test store

Issue/969

Cargo.toml

@@ -47,6 +47,7 @@ murmurhash32 = "0.2"
 chrono = "0.4"
 smallvec = "1"
 rayon = "1"
+env_logger = "0.8"
 lru = "0.6"
 
 [target.'cfg(windows)'.dependencies]

benches/analyzer.rs

@@ -10,7 +10,7 @@ pub fn criterion_benchmark(c: &mut Criterion) {
         b.iter(|| {
             let mut word_count = 0;
             let mut token_stream = tokenizer.token_stream(ALICE_TXT);
-            for token in token_stream {
+            while token_stream.advance() {
                 word_count += 1;
             }
             assert_eq!(word_count, 30_731);

examples/pre_tokenized_text.rs

@@ -17,7 +17,12 @@ use tantivy::{doc, Index, ReloadPolicy};
 use tempfile::TempDir;
 
 fn pre_tokenize_text(text: &str) -> Vec<Token> {
-    SimpleTokenizer.token_stream(text).collect()
+    let mut token_stream = SimpleTokenizer.token_stream(text);
+    let mut tokens = vec![];
+    while token_stream.advance() {
+        tokens.push(token_stream.token().clone());
+    }
+    tokens
 }
 
 fn main() -> tantivy::Result<()> {

examples/snippet.rs

@@ -51,7 +51,7 @@ fn main() -> tantivy::Result<()> {
 
     let top_docs = searcher.search(&query, &TopDocs::with_limit(10))?;
 
-    let mut snippet_generator = SnippetGenerator::create(&searcher, &*query, body)?;
+    let snippet_generator = SnippetGenerator::create(&searcher, &*query, body)?;
 
     for (score, doc_address) in top_docs {
         let doc = searcher.doc(doc_address)?;

examples/stop_words.rs

@@ -50,13 +50,12 @@ 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 = analyzer_builder(SimpleTokenizer)
-        .filter(LowerCaser::new())
+    let tokenizer = TextAnalyzer::from(SimpleTokenizer)
+        .filter(LowerCaser)
         .filter(StopWordFilter::remove(vec![
             "the".to_string(),
             "and".to_string(),
-        ]))
-        .build();
+        ]));
 
     index.tokenizers().register("stoppy", tokenizer);
 

src/core/index.rs

@@ -20,7 +20,7 @@ use crate::reader::IndexReaderBuilder;
 use crate::schema::Field;
 use crate::schema::FieldType;
 use crate::schema::Schema;
-use crate::tokenizer::{TextAnalyzerT, TokenizerManager};
+use crate::tokenizer::{TextAnalyzer, TokenizerManager};
 use crate::IndexWriter;
 use std::collections::HashSet;
 use std::fmt;
@@ -35,12 +35,18 @@ fn load_metas(
     inventory: &SegmentMetaInventory,
 ) -> crate::Result<IndexMeta> {
     let meta_data = directory.atomic_read(&META_FILEPATH)?;
-    let meta_string = String::from_utf8_lossy(&meta_data);
+    let meta_string = String::from_utf8(meta_data)
+        .map_err(|utf8_err| {
+            DataCorruption::new(
+                META_FILEPATH.to_path_buf(),
+                format!("Meta file is not valid utf-8. {:?}", utf8_err)
+            )
+        })?;
     IndexMeta::deserialize(&meta_string, &inventory)
         .map_err(|e| {
             DataCorruption::new(
                 META_FILEPATH.to_path_buf(),
-                format!("Meta file cannot be deserialized. {:?}.", e),
+                format!("Meta file cannot be deserialized. {:?}. content = {}", e, meta_string),
             )
         })
         .map_err(From::from)
@@ -119,12 +125,13 @@ impl Index {
             return Index::create(dir, schema);
         }
         let index = Index::open(dir)?;
-        if index.schema() != schema {
-            return Err(TantivyError::SchemaError(
+        if index.schema() == schema {
+            Ok(index)
+        } else {
+            Err(TantivyError::SchemaError(
                 "An index exists but the schema does not match.".to_string(),
-            ));
+            ))
         }
-        Ok(index)
     }
 
     /// Creates a new index in a temp directory.
@@ -180,11 +187,11 @@ impl Index {
     }
 
     /// Helper to access the tokenizer associated to a specific field.
-    pub fn tokenizer_for_field(&self, field: Field) -> crate::Result<Box<dyn TextAnalyzerT>> {
+    pub fn tokenizer_for_field(&self, field: Field) -> crate::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<Box<dyn TextAnalyzerT>> = 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())

src/core/segment_reader.rs

@@ -310,7 +310,7 @@ impl SegmentReader {
     }
 
     /// Returns an iterator that will iterate over the alive document ids
-    pub fn doc_ids_alive(&self) -> impl Iterator<Item = DocId> + '_ {
+    pub fn doc_ids_alive<'a>(&'a self) -> impl Iterator<Item = DocId> + 'a {
         (0u32..self.max_doc).filter(move |doc| !self.is_deleted(*doc))
     }
 

src/functional_test.rs

@@ -1,31 +1,77 @@
-use rand::thread_rng;
-use std::collections::HashSet;
-
-use crate::schema::*;
 use crate::Index;
 use crate::Searcher;
+use crate::{doc, schema::*};
+use rand::thread_rng;
 use rand::Rng;
+use std::collections::HashSet;
 
-fn check_index_content(searcher: &Searcher, vals: &HashSet<u64>) {
+fn check_index_content(searcher: &Searcher, vals: &[u64]) -> crate::Result<()> {
     assert!(searcher.segment_readers().len() < 20);
     assert_eq!(searcher.num_docs() as usize, vals.len());
+    for segment_reader in searcher.segment_readers() {
+        let store_reader = segment_reader.get_store_reader()?;
+        for doc_id in 0..segment_reader.max_doc() {
+            let _doc = store_reader.get(doc_id)?;
+        }
+    }
+    Ok(())
 }
 
 #[test]
 #[ignore]
-fn test_indexing() {
+fn test_functional_store() -> crate::Result<()> {
+    env_logger::init();
+    let mut schema_builder = Schema::builder();
+
+    let id_field = schema_builder.add_u64_field("id", INDEXED | STORED);
+    let schema = schema_builder.build();
+
+    let index = Index::create_in_ram(schema);
+    let reader = index.reader()?;
+
+    let mut rng = thread_rng();
+
+    let mut index_writer = index.writer_with_num_threads(3, 12_000_000)?;
+
+    let mut doc_set: Vec<u64> = Vec::new();
+
+    let mut doc_id = 0u64;
+    for iteration in 0.. {
+        let num_docs: usize = rng.gen_range(0..4);
+        if doc_set.len() >= 1 {
+            let doc_to_remove_id = rng.gen_range(0..doc_set.len());
+            let removed_doc_id = doc_set.swap_remove(doc_to_remove_id);
+            index_writer.delete_term(Term::from_field_u64(id_field, removed_doc_id));
+        }
+        for _ in 0..num_docs {
+            doc_set.push(doc_id);
+            index_writer.add_document(doc!(id_field=>doc_id));
+            doc_id += 1;
+        }
+        index_writer.commit()?;
+        reader.reload()?;
+        let searcher = reader.searcher();
+        println!("#{} - {}", iteration, searcher.segment_readers().len());
+        check_index_content(&searcher, &doc_set)?;
+    }
+    Ok(())
+}
+
+#[test]
+#[ignore]
+fn test_functional_indexing() -> crate::Result<()> {
     let mut schema_builder = Schema::builder();
 
     let id_field = schema_builder.add_u64_field("id", INDEXED);
     let multiples_field = schema_builder.add_u64_field("multiples", INDEXED);
     let schema = schema_builder.build();
 
-    let index = Index::create_from_tempdir(schema).unwrap();
-    let reader = index.reader().unwrap();
+    let index = Index::create_from_tempdir(schema)?;
+    let reader = index.reader()?;
 
     let mut rng = thread_rng();
 
-    let mut index_writer = index.writer_with_num_threads(3, 120_000_000).unwrap();
+    let mut index_writer = index.writer_with_num_threads(3, 120_000_000)?;
 
     let mut committed_docs: HashSet<u64> = HashSet::new();
     let mut uncommitted_docs: HashSet<u64> = HashSet::new();
@@ -33,13 +79,16 @@ fn test_indexing() {
     for _ in 0..200 {
         let random_val = rng.gen_range(0..20);
         if random_val == 0 {
-            index_writer.commit().expect("Commit failed");
+            index_writer.commit()?;
             committed_docs.extend(&uncommitted_docs);
             uncommitted_docs.clear();
-            reader.reload().unwrap();
+            reader.reload()?;
             let searcher = reader.searcher();
             // check that everything is correct.
-            check_index_content(&searcher, &committed_docs);
+            check_index_content(
+                &searcher,
+                &committed_docs.iter().cloned().collect::<Vec<u64>>(),
+            )?;
         } else {
             if committed_docs.remove(&random_val) || uncommitted_docs.remove(&random_val) {
                 let doc_id_term = Term::from_field_u64(id_field, random_val);
@@ -55,4 +104,5 @@ fn test_indexing() {
             }
         }
     }
+    Ok(())
 }

src/indexer/segment_writer.rs

@@ -10,9 +10,10 @@ use crate::schema::FieldType;
 use crate::schema::Schema;
 use crate::schema::Term;
 use crate::schema::Value;
-use crate::tokenizer::PreTokenizedStream;
-use crate::tokenizer::{DynTokenStreamChain, Tokenizer};
-use crate::tokenizer::{FacetTokenizer, TextAnalyzerT, Token};
+use crate::schema::{Field, FieldEntry};
+use crate::tokenizer::{BoxTokenStream, PreTokenizedStream};
+use crate::tokenizer::{FacetTokenizer, TextAnalyzer};
+use crate::tokenizer::{TokenStreamChain, Tokenizer};
 use crate::Opstamp;
 use crate::{DocId, SegmentComponent};
 
@@ -22,7 +23,7 @@ use crate::{DocId, SegmentComponent};
 fn initial_table_size(per_thread_memory_budget: usize) -> crate::Result<usize> {
     let table_memory_upper_bound = per_thread_memory_budget / 3;
     if let Some(limit) = (10..)
-        .take_while(|&num_bits| compute_table_size(num_bits) < table_memory_upper_bound)
+        .take_while(|num_bits: &usize| compute_table_size(*num_bits) < table_memory_upper_bound)
         .last()
     {
         Ok(limit.min(19)) // we cap it at 2^19 = 512K.
@@ -44,8 +45,7 @@ pub struct SegmentWriter {
     fast_field_writers: FastFieldsWriter,
     fieldnorms_writer: FieldNormsWriter,
     doc_opstamps: Vec<Opstamp>,
-    // TODO: change type
-    tokenizers: Vec<Option<Box<dyn TextAnalyzerT>>>,
+    tokenizers: Vec<Option<TextAnalyzer>>,
     term_buffer: Term,
 }
 
@@ -70,17 +70,17 @@ impl SegmentWriter {
         let multifield_postings = MultiFieldPostingsWriter::new(schema, table_num_bits);
         let tokenizers = schema
             .fields()
-            .map(|(_, field_entry)| match field_entry.field_type() {
-                FieldType::Str(text_options) => {
-                    text_options
+            .map(
+                |(_, field_entry): (Field, &FieldEntry)| match field_entry.field_type() {
+                    FieldType::Str(ref text_options) => text_options
                         .get_indexing_options()
                         .and_then(|text_index_option| {
                             let tokenizer_name = &text_index_option.tokenizer();
                             tokenizer_manager.get(tokenizer_name)
-                        })
-                }
-                _ => None,
-            })
+                        }),
+                    _ => None,
+                },
+            )
             .collect();
         Ok(SegmentWriter {
             max_doc: 0,
@@ -141,13 +141,13 @@ impl SegmentWriter {
             }
             let (term_buffer, multifield_postings) =
                 (&mut self.term_buffer, &mut self.multifield_postings);
-            match field_entry.field_type() {
+            match *field_entry.field_type() {
                 FieldType::HierarchicalFacet => {
                     term_buffer.set_field(field);
                     let facets =
                         field_values
                             .iter()
-                            .flat_map(|field_value| match field_value.value() {
+                            .flat_map(|field_value| match *field_value.value() {
                                 Value::Facet(ref facet) => Some(facet.encoded_str()),
                                 _ => {
                                     panic!("Expected hierarchical facet");
@@ -157,13 +157,12 @@ impl SegmentWriter {
                         let mut unordered_term_id_opt = None;
                         FacetTokenizer
                             .token_stream(facet_str)
-                            .map(|token| {
+                            .process(&mut |token| {
                                 term_buffer.set_text(&token.text);
                                 let unordered_term_id =
                                     multifield_postings.subscribe(doc_id, &term_buffer);
                                 unordered_term_id_opt = Some(unordered_term_id);
-                            })
-                            .count();
+                            });
                         if let Some(unordered_term_id) = unordered_term_id_opt {
                             self.fast_field_writers
                                 .get_multivalue_writer(field)
@@ -173,38 +172,37 @@ impl SegmentWriter {
                     }
                 }
                 FieldType::Str(_) => {
-                    let mut streams_with_offsets = vec![];
+                    let mut token_streams: Vec<BoxTokenStream> = vec![];
+                    let mut offsets = vec![];
                     let mut total_offset = 0;
 
                     for field_value in field_values {
                         match field_value.value() {
                             Value::PreTokStr(tok_str) => {
-                                streams_with_offsets.push((
-                                    Box::new(PreTokenizedStream::from(tok_str.clone()))
-                                        as Box<dyn Iterator<Item = Token>>,
-                                    total_offset,
-                                ));
+                                offsets.push(total_offset);
                                 if let Some(last_token) = tok_str.tokens.last() {
                                     total_offset += last_token.offset_to;
                                 }
+                                token_streams
+                                    .push(PreTokenizedStream::from(tok_str.clone()).into());
                             }
-                            Value::Str(text) => {
+                            Value::Str(ref text) => {
                                 if let Some(ref mut tokenizer) =
                                     self.tokenizers[field.field_id() as usize]
                                 {
-                                    streams_with_offsets
-                                        .push((tokenizer.token_stream(text), total_offset));
+                                    offsets.push(total_offset);
                                     total_offset += text.len();
+                                    token_streams.push(tokenizer.token_stream(text));
                                 }
                             }
                             _ => (),
                         }
                     }
 
