Merge branch 'query-boost' of https://github.com/audunhalland/tantivy into audunhalland-query-boost

* Make Executor public so Searcher::search_in_executor method now can be used

* Fixed cargo fmt

CHANGELOG.md

@@ -1,8 +1,15 @@
 Tantivy 0.12.0
-==========================
+======================
-- By default IndexReader are in `Manual` mode.
+- Removing static dispatch in tokenizers for simplicity. (#762)
+- Added backward iteration for `TermDictionary` stream. (@halvorboe)
+- Fixed a performance issue when searching for the posting lists of a missing term (@audunhalland)
+- Added a configurable maximum number of docs (10M by default) for a segment to be considered for merge (@hntd187, landed by @halvorboe #713) 
 
+## How to update?
 
+Crates relying on custom tokenizer, or registering tokenizer in the manager will require some 
+minor changes. Check https://github.com/tantivy-search/tantivy/blob/master/examples/custom_tokenizer.rs
+to check for some code sample.
 
 Tantivy 0.11.3
 =======================

Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "tantivy"
-version = "0.11.3"
+version = "0.12.0"
 authors = ["Paul Masurel <paul.masurel@gmail.com>"]
 license = "MIT"
 categories = ["database-implementations", "data-structures"]
@@ -18,7 +18,7 @@ byteorder = "1.0"
 crc32fast = "1.2.0"
 once_cell = "1.0"
 regex ={version = "1.3.0", default-features = false, features = ["std"]}
-tantivy-fst = "0.1"
+tantivy-fst = "0.2.1"
 memmap = {version = "0.7", optional=true}
 lz4 = {version="1.20", optional=true}
 snap = {version="0.2"}
@@ -60,7 +60,6 @@ winapi = "0.3"
 rand = "0.7"
 maplit = "1"
 matches = "0.1.8"
-time = "0.1.42"
 
 [dev-dependencies.fail]
 version = "0.3"

examples/pre_tokenized_text.rs

@@ -9,7 +9,7 @@
 // - import tokenized text straight from json,
 // - perform a search on documents with pre-tokenized text
 
-use tantivy::tokenizer::{PreTokenizedString, SimpleTokenizer, Token, TokenStream, Tokenizer};
+use tantivy::tokenizer::{PreTokenizedString, SimpleTokenizer, Token, Tokenizer};
 
 use tantivy::collector::{Count, TopDocs};
 use tantivy::query::TermQuery;

examples/stop_words.rs

@@ -50,7 +50,7 @@ fn main() -> tantivy::Result<()> {
 
     // This tokenizer lowers all of the text (to help with stop word matching)
     // then removes all instances of `the` and `and` from the corpus
-    let tokenizer = SimpleTokenizer
+    let tokenizer = TextAnalyzer::from(SimpleTokenizer)
         .filter(LowerCaser)
         .filter(StopWordFilter::remove(vec![
             "the".to_string(),

src/collector/top_score_collector.rs

@@ -6,6 +6,7 @@ use crate::collector::tweak_score_top_collector::TweakedScoreTopCollector;
 use crate::collector::{
     CustomScorer, CustomSegmentScorer, ScoreSegmentTweaker, ScoreTweaker, SegmentCollector,
 };
+use crate::fastfield::FastFieldReader;
 use crate::schema::Field;
 use crate::DocAddress;
 use crate::DocId;
@@ -61,6 +62,34 @@ impl fmt::Debug for TopDocs {
     }
 }
 
+struct ScorerByFastFieldReader {
+    ff_reader: FastFieldReader<u64>,
+}
+
+impl CustomSegmentScorer<u64> for ScorerByFastFieldReader {
+    fn score(&self, doc: DocId) -> u64 {
+        self.ff_reader.get_u64(u64::from(doc))
+    }
+}
+
+struct ScorerByField {
+    field: Field,
+}
+
+impl CustomScorer<u64> for ScorerByField {
+    type Child = ScorerByFastFieldReader;
+
+    fn segment_scorer(&self, segment_reader: &SegmentReader) -> crate::Result<Self::Child> {
+        let ff_reader = segment_reader
+            .fast_fields()
+            .u64(self.field)
+            .ok_or_else(|| {
+                crate::Error::SchemaError(format!("Field requested is not a i64/u64 fast field."))
+            })?;
+        Ok(ScorerByFastFieldReader { ff_reader })
+    }
+}
+
 impl TopDocs {
     /// Creates a top score collector, with a number of documents equal to "limit".
     ///
@@ -143,14 +172,7 @@ impl TopDocs {
         self,
         field: Field,
     ) -> impl Collector<Fruit = Vec<(u64, DocAddress)>> {
-        self.custom_score(move |segment_reader: &SegmentReader| {
+        self.custom_score(ScorerByField { field })
-            let ff_reader = segment_reader
-                .fast_fields()
-                .u64(field)
-                .expect("Field requested is not a i64/u64 fast field.");
-            //TODO error message missmatch actual behavior for i64
-            move |doc: DocId| ff_reader.get(doc)
-        })
     }
 
     /// Ranks the documents using a custom score.
@@ -572,7 +594,6 @@ mod tests {
     }
 
     #[test]
-    #[should_panic(expected = "Field requested is not a i64/u64 fast field")]
     fn test_field_not_fast_field() {
         let mut schema_builder = Schema::builder();
         let title = schema_builder.add_text_field(TITLE, TEXT);
@@ -587,7 +608,12 @@ mod tests {
         let searcher = index.reader().unwrap().searcher();
         let segment = searcher.segment_reader(0);
         let top_collector = TopDocs::with_limit(4).order_by_u64_field(size);
-        assert!(top_collector.for_segment(0, segment).is_ok());
+        let err = top_collector.for_segment(0, segment);
+        if let Err(crate::Error::SchemaError(msg)) = err {
+            assert_eq!(msg, "Field requested is not a i64/u64 fast field.");
+        } else {
+            assert!(false);
+        }
     }
 
     fn index(

src/core/executor.rs

@@ -10,7 +10,9 @@ use rayon::{ThreadPool, ThreadPoolBuilder};
 /// API of a dependency, knowing it might conflict with a different version
 /// used by the client. Second, we may stop using rayon in the future.
 pub enum Executor {
+    /// Single thread variant of an Executor
     SingleThread,
+    /// Thread pool variant of an Executor
     ThreadPool(ThreadPool),
 }
 
@@ -20,7 +22,7 @@ impl Executor {
         Executor::SingleThread
     }
 
-    // Creates an Executor that dispatches the tasks in a thread pool.
+    /// Creates an Executor that dispatches the tasks in a thread pool.
     pub fn multi_thread(num_threads: usize, prefix: &'static str) -> Result<Executor> {
         let pool = ThreadPoolBuilder::new()
             .num_threads(num_threads)
@@ -29,10 +31,10 @@ impl Executor {
         Ok(Executor::ThreadPool(pool))
     }
 
-    // Perform a map in the thread pool.
+    /// Perform a map in the thread pool.
-    //
+    ///
-    // Regardless of the executor (`SingleThread` or `ThreadPool`), panics in the task
+    /// Regardless of the executor (`SingleThread` or `ThreadPool`), panics in the task
-    // will propagate to the caller.
+    /// will propagate to the caller.
     pub fn map<
         A: Send,
         R: Send,

src/core/index.rs

@@ -1,3 +1,4 @@
+use super::segment::create_segment;
 use super::segment::Segment;
 use crate::core::Executor;
 use crate::core::IndexMeta;
@@ -19,8 +20,7 @@ use crate::reader::IndexReaderBuilder;
 use crate::schema::Field;
 use crate::schema::FieldType;
 use crate::schema::Schema;
-use crate::tokenizer::BoxedTokenizer;
+use crate::tokenizer::{TextAnalyzer, TokenizerManager};
-use crate::tokenizer::TokenizerManager;
 use crate::IndexWriter;
 use crate::Result;
 use num_cpus;
@@ -172,11 +172,11 @@ impl Index {
     }
 
     /// Helper to access the tokenizer associated to a specific field.
-    pub fn tokenizer_for_field(&self, field: Field) -> Result<BoxedTokenizer> {
+    pub fn tokenizer_for_field(&self, field: Field) -> Result<TextAnalyzer> {
         let field_entry = self.schema.get_field_entry(field);
         let field_type = field_entry.field_type();
         let tokenizer_manager: &TokenizerManager = self.tokenizers();
-        let tokenizer_name_opt: Option<BoxedTokenizer> = match field_type {
+        let tokenizer_name_opt: Option<TextAnalyzer> = match field_type {
             FieldType::Str(text_options) => text_options
                 .get_indexing_options()
                 .map(|text_indexing_options| text_indexing_options.tokenizer().to_string())
@@ -330,8 +330,9 @@ impl Index {
             .collect())
     }
 
-    pub(crate) fn segment(&self, segment_meta: SegmentMeta) -> Segment {
+    #[doc(hidden)]
-        Segment::for_index(self.clone(), segment_meta)
+    pub fn segment(&self, segment_meta: SegmentMeta) -> Segment {
+        create_segment(self.clone(), segment_meta)
     }
 
     /// Creates a new segment.
@@ -342,13 +343,6 @@ impl Index {
         self.segment(segment_meta)
     }
 
-    /// Creates a new segment.
-    pub(crate) fn new_segment_unpersisted(&self) -> Segment {
-        let meta = self
-            .inventory
-            .new_segment_meta(SegmentId::generate_random(), 0);
-        Segment::new_volatile(meta, self.schema())
-    }
     /// Return a reference to the index directory.
     pub fn directory(&self) -> &ManagedDirectory {
         &self.directory
@@ -471,7 +465,7 @@ mod tests {
             .try_into()
             .unwrap();
         assert_eq!(reader.searcher().num_docs(), 0);
-        test_index_on_commit_reload_policy_aux(field, index.clone(), &index, &reader);
+        test_index_on_commit_reload_policy_aux(field, &index, &reader);
     }
 
     #[cfg(feature = "mmap")]
@@ -495,7 +489,7 @@ mod tests {
                 .try_into()
                 .unwrap();
             assert_eq!(reader.searcher().num_docs(), 0);
-            test_index_on_commit_reload_policy_aux(field, index.clone(), &index, &reader);
+            test_index_on_commit_reload_policy_aux(field, &index, &reader);
         }
 
         #[test]
@@ -537,11 +531,12 @@ mod tests {
                 .try_into()
                 .unwrap();
             assert_eq!(reader.searcher().num_docs(), 0);
-            test_index_on_commit_reload_policy_aux(field, read_index, &write_index, &reader);
+            test_index_on_commit_reload_policy_aux(field, &write_index, &reader);
         }
     }
 
-    fn test_index_on_commit_reload_policy_aux(field: Field, mut reader_index: Index, index: &Index, reader: &IndexReader) {
+    fn test_index_on_commit_reload_policy_aux(field: Field, index: &Index, reader: &IndexReader) {
+        let mut reader_index = reader.index();
         let (sender, receiver) = crossbeam::channel::unbounded();
         let _watch_handle = reader_index.directory_mut().watch(Box::new(move || {
             let _ = sender.send(());

src/core/inverted_index_reader.rs

@@ -60,7 +60,7 @@ impl InvertedIndexReader {
             .get_index_record_option()
             .unwrap_or(IndexRecordOption::Basic);
         InvertedIndexReader {
-            termdict: TermDictionary::empty(&field_type),
+            termdict: TermDictionary::empty(),
             postings_source: ReadOnlySource::empty(),
             positions_source: ReadOnlySource::empty(),
             positions_idx_source: ReadOnlySource::empty(),

src/core/segment.rs

@@ -3,62 +3,21 @@ use crate::core::Index;
 use crate::core::SegmentId;
 use crate::core::SegmentMeta;
 use crate::directory::error::{OpenReadError, OpenWriteError};
-use crate::directory::{Directory, ManagedDirectory, RAMDirectory};
+use crate::directory::Directory;
 use crate::directory::{ReadOnlySource, WritePtr};
 use crate::indexer::segment_serializer::SegmentSerializer;
 use crate::schema::Schema;
 use crate::Opstamp;
 use crate::Result;
-use failure::_core::ops::DerefMut;
 use std::fmt;
-use std::ops::Deref;
 use std::path::PathBuf;
 use std::result;
 
-#[derive(Clone)]
-pub(crate) enum SegmentDirectory {
-    Persisted(ManagedDirectory),
-    Volatile(RAMDirectory),
-}
-
-impl SegmentDirectory {
-    pub fn new_volatile() -> SegmentDirectory {
-        SegmentDirectory::Volatile(RAMDirectory::default())
-    }
-}
-
-impl From<ManagedDirectory> for SegmentDirectory {
-    fn from(directory: ManagedDirectory) -> Self {
-        SegmentDirectory::Persisted(directory)
-    }
-}
-
-impl Deref for SegmentDirectory {
-    type Target = dyn Directory;
-
-    fn deref(&self) -> &Self::Target {
-        match self {
-            SegmentDirectory::Volatile(dir) => dir,
-            SegmentDirectory::Persisted(dir) => dir,
-        }
-    }
-}
-
-impl DerefMut for SegmentDirectory {
-    fn deref_mut(&mut self) -> &mut Self::Target {
-        match self {
-            SegmentDirectory::Volatile(dir) => dir,
-            SegmentDirectory::Persisted(dir) => dir,
-        }
-    }
-}
-
 /// A segment is a piece of the index.
 #[derive(Clone)]
 pub struct Segment {
-    schema: Schema,
+    index: Index,
     meta: SegmentMeta,
-    directory: SegmentDirectory,
 }
 
 impl fmt::Debug for Segment {
@@ -67,56 +26,23 @@ impl fmt::Debug for Segment {
     }
 }
 
+/// Creates a new segment given an `Index` and a `SegmentId`
+///
+/// The function is here to make it private outside `tantivy`.
+/// #[doc(hidden)]
+pub fn create_segment(index: Index, meta: SegmentMeta) -> Segment {
+    Segment { index, meta }
+}
+
 impl Segment {
+    /// Returns the index the segment belongs to.
+    pub fn index(&self) -> &Index {
+        &self.index
+    }
+
     /// Returns our index's schema.
-    // TODO return a ref.
     pub fn schema(&self) -> Schema {
-        self.schema.clone()
+        self.index.schema()
-    }
-
-    pub(crate) fn new_persisted(
-        meta: SegmentMeta,
-        directory: ManagedDirectory,
-        schema: Schema,
-    ) -> Segment {
-        Segment {
-            meta,
-            schema,
-            directory: SegmentDirectory::from(directory),
-        }
-    }
-
-    /// Creates a new segment that embeds its own `RAMDirectory`.
-    ///
-    /// That segment is entirely dissociated from the index directory.
-    /// It will be persisted by a background thread in charge of IO.
-    pub fn new_volatile(meta: SegmentMeta, schema: Schema) -> Segment {
-        Segment {
-            schema,
-            meta,
-            directory: SegmentDirectory::new_volatile(),
-        }
-    }
-
-    /// Creates a new segment given an `Index` and a `SegmentId`
-    pub(crate) fn for_index(index: Index, meta: SegmentMeta) -> Segment {
-        Segment {
-            directory: SegmentDirectory::Persisted(index.directory().clone()),
-            schema: index.schema(),
-            meta,
-        }
-    }
-
-    pub fn persist(&mut self, mut dest_directory: ManagedDirectory) -> crate::Result<()> {
-        if let SegmentDirectory::Persisted(_) = self.directory {
-            // this segment is already persisted.
-            return Ok(());
-        }
-        if let SegmentDirectory::Volatile(ram_directory) = &self.directory {
-            ram_directory.persist(&mut dest_directory)?;
-        }
-        self.directory = SegmentDirectory::Persisted(dest_directory);
-        Ok(())
     }
 
     /// Returns the segment meta-information
@@ -130,8 +56,7 @@ impl Segment {
     /// as we finalize a fresh new segment.
     pub(crate) fn with_max_doc(self, max_doc: u32) -> Segment {
         Segment {
-            directory: self.directory,
+            index: self.index,
-            schema: self.schema,
             meta: self.meta.with_max_doc(max_doc),
         }
     }
@@ -139,8 +64,7 @@ impl Segment {
     #[doc(hidden)]
     pub fn with_delete_meta(self, num_deleted_docs: u32, opstamp: Opstamp) -> Segment {
         Segment {
-            directory: self.directory,
+            index: self.index,
-            schema: self.schema,
             meta: self.meta.with_delete_meta(num_deleted_docs, opstamp),
         }
     }
@@ -164,7 +88,7 @@ impl Segment {
         component: SegmentComponent,
     ) -> result::Result<ReadOnlySource, OpenReadError> {
         let path = self.relative_path(component);
-        let source = self.directory.open_read(&path)?;
+        let source = self.index.directory().open_read(&path)?;
         Ok(source)
     }
 
@@ -174,7 +98,7 @@ impl Segment {
         component: SegmentComponent,
     ) -> result::Result<WritePtr, OpenWriteError> {
         let path = self.relative_path(component);
-        let write = self.directory.open_write(&path)?;
+        let write = self.index.directory_mut().open_write(&path)?;
         Ok(write)
     }
 }

src/core/segment_reader.rs

@@ -57,68 +57,6 @@ pub struct SegmentReader {
 }
 
 impl SegmentReader {
-    /// Open a new segment for reading.
-    pub fn open(segment: &Segment) -> Result<SegmentReader> {
-        let termdict_source = segment.open_read(SegmentComponent::TERMS)?;
-        let termdict_composite = CompositeFile::open(&termdict_source)?;
-
-        let store_source = segment.open_read(SegmentComponent::STORE)?;
-
-        fail_point!("SegmentReader::open#middle");
-
-        let postings_source = segment.open_read(SegmentComponent::POSTINGS)?;
-        let postings_composite = CompositeFile::open(&postings_source)?;
-
-        let positions_composite = {
-            if let Ok(source) = segment.open_read(SegmentComponent::POSITIONS) {
-                CompositeFile::open(&source)?
-            } else {
-                CompositeFile::empty()
-            }
-        };
-
-        let positions_idx_composite = {
-            if let Ok(source) = segment.open_read(SegmentComponent::POSITIONSSKIP) {
-                CompositeFile::open(&source)?
-            } else {
-                CompositeFile::empty()
-            }
-        };
-
-        let schema = segment.schema();
-
-        let fast_fields_data = segment.open_read(SegmentComponent::FASTFIELDS)?;
-        let fast_fields_composite = CompositeFile::open(&fast_fields_data)?;
-        let fast_field_readers =
-            Arc::new(FastFieldReaders::load_all(&schema, &fast_fields_composite)?);
-
-        let fieldnorms_data = segment.open_read(SegmentComponent::FIELDNORMS)?;
-        let fieldnorms_composite = CompositeFile::open(&fieldnorms_data)?;
-
-        let delete_bitset_opt = if segment.meta().has_deletes() {
-            let delete_data = segment.open_read(SegmentComponent::DELETE)?;
-            Some(DeleteBitSet::open(delete_data))
-        } else {
-            None
-        };
-
-        Ok(SegmentReader {
-            inv_idx_reader_cache: Arc::new(RwLock::new(HashMap::new())),
-            max_doc: segment.meta().max_doc(),
-            num_docs: segment.meta().num_docs(),
-            termdict_composite,
-            postings_composite,
-            fast_fields_readers: fast_field_readers,
-            fieldnorms_composite,
-            segment_id: segment.id(),
-            store_source,
-            delete_bitset_opt,
-            positions_composite,
-            positions_idx_composite,
-            schema,
-        })
-    }
-
     /// Returns the highest document id ever attributed in
     /// this segment + 1.
     /// Today, `tantivy` does not handle deletes, so it happens
@@ -206,6 +144,68 @@ impl SegmentReader {
         StoreReader::from_source(self.store_source.clone())
     }
 
