Merge pull request #93 from Kerollmops/slice-group-by

Use the GroupBy/Mut Traits of the slice-group-by library

Cargo.toml

@@ -19,6 +19,7 @@ sdset = "0.3"
 serde = "1.0"
 serde_derive = "1.0"
 serde_json = { version = "1.0", features = ["preserve_order"] }
+slice-group-by = "0.2"
 unidecode = "0.3"
 
 [dependencies.toml]
@@ -30,16 +31,12 @@ rev = "0372ba6"
 git = "https://github.com/pingcap/rust-rocksdb.git"
 rev = "306e201"
 
-[dependencies.group-by]
-git = "https://github.com/Kerollmops/group-by.git"
-rev = "5a113fe"
-
 [features]
 default = ["simd"]
 i128 = ["bincode/i128", "byteorder/i128"]
 portable = ["rocksdb/portable"]
 simd = ["rocksdb/sse"]
-nightly = ["hashbrown/nightly", "group-by/nightly"]
+nightly = ["hashbrown/nightly", "slice-group-by/nightly"]
 
 [dev-dependencies]
 csv = "1.0"

src/rank/criterion/exact.rs

@@ -2,7 +2,7 @@ use std::cmp::Ordering;
 use std::ops::Deref;
 
 use rocksdb::DB;
-use group_by::GroupBy;
+use slice_group_by::GroupBy;
 
 use crate::rank::{match_query_index, Document};
 use crate::rank::criterion::Criterion;
@@ -16,7 +16,7 @@ fn contains_exact(matches: &&[Match]) -> bool {
 
 #[inline]
 fn number_exact_matches(matches: &[Match]) -> usize {
-    GroupBy::new(matches, match_query_index).filter(contains_exact).count()
+    matches.linear_group_by(match_query_index).filter(contains_exact).count()
 }
 
 #[derive(Debug, Clone, Copy)]

src/rank/criterion/number_of_words.rs

@@ -2,7 +2,7 @@ use std::cmp::Ordering;
 use std::ops::Deref;
 
 use rocksdb::DB;
-use group_by::GroupBy;
+use slice_group_by::GroupBy;
 
 use crate::rank::{match_query_index, Document};
 use crate::rank::criterion::Criterion;
@@ -11,7 +11,7 @@ use crate::Match;
 
 #[inline]
 fn number_of_query_words(matches: &[Match]) -> usize {
-    GroupBy::new(matches, match_query_index).count()
+    matches.linear_group_by(match_query_index).count()
 }
 
 #[derive(Debug, Clone, Copy)]

src/rank/criterion/sum_of_typos.rs

@@ -2,8 +2,7 @@ use std::cmp::Ordering;
 use std::ops::Deref;
 
 use rocksdb::DB;
-
-use group_by::GroupBy;
+use slice_group_by::GroupBy;
 
 use crate::rank::{match_query_index, Document};
 use crate::rank::criterion::Criterion;
@@ -17,7 +16,7 @@ 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) {
+    for group in matches.linear_group_by(match_query_index) {
         sum_typos += unsafe { group.get_unchecked(0).distance as isize };
         number_words += 1;
     }

src/rank/criterion/sum_of_words_attribute.rs

@@ -2,7 +2,7 @@ use std::cmp::Ordering;
 use std::ops::Deref;
 
 use rocksdb::DB;
-use group_by::GroupBy;
+use slice_group_by::GroupBy;
 
 use crate::database::DatabaseView;
 use crate::rank::{match_query_index, Document};
@@ -13,7 +13,7 @@ use crate::Match;
 fn sum_matches_attributes(matches: &[Match]) -> usize {
     // note that GroupBy will never return an empty group
     // so we can do this assumption safely
-    GroupBy::new(matches, match_query_index).map(|group| {
+    matches.linear_group_by(match_query_index).map(|group| {
         unsafe { group.get_unchecked(0).attribute.attribute() as usize }
     }).sum()
 }

src/rank/criterion/sum_of_words_position.rs

@@ -2,7 +2,7 @@ use std::cmp::Ordering;
 use std::ops::Deref;
 
 use rocksdb::DB;
-use group_by::GroupBy;
+use slice_group_by::GroupBy;
 
 use crate::database::DatabaseView;
 use crate::rank::{match_query_index, Document};
@@ -13,7 +13,7 @@ use crate::Match;
 fn sum_matches_attribute_index(matches: &[Match]) -> usize {
     // note that GroupBy will never return an empty group
     // so we can do this assumption safely
-    GroupBy::new(matches, match_query_index).map(|group| {
+    matches.linear_group_by(match_query_index).map(|group| {
         unsafe { group.get_unchecked(0).attribute.word_index() as usize }
     }).sum()
 }

src/rank/criterion/words_proximity.rs

@@ -2,7 +2,7 @@ use std::cmp::{self, Ordering};
 use std::ops::Deref;
 
 use rocksdb::DB;
-use group_by::GroupBy;
+use slice_group_by::GroupBy;
 
 use crate::rank::{match_query_index, Document};
 use crate::rank::criterion::Criterion;
@@ -36,7 +36,7 @@ fn min_proximity(lhs: &[Match], rhs: &[Match]) -> u32 {
 
 fn matches_proximity(matches: &[Match]) -> u32 {
     let mut proximity = 0;
-    let mut iter = GroupBy::new(matches, match_query_index);
+    let mut iter = matches.linear_group_by(match_query_index);
 
     // iterate over groups by windows of size 2
     let mut last = iter.next();

src/rank/query_builder.rs

@@ -4,7 +4,7 @@ use std::error::Error;
 use std::hash::Hash;
 use std::rc::Rc;
 
-use group_by::BinaryGroupByMut;
+use slice_group_by::GroupByMut;
 use hashbrown::HashMap;
 use fst::Streamer;
 use rocksdb::DB;
@@ -164,7 +164,7 @@ where D: Deref<Target=DB>,
                 });
                 info!("criterion {} sort took {}", ci, elapsed);
 
-                for group in BinaryGroupByMut::new(group, |a, b| criterion.eq(a, b, view)) {
+                for group in group.binary_group_by_mut(|a, b| criterion.eq(a, b, view)) {
                     documents_seen += group.len();
                     groups.push(group);
 
@@ -241,7 +241,7 @@ where D: Deref<Target=DB>,
 
                 group.sort_unstable_by(|a, b| criterion.evaluate(a, b, view));
 
-                for group in BinaryGroupByMut::new(group, |a, b| criterion.eq(a, b, view)) {
+                for group in group.binary_group_by_mut(|a, b| criterion.eq(a, b, view)) {
                     // we must compute the real distinguished len of this sub-group
                     for document in group.iter() {
                         let filter_accepted = match &self.inner.filter {