Compare commits

...

8 Commits

Author SHA1 Message Date
Many the fish
af8279a71e
Merge 510ca99996 into 057fcb3993 2024-11-20 09:09:06 +03:00
Lukas Kalbertodt
057fcb3993
Add indices field to _matchesPosition to specify where in an array a match comes from (#5005)
Some checks are pending
Indexing bench (push) / Run and upload benchmarks (push) Waiting to run
Benchmarks of indexing (push) / Run and upload benchmarks (push) Waiting to run
Benchmarks of search for geo (push) / Run and upload benchmarks (push) Waiting to run
Benchmarks of search for songs (push) / Run and upload benchmarks (push) Waiting to run
Benchmarks of search for Wikipedia articles (push) / Run and upload benchmarks (push) Waiting to run
Run the indexing fuzzer / Setup the action (push) Successful in 1h4m31s
* Remove unreachable code

* Add `indices` field to `MatchBounds`

For matches inside arrays, this field holds the indices of the array
elements that matched. For example, searching for `cat` inside
`{ "a": ["dog", "cat", "fox"] }` would return `indices: [1]`. For nested
arrays, this contains multiple indices, starting with the one for the
top-most array. For matches in fields without arrays, `indices` is not
serialized (does not exist) to save space.
2024-11-20 01:00:43 +01:00
ManyTheFish
510ca99996 Fixes #4974 2024-11-18 16:08:55 +01:00
ManyTheFish
8924d486db Add a test reproducing the bug 2024-11-18 16:08:55 +01:00
ManyTheFish
e0c3f3d560 Fix #4984 2024-11-18 16:08:53 +01:00
ManyTheFish
cd796b0f4b Fix SDK test 2024-11-18 11:46:00 +01:00
ManyTheFish
72ba353498 reproduce sdk fail 2024-11-18 10:03:23 +01:00
ManyTheFish
0dd321afc7 reproduce #4984 2024-11-14 10:02:51 +01:00
8 changed files with 319 additions and 104 deletions

