From 66c2c82a18614208b5b47f9597aa8a8509d1e697 Mon Sep 17 00:00:00 2001 From: ManyTheFish Date: Thu, 28 Sep 2023 10:45:25 +0200 Subject: [PATCH] Split wpp in several sorters --- .../extract_word_pair_proximity_docids.rs | 66 ++++++++++++------- 1 file changed, 43 insertions(+), 23 deletions(-) diff --git a/milli/src/update/index_documents/extract/extract_word_pair_proximity_docids.rs b/milli/src/update/index_documents/extract/extract_word_pair_proximity_docids.rs index 847da01c5..70865acbe 100644 --- a/milli/src/update/index_documents/extract/extract_word_pair_proximity_docids.rs +++ b/milli/src/update/index_documents/extract/extract_word_pair_proximity_docids.rs @@ -6,8 +6,8 @@ use std::{cmp, io}; use obkv::KvReaderU16; use super::helpers::{ - create_sorter, merge_cbo_roaring_bitmaps, sorter_into_reader, try_split_array_at, - GrenadParameters, MergeFn, + create_sorter, create_writer, merge_cbo_roaring_bitmaps, sorter_into_reader, + try_split_array_at, writer_into_reader, GrenadParameters, MergeFn, }; use crate::error::SerializationError; use crate::index::db_name::DOCID_WORD_POSITIONS; @@ -27,14 +27,19 @@ pub fn extract_word_pair_proximity_docids( let max_memory = indexer.max_memory_by_thread(); - let mut word_pair_proximity_docids_sorter = create_sorter( - grenad::SortAlgorithm::Unstable, - merge_cbo_roaring_bitmaps, - indexer.chunk_compression_type, - indexer.chunk_compression_level, - indexer.max_nb_chunks, - max_memory.map(|m| m / 2), - ); + let mut word_pair_proximity_docids_sorters: Vec<_> = (1..MAX_DISTANCE) + .into_iter() + .map(|_| { + create_sorter( + grenad::SortAlgorithm::Unstable, + merge_cbo_roaring_bitmaps, + indexer.chunk_compression_type, + indexer.chunk_compression_level, + indexer.max_nb_chunks, + max_memory.map(|m| m / MAX_DISTANCE as usize), + ) + }) + .collect(); let mut word_positions: VecDeque<(String, u16)> = VecDeque::with_capacity(MAX_DISTANCE as usize); @@ -49,6 +54,7 @@ pub fn extract_word_pair_proximity_docids( // if we change document, we fill the sorter if current_document_id.map_or(false, |id| id != document_id) { + puffin::profile_scope!("Document into sorter"); while !word_positions.is_empty() { word_positions_into_word_pair_proximity( &mut word_positions, @@ -59,7 +65,7 @@ pub fn extract_word_pair_proximity_docids( document_word_positions_into_sorter( current_document_id.unwrap(), &word_pair_proximity, - &mut word_pair_proximity_docids_sorter, + &mut word_pair_proximity_docids_sorters, )?; word_pair_proximity.clear(); word_positions.clear(); @@ -85,6 +91,7 @@ pub fn extract_word_pair_proximity_docids( } if let Some(document_id) = current_document_id { + puffin::profile_scope!("Final document into sorter"); while !word_positions.is_empty() { word_positions_into_word_pair_proximity(&mut word_positions, &mut word_pair_proximity)?; } @@ -92,11 +99,23 @@ pub fn extract_word_pair_proximity_docids( document_word_positions_into_sorter( document_id, &word_pair_proximity, - &mut word_pair_proximity_docids_sorter, + &mut word_pair_proximity_docids_sorters, )?; } + { + puffin::profile_scope!("sorter_into_reader"); + let mut writer = create_writer( + indexer.chunk_compression_type, + indexer.chunk_compression_level, + tempfile::tempfile()?, + ); - sorter_into_reader(word_pair_proximity_docids_sorter, indexer) + for sorter in word_pair_proximity_docids_sorters { + sorter.write_into_stream_writer(&mut writer)?; + } + + writer_into_reader(writer) + } } /// Fills the list of all pairs of words with the shortest proximity between 1 and 7 inclusive. @@ -106,9 +125,8 @@ pub fn extract_word_pair_proximity_docids( fn document_word_positions_into_sorter( document_id: DocumentId, word_pair_proximity: &HashMap<(String, String), u8>, - word_pair_proximity_docids_sorter: &mut grenad::Sorter, + word_pair_proximity_docids_sorters: &mut Vec>, ) -> Result<()> { - puffin::profile_function!(); let mut key_buffer = Vec::new(); for ((w1, w2), prox) in word_pair_proximity { key_buffer.clear(); @@ -117,7 +135,8 @@ fn document_word_positions_into_sorter( key_buffer.push(0); key_buffer.extend_from_slice(w2.as_bytes()); - word_pair_proximity_docids_sorter.insert(&key_buffer, document_id.to_ne_bytes())?; + word_pair_proximity_docids_sorters[*prox as usize - 1] + .insert(&key_buffer, document_id.to_ne_bytes())?; } Ok(()) @@ -127,16 +146,17 @@ fn word_positions_into_word_pair_proximity( word_positions: &mut VecDeque<(String, u16)>, word_pair_proximity: &mut HashMap<(String, String), u8>, ) -> Result<()> { - puffin::profile_function!(); let (head_word, head_position) = word_positions.pop_front().unwrap(); for (word, position) in word_positions.iter() { let prox = index_proximity(head_position as u32, *position as u32) as u8; - word_pair_proximity - .entry((head_word.clone(), word.clone())) - .and_modify(|p| { - *p = cmp::min(*p, prox); - }) - .or_insert(prox); + if prox > 0 && prox < MAX_DISTANCE as u8 { + word_pair_proximity + .entry((head_word.clone(), word.clone())) + .and_modify(|p| { + *p = cmp::min(*p, prox); + }) + .or_insert(prox); + } } Ok(()) }