Merge pull request #67 from Kerollmops/reintroduce-stop-words

Reintroduce stop words

examples/create-database.rs

@@ -1,12 +1,13 @@
 #[global_allocator]
 static ALLOC: jemallocator::Jemalloc = jemallocator::Jemalloc;
 
+use std::io::{self, BufRead, BufReader};
 use std::path::{Path, PathBuf};
 use std::error::Error;
 use std::borrow::Cow;
 use std::fs::File;
 
-use hashbrown::HashMap;
+use hashbrown::{HashMap, HashSet};
 use serde_derive::{Serialize, Deserialize};
 use structopt::StructOpt;
 
@@ -26,6 +27,13 @@ pub struct Opt {
     /// The path to the schema.
     #[structopt(long = "schema", parse(from_os_str))]
     pub schema_path: PathBuf,
+
+    /// The path to the list of stop words (one by line).
+    #[structopt(long = "stop-words", parse(from_os_str))]
+    pub stop_words_path: Option<PathBuf>,
+
+    #[structopt(long = "update-group-size")]
+    pub update_group_size: Option<usize>,
 }
 
 #[derive(Serialize, Deserialize)]
@@ -34,20 +42,32 @@ struct Document<'a> (
     HashMap<Cow<'a, str>, Cow<'a, str>>
 );
 
-fn index(schema: Schema, database_path: &Path, csv_data_path: &Path) -> Result<Database, Box<Error>> {
+fn index(
+    schema: Schema,
+    database_path: &Path,
+    csv_data_path: &Path,
+    update_group_size: Option<usize>,
+    stop_words: &HashSet<String>,
+) -> Result<Database, Box<Error>>
+{
     let database = Database::create(database_path, &schema)?;
 
-    println!("start indexing...");
-
-    let tokenizer_builder = DefaultBuilder::new();
-    let update_path = tempfile::NamedTempFile::new()?;
-    let mut update = UpdateBuilder::new(update_path.path().to_path_buf(), schema);
-
     let mut rdr = csv::Reader::from_path(csv_data_path)?;
     let mut raw_record = csv::StringRecord::new();
     let headers = rdr.headers()?.clone();
 
-    while rdr.read_record(&mut raw_record)? {
+    let mut i = 0;
+    let mut end_of_file = false;
+
+    while !end_of_file {
+        let tokenizer_builder = DefaultBuilder::new();
+        let update_path = tempfile::NamedTempFile::new()?;
+        let mut update = UpdateBuilder::new(update_path.path().to_path_buf(), schema.clone());
+
+        loop {
+            end_of_file = !rdr.read_record(&mut raw_record)?;
+            if end_of_file { break }
+
             let document: Document = match raw_record.deserialize(Some(&headers)) {
                 Ok(document) => document,
                 Err(e) => {
@@ -56,15 +76,41 @@ fn index(schema: Schema, database_path: &Path, csv_data_path: &Path) -> Result<D
                 }
             };
 
-        update.update_document(&document, &tokenizer_builder)?;
+            update.update_document(&document, &tokenizer_builder, &stop_words)?;
+
+            print!("\rindexing document {}", i);
+            i += 1;
+
+            if let Some(group_size) = update_group_size {
+                if i % group_size == 0 { break }
+            }
         }
 
+        println!();
+
+        println!("building update...");
         let update = update.build()?;
+        println!("ingesting update...");
         database.ingest_update_file(update)?;
+    }
 
     Ok(database)
 }
 
+fn retrieve_stop_words(path: &Path) -> io::Result<HashSet<String>> {
+    let f = File::open(path)?;
+    let reader = BufReader::new(f);
+    let mut words = HashSet::new();
+
+    for line in reader.lines() {
+        let line = line?;
+        let word = line.trim().to_string();
+        words.insert(word);
+    }
+
+    Ok(words)
+}
+
 fn main() -> Result<(), Box<Error>> {
     let _ = env_logger::init();
     let opt = Opt::from_args();
@@ -74,8 +120,13 @@ fn main() -> Result<(), Box<Error>> {
         Schema::from_toml(file)?
     };
 