+    /// Open a new segment for reading.
+    pub fn open(segment: &Segment) -> Result<SegmentReader> {
+        let termdict_source = segment.open_read(SegmentComponent::TERMS)?;
+        let termdict_composite = CompositeFile::open(&termdict_source)?;
+
+        let store_source = segment.open_read(SegmentComponent::STORE)?;
+
+        fail_point!("SegmentReader::open#middle");
+
+        let postings_source = segment.open_read(SegmentComponent::POSTINGS)?;
+        let postings_composite = CompositeFile::open(&postings_source)?;
+
+        let positions_composite = {
+            if let Ok(source) = segment.open_read(SegmentComponent::POSITIONS) {
+                CompositeFile::open(&source)?
+            } else {
+                CompositeFile::empty()
+            }
+        };
+
+        let positions_idx_composite = {
+            if let Ok(source) = segment.open_read(SegmentComponent::POSITIONSSKIP) {
+                CompositeFile::open(&source)?
+            } else {
+                CompositeFile::empty()
+            }
+        };
+
+        let schema = segment.schema();
+
+        let fast_fields_data = segment.open_read(SegmentComponent::FASTFIELDS)?;
+        let fast_fields_composite = CompositeFile::open(&fast_fields_data)?;
+        let fast_field_readers =
+            Arc::new(FastFieldReaders::load_all(&schema, &fast_fields_composite)?);
+
+        let fieldnorms_data = segment.open_read(SegmentComponent::FIELDNORMS)?;
+        let fieldnorms_composite = CompositeFile::open(&fieldnorms_data)?;
+
+        let delete_bitset_opt = if segment.meta().has_deletes() {
+            let delete_data = segment.open_read(SegmentComponent::DELETE)?;
+            Some(DeleteBitSet::open(delete_data))
+        } else {
+            None
+        };
+
+        Ok(SegmentReader {
+            inv_idx_reader_cache: Arc::new(RwLock::new(HashMap::new())),
+            max_doc: segment.meta().max_doc(),
+            num_docs: segment.meta().num_docs(),
+            termdict_composite,
+            postings_composite,
+            fast_fields_readers: fast_field_readers,
+            fieldnorms_composite,
+            segment_id: segment.id(),
+            store_source,
+            delete_bitset_opt,
+            positions_composite,
+            positions_idx_composite,
+            schema,
+        })
+    }
+
     /// Returns a field reader associated to the field given in argument.
     /// If the field was not present in the index during indexing time,
     /// the InvertedIndexReader is empty.

src/directory/ram_directory.rs

@@ -144,16 +144,6 @@ impl RAMDirectory {
     pub fn total_mem_usage(&self) -> usize {
         self.fs.read().unwrap().total_mem_usage()
     }
-
-    pub fn persist(&self, dest: &mut dyn Directory) -> crate::Result<()> {
-        let wlock = self.fs.write().unwrap();
-        for (path, source) in wlock.fs.iter() {
-            let mut dest_wrt = dest.open_write(path)?;
-            dest_wrt.write_all(source.as_slice())?;
-            dest_wrt.terminate()?;
-        }
-        Ok(())
-    }
 }
 
 impl Directory for RAMDirectory {

src/fastfield/multivalued/mod.rs

@@ -7,9 +7,6 @@ pub use self::writer::MultiValueIntFastFieldWriter;
 #[cfg(test)]
 mod tests {
 
-    use time;
-
-    use self::time::Duration;
     use crate::collector::TopDocs;
     use crate::query::QueryParser;
     use crate::schema::Cardinality;
@@ -17,6 +14,7 @@ mod tests {
     use crate::schema::IntOptions;
     use crate::schema::Schema;
     use crate::Index;
+    use chrono::Duration;
 
     #[test]
     fn test_multivalued_u64() {

src/indexer/index_writer.rs

@@ -8,33 +8,30 @@ use crate::core::SegmentComponent;
 use crate::core::SegmentId;
 use crate::core::SegmentMeta;
 use crate::core::SegmentReader;
+use crate::directory::TerminatingWrite;
 use crate::directory::{DirectoryLock, GarbageCollectionResult};
-use crate::directory::{TerminatingWrite, WatchCallbackList};
 use crate::docset::DocSet;
 use crate::error::TantivyError;
 use crate::fastfield::write_delete_bitset;
 use crate::indexer::delete_queue::{DeleteCursor, DeleteQueue};
 use crate::indexer::doc_opstamp_mapping::DocToOpstampMapping;
 use crate::indexer::operation::DeleteOperation;
-use crate::indexer::segment_manager::SegmentRegisters;
-use crate::indexer::segment_register::SegmentRegister;
 use crate::indexer::stamper::Stamper;
 use crate::indexer::MergePolicy;
 use crate::indexer::SegmentEntry;
 use crate::indexer::SegmentWriter;
-use crate::reader::NRTReader;
 use crate::schema::Document;
 use crate::schema::IndexRecordOption;
 use crate::schema::Term;
-use crate::tokenizer::TokenizerManager;
+use crate::Opstamp;
-use crate::{IndexReader, Opstamp};
 use crossbeam::channel;
 use futures::executor::block_on;
 use futures::future::Future;
-use smallvec::{smallvec, SmallVec};
+use smallvec::smallvec;
+use smallvec::SmallVec;
 use std::mem;
 use std::ops::Range;
-use std::sync::{Arc, RwLock};
+use std::sync::Arc;
 use std::thread;
 use std::thread::JoinHandle;
 
@@ -71,8 +68,6 @@ pub struct IndexWriter {
     // lifetime of the lock with that of the IndexWriter.
     _directory_lock: Option<DirectoryLock>,
 
-    segment_registers: Arc<RwLock<SegmentRegisters>>,
-
     index: Index,
 
     heap_size_in_bytes_per_thread: usize,
@@ -92,8 +87,6 @@ pub struct IndexWriter {
 
     stamper: Stamper,
     committed_opstamp: Opstamp,
-
-    on_commit: WatchCallbackList,
 }
 
 fn compute_deleted_bitset(
@@ -140,6 +133,7 @@ fn compute_deleted_bitset(
 /// For instance, there was no delete operation between the state of the `segment_entry` and
 /// the `target_opstamp`, `segment_entry` is not updated.
 pub(crate) fn advance_deletes(
+    mut segment: Segment,
     segment_entry: &mut SegmentEntry,
     target_opstamp: Opstamp,
 ) -> crate::Result<()> {
@@ -148,38 +142,28 @@ pub(crate) fn advance_deletes(
         return Ok(());
     }
 
-    let delete_bitset_opt = segment_entry.take_delete_bitset();
+    if segment_entry.delete_bitset().is_none() && segment_entry.delete_cursor().get().is_none() {
-
-    // We avoid directly advancing the `SegmentEntry` delete cursor, because
-    // we do not want to end up in an invalid state if the delete bitset
-    // serialization fails.
-    let mut delete_cursor = segment_entry.delete_cursor();
-
-    if delete_bitset_opt.is_none() && delete_cursor.get().is_none() {
         // There has been no `DeleteOperation` between the segment status and `target_opstamp`.
         return Ok(());
     }
 
-    // We open our current serialized segment to compute the new deleted bitset.
-    let segment = segment_entry.segment().clone();
     let segment_reader = SegmentReader::open(&segment)?;
 
     let max_doc = segment_reader.max_doc();
-    let mut delete_bitset: BitSet =
+    let mut delete_bitset: BitSet = match segment_entry.delete_bitset() {
-        delete_bitset_opt.unwrap_or_else(|| BitSet::with_max_value(max_doc));
+        Some(previous_delete_bitset) => (*previous_delete_bitset).clone(),
-
+        None => BitSet::with_max_value(max_doc),
-    let num_deleted_docs_before = segment.meta().num_deleted_docs();
+    };
 
     compute_deleted_bitset(
         &mut delete_bitset,
         &segment_reader,
-        &mut delete_cursor,
+        segment_entry.delete_cursor(),
         &DocToOpstampMapping::None,
         target_opstamp,
     )?;
 
-    // TODO optimize... We are simply manipulating bitsets here.
+    // TODO optimize
-    // We should be able to compute the union much faster.
     if let Some(seg_delete_bitset) = segment_reader.delete_bitset() {
         for doc in 0u32..max_doc {
             if seg_delete_bitset.is_deleted(doc) {
@@ -188,23 +172,15 @@ pub(crate) fn advance_deletes(
         }
     }
 
-    let num_deleted_docs = delete_bitset.len() as u32;
+    let num_deleted_docs = delete_bitset.len();
-
+    if num_deleted_docs > 0 {
-    if num_deleted_docs > num_deleted_docs_before {
+        segment = segment.with_delete_meta(num_deleted_docs as u32, target_opstamp);
-        // We need to write a new delete file.
+        let mut delete_file = segment.open_write(SegmentComponent::DELETE)?;
-        let mut delete_file = segment
-            .with_delete_meta(num_deleted_docs, target_opstamp)
-            .open_write(SegmentComponent::DELETE)?;
         write_delete_bitset(&delete_bitset, max_doc, &mut delete_file)?;
         delete_file.terminate()?;
-        segment_entry.reset_delete_meta(num_deleted_docs as u32, target_opstamp);
     }
 
-    // Regardless of whether we did end up having to write a new file or not
+    segment_entry.set_meta(segment.meta().clone());
-    // we advance the `delete_cursor`. This is an optimisation. We want to ensure we do not
-    // check that a given deleted term does not match any of our docs more than once.
-    segment_entry.set_delete_cursor(delete_cursor);
-
     Ok(())
 }
 
@@ -213,12 +189,11 @@ fn index_documents(
     segment: Segment,
     grouped_document_iterator: &mut dyn Iterator<Item = OperationGroup>,
     segment_updater: &mut SegmentUpdater,
-    tokenizers: &TokenizerManager,
     mut delete_cursor: DeleteCursor,
 ) -> crate::Result<bool> {
-    let mut segment_writer =
-        SegmentWriter::for_segment(memory_budget, segment.clone(), tokenizers)?;
     let schema = segment.schema();
+
+    let mut segment_writer = SegmentWriter::for_segment(memory_budget, segment.clone(), &schema)?;
     for document_group in grouped_document_iterator {
         for doc in document_group {
             segment_writer.add_document(doc, &schema)?;
@@ -256,7 +231,11 @@ fn index_documents(
         last_docstamp,
     )?;
 
-    let segment_entry = SegmentEntry::new(segment_with_max_doc, delete_cursor, delete_bitset_opt);
+    let segment_entry = SegmentEntry::new(
+        segment_with_max_doc.meta().clone(),
+        delete_cursor,
+        delete_bitset_opt,
+    );
     block_on(segment_updater.schedule_add_segment(segment_entry))?;
     Ok(true)
 }
@@ -328,24 +307,16 @@ impl IndexWriter {
 
         let delete_queue = DeleteQueue::new();
 
-        let meta = index.load_metas()?;
+        let current_opstamp = index.load_metas()?.opstamp;
 
-        let stamper = Stamper::new(meta.opstamp);
+        let stamper = Stamper::new(current_opstamp);
-
-        let commited_segments = SegmentRegister::new(
-            index.directory(),
-            &index.schema(),
-            meta.segments,
-            &delete_queue.cursor(),
-        );
-        let segment_registers = Arc::new(RwLock::new(SegmentRegisters::new(commited_segments)));
 
         let segment_updater =
-            SegmentUpdater::create(segment_registers.clone(), index.clone(), stamper.clone())?;
+            SegmentUpdater::create(index.clone(), stamper.clone(), &delete_queue.cursor())?;
 
         let mut index_writer = IndexWriter {
             _directory_lock: Some(directory_lock),
-            segment_registers,
+
             heap_size_in_bytes_per_thread,
             index: index.clone(),
 
@@ -359,12 +330,10 @@ impl IndexWriter {
 
             delete_queue,
 
-            committed_opstamp: meta.opstamp,
+            committed_opstamp: current_opstamp,
             stamper,
 
             worker_id: 0,
-
-            on_commit: Default::default(),
         };
         index_writer.start_workers()?;
         Ok(index_writer)
@@ -404,6 +373,13 @@ impl IndexWriter {
         result
     }
 
+    #[doc(hidden)]
+    pub fn add_segment(&self, segment_meta: SegmentMeta) -> crate::Result<()> {
+        let delete_cursor = self.delete_queue.cursor();
+        let segment_entry = SegmentEntry::new(segment_meta, delete_cursor, None);
+        block_on(self.segment_updater.schedule_add_segment(segment_entry))
+    }
+
     /// Creates a new segment.
     ///
     /// This method is useful only for users trying to do complex
@@ -452,13 +428,12 @@ impl IndexWriter {
                         // was dropped.
                         return Ok(());
                     }
-                    let segment = index.new_segment_unpersisted();
+                    let segment = index.new_segment();
                     index_documents(
                         mem_budget,
                         segment,
                         &mut document_iterator,
                         &mut segment_updater,
-                        index.tokenizers(),
                         delete_cursor.clone(),
                     )?;
                 }
@@ -485,21 +460,6 @@ impl IndexWriter {
         Ok(())
     }
 
-    // TODO move me
-    pub(crate) fn trigger_commit(&self) -> impl Future<Output = ()> {
-        self.on_commit.broadcast()
-    }
-
-    pub fn reader(&self, num_searchers: usize) -> crate::Result<IndexReader> {
-        let nrt_reader = NRTReader::create(
-            num_searchers,
-            self.index.clone(),
-            self.segment_registers.clone(),
-            &self.on_commit,
-        )?;
-        Ok(IndexReader::NRT(nrt_reader))
-    }
-
     /// Detects and removes the files that are not used by the index anymore.
     pub fn garbage_collect_files(
         &self,
@@ -643,7 +603,7 @@ impl IndexWriter {
     /// It is also possible to add a payload to the `commit`
     /// using this API.
     /// See [`PreparedCommit::set_payload()`](PreparedCommit.html)
-    pub fn prepare_commit(&mut self, soft_commit: bool) -> crate::Result<PreparedCommit> {
+    pub fn prepare_commit(&mut self) -> crate::Result<PreparedCommit> {
         // Here, because we join all of the worker threads,
         // all of the segment update for this commit have been
         // sent.
@@ -671,7 +631,7 @@ impl IndexWriter {
         }
 
         let commit_opstamp = self.stamper.stamp();
-        let prepared_commit = PreparedCommit::new(self, commit_opstamp, soft_commit);
+        let prepared_commit = PreparedCommit::new(self, commit_opstamp);
         info!("Prepared commit {}", commit_opstamp);
         Ok(prepared_commit)
     }
@@ -691,11 +651,7 @@ impl IndexWriter {
     /// that made it in the commit.
     ///
     pub fn commit(&mut self) -> crate::Result<Opstamp> {
-        self.prepare_commit(false)?.commit()
+        self.prepare_commit()?.commit()
-    }
-
-    pub fn soft_commit(&mut self) -> crate::Result<Opstamp> {
-        self.prepare_commit(true)?.commit()
     }
 
     pub(crate) fn segment_updater(&self) -> &SegmentUpdater {
@@ -820,6 +776,7 @@ impl Drop for IndexWriter {
 
 #[cfg(test)]
 mod tests {
+
     use super::super::operation::UserOperation;
     use crate::collector::TopDocs;
     use crate::directory::error::LockError;
@@ -940,7 +897,7 @@ mod tests {
         let index_writer = index.writer(3_000_000).unwrap();
         assert_eq!(
             format!("{:?}", index_writer.get_merge_policy()),
-            "LogMergePolicy { min_merge_size: 8, min_layer_size: 10000, \
+            "LogMergePolicy { min_merge_size: 8, max_merge_size: 10000000, min_layer_size: 10000, \
              level_log_size: 0.75 }"
         );
         let merge_policy = Box::new(NoMergePolicy::default());
@@ -1052,8 +1009,7 @@ mod tests {
                 index_writer.add_document(doc!(text_field => "a"));
             }
             {
-                let mut prepared_commit =
+                let mut prepared_commit = index_writer.prepare_commit().expect("commit failed");
-                    index_writer.prepare_commit(false).expect("commit failed");
                 prepared_commit.set_payload("first commit");
                 prepared_commit.commit().expect("commit failed");
             }
@@ -1086,8 +1042,7 @@ mod tests {
                 index_writer.add_document(doc!(text_field => "a"));
             }
             {
-                let mut prepared_commit =
+                let mut prepared_commit = index_writer.prepare_commit().expect("commit failed");
-                    index_writer.prepare_commit(false).expect("commit failed");
                 prepared_commit.set_payload("first commit");
                 prepared_commit.abort().expect("commit failed");
             }
@@ -1262,43 +1217,7 @@ mod tests {
         let index = Index::create_in_ram(schema_builder.build());
         let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
         index_writer.add_document(doc!(idfield=>"myid"));
-        assert!(index_writer.commit().is_ok());
+        let commit = index_writer.commit();
-    }
+        assert!(commit.is_ok());
-
-    #[test]
-    fn test_index_writer_reader() {
-        let mut schema_builder = schema::Schema::builder();
-        let idfield = schema_builder.add_text_field("id", STRING);
-        schema_builder.add_text_field("optfield", STRING);
-        let index = Index::create_in_ram(schema_builder.build());
-        let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
-        index_writer.add_document(doc!(idfield=>"myid"));
-        assert!(index_writer.commit().is_ok());
-        let reader = index_writer.reader(2).unwrap();
-        let searcher = reader.searcher();
-        assert_eq!(searcher.num_docs(), 1u64);
-        index_writer.add_document(doc!(idfield=>"myid"));
-        assert!(index_writer.commit().is_ok());
-        assert_eq!(reader.searcher().num_docs(), 2u64);
-        assert_eq!(searcher.num_docs(), 1u64);
-    }
-
-    #[test]
-    fn test_index_writer_reader_soft_commit() {
-        let mut schema_builder = schema::Schema::builder();
-        let idfield = schema_builder.add_text_field("id", STRING);
-        schema_builder.add_text_field("optfield", STRING);
-        let index = Index::create_in_ram(schema_builder.build());
-        let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
-        index_writer.add_document(doc!(idfield=>"myid"));
-        assert!(index_writer.soft_commit().is_ok());
-        let nrt_reader = index_writer.reader(2).unwrap();
-        let normal_reader = index.reader_builder().try_into().unwrap();
-        assert_eq!(nrt_reader.searcher().num_docs(), 1u64);
-        assert_eq!(normal_reader.searcher().num_docs(), 0u64);
-        assert!(index_writer.commit().is_ok());
-        assert!(normal_reader.reload().is_ok());
-        assert_eq!(nrt_reader.searcher().num_docs(), 1u64);
-        assert_eq!(normal_reader.searcher().num_docs(), 1u64);
     }
 }

src/indexer/log_merge_policy.rs

@@ -6,12 +6,14 @@ use std::f64;
 const DEFAULT_LEVEL_LOG_SIZE: f64 = 0.75;
 const DEFAULT_MIN_LAYER_SIZE: u32 = 10_000;
 const DEFAULT_MIN_MERGE_SIZE: usize = 8;
+const DEFAULT_MAX_MERGE_SIZE: usize = 10_000_000;
 
 /// `LogMergePolicy` tries tries to merge segments that have a similar number of
 /// documents.
 #[derive(Debug, Clone)]
 pub struct LogMergePolicy {
     min_merge_size: usize,
+    max_merge_size: usize,
     min_layer_size: u32,
     level_log_size: f64,
 }
@@ -26,6 +28,12 @@ impl LogMergePolicy {
         self.min_merge_size = min_merge_size;
     }
 
+    /// Set the maximum number docs in a segment for it to be considered for
+    /// merging.
+    pub fn set_max_merge_size(&mut self, max_merge_size: usize) {
+        self.max_merge_size = max_merge_size;
+    }
+
     /// Set the minimum segment size under which all segment belong
     /// to the same level.
     pub fn set_min_layer_size(&mut self, min_layer_size: u32) {
@@ -53,6 +61,7 @@ impl MergePolicy for LogMergePolicy {
         let mut size_sorted_tuples = segments
             .iter()
             .map(SegmentMeta::num_docs)
+            .filter(|s| s <= &(self.max_merge_size as u32))
             .enumerate()
             .collect::<Vec<(usize, u32)>>();
 
