scorer| scorer.is::<TermScorer>());
@@ -34,14 +36,12 @@ pub fn intersect_scorers(mut scorers: Vec<Box<dyn Scorer>>) -> Box<dyn Scorer> {
             left: *(left.downcast::<TermScorer>().map_err(|_| ()).unwrap()),
             right: *(right.downcast::<TermScorer>().map_err(|_| ()).unwrap()),
             others: scorers,
-            num_docsets,
         });
     }
     Box::new(Intersection {
         left,
         right,
         others: scorers,
-        num_docsets,
     })
 }
 
@@ -50,22 +50,34 @@ pub struct Intersection<TDocSet: DocSet, TOtherDocSet: DocSet = Box<dyn Scorer>>
     left: TDocSet,
     right: TDocSet,
     others: Vec<TOtherDocSet>,
-    num_docsets: usize,
+}
+
+fn go_to_first_doc<TDocSet: DocSet>(docsets: &mut [TDocSet]) -> DocId {
+    let mut candidate = 0;
+    'outer: loop {
+        for docset in docsets.iter_mut() {
+            let seek_doc = docset.seek(candidate);
+            if seek_doc > candidate {
+                candidate = docset.doc();
+                continue 'outer;
+            }
+        }
+        return candidate;
+    }
 }
 
 impl<TDocSet: DocSet> Intersection<TDocSet, TDocSet> {
     pub(crate) fn new(mut docsets: Vec<TDocSet>) -> Intersection<TDocSet, TDocSet> {
         let num_docsets = docsets.len();
         assert!(num_docsets >= 2);
-        docsets.sort_by(|left, right| right.size_hint().cmp(&left.size_hint()));
-        let left = docsets.pop().unwrap();
-        let right = docsets.pop().unwrap();
-        docsets.reverse();
+        docsets.sort_by_key(|docset| docset.size_hint());
+        go_to_first_doc(&mut docsets);
+        let left = docsets.remove(0);
+        let right = docsets.remove(0);
         Intersection {
             left,
             right,
             others: docsets,
-            num_docsets,
         }
     }
 }
@@ -80,128 +92,49 @@ impl<TDocSet: DocSet> Intersection<TDocSet, TDocSet> {
     }
 }
 
-impl<TDocSet: DocSet, TOtherDocSet: DocSet> Intersection<TDocSet, TOtherDocSet> {
-    pub(crate) fn docset_mut(&mut self, ord: usize) -> &mut dyn DocSet {
-        match ord {
-            0 => &mut self.left,
-            1 => &mut self.right,
-            n => &mut self.others[n - 2],
-        }
-    }
-}
-
 impl<TDocSet: DocSet, TOtherDocSet: DocSet> DocSet for Intersection<TDocSet, TOtherDocSet> {
-    fn advance(&mut self) -> bool {
+    fn advance(&mut self) -> DocId {
         let (left, right) = (&mut self.left, &mut self.right);
-
-        if !left.advance() {
-            return false;
-        }
-
-        let mut candidate = left.doc();
-        let mut other_candidate_ord: usize = usize::max_value();
+        let mut candidate = left.advance();
 
         'outer: loop {
             // In the first part we look for a document in the intersection
             // of the two rarest `DocSet` in the intersection.
+
             loop {
-                match right.skip_next(candidate) {
-                    SkipResult::Reached => {
-                        break;
-                    }
-                    SkipResult::OverStep => {
-                        candidate = right.doc();
-                        other_candidate_ord = usize::max_value();
-                    }
-                    SkipResult::End => {
-                        return false;
-                    }
-                }
-                match left.skip_next(candidate) {
-                    SkipResult::Reached => {
-                        break;
-                    }
-                    SkipResult::OverStep => {
-                        candidate = left.doc();
-                        other_candidate_ord = usize::max_value();
-                    }
-                    SkipResult::End => {
-                        return false;
-                    }
+                let right_doc = right.seek(candidate);
+                candidate = left.seek(right_doc);
+                if candidate == right_doc {
+                    break;
                 }
             }
+
+            debug_assert_eq!(left.doc(), right.doc());
             // test the remaining scorers;
-            for (ord, docset) in self.others.iter_mut().enumerate() {
-                if ord == other_candidate_ord {
-                    continue;
-                }
+            for docset in self.others.iter_mut() {
                 // `candidate_ord` is already at the
                 // right position.
                 //
                 // Calling `skip_next` would advance this docset
                 // and miss it.
-                match docset.skip_next(candidate) {
-                    SkipResult::Reached => {}
-                    SkipResult::OverStep => {
-                        // this is not in the intersection,
-                        // let's update our candidate.
-                        candidate = docset.doc();
-                        match left.skip_next(candidate) {
-                            SkipResult::Reached => {
-                                other_candidate_ord = ord;
-                            }
-                            SkipResult::OverStep => {
-                                candidate = left.doc();
-                                other_candidate_ord = usize::max_value();
-                            }
-                            SkipResult::End => {
-                                return false;
-                            }
-                        }
-                        continue 'outer;
-                    }
-                    SkipResult::End => {
-                        return false;
-                    }
+                let seek_doc = docset.seek(candidate);
+                if seek_doc > candidate {
+                    candidate = left.seek(seek_doc);
+                    continue 'outer;
                 }
             }
-            return true;
+
+            return candidate;
         }
     }
 
-    fn skip_next(&mut self, target: DocId) -> SkipResult {
-        // We optimize skipping by skipping every single member
-        // of the intersection to target.
-        let mut current_target: DocId = target;
-        let mut current_ord = self.num_docsets;
-
-        'outer: loop {
-            for ord in 0..self.num_docsets {
-                let docset = self.docset_mut(ord);
-                if ord == current_ord {
-                    continue;
-                }
-                match docset.skip_next(current_target) {
-                    SkipResult::End => {
-                        return SkipResult::End;
-                    }
-                    SkipResult::OverStep => {
-                        // update the target
-                        // for the remaining members of the intersection.
-                        current_target = docset.doc();
-                        current_ord = ord;
-                        continue 'outer;
-                    }
-                    SkipResult::Reached => {}
-                }
-            }
-            if target == current_target {
-                return SkipResult::Reached;
-            } else {
-                assert!(current_target > target);
-                return SkipResult::OverStep;
-            }
+    fn seek(&mut self, target: DocId) -> DocId {
+        self.left.seek(target);
+        let mut docsets: Vec<&mut dyn DocSet> = vec![&mut self.left, &mut self.right];
+        for docset in &mut self.others {
+            docsets.push(docset);
         }
+        go_to_first_doc(&mut docsets[..])
     }
 
     fn doc(&self) -> DocId {
@@ -228,7 +161,7 @@ where
 #[cfg(test)]
 mod tests {
     use super::Intersection;
-    use crate::docset::{DocSet, SkipResult};
+    use crate::docset::{DocSet, TERMINATED};
     use crate::postings::tests::test_skip_against_unoptimized;
     use crate::query::VecDocSet;
 
