searchengines-web/MeiliSearch
searchengines-web/MeiliSearch
1
0
Fork 0
mirror of https://github.com/meilisearch/MeiliSearch synced 2024-11-25 22:34:28 +01:00
Code Issues Releases Wiki Activity

Fix panics and issues in exactness graph ranking rule

Browse Source
This commit is contained in:
Louis Dureuil 2023-04-05 14:43:16 +02:00
parent 337e75b0e4
commit f7ecea142e
No known key found for this signature in database
2 changed files with 22 additions and 31 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

48
milli/src/search/new/ranking_rule_graph/exactness/mod.rs
View File

@ -1,24 +1,11 @@
use heed::BytesDecode;
use roaring::RoaringBitmap;
use super::{ComputedCondition, DeadEndsCache, RankingRuleGraph, RankingRuleGraphTrait};
use crate::search::new::interner::{DedupInterner, Interned, MappedInterner};
use crate::search::new::query_graph::{QueryGraph, QueryNode};
use crate::search::new::query_term::{ExactTerm, LocatedQueryTermSubset};
use crate::{CboRoaringBitmapCodec, Result, SearchContext, SearchLogger};
/// - Exactness as first ranking rule: TermsMatchingStrategy? prefer a document that matches 1 word exactly and no other
/// word than a doc that matches 9 words non exactly but none exactly
/// - `TermsMatchingStrategy` as a word + exactness optimization: we could consider
///
/// "naive vision"
/// condition from one node to another:
/// - word exactly present: cost 0
/// - word typo/ngram/prefix/missing: cost 1, not remove from query graph, edge btwn the two nodes, return the universe without condition when resolving, destination query term is inside
///
/// Three strategies:
/// 1. ExactAttribute: word position / word_fid_docid
/// 2. AttributeStart:
/// 3. AttributeContainsExact => implementable via `RankingRuleGraphTrait`
use crate::{Result, RoaringBitmapCodec, SearchContext, SearchLogger};
#[derive(Clone, PartialEq, Eq, Hash)]
pub enum ExactnessCondition {
@ -42,7 +29,7 @@ fn compute_docids(
ExactTerm::Phrase(phrase) => ctx.get_phrase_docids(phrase)?.clone(),
ExactTerm::Word(word) => {
if let Some(word_candidates) = ctx.get_db_word_docids(word)? {
CboRoaringBitmapCodec::deserialize_from(word_candidates)?
RoaringBitmapCodec::bytes_decode(word_candidates).ok_or(heed::Error::Decoding)?
} else {
return Ok(Default::default());
}
@ -86,22 +73,29 @@ impl RankingRuleGraphTrait for ExactnessGraph {
let skip_condition = ExactnessCondition::Skip(dest_node.clone());
let skip_condition = conditions_interner.insert(skip_condition);
Ok(vec![(0, exact_condition), (1, skip_condition)])
Ok(vec![(0, exact_condition), (dest_node.term_ids.len() as u32, skip_condition)])
}
fn log_state(
graph: &RankingRuleGraph<Self>,
paths: &[Vec<Interned<Self::Condition>>],
dead_ends_cache: &DeadEndsCache<Self::Condition>,
universe: &RoaringBitmap,
costs: &MappedInterner<QueryNode, Vec<u64>>,
cost: u64,
logger: &mut dyn SearchLogger<QueryGraph>,
_graph: &RankingRuleGraph<Self>,
_paths: &[Vec<Interned<Self::Condition>>],
_dead_ends_cache: &DeadEndsCache<Self::Condition>,
_niverse: &RoaringBitmap,
_costs: &MappedInterner<QueryNode, Vec<u64>>,
_cost: u64,
_logger: &mut dyn SearchLogger<QueryGraph>,
) {
todo!()
}
fn label_for_condition(ctx: &mut SearchContext, condition: &Self::Condition) -> Result<String> {
todo!()
fn label_for_condition(
_ctx: &mut SearchContext,
condition: &Self::Condition,
) -> Result<String> {
Ok(match condition {
ExactnessCondition::ExactInAttribute(_) => "exact",
ExactnessCondition::Skip(_) => "skip",
}
.to_owned())
}
}

5
milli/src/update/index_documents/extract/extract_word_position_docids.rs
View File

@ -7,10 +7,7 @@ use super::helpers::{
};
use crate::error::SerializationError;
use crate::index::db_name::DOCID_WORD_POSITIONS;
use crate::{
absolute_from_relative_position, bucketed_position, relative_from_absolute_position,
DocumentId, Result,
};
use crate::{bucketed_position, relative_from_absolute_position, DocumentId, Result};
/// Extracts the word positions and the documents ids where this word appear.
///