@@ -86,6 +95,7 @@ impl Default for LogMergePolicy {
     fn default() -> LogMergePolicy {
         LogMergePolicy {
             min_merge_size: DEFAULT_MIN_MERGE_SIZE,
+            max_merge_size: DEFAULT_MAX_MERGE_SIZE,
             min_layer_size: DEFAULT_MIN_LAYER_SIZE,
             level_log_size: DEFAULT_LEVEL_LOG_SIZE,
         }
@@ -104,6 +114,7 @@ mod tests {
     fn test_merge_policy() -> LogMergePolicy {
         let mut log_merge_policy = LogMergePolicy::default();
         log_merge_policy.set_min_merge_size(3);
+        log_merge_policy.set_max_merge_size(100_000);
         log_merge_policy.set_min_layer_size(2);
         log_merge_policy
     }
@@ -141,11 +152,11 @@ mod tests {
             create_random_segment_meta(10),
             create_random_segment_meta(10),
             create_random_segment_meta(10),
-            create_random_segment_meta(1000),
+            create_random_segment_meta(1_000),
-            create_random_segment_meta(1000),
+            create_random_segment_meta(1_000),
-            create_random_segment_meta(1000),
+            create_random_segment_meta(1_000),
-            create_random_segment_meta(10000),
+            create_random_segment_meta(10_000),
-            create_random_segment_meta(10000),
+            create_random_segment_meta(10_000),
             create_random_segment_meta(10),
             create_random_segment_meta(10),
             create_random_segment_meta(10),
@@ -182,4 +193,19 @@ mod tests {
         let result_list = test_merge_policy().compute_merge_candidates(&test_input);
         assert_eq!(result_list.len(), 1);
     }
+
+    #[test]
+    fn test_large_merge_segments() {
+        let test_input = vec![
+            create_random_segment_meta(1_000_000),
+            create_random_segment_meta(100_001),
+            create_random_segment_meta(100_000),
+            create_random_segment_meta(100_000),
+            create_random_segment_meta(100_000),
+        ];
+        let result_list = test_merge_policy().compute_merge_candidates(&test_input);
+        // Do not include large segments
+        assert_eq!(result_list.len(), 1);
+        assert_eq!(result_list[0].0.len(), 3)
+    }
 }

src/indexer/mod.rs

@@ -23,7 +23,6 @@ pub use self::merge_policy::{MergeCandidate, MergePolicy, NoMergePolicy};
 pub use self::prepared_commit::PreparedCommit;
 pub use self::segment_entry::SegmentEntry;
 pub use self::segment_manager::SegmentManager;
-pub(crate) use self::segment_manager::SegmentRegisters;
 pub use self::segment_serializer::SegmentSerializer;
 pub use self::segment_writer::SegmentWriter;
 

src/indexer/operation.rs

@@ -19,6 +19,8 @@ pub struct AddOperation {
 /// UserOperation is an enum type that encapsulates other operation types.
 #[derive(Eq, PartialEq, Debug)]
 pub enum UserOperation {
+    /// Add operation
     Add(Document),
+    /// Delete operation
     Delete(Term),
 }

src/indexer/prepared_commit.rs

@@ -8,20 +8,14 @@ pub struct PreparedCommit<'a> {
     index_writer: &'a mut IndexWriter,
     payload: Option<String>,
     opstamp: Opstamp,
-    soft_commit: bool,
 }
 
 impl<'a> PreparedCommit<'a> {
-    pub(crate) fn new(
+    pub(crate) fn new(index_writer: &'a mut IndexWriter, opstamp: Opstamp) -> PreparedCommit<'_> {
-        index_writer: &'a mut IndexWriter,
-        opstamp: Opstamp,
-        soft_commit: bool,
-    ) -> PreparedCommit {
         PreparedCommit {
             index_writer,
             payload: None,
             opstamp,
-            soft_commit,
         }
     }
 
@@ -39,12 +33,11 @@ impl<'a> PreparedCommit<'a> {
 
     pub fn commit(self) -> Result<Opstamp> {
         info!("committing {}", self.opstamp);
-        block_on(self.index_writer.segment_updater().schedule_commit(
+        let _ = block_on(
-            self.opstamp,
+            self.index_writer
-            self.payload,
+                .segment_updater()
-            self.soft_commit,
+                .schedule_commit(self.opstamp, self.payload),
-        ))?;
+        );
-        block_on(self.index_writer.trigger_commit());
         Ok(self.opstamp)
     }
 }

src/indexer/segment_entry.rs

@@ -1,9 +1,7 @@
 use crate::common::BitSet;
 use crate::core::SegmentId;
 use crate::core::SegmentMeta;
-use crate::directory::ManagedDirectory;
 use crate::indexer::delete_queue::DeleteCursor;
-use crate::{Opstamp, Segment};
 use std::fmt;
 
 /// A segment entry describes the state of
@@ -21,81 +19,55 @@ use std::fmt;
 /// in the .del file or in the `delete_bitset`.
 #[derive(Clone)]
 pub struct SegmentEntry {
-    segment: Segment,
+    meta: SegmentMeta,
     delete_bitset: Option<BitSet>,
     delete_cursor: DeleteCursor,
 }
 
 impl SegmentEntry {
     /// Create a new `SegmentEntry`
-    pub(crate) fn new(
+    pub fn new(
-        segment: Segment,
+        segment_meta: SegmentMeta,
         delete_cursor: DeleteCursor,
         delete_bitset: Option<BitSet>,
     ) -> SegmentEntry {
         SegmentEntry {
-            segment,
+            meta: segment_meta,
             delete_bitset,
             delete_cursor,
         }
     }
 
-    pub fn persist(&mut self, dest_directory: ManagedDirectory) -> crate::Result<()> {
+    /// Return a reference to the segment entry deleted bitset.
-        // TODO take in account delete bitset?
-        self.segment.persist(dest_directory)?;
-        Ok(())
-    }
-
-    pub fn segment(&self) -> &Segment {
-        &self.segment
-    }
-
-    /// `Takes` (as in Option::take) the delete bitset of
-    /// a segment entry.
     ///
     /// `DocId` in this bitset are flagged as deleted.
-    pub fn take_delete_bitset(&mut self) -> Option<BitSet> {
+    pub fn delete_bitset(&self) -> Option<&BitSet> {
-        self.delete_bitset.take()
+        self.delete_bitset.as_ref()
     }
 
-    /// Reset the delete informmationo in this segment.
+    /// Set the `SegmentMeta` for this segment.
-    ///
+    pub fn set_meta(&mut self, segment_meta: SegmentMeta) {
-    /// The `SegmentEntry` segment's `SegmentMeta` gets updated, and
+        self.meta = segment_meta;
-    /// any delete bitset is drop and set to None.
-    pub fn reset_delete_meta(&mut self, num_deleted_docs: u32, target_opstamp: Opstamp) {
-        self.segment = self
-            .segment
-            .clone()
-            .with_delete_meta(num_deleted_docs, target_opstamp);
-        self.delete_bitset = None;
     }
 
-    pub fn set_delete_cursor(&mut self, delete_cursor: DeleteCursor) {
-        self.delete_cursor = delete_cursor;
-    }
     /// Return a reference to the segment_entry's delete cursor
-    pub fn delete_cursor(&mut self) -> DeleteCursor {
+    pub fn delete_cursor(&mut self) -> &mut DeleteCursor {
-        self.delete_cursor.clone()
+        &mut self.delete_cursor
     }
 
     /// Returns the segment id.
     pub fn segment_id(&self) -> SegmentId {
-        self.segment.id()
+        self.meta.id()
     }
 
     /// Accessor to the `SegmentMeta`
     pub fn meta(&self) -> &SegmentMeta {
-        self.segment.meta()
+        &self.meta
     }
 }
 
 impl fmt::Debug for SegmentEntry {
     fn fmt(&self, formatter: &mut fmt::Formatter<'_>) -> fmt::Result {
-        let num_deletes = self.delete_bitset.as_ref().map(|bitset| bitset.len());
+        write!(formatter, "SegmentEntry({:?})", self.meta)
-        write!(
-            formatter,
-            "SegmentEntry(seg={:?}, ndel={:?})",
-            self.segment, num_deletes
-        )
     }
 }

src/indexer/segment_manager.rs

@@ -1,14 +1,17 @@
 use super::segment_register::SegmentRegister;
 use crate::core::SegmentId;
 use crate::core::SegmentMeta;
+use crate::error::TantivyError;
+use crate::indexer::delete_queue::DeleteCursor;
 use crate::indexer::SegmentEntry;
-use crate::Segment;
+use crate::Result as TantivyResult;
 use std::collections::hash_set::HashSet;
-use std::sync::{Arc, RwLock};
+use std::fmt::{self, Debug, Formatter};
+use std::sync::RwLock;
 use std::sync::{RwLockReadGuard, RwLockWriteGuard};
 
 #[derive(Default)]
-pub(crate) struct SegmentRegisters {
+struct SegmentRegisters {
     uncommitted: SegmentRegister,
     committed: SegmentRegister,
 }
@@ -20,17 +23,6 @@ pub(crate) enum SegmentsStatus {
 }
 
 impl SegmentRegisters {
-    pub fn new(committed: SegmentRegister) -> SegmentRegisters {
-        SegmentRegisters {
-            uncommitted: Default::default(),
-            committed,
-        }
-    }
-
-    pub fn committed_segment(&self) -> Vec<Segment> {
-        self.committed.segments()
-    }
-
     /// Check if all the segments are committed or uncommited.
     ///
     /// If some segment is missing or segments are in a different state (this should not happen
@@ -53,7 +45,18 @@ impl SegmentRegisters {
 /// changes (merges especially)
 #[derive(Default)]
 pub struct SegmentManager {
-    registers: Arc<RwLock<SegmentRegisters>>,
+    registers: RwLock<SegmentRegisters>,
+}
+
+impl Debug for SegmentManager {
+    fn fmt(&self, f: &mut Formatter<'_>) -> Result<(), fmt::Error> {
+        let lock = self.read();
+        write!(
+            f,
+            "{{ uncommitted: {:?}, committed: {:?} }}",
+            lock.uncommitted, lock.committed
+        )
+    }
 }
 
 pub fn get_mergeable_segments(
@@ -72,8 +75,16 @@ pub fn get_mergeable_segments(
 }
 
 impl SegmentManager {
-    pub(crate) fn new(registers: Arc<RwLock<SegmentRegisters>>) -> SegmentManager {
+    pub fn from_segments(
-        SegmentManager { registers }
+        segment_metas: Vec<SegmentMeta>,
+        delete_cursor: &DeleteCursor,
+    ) -> SegmentManager {
+        SegmentManager {
+            registers: RwLock::new(SegmentRegisters {
+                uncommitted: SegmentRegister::default(),
+                committed: SegmentRegister::new(segment_metas, delete_cursor),
+            }),
+        }
     }
 
     /// Returns all of the segment entries (committed or uncommitted)
@@ -134,7 +145,7 @@ impl SegmentManager {
     /// Returns an error if some segments are missing, or if
     /// the `segment_ids` are not either all committed or all
     /// uncommitted.
-    pub fn start_merge(&self, segment_ids: &[SegmentId]) -> crate::Result<Vec<SegmentEntry>> {
+    pub fn start_merge(&self, segment_ids: &[SegmentId]) -> TantivyResult<Vec<SegmentEntry>> {
         let registers_lock = self.read();
         let mut segment_entries = vec![];
         if registers_lock.uncommitted.contains_all(segment_ids) {
@@ -155,7 +166,7 @@ impl SegmentManager {
             let error_msg = "Merge operation sent for segments that are not \
                              all uncommited or commited."
                 .to_string();
-            return Err(crate::Error::InvalidArgument(error_msg));
+            return Err(TantivyError::InvalidArgument(error_msg));
         }
         Ok(segment_entries)
     }

src/indexer/segment_register.rs

@@ -1,10 +1,7 @@
 use crate::core::SegmentId;
 use crate::core::SegmentMeta;
-use crate::directory::ManagedDirectory;
 use crate::indexer::delete_queue::DeleteCursor;
 use crate::indexer::segment_entry::SegmentEntry;
-use crate::schema::Schema;
-use crate::Segment;
 use std::collections::HashMap;
 use std::collections::HashSet;
 use std::fmt::{self, Debug, Formatter};
@@ -49,13 +46,6 @@ impl SegmentRegister {
             .collect()
     }
 
-    pub fn segments(&self) -> Vec<Segment> {
-        self.segment_states
-            .values()
-            .map(|segment_entry| segment_entry.segment().clone())
-            .collect()
-    }
-
     pub fn segment_entries(&self) -> Vec<SegmentEntry> {
         self.segment_states.values().cloned().collect()
     }
@@ -89,17 +79,11 @@ impl SegmentRegister {
         self.segment_states.get(segment_id).cloned()
     }
 
-    pub fn new(
+    pub fn new(segment_metas: Vec<SegmentMeta>, delete_cursor: &DeleteCursor) -> SegmentRegister {
-        directory: &ManagedDirectory,
-        schema: &Schema,
-        segment_metas: Vec<SegmentMeta>,
-        delete_cursor: &DeleteCursor,
-    ) -> SegmentRegister {
         let mut segment_states = HashMap::new();
-        for meta in segment_metas {
+        for segment_meta in segment_metas {
-            let segment_id = meta.id();
+            let segment_id = segment_meta.id();
-            let segment = Segment::new_persisted(meta, directory.clone(), schema.clone());
+            let segment_entry = SegmentEntry::new(segment_meta, delete_cursor.clone(), None);
-            let segment_entry = SegmentEntry::new(segment, delete_cursor.clone(), None);
             segment_states.insert(segment_id, segment_entry);
         }
         SegmentRegister { segment_states }
@@ -111,7 +95,6 @@ mod tests {
     use super::*;
     use crate::core::{SegmentId, SegmentMetaInventory};
     use crate::indexer::delete_queue::*;
-    use crate::schema::Schema;
 
     fn segment_ids(segment_register: &SegmentRegister) -> Vec<SegmentId> {
         segment_register
@@ -125,7 +108,6 @@ mod tests {
     fn test_segment_register() {
         let inventory = SegmentMetaInventory::default();
         let delete_queue = DeleteQueue::new();
-        let schema = Schema::builder().build();
 
         let mut segment_register = SegmentRegister::default();
         let segment_id_a = SegmentId::generate_random();
@@ -133,24 +115,21 @@ mod tests {
         let segment_id_merged = SegmentId::generate_random();
 
         {
-            let meta = inventory.new_segment_meta(segment_id_a, 0u32);
+            let segment_meta = inventory.new_segment_meta(segment_id_a, 0u32);
-            let segment = Segment::new_volatile(meta, schema.clone());
+            let segment_entry = SegmentEntry::new(segment_meta, delete_queue.cursor(), None);
-            let segment_entry = SegmentEntry::new(segment, delete_queue.cursor(), None);
             segment_register.add_segment_entry(segment_entry);
         }
         assert_eq!(segment_ids(&segment_register), vec![segment_id_a]);
         {
             let segment_meta = inventory.new_segment_meta(segment_id_b, 0u32);
-            let segment = Segment::new_volatile(segment_meta, schema.clone());
+            let segment_entry = SegmentEntry::new(segment_meta, delete_queue.cursor(), None);
-            let segment_entry = SegmentEntry::new(segment, delete_queue.cursor(), None);
             segment_register.add_segment_entry(segment_entry);
         }
         segment_register.remove_segment(&segment_id_a);
         segment_register.remove_segment(&segment_id_b);
         {
             let segment_meta_merged = inventory.new_segment_meta(segment_id_merged, 0u32);
-            let segment = Segment::new_volatile(segment_meta_merged, schema);
+            let segment_entry = SegmentEntry::new(segment_meta_merged, delete_queue.cursor(), None);
-            let segment_entry = SegmentEntry::new(segment, delete_queue.cursor(), None);
             segment_register.add_segment_entry(segment_entry);
         }
         assert_eq!(segment_ids(&segment_register), vec![segment_id_merged]);

src/indexer/segment_updater.rs

@@ -7,10 +7,11 @@ use crate::core::SegmentMeta;
 use crate::core::SerializableSegment;
 use crate::core::META_FILEPATH;
 use crate::directory::{Directory, DirectoryClone, GarbageCollectionResult};
+use crate::indexer::delete_queue::DeleteCursor;
 use crate::indexer::index_writer::advance_deletes;
 use crate::indexer::merge_operation::MergeOperationInventory;
 use crate::indexer::merger::IndexMerger;
-use crate::indexer::segment_manager::{SegmentRegisters, SegmentsStatus};
+use crate::indexer::segment_manager::SegmentsStatus;
 use crate::indexer::stamper::Stamper;
 use crate::indexer::SegmentEntry;
 use crate::indexer::SegmentSerializer;
@@ -116,7 +117,8 @@ fn merge(
 
     // First we apply all of the delet to the merged segment, up to the target opstamp.
     for segment_entry in &mut segment_entries {
-        advance_deletes(segment_entry, target_opstamp)?;
+        let segment = index.segment(segment_entry.meta().clone());
+        advance_deletes(segment, segment_entry, target_opstamp)?;
     }
 
     let delete_cursor = segment_entries[0].delete_cursor().clone();
@@ -132,13 +134,11 @@ fn merge(
     // ... we just serialize this index merger in our new segment to merge the two segments.
     let segment_serializer = SegmentSerializer::for_segment(&mut merged_segment)?;
 
-    let max_doc = merger.write(segment_serializer)?;
+    let num_docs = merger.write(segment_serializer)?;
 
-    Ok(SegmentEntry::new(
+    let segment_meta = index.new_segment_meta(merged_segment.id(), num_docs);
-        merged_segment.with_max_doc(max_doc),
+
-        delete_cursor,
+    Ok(SegmentEntry::new(segment_meta, delete_cursor, None))
-        None,
-    ))
 }
 
 pub(crate) struct InnerSegmentUpdater {
@@ -162,11 +162,12 @@ pub(crate) struct InnerSegmentUpdater {
 
 impl SegmentUpdater {
     pub fn create(
-        segment_registers: Arc<RwLock<SegmentRegisters>>,
         index: Index,
         stamper: Stamper,
+        delete_cursor: &DeleteCursor,
     ) -> crate::Result<SegmentUpdater> {
-        let segment_manager = SegmentManager::new(segment_registers);
+        let segments = index.searchable_segment_metas()?;
+        let segment_manager = SegmentManager::from_segments(segments, delete_cursor);
         let pool = ThreadPoolBuilder::new()
             .name_prefix("segment_updater")
             .pool_size(1)
@@ -229,7 +230,6 @@ impl SegmentUpdater {
         &self,
         segment_entry: SegmentEntry,
     ) -> impl Future<Output = crate::Result<()>> {
-        // TODO temporary: serializing the segment at this point.
         let segment_updater = self.clone();
         self.schedule_future(async move {
             segment_updater.segment_manager.add_segment(segment_entry);
@@ -258,7 +258,8 @@ impl SegmentUpdater {
     fn purge_deletes(&self, target_opstamp: Opstamp) -> crate::Result<Vec<SegmentEntry>> {
         let mut segment_entries = self.segment_manager.segment_entries();
         for segment_entry in &mut segment_entries {
-            advance_deletes(segment_entry, target_opstamp)?;
+            let segment = self.index.segment(segment_entry.meta().clone());
+            advance_deletes(segment, segment_entry, target_opstamp)?;
         }
         Ok(segment_entries)
     }
@@ -326,21 +327,12 @@ impl SegmentUpdater {
         &self,
         opstamp: Opstamp,
         payload: Option<String>,
-        soft_commit: bool,
     ) -> impl Future<Output = crate::Result<()>> {
         let segment_updater: SegmentUpdater = self.clone();
-        let directory = self.index.directory().clone();
         self.schedule_future(async move {
-            let mut segment_entries = segment_updater.purge_deletes(opstamp)?;
+            let segment_entries = segment_updater.purge_deletes(opstamp)?;
-            if !soft_commit {
-                for segment_entry in &mut segment_entries {
-                    segment_entry.persist(directory.clone())?;
-                }
-            }
             segment_updater.segment_manager.commit(segment_entries);
-            if !soft_commit {
+            segment_updater.save_metas(opstamp, payload)?;
-                segment_updater.save_metas(opstamp, payload)?;
-            }
             let _ = garbage_collect_files(segment_updater.clone()).await;
             segment_updater.consider_merge_options().await;
             Ok(())
@@ -478,14 +470,17 @@ impl SegmentUpdater {
         let end_merge_future = self.schedule_future(async move {
             info!("End merge {:?}", after_merge_segment_entry.meta());
             {
-                let mut delete_cursor = after_merge_segment_entry.delete_cursor();
+                let mut delete_cursor = after_merge_segment_entry.delete_cursor().clone();
                 if let Some(delete_operation) = delete_cursor.get() {
                     let committed_opstamp = segment_updater.load_metas().opstamp;
                     if delete_operation.opstamp < committed_opstamp {
-                        let _index = &segment_updater.index;
+                        let index = &segment_updater.index;
-                        if let Err(e) =
+                        let segment = index.segment(after_merge_segment_entry.meta().clone());
-                            advance_deletes(&mut after_merge_segment_entry, committed_opstamp)
+                        if let Err(e) = advance_deletes(
-                        {
+                            segment,
+                            &mut after_merge_segment_entry,
+                            committed_opstamp,
+                        ) {
                             error!(
                                 "Merge of {:?} was cancelled (advancing deletes failed): {:?}",
                                 merge_operation.segment_ids(),