-                    let num_tokens = if streams_with_offsets.is_empty() {
+                    let num_tokens = if token_streams.is_empty() {
                         0
                     } else {
-                        let mut token_stream = DynTokenStreamChain::from_vec(streams_with_offsets);
+                        let mut token_stream = TokenStreamChain::new(offsets, token_streams);
                         multifield_postings.index_text(
                             doc_id,
                             field,
@@ -215,62 +213,71 @@ impl SegmentWriter {
 
                     self.fieldnorms_writer.record(doc_id, field, num_tokens);
                 }
-                FieldType::U64(int_option) if int_option.is_indexed() => {
-                    for field_value in field_values {
-                        term_buffer.set_field(field_value.field());
-                        let u64_val = field_value
-                            .value()
-                            .u64_value()
-                            .ok_or_else(make_schema_error)?;
-                        term_buffer.set_u64(u64_val);
-                        multifield_postings.subscribe(doc_id, &term_buffer);
+                FieldType::U64(ref int_option) => {
+                    if int_option.is_indexed() {
+                        for field_value in field_values {
+                            term_buffer.set_field(field_value.field());
+                            let u64_val = field_value
+                                .value()
+                                .u64_value()
+                                .ok_or_else(make_schema_error)?;
+                            term_buffer.set_u64(u64_val);
+                            multifield_postings.subscribe(doc_id, &term_buffer);
+                        }
                     }
                 }
-                FieldType::Date(int_option) if int_option.is_indexed() => {
-                    for field_value in field_values {
-                        term_buffer.set_field(field_value.field());
-                        let date_val = field_value
-                            .value()
-                            .date_value()
-                            .ok_or_else(make_schema_error)?;
-                        term_buffer.set_i64(date_val.timestamp());
-                        multifield_postings.subscribe(doc_id, &term_buffer);
+                FieldType::Date(ref int_option) => {
+                    if int_option.is_indexed() {
+                        for field_value in field_values {
+                            term_buffer.set_field(field_value.field());
+                            let date_val = field_value
+                                .value()
+                                .date_value()
+                                .ok_or_else(make_schema_error)?;
+                            term_buffer.set_i64(date_val.timestamp());
+                            multifield_postings.subscribe(doc_id, &term_buffer);
+                        }
                     }
                 }
-                FieldType::I64(int_option) if int_option.is_indexed() => {
-                    for field_value in field_values {
-                        term_buffer.set_field(field_value.field());
-                        let i64_val = field_value
-                            .value()
-                            .i64_value()
-                            .ok_or_else(make_schema_error)?;
-                        term_buffer.set_i64(i64_val);
-                        multifield_postings.subscribe(doc_id, &term_buffer);
+                FieldType::I64(ref int_option) => {
+                    if int_option.is_indexed() {
+                        for field_value in field_values {
+                            term_buffer.set_field(field_value.field());
+                            let i64_val = field_value
+                                .value()
+                                .i64_value()
+                                .ok_or_else(make_schema_error)?;
+                            term_buffer.set_i64(i64_val);
+                            multifield_postings.subscribe(doc_id, &term_buffer);
+                        }
                     }
                 }
-                FieldType::F64(int_option) if int_option.is_indexed() => {
-                    for field_value in field_values {
-                        term_buffer.set_field(field_value.field());
-                        let f64_val = field_value
-                            .value()
-                            .f64_value()
-                            .ok_or_else(make_schema_error)?;
-                        term_buffer.set_f64(f64_val);
-                        multifield_postings.subscribe(doc_id, &term_buffer);
+                FieldType::F64(ref int_option) => {
+                    if int_option.is_indexed() {
+                        for field_value in field_values {
+                            term_buffer.set_field(field_value.field());
+                            let f64_val = field_value
+                                .value()
+                                .f64_value()
+                                .ok_or_else(make_schema_error)?;
+                            term_buffer.set_f64(f64_val);
+                            multifield_postings.subscribe(doc_id, &term_buffer);
+                        }
                     }
                 }
-                FieldType::Bytes(option) if option.is_indexed() => {
-                    for field_value in field_values {
-                        term_buffer.set_field(field_value.field());
-                        let bytes = field_value
-                            .value()
-                            .bytes_value()
-                            .ok_or_else(make_schema_error)?;
-                        term_buffer.set_bytes(bytes);
-                        self.multifield_postings.subscribe(doc_id, &term_buffer);
+                FieldType::Bytes(ref option) => {
+                    if option.is_indexed() {
+                        for field_value in field_values {
+                            term_buffer.set_field(field_value.field());
+                            let bytes = field_value
+                                .value()
+                                .bytes_value()
+                                .ok_or_else(make_schema_error)?;
+                            term_buffer.set_bytes(bytes);
+                            self.multifield_postings.subscribe(doc_id, &term_buffer);
+                        }
                     }
                 }
-                _ => {}
             }
         }
         doc.filter_fields(|field| schema.get_field_entry(field).is_stored());

src/positions/reader.rs

@@ -132,7 +132,7 @@ impl PositionReader {
             "offset arguments should be increasing."
         );
         let delta_to_block_offset = offset as i64 - self.block_offset as i64;
-        if !(0..128).contains(&delta_to_block_offset) {
+        if delta_to_block_offset < 0 || delta_to_block_offset >= 128 {
             // The first position is not within the first block.
             // We need to decompress the first block.
             let delta_to_anchor_offset = offset - self.anchor_offset;

src/postings/block_search.rs

@@ -109,9 +109,9 @@ impl BlockSearcher {
     /// The results should be equivalent to
     /// ```compile_fail
     /// block[..]
-    ///       .iter()
-    ///       .take_while(|&&val| val < target)
-    ///       .count()
+    //       .iter()
+    //       .take_while(|&&val| val < target)
+    //       .count()
     /// ```
     ///
     /// The `start` argument is just used to hint that the response is

src/postings/postings_writer.rs

@@ -9,6 +9,7 @@ use crate::postings::{FieldSerializer, InvertedIndexSerializer};
 use crate::schema::IndexRecordOption;
 use crate::schema::{Field, FieldEntry, FieldType, Schema, Term};
 use crate::termdict::TermOrdinal;
+use crate::tokenizer::TokenStream;
 use crate::tokenizer::{Token, MAX_TOKEN_LEN};
 use crate::DocId;
 use fnv::FnvHashMap;
@@ -99,10 +100,12 @@ impl MultiFieldPostingsWriter {
         &mut self,
         doc: DocId,
         field: Field,
-        token_stream: &mut dyn Iterator<Item = Token>,
+        token_stream: &mut dyn TokenStream,
         term_buffer: &mut Term,
     ) -> u32 {
-        self.per_field_postings_writers[field.field_id() as usize].index_text(
+        let postings_writer =
+            self.per_field_postings_writers[field.field_id() as usize].deref_mut();
+        postings_writer.index_text(
             &mut self.term_index,
             doc,
             field,
@@ -214,7 +217,7 @@ pub trait PostingsWriter {
         term_index: &mut TermHashMap,
         doc_id: DocId,
         field: Field,
-        token_stream: &mut dyn Iterator<Item = Token>,
+        token_stream: &mut dyn TokenStream,
         heap: &mut MemoryArena,
         term_buffer: &mut Term,
     ) -> u32 {
@@ -239,7 +242,7 @@ pub trait PostingsWriter {
                 );
             }
         };
-        token_stream.map(|tok| sink(&tok)).count() as u32
+        token_stream.process(&mut sink)
     }
 
     fn total_num_tokens(&self) -> u64;

src/query/query_parser/query_parser.rs

@@ -289,7 +289,7 @@ impl QueryParser {
             let field_name = field_entry.name().to_string();
             return Err(QueryParserError::FieldNotIndexed(field_name));
         }
-        match field_type {
+        match *field_type {
             FieldType::I64(_) => {
                 let val: i64 = i64::from_str(phrase)?;
                 let term = Term::from_field_i64(field, val);
@@ -312,7 +312,7 @@ impl QueryParser {
                 let term = Term::from_field_u64(field, val);
                 Ok(vec![(0, term)])
             }
-            FieldType::Str(str_options) => {
+            FieldType::Str(ref str_options) => {
                 if let Some(option) = str_options.get_indexing_options() {
                     let tokenizer =
                         self.tokenizer_manager
@@ -323,14 +323,15 @@ impl QueryParser {
                                     option.tokenizer().to_string(),
                                 )
                             })?;
-                    let token_stream = tokenizer.token_stream(phrase);
-                    let terms: Vec<_> = token_stream
-                        .map(|token| {
-                            let term = Term::from_field_text(field, &token.text);
-                            (token.position, term)
-                        })
-                        .collect();
-                    if terms.len() <= 1 {
+                    let mut terms: Vec<(usize, Term)> = Vec::new();
+                    let mut token_stream = tokenizer.token_stream(phrase);
+                    token_stream.process(&mut |token| {
+                        let term = Term::from_field_text(field, &token.text);
+                        terms.push((token.position, term));
+                    });
+                    if terms.is_empty() {
+                        Ok(vec![])
+                    } else if terms.len() == 1 {
                         Ok(terms)
                     } else {
                         let field_entry = self.schema.get_field_entry(field);
@@ -413,7 +414,7 @@ impl QueryParser {
         &self,
         given_field: &Option<String>,
     ) -> Result<Cow<'_, [Field]>, QueryParserError> {
-        match given_field {
+        match *given_field {
             None => {
                 if self.default_fields.is_empty() {
                     Err(QueryParserError::NoDefaultFieldDeclared)
@@ -421,7 +422,7 @@ impl QueryParser {
                     Ok(Cow::from(&self.default_fields[..]))
                 }
             }
-            Some(field) => Ok(Cow::from(vec![self.resolve_field_name(&*field)?])),
+            Some(ref field) => Ok(Cow::from(vec![self.resolve_field_name(&*field)?])),
         }
     }
 
@@ -573,12 +574,15 @@ fn convert_to_query(logical_ast: LogicalAST) -> Box<dyn Query> {
 #[cfg(test)]
 mod test {
     use super::super::logical_ast::*;
-    use super::*;
+    use super::QueryParser;
+    use super::QueryParserError;
     use crate::query::Query;
     use crate::schema::Field;
     use crate::schema::{IndexRecordOption, TextFieldIndexing, TextOptions};
     use crate::schema::{Schema, Term, INDEXED, STORED, STRING, TEXT};
-    use crate::tokenizer::{analyzer_builder, LowerCaser, SimpleTokenizer, StopWordFilter};
+    use crate::tokenizer::{
+        LowerCaser, SimpleTokenizer, StopWordFilter, TextAnalyzer, TokenizerManager,
+    };
     use crate::Index;
     use matches::assert_matches;
 
@@ -616,10 +620,9 @@ mod test {
         let tokenizer_manager = TokenizerManager::default();
         tokenizer_manager.register(
             "en_with_stop_words",
-            analyzer_builder(SimpleTokenizer)
-                .filter(LowerCaser::new())
-                .filter(StopWordFilter::remove(vec!["the".to_string()]))
-                .build(),
+            TextAnalyzer::from(SimpleTokenizer)
+                .filter(LowerCaser)
+                .filter(StopWordFilter::remove(vec!["the".to_string()])),
         );
         QueryParser::new(schema, default_fields, tokenizer_manager)
     }

src/snippet/mod.rs

@@ -1,7 +1,7 @@
 use crate::query::Query;
 use crate::schema::Field;
 use crate::schema::Value;
-use crate::tokenizer::{TextAnalyzerT, Token};
+use crate::tokenizer::{TextAnalyzer, Token};
 use crate::Searcher;
 use crate::{Document, Score};
 use htmlescape::encode_minimal;
@@ -139,9 +139,9 @@ impl Snippet {
 ///
 /// Fragments must be valid in the sense that `&text[fragment.start..fragment.stop]`\
 /// has to be a valid string.
-fn search_fragments(
-    tokenizer: &dyn TextAnalyzerT,
-    text: &str,
+fn search_fragments<'a>(
+    tokenizer: &TextAnalyzer,
+    text: &'a str,
     terms: &BTreeMap<String, Score>,
     max_num_chars: usize,
 ) -> Vec<FragmentCandidate> {
@@ -155,7 +155,7 @@ fn search_fragments(
             };
             fragment = FragmentCandidate::new(next.offset_from);
         }
-        fragment.try_add_token(&next, &terms);
+        fragment.try_add_token(next, &terms);
     }
     if fragment.score > 0.0 {
         fragments.push(fragment)
@@ -249,7 +249,7 @@ fn select_best_fragment_combination(fragments: &[FragmentCandidate], text: &str)
 /// ```
 pub struct SnippetGenerator {
     terms_text: BTreeMap<String, Score>,
-    tokenizer: Box<dyn TextAnalyzerT>,
+    tokenizer: TextAnalyzer,
     field: Field,
     max_num_chars: usize,
 }
@@ -297,37 +297,33 @@ impl SnippetGenerator {
     ///
     /// This method extract the text associated to the `SnippetGenerator`'s field
     /// and computes a snippet.
-    pub fn snippet_from_doc(&mut self, doc: &Document) -> Snippet {
+    pub fn snippet_from_doc(&self, doc: &Document) -> Snippet {
         let text: String = doc
             .get_all(self.field)
             .flat_map(Value::text)
             .collect::<Vec<&str>>()
             .join(" ");
-        self.snippet(text.as_ref())
+        self.snippet(&text)
     }
 
     /// Generates a snippet for the given text.
-    pub fn snippet(&mut self, text: &str) -> Snippet {
-        let fragment_candidates = search_fragments(
-            &mut *self.tokenizer,
-            text,
-            &self.terms_text,
-            self.max_num_chars,
-        );
-        select_best_fragment_combination(&fragment_candidates[..], text)
+    pub fn snippet(&self, text: &str) -> Snippet {
+        let fragment_candidates =
+            search_fragments(&self.tokenizer, &text, &self.terms_text, self.max_num_chars);
+        select_best_fragment_combination(&fragment_candidates[..], &text)
     }
 }
 
