searchengines-web/MeiliSearch
searchengines-web/MeiliSearch
1
0
Fork 0
mirror of https://github.com/meilisearch/MeiliSearch synced 2024-12-23 21:20:24 +01:00
Code Issues Releases Wiki Activity

Remove unused files

Browse Source
This commit is contained in:
Kerollmops 2021-03-02 11:02:09 +01:00
parent f118d7e067
commit 6bf6b40495
No known key found for this signature in database
GPG Key ID: 92ADA4E935E71FA4
3 changed files with 0 additions and 382 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
milli/src/lib.rs
View File

@ -3,8 +3,6 @@
mod criterion;
mod external_documents_ids;
mod fields_ids_map;
mod mdfs;
mod query_tokens;
mod search;
mod update_store;
pub mod facet;

163
milli/src/mdfs.rs
View File

@ -1,163 +0,0 @@
use std::collections::hash_map::Entry::{Occupied, Vacant};
use std::collections::HashMap;
use std::mem;
use roaring::RoaringBitmap;
use crate::Index;
/// A mana depth first search implementation.
pub struct Mdfs<'a> {
index: &'a Index,
rtxn: &'a heed::RoTxn<'a>,
words: &'a [(HashMap<String, (u8, RoaringBitmap)>, RoaringBitmap)],
union_cache: HashMap<(usize, u8), RoaringBitmap>,
candidates: RoaringBitmap,
mana: u32,
max_mana: u32,
}
impl<'a> Mdfs<'a> {
pub fn new(
index: &'a Index,
rtxn: &'a heed::RoTxn,
words: &'a [(HashMap<String, (u8, RoaringBitmap)>, RoaringBitmap)],
candidates: RoaringBitmap,
) -> Mdfs<'a>
{
// Compute the number of pairs (windows) we have for this list of words.
let mana = words.len().saturating_sub(1) as u32;
let max_mana = mana * 8;
Mdfs { index, rtxn, words, union_cache: HashMap::new(), candidates, mana, max_mana }
}
}
impl<'a> Iterator for Mdfs<'a> {
type Item = anyhow::Result<(u32, RoaringBitmap)>;
fn next(&mut self) -> Option<Self::Item> {
// If there is less or only one word therefore the only
// possible documents that we can return are the candidates.
if self.words.len() <= 1 {
if self.candidates.is_empty() { return None }
return Some(Ok((0, mem::take(&mut self.candidates))));
}
while self.mana <= self.max_mana {
let mut answer = RoaringBitmap::new();
let result = mdfs_step(
&self.index,
&self.rtxn,
self.mana,
self.words,
&self.candidates,
&self.candidates,
&mut self.union_cache,
&mut answer,
);
match result {
Ok(()) => {
// We always increase the mana for the next loop.
let proximity = self.mana;
self.mana += 1;
// If no documents were found we must not return and continue
// the search with more mana.
if !answer.is_empty() {
// We remove the answered documents from the list of
// candidates to be sure we don't search for them again.
self.candidates.difference_with(&answer);
// We return the answer.
return Some(Ok((proximity, answer)));
}
},
Err(e) => return Some(Err(e)),
}
}
None
}
}
fn mdfs_step(
index: &Index,
rtxn: &heed::RoTxn,
mana: u32,
words: &[(HashMap<String, (u8, RoaringBitmap)>, RoaringBitmap)],
candidates: &RoaringBitmap,
parent_docids: &RoaringBitmap,
union_cache: &mut HashMap<(usize, u8), RoaringBitmap>,
answer: &mut RoaringBitmap,
) -> anyhow::Result<()>
{
use std::cmp::{min, max};
let (words1, words2) = (&words[0].0, &words[1].0);
let pairs = words_pair_combinations(words1, words2);
let tail = &words[1..];
let nb_children = tail.len() as u32 - 1;
// The minimum amount of mana that you must consume is at least 1 and the
// amount of mana that your children can consume. Because the last child must
// consume the remaining mana, it is mandatory that there not too much at the end.
let min_proximity = max(1, mana.saturating_sub(nb_children * 8)) as u8;
// The maximum amount of mana that you can use is 8 or the remaining amount of
// mana minus your children, as you can't just consume all the mana,
// your children must have at least 1 mana.
let max_proximity = min(8, mana - nb_children) as u8;
for proximity in min_proximity..=max_proximity {
let mut docids = match union_cache.entry((words.len(), proximity)) {
Occupied(entry) => entry.get().clone(),
Vacant(entry) => {
let mut docids = RoaringBitmap::new();
if proximity == 8 {
docids = candidates.clone();
} else {
for (w1, w2) in pairs.iter().cloned() {
let key = (w1, w2, proximity);
if let Some(di) = index.word_pair_proximity_docids.get(rtxn, &key)? {
docids.union_with(&di);
}
}
}
entry.insert(docids).clone()
}
};
// We must be sure that we only return docids that are present in the candidates.
docids.intersect_with(parent_docids);
if !docids.is_empty() {
let mana = mana.checked_sub(proximity as u32).unwrap();
if tail.len() < 2 {
// We are the last pair, we return without recuring as we don't have any child.
answer.union_with(&docids);
return Ok(());
} else {
return mdfs_step(index, rtxn, mana, tail, candidates, &docids, union_cache, answer);
}
}
}
Ok(())
}
fn words_pair_combinations<'h>(
w1: &'h HashMap<String, (u8, RoaringBitmap)>,
w2: &'h HashMap<String, (u8, RoaringBitmap)>,
) -> Vec<(&'h str, &'h str)>
{
let mut pairs = Vec::new();
for (w1, (_typos, docids1)) in w1 {
for (w2, (_typos, docids2)) in w2 {
if !docids1.is_disjoint(&docids2) {
pairs.push((w1.as_str(), w2.as_str()));
}
}
}
pairs
}

