snowykami/meilisearch
1
0
Fork 0
mirror of https://github.com/meilisearch/meilisearch.git synced 2024-11-27 04:25:06 +08:00
Code Issues Actions 4 Packages Projects Releases Wiki Activity

Avoid iterating on big databases when useless

Browse Source
This commit is contained in:
Kerollmops 2022-03-01 18:02:12 +01:00
parent a8d28e364d
commit 1ae13c1374
No known key found for this signature in database
GPG Key ID: 92ADA4E935E71FA4
3 changed files with 111 additions and 98 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
milli/src/update/word_prefix_docids.rs
View File

@ -50,6 +50,7 @@ impl<'t, 'u, 'i> WordPrefixDocids<'t, 'u, 'i> {
self.max_memory,
);
if !common_prefix_fst_words.is_empty() {
let mut new_word_docids_iter = new_word_docids.into_cursor()?;
let mut current_prefixes: Option<&&[String]> = None;
let mut prefixes_cache = HashMap::new();
@ -70,7 +71,8 @@ impl<'t, 'u, 'i> WordPrefixDocids<'t, 'u, 'i> {
match prefixes_cache.get_mut(prefix.as_bytes()) {
Some(value) => value.push(data.to_owned()),
None => {
prefixes_cache.insert(prefix.clone().into(), vec![data.to_owned()]);
prefixes_cache
.insert(prefix.clone().into(), vec![data.to_owned()]);
}
}
}
@ -79,6 +81,7 @@ impl<'t, 'u, 'i> WordPrefixDocids<'t, 'u, 'i> {
}
write_prefixes_in_sorter(&mut prefixes_cache, &mut prefix_docids_sorter)?;
}
// We fetch the docids associated to the newly added word prefix fst only.
let db = self.index.word_docids.remap_data_type::<ByteSlice>();

16
milli/src/update/word_prefix_pair_proximity_docids.rs
View File

@ -83,13 +83,15 @@ impl<'t, 'u, 'i> WordPrefixPairProximityDocids<'t, 'u, 'i> {
self.max_memory,
);
if !common_prefix_fst_words.is_empty() {
// We compute the prefix docids associated with the common prefixes between
// the old and new word prefix fst.
let mut buffer = Vec::new();
let mut current_prefixes: Option<&&[String]> = None;
let mut prefixes_cache = HashMap::new();
while let Some((key, data)) = new_wppd_iter.move_on_next()? {
let (w1, w2, prox) = StrStrU8Codec::bytes_decode(key).ok_or(heed::Error::Decoding)?;
let (w1, w2, prox) =
StrStrU8Codec::bytes_decode(key).ok_or(heed::Error::Decoding)?;
if prox > self.max_proximity {
continue;
}
@ -112,11 +114,16 @@ impl<'t, 'u, 'i> WordPrefixPairProximityDocids<'t, 'u, 'i> {
&mut prefixes_cache,
&mut word_prefix_pair_proximity_docids_sorter,
)?;
}
if !new_prefix_fst_words.is_empty() {
// We compute the prefix docids associated with the newly added prefixes
// in the new word prefix fst.
let mut db_iter =
self.index.word_pair_proximity_docids.remap_data_type::<ByteSlice>().iter(self.wtxn)?;
let mut db_iter = self
.index
.word_pair_proximity_docids
.remap_data_type::<ByteSlice>()
.iter(self.wtxn)?;
let mut buffer = Vec::new();
let mut current_prefixes: Option<&&[String]> = None;
@ -144,8 +151,7 @@ impl<'t, 'u, 'i> WordPrefixPairProximityDocids<'t, 'u, 'i> {
&mut prefixes_cache,
&mut word_prefix_pair_proximity_docids_sorter,
)?;
drop(db_iter);
}
// All of the word prefix pairs in the database that have a w2
// that is contained in the `suppr_pw` set must be removed as well.

6
milli/src/update/words_prefix_position_docids.rs
View File

@ -74,6 +74,7 @@ impl<'t, 'u, 'i> WordPrefixPositionDocids<'t, 'u, 'i> {
let mut new_word_position_docids_iter = new_word_position_docids.into_cursor()?;
if !common_prefix_fst_words.is_empty() {
// We fetch all the new common prefixes between the previous and new prefix fst.
let mut buffer = Vec::new();
let mut current_prefixes: Option<&&[String]> = None;
@ -88,7 +89,9 @@ impl<'t, 'u, 'i> WordPrefixPositionDocids<'t, 'u, 'i> {
&mut prefixes_cache,
&mut prefix_position_docids_sorter,
)?;
common_prefix_fst_words.iter().find(|prefixes| word.starts_with(&prefixes[0]))
common_prefix_fst_words
.iter()
.find(|prefixes| word.starts_with(&prefixes[0]))
}
};
@ -110,6 +113,7 @@ impl<'t, 'u, 'i> WordPrefixPositionDocids<'t, 'u, 'i> {
}
write_prefixes_in_sorter(&mut prefixes_cache, &mut prefix_position_docids_sorter)?;
}
// We fetch the docids associated to the newly added word prefix fst only.
let db = self.index.word_position_docids.remap_data_type::<ByteSlice>();