 #[cfg(test)]
 mod tests {
-    use super::*;
+    use super::{search_fragments, select_best_fragment_combination};
     use crate::query::QueryParser;
     use crate::schema::{IndexRecordOption, Schema, TextFieldIndexing, TextOptions, TEXT};
     use crate::tokenizer::SimpleTokenizer;
-    use crate::tokenizer::TextAnalyzer;
     use crate::Index;
     use crate::SnippetGenerator;
     use maplit::btreemap;
+    use std::collections::BTreeMap;
     use std::iter::Iterator;
 
     const TEST_TEXT: &'static str = r#"Rust is a systems programming language sponsored by
@@ -350,13 +346,7 @@ Survey in 2016, 2017, and 2018."#;
             String::from("rust") => 1.0,
             String::from("language") => 0.9
         };
-
-        let fragments = search_fragments(
-            &Into::<TextAnalyzer<_>>::into(SimpleTokenizer),
-            TEST_TEXT,
-            &terms,
-            100,
-        );
+        let fragments = search_fragments(&From::from(SimpleTokenizer), TEST_TEXT, &terms, 100);
         assert_eq!(fragments.len(), 7);
         {
             let first = &fragments[0];
@@ -383,12 +373,7 @@ Survey in 2016, 2017, and 2018."#;
                 String::from("rust") =>1.0,
                 String::from("language") => 0.9
             };
-            let fragments = search_fragments(
-                &Into::<TextAnalyzer<_>>::into(SimpleTokenizer),
-                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);
@@ -402,12 +387,7 @@ Survey in 2016, 2017, and 2018."#;
                 String::from("rust") =>0.9,
                 String::from("language") => 1.0
             };
-            let fragments = search_fragments(
-                &Into::<TextAnalyzer<_>>::into(SimpleTokenizer),
-                TEST_TEXT,
-                &terms,
-                20,
-            );
+            let fragments = search_fragments(&From::from(SimpleTokenizer), TEST_TEXT, &terms, 20);
             //assert_eq!(fragments.len(), 7);
             {
                 let first = &fragments[0];
@@ -426,12 +406,7 @@ Survey in 2016, 2017, and 2018."#;
         let mut terms = BTreeMap::new();
         terms.insert(String::from("c"), 1.0);
 
-        let fragments = search_fragments(
-            &Into::<TextAnalyzer<_>>::into(SimpleTokenizer),
-            &text,
-            &terms,
-            3,
-        );
+        let fragments = search_fragments(&From::from(SimpleTokenizer), &text, &terms, 3);
 
         assert_eq!(fragments.len(), 1);
         {
@@ -453,12 +428,7 @@ Survey in 2016, 2017, and 2018."#;
         let mut terms = BTreeMap::new();
         terms.insert(String::from("f"), 1.0);
 
-        let fragments = search_fragments(
-            &Into::<TextAnalyzer<_>>::into(SimpleTokenizer),
-            &text,
-            &terms,
-            3,
-        );
+        let fragments = search_fragments(&From::from(SimpleTokenizer), &text, &terms, 3);
 
         assert_eq!(fragments.len(), 2);
         {
@@ -481,12 +451,7 @@ Survey in 2016, 2017, and 2018."#;
         terms.insert(String::from("f"), 1.0);
         terms.insert(String::from("a"), 0.9);
 
-        let fragments = search_fragments(
-            &Into::<TextAnalyzer<_>>::into(SimpleTokenizer),
-            &text,
-            &terms,
-            7,
-        );
+        let fragments = search_fragments(&From::from(SimpleTokenizer), &text, &terms, 7);
 
         assert_eq!(fragments.len(), 2);
         {
@@ -508,12 +473,7 @@ Survey in 2016, 2017, and 2018."#;
         let mut terms = BTreeMap::new();
         terms.insert(String::from("z"), 1.0);
 
-        let fragments = search_fragments(
-            &Into::<TextAnalyzer<_>>::into(SimpleTokenizer),
-            &text,
-            &terms,
-            3,
-        );
+        let fragments = search_fragments(&From::from(SimpleTokenizer), &text, &terms, 3);
 
         assert_eq!(fragments.len(), 0);
 
@@ -527,12 +487,7 @@ Survey in 2016, 2017, and 2018."#;
         let text = "a b c d";
 
         let terms = BTreeMap::new();
-        let fragments = search_fragments(
-            &Into::<TextAnalyzer<_>>::into(SimpleTokenizer),
-            &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);
@@ -617,12 +572,12 @@ Survey in 2016, 2017, and 2018."#;
         let mut snippet_generator =
             SnippetGenerator::create(&searcher, &*query, text_field).unwrap();
         {
-            let snippet = snippet_generator.snippet(TEST_TEXT.into());
+            let snippet = snippet_generator.snippet(TEST_TEXT);
             assert_eq!(snippet.to_html(), "imperative-procedural paradigms. <b>Rust</b> is syntactically similar to C++[according to whom?],\nbut its <b>designers</b> intend it to provide better memory safety");
         }
         {
             snippet_generator.set_max_num_chars(90);
-            let snippet = snippet_generator.snippet(TEST_TEXT.into());
+            let snippet = snippet_generator.snippet(TEST_TEXT);
             assert_eq!(snippet.to_html(), "<b>Rust</b> is syntactically similar to C++[according to whom?],\nbut its <b>designers</b> intend it to");
         }
     }

src/store/index/block.rs

@@ -43,6 +43,9 @@ impl CheckpointBlock {
 
     /// Adding another checkpoint in the block.
     pub fn push(&mut self, checkpoint: Checkpoint) {
+        if let Some(prev_checkpoint) = self.checkpoints.last() {
+            assert!(checkpoint.follows(prev_checkpoint));
+        }
         self.checkpoints.push(checkpoint);
     }
 

src/store/index/mod.rs

@@ -1,4 +1,4 @@
-const CHECKPOINT_PERIOD: usize = 8;
+const CHECKPOINT_PERIOD: usize = 2;
 
 use std::fmt;
 mod block;
@@ -26,6 +26,13 @@ pub struct Checkpoint {
     pub end_offset: u64,
 }
 
+impl Checkpoint {
+    pub(crate) fn follows(&self, other: &Checkpoint) -> bool {
+        (self.start_doc == other.end_doc) &&
+        (self.start_offset == other.end_offset)
+    }
+}
+
 impl fmt::Debug for Checkpoint {
     fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
         write!(
@@ -39,13 +46,16 @@ impl fmt::Debug for Checkpoint {
 #[cfg(test)]
 mod tests {
 
-    use std::io;
+    use std::{io, iter};
 
+    use futures::executor::block_on;
     use proptest::strategy::{BoxedStrategy, Strategy};
 
     use crate::directory::OwnedBytes;
+    use crate::indexer::NoMergePolicy;
+    use crate::schema::{SchemaBuilder, STORED, STRING};
     use crate::store::index::Checkpoint;
-    use crate::DocId;
+    use crate::{DocAddress, DocId, Index, Term};
 
     use super::{SkipIndex, SkipIndexBuilder};
 
@@ -54,7 +64,7 @@ mod tests {
         let mut output: Vec<u8> = Vec::new();
         let skip_index_builder: SkipIndexBuilder = SkipIndexBuilder::new();
         skip_index_builder.write(&mut output)?;
-        let skip_index: SkipIndex = SkipIndex::from(OwnedBytes::new(output));
+        let skip_index: SkipIndex = SkipIndex::open(OwnedBytes::new(output));
         let mut skip_cursor = skip_index.checkpoints();
         assert!(skip_cursor.next().is_none());
         Ok(())
@@ -72,7 +82,7 @@ mod tests {
         };
         skip_index_builder.insert(checkpoint);
         skip_index_builder.write(&mut output)?;
-        let skip_index: SkipIndex = SkipIndex::from(OwnedBytes::new(output));
+        let skip_index: SkipIndex = SkipIndex::open(OwnedBytes::new(output));
         let mut skip_cursor = skip_index.checkpoints();
         assert_eq!(skip_cursor.next(), Some(checkpoint));
         assert_eq!(skip_cursor.next(), None);
@@ -121,7 +131,7 @@ mod tests {
         }
         skip_index_builder.write(&mut output)?;
 
-        let skip_index: SkipIndex = SkipIndex::from(OwnedBytes::new(output));
+        let skip_index: SkipIndex = SkipIndex::open(OwnedBytes::new(output));
         assert_eq!(
             &skip_index.checkpoints().collect::<Vec<_>>()[..],
             &checkpoints[..]
@@ -133,6 +143,40 @@ mod tests {
         (doc as u64) * (doc as u64)
     }
 
+    #[test]
+    fn test_merge_store_with_stacking_reproducing_issue969() -> crate::Result<()> {
+        let mut schema_builder = SchemaBuilder::default();
+        let text = schema_builder.add_text_field("text", STORED | STRING);
+        let body = schema_builder.add_text_field("body", STORED);
+        let schema = schema_builder.build();
+        let index = Index::create_in_ram(schema);
+        let mut index_writer = index.writer_for_tests()?;
+        index_writer.set_merge_policy(Box::new(NoMergePolicy));
+        let long_text: String = iter::repeat("abcdefghijklmnopqrstuvwxyz")
+            .take(1_000)
+            .collect();
+        for _ in 0..20 {
+            index_writer.add_document(doc!(body=>long_text.clone()));
+        }
+        index_writer.commit()?;
+        index_writer.add_document(doc!(text=>"testb"));
+        for _ in 0..10 {
+            index_writer.add_document(doc!(text=>"testd", body=>long_text.clone()));
+        }
+        index_writer.commit()?;
+        index_writer.delete_term(Term::from_field_text(text, "testb"));
+        index_writer.commit()?;
+        let segment_ids = index.searchable_segment_ids()?;
+        block_on(index_writer.merge(&segment_ids))?;
+        let reader = index.reader()?;
+        let searcher = reader.searcher();
+        assert_eq!(searcher.num_docs(), 30);
+        for i in 0..searcher.num_docs() as u32 {
+            let _doc = searcher.doc(DocAddress(0u32, i))?;
+        }
+        Ok(())
+    }
+
     #[test]
     fn test_skip_index_long() -> io::Result<()> {
         let mut output: Vec<u8> = Vec::new();
@@ -150,7 +194,7 @@ mod tests {
         }
         skip_index_builder.write(&mut output)?;
         assert_eq!(output.len(), 4035);
-        let resulting_checkpoints: Vec<Checkpoint> = SkipIndex::from(OwnedBytes::new(output))
+        let resulting_checkpoints: Vec<Checkpoint> = SkipIndex::open(OwnedBytes::new(output))
             .checkpoints()
             .collect();
         assert_eq!(&resulting_checkpoints, &checkpoints);
@@ -221,7 +265,7 @@ mod tests {
              }
              let mut buffer = Vec::new();
              skip_index_builder.write(&mut buffer).unwrap();
-             let skip_index = SkipIndex::from(OwnedBytes::new(buffer));
+             let skip_index = SkipIndex::open(OwnedBytes::new(buffer));
              let iter_checkpoints: Vec<Checkpoint> = skip_index.checkpoints().collect();
              assert_eq!(&checkpoints[..], &iter_checkpoints[..]);
              test_skip_index_aux(skip_index, &checkpoints[..]);

src/store/index/skip_index.rs

@@ -35,11 +35,11 @@ struct Layer {
 }
 
 impl Layer {
-    fn cursor(&self) -> impl Iterator<Item = Checkpoint> + '_ {
+    fn cursor<'a>(&'a self) -> impl Iterator<Item = Checkpoint> + 'a {
         self.cursor_at_offset(0u64)
     }
 
-    fn cursor_at_offset(&self, start_offset: u64) -> impl Iterator<Item = Checkpoint> + '_ {
+    fn cursor_at_offset<'a>(&'a self, start_offset: u64) -> impl Iterator<Item = Checkpoint> + 'a {
         let data = &self.data.as_slice();
         LayerCursor {
             remaining: &data[start_offset as usize..],
@@ -59,7 +59,25 @@ pub struct SkipIndex {
 }
 
 impl SkipIndex {
-    pub(crate) fn checkpoints(&self) -> impl Iterator<Item = Checkpoint> + '_ {
+    pub fn open(mut data: OwnedBytes) -> SkipIndex {
+        let offsets: Vec<u64> = Vec::<VInt>::deserialize(&mut data)
+            .unwrap()
+            .into_iter()
+            .map(|el| el.0)
+            .collect();
+        let mut start_offset = 0;
+        let mut layers = Vec::new();
+        for end_offset in offsets {
+            let layer = Layer {
+                data: data.slice(start_offset as usize, end_offset as usize),
+            };
+            layers.push(layer);
+            start_offset = end_offset;
+        }
+        SkipIndex { layers }
+    }
+
+    pub(crate) fn checkpoints<'a>(&'a self) -> impl Iterator<Item = Checkpoint> + 'a {
         self.layers
             .last()
             .into_iter()
@@ -90,22 +108,3 @@ impl SkipIndex {
         Some(cur_checkpoint)
     }
 }
-
-impl From<OwnedBytes> for SkipIndex {
-    fn from(mut data: OwnedBytes) -> SkipIndex {
-        let offsets: Vec<u64> = Vec::<VInt>::deserialize(&mut data)
-            .unwrap()
-            .into_iter()
-            .map(|el| el.0)
-            .collect();
-        let mut start_offset = 0;
-        let mut layers = Vec::new();
-        for end_offset in offsets {
-            layers.push(Layer {
-                data: data.slice(start_offset as usize, end_offset as usize),
-            });
-            start_offset = end_offset;
-        }
-        SkipIndex { layers }
-    }
-}

src/store/index/skip_index_builder.rs

@@ -28,18 +28,20 @@ impl LayerBuilder {
     ///
     /// If the block was empty to begin with, simply return None.
     fn flush_block(&mut self) -> Option<Checkpoint> {
-        self.block.doc_interval().map(|(start_doc, end_doc)| {
+        if let Some((start_doc, end_doc)) = self.block.doc_interval() {
             let start_offset = self.buffer.len() as u64;
             self.block.serialize(&mut self.buffer);
             let end_offset = self.buffer.len() as u64;
             self.block.clear();
-            Checkpoint {
+            Some(Checkpoint {
                 start_doc,
                 end_doc,
                 start_offset,
                 end_offset,
-            }
-        })
+            })
+        } else {
+            None
+        }
     }
 
     fn push(&mut self, checkpoint: Checkpoint) {
@@ -48,7 +50,7 @@ impl LayerBuilder {
 
     fn insert(&mut self, checkpoint: Checkpoint) -> Option<Checkpoint> {
         self.push(checkpoint);
-        let emit_skip_info = (self.block.len() % CHECKPOINT_PERIOD) == 0;
+        let emit_skip_info = self.block.len() >= CHECKPOINT_PERIOD;
         if emit_skip_info {
             self.flush_block()
         } else {

src/store/reader.rs

@@ -35,7 +35,7 @@ impl StoreReader {
         let (data_file, offset_index_file) = split_file(store_file)?;
         let index_data = offset_index_file.read_bytes()?;
         let space_usage = StoreSpaceUsage::new(data_file.len(), offset_index_file.len());
-        let skip_index = SkipIndex::from(index_data);
+        let skip_index = SkipIndex::open(index_data);
         Ok(StoreReader {
             data: data_file,
             cache: Arc::new(Mutex::new(LruCache::new(LRU_CACHE_CAPACITY))),
@@ -46,7 +46,7 @@ impl StoreReader {
         })
     }
 