@@ -238,20 +171,18 @@ mod tests {
             let left = VecDocSet::from(vec![1, 3, 9]);
             let right = VecDocSet::from(vec![3, 4, 9, 18]);
             let mut intersection = Intersection::new(vec![left, right]);
-            assert!(intersection.advance());
             assert_eq!(intersection.doc(), 3);
-            assert!(intersection.advance());
+            assert_eq!(intersection.advance(), 9);
             assert_eq!(intersection.doc(), 9);
-            assert!(!intersection.advance());
+            assert_eq!(intersection.advance(), TERMINATED);
         }
         {
             let a = VecDocSet::from(vec![1, 3, 9]);
             let b = VecDocSet::from(vec![3, 4, 9, 18]);
             let c = VecDocSet::from(vec![1, 5, 9, 111]);
             let mut intersection = Intersection::new(vec![a, b, c]);
-            assert!(intersection.advance());
             assert_eq!(intersection.doc(), 9);
-            assert!(!intersection.advance());
+            assert_eq!(intersection.advance(), TERMINATED);
         }
     }
 
@@ -260,8 +191,8 @@ mod tests {
         let left = VecDocSet::from(vec![0]);
         let right = VecDocSet::from(vec![0]);
         let mut intersection = Intersection::new(vec![left, right]);
-        assert!(intersection.advance());
         assert_eq!(intersection.doc(), 0);
+        assert_eq!(intersection.advance(), TERMINATED);
     }
 
     #[test]
@@ -269,7 +200,7 @@ mod tests {
         let left = VecDocSet::from(vec![0, 1, 2, 4]);
         let right = VecDocSet::from(vec![2, 5]);
         let mut intersection = Intersection::new(vec![left, right]);
-        assert_eq!(intersection.skip_next(2), SkipResult::Reached);
+        assert_eq!(intersection.seek(2), 2);
         assert_eq!(intersection.doc(), 2);
     }
 
@@ -312,7 +243,7 @@ mod tests {
         let a = VecDocSet::from(vec![1, 3]);
         let b = VecDocSet::from(vec![1, 4]);
         let c = VecDocSet::from(vec![3, 9]);
-        let mut intersection = Intersection::new(vec![a, b, c]);
-        assert!(!intersection.advance());
+        let intersection = Intersection::new(vec![a, b, c]);
+        assert_eq!(intersection.doc(), TERMINATED);
     }
 }

src/query/mod.rs

@@ -3,8 +3,6 @@
 mod all_query;
 mod automaton_weight;
 mod bitset;
-mod block_max_scorer;
-mod block_max_wand;
 mod bm25;
 mod boolean_query;
 mod boost_query;
@@ -37,7 +35,6 @@ pub use self::vec_docset::VecDocSet;
 pub use self::all_query::{AllQuery, AllScorer, AllWeight};
 pub use self::automaton_weight::AutomatonWeight;
 pub use self::bitset::BitSetDocSet;
-pub use self::block_max_scorer::BlockMaxScorer;
 pub use self::boolean_query::BooleanQuery;
 pub use self::boost_query::BoostQuery;
 pub use self::empty_query::{EmptyQuery, EmptyScorer, EmptyWeight};
@@ -59,8 +56,6 @@ pub use self::scorer::Scorer;
 pub use self::term_query::TermQuery;
 pub use self::weight::Weight;
 pub use tantivy_query_grammar::Occur;
-pub use self::weight::PruningScorerIfPossible;
-
 
 #[cfg(test)]
 mod tests {

src/query/phrase_query/mod.rs

@@ -60,8 +60,8 @@ pub mod tests {
                 .map(|docaddr| docaddr.1)
                 .collect::<Vec<_>>()
         };
-        assert_eq!(test_query(vec!["a", "b", "c"]), vec![2, 4]);
         assert_eq!(test_query(vec!["a", "b"]), vec![1, 2, 3, 4]);
+        assert_eq!(test_query(vec!["a", "b", "c"]), vec![2, 4]);
         assert_eq!(test_query(vec!["b", "b"]), vec![0, 1]);
         assert!(test_query(vec!["g", "ewrwer"]).is_empty());
         assert!(test_query(vec!["g", "a"]).is_empty());

src/query/phrase_query/phrase_scorer.rs

@@ -1,4 +1,4 @@
-use crate::docset::{DocSet, SkipResult};
+use crate::docset::{DocSet, TERMINATED};
 use crate::fieldnorm::FieldNormReader;
 use crate::postings::Postings;
 use crate::query::bm25::BM25Weight;
@@ -25,12 +25,12 @@ impl<TPostings: Postings> PostingsWithOffset<TPostings> {
 }
 
 impl<TPostings: Postings> DocSet for PostingsWithOffset<TPostings> {
-    fn advance(&mut self) -> bool {
+    fn advance(&mut self) -> DocId {
         self.postings.advance()
     }
 
-    fn skip_next(&mut self, target: DocId) -> SkipResult {
-        self.postings.skip_next(target)
+    fn seek(&mut self, target: DocId) -> DocId {
+        self.postings.seek(target)
     }
 
     fn doc(&self) -> DocId {
@@ -149,7 +149,7 @@ impl<TPostings: Postings> PhraseScorer<TPostings> {
                 PostingsWithOffset::new(postings, (max_offset - offset) as u32)
             })
             .collect::<Vec<_>>();
-        PhraseScorer {
+        let mut scorer = PhraseScorer {
             intersection_docset: Intersection::new(postings_with_offsets),
             num_terms: num_docsets,
             left: Vec::with_capacity(100),
@@ -158,7 +158,11 @@ impl<TPostings: Postings> PhraseScorer<TPostings> {
             similarity_weight,
             fieldnorm_reader,
             score_needed,
+        };
+        if scorer.doc() != TERMINATED && !scorer.phrase_match() {
+            scorer.advance();
         }
+        scorer
     }
 
     pub fn phrase_count(&self) -> u32 {
@@ -225,31 +229,21 @@ impl<TPostings: Postings> PhraseScorer<TPostings> {
 }
 
 impl<TPostings: Postings> DocSet for PhraseScorer<TPostings> {
-    fn advance(&mut self) -> bool {
-        while self.intersection_docset.advance() {
-            if self.phrase_match() {
-                return true;
+    fn advance(&mut self) -> DocId {
+        loop {
+            let doc = self.intersection_docset.advance();
+            if doc == TERMINATED || self.phrase_match() {
+                return doc;
             }
         }
-        false
     }
 