+    let stop_words = match opt.stop_words_path {
+        Some(ref path) => retrieve_stop_words(path)?,
+        None           => HashSet::new(),
+    };
+
     let (elapsed, result) = elapsed::measure_time(|| {
-        index(schema, &opt.database_path, &opt.csv_data_path)
+        index(schema, &opt.database_path, &opt.csv_data_path, opt.update_group_size, &stop_words)
     });
 
     if let Err(e) = result {

misc/en.stopwords.txt

@@ -95,7 +95,8 @@ or
 other
 ought
 our
-ours    ourselves
+ours
+ourselves
 out
 over
 own

misc/fr.stopwords.txt

@@ -0,0 +1,163 @@
+au
+aux
+avec
+ce
+ces
+dans
+de
+des
+du
+elle
+en
+et
+eux
+il
+je
+la
+le
+leur
+lui
+ma
+mais
+me
+même
+mes
+moi
+mon
+ne
+nos
+notre
+nous
+on
+ou
+par
+pas
+pour
+qu
+que
+qui
+sa
+se
+ses
+son
+sur
+ta
+te
+tes
+toi
+ton
+tu
+un
+une
+vos
+votre
+vous
+c
+d
+j
+l
+à
+m
+n
+s
+t
+y
+été
+étée
+étées
+étés
+étant
+suis
+es
+est
+sommes
+êtes
+sont
+serai
+seras
+sera
+serons
+serez
+seront
+serais
+serait
+serions
+seriez
+seraient
+étais
+était
+étions
+étiez
+étaient
+fus
+fut
+fûmes
+fûtes
+furent
+sois
+soit
+soyons
+soyez
+soient
+fusse
+fusses
+fût
+fussions
+fussiez
+fussent
+ayant
+eu
+eue
+eues
+eus
+ai
+as
+avons
+avez
+ont
+aurai
+auras
+aura
+aurons
+aurez
+auront
+aurais
+aurait
+aurions
+auriez
+auraient
+avais
+avait
+avions
+aviez
+avaient
+eut
+eûmes
+eûtes
+eurent
+aie
+aies
+ait
+ayons
+ayez
+aient
+eusse
+eusses
+eût
+eussions
+eussiez
+eussent
+ceci
+celà
+cet
+cette
+ici
+ils
+les
+leurs
+quel
+quels
+quelle
+quelles
+sans
+soi

src/database/mod.rs

@@ -7,6 +7,7 @@ use rocksdb::rocksdb_options::{DBOptions, IngestExternalFileOptions, ColumnFamil
 use rocksdb::rocksdb::{Writable, Snapshot};
 use rocksdb::{DB, DBVector, MergeOperands};
 use crossbeam::atomic::ArcCell;
+use log::debug;
 
 pub use self::document_key::{DocumentKey, DocumentKeyAttr};
 pub use self::view::{DatabaseView, DocumentIter};
@@ -147,9 +148,11 @@ impl Database {
             let options = IngestExternalFileOptions::new();
             // options.move_files(move_update);
 
+            debug!("ingest update file");
             let cf_handle = db.cf_handle("default").expect("\"default\" column family not found");
             db.ingest_external_file_optimized(&cf_handle, &options, &[&path])?;
 
+            debug!("compacting index range");
             // Compacting to trigger the merge operator only one time
             // while ingesting the update and not each time searching
             db.compact_range(Some(DATA_INDEX), Some(DATA_INDEX));
@@ -185,6 +188,7 @@ mod tests {
     use std::error::Error;
 
     use serde_derive::{Serialize, Deserialize};
+    use hashbrown::HashSet;
     use tempfile::tempdir;
 
     use crate::database::schema::{SchemaBuilder, STORED, INDEXED};
@@ -194,6 +198,7 @@ mod tests {
     #[test]
     fn ingest_one_update_file() -> Result<(), Box<Error>> {
         let dir = tempdir()?;
+        let stop_words = HashSet::new();
 
         let rocksdb_path = dir.path().join("rocksdb.rdb");
 