src/indexer/segment_writer.rs

@@ -11,10 +11,9 @@ use crate::schema::Schema;
 use crate::schema::Term;
 use crate::schema::Value;
 use crate::schema::{Field, FieldEntry};
-use crate::tokenizer::FacetTokenizer;
+use crate::tokenizer::{BoxTokenStream, PreTokenizedStream};
-use crate::tokenizer::PreTokenizedStream;
+use crate::tokenizer::{FacetTokenizer, TextAnalyzer};
-use crate::tokenizer::{BoxedTokenizer, TokenizerManager};
+use crate::tokenizer::{TokenStreamChain, Tokenizer};
-use crate::tokenizer::{TokenStream, TokenStreamChain, Tokenizer};
 use crate::DocId;
 use crate::Opstamp;
 use crate::Result;
@@ -50,7 +49,7 @@ pub struct SegmentWriter {
     fast_field_writers: FastFieldsWriter,
     fieldnorms_writer: FieldNormsWriter,
     doc_opstamps: Vec<Opstamp>,
-    tokenizers: Vec<Option<BoxedTokenizer>>,
+    tokenizers: Vec<Option<TextAnalyzer>>,
 }
 
 impl SegmentWriter {
@@ -66,12 +65,11 @@ impl SegmentWriter {
     pub fn for_segment(
         memory_budget: usize,
         mut segment: Segment,
-        tokenizers: &TokenizerManager,
+        schema: &Schema,
     ) -> Result<SegmentWriter> {
-        let schema = segment.schema();
         let table_num_bits = initial_table_size(memory_budget)?;
         let segment_serializer = SegmentSerializer::for_segment(&mut segment)?;
-        let multifield_postings = MultiFieldPostingsWriter::new(&schema, table_num_bits);
+        let multifield_postings = MultiFieldPostingsWriter::new(schema, table_num_bits);
         let tokenizers = schema
             .fields()
             .map(
@@ -80,7 +78,7 @@ impl SegmentWriter {
                         .get_indexing_options()
                         .and_then(|text_index_option| {
                             let tokenizer_name = &text_index_option.tokenizer();
-                            tokenizers.get(tokenizer_name)
+                            segment.index().tokenizers().get(tokenizer_name)
                         }),
                     _ => None,
                 },
@@ -89,9 +87,9 @@ impl SegmentWriter {
         Ok(SegmentWriter {
             max_doc: 0,
             multifield_postings,
-            fieldnorms_writer: FieldNormsWriter::for_schema(&schema),
+            fieldnorms_writer: FieldNormsWriter::for_schema(schema),
             segment_serializer,
-            fast_field_writers: FastFieldsWriter::from_schema(&schema),
+            fast_field_writers: FastFieldsWriter::from_schema(schema),
             doc_opstamps: Vec::with_capacity(1_000),
             tokenizers,
         })
@@ -160,7 +158,7 @@ impl SegmentWriter {
                     }
                 }
                 FieldType::Str(_) => {
-                    let mut token_streams: Vec<Box<dyn TokenStream>> = vec![];
+                    let mut token_streams: Vec<BoxTokenStream> = vec![];
                     let mut offsets = vec![];
                     let mut total_offset = 0;
 
@@ -173,7 +171,7 @@ impl SegmentWriter {
                                 }
 
                                 token_streams
-                                    .push(Box::new(PreTokenizedStream::from(tok_str.clone())));
+                                    .push(PreTokenizedStream::from(tok_str.clone()).into());
                             }
                             Value::Str(ref text) => {
                                 if let Some(ref mut tokenizer) =
@@ -192,8 +190,7 @@ impl SegmentWriter {
                     let num_tokens = if token_streams.is_empty() {
                         0
                     } else {
-                        let mut token_stream: Box<dyn TokenStream> =
+                        let mut token_stream = TokenStreamChain::new(offsets, token_streams);
-                            Box::new(TokenStreamChain::new(offsets, token_streams));
                         self.multifield_postings
                             .index_text(doc_id, field, &mut token_stream)
                     };

src/indexer/stamper.rs

@@ -1,18 +1,76 @@
 use crate::Opstamp;
 use std::ops::Range;
-use std::sync::atomic::{AtomicU64, Ordering};
+use std::sync::atomic::Ordering;
 use std::sync::Arc;
 
+#[cfg(not(target_arch = "arm"))]
+mod atomic_impl {
+
+    use crate::Opstamp;
+    use std::sync::atomic::{AtomicU64, Ordering};
+
+    #[derive(Default)]
+    pub struct AtomicU64Wrapper(AtomicU64);
+
+    impl AtomicU64Wrapper {
+        pub fn new(first_opstamp: Opstamp) -> AtomicU64Wrapper {
+            AtomicU64Wrapper(AtomicU64::new(first_opstamp as u64))
+        }
+
+        pub fn fetch_add(&self, val: u64, order: Ordering) -> u64 {
+            self.0.fetch_add(val as u64, order) as u64
+        }
+
+        pub fn revert(&self, val: u64, order: Ordering) -> u64 {
+            self.0.store(val, order);
+            val
+        }
+    }
+}
+
+#[cfg(target_arch = "arm")]
+mod atomic_impl {
+
+    use crate::Opstamp;
+    /// Under other architecture, we rely on a mutex.
+    use std::sync::atomic::Ordering;
+    use std::sync::RwLock;
+
+    #[derive(Default)]
+    pub struct AtomicU64Wrapper(RwLock<u64>);
+
+    impl AtomicU64Wrapper {
+        pub fn new(first_opstamp: Opstamp) -> AtomicU64Wrapper {
+            AtomicU64Wrapper(RwLock::new(first_opstamp))
+        }
+
+        pub fn fetch_add(&self, incr: u64, _order: Ordering) -> u64 {
+            let mut lock = self.0.write().unwrap();
+            let previous_val = *lock;
+            *lock = previous_val + incr;
+            previous_val
+        }
+
+        pub fn revert(&self, val: u64, _order: Ordering) -> u64 {
+            let mut lock = self.0.write().unwrap();
+            *lock = val;
+            val
+        }
+    }
+}
+
+use self::atomic_impl::AtomicU64Wrapper;
+
 /// Stamper provides Opstamps, which is just an auto-increment id to label
 /// an operation.
 ///
 /// Cloning does not "fork" the stamp generation. The stamper actually wraps an `Arc`.
 #[derive(Clone, Default)]
-pub struct Stamper(Arc<AtomicU64>);
+pub struct Stamper(Arc<AtomicU64Wrapper>);
 
 impl Stamper {
     pub fn new(first_opstamp: Opstamp) -> Stamper {
-        Stamper(Arc::new(AtomicU64::new(first_opstamp)))
+        Stamper(Arc::new(AtomicU64Wrapper::new(first_opstamp)))
     }
 
     pub fn stamp(&self) -> Opstamp {
@@ -31,8 +89,7 @@ impl Stamper {
 
     /// Reverts the stamper to a given `Opstamp` value and returns it
     pub fn revert(&self, to_opstamp: Opstamp) -> Opstamp {
-        self.0.store(to_opstamp, Ordering::SeqCst);
+        self.0.revert(to_opstamp, Ordering::SeqCst)
-        to_opstamp
     }
 }
 

src/lib.rs

@@ -161,10 +161,11 @@ pub use self::snippet::{Snippet, SnippetGenerator};
 mod docset;
 pub use self::docset::{DocSet, SkipResult};
 pub use crate::common::{f64_to_u64, i64_to_u64, u64_to_f64, u64_to_i64};
-pub use crate::core::SegmentComponent;
+pub use crate::core::{Executor, SegmentComponent};
 pub use crate::core::{Index, IndexMeta, Searcher, Segment, SegmentId, SegmentMeta};
 pub use crate::core::{InvertedIndexReader, SegmentReader};
 pub use crate::directory::Directory;
+pub use crate::indexer::operation::UserOperation;
 pub use crate::indexer::IndexWriter;
 pub use crate::postings::Postings;
 pub use crate::reader::LeasedItem;

src/postings/mod.rs

@@ -220,7 +220,7 @@ pub mod tests {
 
         {
             let mut segment_writer =
-                SegmentWriter::for_segment(3_000_000, segment.clone(), index.tokenizers()).unwrap();
+                SegmentWriter::for_segment(3_000_000, segment.clone(), &schema).unwrap();
             {
                 let mut doc = Document::default();
                 // checking that position works if the field has two values

src/postings/serializer.rs

@@ -148,8 +148,7 @@ impl<'a> FieldSerializer<'a> {
             }
             _ => (false, false),
         };
-        let term_dictionary_builder =
+        let term_dictionary_builder = TermDictionaryBuilder::create(term_dictionary_write)?;
-            TermDictionaryBuilder::create(term_dictionary_write, &field_type)?;
         let postings_serializer =
             PostingsSerializer::new(postings_write, term_freq_enabled, position_enabled);
         let positions_serializer_opt = if position_enabled {

src/query/automaton_weight.rs

@@ -15,6 +15,7 @@ use tantivy_fst::Automaton;
 pub struct AutomatonWeight<A> {
     field: Field,
     automaton: Arc<A>,
+    boost: f32,
 }
 
 impl<A> AutomatonWeight<A>
@@ -26,9 +27,15 @@ where
         AutomatonWeight {
             field,
             automaton: automaton.into(),
+            boost: 1.0,
         }
     }
 
+    /// Boost the scorer by the given factor.
+    pub fn boost_by(self, boost: f32) -> Self {
+        Self { boost, ..self }
+    }
+
     fn automaton_stream<'a>(&'a self, term_dict: &'a TermDictionary) -> TermStreamer<'a, &'a A> {
         let automaton: &A = &*self.automaton;
         let term_stream_builder = term_dict.search(automaton);
@@ -58,7 +65,7 @@ where
             }
         }
         let doc_bitset = BitSetDocSet::from(doc_bitset);
-        Ok(Box::new(ConstScorer::new(doc_bitset)))
+        Ok(Box::new(ConstScorer::with_score(doc_bitset, self.boost)))
     }
 
     fn explain(&self, reader: &SegmentReader, doc: DocId) -> Result<Explanation> {

src/query/bm25.rs

@@ -34,7 +34,7 @@ pub struct BM25Weight {
 }
 
 impl BM25Weight {
-    pub fn for_terms(searcher: &Searcher, terms: &[Term]) -> BM25Weight {
+    pub fn for_terms(searcher: &Searcher, terms: &[Term], boost: f32) -> BM25Weight {
         assert!(!terms.is_empty(), "BM25 requires at least one term");
         let field = terms[0].field();
         for term in &terms[1..] {
@@ -75,11 +75,11 @@ impl BM25Weight {
                 .sum::<f32>();
             idf_explain = Explanation::new("idf", idf);
         }
-        BM25Weight::new(idf_explain, average_fieldnorm)
+        BM25Weight::new(idf_explain, average_fieldnorm, boost)
     }
 
-    fn new(idf_explain: Explanation, average_fieldnorm: f32) -> BM25Weight {
+    fn new(idf_explain: Explanation, average_fieldnorm: f32, boost: f32) -> BM25Weight {
-        let weight = idf_explain.value() * (1f32 + K1);
+        let weight = idf_explain.value() * (1f32 + K1) * boost;
         BM25Weight {
             idf_explain,
             weight,

src/query/fuzzy_query.rs

@@ -79,6 +79,7 @@ pub struct FuzzyTermQuery {
     transposition_cost_one: bool,
     ///
     prefix: bool,
+    boost: f32,
 }
 
 impl FuzzyTermQuery {
@@ -89,6 +90,7 @@ impl FuzzyTermQuery {
             distance,
             transposition_cost_one,
             prefix: false,
+            boost: 1.0,
         }
     }
 
@@ -99,16 +101,22 @@ impl FuzzyTermQuery {
             distance,
             transposition_cost_one,
             prefix: true,
+            boost: 1.0,
         }
     }
 
+    /// Boost the query score by the given factor.
+    pub fn boost_by(self, boost: f32) -> Self {
+        Self { boost, ..self }
+    }
+
     fn specialized_weight(&self) -> Result<AutomatonWeight<DFA>> {
         // LEV_BUILDER is a HashMap, whose `get` method returns an Option
         match LEV_BUILDER.get(&(self.distance, false)) {
             // Unwrap the option and build the Ok(AutomatonWeight)
             Some(automaton_builder) => {
                 let automaton = automaton_builder.build_dfa(self.term.text());
-                Ok(AutomatonWeight::new(self.term.field(), automaton))
+                Ok(AutomatonWeight::new(self.term.field(), automaton).boost_by(self.boost))
             }
             None => Err(InvalidArgument(format!(
                 "Levenshtein distance of {} is not allowed. Choose a value in the {:?} range",

src/query/phrase_query/phrase_query.rs

@@ -27,6 +27,7 @@ use std::collections::BTreeSet;
 pub struct PhraseQuery {
     field: Field,
     phrase_terms: Vec<(usize, Term)>,
+    boost: f32,
 }
 
 impl PhraseQuery {
@@ -57,9 +58,15 @@ impl PhraseQuery {
         PhraseQuery {
             field,
             phrase_terms: terms,
+            boost: 1.0,
         }
     }
 
+    /// Boost the query score by the given factor.
+    pub fn boost_by(self, boost: f32) -> Self {
+        Self { boost, ..self }
+    }
+
     /// The `Field` this `PhraseQuery` is targeting.
     pub fn field(&self) -> Field {
         self.field
@@ -97,7 +104,7 @@ impl PhraseQuery {
             )));
         }
         let terms = self.phrase_terms();
-        let bm25_weight = BM25Weight::for_terms(searcher, &terms);
+        let bm25_weight = BM25Weight::for_terms(searcher, &terms, self.boost);
         Ok(PhraseWeight::new(
             self.phrase_terms.clone(),
             bm25_weight,

src/query/query_parser/query_parser.rs

@@ -533,7 +533,7 @@ mod test {
     use crate::schema::{IndexRecordOption, TextFieldIndexing, TextOptions};
     use crate::schema::{Schema, Term, INDEXED, STORED, STRING, TEXT};
     use crate::tokenizer::{
-        LowerCaser, SimpleTokenizer, StopWordFilter, Tokenizer, TokenizerManager,
+        LowerCaser, SimpleTokenizer, StopWordFilter, TextAnalyzer, TokenizerManager,
     };
     use crate::Index;
     use matches::assert_matches;
@@ -563,7 +563,7 @@ mod test {
         let tokenizer_manager = TokenizerManager::default();
         tokenizer_manager.register(
             "en_with_stop_words",
-            SimpleTokenizer
+            TextAnalyzer::from(SimpleTokenizer)
                 .filter(LowerCaser)
                 .filter(StopWordFilter::remove(vec!["the".to_string()])),
         );

src/query/regex_query.rs

@@ -54,6 +54,7 @@ use tantivy_fst::Regex;
 pub struct RegexQuery {
     regex: Arc<Regex>,
     field: Field,
+    boost: f32,
 }
 
 impl RegexQuery {
@@ -69,11 +70,17 @@ impl RegexQuery {
         RegexQuery {
             regex: regex.into(),
             field,
+            boost: 1.0,
         }
     }
 
+    /// Boost the query score by the given factor.
+    pub fn boost_by(self, boost: f32) -> Self {
+        Self { boost, ..self }
+    }
+
     fn specialized_weight(&self) -> AutomatonWeight<Regex> {
-        AutomatonWeight::new(self.field, self.regex.clone())
+        AutomatonWeight::new(self.field, self.regex.clone()).boost_by(self.boost)
     }
 }
 

src/query/scorer.rs

@@ -56,6 +56,11 @@ impl<TDocSet: DocSet> ConstScorer<TDocSet> {
         }
     }
 
+    /// Creates a new `ConstScorer` with a custom score value
+    pub fn with_score(docset: TDocSet, score: f32) -> ConstScorer<TDocSet> {
+        ConstScorer { docset, score }
+    }
+
     /// Sets the constant score to a different value.
     pub fn set_score(&mut self, score: Score) {
         self.score = score;

src/query/term_query/mod.rs

@@ -45,6 +45,35 @@ mod tests {
         assert_eq!(term_scorer.score(), 0.28768212);
     }
 
+    #[test]
+    pub fn test_term_query_boost_by() {
+        let mut schema_builder = Schema::builder();
+        let text_field = schema_builder.add_text_field("text", STRING);
+        let schema = schema_builder.build();
+        let index = Index::create_in_ram(schema);
+        {
+            // 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);
+            }
+            assert!(index_writer.commit().is_ok());
+        }
+        let searcher = index.reader().unwrap().searcher();
+        let term_query = TermQuery::new(
+            Term::from_field_text(text_field, "a"),
+            IndexRecordOption::Basic,
+        )
+        .boost_by(42.0);
+        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).unwrap();
+        assert!(term_scorer.advance());
+        assert_eq!(term_scorer.doc(), 0);
+        assert_nearly_equals(0.28768212 * 42.0, term_scorer.score());
+    }
+
     #[test]
     pub fn test_term_weight() {
         let mut schema_builder = Schema::builder();
@@ -112,7 +141,6 @@ mod tests {
         let term_a = Term::from_field_text(text_field, "a");
         let term_query = TermQuery::new(term_a, IndexRecordOption::Basic);
         let reader = index.reader().unwrap();
-        assert_eq!(reader.searcher().segment_readers().len(), 1);
         assert_eq!(term_query.count(&*reader.searcher()).unwrap(), 1);
     }
 

src/query/term_query/term_query.rs

@@ -61,6 +61,7 @@ use std::fmt;
 pub struct TermQuery {
     term: Term,
     index_record_option: IndexRecordOption,
+    boost: f32,
 }
 
 impl fmt::Debug for TermQuery {
@@ -75,9 +76,15 @@ impl TermQuery {
         TermQuery {
             term,
             index_record_option: segment_postings_options,
+            boost: 1.0,
         }
     }
 