-    fn skip_next(&mut self, target: DocId) -> SkipResult {
-        if self.intersection_docset.skip_next(target) == SkipResult::End {
-            return SkipResult::End;
-        }
-        if self.phrase_match() {
-            if self.doc() == target {
-                return SkipResult::Reached;
-            } else {
-                return SkipResult::OverStep;
-            }
-        }
-        if self.advance() {
-            SkipResult::OverStep
-        } else {
-            SkipResult::End
+    fn seek(&mut self, target: DocId) -> DocId {
+        let doc = self.intersection_docset.seek(target);
+        if doc == TERMINATED || self.phrase_match() {
+            return doc;
         }
+        self.advance()
     }
 
     fn doc(&self) -> DocId {

src/query/phrase_query/phrase_weight.rs

@@ -9,8 +9,8 @@ use crate::query::Weight;
 use crate::query::{EmptyScorer, Explanation};
 use crate::schema::IndexRecordOption;
 use crate::schema::Term;
+use crate::Result;
 use crate::{DocId, DocSet};
-use crate::{Result, SkipResult};
 
 pub struct PhraseWeight {
     phrase_terms: Vec<(usize, Term)>,
@@ -99,7 +99,7 @@ impl Weight for PhraseWeight {
             return Err(does_not_match(doc));
         }
         let mut scorer = scorer_opt.unwrap();
-        if scorer.skip_next(doc) != SkipResult::Reached {
+        if scorer.seek(doc) != doc {
             return Err(does_not_match(doc));
         }
         let fieldnorm_reader = self.fieldnorm_reader(reader);
@@ -114,6 +114,7 @@ impl Weight for PhraseWeight {
 #[cfg(test)]
 mod tests {
     use super::super::tests::create_index;
+    use crate::docset::TERMINATED;
     use crate::query::PhraseQuery;
     use crate::{DocSet, Term};
 
@@ -132,12 +133,11 @@ mod tests {
             .phrase_scorer(searcher.segment_reader(0u32), 1.0f32)
             .unwrap()
             .unwrap();
-        assert!(phrase_scorer.advance());
         assert_eq!(phrase_scorer.doc(), 1);
         assert_eq!(phrase_scorer.phrase_count(), 2);
-        assert!(phrase_scorer.advance());
+        assert_eq!(phrase_scorer.advance(), 2);
         assert_eq!(phrase_scorer.doc(), 2);
         assert_eq!(phrase_scorer.phrase_count(), 1);
-        assert!(!phrase_scorer.advance());
+        assert_eq!(phrase_scorer.advance(), TERMINATED);
     }
 }

src/query/query_parser/query_parser.rs

@@ -113,8 +113,9 @@ fn trim_ast(logical_ast: LogicalAST) -> Option<LogicalAST> {
 /// The language covered by the current parser is extremely simple.
 ///
 /// * simple terms: "e.g.: `Barack Obama` are simply tokenized using
-///   tantivy's `StandardTokenizer`, hence becoming `["barack", "obama"]`.
-///   The terms are then searched within the default terms of the query parser.
+///   tantivy's [`SimpleTokenizer`](tantivy::tokenizer::SimpleTokenizer), hence
+///   becoming `["barack", "obama"]`. The terms are then searched within
+///   the default terms of the query parser.
 ///
 ///   e.g. If `body` and `title` are default fields, our example terms are
 ///   `["title:barack", "body:barack", "title:obama", "body:obama"]`.

src/query/range_query.rs

@@ -10,7 +10,7 @@ use crate::schema::Type;
 use crate::schema::{Field, IndexRecordOption, Term};
 use crate::termdict::{TermDictionary, TermStreamer};
 use crate::DocId;
-use crate::{Result, SkipResult};
+use crate::Result;
 use std::collections::Bound;
 use std::ops::Range;
 
@@ -312,7 +312,7 @@ impl Weight for RangeWeight {
 
     fn explain(&self, reader: &SegmentReader, doc: DocId) -> Result<Explanation> {
         let mut scorer = self.scorer(reader, 1.0f32)?;
-        if scorer.skip_next(doc) != SkipResult::Reached {
+        if scorer.seek(doc) != doc {
             return Err(does_not_match(doc));
         }
         Ok(Explanation::new("RangeQuery", 1.0f32))

src/query/reqopt_scorer.rs

@@ -1,9 +1,8 @@
-use crate::docset::{DocSet, SkipResult};
+use crate::docset::DocSet;
 use crate::query::score_combiner::ScoreCombiner;
 use crate::query::Scorer;
 use crate::DocId;
 use crate::Score;
-use std::cmp::Ordering;
 use std::marker::PhantomData;
 
 /// Given a required scorer and an optional scorer
@@ -17,7 +16,6 @@ pub struct RequiredOptionalScorer<TReqScorer, TOptScorer, TScoreCombiner> {
     req_scorer: TReqScorer,
     opt_scorer: TOptScorer,
     score_cache: Option<Score>,
-    opt_finished: bool,
     _phantom: PhantomData<TScoreCombiner>,
 }
 
@@ -29,14 +27,12 @@ where
     /// Creates a new `RequiredOptionalScorer`.
     pub fn new(
         req_scorer: TReqScorer,
-        mut opt_scorer: TOptScorer,
+        opt_scorer: TOptScorer,
     ) -> RequiredOptionalScorer<TReqScorer, TOptScorer, TScoreCombiner> {
-        let opt_finished = !opt_scorer.advance();
         RequiredOptionalScorer {
             req_scorer,
             opt_scorer,
             score_cache: None,
-            opt_finished,
             _phantom: PhantomData,
         }
     }
@@ -48,7 +44,7 @@ where
     TReqScorer: DocSet,
     TOptScorer: DocSet,
 {
-    fn advance(&mut self) -> bool {
+    fn advance(&mut self) -> DocId {
         self.score_cache = None;
         self.req_scorer.advance()
     }
@@ -76,22 +72,8 @@ where
         let doc = self.doc();
         let mut score_combiner = TScoreCombiner::default();
         score_combiner.update(&mut self.req_scorer);
-        if !self.opt_finished {
-            match self.opt_scorer.doc().cmp(&doc) {
-                Ordering::Greater => {}
-                Ordering::Equal => {
-                    score_combiner.update(&mut self.opt_scorer);
-                }
-                Ordering::Less => match self.opt_scorer.skip_next(doc) {
-                    SkipResult::Reached => {
-                        score_combiner.update(&mut self.opt_scorer);
-                    }
-                    SkipResult::End => {
-                        self.opt_finished = true;
-                    }
-                    SkipResult::OverStep => {}
-                },
-            }
+        if self.opt_scorer.seek(doc) == doc {
+            score_combiner.update(&mut self.opt_scorer);
         }
         let score = score_combiner.score();
         self.score_cache = Some(score);
@@ -102,7 +84,7 @@ where
 #[cfg(test)]
 mod tests {
     use super::RequiredOptionalScorer;
-    use crate::docset::DocSet;
+    use crate::docset::{DocSet, TERMINATED};
     use crate::postings::tests::test_skip_against_unoptimized;
     use crate::query::score_combiner::{DoNothingCombiner, SumCombiner};
     use crate::query::ConstScorer;
@@ -119,9 +101,7 @@ mod tests {
                 ConstScorer::from(VecDocSet::from(vec![])),
             );
         let mut docs = vec![];
-        while reqoptscorer.advance() {
-            docs.push(reqoptscorer.doc());
-        }
+        reqoptscorer.for_each(&mut |doc, _| docs.push(doc));
         assert_eq!(docs, req);
     }
 