-    pub(crate) fn block_checkpoints(&self) -> impl Iterator<Item = Checkpoint> + '_ {
+    pub(crate) fn block_checkpoints<'a>(&'a self) -> impl Iterator<Item = Checkpoint> + 'a {
         self.skip_index.checkpoints()
     }
 

src/store/tests_store.rs

@@ -0,0 +1,50 @@
+use std::path::Path;
+
+use crate::HasLen;
+use crate::directory::{Directory, ManagedDirectory, MmapDirectory, RAMDirectory};
+use crate::fastfield::DeleteBitSet;
+
+use super::{StoreReader, StoreWriter};
+
+#[test]
+fn test_toto2() -> crate::Result<()> {
+    let directory = ManagedDirectory::wrap(MmapDirectory::open("src/store/broken_seg")?)?;
+    let path = Path::new("b6029ade1b954ea1acad15b432eaacb9.store");
+    assert!(directory.validate_checksum(path)?);
+    let store_file = directory.open_read(path)?;
+    let store = StoreReader::open(store_file)?;
+    let documents = store.documents();
+    // for doc in documents {
+    //     println!("{:?}", doc);
+    // }
+    let doc= store.get(15_086)?;
+    Ok(())
+}
+
+#[test]
+fn test_toto() -> crate::Result<()> {
+    let directory = ManagedDirectory::wrap(MmapDirectory::open("src/store/broken_seg")?)?;
+    assert!(directory.validate_checksum(Path::new("e6ece22e5bca4e0dbe7ce3e4dcbd5bbf.store"))?);
+    let store_file = directory.open_read(Path::new("e6ece22e5bca4e0dbe7ce3e4dcbd5bbf.store.patched"))?;
+    let store = StoreReader::open(store_file)?;
+    let doc= store.get(53)?;
+    println!("{:?}", doc);
+    // let documents = store.documents();
+    // let ram_directory = RAMDirectory::create();
+    // let path = Path::new("store");
+
+    // let store_wrt = ram_directory.open_write(path)?;
+    // let mut store_writer = StoreWriter::new(store_wrt);
+    // for doc in &documents {
+    //     store_writer.store(doc)?;
+    // }
+    // store_writer.close()?;
+    // let store_data = ram_directory.open_read(path)?;
+    // let new_store = StoreReader::open(store_data)?;
+    // for doc in 0..59 {
+    //     println!("{}", doc);
+    //     let doc = new_store.get(doc)?;
+    //     println!("{:?}", doc);
+    // }
+    Ok(())
+}

src/store/writer.rs

@@ -10,7 +10,7 @@ use crate::store::index::Checkpoint;
 use crate::DocId;
 use std::io::{self, Write};
 
-const BLOCK_SIZE: usize = 16_384;
+const BLOCK_SIZE: usize = 30;
 
 /// Write tantivy's [`Store`](./index.html)
 ///
@@ -72,6 +72,7 @@ impl StoreWriter {
         if !self.current_block.is_empty() {
             self.write_and_compress_block()?;
         }
+        assert_eq!(self.first_doc_in_block, self.doc);
         let doc_shift = self.doc;
         let start_shift = self.writer.written_bytes() as u64;
 
@@ -86,12 +87,17 @@ impl StoreWriter {
             checkpoint.end_doc += doc_shift;
             checkpoint.start_offset += start_shift;
             checkpoint.end_offset += start_shift;
-            self.offset_index_writer.insert(checkpoint);
-            self.doc = checkpoint.end_doc;
+            self.register_checkpoint(checkpoint);
         }
         Ok(())
     }
 
+    fn register_checkpoint(&mut self, checkpoint: Checkpoint) {
+        self.offset_index_writer.insert(checkpoint);
+        self.first_doc_in_block = checkpoint.end_doc;
+        self.doc = checkpoint.end_doc;
+    }
+
     fn write_and_compress_block(&mut self) -> io::Result<()> {
         assert!(self.doc > 0);
         self.intermediary_buffer.clear();
@@ -100,14 +106,13 @@ impl StoreWriter {
         self.writer.write_all(&self.intermediary_buffer)?;
         let end_offset = self.writer.written_bytes();
         let end_doc = self.doc;
-        self.offset_index_writer.insert(Checkpoint {
+        self.register_checkpoint(Checkpoint {
             start_doc: self.first_doc_in_block,
             end_doc,
             start_offset,
             end_offset,
         });
         self.current_block.clear();
-        self.first_doc_in_block = self.doc;
         Ok(())
     }
 

src/tokenizer/alphanum_only.rs

@@ -2,16 +2,16 @@
 //! ```rust
 //! use tantivy::tokenizer::*;
 //!
-//! let tokenizer = analyzer_builder(RawTokenizer)
-//!   .filter(AlphaNumOnlyFilter).build();
+//! let tokenizer = TextAnalyzer::from(RawTokenizer)
+//!   .filter(AlphaNumOnlyFilter);
 //!
 //! let mut stream = tokenizer.token_stream("hello there");
 //! // is none because the raw filter emits one token that
 //! // contains a space
 //! assert!(stream.next().is_none());
 //!
-//! let tokenizer = analyzer_builder(SimpleTokenizer)
-//!   .filter(AlphaNumOnlyFilter).build();
+//! let tokenizer = TextAnalyzer::from(SimpleTokenizer)
+//!   .filter(AlphaNumOnlyFilter);
 //!
 //! let mut stream = tokenizer.token_stream("hello there 💣");
 //! assert!(stream.next().is_some());
@@ -19,18 +19,45 @@
 //! // the "emoji" is dropped because its not an alphanum
 //! assert!(stream.next().is_none());
 //! ```
-use super::{Token, TokenFilter};
+use super::{BoxTokenStream, Token, TokenFilter, TokenStream};
 
 /// `TokenFilter` that removes all tokens that contain non
 /// ascii alphanumeric characters.
-#[derive(Clone, Debug, Default)]
+#[derive(Clone)]
 pub struct AlphaNumOnlyFilter;
 
-impl TokenFilter for AlphaNumOnlyFilter {
-    fn transform(&mut self, token: Token) -> Option<Token> {
-        if token.text.chars().all(|c| c.is_ascii_alphanumeric()) {
-            return Some(token);
-        }
-        None
+pub struct AlphaNumOnlyFilterStream<'a> {
+    tail: BoxTokenStream<'a>,
+}
+
+impl<'a> AlphaNumOnlyFilterStream<'a> {
+    fn predicate(&self, token: &Token) -> bool {
+        token.text.chars().all(|c| c.is_ascii_alphanumeric())
+    }
+}
+
+impl TokenFilter for AlphaNumOnlyFilter {
+    fn transform<'a>(&self, token_stream: BoxTokenStream<'a>) -> BoxTokenStream<'a> {
+        BoxTokenStream::from(AlphaNumOnlyFilterStream { tail: token_stream })
+    }
+}
+
+impl<'a> TokenStream for AlphaNumOnlyFilterStream<'a> {
+    fn advance(&mut self) -> bool {
+        while self.tail.advance() {
+            if self.predicate(self.tail.token()) {
+                return true;
+            }
+        }
+
+        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

@@ -1,31 +1,45 @@
-use super::{Token, TokenFilter};
+use super::{BoxTokenStream, Token, TokenFilter, TokenStream};
 use std::mem;
 
 /// This class converts alphabetic, numeric, and symbolic Unicode characters
 /// which are not in the first 127 ASCII characters (the "Basic Latin" Unicode
 /// block) into their ASCII equivalents, if one exists.
-#[derive(Clone, Debug, Default)]
-pub struct AsciiFolding {
-    buffer: String,
-}
+#[derive(Clone)]
+pub struct AsciiFoldingFilter;
 
-impl AsciiFolding {
-    /// Construct a new `AsciiFolding` filter.
-    pub fn new() -> Self {
-        Self {
+impl TokenFilter for AsciiFoldingFilter {
+    fn transform<'a>(&self, token_stream: BoxTokenStream<'a>) -> BoxTokenStream<'a> {
+        From::from(AsciiFoldingFilterTokenStream {
+            tail: token_stream,
             buffer: String::with_capacity(100),
-        }
+        })
     }
 }
 
-impl TokenFilter for AsciiFolding {
-    fn transform(&mut self, mut token: Token) -> Option<Token> {
-        if !token.text.is_ascii() {
-            // ignore its already ascii
-            to_ascii(&token.text, &mut self.buffer);
-            mem::swap(&mut token.text, &mut self.buffer);
+pub struct AsciiFoldingFilterTokenStream<'a> {
+    buffer: String,
+    tail: BoxTokenStream<'a>,
+}
+
+impl<'a> TokenStream for AsciiFoldingFilterTokenStream<'a> {
+    fn advance(&mut self) -> bool {
+        if !self.tail.advance() {
+            return false;
         }
-        Some(token)
+        if !self.token_mut().text.is_ascii() {
+            // ignore its already ascii
+            to_ascii(&mut self.tail.token_mut().text, &mut self.buffer);
+            mem::swap(&mut self.tail.token_mut().text, &mut self.buffer);
+        }
+        true
+    }
+
+    fn token(&self) -> &Token {
+        self.tail.token()
+    }
+
+    fn token_mut(&mut self) -> &mut Token {
+        self.tail.token_mut()
     }
 }
 
@@ -1512,7 +1526,7 @@ fn fold_non_ascii_char(c: char) -> Option<&'static str> {
 }
 
 // https://github.com/apache/lucene-solr/blob/master/lucene/analysis/common/src/java/org/apache/lucene/analysis/miscellaneous/ASCIIFoldingFilter.java#L187
-fn to_ascii(text: &String, output: &mut String) {
+fn to_ascii(text: &mut String, output: &mut String) {
     output.clear();
 
     for c in text.chars() {
@@ -1526,8 +1540,11 @@ fn to_ascii(text: &String, output: &mut String) {
 
 #[cfg(test)]
 mod tests {
-    use super::super::*;
-    use super::*;
+    use super::to_ascii;
+    use crate::tokenizer::AsciiFoldingFilter;
+    use crate::tokenizer::RawTokenizer;
+    use crate::tokenizer::SimpleTokenizer;
+    use crate::tokenizer::TextAnalyzer;
     use std::iter;
 
     #[test]
@@ -1543,22 +1560,22 @@ mod tests {
     }
 
     fn folding_helper(text: &str) -> Vec<String> {
-        let tokens = analyzer_builder(SimpleTokenizer)
-            .filter(AsciiFolding::new())
-            .build()
+        let mut tokens = Vec::new();
+        TextAnalyzer::from(SimpleTokenizer)
+            .filter(AsciiFoldingFilter)
             .token_stream(text)
-            .map(|token| token.text.clone())
-            .collect();
+            .process(&mut |token| {
+                tokens.push(token.text.clone());
+            });
         tokens
     }
 
     fn folding_using_raw_tokenizer_helper(text: &str) -> String {
-        let mut token_stream = analyzer_builder(RawTokenizer)
-            .filter(AsciiFolding::new())
-            .build()
+        let mut token_stream = TextAnalyzer::from(RawTokenizer)
+            .filter(AsciiFoldingFilter)
             .token_stream(text);
-        let Token { text, .. } = token_stream.next().unwrap();
-        text
+        token_stream.advance();
+        token_stream.token().text.clone()
     }
 
     #[test]
@@ -1609,9 +1626,9 @@ mod tests {
 
     #[test]
     fn test_to_ascii() {
-        let input = "Rámon".to_string();
+        let mut input = "Rámon".to_string();
         let mut buffer = String::new();
-        to_ascii(&input, &mut buffer);
+        to_ascii(&mut input, &mut buffer);
         assert_eq!("Ramon", buffer);
     }
 

src/tokenizer/facet_tokenizer.rs

@@ -1,4 +1,4 @@
-use super::{Token, Tokenizer};
+use super::{BoxTokenStream, Token, TokenStream, Tokenizer};
 use crate::schema::FACET_SEP_BYTE;
 
 /// The `FacetTokenizer` process a `Facet` binary representation
@@ -9,63 +9,72 @@ use crate::schema::FACET_SEP_BYTE;
 ///     - `/america/north_america/canada`
 ///     - `/america/north_america`
 ///     - `/america`
-#[derive(Clone, Debug, Default)]
+#[derive(Clone)]
 pub struct FacetTokenizer;
 
-#[derive(Clone, Debug)]
+#[derive(Debug)]
 enum State {
     RootFacetNotEmitted,
     UpToPosition(usize), //< we already emitted facet prefix up to &text[..cursor]
     Terminated,
 }
 
-#[derive(Clone, Debug)]
-pub struct FacetTokenStream {
-    text: String,
+pub struct FacetTokenStream<'a> {
+    text: &'a str,
     state: State,
     token: Token,
 }
 
 impl Tokenizer for FacetTokenizer {
-    type Iter = FacetTokenStream;
-    fn token_stream(&self, text: &str) -> Self::Iter {
+    fn token_stream<'a>(&self, text: &'a str) -> BoxTokenStream<'a> {
         FacetTokenStream {
-            text: text.to_string(),
+            text,
             state: State::RootFacetNotEmitted, //< pos is the first char that has not been processed yet.
             token: Token::default(),
         }
+        .into()
     }
 }
 
-impl Iterator for FacetTokenStream {
-    type Item = Token;
-    fn next(&mut self) -> Option<Self::Item> {
-        self.state = match self.state {
+impl<'a> TokenStream for FacetTokenStream<'a> {
+    fn advance(&mut self) -> bool {
+        match self.state {
             State::RootFacetNotEmitted => {
-                if self.text.is_empty() {
+                self.state = if self.text.is_empty() {
                     State::Terminated
                 } else {
                     State::UpToPosition(0)
-                }
+                };
+                true
             }
             State::UpToPosition(cursor) => {
-                if let Some(next_sep_pos) = self.text.as_bytes()[cursor + 1..]
+                let bytes: &[u8] = self.text.as_bytes();
+                if let Some(next_sep_pos) = bytes[cursor + 1..]
                     .iter()
-                    .position(|&b| b == FACET_SEP_BYTE)
+                    .cloned()
+                    .position(|b| b == FACET_SEP_BYTE)
                     .map(|pos| cursor + 1 + pos)
                 {
                     let facet_part = &self.text[cursor..next_sep_pos];
                     self.token.text.push_str(facet_part);
-                    State::UpToPosition(next_sep_pos)
+                    self.state = State::UpToPosition(next_sep_pos);
                 } else {
                     let facet_part = &self.text[cursor..];
                     self.token.text.push_str(facet_part);
-                    State::Terminated
+                    self.state = State::Terminated;
                 }
+                true
             }
-            State::Terminated => return None,
-        };
-        Some(self.token.clone())
+            State::Terminated => false,
+        }
+    }
+
+    fn token(&self) -> &Token {
+        &self.token
+    }
+
+    fn token_mut(&mut self) -> &mut Token {
+        &mut self.token
     }
 }
 