217
milli/src/query_tokens.rs
View File

@ -1,217 +0,0 @@
use meilisearch_tokenizer::{Token, TokenKind};
#[derive(Debug)]
enum State {
Free,
Quoted,
}
impl State {
fn swap(&mut self) {
match self {
State::Quoted => *self = State::Free,
State::Free => *self = State::Quoted,
}
}
}
#[derive(Debug, PartialEq, Eq)]
pub enum QueryToken<'a> {
Free(Token<'a>),
Quoted(Token<'a>),
}
pub fn query_tokens<'a>(mut tokens: impl Iterator<Item = Token<'a>>) -> impl Iterator<Item = QueryToken<'a>> {
let mut state = State::Free;
let f = move || {
loop {
let token = tokens.next()?;
match token.kind() {
_ if token.text().trim() == "\"" => state.swap(),
TokenKind::Word => {
let token = match state {
State::Quoted => QueryToken::Quoted(token),
State::Free => QueryToken::Free(token),
};
return Some(token);
},
_ => (),
}
}
};
std::iter::from_fn(f)
}
#[cfg(test)]
mod tests {
use super::*;
use QueryToken::{Quoted, Free};
use meilisearch_tokenizer::{Analyzer, AnalyzerConfig};
use fst::Set;
macro_rules! assert_eq_query_token {
($test:expr, Quoted($val:literal)) => {
match $test {
Quoted(val) => assert_eq!(val.text(), $val),
Free(val) => panic!("expected Quoted(\"{}\"), found Free(\"{}\")", $val, val.text()),
}
};
($test:expr, Free($val:literal)) => {
match $test {
Quoted(val) => panic!("expected Free(\"{}\"), found Quoted(\"{}\")", $val, val.text()),
Free(val) => assert_eq!(val.text(), $val),
}
};
}
#[test]
fn empty() {
let stop_words = Set::default();
let analyzer = Analyzer::new(AnalyzerConfig::default_with_stopwords(&stop_words));
let query = "";
let analyzed = analyzer.analyze(query);
let tokens = analyzed.tokens();
let mut iter = query_tokens(tokens);
assert!(iter.next().is_none());
let query = " ";
let analyzed = analyzer.analyze(query);
let tokens = analyzed.tokens();
let mut iter = query_tokens(tokens);
assert!(iter.next().is_none());
}
#[test]
fn one_quoted_string() {
let stop_words = Set::default();
let analyzer = Analyzer::new(AnalyzerConfig::default_with_stopwords(&stop_words));
let query = "\"hello\"";
let analyzed = analyzer.analyze(query);
let tokens = analyzed.tokens();
let mut iter = query_tokens(tokens);
assert_eq_query_token!(iter.next().unwrap(), Quoted("hello"));
assert!(iter.next().is_none());
}
#[test]
fn one_pending_quoted_string() {
let stop_words = Set::default();
let analyzer = Analyzer::new(AnalyzerConfig::default_with_stopwords(&stop_words));
let query = "\"hello";
let analyzed = analyzer.analyze(query);
let tokens = analyzed.tokens();
let mut iter = query_tokens(tokens);
assert_eq_query_token!(iter.next().unwrap(), Quoted("hello"));
assert!(iter.next().is_none());
}
#[test]
fn one_non_quoted_string() {
let stop_words = Set::default();
let analyzer = Analyzer::new(AnalyzerConfig::default_with_stopwords(&stop_words));
let query = "hello";
let analyzed = analyzer.analyze(query);
let tokens = analyzed.tokens();
let mut iter = query_tokens(tokens);
assert_eq_query_token!(iter.next().unwrap(), Free("hello"));
assert!(iter.next().is_none());
}
#[test]
fn quoted_directly_followed_by_free_strings() {
let stop_words = Set::default();
let analyzer = Analyzer::new(AnalyzerConfig::default_with_stopwords(&stop_words));
let query = "\"hello\"world";
let analyzed = analyzer.analyze(query);
let tokens = analyzed.tokens();
let mut iter = query_tokens(tokens);