@@ -237,8 +242,8 @@ mod tests {
             let tokenizer_builder = DefaultBuilder::new();
             let mut builder = UpdateBuilder::new(update_path, schema);
 
-            docid0 = builder.update_document(&doc0, &tokenizer_builder)?;
+            docid0 = builder.update_document(&doc0, &tokenizer_builder, &stop_words)?;
-            docid1 = builder.update_document(&doc1, &tokenizer_builder)?;
+            docid1 = builder.update_document(&doc1, &tokenizer_builder, &stop_words)?;
 
             builder.build()?
         };
@@ -258,6 +263,7 @@ mod tests {
     #[test]
     fn ingest_two_update_files() -> Result<(), Box<Error>> {
         let dir = tempdir()?;
+        let stop_words = HashSet::new();
 
         let rocksdb_path = dir.path().join("rocksdb.rdb");
 
@@ -312,8 +318,8 @@ mod tests {
             let update_path = dir.path().join("update-000.sst");
             let mut builder = UpdateBuilder::new(update_path, schema.clone());
 
-            docid0 = builder.update_document(&doc0, &tokenizer_builder)?;
+            docid0 = builder.update_document(&doc0, &tokenizer_builder, &stop_words)?;
-            docid1 = builder.update_document(&doc1, &tokenizer_builder)?;
+            docid1 = builder.update_document(&doc1, &tokenizer_builder, &stop_words)?;
 
             builder.build()?
         };
@@ -325,8 +331,8 @@ mod tests {
             let update_path = dir.path().join("update-001.sst");
             let mut builder = UpdateBuilder::new(update_path, schema);
 
-            docid2 = builder.update_document(&doc2, &tokenizer_builder)?;
+            docid2 = builder.update_document(&doc2, &tokenizer_builder, &stop_words)?;
-            docid3 = builder.update_document(&doc3, &tokenizer_builder)?;
+            docid3 = builder.update_document(&doc3, &tokenizer_builder, &stop_words)?;
 
             builder.build()?
         };
@@ -364,8 +370,9 @@ mod bench {
     use rand::distributions::Alphanumeric;
     use rand_xorshift::XorShiftRng;
     use rand::{Rng, SeedableRng};
-    use rand::seq::SliceRandom;
     use serde_derive::Serialize;
+    use rand::seq::SliceRandom;
+    use hashbrown::HashSet;
 
     use crate::tokenizer::DefaultBuilder;
     use crate::database::update::UpdateBuilder;
@@ -394,6 +401,7 @@ mod bench {
     #[bench]
     fn open_little_database(bench: &mut Bencher) -> Result<(), Box<Error>> {
         let dir = tempfile::tempdir()?;
+        let stop_words = HashSet::new();
 
         let mut builder = SchemaBuilder::with_identifier("id");
         builder.new_attribute("title", STORED | INDEXED);
@@ -421,7 +429,7 @@ mod bench {
                 title: random_sentences(rng.gen_range(1, 8), &mut rng),
                 description: random_sentences(rng.gen_range(20, 200), &mut rng),
             };
-            builder.update_document(&document, &tokenizer_builder)?;
+            builder.update_document(&document, &tokenizer_builder, &stop_words)?;
         }
 
         let update = builder.build()?;
@@ -440,6 +448,7 @@ mod bench {
     #[bench]
     fn open_medium_database(bench: &mut Bencher) -> Result<(), Box<Error>> {
         let dir = tempfile::tempdir()?;
+        let stop_words = HashSet::new();
 
         let mut builder = SchemaBuilder::with_identifier("id");
         builder.new_attribute("title", STORED | INDEXED);
@@ -467,7 +476,7 @@ mod bench {
                 title: random_sentences(rng.gen_range(1, 8), &mut rng),
                 description: random_sentences(rng.gen_range(20, 200), &mut rng),
             };
-            builder.update_document(&document, &tokenizer_builder)?;
+            builder.update_document(&document, &tokenizer_builder, &stop_words)?;
         }
 
