searchengines-web/MeiliSearch
searchengines-web/MeiliSearch
1
0
Fork 0
mirror of https://github.com/meilisearch/MeiliSearch synced 2024-11-26 23:04:26 +01:00
Code Issues Releases Wiki Activity

Add more comments to WordPrefixPairProximityDocids::execute

Browse Source
This commit is contained in:
Loïc Lecrenier 2022-07-18 15:57:44 +02:00
parent 474500362c
commit 06f3fd8c6d
1 changed files with 17 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

17
milli/src/update/word_prefix_pair_proximity_docids.rs
View File

@ -351,23 +351,34 @@ fn execute_on_word_pairs_and_prefixes<Iter>(
let mut batch = PrefixAndProximityBatch::default(); let mut batch = PrefixAndProximityBatch::default();
let mut prev_word2_start = 0; let mut prev_word2_start = 0;
// Optimisation: the index at the root of the prefix trie where to search for
let mut prefix_search_start = PrefixTrieNodeSearchStart(0); let mut prefix_search_start = PrefixTrieNodeSearchStart(0);
// Optimisation: true if there are no potential prefixes for the current word2 based on its first letter
let mut empty_prefixes = false; let mut empty_prefixes = false;
let mut prefix_buffer = allocations.take_byte_vector(); let mut prefix_buffer = allocations.take_byte_vector();
let mut merge_buffer = allocations.take_byte_vector(); let mut merge_buffer = allocations.take_byte_vector();
while let Some(((word1, word2, proximity), data)) = next_word_pair_proximity(iter)? { while let Some(((word1, word2, proximity), data)) = next_word_pair_proximity(iter)? {
// skip this iteration if the proximity is over the threshold
if proximity > max_proximity { if proximity > max_proximity {
continue; continue;
}; };
let word2_start_different_than_prev = word2[0] != prev_word2_start; let word2_start_different_than_prev = word2[0] != prev_word2_start;
// if there were no potential prefixes for the previous word2 based on its first letter,
// and if the current word2 starts with the same letter, then there is also no potential
// prefixes for the current word2, and we can skip to the next iteration
if empty_prefixes && !word2_start_different_than_prev { if empty_prefixes && !word2_start_different_than_prev {
continue; continue;
} }
// if word1 is different than the previous word1 OR if the start of word2 is different
// than the previous start of word2, then we'll need to flush the batch
let word1_different_than_prev = word1 != batch.word1; let word1_different_than_prev = word1 != batch.word1;
if word1_different_than_prev || word2_start_different_than_prev { if word1_different_than_prev || word2_start_different_than_prev {
batch.flush(allocations, &mut merge_buffer, &mut insert)?; batch.flush(allocations, &mut merge_buffer, &mut insert)?;
// don't forget to reset the value of batch.word1 and prev_word2_start
if word1_different_than_prev { if word1_different_than_prev {
prefix_search_start.0 = 0; prefix_search_start.0 = 0;
batch.word1.clear(); batch.word1.clear();
@ -377,10 +388,12 @@ fn execute_on_word_pairs_and_prefixes<Iter>(
// word2_start_different_than_prev == true // word2_start_different_than_prev == true
prev_word2_start = word2[0]; prev_word2_start = word2[0];
} }
// Optimisation: find the search start in the prefix trie to iterate over the prefixes of word2
empty_prefixes = !prefixes.set_search_start(word2, &mut prefix_search_start); empty_prefixes = !prefixes.set_search_start(word2, &mut prefix_search_start);
} }
if !empty_prefixes { if !empty_prefixes {
// All conditions are satisfied, we can now insert each new prefix of word2 into the batch
prefixes.for_each_prefix_of( prefixes.for_each_prefix_of(
word2, word2,
&mut prefix_buffer, &mut prefix_buffer,
@ -618,6 +631,10 @@ impl PrefixTrieNode {
self.is_end_node = true; self.is_end_node = true;
} }
} }
/// Call the given closure on each prefix of the word contained in the prefix trie.
///
/// The search starts from the given `search_start`.
fn for_each_prefix_of( fn for_each_prefix_of(
&self, &self,
word: &[u8], word: &[u8],