@@ -133,46 +113,45 @@ mod tests {
                 ConstScorer::new(VecDocSet::from(vec![1, 2, 7, 11, 12, 15]), 1.0f32),
             );
         {
-            assert!(reqoptscorer.advance());
             assert_eq!(reqoptscorer.doc(), 1);
             assert_eq!(reqoptscorer.score(), 2f32);
         }
         {
-            assert!(reqoptscorer.advance());
+            assert_eq!(reqoptscorer.advance(), 3);
             assert_eq!(reqoptscorer.doc(), 3);
             assert_eq!(reqoptscorer.score(), 1f32);
         }
         {
-            assert!(reqoptscorer.advance());
+            assert_eq!(reqoptscorer.advance(), 7);
             assert_eq!(reqoptscorer.doc(), 7);
             assert_eq!(reqoptscorer.score(), 2f32);
         }
         {
-            assert!(reqoptscorer.advance());
+            assert_eq!(reqoptscorer.advance(), 8);
             assert_eq!(reqoptscorer.doc(), 8);
             assert_eq!(reqoptscorer.score(), 1f32);
         }
         {
-            assert!(reqoptscorer.advance());
+            assert_eq!(reqoptscorer.advance(), 9);
             assert_eq!(reqoptscorer.doc(), 9);
             assert_eq!(reqoptscorer.score(), 1f32);
         }
         {
-            assert!(reqoptscorer.advance());
+            assert_eq!(reqoptscorer.advance(), 10);
             assert_eq!(reqoptscorer.doc(), 10);
             assert_eq!(reqoptscorer.score(), 1f32);
         }
         {
-            assert!(reqoptscorer.advance());
+            assert_eq!(reqoptscorer.advance(), 13);
             assert_eq!(reqoptscorer.doc(), 13);
             assert_eq!(reqoptscorer.score(), 1f32);
         }
         {
-            assert!(reqoptscorer.advance());
+            assert_eq!(reqoptscorer.advance(), 15);
             assert_eq!(reqoptscorer.doc(), 15);
             assert_eq!(reqoptscorer.score(), 2f32);
         }
-        assert!(!reqoptscorer.advance());
+        assert_eq!(reqoptscorer.advance(), TERMINATED);
     }
 
     #[test]

src/query/scorer.rs

@@ -1,5 +1,4 @@
-use crate::common::BitSet;
-use crate::docset::{DocSet, SkipResult};
+use crate::docset::{DocSet, TERMINATED};
 use crate::DocId;
 use crate::Score;
 use downcast_rs::impl_downcast;
@@ -17,8 +16,35 @@ pub trait Scorer: downcast_rs::Downcast + DocSet + 'static {
     /// Iterates through all of the document matched by the DocSet
     /// `DocSet` and push the scored documents to the collector.
     fn for_each(&mut self, callback: &mut dyn FnMut(DocId, Score)) {
-        while self.advance() {
-            callback(self.doc(), self.score());
+        let mut doc = self.doc();
+        while doc != TERMINATED {
+            callback(doc, self.score());
+            doc = self.advance();
+        }
+    }
+
+    /// Calls `callback` with all of the `(doc, score)` for which score
+    /// is exceeding a given threshold.
+    ///
+    /// This method is useful for the TopDocs collector.
+    /// For all docsets, the blanket implementation has the benefit
+    /// of prefiltering (doc, score) pairs, avoiding the
+    /// virtual dispatch cost.
+    ///
+    /// More importantly, it makes it possible for scorers to implement
+    /// important optimization (e.g. BlockWAND for union).
+    fn for_each_pruning(
+        &mut self,
+        mut threshold: f32,
+        callback: &mut dyn FnMut(DocId, Score) -> Score,
+    ) {
+        let mut doc = self.doc();
+        while doc != TERMINATED {
+            let score = self.score();
+            if score > threshold {
+                threshold = callback(doc, score);
+            }
+            doc = self.advance();
         }
     }
 }
@@ -61,12 +87,12 @@ impl<TDocSet: DocSet> From<TDocSet> for ConstScorer<TDocSet> {
 }
 
 impl<TDocSet: DocSet> DocSet for ConstScorer<TDocSet> {
-    fn advance(&mut self) -> bool {
+    fn advance(&mut self) -> DocId {
         self.docset.advance()
     }
 
-    fn skip_next(&mut self, target: DocId) -> SkipResult {
-        self.docset.skip_next(target)
+    fn seek(&mut self, target: DocId) -> DocId {
+        self.docset.seek(target)
     }
 
     fn fill_buffer(&mut self, buffer: &mut [DocId]) -> usize {
@@ -80,10 +106,6 @@ impl<TDocSet: DocSet> DocSet for ConstScorer<TDocSet> {
     fn size_hint(&self) -> u32 {
         self.docset.size_hint()
     }
-
-    fn append_to_bitset(&mut self, bitset: &mut BitSet) {
-        self.docset.append_to_bitset(bitset);
-    }
 }
 