         let update = builder.build()?;
@@ -487,6 +496,7 @@ mod bench {
     #[ignore]
     fn open_big_database(bench: &mut Bencher) -> Result<(), Box<Error>> {
         let dir = tempfile::tempdir()?;
+        let stop_words = HashSet::new();
 
         let mut builder = SchemaBuilder::with_identifier("id");
         builder.new_attribute("title", STORED | INDEXED);
@@ -514,7 +524,7 @@ mod bench {
                 title: random_sentences(rng.gen_range(1, 8), &mut rng),
                 description: random_sentences(rng.gen_range(20, 200), &mut rng),
             };
-            builder.update_document(&document, &tokenizer_builder)?;
+            builder.update_document(&document, &tokenizer_builder, &stop_words)?;
         }
 
         let update = builder.build()?;
@@ -533,6 +543,7 @@ mod bench {
     #[bench]
     fn search_oneletter_little_database(bench: &mut Bencher) -> Result<(), Box<Error>> {
         let dir = tempfile::tempdir()?;
+        let stop_words = HashSet::new();
 
         let mut builder = SchemaBuilder::with_identifier("id");
         builder.new_attribute("title", STORED | INDEXED);
@@ -560,7 +571,7 @@ mod bench {
                 title: random_sentences(rng.gen_range(1, 8), &mut rng),
                 description: random_sentences(rng.gen_range(20, 200), &mut rng),
             };
-            builder.update_document(&document, &tokenizer_builder)?;
+            builder.update_document(&document, &tokenizer_builder, &stop_words)?;
         }
 
         let update = builder.build()?;
@@ -579,6 +590,7 @@ mod bench {
     #[bench]
     fn search_oneletter_medium_database(bench: &mut Bencher) -> Result<(), Box<Error>> {
         let dir = tempfile::tempdir()?;
+        let stop_words = HashSet::new();
 
         let mut builder = SchemaBuilder::with_identifier("id");
         builder.new_attribute("title", STORED | INDEXED);
@@ -606,7 +618,7 @@ mod bench {
                 title: random_sentences(rng.gen_range(1, 8), &mut rng),
                 description: random_sentences(rng.gen_range(20, 200), &mut rng),
             };
-            builder.update_document(&document, &tokenizer_builder)?;
+            builder.update_document(&document, &tokenizer_builder, &stop_words)?;
         }
 
         let update = builder.build()?;
@@ -626,6 +638,7 @@ mod bench {
     #[ignore]
     fn search_oneletter_big_database(bench: &mut Bencher) -> Result<(), Box<Error>> {
         let dir = tempfile::tempdir()?;
+        let stop_words = HashSet::new();
 
         let mut builder = SchemaBuilder::with_identifier("id");
         builder.new_attribute("title", STORED | INDEXED);
@@ -653,7 +666,7 @@ mod bench {
                 title: random_sentences(rng.gen_range(1, 8), &mut rng),
                 description: random_sentences(rng.gen_range(20, 200), &mut rng),
             };
-            builder.update_document(&document, &tokenizer_builder)?;
+            builder.update_document(&document, &tokenizer_builder, &stop_words)?;
         }
 
         let update = builder.build()?;

src/database/serde/indexer_serializer.rs

@@ -5,6 +5,7 @@ use crate::tokenizer::TokenizerBuilder;
 use crate::tokenizer::Token;
 use crate::{DocumentId, DocIndex, Attribute, WordArea};
 
+use hashbrown::HashSet;
 use serde::Serialize;
 use serde::ser;
 
@@ -13,6 +14,7 @@ pub struct IndexerSerializer<'a, B> {
     pub update: &'a mut DocumentUpdate,
     pub document_id: DocumentId,
     pub attribute: SchemaAttr,
+    pub stop_words: &'a HashSet<String>,
 }
 
 impl<'a, B> ser::Serializer for IndexerSerializer<'a, B>
@@ -48,6 +50,7 @@ where B: TokenizerBuilder
 
     fn serialize_str(self, v: &str) -> Result<Self::Ok, Self::Error> {
         for Token { word, word_index, char_index } in self.tokenizer_builder.build(v) {
+
             // FIXME must u32::try_from instead
             let attribute = match Attribute::new(self.attribute.0, word_index as u32) {
                 Ok(attribute) => attribute,
@@ -69,6 +72,8 @@ where B: TokenizerBuilder
             // insert the exact representation
             let word_lower = word.to_lowercase();
 