@@ -74,19 +83,21 @@ mod tests {
 
     use super::FacetTokenizer;
     use crate::schema::Facet;
-    use crate::tokenizer::Tokenizer;
+    use crate::tokenizer::{Token, Tokenizer};
 
     #[test]
     fn test_facet_tokenizer() {
         let facet = Facet::from_path(vec!["top", "a", "b"]);
-        let tokens: Vec<_> = FacetTokenizer
-            .token_stream(facet.encoded_str())
-            .map(|token| {
-                Facet::from_encoded(token.text.as_bytes().to_owned())
-                    .unwrap()
-                    .to_string()
-            })
-            .collect();
+        let mut tokens = vec![];
+        {
+            let mut add_token = |token: &Token| {
+                let facet = Facet::from_encoded(token.text.as_bytes().to_owned()).unwrap();
+                tokens.push(format!("{}", facet));
+            };
+            FacetTokenizer
+                .token_stream(facet.encoded_str())
+                .process(&mut add_token);
+        }
         assert_eq!(tokens.len(), 4);
         assert_eq!(tokens[0], "/");
         assert_eq!(tokens[1], "/top");
@@ -97,14 +108,16 @@ mod tests {
     #[test]
     fn test_facet_tokenizer_root_facets() {
         let facet = Facet::root();
-        let tokens: Vec<_> = FacetTokenizer
-            .token_stream(facet.encoded_str())
-            .map(|token| {
-                Facet::from_encoded(token.text.as_bytes().to_owned())
-                    .unwrap()
-                    .to_string()
-            })
-            .collect();
+        let mut tokens = vec![];
+        {
+            let mut add_token = |token: &Token| {
+                let facet = Facet::from_encoded(token.text.as_bytes().to_owned()).unwrap(); // ok test
+                tokens.push(format!("{}", facet));
+            };
+            FacetTokenizer
+                .token_stream(facet.encoded_str()) // ok test
+                .process(&mut add_token);
+        }
         assert_eq!(tokens.len(), 1);
         assert_eq!(tokens[0], "/");
     }

src/tokenizer/lower_caser.rs

@@ -1,36 +1,27 @@
-use super::{Token, TokenFilter};
+use super::{Token, TokenFilter, TokenStream};
+use crate::tokenizer::BoxTokenStream;
 use std::mem;
 
 impl TokenFilter for LowerCaser {
-    fn transform(&mut self, mut token: Token) -> Option<Token> {
-        if token.text.is_ascii() {
-            // fast track for ascii.
-            token.text.make_ascii_lowercase();
-        } else {
-            to_lowercase_unicode(&token.text, &mut self.buffer);
-            mem::swap(&mut token.text, &mut self.buffer);
-        }
-        Some(token)
+    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, Debug, Default)]
-pub struct LowerCaser {
-    buffer: String,
-}
+#[derive(Clone)]
+pub struct LowerCaser;
 
-impl LowerCaser {
-    /// Initialize the `LowerCaser`
-    pub fn new() -> Self {
-        LowerCaser {
-            buffer: String::with_capacity(100),
-        }
-    }
+pub struct LowerCaserTokenStream<'a> {
+    buffer: String,
+    tail: BoxTokenStream<'a>,
 }
 
 // writes a lowercased version of text into output.
-fn to_lowercase_unicode(text: &String, output: &mut String) {
+fn to_lowercase_unicode(text: &mut String, output: &mut String) {
     output.clear();
     for c in text.chars() {
         // Contrary to the std, we do not take care of sigma special case.
@@ -39,31 +30,57 @@ fn to_lowercase_unicode(text: &String, output: &mut String) {
     }
 }
 
+impl<'a> TokenStream for LowerCaserTokenStream<'a> {
+    fn advance(&mut self) -> bool {
+        if !self.tail.advance() {
+            return false;
+        }
+        if self.token_mut().text.is_ascii() {
+            // fast track for ascii.
+            self.token_mut().text.make_ascii_lowercase();
+        } else {
+            to_lowercase_unicode(&mut self.tail.token_mut().text, &mut self.buffer);
+            mem::swap(&mut self.tail.token_mut().text, &mut self.buffer);
+        }
+        true
+    }
+
+    fn token(&self) -> &Token {
+        self.tail.token()
+    }
+
+    fn token_mut(&mut self) -> &mut Token {
+        self.tail.token_mut()
+    }
+}
+
 #[cfg(test)]
 mod tests {
-    use super::*;
-    use crate::tokenizer::{analyzer_builder, LowerCaser, SimpleTokenizer, TextAnalyzerT};
+    use crate::tokenizer::{LowerCaser, SimpleTokenizer, TextAnalyzer};
 
     #[test]
     fn test_to_lower_case() {
-        assert_eq!(lowercase_helper("Русский текст"), vec!["русский", "текст"]);
+        assert_eq!(
+            lowercase_helper("Русский текст"),
+            vec!["русский".to_string(), "текст".to_string()]
+        );
     }
 
     fn lowercase_helper(text: &str) -> Vec<String> {
-        analyzer_builder(SimpleTokenizer)
-            .filter(LowerCaser::new())
-            .build()
-            .token_stream(text)
-            .map(|token| {
-                let Token { text, .. } = token;
-                text
-            })
-            .collect()
+        let mut tokens = vec![];
+        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);
+        }
+        tokens
     }
 
     #[test]
     fn test_lowercaser() {
-        assert_eq!(lowercase_helper("Tree"), vec!["tree"]);
-        assert_eq!(lowercase_helper("Русский"), vec!["русский"]);
+        assert_eq!(lowercase_helper("Tree"), vec!["tree".to_string()]);
+        assert_eq!(lowercase_helper("Русский"), vec!["русский".to_string()]);
     }
 }

src/tokenizer/mod.rs

@@ -64,10 +64,10 @@
 //! ```rust
 //! use tantivy::tokenizer::*;
 //!
-//! let en_stem = analyzer_builder(SimpleTokenizer)
+//! let en_stem = TextAnalyzer::from(SimpleTokenizer)
 //!     .filter(RemoveLongFilter::limit(40))
-//!     .filter(LowerCaser::new())
-//!     .filter(Stemmer::new(Language::English)).build();
+//!     .filter(LowerCaser)
+//!     .filter(Stemmer::new(Language::English));
 //! ```
 //!
 //! Once your tokenizer is defined, you need to
@@ -109,9 +109,9 @@
 //! let index = Index::create_in_ram(schema);
 //!
 //! // We need to register our tokenizer :
-//! let custom_en_tokenizer = analyzer_builder(SimpleTokenizer)
+//! let custom_en_tokenizer = TextAnalyzer::from(SimpleTokenizer)
 //!     .filter(RemoveLongFilter::limit(40))
-//!     .filter(LowerCaser::new()).build();
+//!     .filter(LowerCaser);
 //! index
 //!     .tokenizers()
 //!     .register("custom_en", custom_en_tokenizer);
@@ -133,7 +133,7 @@ mod tokenizer;
 mod tokenizer_manager;
 
 pub use self::alphanum_only::AlphaNumOnlyFilter;
-pub use self::ascii_folding_filter::AsciiFolding;
+pub use self::ascii_folding_filter::AsciiFoldingFilter;
 pub use self::facet_tokenizer::FacetTokenizer;
 pub use self::lower_caser::LowerCaser;
 pub use self::ngram_tokenizer::NgramTokenizer;
@@ -142,11 +142,11 @@ pub use self::remove_long::RemoveLongFilter;
 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::{DynTokenStreamChain, TokenStreamChain};
+pub(crate) use self::token_stream_chain::TokenStreamChain;
 
 pub use self::tokenized_string::{PreTokenizedStream, PreTokenizedString};
 pub use self::tokenizer::{
-    analyzer_builder, Identity, TextAnalyzer, TextAnalyzerT, Token, TokenFilter, Tokenizer,
+    BoxTokenFilter, BoxTokenStream, TextAnalyzer, Token, TokenFilter, TokenStream, Tokenizer,
 };
 
 pub use self::tokenizer_manager::TokenizerManager;
@@ -160,7 +160,10 @@ pub const MAX_TOKEN_LEN: usize = u16::max_value() as usize - 4;
 
 #[cfg(test)]
 pub mod tests {
-    use super::*;
+    use super::{
+        Language, LowerCaser, RemoveLongFilter, SimpleTokenizer, Stemmer, Token, TokenizerManager,
+    };
+    use crate::tokenizer::TextAnalyzer;
 
     /// This is a function that can be used in tests and doc tests
     /// to assert a token's correctness.
@@ -187,9 +190,15 @@ pub mod tests {
     fn test_raw_tokenizer() {
         let tokenizer_manager = TokenizerManager::default();
         let en_tokenizer = tokenizer_manager.get("raw").unwrap();
-        let tokens: Vec<Token> = en_tokenizer
-            .token_stream("Hello, happy tax payer!")
-            .collect();
+        let mut tokens: Vec<Token> = vec![];
+        {
+            let mut add_token = |token: &Token| {
+                tokens.push(token.clone());
+            };
+            en_tokenizer
+                .token_stream("Hello, happy tax payer!")
+                .process(&mut add_token);
+        }
         assert_eq!(tokens.len(), 1);
         assert_token(&tokens[0], 0, "Hello, happy tax payer!", 0, 23);
     }
@@ -199,9 +208,15 @@ pub mod tests {
         let tokenizer_manager = TokenizerManager::default();
         assert!(tokenizer_manager.get("en_doesnotexist").is_none());
         let en_tokenizer = tokenizer_manager.get("en_stem").unwrap();
-        let tokens: Vec<Token> = en_tokenizer
-            .token_stream("Hello, happy tax payer!")
-            .collect();
+        let mut tokens: Vec<Token> = vec![];
+        {
+            let mut add_token = |token: &Token| {
+                tokens.push(token.clone());
+            };
+            en_tokenizer
+                .token_stream("Hello, happy tax payer!")
+                .process(&mut add_token);
+        }
 
         assert_eq!(tokens.len(), 4);
         assert_token(&tokens[0], 0, "hello", 0, 5);
@@ -215,16 +230,21 @@ pub mod tests {
         let tokenizer_manager = TokenizerManager::default();
         tokenizer_manager.register(
             "el_stem",
-            analyzer_builder(SimpleTokenizer)
+            TextAnalyzer::from(SimpleTokenizer)
                 .filter(RemoveLongFilter::limit(40))
-                .filter(LowerCaser::new())
-                .filter(Stemmer::new(Language::Greek))
-                .build(),
+                .filter(LowerCaser)
+                .filter(Stemmer::new(Language::Greek)),
         );
         let en_tokenizer = tokenizer_manager.get("el_stem").unwrap();
-        let tokens: Vec<Token> = en_tokenizer
-            .token_stream("Καλημέρα, χαρούμενε φορολογούμενε!")
-            .collect();
+        let mut tokens: Vec<Token> = vec![];
+        {
+            let mut add_token = |token: &Token| {
+                tokens.push(token.clone());
+            };
+            en_tokenizer
+                .token_stream("Καλημέρα, χαρούμενε φορολογούμενε!")
+                .process(&mut add_token);
+        }
 
         assert_eq!(tokens.len(), 3);
         assert_token(&tokens[0], 0, "καλημερ", 0, 16);
@@ -236,9 +256,25 @@ pub mod tests {
     fn test_tokenizer_empty() {
         let tokenizer_manager = TokenizerManager::default();
         let en_tokenizer = tokenizer_manager.get("en_stem").unwrap();
-        let tokens: Vec<Token> = en_tokenizer.token_stream(" ").collect();
-        assert!(tokens.is_empty());
-        let tokens: Vec<Token> = en_tokenizer.token_stream(" ").collect();
-        assert!(tokens.is_empty());
+        {
+            let mut tokens: Vec<Token> = vec![];
+            {
+                let mut add_token = |token: &Token| {
+                    tokens.push(token.clone());
+                };
+                en_tokenizer.token_stream(" ").process(&mut add_token);
+            }
+            assert!(tokens.is_empty());
+        }
+        {
+            let mut tokens: Vec<Token> = vec![];
+            {
+                let mut add_token = |token: &Token| {
+                    tokens.push(token.clone());
+                };
+                en_tokenizer.token_stream(" ").process(&mut add_token);
+            }
+            assert!(tokens.is_empty());
+        }
     }
 }

src/tokenizer/ngram_tokenizer.rs

@@ -1,4 +1,5 @@
-use super::{Token, Tokenizer};
+use super::{Token, TokenStream, Tokenizer};
+use crate::tokenizer::BoxTokenStream;
 
 /// Tokenize the text by splitting words into n-grams of the given size(s)
 ///
@@ -78,7 +79,7 @@ use super::{Token, Tokenizer};
 /// }
 /// assert!(stream.next().is_none());
 /// ```
-#[derive(Clone, Debug, Default)]
+#[derive(Clone)]
 pub struct NgramTokenizer {
     /// min size of the n-gram
     min_gram: usize,
@@ -118,48 +119,54 @@ impl NgramTokenizer {
 }
 
 /// TokenStream associate to the `NgramTokenizer`
-pub struct NgramTokenStream {
+pub struct NgramTokenStream<'a> {
     /// parameters
-    ngram_charidx_iterator: StutteringIterator<CodepointFrontiers>,
+    ngram_charidx_iterator: StutteringIterator<CodepointFrontiers<'a>>,
     /// true if the NgramTokenStream is in prefix mode.
     prefix_only: bool,
     /// input
-    text: String,
+    text: &'a str,
     /// output
     token: Token,
 }
 
 impl Tokenizer for NgramTokenizer {
-    type Iter = NgramTokenStream;
-    fn token_stream(&self, text: &str) -> Self::Iter {
-        NgramTokenStream {
+    fn token_stream<'a>(&self, text: &'a str) -> BoxTokenStream<'a> {
+        From::from(NgramTokenStream {
             ngram_charidx_iterator: StutteringIterator::new(
                 CodepointFrontiers::for_str(text),
                 self.min_gram,
                 self.max_gram,
             ),
             prefix_only: self.prefix_only,
-            text: text.to_string(),
+            text,
             token: Token::default(),
-        }
+        })
     }
 }
 