 impl<TDocSet: DocSet + 'static> Scorer for ConstScorer<TDocSet> {

src/query/term_query/block_max_term_scorer.rs

@@ -1,98 +0,0 @@
-use crate::docset::{DocSet, SkipResult};
-use crate::query::{Explanation, Scorer};
-use crate::DocId;
-use crate::Score;
-
-use crate::fieldnorm::FieldNormReader;
-use crate::postings::Postings;
-use crate::postings::{BlockMaxPostings, BlockMaxSegmentPostings};
-use crate::query::bm25::BM25Weight;
-use crate::query::BlockMaxScorer;
-
-pub struct BlockMaxTermScorer {
-    postings: BlockMaxSegmentPostings,
-    fieldnorm_reader: FieldNormReader,
-    similarity_weight: BM25Weight,
-}
-
-impl BlockMaxTermScorer {
-    pub fn new(
-        postings: BlockMaxSegmentPostings,
-        fieldnorm_reader: FieldNormReader,
-        similarity_weight: BM25Weight,
-    ) -> Self {
-        Self {
-            postings,
-            fieldnorm_reader,
-            similarity_weight,
-        }
-    }
-}
-
-impl BlockMaxTermScorer {
-    fn _score(&self, fieldnorm_id: u8, term_freq: u32) -> Score {
-        self.similarity_weight.score(fieldnorm_id, term_freq)
-    }
-
-    pub fn term_freq(&self) -> u32 {
-        self.postings.term_freq()
-    }
-
-    pub fn fieldnorm_id(&self) -> u8 {
-        self.fieldnorm_reader.fieldnorm_id(self.doc())
-    }
-
-    pub fn explain(&self) -> Explanation {
-        let fieldnorm_id = self.fieldnorm_id();
-        let term_freq = self.term_freq();
-        self.similarity_weight.explain(fieldnorm_id, term_freq)
-    }
-}
-
-impl DocSet for BlockMaxTermScorer {
-    fn advance(&mut self) -> bool {
-        self.postings.advance()
-    }
-
-    fn skip_next(&mut self, target: DocId) -> SkipResult {
-        self.postings.skip_next(target)
-    }
-
-    fn doc(&self) -> DocId {
-        self.postings.doc()
-    }
-
-    fn size_hint(&self) -> u32 {
-        self.postings.size_hint()
-    }
-}
-
-impl Scorer for BlockMaxTermScorer {
-    fn score(&mut self) -> Score {
-        self._score(
-            self.fieldnorm_reader.fieldnorm_id(self.doc()),
-            self.postings.term_freq(),
-        )
-    }
-}
-
-impl BlockMaxScorer for BlockMaxTermScorer {
-    fn block_max_score(&mut self) -> Score {
-        self._score(
-            self.fieldnorm_reader
-                .fieldnorm_id(self.postings.block_max_doc()),
-            self.postings.term_freq(),
-        )
-    }
-
-    fn block_max_doc(&mut self) -> DocId {
-        self.postings.block_max_doc()
-    }
-
-    fn max_score(&self) -> Score {
-        self._score(
-            self.fieldnorm_reader.fieldnorm_id(self.postings.max_doc()),
-            self.postings.max_term_freq(),
-        )
-    }
-}

src/query/term_query/mod.rs

@@ -1,9 +1,7 @@
-mod block_max_term_scorer;
 mod term_query;
 mod term_scorer;
 mod term_weight;
 
-pub use self::block_max_term_scorer::BlockMaxTermScorer;
 pub use self::term_query::TermQuery;
 pub use self::term_scorer::TermScorer;
 pub use self::term_weight::TermWeight;
@@ -28,10 +26,8 @@ mod tests {
         {
             // writing the segment
             let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
-            {
-                let doc = doc!(text_field => "a");
-                index_writer.add_document(doc);
-            }
+            let doc = doc!(text_field => "a");
+            index_writer.add_document(doc);
             assert!(index_writer.commit().is_ok());
         }
         let searcher = index.reader().unwrap().searcher();
@@ -42,7 +38,6 @@ mod tests {
         let term_weight = term_query.weight(&searcher, true).unwrap();
         let segment_reader = searcher.segment_reader(0);
         let mut term_scorer = term_weight.scorer(segment_reader, 1.0f32).unwrap();
-        assert!(term_scorer.advance());
         assert_eq!(term_scorer.doc(), 0);
         assert_eq!(term_scorer.score(), 0.28768212);
     }

src/query/term_query/term_scorer.rs

@@ -1,4 +1,4 @@
-use crate::docset::{DocSet, SkipResult};
+use crate::docset::DocSet;
 use crate::query::{Explanation, Scorer};
 use crate::DocId;
 use crate::Score;
@@ -45,12 +45,12 @@ impl TermScorer {
 }
 
 impl DocSet for TermScorer {
-    fn advance(&mut self) -> bool {
+    fn advance(&mut self) -> DocId {
         self.postings.advance()
     }
 
-    fn skip_next(&mut self, target: DocId) -> SkipResult {
-        self.postings.skip_next(target)
+    fn seek(&mut self, target: DocId) -> DocId {
+        self.postings.seek(target)
     }
 
     fn doc(&self) -> DocId {

src/query/term_query/term_weight.rs

@@ -8,8 +8,8 @@ use crate::query::Weight;
 use crate::query::{Explanation, Scorer};
 use crate::schema::IndexRecordOption;
 use crate::DocId;
+use crate::Result;
 use crate::Term;
-use crate::{Result, SkipResult};
 
 pub struct TermWeight {
     term: Term,
@@ -25,7 +25,7 @@ impl Weight for TermWeight {
 
     fn explain(&self, reader: &SegmentReader, doc: DocId) -> Result<Explanation> {
         let mut scorer = self.scorer_specialized(reader, 1.0f32)?;
-        if scorer.skip_next(doc) != SkipResult::Reached {
+        if scorer.seek(doc) != doc {
             return Err(does_not_match(doc));
         }
         Ok(scorer.explain())

src/query/union.rs

@@ -1,107 +1,13 @@
 use crate::common::TinySet;
-use crate::docset::{DocSet, SkipResult};
+use crate::docset::{DocSet, TERMINATED};
 use crate::query::score_combiner::{DoNothingCombiner, ScoreCombiner};
-use crate::query::{Scorer, BlockMaxScorer};
+use crate::query::Scorer;
 use crate::DocId;
 use crate::Score;
-use std::cmp::Ordering;
 
 const HORIZON_NUM_TINYBITSETS: usize = 64;
 const HORIZON: u32 = 64u32 * HORIZON_NUM_TINYBITSETS as u32;
 
