Merge pull request #543 from MarinPostma/aligned-search-crops

adds support for aligned crop in search result
This commit is contained in:
Clément Renault 2020-03-27 18:58:45 +01:00 committed by GitHub
commit c82eed010a
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
3 changed files with 65 additions and 2 deletions

1
Cargo.lock generated
View File

@ -1081,6 +1081,7 @@ dependencies = [
"main_error 0.1.0 (registry+https://github.com/rust-lang/crates.io-index)", "main_error 0.1.0 (registry+https://github.com/rust-lang/crates.io-index)",
"meilisearch-core 0.9.0", "meilisearch-core 0.9.0",
"meilisearch-schema 0.9.0", "meilisearch-schema 0.9.0",
"meilisearch-tokenizer 0.9.0",
"mime 0.3.16 (registry+https://github.com/rust-lang/crates.io-index)", "mime 0.3.16 (registry+https://github.com/rust-lang/crates.io-index)",
"pretty-bytes 0.2.2 (registry+https://github.com/rust-lang/crates.io-index)", "pretty-bytes 0.2.2 (registry+https://github.com/rust-lang/crates.io-index)",
"rand 0.7.2 (registry+https://github.com/rust-lang/crates.io-index)", "rand 0.7.2 (registry+https://github.com/rust-lang/crates.io-index)",

View File

@ -27,6 +27,7 @@ log = "0.4.8"
main_error = "0.1.0" main_error = "0.1.0"
meilisearch-core = { path = "../meilisearch-core", version = "0.9.0" } meilisearch-core = { path = "../meilisearch-core", version = "0.9.0" }
meilisearch-schema = { path = "../meilisearch-schema", version = "0.9.0" } meilisearch-schema = { path = "../meilisearch-schema", version = "0.9.0" }
meilisearch-tokenizer = {path = "../meilisearch-tokenizer", version = "0.9.0"}
mime = "0.3.16" mime = "0.3.16"
pretty-bytes = "0.2.2" pretty-bytes = "0.2.2"
rand = "0.7.2" rand = "0.7.2"

View File

@ -11,6 +11,7 @@ use log::error;
use meilisearch_core::criterion::*; use meilisearch_core::criterion::*;
use meilisearch_core::settings::RankingRule; use meilisearch_core::settings::RankingRule;
use meilisearch_core::{Highlight, Index, MainT, RankedMap}; use meilisearch_core::{Highlight, Index, MainT, RankedMap};
use meilisearch_tokenizer::is_cjk;
use meilisearch_schema::{FieldId, Schema}; use meilisearch_schema::{FieldId, Schema};
use serde::{Deserialize, Serialize}; use serde::{Deserialize, Serialize};
use serde_json::Value; use serde_json::Value;
@ -372,6 +373,30 @@ pub struct SearchResult {
pub query: String, pub query: String,
} }
/// returns the start index and the length on the crop.
fn aligned_crop(text: &str, match_index: usize, context: usize) -> (usize, usize) {
let is_word_component = |c: &char| c.is_alphanumeric() && !is_cjk(*c);
let word_end_index = |mut index| {
if text.chars().nth(index - 1).map_or(false, |c| is_word_component(&c)) {
index += text.chars().skip(index).take_while(is_word_component).count();
}
index
};
if context == 0 {
// count need to be at least 1 for cjk queries to return something
return (match_index, 1 + text.chars().skip(match_index).take_while(is_word_component).count());
}
let start = match match_index.saturating_sub(context) {
n if n == 0 => n,
n => word_end_index(n)
};
let end = word_end_index(start + 2 * context);
(start, end - start)
}
fn crop_text( fn crop_text(
text: &str, text: &str,
matches: impl IntoIterator<Item = Highlight>, matches: impl IntoIterator<Item = Highlight>,
@ -380,9 +405,12 @@ fn crop_text(
let mut matches = matches.into_iter().peekable(); let mut matches = matches.into_iter().peekable();
let char_index = matches.peek().map(|m| m.char_index as usize).unwrap_or(0); let char_index = matches.peek().map(|m| m.char_index as usize).unwrap_or(0);
let start = char_index.saturating_sub(context); let (start, count) = aligned_crop(text, char_index, context);
let text = text.chars().skip(start).take(context * 2).collect();
//TODO do something about the double allocation
let text = text.chars().skip(start).take(count).collect::<String>().trim().to_string();
// update matches index to match the new cropped text
let matches = matches let matches = matches
.take_while(|m| (m.char_index as usize) + (m.char_length as usize) <= start + (context * 2)) .take_while(|m| (m.char_index as usize) + (m.char_length as usize) <= start + (context * 2))
.map(|match_| Highlight { .map(|match_| Highlight {
@ -504,6 +532,39 @@ fn calculate_highlights(
mod tests { mod tests {
use super::*; use super::*;
#[test]
fn aligned_crops() {
let text = r#"En ce début de trentième millénaire, l'Empire n'a jamais été aussi puissant, aussi étendu à travers toute la galaxie. C'est dans sa capitale, Trantor, que l'éminent savant Hari Seldon invente la psychohistoire, une science toute nouvelle, à base de psychologie et de mathématiques, qui lui permet de prédire l'avenir... C'est-à-dire l'effondrement de l'Empire d'ici cinq siècles et au-delà, trente mille années de chaos et de ténèbres. Pour empêcher cette catastrophe et sauver la civilisation, Seldon crée la Fondation."#;
// simple test
let (start, length) = aligned_crop(&text, 6, 2);
let cropped = text.chars().skip(start).take(length).collect::<String>().trim().to_string();
assert_eq!("début", cropped);
// first word test
let (start, length) = aligned_crop(&text, 0, 1);
let cropped = text.chars().skip(start).take(length).collect::<String>().trim().to_string();
assert_eq!("En", cropped);
// last word test
let (start, length) = aligned_crop(&text, 510, 2);
let cropped = text.chars().skip(start).take(length).collect::<String>().trim().to_string();
assert_eq!("Fondation", cropped);
// CJK tests
let text = "this isのス foo myタイリ test";
// mixed charset
let (start, length) = aligned_crop(&text, 5, 3);
let cropped = text.chars().skip(start).take(length).collect::<String>().trim().to_string();
assert_eq!("isのス", cropped);
// split regular word / CJK word, no space
let (start, length) = aligned_crop(&text, 7, 1);
let cropped = text.chars().skip(start).take(length).collect::<String>().trim().to_string();
assert_eq!("のス", cropped);
}
#[test] #[test]
fn calculate_highlights() { fn calculate_highlights() {
let data = r#"{ let data = r#"{