+    /// Boost the query score by the given factor.
+    pub fn boost_by(self, boost: f32) -> Self {
+        Self { boost, ..self }
+    }
+
     /// The `Term` this query is built out of.
     pub fn term(&self) -> &Term {
         &self.term
@@ -90,7 +97,7 @@ impl TermQuery {
     /// This is useful for optimization purpose.
     pub fn specialized_weight(&self, searcher: &Searcher, scoring_enabled: bool) -> TermWeight {
         let term = self.term.clone();
-        let bm25_weight = BM25Weight::for_terms(searcher, &[term]);
+        let bm25_weight = BM25Weight::for_terms(searcher, &[term], self.boost);
         let index_record_option = if scoring_enabled {
             self.index_record_option
         } else {

src/reader/index_writer_reader.rs

@@ -1,83 +0,0 @@
-use crate::directory::{WatchCallbackList, WatchHandle};
-use crate::indexer::SegmentRegisters;
-use crate::reader::pool::Pool;
-use crate::{Index, LeasedItem, Searcher, Segment, SegmentReader};
-use std::iter::repeat_with;
-use std::sync::{Arc, RwLock, Weak};
-
-struct InnerNRTReader {
-    num_searchers: usize,
-    index: Index,
-    searcher_pool: Pool<Searcher>,
-    segment_registers: Arc<RwLock<SegmentRegisters>>,
-}
-
-impl InnerNRTReader {
-    fn load_segment_readers(&self) -> crate::Result<Vec<SegmentReader>> {
-        let segments: Vec<Segment> = {
-            let rlock = self.segment_registers.read().unwrap();
-            rlock.committed_segment()
-        };
-        segments
-            .iter()
-            .map(SegmentReader::open)
-            .collect::<crate::Result<Vec<SegmentReader>>>()
-    }
-
-    pub fn reload(&self) -> crate::Result<()> {
-        let segment_readers: Vec<SegmentReader> = self.load_segment_readers()?;
-        let schema = self.index.schema();
-        let searchers = repeat_with(|| {
-            Searcher::new(schema.clone(), self.index.clone(), segment_readers.clone())
-        })
-        .take(self.num_searchers)
-        .collect();
-        self.searcher_pool.publish_new_generation(searchers);
-        Ok(())
-    }
-
-    pub fn searcher(&self) -> LeasedItem<Searcher> {
-        self.searcher_pool.acquire()
-    }
-}
-
-#[derive(Clone)]
-pub struct NRTReader {
-    inner: Arc<InnerNRTReader>,
-    watch_handle: WatchHandle,
-}
-
-impl NRTReader {
-    pub fn reload(&self) -> crate::Result<()> {
-        self.inner.reload()
-    }
-
-    pub fn searcher(&self) -> LeasedItem<Searcher> {
-        self.inner.searcher()
-    }
-
-    pub(crate) fn create(
-        num_searchers: usize,
-        index: Index,
-        segment_registers: Arc<RwLock<SegmentRegisters>>,
-        watch_callback_list: &WatchCallbackList,
-    ) -> crate::Result<Self> {
-        let inner_reader: Arc<InnerNRTReader> = Arc::new(InnerNRTReader {
-            num_searchers,
-            index,
-            searcher_pool: Pool::new(),
-            segment_registers,
-        });
-        let inner_reader_weak: Weak<InnerNRTReader> = Arc::downgrade(&inner_reader);
-        let watch_handle = watch_callback_list.subscribe(Box::new(move || {
-            if let Some(nrt_reader_arc) = inner_reader_weak.upgrade() {
-                let _ = nrt_reader_arc.reload();
-            }
-        }));
-        inner_reader.reload()?;
-        Ok(NRTReader {
-            inner: inner_reader,
-            watch_handle,
-        })
-    }
-}

src/reader/meta_file_reader.rs

@@ -1,180 +0,0 @@
-use super::pool::Pool;
-use crate::core::Segment;
-use crate::directory::Directory;
-use crate::directory::WatchHandle;
-use crate::directory::META_LOCK;
-use crate::Searcher;
-use crate::SegmentReader;
-use crate::{Index, LeasedItem};
-use crate::{IndexReader, Result};
-use std::iter::repeat_with;
-use std::sync::Arc;
-
-/// Defines when a new version of the index should be reloaded.
-///
-/// Regardless of whether you search and index in the same process, tantivy does not necessarily
-/// reflects the change that are commited to your index. `ReloadPolicy` precisely helps you define
-/// when you want your index to be reloaded.
-#[derive(Clone, Copy)]
-pub enum ReloadPolicy {
-    /// The index is entirely reloaded manually.
-    /// All updates of the index should be manual.
-    ///
-    /// No change is reflected automatically. You are required to call `.load_seacher()` manually.
-    Manual,
-    /// The index is reloaded within milliseconds after a new commit is available.
-    /// This is made possible by watching changes in the `meta.json` file.
-    OnCommit, // TODO add NEAR_REAL_TIME(target_ms)
-}
-
-/// `IndexReader` builder
-///
-/// It makes it possible to set the following values.
-///
-/// - `num_searchers` (by default, the number of detected CPU threads):
-///
-///   When `num_searchers` queries are requested at the same time, the `num_searchers` will block
-///   until the one of the searcher in-use gets released.
-/// - `reload_policy` (by default `ReloadPolicy::OnCommit`):
-///
-///   See [`ReloadPolicy`](./enum.ReloadPolicy.html) for more details.
-#[derive(Clone)]
-pub struct IndexReaderBuilder {
-    num_searchers: usize,
-    reload_policy: ReloadPolicy,
-    index: Index,
-}
-
-impl IndexReaderBuilder {
-    pub(crate) fn new(index: Index) -> IndexReaderBuilder {
-        IndexReaderBuilder {
-            num_searchers: num_cpus::get(),
-            reload_policy: ReloadPolicy::Manual,
-            index,
-        }
-    }
-
-    /// Builds the reader.
-    ///
-    /// Building the reader is a non-trivial operation that requires
-    /// to open different segment readers. It may take hundreds of milliseconds
-    /// of time and it may return an error.
-    /// TODO(pmasurel) Use the `TryInto` trait once it is available in stable.
-    pub fn try_into(self) -> Result<IndexReader> {
-        let inner_reader = MetaFileIndexReaderInner {
-            index: self.index,
-            num_searchers: self.num_searchers,
-            searcher_pool: Pool::new(),
-        };
-        inner_reader.reload()?;
-        let inner_reader_arc = Arc::new(inner_reader);
-        let watch_handle_opt: Option<WatchHandle>;
-        match self.reload_policy {
-            ReloadPolicy::Manual => {
-                // No need to set anything...
-                watch_handle_opt = None;
-            }
-            ReloadPolicy::OnCommit => {
-                let inner_reader_arc_clone = inner_reader_arc.clone();
-                let callback = move || {
-                    if let Err(err) = inner_reader_arc_clone.reload() {
-                        error!(
-                            "Error while loading searcher after commit was detected. {:?}",
-                            err
-                        );
-                    }
-                };
-                let watch_handle = inner_reader_arc
-                    .index
-                    .directory()
-                    .watch(Box::new(callback))?;
-                watch_handle_opt = Some(watch_handle);
-            }
-        }
-        Ok(IndexReader::from(MetaFileIndexReader {
-            inner: inner_reader_arc,
-            watch_handle_opt,
-        }))
-    }
-
-    /// Sets the reload_policy.
-    ///
-    /// See [`ReloadPolicy`](./enum.ReloadPolicy.html) for more details.
-    pub fn reload_policy(mut self, reload_policy: ReloadPolicy) -> IndexReaderBuilder {
-        self.reload_policy = reload_policy;
-        self
-    }
-
-    /// Sets the number of `Searcher` in the searcher pool.
-    pub fn num_searchers(mut self, num_searchers: usize) -> IndexReaderBuilder {
-        self.num_searchers = num_searchers;
-        self
-    }
-}
-
-struct MetaFileIndexReaderInner {
-    num_searchers: usize,
-    searcher_pool: Pool<Searcher>,
-    index: Index,
-}
-
-impl MetaFileIndexReaderInner {
-    fn load_segment_readers(&self) -> crate::Result<Vec<SegmentReader>> {
-        // We keep the lock until we have effectively finished opening the
-        // the `SegmentReader` because it prevents a diffferent process
-        // to garbage collect these file while we open them.
-        //
-        // Once opened, on linux & mac, the mmap will remain valid after
-        // the file has been deleted
-        // On windows, the file deletion will fail.
-        let _meta_lock = self.index.directory().acquire_lock(&META_LOCK)?;
-        let searchable_segments = self.searchable_segments()?;
-        searchable_segments
-            .iter()
-            .map(SegmentReader::open)
-            .collect::<Result<_>>()
-    }
-
-    fn reload(&self) -> crate::Result<()> {
-        let segment_readers: Vec<SegmentReader> = self.load_segment_readers()?;
-        let schema = self.index.schema();
-        let searchers = repeat_with(|| {
-            Searcher::new(schema.clone(), self.index.clone(), segment_readers.clone())
-        })
-        .take(self.num_searchers)
-        .collect();
-        self.searcher_pool.publish_new_generation(searchers);
-        Ok(())
-    }
-
-    /// Returns the list of segments that are searchable
-    fn searchable_segments(&self) -> crate::Result<Vec<Segment>> {
-        self.index.searchable_segments()
-    }
-
-    fn searcher(&self) -> LeasedItem<Searcher> {
-        self.searcher_pool.acquire()
-    }
-}
-
-/// `IndexReader` is your entry point to read and search the index.
-///
-/// It controls when a new version of the index should be loaded and lends
-/// you instances of `Searcher` for the last loaded version.
-///
-/// `Clone` does not clone the different pool of searcher. `IndexReader`
-/// just wraps and `Arc`.
-#[derive(Clone)]
-pub struct MetaFileIndexReader {
-    inner: Arc<MetaFileIndexReaderInner>,
-    watch_handle_opt: Option<WatchHandle>,
-}
-
-impl MetaFileIndexReader {
-    pub fn reload(&self) -> crate::Result<()> {
-        self.inner.reload()
-    }
-    pub fn searcher(&self) -> LeasedItem<Searcher> {
-        self.inner.searcher()
-    }
-}

src/reader/mod.rs

@@ -1,20 +1,172 @@
-mod index_writer_reader;
-mod meta_file_reader;
 mod pool;
 
-use self::meta_file_reader::MetaFileIndexReader;
-pub use self::meta_file_reader::{IndexReaderBuilder, ReloadPolicy};
 pub use self::pool::LeasedItem;
-pub(crate) use crate::reader::index_writer_reader::{NRTReader};
+use self::pool::Pool;
+use crate::core::Segment;
+use crate::directory::Directory;
+use crate::directory::WatchHandle;
+use crate::directory::META_LOCK;
+use crate::Index;
+use crate::Result;
 use crate::Searcher;
+use crate::SegmentReader;
+use std::sync::Arc;
 
+/// Defines when a new version of the index should be reloaded.
+///
+/// Regardless of whether you search and index in the same process, tantivy does not necessarily
+/// reflects the change that are commited to your index. `ReloadPolicy` precisely helps you define
+/// when you want your index to be reloaded.
+#[derive(Clone, Copy)]
+pub enum ReloadPolicy {
+    /// The index is entirely reloaded manually.
+    /// All updates of the index should be manual.
+    ///
+    /// No change is reflected automatically. You are required to call `.load_seacher()` manually.
+    Manual,
+    /// The index is reloaded within milliseconds after a new commit is available.
+    /// This is made possible by watching changes in the `meta.json` file.
+    OnCommit, // TODO add NEAR_REAL_TIME(target_ms)
+}
+
+/// `IndexReader` builder
+///
+/// It makes it possible to set the following values.
+///
+/// - `num_searchers` (by default, the number of detected CPU threads):
+///
+///   When `num_searchers` queries are requested at the same time, the `num_searchers` will block
+///   until the one of the searcher in-use gets released.
+/// - `reload_policy` (by default `ReloadPolicy::OnCommit`):
+///
+///   See [`ReloadPolicy`](./enum.ReloadPolicy.html) for more details.
 #[derive(Clone)]
-pub enum IndexReader {
+pub struct IndexReaderBuilder {
-    FromMetaFile(MetaFileIndexReader),
+    num_searchers: usize,
-    NRT(NRTReader),
+    reload_policy: ReloadPolicy,
+    index: Index,
+}
+
+impl IndexReaderBuilder {
+    pub(crate) fn new(index: Index) -> IndexReaderBuilder {
+        IndexReaderBuilder {
+            num_searchers: num_cpus::get(),
+            reload_policy: ReloadPolicy::OnCommit,
+            index,
+        }
+    }
+
+    /// Builds the reader.
+    ///
+    /// Building the reader is a non-trivial operation that requires
+    /// to open different segment readers. It may take hundreds of milliseconds
+    /// of time and it may return an error.
+    /// TODO(pmasurel) Use the `TryInto` trait once it is available in stable.
+    pub fn try_into(self) -> Result<IndexReader> {
+        let inner_reader = InnerIndexReader {
+            index: self.index,
+            num_searchers: self.num_searchers,
+            searcher_pool: Pool::new(),
+        };
+        inner_reader.reload()?;
+        let inner_reader_arc = Arc::new(inner_reader);
+        let watch_handle_opt: Option<WatchHandle>;
+        match self.reload_policy {
+            ReloadPolicy::Manual => {
+                // No need to set anything...
+                watch_handle_opt = None;
+            }
+            ReloadPolicy::OnCommit => {
+                let inner_reader_arc_clone = inner_reader_arc.clone();
+                let callback = move || {
+                    if let Err(err) = inner_reader_arc_clone.reload() {
+                        error!(
+                            "Error while loading searcher after commit was detected. {:?}",
+                            err
+                        );
+                    }
+                };
+                let watch_handle = inner_reader_arc
+                    .index
+                    .directory()
+                    .watch(Box::new(callback))?;
+                watch_handle_opt = Some(watch_handle);
+            }
+        }
+        Ok(IndexReader {
+            inner: inner_reader_arc,
+            watch_handle_opt,
+        })
+    }
+
+    /// Sets the reload_policy.
+    ///
+    /// See [`ReloadPolicy`](./enum.ReloadPolicy.html) for more details.
+    pub fn reload_policy(mut self, reload_policy: ReloadPolicy) -> IndexReaderBuilder {
+        self.reload_policy = reload_policy;
+        self
+    }
+
+    /// Sets the number of `Searcher` in the searcher pool.
+    pub fn num_searchers(mut self, num_searchers: usize) -> IndexReaderBuilder {
+        self.num_searchers = num_searchers;
+        self
+    }
+}
+
+struct InnerIndexReader {
+    num_searchers: usize,
+    searcher_pool: Pool<Searcher>,
+    index: Index,
+}
+
+impl InnerIndexReader {
+    fn reload(&self) -> Result<()> {
+        let segment_readers: Vec<SegmentReader> = {
+            let _meta_lock = self.index.directory().acquire_lock(&META_LOCK)?;
+            let searchable_segments = self.searchable_segments()?;
+            searchable_segments
+                .iter()
+                .map(SegmentReader::open)
+                .collect::<Result<_>>()?
+        };
+        let schema = self.index.schema();
+        let searchers = (0..self.num_searchers)
+            .map(|_| Searcher::new(schema.clone(), self.index.clone(), segment_readers.clone()))
+            .collect();
+        self.searcher_pool.publish_new_generation(searchers);
+        Ok(())
+    }
+
+    /// Returns the list of segments that are searchable
+    fn searchable_segments(&self) -> Result<Vec<Segment>> {
+        self.index.searchable_segments()
+    }
+
+    fn searcher(&self) -> LeasedItem<Searcher> {
+        self.searcher_pool.acquire()
+    }
+}
+
+/// `IndexReader` is your entry point to read and search the index.
+///
+/// It controls when a new version of the index should be loaded and lends
+/// you instances of `Searcher` for the last loaded version.
+///
+/// `Clone` does not clone the different pool of searcher. `IndexReader`
+/// just wraps and `Arc`.
+#[derive(Clone)]
+pub struct IndexReader {
+    inner: Arc<InnerIndexReader>,
+    watch_handle_opt: Option<WatchHandle>,
 }
 
 impl IndexReader {
+    #[cfg(test)]
+    pub(crate) fn index(&self) -> Index {
+        self.inner.index.clone()
+    }
+
     /// Update searchers so that they reflect the state of the last
     /// `.commit()`.
     ///
@@ -24,11 +176,8 @@ impl IndexReader {
     ///
     /// This automatic reload can take 10s of milliseconds to kick in however, and in unit tests
     /// it can be nice to deterministically force the reload of searchers.
-    pub fn reload(&self) -> crate::Result<()> {
+    pub fn reload(&self) -> Result<()> {
-        match self {
+        self.inner.reload()
-            IndexReader::FromMetaFile(meta_file_reader) => meta_file_reader.reload(),
-            IndexReader::NRT(nrt_reader) => nrt_reader.reload(),
-        }
     }
 
     /// Returns a searcher
@@ -42,21 +191,6 @@ impl IndexReader {
     /// The same searcher must be used for a given query, as it ensures
     /// the use of a consistent segment set.
     pub fn searcher(&self) -> LeasedItem<Searcher> {
-        match self {
+        self.inner.searcher()
-            IndexReader::FromMetaFile(meta_file_reader) => meta_file_reader.searcher(),
-            IndexReader::NRT(nrt_reader) => nrt_reader.searcher(),
-        }
-    }
-}
-
-impl From<MetaFileIndexReader> for IndexReader {
-    fn from(meta_file_reader: MetaFileIndexReader) -> Self {
-        IndexReader::FromMetaFile(meta_file_reader)
-    }
-}
-
-impl From<NRTReader> for IndexReader {
-    fn from(nrt_reader: NRTReader) -> Self {
-        IndexReader::NRT(nrt_reader)
     }
 }

src/schema/facet.rs

@@ -122,6 +122,11 @@ impl Facet {
     pub fn to_path(&self) -> Vec<&str> {
         self.encoded_str().split(|c| c == FACET_SEP_CHAR).collect()
     }
+
+    /// This function is the inverse of Facet::from(&str).
+    pub fn to_path_string(&self) -> String {
+        format!("{}", self.to_string())
+    }
 }
 
 impl Borrow<str> for Facet {
@@ -265,4 +270,21 @@ mod tests {
         let facet = Facet::from_path(v.iter());
         assert_eq!(facet.to_path(), v);
     }
+
+    #[test]
+    fn test_to_path_string() {
+        let v = ["first", "second", "third/not_fourth"];
+        let facet = Facet::from_path(v.iter());
+        assert_eq!(
+            facet.to_path_string(),
+            String::from("/first/second/third\\/not_fourth")
+        );
+    }
+
+    #[test]
+    fn test_to_path_string_empty() {
+        let v: Vec<&str> = vec![];
+        let facet = Facet::from_path(v.iter());
+        assert_eq!(facet.to_path_string(), "/");
+    }
 }

src/snippet/mod.rs

@@ -1,8 +1,7 @@
 use crate::query::Query;
 use crate::schema::Field;
 use crate::schema::Value;
-use crate::tokenizer::BoxedTokenizer;
+use crate::tokenizer::{TextAnalyzer, Token};
-use crate::tokenizer::{Token, TokenStream};
 use crate::Document;
 use crate::Result;
 use crate::Searcher;
@@ -142,7 +141,7 @@ impl Snippet {
 /// Fragments must be valid in the sense that `&text[fragment.start..fragment.stop]`\
 /// has to be a valid string.
 fn search_fragments<'a>(
-    tokenizer: &BoxedTokenizer,
+    tokenizer: &TextAnalyzer,
     text: &'a str,
     terms: &BTreeMap<String, f32>,
     max_num_chars: usize,
@@ -251,7 +250,7 @@ fn select_best_fragment_combination(fragments: &[FragmentCandidate], text: &str)
 /// ```
 pub struct SnippetGenerator {
     terms_text: BTreeMap<String, f32>,
-    tokenizer: BoxedTokenizer,
+    tokenizer: TextAnalyzer,
     field: Field,
     max_num_chars: usize,
 }
@@ -347,12 +346,11 @@ Survey in 2016, 2017, and 2018."#;
 
     #[test]
     fn test_snippet() {
-        let boxed_tokenizer = SimpleTokenizer.into();
         let terms = btreemap! {
             String::from("rust") => 1.0,
             String::from("language") => 0.9
         };
-        let fragments = search_fragments(&boxed_tokenizer, TEST_TEXT, &terms, 100);
+        let fragments = search_fragments(&From::from(SimpleTokenizer), TEST_TEXT, &terms, 100);
         assert_eq!(fragments.len(), 7);
         {
             let first = &fragments[0];
@@ -374,13 +372,12 @@ Survey in 2016, 2017, and 2018."#;
 
     #[test]
     fn test_snippet_scored_fragment() {
-        let boxed_tokenizer = SimpleTokenizer.into();
         {
             let terms = btreemap! {
                 String::from("rust") =>1.0f32,
                 String::from("language") => 0.9f32
             };
-            let fragments = search_fragments(&boxed_tokenizer, TEST_TEXT, &terms, 20);
+            let fragments = search_fragments(&From::from(SimpleTokenizer), TEST_TEXT, &terms, 20);
             {
                 let first = &fragments[0];
                 assert_eq!(first.score, 1.0);
@@ -389,13 +386,12 @@ Survey in 2016, 2017, and 2018."#;
             let snippet = select_best_fragment_combination(&fragments[..], &TEST_TEXT);
             assert_eq!(snippet.to_html(), "<b>Rust</b> is a systems")
         }
-        let boxed_tokenizer = SimpleTokenizer.into();
         {
             let terms = btreemap! {
                 String::from("rust") =>0.9f32,
                 String::from("language") => 1.0f32
             };
-            let fragments = search_fragments(&boxed_tokenizer, TEST_TEXT, &terms, 20);
+            let fragments = search_fragments(&From::from(SimpleTokenizer), TEST_TEXT, &terms, 20);
             //assert_eq!(fragments.len(), 7);
             {
                 let first = &fragments[0];
@@ -409,14 +405,12 @@ Survey in 2016, 2017, and 2018."#;
 
     #[test]
     fn test_snippet_in_second_fragment() {
-        let boxed_tokenizer = SimpleTokenizer.into();
-
         let text = "a b c d e f g";
 
         let mut terms = BTreeMap::new();
         terms.insert(String::from("c"), 1.0);
 
-        let fragments = search_fragments(&boxed_tokenizer, &text, &terms, 3);
+        let fragments = search_fragments(&From::from(SimpleTokenizer), &text, &terms, 3);
 
         assert_eq!(fragments.len(), 1);
         {
@@ -433,14 +427,12 @@ Survey in 2016, 2017, and 2018."#;
 