-
-#[derive(Debug, Copy, Clone, PartialEq, Eq)]
-struct Pivot {
-    position: usize,
-    first_occurrence: usize,
-    doc: DocId,
-}
-
-/// Find the position in the sorted list of posting lists of the **pivot**.
-///
-/// docsets need to be advanced, and are required to be sorted by the doc they point to.
-///
-/// The pivot is then defined as the lowest DocId that has a chance of matching our condition.
-fn find_pivot_position<'a, TScorer>(
-    mut docsets: impl Iterator<Item = &'a TScorer>,
-    lower_bound_score: Score,
-) -> Option<Pivot>
-    where TScorer: BlockMaxScorer
-{
-    let mut position = 0;
-    let mut upper_bound = Score::default();
-    while let Some(docset) = docsets.next() {
-        upper_bound += docset.max_score();
-        if lower_bound_score < upper_bound {
-            let pivot_doc = docset.doc();
-            let first_occurrence = position;
-            while let Some(docset) = docsets.next() {
-                if docset.doc() != pivot_doc {
-                    break;
-                } else {
-                    position += 1;
-                }
-            }
-            return Some(Pivot {
-                position,
-                doc: pivot_doc,
-                first_occurrence,
-            });
-        }
-        position += 1;
-    }
-    None
-}
-
-/// Sifts down the first element of the slice.
-///
-/// `docsets[1..]` are assumed sorted.
-/// This function swaps `docsets[0]` with its right
-/// neighbor successively -bubble sort style- until it reaches the first
-/// position such that `docsets` is sorted.
-fn sift_down<TScorer>(docsets: &mut [TScorer])
-    where
-        TScorer: BlockMaxScorer + Scorer,
-{
-    for idx in 1..docsets.len() {
-        if docsets[idx].doc() >= docsets[idx - 1].doc() {
-            return;
-        }
-        docsets.swap(idx, idx - 1);
-    }
-}
-
-
-/// Given an iterator over all ordered lists up to the pivot (inclusive) and the following list (if
-/// exists), it returns the next document ID that can be possibly relevant, based on the block max
-/// scores.
-fn find_next_relevant_doc<TScorer>(
-    docsets_up_to_pivot: &mut [TScorer],
-    pivot_docset: &mut TScorer,
-    docset_after_pivot: Option<&mut TScorer>,
-) -> DocId
-    where
-        TScorer: BlockMaxScorer + Scorer,
-{
-    let mut next_doc = 1 + docsets_up_to_pivot
-        .iter_mut()
-        .map(|docset| docset.block_max_doc())
-        .chain(std::iter::once(pivot_docset.block_max_doc()))
-        .min()
-        .unwrap();
-    if let Some(docset) = docset_after_pivot {
-        let doc = docset.doc();
-        if doc < next_doc {
-            next_doc = doc;
-        }
-    }
-    if next_doc <= pivot_docset.doc() {
-        pivot_docset.doc() + 1
-    } else {
-        next_doc
-    }
-}
-
 // `drain_filter` is not stable yet.
 // This function is similar except that it does is not unstable, and
 // it does not keep the original vector ordering.
@@ -132,7 +38,6 @@ pub struct Union<TScorer, TScoreCombiner = DoNothingCombiner> {
     score: Score,
 }
 
-
 impl<TScorer, TScoreCombiner> From<Vec<TScorer>> for Union<TScorer, TScoreCombiner>
 where
     TScoreCombiner: ScoreCombiner,
@@ -141,17 +46,9 @@ where
     fn from(docsets: Vec<TScorer>) -> Union<TScorer, TScoreCombiner> {
         let non_empty_docsets: Vec<TScorer> = docsets
             .into_iter()
-            .flat_map(
-                |mut docset| {
-                    if docset.advance() {
-                        Some(docset)
-                    } else {
-                        None
-                    }
-                },
-            )
+            .filter(|docset| docset.doc() != TERMINATED)
             .collect();
-        Union {
+        let mut union = Union {
             docsets: non_empty_docsets,
             bitsets: Box::new([TinySet::empty(); HORIZON_NUM_TINYBITSETS]),
             scores: Box::new([TScoreCombiner::default(); HORIZON as usize]),
@@ -159,7 +56,13 @@ where
             offset: 0,
             doc: 0,
             score: 0f32,
+        };
+        if union.refill() {
+            union.advance();
+        } else {
+            union.doc = TERMINATED;
         }
+        union
     }
 }
 
@@ -180,7 +83,7 @@ fn refill<TScorer: Scorer, TScoreCombiner: ScoreCombiner>(
             let delta = doc - min_doc;
             bitsets[(delta / 64) as usize].insert_mut(delta % 64u32);
             score_combiner[delta as usize].update(scorer);
-            if !scorer.advance() {
+            if scorer.advance() == TERMINATED {
                 // remove the docset, it has been entirely consumed.
                 return true;
             }
@@ -193,6 +96,7 @@ impl<TScorer: Scorer, TScoreCombiner: ScoreCombiner> Union<TScorer, TScoreCombin
         if let Some(min_doc) = self.docsets.iter().map(DocSet::doc).min() {
             self.offset = min_doc;
             self.cursor = 0;
+            self.doc = min_doc;
             refill(
                 &mut self.docsets,
                 &mut *self.bitsets,
@@ -220,76 +124,6 @@ impl<TScorer: Scorer, TScoreCombiner: ScoreCombiner> Union<TScorer, TScoreCombin
         }
         false
     }
-
-
-}
-
-
-impl<TScorer: BlockMaxScorer, TScoreCombiner: ScoreCombiner> Union<TScorer, TScoreCombiner> {
-    fn advance_with_pivot(&mut self, pivot: Pivot, lower_bound_score: Score) -> SkipResult {
-        let block_upper_bound: Score = self.docsets[..=pivot.position]
-            .iter_mut()
-            .map(|docset| docset.block_max_score())
-            .sum();
-        if block_upper_bound > lower_bound_score {
-            if pivot.doc == self.docsets[0].doc() {
-                // Since self.docsets is sorted by their current doc, in this branch, all
-                // docsets in [0..=pivot] are positioned on pivot.doc.
-                //
-                // Lets compute the actual score for this doc.
-                //
-                // NOTE(elshize): One additional check needs to be done to improve performance:
-                // update block-wise bound while accumulating score with the actual score,
-                // and check each time if still above threshold.
-                let mut combiner = TScoreCombiner::default();
-                for idx in (0..=pivot.position).rev() {
-                    combiner.update(&mut self.docsets[idx]);
-                    if !self.docsets[idx].advance() {
-                        self.docsets.swap_remove(idx);
-                    }
-                }
-                self.score = combiner.score();
-                self.doc = pivot.doc;
-                self.docsets.sort_by_key(TScorer::doc);
-                SkipResult::Reached
-            } else {
-                // The substraction does not underflow because otherwise we would go to the other
-                // branch.
-                //
-                // `advanced_idx` is the last idx that is not positionned on the pivot yet.
-                let advanced_idx = pivot.first_occurrence - 1;
-                if !self.docsets[advanced_idx].advance() {
-                    self.docsets.swap_remove(advanced_idx);
-                }
-                if self.docsets.is_empty() {
-                    return SkipResult::End;
-                }
-                sift_down(&mut self.docsets[advanced_idx..]);
-                SkipResult::OverStep
-            }
-        } else {
-            let (up_to_pivot, pivot_and_rest) = self.docsets.split_at_mut(pivot.position as usize);
-            let (pivot, after_pivot) = pivot_and_rest.split_first_mut().unwrap();
-            let next_doc = find_next_relevant_doc(up_to_pivot, pivot, after_pivot.first_mut());
-            // NOTE(elshize): It might be more efficient to advance the list with the higher
-            // max score, but let's advance the first one for now for simplicity.
-            if self.docsets[0].skip_next(next_doc) == SkipResult::End {
-                self.docsets.swap_remove(0);
-            }
-            if self.docsets.is_empty() {
-                return SkipResult::End;
-            }
-            sift_down(&mut self.docsets[..]);
-            SkipResult::OverStep
-        }
-    }
-
-    /// Find the position in the sorted list of posting lists of the **pivot**.
-    fn find_pivot_position(&self, lower_bound_score: Score) -> Option<Pivot> {
-        find_pivot_position(
-            self.docsets.iter().map(|docset| docset),
-            lower_bound_score)
-    }
 }
 