assert_eq_query_token!(iter.next().unwrap(), Quoted("hello"));
assert_eq_query_token!(iter.next().unwrap(), Free("world"));
assert!(iter.next().is_none());
}
#[test]
fn free_directly_followed_by_quoted_strings() {
let stop_words = Set::default();
let analyzer = Analyzer::new(AnalyzerConfig::default_with_stopwords(&stop_words));
let query = "hello\"world\"";
let analyzed = analyzer.analyze(query);
let tokens = analyzed.tokens();
let mut iter = query_tokens(tokens);
assert_eq_query_token!(iter.next().unwrap(), Free("hello"));
assert_eq_query_token!(iter.next().unwrap(), Quoted("world"));
assert!(iter.next().is_none());
}
#[test]
fn free_followed_by_quoted_strings() {
let stop_words = Set::default();
let analyzer = Analyzer::new(AnalyzerConfig::default_with_stopwords(&stop_words));
let query = "hello \"world\"";
let analyzed = analyzer.analyze(query);
let tokens = analyzed.tokens();
let mut iter = query_tokens(tokens);
assert_eq_query_token!(iter.next().unwrap(), Free("hello"));
assert_eq_query_token!(iter.next().unwrap(), Quoted("world"));
assert!(iter.next().is_none());
}
#[test]
fn multiple_spaces_separated_strings() {
let stop_words = Set::default();
let analyzer = Analyzer::new(AnalyzerConfig::default_with_stopwords(&stop_words));
let query = "hello world ";
let analyzed = analyzer.analyze(query);
let tokens = analyzed.tokens();
let mut iter = query_tokens(tokens);
assert_eq_query_token!(iter.next().unwrap(), Free("hello"));
assert_eq_query_token!(iter.next().unwrap(), Free("world"));
assert!(iter.next().is_none());
}
#[test]
fn multi_interleaved_quoted_free_strings() {
let stop_words = Set::default();
let analyzer = Analyzer::new(AnalyzerConfig::default_with_stopwords(&stop_words));
let query = "hello \"world\" coucou \"monde\"";
let analyzed = analyzer.analyze(query);
let tokens = analyzed.tokens();
let mut iter = query_tokens(tokens);
assert_eq_query_token!(iter.next().unwrap(), Free("hello"));
assert_eq_query_token!(iter.next().unwrap(), Quoted("world"));
assert_eq_query_token!(iter.next().unwrap(), Free("coucou"));
assert_eq_query_token!(iter.next().unwrap(), Quoted("monde"));
assert!(iter.next().is_none());
}
#[test]
fn multi_quoted_strings() {
let stop_words = Set::default();
let analyzer = Analyzer::new(AnalyzerConfig::default_with_stopwords(&stop_words));
let query = "\"hello world\" coucou \"monde est beau\"";
let analyzed = analyzer.analyze(query);
let tokens = analyzed.tokens();
let mut iter = query_tokens(tokens);
assert_eq_query_token!(iter.next().unwrap(), Quoted("hello"));
assert_eq_query_token!(iter.next().unwrap(), Quoted("world"));
assert_eq_query_token!(iter.next().unwrap(), Free("coucou"));
assert_eq_query_token!(iter.next().unwrap(), Quoted("monde"));
assert_eq_query_token!(iter.next().unwrap(), Quoted("est"));
assert_eq_query_token!(iter.next().unwrap(), Quoted("beau"));
assert!(iter.next().is_none());
}
#[test]
fn chinese() {
let stop_words = Set::default();
let analyzer = Analyzer::new(AnalyzerConfig::default_with_stopwords(&stop_words));
let query = "汽车男生";
let analyzed = analyzer.analyze(query);
let tokens = analyzed.tokens();
let mut iter = query_tokens(tokens);
assert_eq_query_token!(iter.next().unwrap(), Free("汽车"));
assert_eq_query_token!(iter.next().unwrap(), Free("男生"));
assert!(iter.next().is_none());
}
}