-impl Iterator for NgramTokenStream {
-    type Item = Token;
-    fn next(&mut self) -> Option<Self::Item> {
+impl<'a> TokenStream for NgramTokenStream<'a> {
+    fn advance(&mut self) -> bool {
         if let Some((offset_from, offset_to)) = self.ngram_charidx_iterator.next() {
             if self.prefix_only && offset_from > 0 {
-                return None;
+                return false;
             }
             self.token.position = 0;
             self.token.offset_from = offset_from;
             self.token.offset_to = offset_to;
             self.token.text.clear();
             self.token.text.push_str(&self.text[offset_from..offset_to]);
-            return Some(self.token.clone());
-        };
-        None
+            true
+        } else {
+            false
+        }
+    }
+
+    fn token(&self) -> &Token {
+        &self.token
+    }
+    fn token_mut(&mut self) -> &mut Token {
+        &mut self.token
     }
 }
 
@@ -250,21 +257,21 @@ where
 /// or a codepoint ends.
 ///
 /// By convention, we emit [0] for the empty string.
-struct CodepointFrontiers {
-    s: String,
+struct CodepointFrontiers<'a> {
+    s: &'a str,
     next_el: Option<usize>,
 }
 
-impl CodepointFrontiers {
-    fn for_str(s: &str) -> Self {
+impl<'a> CodepointFrontiers<'a> {
+    fn for_str(s: &'a str) -> Self {
         CodepointFrontiers {
-            s: s.to_string(),
+            s,
             next_el: Some(0),
         }
     }
 }
 
-impl<'a> Iterator for CodepointFrontiers {
+impl<'a> Iterator for CodepointFrontiers<'a> {
     type Item = usize;
 
     fn next(&mut self) -> Option<usize> {
@@ -273,7 +280,7 @@ impl<'a> Iterator for CodepointFrontiers {
                 self.next_el = None;
             } else {
                 let first_codepoint_width = utf8_codepoint_width(self.s.as_bytes()[0]);
-                self.s = (&self.s[first_codepoint_width..]).to_string();
+                self.s = &self.s[first_codepoint_width..];
                 self.next_el = Some(offset + first_codepoint_width);
             }
             offset
@@ -294,8 +301,20 @@ fn utf8_codepoint_width(b: u8) -> usize {
 
 #[cfg(test)]
 mod tests {
-    use super::*;
+
+    use super::utf8_codepoint_width;
+    use super::CodepointFrontiers;
+    use super::NgramTokenizer;
+    use super::StutteringIterator;
     use crate::tokenizer::tests::assert_token;
+    use crate::tokenizer::tokenizer::Tokenizer;
+    use crate::tokenizer::{BoxTokenStream, 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
+    }
 
     #[test]
     fn test_utf8_codepoint_width() {
@@ -332,9 +351,7 @@ mod tests {
 
     #[test]
     fn test_ngram_tokenizer_1_2_false() {
-        let tokens: Vec<_> = NgramTokenizer::all_ngrams(1, 2)
-            .token_stream("hello")
-            .collect();
+        let tokens = test_helper(NgramTokenizer::all_ngrams(1, 2).token_stream("hello"));
         assert_eq!(tokens.len(), 9);
         assert_token(&tokens[0], 0, "h", 0, 1);
         assert_token(&tokens[1], 0, "he", 0, 2);
@@ -349,9 +366,7 @@ mod tests {
 
     #[test]
     fn test_ngram_tokenizer_min_max_equal() {
-        let tokens: Vec<_> = NgramTokenizer::all_ngrams(3, 3)
-            .token_stream("hello")
-            .collect();
+        let tokens = test_helper(NgramTokenizer::all_ngrams(3, 3).token_stream("hello"));
         assert_eq!(tokens.len(), 3);
         assert_token(&tokens[0], 0, "hel", 0, 3);
         assert_token(&tokens[1], 0, "ell", 1, 4);
@@ -360,9 +375,7 @@ mod tests {
 
     #[test]
     fn test_ngram_tokenizer_2_5_prefix() {
-        let tokens: Vec<_> = NgramTokenizer::prefix_only(2, 5)
-            .token_stream("frankenstein")
-            .collect();
+        let tokens = test_helper(NgramTokenizer::prefix_only(2, 5).token_stream("frankenstein"));
         assert_eq!(tokens.len(), 4);
         assert_token(&tokens[0], 0, "fr", 0, 2);
         assert_token(&tokens[1], 0, "fra", 0, 3);
@@ -372,9 +385,7 @@ mod tests {
 
     #[test]
     fn test_ngram_non_ascii_1_2() {
-        let tokens: Vec<_> = NgramTokenizer::all_ngrams(1, 2)
-            .token_stream("hεllo")
-            .collect();
+        let tokens = test_helper(NgramTokenizer::all_ngrams(1, 2).token_stream("hεllo"));
         assert_eq!(tokens.len(), 9);
         assert_token(&tokens[0], 0, "h", 0, 1);
         assert_token(&tokens[1], 0, "hε", 0, 3);
@@ -389,9 +400,7 @@ mod tests {
 
     #[test]
     fn test_ngram_non_ascii_2_5_prefix() {
-        let tokens: Vec<_> = NgramTokenizer::prefix_only(2, 5)
-            .token_stream("hεllo")
-            .collect();
+        let tokens = test_helper(NgramTokenizer::prefix_only(2, 5).token_stream("hεllo"));
         assert_eq!(tokens.len(), 4);
         assert_token(&tokens[0], 0, "hε", 0, 3);
         assert_token(&tokens[1], 0, "hεl", 0, 4);
@@ -401,16 +410,16 @@ mod tests {
 
     #[test]
     fn test_ngram_empty() {
-        let tokens: Vec<_> = NgramTokenizer::all_ngrams(1, 5).token_stream("").collect();
+        let tokens = test_helper(NgramTokenizer::all_ngrams(1, 5).token_stream(""));
         assert!(tokens.is_empty());
-        let tokens: Vec<_> = NgramTokenizer::all_ngrams(2, 5).token_stream("").collect();
+        let tokens = test_helper(NgramTokenizer::all_ngrams(2, 5).token_stream(""));
         assert!(tokens.is_empty());
     }
 
     #[test]
     #[should_panic(expected = "min_gram must be greater than 0")]
     fn test_ngram_min_max_interval_empty() {
-        NgramTokenizer::all_ngrams(0, 2).token_stream("hellossss");
+        test_helper(NgramTokenizer::all_ngrams(0, 2).token_stream("hellossss"));
     }
 
     #[test]

src/tokenizer/raw_tokenizer.rs

@@ -1,17 +1,17 @@
-use super::{Token, Tokenizer};
+use super::{Token, TokenStream, Tokenizer};
+use crate::tokenizer::BoxTokenStream;
 
 /// For each value of the field, emit a single unprocessed token.
-#[derive(Clone, Debug, Default)]
+#[derive(Clone)]
 pub struct RawTokenizer;
 
-#[derive(Clone, Debug)]
 pub struct RawTokenStream {
-    token: Option<Token>,
+    token: Token,
+    has_token: bool,
 }
 
 impl Tokenizer for RawTokenizer {
-    type Iter = RawTokenStream;
-    fn token_stream(&self, text: &str) -> Self::Iter {
+    fn token_stream<'a>(&self, text: &'a str) -> BoxTokenStream<'a> {
         let token = Token {
             offset_from: 0,
             offset_to: text.len(),
@@ -19,13 +19,26 @@ impl Tokenizer for RawTokenizer {
             text: text.to_string(),
             position_length: 1,
         };
-        RawTokenStream { token: Some(token) }
+        RawTokenStream {
+            token,
+            has_token: true,
+        }
+        .into()
     }
 }
 
-impl Iterator for RawTokenStream {
-    type Item = Token;
-    fn next(&mut self) -> Option<Token> {
-        self.token.take()
+impl TokenStream for RawTokenStream {
+    fn advance(&mut self) -> bool {
+        let result = self.has_token;
+        self.has_token = false;
+        result
+    }
+
+    fn token(&self) -> &Token {
+        &self.token
+    }
+
+    fn token_mut(&mut self) -> &mut Token {
+        &mut self.token
     }
 }

src/tokenizer/remove_long.rs

@@ -2,8 +2,8 @@
 //! ```rust
 //! use tantivy::tokenizer::*;
 //!
-//! let tokenizer = analyzer_builder(SimpleTokenizer)
-//!   .filter(RemoveLongFilter::limit(5)).build();
+//! let tokenizer = TextAnalyzer::from(SimpleTokenizer)
+//!   .filter(RemoveLongFilter::limit(5));
 //!
 //! let mut stream = tokenizer.token_stream("toolong nice");
 //! // because `toolong` is more than 5 characters, it is filtered
@@ -12,30 +12,61 @@
 //! assert!(stream.next().is_none());
 //! ```
 //!
-use super::{Token, TokenFilter};
+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).
 ///
 /// It is especially useful when indexing unconstrained content.
 /// e.g. Mail containing base-64 encoded pictures etc.
-#[derive(Clone, Debug)]
+#[derive(Clone)]
 pub struct RemoveLongFilter {
-    limit: usize,
+    length_limit: usize,
 }
 
 impl RemoveLongFilter {
     /// Creates a `RemoveLongFilter` given a limit in bytes of the UTF-8 representation.
-    pub fn limit(limit: usize) -> RemoveLongFilter {
-        RemoveLongFilter { limit }
+    pub fn limit(length_limit: usize) -> RemoveLongFilter {
+        RemoveLongFilter { length_limit }
+    }
+}
+
+impl<'a> RemoveLongFilterStream<'a> {
+    fn predicate(&self, token: &Token) -> bool {
+        token.text.len() < self.token_length_limit
     }
 }
 
 impl TokenFilter for RemoveLongFilter {
-    fn transform(&mut self, token: Token) -> Option<Token> {
-        if token.text.len() >= self.limit {
-            return None;
-        }
-        Some(token)
+    fn transform<'a>(&self, token_stream: BoxTokenStream<'a>) -> BoxTokenStream<'a> {
+        BoxTokenStream::from(RemoveLongFilterStream {
+            token_length_limit: self.length_limit,
+            tail: token_stream,
+        })
+    }
+}
+
+pub struct RemoveLongFilterStream<'a> {
+    token_length_limit: usize,
+    tail: BoxTokenStream<'a>,
+}
+
+impl<'a> TokenStream for RemoveLongFilterStream<'a> {
+    fn advance(&mut self) -> bool {
+        while self.tail.advance() {
+            if self.predicate(self.tail.token()) {
+                return true;
+            }
+        }
+        false
+    }
+
+    fn token(&self) -> &Token {
+        self.tail.token()
+    }
+
+    fn token_mut(&mut self) -> &mut Token {
+        self.tail.token_mut()
     }
 }

src/tokenizer/simple_tokenizer.rs

@@ -1,74 +1,59 @@
-use super::{Token, Tokenizer};
+use super::BoxTokenStream;
+use super::{Token, TokenStream, Tokenizer};
+use std::str::CharIndices;
 
 /// Tokenize the text by splitting on whitespaces and punctuation.
-#[derive(Clone, Debug)]
+#[derive(Clone)]
 pub struct SimpleTokenizer;
+
+pub struct SimpleTokenStream<'a> {
+    text: &'a str,
+    chars: CharIndices<'a>,
+    token: Token,
+}
+
 impl Tokenizer for SimpleTokenizer {
-    type Iter = SimpleTokenizerStream;
-    fn token_stream(&self, text: &str) -> Self::Iter {
-        let vec: Vec<_> = text.char_indices().collect();
-        SimpleTokenizerStream {
-            text: text.to_string(),
-            chars: vec.into_iter(),
-            position: usize::max_value(),
-        }
+    fn token_stream<'a>(&self, text: &'a str) -> BoxTokenStream<'a> {
+        BoxTokenStream::from(SimpleTokenStream {
+            text,
+            chars: text.char_indices(),
+            token: Token::default(),
+        })
     }
 }
 
-#[derive(Clone, Debug)]
-pub struct SimpleTokenizerStream {
-    text: String,
-    chars: std::vec::IntoIter<(usize, char)>,
-    position: usize,
-}
-
-impl SimpleTokenizerStream {
+impl<'a> SimpleTokenStream<'a> {
     // search for the end of the current token.
     fn search_token_end(&mut self) -> usize {
         (&mut self.chars)
-            .filter(|&(_, c)| !c.is_alphanumeric())
+            .filter(|&(_, ref c)| !c.is_alphanumeric())
             .map(|(offset, _)| offset)
             .next()
             .unwrap_or_else(|| self.text.len())
     }
 }
 
-impl Iterator for SimpleTokenizerStream {
-    type Item = Token;
-    fn next(&mut self) -> Option<Self::Item> {
-        self.position = self.position.wrapping_add(1);
+impl<'a> TokenStream for SimpleTokenStream<'a> {
+    fn advance(&mut self) -> bool {
+        self.token.text.clear();
+        self.token.position = self.token.position.wrapping_add(1);
         while let Some((offset_from, c)) = self.chars.next() {
             if c.is_alphanumeric() {
                 let offset_to = self.search_token_end();
-                let token = Token {
-                    text: self.text[offset_from..offset_to].into(),
-                    offset_from,
-                    offset_to,
-                    position: self.position,
-                    ..Default::default()
-                };
-                return Some(token);
+                self.token.offset_from = offset_from;
+                self.token.offset_to = offset_to;
+                self.token.text.push_str(&self.text[offset_from..offset_to]);
+                return true;
             }
         }
-        None
-    }
-}
-
-#[cfg(test)]
-mod tests {
-    use super::*;
-
-    #[test]
-    fn test_empty() {
-        let mut empty = SimpleTokenizer.token_stream("");
-        assert_eq!(empty.next(), None);
-    }
-
-    #[test]
-    fn simple_tokenizer() {
-        let mut simple = SimpleTokenizer.token_stream("tokenizer hello world");
-        assert_eq!(simple.next().unwrap().text, "tokenizer");
-        assert_eq!(simple.next().unwrap().text, "hello");
-        assert_eq!(simple.next().unwrap().text, "world");
+        false
+    }
+
+    fn token(&self) -> &Token {
+        &self.token
+    }
+
+    fn token_mut(&mut self) -> &mut Token {
+        &mut self.token
     }
 }

src/tokenizer/stemmer.rs

@@ -1,6 +1,5 @@
-use std::sync::Arc;
-
-use super::{Token, TokenFilter};
+use super::{Token, TokenFilter, TokenStream};
+use crate::tokenizer::BoxTokenStream;
 use rust_stemmers::{self, Algorithm};
 use serde::{Deserialize, Serialize};
 