 impl<TScorer, TScoreCombiner> DocSet for Union<TScorer, TScoreCombiner>
@@ -297,30 +131,23 @@ where
     TScorer: Scorer,
     TScoreCombiner: ScoreCombiner,
 {
-    fn advance(&mut self) -> bool {
+    fn advance(&mut self) -> DocId {
         if self.advance_buffered() {
-            return true;
+            return self.doc;
         }
-        if self.refill() {
-            self.advance();
-            true
-        } else {
-            false
+        if !self.refill() {
+            self.doc = TERMINATED;
+            return TERMINATED;
         }
+        if !self.advance_buffered() {
+            return TERMINATED;
+        }
+        self.doc
     }
 
-    fn skip_next(&mut self, target: DocId) -> SkipResult {
-        if !self.advance() {
-            return SkipResult::End;
-        }
-        match self.doc.cmp(&target) {
-            Ordering::Equal => {
-                return SkipResult::Reached;
-            }
-            Ordering::Greater => {
-                return SkipResult::OverStep;
-            }
-            Ordering::Less => {}
+    fn seek(&mut self, target: DocId) -> DocId {
+        if self.doc >= target {
+            return self.doc;
         }
         let gap = target - self.offset;
         if gap < HORIZON {
@@ -338,18 +165,11 @@ where
 
             // Advancing until we reach the end of the bucket
             // or we reach a doc greater or equal to the target.
-            while self.advance() {
-                match self.doc().cmp(&target) {
-                    Ordering::Equal => {
-                        return SkipResult::Reached;
-                    }
-                    Ordering::Greater => {
-                        return SkipResult::OverStep;
-                    }
-                    Ordering::Less => {}
-                }
+            let mut doc = self.doc();
+            while doc < target {
+                doc = self.advance();
             }
-            SkipResult::End
+            doc
         } else {
             // clear the buffered info.
             for obsolete_tinyset in self.bitsets.iter_mut() {
@@ -363,45 +183,30 @@ where
             // advance all docsets to a doc >= to the target.
             #[cfg_attr(feature = "cargo-clippy", allow(clippy::clippy::collapsible_if))]
             unordered_drain_filter(&mut self.docsets, |docset| {
-                if docset.doc() < target {
-                    if docset.skip_next(target) == SkipResult::End {
-                        return true;
-                    }
-                }
-                false
+                docset.seek(target) == TERMINATED
             });
 
             // at this point all of the docsets
             // are positionned on a doc >= to the target.
-            if self.refill() {
-                self.advance();
-                if self.doc() == target {
-                    SkipResult::Reached
-                } else {
-                    debug_assert!(self.doc() > target);
-                    SkipResult::OverStep
-                }
-            } else {
-                SkipResult::End
+            if !self.refill() {
+                self.doc = TERMINATED;
+                return TERMINATED;
             }
+            self.advance()
         }
     }
 
-    // TODO implement `count` efficiently.
-
-    fn doc(&self) -> DocId {
-        self.doc
-    }
-
-    fn size_hint(&self) -> u32 {
-        0u32
-    }
+    // TODO Also implement `count` with deletes efficiently.
 
     fn count_including_deleted(&mut self) -> u32 {
+        if self.doc == TERMINATED {
+            return 0;
+        }
         let mut count = self.bitsets[self.cursor..HORIZON_NUM_TINYBITSETS]
             .iter()
             .map(|bitset| bitset.len())
-            .sum::<u32>();
+            .sum::<u32>()
+            + 1;
         for bitset in self.bitsets.iter_mut() {
             bitset.clear();
         }
@@ -414,6 +219,18 @@ where
         self.cursor = HORIZON_NUM_TINYBITSETS;
         count
     }
+
+    fn doc(&self) -> DocId {
+        self.doc
+    }
+
+    fn size_hint(&self) -> u32 {
+        self.docsets
+            .iter()
+            .map(|docset| docset.size_hint())
+            .max()
+            .unwrap_or(0u32)
+    }
 }
 
 impl<TScorer, TScoreCombiner> Scorer for Union<TScorer, TScoreCombiner>
@@ -431,7 +248,7 @@ mod tests {
 
     use super::Union;
     use super::HORIZON;
-    use crate::docset::{DocSet, SkipResult};
+    use crate::docset::{DocSet, TERMINATED};
     use crate::postings::tests::test_skip_against_unoptimized;
     use crate::query::score_combiner::DoNothingCombiner;
     use crate::query::ConstScorer;
@@ -460,12 +277,12 @@ mod tests {
         };
         let mut union: Union<_, DoNothingCombiner> = make_union();
         let mut count = 0;
-        while union.advance() {
-            assert!(union_expected.advance());
+        while union.doc() != TERMINATED {
             assert_eq!(union_expected.doc(), union.doc());
+            assert_eq!(union_expected.advance(), union.advance());
             count += 1;
         }
-        assert!(!union_expected.advance());
+        assert_eq!(union_expected.advance(), TERMINATED);
         assert_eq!(count, make_union().count_including_deleted());
     }
 