     #[test]
     fn test_snippet_with_term_at_the_end_of_fragment() {
-        let boxed_tokenizer = SimpleTokenizer.into();
-
         let text = "a b c d e f f g";
 
         let mut terms = BTreeMap::new();
         terms.insert(String::from("f"), 1.0);
 
-        let fragments = search_fragments(&boxed_tokenizer, &text, &terms, 3);
+        let fragments = search_fragments(&From::from(SimpleTokenizer), &text, &terms, 3);
 
         assert_eq!(fragments.len(), 2);
         {
@@ -457,15 +449,13 @@ Survey in 2016, 2017, and 2018."#;
 
     #[test]
     fn test_snippet_with_second_fragment_has_the_highest_score() {
-        let boxed_tokenizer = SimpleTokenizer.into();
-
         let text = "a b c d e f g";
 
         let mut terms = BTreeMap::new();
         terms.insert(String::from("f"), 1.0);
         terms.insert(String::from("a"), 0.9);
 
-        let fragments = search_fragments(&boxed_tokenizer, &text, &terms, 7);
+        let fragments = search_fragments(&From::from(SimpleTokenizer), &text, &terms, 7);
 
         assert_eq!(fragments.len(), 2);
         {
@@ -482,14 +472,12 @@ Survey in 2016, 2017, and 2018."#;
 
     #[test]
     fn test_snippet_with_term_not_in_text() {
-        let boxed_tokenizer = SimpleTokenizer.into();
-
         let text = "a b c d";
 
         let mut terms = BTreeMap::new();
         terms.insert(String::from("z"), 1.0);
 
-        let fragments = search_fragments(&boxed_tokenizer, &text, &terms, 3);
+        let fragments = search_fragments(&From::from(SimpleTokenizer), &text, &terms, 3);
 
         assert_eq!(fragments.len(), 0);
 
@@ -500,12 +488,10 @@ Survey in 2016, 2017, and 2018."#;
 
     #[test]
     fn test_snippet_with_no_terms() {
-        let boxed_tokenizer = SimpleTokenizer.into();
-
         let text = "a b c d";
 
         let terms = BTreeMap::new();
-        let fragments = search_fragments(&boxed_tokenizer, &text, &terms, 3);
+        let fragments = search_fragments(&From::from(SimpleTokenizer), &text, &terms, 3);
         assert_eq!(fragments.len(), 0);
 
         let snippet = select_best_fragment_combination(&fragments[..], &text);

src/termdict/mod.rs

@@ -38,7 +38,7 @@ mod tests {
     use crate::core::Index;
     use crate::directory::{Directory, RAMDirectory, ReadOnlySource};
     use crate::postings::TermInfo;
-    use crate::schema::{Document, FieldType, Schema, TEXT};
+    use crate::schema::{Document, Schema, TEXT};
     use std::path::PathBuf;
     use std::str;
 
@@ -52,6 +52,12 @@ mod tests {
         }
     }
 
+    #[test]
+    fn test_empty_term_dictionary() {
+        let empty = TermDictionary::empty();
+        assert!(empty.stream().next().is_none());
+    }
+
     #[test]
     fn test_term_ordinals() {
         const COUNTRIES: [&'static str; 7] = [
@@ -67,9 +73,7 @@ mod tests {
         let path = PathBuf::from("TermDictionary");
         {
             let write = directory.open_write(&path).unwrap();
-            let field_type = FieldType::Str(TEXT);
+            let mut term_dictionary_builder = TermDictionaryBuilder::create(write).unwrap();
-            let mut term_dictionary_builder =
-                TermDictionaryBuilder::create(write, &field_type).unwrap();
             for term in COUNTRIES.iter() {
                 term_dictionary_builder
                     .insert(term.as_bytes(), &make_term_info(0u64))
@@ -93,9 +97,7 @@ mod tests {
         let path = PathBuf::from("TermDictionary");
         {
             let write = directory.open_write(&path).unwrap();
-            let field_type = FieldType::Str(TEXT);
+            let mut term_dictionary_builder = TermDictionaryBuilder::create(write).unwrap();
-            let mut term_dictionary_builder =
-                TermDictionaryBuilder::create(write, &field_type).unwrap();
             term_dictionary_builder
                 .insert("abc".as_bytes(), &make_term_info(34u64))
                 .unwrap();
@@ -179,10 +181,8 @@ mod tests {
         let ids: Vec<_> = (0u32..10_000u32)
             .map(|i| (format!("doc{:0>6}", i), i))
             .collect();
-        let field_type = FieldType::Str(TEXT);
         let buffer: Vec<u8> = {
-            let mut term_dictionary_builder =
+            let mut term_dictionary_builder = TermDictionaryBuilder::create(vec![]).unwrap();
-                TermDictionaryBuilder::create(vec![], &field_type).unwrap();
             for &(ref id, ref i) in &ids {
                 term_dictionary_builder
                     .insert(id.as_bytes(), &make_term_info(*i as u64))
@@ -209,10 +209,8 @@ mod tests {
 
     #[test]
     fn test_stream_high_range_prefix_suffix() {
-        let field_type = FieldType::Str(TEXT);
         let buffer: Vec<u8> = {
-            let mut term_dictionary_builder =
+            let mut term_dictionary_builder = TermDictionaryBuilder::create(vec![]).unwrap();
-                TermDictionaryBuilder::create(vec![], &field_type).unwrap();
             // term requires more than 16bits
             term_dictionary_builder
                 .insert("abcdefghijklmnopqrstuvwxy", &make_term_info(1))
@@ -244,10 +242,8 @@ mod tests {
         let ids: Vec<_> = (0u32..10_000u32)
             .map(|i| (format!("doc{:0>6}", i), i))
             .collect();
-        let field_type = FieldType::Str(TEXT);
         let buffer: Vec<u8> = {
-            let mut term_dictionary_builder =
+            let mut term_dictionary_builder = TermDictionaryBuilder::create(vec![]).unwrap();
-                TermDictionaryBuilder::create(vec![], &field_type).unwrap();
             for &(ref id, ref i) in &ids {
                 term_dictionary_builder
                     .insert(id.as_bytes(), &make_term_info(*i as u64))
@@ -313,10 +309,8 @@ mod tests {
 
     #[test]
     fn test_empty_string() {
-        let field_type = FieldType::Str(TEXT);
         let buffer: Vec<u8> = {
-            let mut term_dictionary_builder =
+            let mut term_dictionary_builder = TermDictionaryBuilder::create(vec![]).unwrap();
-                TermDictionaryBuilder::create(vec![], &field_type).unwrap();
             term_dictionary_builder
                 .insert(&[], &make_term_info(1 as u64))
                 .unwrap();
@@ -337,10 +331,8 @@ mod tests {
 
     #[test]
     fn test_stream_range_boundaries() {
-        let field_type = FieldType::Str(TEXT);
         let buffer: Vec<u8> = {
-            let mut term_dictionary_builder =
+            let mut term_dictionary_builder = TermDictionaryBuilder::create(vec![]).unwrap();
-                TermDictionaryBuilder::create(vec![], &field_type).unwrap();
             for i in 0u8..10u8 {
                 let number_arr = [i; 1];
                 term_dictionary_builder
@@ -352,41 +344,91 @@ mod tests {
         let source = ReadOnlySource::from(buffer);
         let term_dictionary: TermDictionary = TermDictionary::from_source(&source);
 
-        let value_list = |mut streamer: TermStreamer<'_>| {
+        let value_list = |mut streamer: TermStreamer<'_>, backwards: bool| {
             let mut res: Vec<u32> = vec![];
             while let Some((_, ref v)) = streamer.next() {
                 res.push(v.doc_freq);
             }
+            if backwards {
+                res.reverse();
+            }
             res
         };
+        {
+            let range = term_dictionary.range().backward().into_stream();
+            assert_eq!(
+                value_list(range, true),
+                vec![0u32, 1u32, 2u32, 3u32, 4u32, 5u32, 6u32, 7u32, 8u32, 9u32]
+            );
+        }
         {
             let range = term_dictionary.range().ge([2u8]).into_stream();
             assert_eq!(
-                value_list(range),
+                value_list(range, false),
+                vec![2u32, 3u32, 4u32, 5u32, 6u32, 7u32, 8u32, 9u32]
+            );
+        }
+        {
+            let range = term_dictionary.range().ge([2u8]).backward().into_stream();
+            assert_eq!(
+                value_list(range, true),
                 vec![2u32, 3u32, 4u32, 5u32, 6u32, 7u32, 8u32, 9u32]
             );
         }
         {
             let range = term_dictionary.range().gt([2u8]).into_stream();
             assert_eq!(
-                value_list(range),
+                value_list(range, false),
+                vec![3u32, 4u32, 5u32, 6u32, 7u32, 8u32, 9u32]
+            );
+        }
+        {
+            let range = term_dictionary.range().gt([2u8]).backward().into_stream();
+            assert_eq!(
+                value_list(range, true),
                 vec![3u32, 4u32, 5u32, 6u32, 7u32, 8u32, 9u32]
             );
         }
         {
             let range = term_dictionary.range().lt([6u8]).into_stream();
-            assert_eq!(value_list(range), vec![0u32, 1u32, 2u32, 3u32, 4u32, 5u32]);
+            assert_eq!(
+                value_list(range, false),
+                vec![0u32, 1u32, 2u32, 3u32, 4u32, 5u32]
+            );
+        }
+        {
+            let range = term_dictionary.range().lt([6u8]).backward().into_stream();
+            assert_eq!(
+                value_list(range, true),
+                vec![0u32, 1u32, 2u32, 3u32, 4u32, 5u32]
+            );
         }
         {
             let range = term_dictionary.range().le([6u8]).into_stream();
             assert_eq!(
-                value_list(range),
+                value_list(range, false),
+                vec![0u32, 1u32, 2u32, 3u32, 4u32, 5u32, 6u32]
+            );
+        }
+        {
+            let range = term_dictionary.range().le([6u8]).backward().into_stream();
+            assert_eq!(
+                value_list(range, true),
                 vec![0u32, 1u32, 2u32, 3u32, 4u32, 5u32, 6u32]
             );
         }
         {
             let range = term_dictionary.range().ge([0u8]).lt([5u8]).into_stream();
-            assert_eq!(value_list(range), vec![0u32, 1u32, 2u32, 3u32, 4u32]);
+            assert_eq!(value_list(range, false), vec![0u32, 1u32, 2u32, 3u32, 4u32]);
+        }
+        {
+            let range = term_dictionary
+                .range()
+                .ge([0u8])
+                .lt([5u8])
+                .backward()
+                .into_stream();
+            assert_eq!(value_list(range, true), vec![0u32, 1u32, 2u32, 3u32, 4u32]);
         }
     }
 
@@ -408,9 +450,7 @@ mod tests {
         let path = PathBuf::from("TermDictionary");
         {
             let write = directory.open_write(&path).unwrap();
-            let field_type = FieldType::Str(TEXT);
+            let mut term_dictionary_builder = TermDictionaryBuilder::create(write).unwrap();
-            let mut term_dictionary_builder =
-                TermDictionaryBuilder::create(write, &field_type).unwrap();
             for term in COUNTRIES.iter() {
                 term_dictionary_builder
                     .insert(term.as_bytes(), &make_term_info(0u64))

src/termdict/streamer.rs

@@ -51,6 +51,12 @@ where
         self
     }
 
+    /// Iterate over the range backwards.
+    pub fn backward(mut self) -> Self {
+        self.stream_builder = self.stream_builder.backward();
+        self
+    }
+
     /// Creates the stream corresponding to the range
     /// of terms defined using the `TermStreamerBuilder`.
     pub fn into_stream(self) -> TermStreamer<'a, A> {

src/termdict/termdict.rs

@@ -4,8 +4,8 @@ use crate::common::BinarySerializable;
 use crate::common::CountingWriter;
 use crate::directory::ReadOnlySource;
 use crate::postings::TermInfo;
-use crate::schema::FieldType;
 use crate::termdict::TermOrdinal;
+use once_cell::sync::Lazy;
 use std::io::{self, Write};
 use tantivy_fst;
 use tantivy_fst::raw::Fst;
@@ -29,7 +29,7 @@ where
     W: Write,
 {
     /// Creates a new `TermDictionaryBuilder`
-    pub fn create(w: W, _field_type: &FieldType) -> io::Result<Self> {
+    pub fn create(w: W) -> io::Result<Self> {
         let fst_builder = tantivy_fst::MapBuilder::new(w).map_err(convert_fst_error)?;
         Ok(TermDictionaryBuilder {
             fst_builder,
@@ -92,6 +92,14 @@ fn open_fst_index(source: ReadOnlySource) -> tantivy_fst::Map<ReadOnlySource> {
     tantivy_fst::Map::from(fst)
 }
 
+static EMPTY_DATA_SOURCE: Lazy<ReadOnlySource> = Lazy::new(|| {
+    let term_dictionary_data: Vec<u8> = TermDictionaryBuilder::create(Vec::<u8>::new())
+        .expect("Creating a TermDictionaryBuilder in a Vec<u8> should never fail")
+        .finish()
+        .expect("Writing in a Vec<u8> should never fail");
+    ReadOnlySource::from(term_dictionary_data)
+});
+
 /// The term dictionary contains all of the terms in
 /// `tantivy index` in a sorted manner.
 ///
@@ -122,14 +130,8 @@ impl TermDictionary {
     }
 
     /// Creates an empty term dictionary which contains no terms.
-    pub fn empty(field_type: &FieldType) -> Self {
+    pub fn empty() -> Self {
-        let term_dictionary_data: Vec<u8> =
+        TermDictionary::from_source(&*EMPTY_DATA_SOURCE)
-            TermDictionaryBuilder::create(Vec::<u8>::new(), &field_type)
-                .expect("Creating a TermDictionaryBuilder in a Vec<u8> should never fail")
-                .finish()
-                .expect("Writing in a Vec<u8> should never fail");
-        let source = ReadOnlySource::from(term_dictionary_data);
-        Self::from_source(&source)
     }
 
     /// Returns the number of terms in the dictionary.

src/tokenizer/alphanum_only.rs

@@ -2,7 +2,7 @@
 //! ```rust
 //! use tantivy::tokenizer::*;
 //!
-//! let tokenizer = RawTokenizer
+//! let tokenizer = TextAnalyzer::from(RawTokenizer)
 //!   .filter(AlphaNumOnlyFilter);
 //!
 //! let mut stream = tokenizer.token_stream("hello there");
@@ -10,7 +10,7 @@
 //! // contains a space
 //! assert!(stream.next().is_none());
 //!
-//! let tokenizer = SimpleTokenizer
+//! let tokenizer = TextAnalyzer::from(SimpleTokenizer)
 //!   .filter(AlphaNumOnlyFilter);
 //!
 //! let mut stream = tokenizer.token_stream("hello there 💣");
@@ -19,56 +19,30 @@
 //! // the "emoji" is dropped because its not an alphanum
 //! assert!(stream.next().is_none());
 //! ```
-use super::{Token, TokenFilter, TokenStream};
+use super::{BoxTokenStream, Token, TokenFilter, TokenStream};
 
 /// `TokenFilter` that removes all tokens that contain non
 /// ascii alphanumeric characters.
 #[derive(Clone)]
 pub struct AlphaNumOnlyFilter;
 
-pub struct AlphaNumOnlyFilterStream<TailTokenStream>
+pub struct AlphaNumOnlyFilterStream<'a> {
-where
+    tail: BoxTokenStream<'a>,
-    TailTokenStream: TokenStream,
-{
-    tail: TailTokenStream,
 }
 
-impl<TailTokenStream> AlphaNumOnlyFilterStream<TailTokenStream>
+impl<'a> AlphaNumOnlyFilterStream<'a> {
-where
-    TailTokenStream: TokenStream,
-{
     fn predicate(&self, token: &Token) -> bool {
         token.text.chars().all(|c| c.is_ascii_alphanumeric())
     }
+}
 
-    fn wrap(tail: TailTokenStream) -> AlphaNumOnlyFilterStream<TailTokenStream> {
+impl TokenFilter for AlphaNumOnlyFilter {
-        AlphaNumOnlyFilterStream { tail }
+    fn transform<'a>(&self, token_stream: BoxTokenStream<'a>) -> BoxTokenStream<'a> {
+        BoxTokenStream::from(AlphaNumOnlyFilterStream { tail: token_stream })
     }
 }
 
-impl<TailTokenStream> TokenFilter<TailTokenStream> for AlphaNumOnlyFilter
+impl<'a> TokenStream for AlphaNumOnlyFilterStream<'a> {
-where
-    TailTokenStream: TokenStream,
-{
-    type ResultTokenStream = AlphaNumOnlyFilterStream<TailTokenStream>;
-
-    fn transform(&self, token_stream: TailTokenStream) -> Self::ResultTokenStream {
-        AlphaNumOnlyFilterStream::wrap(token_stream)
-    }
-}
-
-impl<TailTokenStream> TokenStream for AlphaNumOnlyFilterStream<TailTokenStream>
-where
-    TailTokenStream: TokenStream,
-{
-    fn token(&self) -> &Token {
-        self.tail.token()
-    }
-
-    fn token_mut(&mut self) -> &mut Token {
-        self.tail.token_mut()
-    }
-
     fn advance(&mut self) -> bool {
         while self.tail.advance() {
             if self.predicate(self.tail.token()) {
@@ -78,4 +52,12 @@ where
 
         false
     }
+
+    fn token(&self) -> &Token {
+        self.tail.token()
+    }
+
+    fn token_mut(&mut self) -> &mut Token {
+        self.tail.token_mut()
+    }
 }

src/tokenizer/ascii_folding_filter.rs

@@ -1,4 +1,4 @@
-use super::{Token, TokenFilter, TokenStream};
+use super::{BoxTokenStream, Token, TokenFilter, TokenStream};
 use std::mem;
 
 /// This class converts alphabetic, numeric, and symbolic Unicode characters
@@ -7,26 +7,21 @@ use std::mem;
 #[derive(Clone)]
 pub struct AsciiFoldingFilter;
 
-impl<TailTokenStream> TokenFilter<TailTokenStream> for AsciiFoldingFilter
+impl TokenFilter for AsciiFoldingFilter {
-where
+    fn transform<'a>(&self, token_stream: BoxTokenStream<'a>) -> BoxTokenStream<'a> {
-    TailTokenStream: TokenStream,
+        From::from(AsciiFoldingFilterTokenStream {
-{
+            tail: token_stream,
-    type ResultTokenStream = AsciiFoldingFilterTokenStream<TailTokenStream>;
+            buffer: String::with_capacity(100),
-
+        })
-    fn transform(&self, token_stream: TailTokenStream) -> Self::ResultTokenStream {
-        AsciiFoldingFilterTokenStream::wrap(token_stream)
     }
 }
 
-pub struct AsciiFoldingFilterTokenStream<TailTokenStream> {
+pub struct AsciiFoldingFilterTokenStream<'a> {
     buffer: String,
-    tail: TailTokenStream,
+    tail: BoxTokenStream<'a>,
 }
 
-impl<TailTokenStream> TokenStream for AsciiFoldingFilterTokenStream<TailTokenStream>
+impl<'a> TokenStream for AsciiFoldingFilterTokenStream<'a> {
-where
-    TailTokenStream: TokenStream,
-{
     fn advance(&mut self) -> bool {
         if !self.tail.advance() {
             return false;
@@ -48,18 +43,6 @@ where
     }
 }
 
-impl<TailTokenStream> AsciiFoldingFilterTokenStream<TailTokenStream>
-where
-    TailTokenStream: TokenStream,
-{
-    fn wrap(tail: TailTokenStream) -> AsciiFoldingFilterTokenStream<TailTokenStream> {
-        AsciiFoldingFilterTokenStream {
-            tail,
-            buffer: String::with_capacity(100),
-        }
-    }
-}
-
 // Returns a string that represents the ascii folded version of
 // the character. If the `char` does not require ascii folding
 // (e.g. simple ASCII chars like `A`) or if the `char`
@@ -1561,8 +1544,7 @@ mod tests {
     use crate::tokenizer::AsciiFoldingFilter;
     use crate::tokenizer::RawTokenizer;
     use crate::tokenizer::SimpleTokenizer;
-    use crate::tokenizer::TokenStream;
+    use crate::tokenizer::TextAnalyzer;
-    use crate::tokenizer::Tokenizer;
     use std::iter;
 