@@ -59,15 +58,14 @@ impl Language {
 /// Tokens are expected to be lowercased beforehand.
 #[derive(Clone)]
 pub struct Stemmer {
-    stemmer: Arc<rust_stemmers::Stemmer>,
+    stemmer_algorithm: Algorithm,
 }
 
 impl Stemmer {
     /// Creates a new Stemmer `TokenFilter` for a given language algorithm.
     pub fn new(language: Language) -> Stemmer {
-        let stemmer = rust_stemmers::Stemmer::create(language.algorithm());
         Stemmer {
-            stemmer: Arc::new(stemmer),
+            stemmer_algorithm: language.algorithm(),
         }
     }
 }
@@ -80,12 +78,37 @@ impl Default for Stemmer {
 }
 
 impl TokenFilter for Stemmer {
-    fn transform(&mut self, mut token: Token) -> Option<Token> {
-        // TODO remove allocation
-        let stemmed_str: String = self.stemmer.stem(&token.text).into_owned();
-        // TODO remove clear
-        token.text.clear();
-        token.text.push_str(&stemmed_str);
-        Some(token)
+    fn transform<'a>(&self, token_stream: BoxTokenStream<'a>) -> BoxTokenStream<'a> {
+        let inner_stemmer = rust_stemmers::Stemmer::create(self.stemmer_algorithm);
+        BoxTokenStream::from(StemmerTokenStream {
+            tail: token_stream,
+            stemmer: inner_stemmer,
+        })
+    }
+}
+
+pub struct StemmerTokenStream<'a> {
+    tail: BoxTokenStream<'a>,
+    stemmer: rust_stemmers::Stemmer,
+}
+
+impl<'a> TokenStream for StemmerTokenStream<'a> {
+    fn advance(&mut self) -> bool {
+        if !self.tail.advance() {
+            return false;
+        }
+        // TODO remove allocation
+        let stemmed_str: String = self.stemmer.stem(&self.token().text).into_owned();
+        self.token_mut().text.clear();
+        self.token_mut().text.push_str(&stemmed_str);
+        true
+    }
+
+    fn token(&self) -> &Token {
+        self.tail.token()
+    }
+
+    fn token_mut(&mut self) -> &mut Token {
+        self.tail.token_mut()
     }
 }

src/tokenizer/stop_word_filter.rs

@@ -2,15 +2,16 @@
 //! ```rust
 //! use tantivy::tokenizer::*;
 //!
-//! let tokenizer = analyzer_builder(SimpleTokenizer)
-//!   .filter(StopWordFilter::remove(vec!["the".to_string(), "is".to_string()])).build();
+//! 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");
 //! assert_eq!(stream.next().unwrap().text, "fox");
 //! assert_eq!(stream.next().unwrap().text, "crafty");
 //! assert!(stream.next().is_none());
 //! ```
-use super::{Token, TokenFilter};
+use super::{Token, TokenFilter, TokenStream};
+use crate::tokenizer::BoxTokenStream;
 use fnv::FnvHasher;
 use std::collections::HashSet;
 use std::hash::BuildHasherDefault;
@@ -48,12 +49,42 @@ impl StopWordFilter {
     }
 }
 
+pub struct StopWordFilterStream<'a> {
+    words: StopWordHashSet,
+    tail: BoxTokenStream<'a>,
+}
+
 impl TokenFilter for StopWordFilter {
-    fn transform(&mut self, token: Token) -> Option<Token> {
-        if self.words.contains(&token.text) {
-            return None;
+    fn transform<'a>(&self, token_stream: BoxTokenStream<'a>) -> BoxTokenStream<'a> {
+        BoxTokenStream::from(StopWordFilterStream {
+            words: self.words.clone(),
+            tail: token_stream,
+        })
+    }
+}
+
+impl<'a> StopWordFilterStream<'a> {
+    fn predicate(&self, token: &Token) -> bool {
+        !self.words.contains(&token.text)
+    }
+}
+
+impl<'a> TokenStream for StopWordFilterStream<'a> {
+    fn advance(&mut self) -> bool {
+        while self.tail.advance() {
+            if self.predicate(self.tail.token()) {
+                return true;
+            }
         }
-        Some(token)
+        false
+    }
+
+    fn token(&self) -> &Token {
+        self.tail.token()
+    }
+
+    fn token_mut(&mut self) -> &mut Token {
+        self.tail.token_mut()
     }
 }
 

src/tokenizer/token_stream_chain.rs

@@ -1,121 +1,95 @@
-use crate::tokenizer::Token;
+use crate::tokenizer::{BoxTokenStream, Token, TokenStream};
+use std::ops::DerefMut;
 
 const POSITION_GAP: usize = 2;
 
-pub(crate) struct TokenStreamChain<Inner, Outer> {
-    streams_with_offsets: Outer,
-    current: Option<(Inner, usize)>,
-    position: usize,
+pub(crate) struct TokenStreamChain<'a> {
+    offsets: Vec<usize>,
+    token_streams: Vec<BoxTokenStream<'a>>,
     position_shift: usize,
+    stream_idx: usize,
+    token: Token,
 }
 
-impl<'a, Inner, Outer> TokenStreamChain<Inner, Outer>
-where
-    Inner: Iterator<Item = Token>,
-    Outer: Iterator<Item = (Inner, usize)>,
-{
-    pub fn new(mut streams_with_offsets: Outer) -> TokenStreamChain<Inner, Outer> {
-        let current = streams_with_offsets.next();
+impl<'a> TokenStreamChain<'a> {
+    pub fn new(
+        offsets: Vec<usize>,
+        token_streams: Vec<BoxTokenStream<'a>>,
+    ) -> TokenStreamChain<'a> {
         TokenStreamChain {
-            streams_with_offsets: streams_with_offsets,
-            current,
-            position: usize::max_value(),
+            offsets,
+            stream_idx: 0,
+            token_streams,
             position_shift: 0,
+            token: Token::default(),
         }
     }
 }
 
-impl<'a, Inner, Outer> Iterator for TokenStreamChain<Inner, Outer>
-where
-    Inner: Iterator<Item = Token>,
-    Outer: Iterator<Item = (Inner, usize)>,
-{
-    type Item = Token;
-    fn next(&mut self) -> Option<Token> {
-        while let Some((ref mut token_stream, offset_offset)) = self.current {
-            if let Some(mut token) = token_stream.next() {
-                token.offset_from += offset_offset;
-                token.offset_to += offset_offset;
-                token.position += self.position_shift;
-                self.position = token.position;
-                return Some(token);
+impl<'a> TokenStream for TokenStreamChain<'a> {
+    fn advance(&mut self) -> bool {
+        while self.stream_idx < self.token_streams.len() {
+            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];
+                self.token.offset_from = token.offset_from + offset_offset;
+                self.token.offset_to = token.offset_to + offset_offset;
+                self.token.position = token.position + self.position_shift;
+                self.token.text.clear();
+                self.token.text.push_str(token.text.as_str());
+                return true;
+            } else {
+                self.stream_idx += 1;
+                self.position_shift = self.token.position.wrapping_add(POSITION_GAP);
             }
-            self.position_shift = self.position.wrapping_add(POSITION_GAP);
-            self.current = self.streams_with_offsets.next();
         }
-        None
+        false
     }
-}
 
-impl DynTokenStreamChain {
-    pub fn from_vec(
-        streams_with_offsets: Vec<(Box<dyn Iterator<Item = Token>>, usize)>,
-    ) -> impl Iterator<Item = Token> {
-        DynTokenStreamChain {
-            streams_with_offsets,
-            idx: 0,
-            position: usize::max_value(),
-            position_shift: 0,
-        }
+    fn token(&self) -> &Token {
+        assert!(
+            self.stream_idx <= self.token_streams.len(),
+            "You called .token(), after the end of the token stream has been reached"
+        );
+        &self.token
     }
-}
 
-pub(crate) struct DynTokenStreamChain {
-    streams_with_offsets: Vec<(Box<dyn Iterator<Item = Token>>, usize)>,
-    idx: usize,
-    position: usize,
-    position_shift: usize,
-}
-
-impl Iterator for DynTokenStreamChain {
-    type Item = Token;
-    fn next(&mut self) -> Option<Token> {
-        while let Some((token_stream, offset_offset)) = self.streams_with_offsets.get_mut(self.idx)
-        {
-            if let Some(mut token) = token_stream.next() {
-                token.offset_from += *offset_offset;
-                token.offset_to += *offset_offset;
-                token.position += self.position_shift;
-                self.position = token.position;
-                return Some(token);
-            }
-            self.idx += 1;
-            self.position_shift = self.position.wrapping_add(POSITION_GAP);
-        }
-        None
+    fn token_mut(&mut self) -> &mut Token {
+        assert!(
+            self.stream_idx <= self.token_streams.len(),
+            "You called .token(), after the end of the token stream has been reached"
+        );
+        &mut self.token
     }
 }
 
 #[cfg(test)]
 mod tests {
-    use super::super::tokenizer::Tokenizer;
-    use super::super::SimpleTokenizer;
-    use super::*;
+    use super::super::{SimpleTokenizer, TokenStream, Tokenizer};
+    use super::TokenStreamChain;
+    use super::POSITION_GAP;
 
     #[test]
     fn test_chain_first_emits_no_tokens() {
         let token_streams = vec![
-            (SimpleTokenizer.token_stream(""), 0),
-            (SimpleTokenizer.token_stream("hello world"), 0),
+            SimpleTokenizer.token_stream(""),
+            SimpleTokenizer.token_stream("hello world"),
         ];
-        let mut token_chain = TokenStreamChain::new(token_streams.into_iter());
-        let token = token_chain.next();
+        let mut token_chain = TokenStreamChain::new(vec![0, 0], token_streams);
 
-        let expect = Token {
-            offset_from: 0,
-            offset_to: 5,
-            position: POSITION_GAP - 1,
-            text: "hello".into(),
-            ..Token::default()
-        };
-        assert_eq!(token.unwrap(), expect);
+        assert!(token_chain.advance());
+        assert_eq!(token_chain.token().text, "hello");
+        assert_eq!(token_chain.token().offset_from, 0);
+        assert_eq!(token_chain.token().offset_to, 5);
+        assert_eq!(token_chain.token().position, POSITION_GAP - 1);
 
-        let token = token_chain.next().unwrap();
-        assert_eq!(token.text, "world");
-        assert_eq!(token.offset_from, 6);
-        assert_eq!(token.offset_to, 11);
-        assert_eq!(token.position, POSITION_GAP);
+        assert!(token_chain.advance());
+        assert_eq!(token_chain.token().text, "world");
+        assert_eq!(token_chain.token().offset_from, 6);
+        assert_eq!(token_chain.token().offset_to, 11);
+        assert_eq!(token_chain.token().position, POSITION_GAP);
 
-        assert!(token_chain.next().is_none());
+        assert!(!token_chain.advance());
     }
 }

src/tokenizer/tokenized_string.rs

@@ -1,4 +1,4 @@
-use crate::tokenizer::{Token, TokenStreamChain};
+use crate::tokenizer::{BoxTokenStream, Token, TokenStream, TokenStreamChain};
 use serde::{Deserialize, Serialize};
 use std::cmp::Ordering;
 
@@ -26,14 +26,14 @@ impl PartialOrd for PreTokenizedString {
 /// TokenStream implementation which wraps PreTokenizedString
 pub struct PreTokenizedStream {
     tokenized_string: PreTokenizedString,
-    current_token: usize,
+    current_token: i64,
 }
 
 impl From<PreTokenizedString> for PreTokenizedStream {
     fn from(s: PreTokenizedString) -> PreTokenizedStream {
         PreTokenizedStream {
             tokenized_string: s,
-            current_token: 0,
+            current_token: -1,
         }
     }
 }
@@ -41,28 +41,49 @@ impl From<PreTokenizedString> for PreTokenizedStream {
 impl PreTokenizedStream {
     /// Creates a TokenStream from PreTokenizedString array
     pub fn chain_tokenized_strings<'a>(
-        tok_strings: &'a [&PreTokenizedString],
-    ) -> impl Iterator<Item = Token> + 'a {
-        let streams_with_offsets = tok_strings.iter().scan(0, |total_offset, tok_string| {
-            let next = Some((
-                PreTokenizedStream::from((*tok_string).to_owned()),
-                *total_offset,
-            ));
-            if let Some(last_token) = tok_string.tokens.last() {
-                *total_offset += last_token.offset_to;
+        tok_strings: &'a [&'a PreTokenizedString],
+    ) -> BoxTokenStream {
+        if tok_strings.len() == 1 {
+            PreTokenizedStream::from((*tok_strings[0]).clone()).into()
+        } else {
+            let mut offsets = vec![];
+            let mut total_offset = 0;
+            for &tok_string in tok_strings {
+                offsets.push(total_offset);
+                if let Some(last_token) = tok_string.tokens.last() {
+                    total_offset += last_token.offset_to;
+                }
             }
-            next
-        });
-        TokenStreamChain::new(streams_with_offsets)
+            // TODO remove the string cloning.
+            let token_streams: Vec<BoxTokenStream<'static>> = tok_strings
+                .iter()
+                .map(|&tok_string| PreTokenizedStream::from((*tok_string).clone()).into())
+                .collect();
+            TokenStreamChain::new(offsets, token_streams).into()
+        }
     }
 }
 
-impl Iterator for PreTokenizedStream {
-    type Item = Token;
-    fn next(&mut self) -> Option<Token> {
-        let token = self.tokenized_string.tokens.get(self.current_token)?;
+impl TokenStream for PreTokenizedStream {
+    fn advance(&mut self) -> bool {
         self.current_token += 1;
-        Some(token.clone())
+        self.current_token < self.tokenized_string.tokens.len() as i64
+    }
+
+    fn token(&self) -> &Token {
+        assert!(
+            self.current_token >= 0,
+            "TokenStream not initialized. You should call advance() at least once."
+        );
+        &self.tokenized_string.tokens[self.current_token as usize]
+    }
+
+    fn token_mut(&mut self) -> &mut Token {
+        assert!(
+            self.current_token >= 0,
+            "TokenStream not initialized. You should call advance() at least once."
+        );
+        &mut self.tokenized_string.tokens[self.current_token as usize]
     }
 }
 
@@ -98,9 +119,10 @@ mod tests {
         let mut token_stream = PreTokenizedStream::from(tok_text.clone());
 
         for expected_token in tok_text.tokens {
-            assert_eq!(token_stream.next().unwrap(), expected_token);
+            assert!(token_stream.advance());
+            assert_eq!(token_stream.token(), &expected_token);
         }
-        assert!(token_stream.next().is_none());
+        assert!(!token_stream.advance());
     }
 