@@ -493,9 +310,7 @@ mod tests {
     fn test_aux_union_skip(docs_list: &[Vec<DocId>], skip_targets: Vec<DocId>) {
         let mut btree_set = BTreeSet::new();
         for docs in docs_list {
-            for &doc in docs.iter() {
-                btree_set.insert(doc);
-            }
+            btree_set.extend(docs.iter().cloned());
         }
         let docset_factory = || {
             let res: Box<dyn DocSet> = Box::new(Union::<_, DoNothingCombiner>::from(
@@ -510,10 +325,10 @@ mod tests {
         };
         let mut docset = docset_factory();
         for el in btree_set {
-            assert!(docset.advance());
             assert_eq!(el, docset.doc());
+            docset.advance();
         }
-        assert!(!docset.advance());
+        assert_eq!(docset.doc(), TERMINATED);
         test_skip_against_unoptimized(docset_factory, skip_targets);
     }
 
@@ -536,10 +351,10 @@ mod tests {
             ConstScorer::from(VecDocSet::from(vec![0u32, 5u32])),
             ConstScorer::from(VecDocSet::from(vec![1u32, 4u32])),
         ]);
-        assert!(docset.advance());
         assert_eq!(docset.doc(), 0u32);
-        assert_eq!(docset.skip_next(0u32), SkipResult::OverStep);
-        assert_eq!(docset.doc(), 1u32)
+        assert_eq!(docset.seek(0u32), 0u32);
+        assert_eq!(docset.seek(0u32), 0u32);
+        assert_eq!(docset.doc(), 0u32)
     }
 
     #[test]

src/query/vec_docset.rs

@@ -1,9 +1,8 @@
 #![allow(dead_code)]
 
 use crate::common::HasLen;
-use crate::docset::DocSet;
+use crate::docset::{DocSet, TERMINATED};
 use crate::DocId;
-use std::num::Wrapping;
 
 /// Simulate a `Postings` objects from a `VecPostings`.
 /// `VecPostings` only exist for testing purposes.
@@ -12,26 +11,30 @@ use std::num::Wrapping;
 /// No positions are returned.
 pub struct VecDocSet {
     doc_ids: Vec<DocId>,
-    cursor: Wrapping<usize>,
+    cursor: usize,
 }
 
 impl From<Vec<DocId>> for VecDocSet {
     fn from(doc_ids: Vec<DocId>) -> VecDocSet {
-        VecDocSet {
-            doc_ids,
-            cursor: Wrapping(usize::max_value()),
-        }
+        VecDocSet { doc_ids, cursor: 0 }
     }
 }
 
 impl DocSet for VecDocSet {
-    fn advance(&mut self) -> bool {
-        self.cursor += Wrapping(1);
-        self.doc_ids.len() > self.cursor.0
+    fn advance(&mut self) -> DocId {
+        self.cursor += 1;
+        if self.cursor >= self.doc_ids.len() {
+            self.cursor = self.doc_ids.len();
+            return TERMINATED;
+        }
+        self.doc()
     }
 
     fn doc(&self) -> DocId {
-        self.doc_ids[self.cursor.0]
+        if self.cursor == self.doc_ids.len() {
+            return TERMINATED;
+        }
+        self.doc_ids[self.cursor]
     }
 
     fn size_hint(&self) -> u32 {
@@ -49,22 +52,21 @@ impl HasLen for VecDocSet {
 pub mod tests {
 
     use super::*;
-    use crate::docset::{DocSet, SkipResult};
+    use crate::docset::DocSet;
     use crate::DocId;
 
     #[test]
     pub fn test_vec_postings() {
         let doc_ids: Vec<DocId> = (0u32..1024u32).map(|e| e * 3).collect();
         let mut postings = VecDocSet::from(doc_ids);
-        assert!(postings.advance());
         assert_eq!(postings.doc(), 0u32);
-        assert!(postings.advance());
+        assert_eq!(postings.advance(), 3u32);
         assert_eq!(postings.doc(), 3u32);
-        assert_eq!(postings.skip_next(14u32), SkipResult::OverStep);
+        assert_eq!(postings.seek(14u32), 15u32);
         assert_eq!(postings.doc(), 15u32);
-        assert_eq!(postings.skip_next(300u32), SkipResult::Reached);
+        assert_eq!(postings.seek(300u32), 300u32);
         assert_eq!(postings.doc(), 300u32);
-        assert_eq!(postings.skip_next(6000u32), SkipResult::End);
+        assert_eq!(postings.seek(6000u32), TERMINATED);
     }
 
     #[test]

src/query/weight.rs

@@ -1,26 +1,7 @@
 use super::Scorer;
 use crate::core::SegmentReader;
 use crate::query::Explanation;
-use crate::{DocId, Score};
-
-pub trait PruningScorer {
-
-    fn doc(&self) -> DocId;
-
-    fn score(&self) -> Score;
-
-    /// Advance to the next document that has a score strictly greater than
-    /// `lower_bound_score`.
-    fn advance_with_pruning(&mut self, score_lower_bound: f32) -> bool;
-    fn advance(&mut self) -> bool {
-        self.advance_with_pruning(std::f32::NEG_INFINITY)
-    }
-}
-
-pub enum PruningScorerIfPossible {
-    Pruning(Box<dyn PruningScorer>),
-    NonPruning(Box<dyn Scorer>)
-}
+use crate::DocId;
 
 /// A Weight is the specialization of a Query
 /// for a given set of segments.
@@ -34,11 +15,6 @@ pub trait Weight: Send + Sync + 'static {
     /// See [`Query`](./trait.Query.html).
     fn scorer(&self, reader: &SegmentReader, boost: f32) -> crate::Result<Box<dyn Scorer>>;
 
-    fn pruning_scorer(&self, reader: &SegmentReader, boost: f32) -> crate::Result<PruningScorerIfPossible> {
-        let scorer = self.scorer(reader, boost)?;
-        Ok(PruningScorerIfPossible::NonPruning(Box::new(scorer)))
-    }
-
     /// Returns an `Explanation` for the given document.
     fn explain(&self, reader: &SegmentReader, doc: DocId) -> crate::Result<Explanation>;
 

src/schema/field.rs

@@ -12,13 +12,13 @@ pub struct Field(u32);
 
 impl Field {
     /// Create a new field object for the given FieldId.
-    pub fn from_field_id(field_id: u32) -> Field {
+    pub const fn from_field_id(field_id: u32) -> Field {
         Field(field_id)
     }
 
     /// Returns a u32 identifying uniquely a field within a schema.
     #[allow(clippy::trivially_copy_pass_by_ref)]
-    pub fn field_id(&self) -> u32 {
+    pub const fn field_id(&self) -> u32 {
         self.0
     }
 }