     #[test]
@@ -1579,7 +1561,7 @@ mod tests {
 
     fn folding_helper(text: &str) -> Vec<String> {
         let mut tokens = Vec::new();
-        SimpleTokenizer
+        TextAnalyzer::from(SimpleTokenizer)
             .filter(AsciiFoldingFilter)
             .token_stream(text)
             .process(&mut |token| {
@@ -1589,7 +1571,9 @@ mod tests {
     }
 
     fn folding_using_raw_tokenizer_helper(text: &str) -> String {
-        let mut token_stream = RawTokenizer.filter(AsciiFoldingFilter).token_stream(text);
+        let mut token_stream = TextAnalyzer::from(RawTokenizer)
+            .filter(AsciiFoldingFilter)
+            .token_stream(text);
         token_stream.advance();
         token_stream.token().text.clone()
     }

src/tokenizer/facet_tokenizer.rs

@@ -1,4 +1,4 @@
-use super::{Token, TokenStream, Tokenizer};
+use super::{BoxTokenStream, Token, TokenStream, Tokenizer};
 use crate::schema::FACET_SEP_BYTE;
 
 /// The `FacetTokenizer` process a `Facet` binary representation
@@ -25,15 +25,14 @@ pub struct FacetTokenStream<'a> {
     token: Token,
 }
 
-impl<'a> Tokenizer<'a> for FacetTokenizer {
+impl Tokenizer for FacetTokenizer {
-    type TokenStreamImpl = FacetTokenStream<'a>;
+    fn token_stream<'a>(&self, text: &'a str) -> BoxTokenStream<'a> {
-
-    fn token_stream(&self, text: &'a str) -> Self::TokenStreamImpl {
         FacetTokenStream {
             text,
             state: State::RootFacetNotEmitted, //< pos is the first char that has not been processed yet.
             token: Token::default(),
         }
+        .into()
     }
 }
 
@@ -84,7 +83,7 @@ mod tests {
 
     use super::FacetTokenizer;
     use crate::schema::Facet;
-    use crate::tokenizer::{Token, TokenStream, Tokenizer};
+    use crate::tokenizer::{Token, Tokenizer};
 
     #[test]
     fn test_facet_tokenizer() {

src/tokenizer/lower_caser.rs

@@ -1,24 +1,23 @@
 use super::{Token, TokenFilter, TokenStream};
+use crate::tokenizer::BoxTokenStream;
 use std::mem;
 
+impl TokenFilter for LowerCaser {
+    fn transform<'a>(&self, token_stream: BoxTokenStream<'a>) -> BoxTokenStream<'a> {
+        BoxTokenStream::from(LowerCaserTokenStream {
+            tail: token_stream,
+            buffer: String::with_capacity(100),
+        })
+    }
+}
+
 /// Token filter that lowercase terms.
 #[derive(Clone)]
 pub struct LowerCaser;
 
-impl<TailTokenStream> TokenFilter<TailTokenStream> for LowerCaser
+pub struct LowerCaserTokenStream<'a> {
-where
-    TailTokenStream: TokenStream,
-{
-    type ResultTokenStream = LowerCaserTokenStream<TailTokenStream>;
-
-    fn transform(&self, token_stream: TailTokenStream) -> Self::ResultTokenStream {
-        LowerCaserTokenStream::wrap(token_stream)
-    }
-}
-
-pub struct LowerCaserTokenStream<TailTokenStream> {
     buffer: String,
-    tail: TailTokenStream,
+    tail: BoxTokenStream<'a>,
 }
 
 // writes a lowercased version of text into output.
@@ -31,18 +30,7 @@ fn to_lowercase_unicode(text: &mut String, output: &mut String) {
     }
 }
 
-impl<TailTokenStream> TokenStream for LowerCaserTokenStream<TailTokenStream>
+impl<'a> TokenStream for LowerCaserTokenStream<'a> {
-where
-    TailTokenStream: TokenStream,
-{
-    fn token(&self) -> &Token {
-        self.tail.token()
-    }
-
-    fn token_mut(&mut self) -> &mut Token {
-        self.tail.token_mut()
-    }
-
     fn advance(&mut self) -> bool {
         if !self.tail.advance() {
             return false;
@@ -56,26 +44,19 @@ where
         }
         true
     }
-}
 
-impl<TailTokenStream> LowerCaserTokenStream<TailTokenStream>
+    fn token(&self) -> &Token {
-where
+        self.tail.token()
-    TailTokenStream: TokenStream,
+    }
-{
+
-    fn wrap(tail: TailTokenStream) -> LowerCaserTokenStream<TailTokenStream> {
+    fn token_mut(&mut self) -> &mut Token {
-        LowerCaserTokenStream {
+        self.tail.token_mut()
-            tail,
-            buffer: String::with_capacity(100),
-        }
     }
 }
 
 #[cfg(test)]
 mod tests {
-    use crate::tokenizer::LowerCaser;
+    use crate::tokenizer::{LowerCaser, SimpleTokenizer, TextAnalyzer};
-    use crate::tokenizer::SimpleTokenizer;
-    use crate::tokenizer::TokenStream;
-    use crate::tokenizer::Tokenizer;
 
     #[test]
     fn test_to_lower_case() {
@@ -87,7 +68,9 @@ mod tests {
 
     fn lowercase_helper(text: &str) -> Vec<String> {
         let mut tokens = vec![];
-        let mut token_stream = SimpleTokenizer.filter(LowerCaser).token_stream(text);
+        let mut token_stream = TextAnalyzer::from(SimpleTokenizer)
+            .filter(LowerCaser)
+            .token_stream(text);
         while token_stream.advance() {
             let token_text = token_stream.token().text.clone();
             tokens.push(token_text);

src/tokenizer/mod.rs

@@ -64,7 +64,7 @@
 //! ```rust
 //! use tantivy::tokenizer::*;
 //!
-//! let en_stem = SimpleTokenizer
+//! let en_stem = TextAnalyzer::from(SimpleTokenizer)
 //!     .filter(RemoveLongFilter::limit(40))
 //!     .filter(LowerCaser)
 //!     .filter(Stemmer::new(Language::English));
@@ -109,7 +109,7 @@
 //! let index = Index::create_in_ram(schema);
 //!
 //! // We need to register our tokenizer :
-//! let custom_en_tokenizer = SimpleTokenizer
+//! let custom_en_tokenizer = TextAnalyzer::from(SimpleTokenizer)
 //!     .filter(RemoveLongFilter::limit(40))
 //!     .filter(LowerCaser);
 //! index
@@ -143,10 +143,11 @@ pub use self::simple_tokenizer::SimpleTokenizer;
 pub use self::stemmer::{Language, Stemmer};
 pub use self::stop_word_filter::StopWordFilter;
 pub(crate) use self::token_stream_chain::TokenStreamChain;
-pub use self::tokenizer::BoxedTokenizer;
 
 pub use self::tokenized_string::{PreTokenizedStream, PreTokenizedString};
-pub use self::tokenizer::{Token, TokenFilter, TokenStream, Tokenizer};
+pub use self::tokenizer::{
+    BoxTokenFilter, BoxTokenStream, TextAnalyzer, Token, TokenFilter, TokenStream, Tokenizer,
+};
 
 pub use self::tokenizer_manager::TokenizerManager;
 
@@ -160,9 +161,9 @@ pub const MAX_TOKEN_LEN: usize = u16::max_value() as usize - 4;
 #[cfg(test)]
 pub mod tests {
     use super::{
-        Language, LowerCaser, RemoveLongFilter, SimpleTokenizer, Stemmer, Token, Tokenizer,
+        Language, LowerCaser, RemoveLongFilter, SimpleTokenizer, Stemmer, Token, TokenizerManager,
-        TokenizerManager,
     };
+    use crate::tokenizer::TextAnalyzer;
 
     /// This is a function that can be used in tests and doc tests
     /// to assert a token's correctness.
@@ -229,7 +230,7 @@ pub mod tests {
         let tokenizer_manager = TokenizerManager::default();
         tokenizer_manager.register(
             "el_stem",
-            SimpleTokenizer
+            TextAnalyzer::from(SimpleTokenizer)
                 .filter(RemoveLongFilter::limit(40))
                 .filter(LowerCaser)
                 .filter(Stemmer::new(Language::Greek)),

src/tokenizer/ngram_tokenizer.rs

@@ -1,4 +1,5 @@
 use super::{Token, TokenStream, Tokenizer};
+use crate::tokenizer::BoxTokenStream;
 
 /// Tokenize the text by splitting words into n-grams of the given size(s)
 ///
@@ -129,11 +130,9 @@ pub struct NgramTokenStream<'a> {
     token: Token,
 }
 
-impl<'a> Tokenizer<'a> for NgramTokenizer {
+impl Tokenizer for NgramTokenizer {
-    type TokenStreamImpl = NgramTokenStream<'a>;
+    fn token_stream<'a>(&self, text: &'a str) -> BoxTokenStream<'a> {
-
+        From::from(NgramTokenStream {
-    fn token_stream(&self, text: &'a str) -> Self::TokenStreamImpl {
-        NgramTokenStream {
             ngram_charidx_iterator: StutteringIterator::new(
                 CodepointFrontiers::for_str(text),
                 self.min_gram,
@@ -142,7 +141,7 @@ impl<'a> Tokenizer<'a> for NgramTokenizer {
             prefix_only: self.prefix_only,
             text,
             token: Token::default(),
-        }
+        })
     }
 }
 
@@ -308,10 +307,10 @@ mod tests {
     use super::NgramTokenizer;
     use super::StutteringIterator;
     use crate::tokenizer::tests::assert_token;
-    use crate::tokenizer::tokenizer::{TokenStream, Tokenizer};
+    use crate::tokenizer::tokenizer::Tokenizer;
-    use crate::tokenizer::Token;
+    use crate::tokenizer::{BoxTokenStream, Token};
 
-    fn test_helper<T: TokenStream>(mut tokenizer: T) -> Vec<Token> {
+    fn test_helper(mut tokenizer: BoxTokenStream) -> Vec<Token> {
         let mut tokens: Vec<Token> = vec![];
         tokenizer.process(&mut |token: &Token| tokens.push(token.clone()));
         tokens

src/tokenizer/raw_tokenizer.rs

@@ -1,4 +1,5 @@
 use super::{Token, TokenStream, Tokenizer};
+use crate::tokenizer::BoxTokenStream;
 
 /// For each value of the field, emit a single unprocessed token.
 #[derive(Clone)]
@@ -9,10 +10,8 @@ pub struct RawTokenStream {
     has_token: bool,
 }
 
-impl<'a> Tokenizer<'a> for RawTokenizer {
+impl Tokenizer for RawTokenizer {
-    type TokenStreamImpl = RawTokenStream;
+    fn token_stream<'a>(&self, text: &'a str) -> BoxTokenStream<'a> {
-
-    fn token_stream(&self, text: &'a str) -> Self::TokenStreamImpl {
         let token = Token {
             offset_from: 0,
             offset_to: text.len(),
@@ -24,6 +23,7 @@ impl<'a> Tokenizer<'a> for RawTokenizer {
             token,
             has_token: true,
         }
+        .into()
     }
 }
 

src/tokenizer/remove_long.rs

@@ -2,7 +2,7 @@
 //! ```rust
 //! use tantivy::tokenizer::*;
 //!
-//! let tokenizer = SimpleTokenizer
+//! let tokenizer = TextAnalyzer::from(SimpleTokenizer)
 //!   .filter(RemoveLongFilter::limit(5));
 //!
 //! let mut stream = tokenizer.token_stream("toolong nice");
@@ -13,6 +13,7 @@
 //! ```
 //!
 use super::{Token, TokenFilter, TokenStream};
+use crate::tokenizer::BoxTokenStream;
 
 /// `RemoveLongFilter` removes tokens that are longer
 /// than a given number of bytes (in UTF-8 representation).
@@ -31,56 +32,27 @@ impl RemoveLongFilter {
     }
 }
 
-impl<TailTokenStream> RemoveLongFilterStream<TailTokenStream>
+impl<'a> RemoveLongFilterStream<'a> {
-where
-    TailTokenStream: TokenStream,
-{
     fn predicate(&self, token: &Token) -> bool {
         token.text.len() < self.token_length_limit
     }
+}
 
-    fn wrap(
+impl TokenFilter for RemoveLongFilter {
-        token_length_limit: usize,
+    fn transform<'a>(&self, token_stream: BoxTokenStream<'a>) -> BoxTokenStream<'a> {
-        tail: TailTokenStream,
+        BoxTokenStream::from(RemoveLongFilterStream {
-    ) -> RemoveLongFilterStream<TailTokenStream> {
+            token_length_limit: self.length_limit,
-        RemoveLongFilterStream {
+            tail: token_stream,
-            token_length_limit,
+        })
-            tail,
-        }
     }
 }
 
-impl<TailTokenStream> TokenFilter<TailTokenStream> for RemoveLongFilter
+pub struct RemoveLongFilterStream<'a> {
-where
-    TailTokenStream: TokenStream,
-{
-    type ResultTokenStream = RemoveLongFilterStream<TailTokenStream>;
-
-    fn transform(&self, token_stream: TailTokenStream) -> Self::ResultTokenStream {
-        RemoveLongFilterStream::wrap(self.length_limit, token_stream)
-    }
-}
-
-pub struct RemoveLongFilterStream<TailTokenStream>
-where
-    TailTokenStream: TokenStream,
-{
     token_length_limit: usize,
-    tail: TailTokenStream,
+    tail: BoxTokenStream<'a>,
 }
 
-impl<TailTokenStream> TokenStream for RemoveLongFilterStream<TailTokenStream>
+impl<'a> TokenStream for RemoveLongFilterStream<'a> {
-where
-    TailTokenStream: TokenStream,
-{
-    fn token(&self) -> &Token {
-        self.tail.token()
-    }
-
-    fn token_mut(&mut self) -> &mut Token {
-        self.tail.token_mut()
-    }
-
     fn advance(&mut self) -> bool {
         while self.tail.advance() {
             if self.predicate(self.tail.token()) {
@@ -89,4 +61,12 @@ where
         }
         false
     }
+
+    fn token(&self) -> &Token {
+        self.tail.token()
+    }
+
+    fn token_mut(&mut self) -> &mut Token {
+        self.tail.token_mut()
+    }
 }

src/tokenizer/simple_tokenizer.rs

@@ -1,3 +1,4 @@
+use super::BoxTokenStream;
 use super::{Token, TokenStream, Tokenizer};
 use std::str::CharIndices;
 
@@ -11,15 +12,13 @@ pub struct SimpleTokenStream<'a> {
     token: Token,
 }
 
-impl<'a> Tokenizer<'a> for SimpleTokenizer {
+impl Tokenizer for SimpleTokenizer {
-    type TokenStreamImpl = SimpleTokenStream<'a>;
+    fn token_stream<'a>(&self, text: &'a str) -> BoxTokenStream<'a> {
-
+        BoxTokenStream::from(SimpleTokenStream {
-    fn token_stream(&self, text: &'a str) -> Self::TokenStreamImpl {
-        SimpleTokenStream {
             text,
             chars: text.char_indices(),
             token: Token::default(),
-        }
+        })
     }
 }
 

src/tokenizer/stemmer.rs

@@ -1,4 +1,5 @@
 use super::{Token, TokenFilter, TokenStream};
+use crate::tokenizer::BoxTokenStream;
 use rust_stemmers::{self, Algorithm};
 
 /// Available stemmer languages.
@@ -75,38 +76,22 @@ impl Default for Stemmer {
     }
 }
 
-impl<TailTokenStream> TokenFilter<TailTokenStream> for Stemmer
+impl TokenFilter for Stemmer {
-where
+    fn transform<'a>(&self, token_stream: BoxTokenStream<'a>) -> BoxTokenStream<'a> {
-    TailTokenStream: TokenStream,
-{
-    type ResultTokenStream = StemmerTokenStream<TailTokenStream>;
-
-    fn transform(&self, token_stream: TailTokenStream) -> Self::ResultTokenStream {
         let inner_stemmer = rust_stemmers::Stemmer::create(self.stemmer_algorithm);
-        StemmerTokenStream::wrap(inner_stemmer, token_stream)
+        BoxTokenStream::from(StemmerTokenStream {
+            tail: token_stream,
+            stemmer: inner_stemmer,
+        })
     }
 }
 
-pub struct StemmerTokenStream<TailTokenStream>
+pub struct StemmerTokenStream<'a> {
-where
+    tail: BoxTokenStream<'a>,
-    TailTokenStream: TokenStream,
-{
-    tail: TailTokenStream,
     stemmer: rust_stemmers::Stemmer,
 }
 
-impl<TailTokenStream> TokenStream for StemmerTokenStream<TailTokenStream>
+impl<'a> TokenStream for StemmerTokenStream<'a> {
-where
-    TailTokenStream: TokenStream,
-{
-    fn token(&self) -> &Token {
-        self.tail.token()
-    }
-
-    fn token_mut(&mut self) -> &mut Token {
-        self.tail.token_mut()
-    }
-
     fn advance(&mut self) -> bool {
         if !self.tail.advance() {
             return false;
@@ -117,16 +102,12 @@ where
         self.token_mut().text.push_str(&stemmed_str);
         true
     }
-}
 
-impl<TailTokenStream> StemmerTokenStream<TailTokenStream>
+    fn token(&self) -> &Token {
-where
+        self.tail.token()
-    TailTokenStream: TokenStream,
+    }
-{
+
-    fn wrap(
+    fn token_mut(&mut self) -> &mut Token {
-        stemmer: rust_stemmers::Stemmer,
+        self.tail.token_mut()
-        tail: TailTokenStream,
-    ) -> StemmerTokenStream<TailTokenStream> {
-        StemmerTokenStream { tail, stemmer }
     }
 }

src/tokenizer/stop_word_filter.rs

@@ -2,7 +2,7 @@
 //! ```rust
 //! use tantivy::tokenizer::*;
 //!
-//! let tokenizer = SimpleTokenizer
+//! let tokenizer = TextAnalyzer::from(SimpleTokenizer)
 //!   .filter(StopWordFilter::remove(vec!["the".to_string(), "is".to_string()]));
 //!
 //! let mut stream = tokenizer.token_stream("the fox is crafty");
@@ -11,6 +11,7 @@
 //! assert!(stream.next().is_none());
 //! ```
 use super::{Token, TokenFilter, TokenStream};
+use crate::tokenizer::BoxTokenStream;
 use fnv::FnvHasher;
 use std::collections::HashSet;
 use std::hash::BuildHasherDefault;
@@ -48,53 +49,27 @@ impl StopWordFilter {
     }
 }
 
-pub struct StopWordFilterStream<TailTokenStream>
+pub struct StopWordFilterStream<'a> {
-where
-    TailTokenStream: TokenStream,
-{
     words: StopWordHashSet,
-    tail: TailTokenStream,
+    tail: BoxTokenStream<'a>,
 }
 
-impl<TailTokenStream> TokenFilter<TailTokenStream> for StopWordFilter
+impl TokenFilter for StopWordFilter {
-where
+    fn transform<'a>(&self, token_stream: BoxTokenStream<'a>) -> BoxTokenStream<'a> {
-    TailTokenStream: TokenStream,
+        BoxTokenStream::from(StopWordFilterStream {
-{
+            words: self.words.clone(),
-    type ResultTokenStream = StopWordFilterStream<TailTokenStream>;
+            tail: token_stream,
-
+        })
-    fn transform(&self, token_stream: TailTokenStream) -> Self::ResultTokenStream {
-        StopWordFilterStream::wrap(self.words.clone(), token_stream)
     }
 }
 
-impl<TailTokenStream> StopWordFilterStream<TailTokenStream>
+impl<'a> StopWordFilterStream<'a> {
-where
-    TailTokenStream: TokenStream,
-{
     fn predicate(&self, token: &Token) -> bool {
         !self.words.contains(&token.text)
     }
-
-    fn wrap(
-        words: StopWordHashSet,
-        tail: TailTokenStream,
-    ) -> StopWordFilterStream<TailTokenStream> {
-        StopWordFilterStream { words, tail }
-    }
 }
 