+            if self.stop_words.contains(&word_lower) { continue }
+
             // and the unidecoded lowercased version
             let word_unidecoded = unidecode::unidecode(word).to_lowercase();
             if word_lower != word_unidecoded {

src/database/serde/serializer.rs

@@ -1,3 +1,4 @@
+use hashbrown::HashSet;
 use serde::Serialize;
 use serde::ser;
 
@@ -14,6 +15,7 @@ pub struct Serializer<'a, B> {
     pub update: &'a mut DocumentUpdate,
     pub document_id: DocumentId,
     pub tokenizer_builder: &'a B,
+    pub stop_words: &'a HashSet<String>,
 }
 
 impl<'a, B> ser::Serializer for Serializer<'a, B>
@@ -139,6 +141,7 @@ where B: TokenizerBuilder
             document_id: self.document_id,
             update: self.update,
             tokenizer_builder: self.tokenizer_builder,
+            stop_words: self.stop_words,
             current_key_name: None,
         })
     }
@@ -154,6 +157,7 @@ where B: TokenizerBuilder
             update: self.update,
             document_id: self.document_id,
             tokenizer_builder: self.tokenizer_builder,
+            stop_words: self.stop_words,
         })
     }
 
@@ -174,6 +178,7 @@ pub struct MapSerializer<'a, B> {
     pub document_id: DocumentId,
     pub update: &'a mut DocumentUpdate,
     pub tokenizer_builder: &'a B,
+    pub stop_words: &'a HashSet<String>,
     pub current_key_name: Option<String>,
 }
 
@@ -219,6 +224,7 @@ where B: TokenizerBuilder
                     tokenizer_builder: self.tokenizer_builder,
                     document_id: self.document_id,
                     attribute: attr,
+                    stop_words: self.stop_words,
                 };
                 value.serialize(serializer)?;
             }
@@ -237,6 +243,7 @@ pub struct StructSerializer<'a, B> {
     pub document_id: DocumentId,
     pub update: &'a mut DocumentUpdate,
     pub tokenizer_builder: &'a B,
+    pub stop_words: &'a HashSet<String>,
 }
 
 impl<'a, B> ser::SerializeStruct for StructSerializer<'a, B>
@@ -264,6 +271,7 @@ where B: TokenizerBuilder
                     tokenizer_builder: self.tokenizer_builder,
                     document_id: self.document_id,
                     attribute: attr,
+                    stop_words: self.stop_words,
                 };
                 value.serialize(serializer)?;
             }

src/database/update/builder.rs

@@ -1,6 +1,7 @@
 use std::path::PathBuf;
 use std::error::Error;
 
+use hashbrown::HashSet;
 use serde::Serialize;
 
 use crate::database::serde::serializer::Serializer;
@@ -28,6 +29,7 @@ impl UpdateBuilder {
         &mut self,
         document: T,
         tokenizer_builder: &B,
+        stop_words: &HashSet<String>,
     ) -> Result<DocumentId, SerializerError>
     where T: Serialize,
           B: TokenizerBuilder,
@@ -40,6 +42,7 @@ impl UpdateBuilder {
             document_id: document_id,
             tokenizer_builder: tokenizer_builder,
             update: update,
+            stop_words: stop_words,
         };
 
         document.serialize(serializer)?;

src/rank/criterion/sum_of_typos.rs

@@ -18,7 +18,7 @@ fn sum_matches_typos(matches: &[Match]) -> isize {
     // note that GroupBy will never return an empty group
     // so we can do this assumption safely
     for group in GroupBy::new(matches, match_query_index) {
-        sum_typos += unsafe { group.get_unchecked(0).distance } as isize;
+        sum_typos += unsafe { group.get_unchecked(0).distance as isize };
         number_words += 1;
     }