     #[test]
@@ -161,8 +183,9 @@ mod tests {
         ];
 
         for expected_token in expected_tokens {
-            assert_eq!(token_stream.next().unwrap(), expected_token);
+            assert!(token_stream.advance());
+            assert_eq!(token_stream.token(), &expected_token);
         }
-        assert!(token_stream.next().is_none());
+        assert!(!token_stream.advance());
     }
 }

src/tokenizer/tokenizer.rs

@@ -2,23 +2,8 @@ use crate::tokenizer::TokenStreamChain;
 use serde::{Deserialize, Serialize};
 /// The tokenizer module contains all of the tools used to process
 /// text in `tantivy`.
-
-pub trait TextAnalyzerClone {
-    fn box_clone(&self) -> Box<dyn TextAnalyzerT>;
-}
-
-/// 'Top-level' trait hiding concrete types, below which static dispatch occurs.
-pub trait TextAnalyzerT: 'static + Send + Sync + TextAnalyzerClone {
-    /// 'Top-level' dynamic dispatch function hiding concrete types of the staticly
-    /// dispatched `token_stream` from the `Tokenizer` trait.
-    fn token_stream(&self, text: &str) -> Box<dyn Iterator<Item = Token>>;
-}
-
-impl Clone for Box<dyn TextAnalyzerT> {
-    fn clone(&self) -> Self {
-        (**self).box_clone()
-    }
-}
+use std::borrow::{Borrow, BorrowMut};
+use std::ops::{Deref, DerefMut};
 
 /// Token
 #[derive(Debug, Clone, Serialize, Deserialize, Eq, PartialEq)]
@@ -50,116 +35,35 @@ impl Default for Token {
     }
 }
 
-/// Trait for the pluggable components of `Tokenizer`s.
-pub trait TokenFilter: 'static + Send + Sync + Clone {
-    /// Take a `Token` and transform it or return `None` if it's to be removed
-    /// from the output stream.
-    fn transform(&mut self, token: Token) -> Option<Token>;
-}
-
-/// `Tokenizer` are in charge of splitting text into a stream of token
-/// before indexing.
-///
-/// See the [module documentation](./index.html) for more detail.
-pub trait Tokenizer: 'static + Send + Sync + Clone {
-    /// An iteratable type is returned.
-    type Iter: Iterator<Item = Token>;
-    /// Creates a token stream for a given `str`.
-    fn token_stream(&self, text: &str) -> Self::Iter;
-    /// Tokenize an array`&str`
-    ///
-    /// The resulting `Token` stream 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<'a>(&'a self, texts: &'a [&str]) -> Box<dyn Iterator<Item = Token> + 'a> {
-        let streams_with_offsets = texts.iter().scan(0, move |total_offset, &text| {
-            let temp = *total_offset;
-            *total_offset += text.len();
-            Some((self.token_stream(text), temp))
-        });
-        Box::new(TokenStreamChain::new(streams_with_offsets))
-    }
-}
-
-/// `TextAnalyzer` wraps the tokenization of an input text and its modification by any filters applied onto it.
+/// `TextAnalyzer` tokenizes an input text into tokens and modifies the resulting `TokenStream`.
 ///
 /// It simply wraps a `Tokenizer` and a list of `TokenFilter` that are applied sequentially.
-#[derive(Clone, Debug, Default)]
-pub struct TextAnalyzer<T>(T);
+pub struct TextAnalyzer {
+    tokenizer: Box<dyn Tokenizer>,
+    token_filters: Vec<BoxTokenFilter>,
+}
 
-impl<T: Tokenizer> From<T> for TextAnalyzer<T> {
-    fn from(src: T) -> TextAnalyzer<T> {
-        TextAnalyzer(src)
+impl<T: Tokenizer> From<T> for TextAnalyzer {
+    fn from(tokenizer: T) -> Self {
+        TextAnalyzer::new(tokenizer, Vec::new())
     }
 }
 
-impl<T: Tokenizer> TextAnalyzerClone for TextAnalyzer<T> {
-    fn box_clone(&self) -> Box<dyn TextAnalyzerT> {
-        Box::new(TextAnalyzer(self.0.clone()))
-    }
-}
-
-impl<T: Tokenizer> TextAnalyzerT for TextAnalyzer<T> {
-    fn token_stream(&self, text: &str) -> Box<dyn Iterator<Item = Token>> {
-        Box::new(self.0.token_stream(text))
-    }
-}
-
-/// Identity `TokenFilter`
-#[derive(Clone, Debug, Default)]
-pub struct Identity;
-
-impl TokenFilter for Identity {
-    fn transform(&mut self, token: Token) -> Option<Token> {
-        Some(token)
-    }
-}
-
-/// `Filter` is a wrapper around a `Token` stream and a `TokenFilter` which modifies it.
-#[derive(Clone, Default, Debug)]
-pub struct Filter<I, F> {
-    iter: I,
-    f: F,
-}
-
-impl<I, F> Iterator for Filter<I, F>
-where
-    I: Iterator<Item = Token>,
-    F: TokenFilter,
-{
-    type Item = Token;
-    fn next(&mut self) -> Option<Token> {
-        while let Some(token) = self.iter.next() {
-            if let Some(tok) = self.f.transform(token) {
-                return Some(tok);
-            }
+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,
         }
-        None
     }
-}
 
-#[derive(Clone, Debug, Default)]
-pub struct AnalyzerBuilder<T, F> {
-    tokenizer: T,
-    f: F,
-}
-
-/// Construct an `AnalyzerBuilder` on which to apply `TokenFilter`.
-pub fn analyzer_builder<T: Tokenizer>(tokenizer: T) -> AnalyzerBuilder<T, Identity> {
-    AnalyzerBuilder {
-        tokenizer,
-        f: Identity,
-    }
-}
-
-impl<T, F> AnalyzerBuilder<T, F>
-where
-    T: Tokenizer,
-    F: TokenFilter,
-{
     /// Appends a token filter to the current tokenizer.
     ///
-    /// The method consumes the current `Token` and returns a
+    /// The method consumes the current `TokenStream` and returns a
     /// new one.
     ///
     /// # Example
@@ -167,35 +71,248 @@ where
     /// ```rust
     /// use tantivy::tokenizer::*;
     ///
-    /// let en_stem = analyzer_builder(SimpleTokenizer)
+    /// let en_stem = TextAnalyzer::from(SimpleTokenizer)
     ///     .filter(RemoveLongFilter::limit(40))
-    ///     .filter(LowerCaser::new())
-    ///     .filter(Stemmer::default()).build();
+    ///     .filter(LowerCaser)
+    ///     .filter(Stemmer::default());
     /// ```
     ///
-    pub fn filter<G: TokenFilter>(self, f: G) -> AnalyzerBuilder<AnalyzerBuilder<T, F>, G> {
-        AnalyzerBuilder { tokenizer: self, f }
+    pub fn filter<F: Into<BoxTokenFilter>>(mut self, token_filter: F) -> Self {
+        self.token_filters.push(token_filter.into());
+        self
     }
-    /// Finalize the build process.
-    pub fn build(self) -> TextAnalyzer<AnalyzerBuilder<T, F>> {
-        TextAnalyzer(self)
+
+    /// Tokenize an array`&str`
+    ///
+    /// 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<'a>(&self, texts: &'a [&'a str]) -> BoxTokenStream<'a> {
+        assert!(!texts.is_empty());
+        if texts.len() == 1 {
+            self.token_stream(texts[0])
+        } else {
+            let mut offsets = vec![];
+            let mut total_offset = 0;
+            for &text in texts {
+                offsets.push(total_offset);
+                total_offset += text.len();
+            }
+            let token_streams: Vec<BoxTokenStream<'a>> = texts
+                .iter()
+                .cloned()
+                .map(|text| self.token_stream(text))
+                .collect();
+            From::from(TokenStreamChain::new(offsets, token_streams))
+        }
+    }
+
+    /// Creates a token stream for a given `str`.
+    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<T: Tokenizer, F: TokenFilter> Tokenizer for AnalyzerBuilder<T, F> {
-    type Iter = Filter<T::Iter, F>;
-    fn token_stream(&self, text: &str) -> Self::Iter {
-        Filter {
-            iter: self.tokenizer.token_stream(text),
-            f: self.f.clone(),
+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<'b>(&'b self) -> &'b Token {
+        let token_stream: &'b (dyn TokenStream + 'a) = self.borrow();
+        token_stream.token()
+    }
+
+    fn token_mut<'b>(&'b mut self) -> &'b mut Token {
+        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.
+///
+/// # Example
+///
+/// ```
+/// use tantivy::tokenizer::*;
+///
+/// let tokenizer = TextAnalyzer::from(SimpleTokenizer)
+///        .filter(RemoveLongFilter::limit(40))
+///        .filter(LowerCaser);
+/// let mut token_stream = tokenizer.token_stream("Hello, happy tax payer");
+/// {
+///     let token = token_stream.next().unwrap();
+///     assert_eq!(&token.text, "hello");
+///     assert_eq!(token.offset_from, 0);
+///     assert_eq!(token.offset_to, 5);
+///     assert_eq!(token.position, 0);
+/// }
+/// {
+///     let token = token_stream.next().unwrap();
+///     assert_eq!(&token.text, "happy");
+///     assert_eq!(token.offset_from, 7);
+///     assert_eq!(token.offset_to, 12);
+///     assert_eq!(token.position, 1);
+/// }
+/// ```
+///
+pub trait TokenStream {
+    /// Advance to the next token
+    ///
+    /// Returns false if there are no other tokens.
+    fn advance(&mut self) -> bool;
+
+    /// Returns a reference to the current token.
+    fn token(&self) -> &Token;
+
+    /// Returns a mutable reference to the current token.
+    fn token_mut(&mut self) -> &mut Token;
+
+    /// Helper to iterate over tokens. It
+    /// simply combines a call to `.advance()`
+    /// and `.token()`.
+    ///
+    /// ```
+    /// use tantivy::tokenizer::*;
+    ///
+    /// let tokenizer = TextAnalyzer::from(SimpleTokenizer)
+    ///       .filter(RemoveLongFilter::limit(40))
+    ///       .filter(LowerCaser);
+    /// let mut token_stream = tokenizer.token_stream("Hello, happy tax payer");
+    /// while let Some(token) = token_stream.next() {
+    ///     println!("Token {:?}", token.text);
+    /// }
+    /// ```
+    fn next(&mut self) -> Option<&Token> {
+        if self.advance() {
+            Some(self.token())
+        } else {
+            None
+        }
+    }
+
+    /// 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() {
+            sink(self.token());
+            num_tokens_pushed += 1u32;
+        }
+        num_tokens_pushed
+    }
+}
+
+pub trait TokenFilterClone {
+    fn box_clone(&self) -> BoxTokenFilter;
+}
+
+/// Trait for the pluggable components of `Tokenizer`s.
+pub trait TokenFilter: 'static + Send + Sync + TokenFilterClone {
+    /// Wraps a token stream and returns the modified one.
+    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)]
 mod test {
-    use super::*;
-    use crate::tokenizer::SimpleTokenizer;
+    use super::Token;
 
     #[test]
     fn clone() {
@@ -213,15 +330,4 @@ mod test {
         assert_eq!(t1.offset_to, t2.offset_to);
         assert_eq!(t1.text, t2.text);
     }
-
-    #[test]
-    fn text_analyzer() {
-        let mut stream = SimpleTokenizer.token_stream("tokenizer hello world");
-        dbg!(stream.next());
-        dbg!(stream.next());
-        dbg!(stream.next());
-        dbg!(stream.next());
-        dbg!(stream.next());
-        dbg!(stream.next());
-    }
 }

src/tokenizer/tokenizer_manager.rs

@@ -1,5 +1,5 @@
 use crate::tokenizer::stemmer::Language;
-use crate::tokenizer::tokenizer::{analyzer_builder, TextAnalyzer, TextAnalyzerT, Tokenizer};
+use crate::tokenizer::tokenizer::TextAnalyzer;
 use crate::tokenizer::LowerCaser;
 use crate::tokenizer::RawTokenizer;
 use crate::tokenizer::RemoveLongFilter;
@@ -22,23 +22,24 @@ use std::sync::{Arc, RwLock};
 ///  search engine.
 #[derive(Clone)]
 pub struct TokenizerManager {
-    tokenizers: Arc<RwLock<HashMap<String, Box<dyn TextAnalyzerT>>>>,
+    tokenizers: Arc<RwLock<HashMap<String, TextAnalyzer>>>,
 }
 
 impl TokenizerManager {
     /// Registers a new tokenizer associated with a given name.
-    pub fn register<U: Tokenizer, T>(&self, tokenizer_name: &str, tokenizer: T)
+    pub fn register<T>(&self, tokenizer_name: &str, tokenizer: T)
     where
-        T: Into<TextAnalyzer<U>>,
+        TextAnalyzer: From<T>,
     {
+        let boxed_tokenizer: TextAnalyzer = TextAnalyzer::from(tokenizer);
         self.tokenizers
             .write()
             .expect("Acquiring the lock should never fail")
-            .insert(tokenizer_name.to_string(), Box::new(tokenizer.into()));
+            .insert(tokenizer_name.to_string(), boxed_tokenizer);
     }
 
     /// Accessing a tokenizer given its name.
-    pub fn get(&self, tokenizer_name: &str) -> Option<Box<dyn TextAnalyzerT>> {
+    pub fn get(&self, tokenizer_name: &str) -> Option<TextAnalyzer> {
         self.tokenizers
             .read()
             .expect("Acquiring the lock should never fail")
@@ -53,25 +54,23 @@ impl Default for TokenizerManager {
     /// - simple
     /// - en_stem
     /// - ja
-    fn default() -> Self {
+    fn default() -> TokenizerManager {
         let manager = TokenizerManager {
             tokenizers: Arc::new(RwLock::new(HashMap::new())),
         };
         manager.register("raw", RawTokenizer);
         manager.register(
             "default",
-            analyzer_builder(SimpleTokenizer)
+            TextAnalyzer::from(SimpleTokenizer)
                 .filter(RemoveLongFilter::limit(40))
-                .filter(LowerCaser::new())
-                .build(),
+                .filter(LowerCaser),
         );
         manager.register(
             "en_stem",
-            analyzer_builder(SimpleTokenizer)
+            TextAnalyzer::from(SimpleTokenizer)
                 .filter(RemoveLongFilter::limit(40))
-                .filter(LowerCaser::new())
-                .filter(Stemmer::new(Language::English))
-                .build(),
+                .filter(LowerCaser)
+                .filter(Stemmer::new(Language::English)),
         );
         manager
     }