-impl<TailTokenStream> TokenStream for StopWordFilterStream<TailTokenStream>
+impl<'a> TokenStream for StopWordFilterStream<'a> {
-where
-    TailTokenStream: TokenStream,
-{
-    fn token(&self) -> &Token {
-        self.tail.token()
-    }
-
-    fn token_mut(&mut self) -> &mut Token {
-        self.tail.token_mut()
-    }
-
     fn advance(&mut self) -> bool {
         while self.tail.advance() {
             if self.predicate(self.tail.token()) {
@@ -103,6 +78,14 @@ where
         }
         false
     }
+
+    fn token(&self) -> &Token {
+        self.tail.token()
+    }
+
+    fn token_mut(&mut self) -> &mut Token {
+        self.tail.token_mut()
+    }
 }
 
 impl Default for StopWordFilter {

src/tokenizer/token_stream_chain.rs

@@ -1,23 +1,21 @@
-use crate::tokenizer::{Token, TokenStream};
+use crate::tokenizer::{BoxTokenStream, Token, TokenStream};
+use std::ops::DerefMut;
 
 const POSITION_GAP: usize = 2;
 
-pub(crate) struct TokenStreamChain<TTokenStream: TokenStream> {
+pub(crate) struct TokenStreamChain<'a> {
     offsets: Vec<usize>,
-    token_streams: Vec<TTokenStream>,
+    token_streams: Vec<BoxTokenStream<'a>>,
     position_shift: usize,
     stream_idx: usize,
     token: Token,
 }
 
-impl<'a, TTokenStream> TokenStreamChain<TTokenStream>
+impl<'a> TokenStreamChain<'a> {
-where
-    TTokenStream: TokenStream,
-{
     pub fn new(
         offsets: Vec<usize>,
-        token_streams: Vec<TTokenStream>,
+        token_streams: Vec<BoxTokenStream<'a>>,
-    ) -> TokenStreamChain<TTokenStream> {
+    ) -> TokenStreamChain<'a> {
         TokenStreamChain {
             offsets,
             stream_idx: 0,
@@ -28,13 +26,10 @@ where
     }
 }
 
-impl<'a, TTokenStream> TokenStream for TokenStreamChain<TTokenStream>
+impl<'a> TokenStream for TokenStreamChain<'a> {
-where
-    TTokenStream: TokenStream,
-{
     fn advance(&mut self) -> bool {
         while self.stream_idx < self.token_streams.len() {
-            let token_stream = &mut self.token_streams[self.stream_idx];
+            let token_stream = self.token_streams[self.stream_idx].deref_mut();
             if token_stream.advance() {
                 let token = token_stream.token();
                 let offset_offset = self.offsets[self.stream_idx];

src/tokenizer/tokenized_string.rs

@@ -1,4 +1,4 @@
-use crate::tokenizer::{Token, TokenStream, TokenStreamChain};
+use crate::tokenizer::{BoxTokenStream, Token, TokenStream, TokenStreamChain};
 use std::cmp::Ordering;
 
 /// Struct representing pre-tokenized text
@@ -41,9 +41,9 @@ impl PreTokenizedStream {
     /// Creates a TokenStream from PreTokenizedString array
     pub fn chain_tokenized_strings<'a>(
         tok_strings: &'a [&'a PreTokenizedString],
-    ) -> Box<dyn TokenStream + 'a> {
+    ) -> BoxTokenStream {
         if tok_strings.len() == 1 {
-            Box::new(PreTokenizedStream::from((*tok_strings[0]).clone()))
+            PreTokenizedStream::from((*tok_strings[0]).clone()).into()
         } else {
             let mut offsets = vec![];
             let mut total_offset = 0;
@@ -53,11 +53,12 @@ impl PreTokenizedStream {
                     total_offset += last_token.offset_to;
                 }
             }
-            let token_streams: Vec<_> = tok_strings
+            // TODO remove the string cloning.
+            let token_streams: Vec<BoxTokenStream<'static>> = tok_strings
                 .iter()
-                .map(|tok_string| PreTokenizedStream::from((*tok_string).clone()))
+                .map(|&tok_string| PreTokenizedStream::from((*tok_string).clone()).into())
                 .collect();
-            Box::new(TokenStreamChain::new(offsets, token_streams))
+            TokenStreamChain::new(offsets, token_streams).into()
         }
     }
 }

src/tokenizer/tokenizer.rs

@@ -2,6 +2,7 @@ use crate::tokenizer::TokenStreamChain;
 /// The tokenizer module contains all of the tools used to process
 /// text in `tantivy`.
 use std::borrow::{Borrow, BorrowMut};
+use std::ops::{Deref, DerefMut};
 
 /// Token
 #[derive(Debug, Clone, Serialize, Deserialize, Eq, PartialEq)]
@@ -33,20 +34,31 @@ impl Default for Token {
     }
 }
 
-/// `Tokenizer` are in charge of splitting text into a stream of token
+/// `TextAnalyzer` tokenizes an input text into tokens and modifies the resulting `TokenStream`.
-/// before indexing.
 ///
-/// See the [module documentation](./index.html) for more detail.
+/// It simply wraps a `Tokenizer` and a list of `TokenFilter` that are applied sequentially.
-///
+pub struct TextAnalyzer {
-/// # Warning
+    tokenizer: Box<dyn Tokenizer>,
-///
+    token_filters: Vec<BoxTokenFilter>,
-/// This API may change to use associated types.
+}
-pub trait Tokenizer<'a>: Sized + Clone {
-    /// Type associated to the resulting tokenstream tokenstream.
-    type TokenStreamImpl: TokenStream;
 
-    /// Creates a token stream for a given `str`.
+impl<T: Tokenizer> From<T> for TextAnalyzer {
-    fn token_stream(&self, text: &'a str) -> Self::TokenStreamImpl;
+    fn from(tokenizer: T) -> Self {
+        TextAnalyzer::new(tokenizer, Vec::new())
+    }
+}
+
+impl TextAnalyzer {
+    /// Creates a new `TextAnalyzer` given a tokenizer and a vector of `BoxTokenFilter`.
+    ///
+    /// When creating a `TextAnalyzer` from a `Tokenizer` alone, prefer using
+    /// `TextAnalyzer::from(tokenizer)`.
+    pub fn new<T: Tokenizer>(tokenizer: T, token_filters: Vec<BoxTokenFilter>) -> TextAnalyzer {
+        TextAnalyzer {
+            tokenizer: Box::new(tokenizer),
+            token_filters,
+        }
+    }
 
     /// Appends a token filter to the current tokenizer.
     ///
@@ -58,90 +70,26 @@ pub trait Tokenizer<'a>: Sized + Clone {
     /// ```rust
     /// use tantivy::tokenizer::*;
     ///
-    /// let en_stem = SimpleTokenizer
+    /// let en_stem = TextAnalyzer::from(SimpleTokenizer)
     ///     .filter(RemoveLongFilter::limit(40))
     ///     .filter(LowerCaser)
     ///     .filter(Stemmer::default());
     /// ```
     ///
-    fn filter<NewFilter>(self, new_filter: NewFilter) -> ChainTokenizer<NewFilter, Self>
+    pub fn filter<F: Into<BoxTokenFilter>>(mut self, token_filter: F) -> Self {
-    where
+        self.token_filters.push(token_filter.into());
-        NewFilter: TokenFilter<<Self as Tokenizer<'a>>::TokenStreamImpl>,
+        self
-    {
-        ChainTokenizer {
-            head: new_filter,
-            tail: self,
-        }
-    }
-}
-
-/// A boxed tokenizer
-trait BoxedTokenizerTrait: Send + Sync {
-    /// Tokenize a `&str`
-    fn token_stream<'a>(&self, text: &'a str) -> Box<dyn TokenStream + 'a>;
-
-    /// Tokenize an array`&str`
-    ///
-    /// The resulting `TokenStream` is equivalent to what would be obtained if the &str were
-    /// one concatenated `&str`, with an artificial position gap of `2` between the different fields
-    /// to prevent accidental `PhraseQuery` to match accross two terms.
-    fn token_stream_texts<'b>(&self, texts: &'b [&'b str]) -> Box<dyn TokenStream + 'b>;
-
-    /// Return a boxed clone of the tokenizer
-    fn boxed_clone(&self) -> BoxedTokenizer;
-}
-
-/// A boxed tokenizer
-pub struct BoxedTokenizer(Box<dyn BoxedTokenizerTrait>);
-
-impl<T> From<T> for BoxedTokenizer
-where
-    T: 'static + Send + Sync + for<'a> Tokenizer<'a>,
-{
-    fn from(tokenizer: T) -> BoxedTokenizer {
-        BoxedTokenizer(Box::new(BoxableTokenizer(tokenizer)))
-    }
-}
-
-impl BoxedTokenizer {
-    /// Tokenize a `&str`
-    pub fn token_stream<'a>(&self, text: &'a str) -> Box<dyn TokenStream + 'a> {
-        self.0.token_stream(text)
     }
 
     /// Tokenize an array`&str`
     ///
-    /// The resulting `TokenStream` is equivalent to what would be obtained if the &str were
+    /// The resulting `BoxTokenStream` is equivalent to what would be obtained if the &str were
     /// one concatenated `&str`, with an artificial position gap of `2` between the different fields
     /// to prevent accidental `PhraseQuery` to match accross two terms.
-    pub fn token_stream_texts<'b>(&self, texts: &'b [&'b str]) -> Box<dyn TokenStream + 'b> {
+    pub fn token_stream_texts<'a>(&self, texts: &'a [&'a str]) -> BoxTokenStream<'a> {
-        self.0.token_stream_texts(texts)
-    }
-}
-
-impl Clone for BoxedTokenizer {
-    fn clone(&self) -> BoxedTokenizer {
-        self.0.boxed_clone()
-    }
-}
-
-#[derive(Clone)]
-struct BoxableTokenizer<A>(A)
-where
-    A: for<'a> Tokenizer<'a> + Send + Sync;
-
-impl<A> BoxedTokenizerTrait for BoxableTokenizer<A>
-where
-    A: 'static + Send + Sync + for<'a> Tokenizer<'a>,
-{
-    fn token_stream<'a>(&self, text: &'a str) -> Box<dyn TokenStream + 'a> {
-        Box::new(self.0.token_stream(text))
-    }
-
-    fn token_stream_texts<'b>(&self, texts: &'b [&'b str]) -> Box<dyn TokenStream + 'b> {
         assert!(!texts.is_empty());
         if texts.len() == 1 {
-            Box::new(self.0.token_stream(texts[0]))
+            self.token_stream(texts[0])
         } else {
             let mut offsets = vec![];
             let mut total_offset = 0;
@@ -149,34 +97,124 @@ where
                 offsets.push(total_offset);
                 total_offset += text.len();
             }
-            let token_streams: Vec<_> =
+            let token_streams: Vec<BoxTokenStream<'a>> = texts
-                texts.iter().map(|text| self.0.token_stream(text)).collect();
+                .iter()
-            Box::new(TokenStreamChain::new(offsets, token_streams))
+                .cloned()
+                .map(|text| self.token_stream(text))
+                .collect();
+            From::from(TokenStreamChain::new(offsets, token_streams))
         }
     }
 
-    fn boxed_clone(&self) -> BoxedTokenizer {
+    /// Creates a token stream for a given `str`.
-        self.0.clone().into()
+    pub fn token_stream<'a>(&self, text: &'a str) -> BoxTokenStream<'a> {
+        let mut token_stream = self.tokenizer.token_stream(text);
+        for token_filter in &self.token_filters {
+            token_stream = token_filter.transform(token_stream);
+        }
+        token_stream
     }
 }
 
-impl<'b> TokenStream for Box<dyn TokenStream + 'b> {
+impl Clone for TextAnalyzer {
+    fn clone(&self) -> Self {
+        TextAnalyzer {
+            tokenizer: self.tokenizer.box_clone(),
+            token_filters: self
+                .token_filters
+                .iter()
+                .map(|token_filter| token_filter.box_clone())
+                .collect(),
+        }
+    }
+}
+
+/// `Tokenizer` are in charge of splitting text into a stream of token
+/// before indexing.
+///
+/// See the [module documentation](./index.html) for more detail.
+///
+/// # Warning
+///
+/// This API may change to use associated types.
+pub trait Tokenizer: 'static + Send + Sync + TokenizerClone {
+    /// Creates a token stream for a given `str`.
+    fn token_stream<'a>(&self, text: &'a str) -> BoxTokenStream<'a>;
+}
+
+pub trait TokenizerClone {
+    fn box_clone(&self) -> Box<dyn Tokenizer>;
+}
+
+impl<T: Tokenizer + Clone> TokenizerClone for T {
+    fn box_clone(&self) -> Box<dyn Tokenizer> {
+        Box::new(self.clone())
+    }
+}
+
+impl<'a> TokenStream for Box<dyn TokenStream + 'a> {
     fn advance(&mut self) -> bool {
         let token_stream: &mut dyn TokenStream = self.borrow_mut();
         token_stream.advance()
     }
 
-    fn token(&self) -> &Token {
+    fn token<'b>(&'b self) -> &'b Token {
-        let token_stream: &dyn TokenStream = self.borrow();
+        let token_stream: &'b (dyn TokenStream + 'a) = self.borrow();
         token_stream.token()
     }
 
-    fn token_mut(&mut self) -> &mut Token {
+    fn token_mut<'b>(&'b mut self) -> &'b mut Token {
-        let token_stream: &mut dyn TokenStream = self.borrow_mut();
+        let token_stream: &'b mut (dyn TokenStream + 'a) = self.borrow_mut();
         token_stream.token_mut()
     }
 }
 
+/// Simple wrapper of `Box<dyn TokenStream + 'a>`.
+///
+/// See `TokenStream` for more information.
+pub struct BoxTokenStream<'a>(Box<dyn TokenStream + 'a>);
+
+impl<'a, T> From<T> for BoxTokenStream<'a>
+where
+    T: TokenStream + 'a,
+{
+    fn from(token_stream: T) -> BoxTokenStream<'a> {
+        BoxTokenStream(Box::new(token_stream))
+    }
+}
+
+impl<'a> Deref for BoxTokenStream<'a> {
+    type Target = dyn TokenStream + 'a;
+
+    fn deref(&self) -> &Self::Target {
+        &*self.0
+    }
+}
+impl<'a> DerefMut for BoxTokenStream<'a> {
+    fn deref_mut(&mut self) -> &mut Self::Target {
+        &mut *self.0
+    }
+}
+
+/// Simple wrapper of `Box<dyn TokenFilter + 'a>`.
+///
+/// See `TokenStream` for more information.
+pub struct BoxTokenFilter(Box<dyn TokenFilter>);
+
+impl Deref for BoxTokenFilter {
+    type Target = dyn TokenFilter;
+
+    fn deref(&self) -> &dyn TokenFilter {
+        &*self.0
+    }
+}
+
+impl<T: TokenFilter> From<T> for BoxTokenFilter {
+    fn from(tokenizer: T) -> BoxTokenFilter {
+        BoxTokenFilter(Box::new(tokenizer))
+    }
+}
+
 /// `TokenStream` is the result of the tokenization.
 ///
 /// It consists consumable stream of `Token`s.
@@ -186,7 +224,7 @@ impl<'b> TokenStream for Box<dyn TokenStream + 'b> {
 /// ```
 /// use tantivy::tokenizer::*;
 ///
-/// let tokenizer = SimpleTokenizer
+/// let tokenizer = TextAnalyzer::from(SimpleTokenizer)
 ///        .filter(RemoveLongFilter::limit(40))
 ///        .filter(LowerCaser);
 /// let mut token_stream = tokenizer.token_stream("Hello, happy tax payer");
@@ -225,7 +263,7 @@ pub trait TokenStream {
     /// ```
     /// use tantivy::tokenizer::*;
     ///
-    /// let tokenizer = SimpleTokenizer
+    /// let tokenizer = TextAnalyzer::from(SimpleTokenizer)
     ///       .filter(RemoveLongFilter::limit(40))
     ///       .filter(LowerCaser);
     /// let mut token_stream = tokenizer.token_stream("Hello, happy tax payer");
@@ -243,6 +281,8 @@ pub trait TokenStream {
 
     /// Helper function to consume the entire `TokenStream`
     /// and push the tokens to a sink function.
+    ///
+    /// Remove this.
     fn process(&mut self, sink: &mut dyn FnMut(&Token)) -> u32 {
         let mut num_tokens_pushed = 0u32;
         while self.advance() {
@@ -253,33 +293,20 @@ pub trait TokenStream {
     }
 }
 
-#[derive(Clone)]
+pub trait TokenFilterClone {
-pub struct ChainTokenizer<HeadTokenFilterFactory, TailTokenizer> {
+    fn box_clone(&self) -> BoxTokenFilter;
-    head: HeadTokenFilterFactory,
-    tail: TailTokenizer,
-}
-
-impl<'a, HeadTokenFilterFactory, TailTokenizer> Tokenizer<'a>
-    for ChainTokenizer<HeadTokenFilterFactory, TailTokenizer>
-where
-    HeadTokenFilterFactory: TokenFilter<TailTokenizer::TokenStreamImpl>,
-    TailTokenizer: Tokenizer<'a>,
-{
-    type TokenStreamImpl = HeadTokenFilterFactory::ResultTokenStream;
-
-    fn token_stream(&self, text: &'a str) -> Self::TokenStreamImpl {
-        let tail_token_stream = self.tail.token_stream(text);
-        self.head.transform(tail_token_stream)
-    }
 }
 
 /// Trait for the pluggable components of `Tokenizer`s.
-pub trait TokenFilter<TailTokenStream: TokenStream>: Clone {
+pub trait TokenFilter: 'static + Send + Sync + TokenFilterClone {
-    /// The resulting `TokenStream` type.
-    type ResultTokenStream: TokenStream;
-
     /// Wraps a token stream and returns the modified one.
-    fn transform(&self, token_stream: TailTokenStream) -> Self::ResultTokenStream;
+    fn transform<'a>(&self, token_stream: BoxTokenStream<'a>) -> BoxTokenStream<'a>;
+}
+
+impl<T: TokenFilter + Clone> TokenFilterClone for T {
+    fn box_clone(&self) -> BoxTokenFilter {
+        BoxTokenFilter::from(self.clone())
+    }
 }
 
 #[cfg(test)]

src/tokenizer/tokenizer_manager.rs

@@ -1,11 +1,10 @@
 use crate::tokenizer::stemmer::Language;
-use crate::tokenizer::BoxedTokenizer;
+use crate::tokenizer::tokenizer::TextAnalyzer;
 use crate::tokenizer::LowerCaser;
 use crate::tokenizer::RawTokenizer;
 use crate::tokenizer::RemoveLongFilter;
 use crate::tokenizer::SimpleTokenizer;
 use crate::tokenizer::Stemmer;
-use crate::tokenizer::Tokenizer;
 use std::collections::HashMap;
 use std::sync::{Arc, RwLock};
 
@@ -23,16 +22,16 @@ use std::sync::{Arc, RwLock};
 ///  search engine.
 #[derive(Clone)]
 pub struct TokenizerManager {
-    tokenizers: Arc<RwLock<HashMap<String, BoxedTokenizer>>>,
+    tokenizers: Arc<RwLock<HashMap<String, TextAnalyzer>>>,
 }
 
 impl TokenizerManager {
     /// Registers a new tokenizer associated with a given name.
-    pub fn register<A>(&self, tokenizer_name: &str, tokenizer: A)
+    pub fn register<T>(&self, tokenizer_name: &str, tokenizer: T)
     where
-        A: Into<BoxedTokenizer>,
+        TextAnalyzer: From<T>,
     {
-        let boxed_tokenizer = tokenizer.into();
+        let boxed_tokenizer: TextAnalyzer = TextAnalyzer::from(tokenizer);
         self.tokenizers
             .write()
             .expect("Acquiring the lock should never fail")
@@ -40,7 +39,7 @@ impl TokenizerManager {
     }
 
     /// Accessing a tokenizer given its name.
-    pub fn get(&self, tokenizer_name: &str) -> Option<BoxedTokenizer> {
+    pub fn get(&self, tokenizer_name: &str) -> Option<TextAnalyzer> {
         self.tokenizers
             .read()
             .expect("Acquiring the lock should never fail")
@@ -62,13 +61,13 @@ impl Default for TokenizerManager {
         manager.register("raw", RawTokenizer);
         manager.register(
             "default",
-            SimpleTokenizer
+            TextAnalyzer::from(SimpleTokenizer)
                 .filter(RemoveLongFilter::limit(40))
                 .filter(LowerCaser),
         );
         manager.register(
             "en_stem",
-            SimpleTokenizer
+            TextAnalyzer::from(SimpleTokenizer)
                 .filter(RemoveLongFilter::limit(40))
                 .filter(LowerCaser)
                 .filter(Stemmer::new(Language::English)),