View File

@ -1733,46 +1733,51 @@ fn format_fields(
// select the attributes to retrieve
let displayable_names =
displayable_ids.iter().map(|&fid| field_ids_map.name(fid).expect("Missing field name"));
permissive_json_pointer::map_leaf_values(&mut document, displayable_names, |key, value| {
// To get the formatting option of each key we need to see all the rules that applies
// to the value and merge them together. eg. If a user said he wanted to highlight `doggo`
// and crop `doggo.name`. `doggo.name` needs to be highlighted + cropped while `doggo.age` is only
// highlighted.
// Warn: The time to compute the format list scales with the number of fields to format;
// cumulated with map_leaf_values that iterates over all the nested fields, it gives a quadratic complexity:
// d*f where d is the total number of fields to display and f is the total number of fields to format.
let format = formatting_fields_options
.iter()
.filter(|(name, _option)| {
milli::is_faceted_by(name, key) || milli::is_faceted_by(key, name)
})
.map(|(_, option)| **option)
.reduce(|acc, option| acc.merge(option));
let mut infos = Vec::new();
// if no locales has been provided, we try to find the locales in the localized_attributes.
let locales = locales.or_else(|| {
localized_attributes
permissive_json_pointer::map_leaf_values(
&mut document,
displayable_names,
|key, array_indices, value| {
// To get the formatting option of each key we need to see all the rules that applies
// to the value and merge them together. eg. If a user said he wanted to highlight `doggo`
// and crop `doggo.name`. `doggo.name` needs to be highlighted + cropped while `doggo.age` is only
// highlighted.
// Warn: The time to compute the format list scales with the number of fields to format;
// cumulated with map_leaf_values that iterates over all the nested fields, it gives a quadratic complexity:
// d*f where d is the total number of fields to display and f is the total number of fields to format.
let format = formatting_fields_options
.iter()
.find(|rule| rule.match_str(key))
.map(LocalizedAttributesRule::locales)
});
.filter(|(name, _option)| {
milli::is_faceted_by(name, key) || milli::is_faceted_by(key, name)
})
.map(|(_, option)| **option)
.reduce(|acc, option| acc.merge(option));
let mut infos = Vec::new();
*value = format_value(
std::mem::take(value),
builder,
format,
&mut infos,
compute_matches,
locales,
);
// if no locales has been provided, we try to find the locales in the localized_attributes.
let locales = locales.or_else(|| {
localized_attributes
.iter()
.find(|rule| rule.match_str(key))
.map(LocalizedAttributesRule::locales)
});
if let Some(matches) = matches_position.as_mut() {
if !infos.is_empty() {
matches.insert(key.to_owned(), infos);
*value = format_value(
std::mem::take(value),
builder,
format,
&mut infos,
compute_matches,
array_indices,
locales,
);
if let Some(matches) = matches_position.as_mut() {
if !infos.is_empty() {
matches.insert(key.to_owned(), infos);
}
}
}
});
},
);
let selectors = formatted_options
.keys()
@ -1790,13 +1795,14 @@ fn format_value(
format_options: Option<FormatOptions>,
infos: &mut Vec<MatchBounds>,
compute_matches: bool,
array_indices: &[usize],
locales: Option<&[Language]>,
) -> Value {
match value {
Value::String(old_string) => {
let mut matcher = builder.build(&old_string, locales);
if compute_matches {
let matches = matcher.matches();
let matches = matcher.matches(array_indices);
infos.extend_from_slice(&matches[..]);
}
@ -1808,51 +1814,15 @@ fn format_value(
None => Value::String(old_string),
}
}
Value::Array(values) => Value::Array(
values
.into_iter()
.map(|v| {
format_value(
v,
builder,
format_options.map(|format_options| FormatOptions {
highlight: format_options.highlight,
crop: None,
}),
infos,
compute_matches,
locales,
)
})
.collect(),
),
Value::Object(object) => Value::Object(
object
.into_iter()
.map(|(k, v)| {
(
k,
format_value(
v,
builder,
format_options.map(|format_options| FormatOptions {
highlight: format_options.highlight,
crop: None,
}),
infos,
compute_matches,
locales,
),
)
})
.collect(),
),
// `map_leaf_values` makes sure this is only called for leaf fields
Value::Array(_) => unreachable!(),
Value::Object(_) => unreachable!(),
Value::Number(number) => {
let s = number.to_string();
let mut matcher = builder.build(&s, locales);
if compute_matches {
let matches = matcher.matches();
let matches = matcher.matches(array_indices);
infos.extend_from_slice(&matches[..]);
}

View File

@ -4,6 +4,58 @@ use super::*;
use crate::common::Server;
use crate::json;
#[actix_rt::test]
async fn search_formatted_from_sdk() {
let server = Server::new_shared();
let index = server.unique_index();
index
.update_settings(
json!({ "filterableAttributes": ["genre"], "searchableAttributes": ["title"] }),
)
.await;
let documents = json!([
{ "id": 123, "title": "Pride and Prejudice", "genre": "romance" },
{ "id": 456, "title": "Le Petit Prince", "genre": "adventure" },
{ "id": 1, "title": "Alice In Wonderland", "genre": "adventure" },
{ "id": 2, "title": "Le Rouge et le Noir", "genre": "romance" },
{ "id": 1344, "title": "The Hobbit", "genre": "adventure" },
{ "id": 4, "title": "Harry Potter and the Half-Blood Prince", "genre": "fantasy" },
{ "id": 7, "title": "Harry Potter and the Chamber of Secrets", "genre": "fantasy" },
{ "id": 42, "title": "The Hitchhiker's Guide to the Galaxy" }
]);
let (response, _) = index.add_documents(documents, None).await;
index.wait_task(response.uid()).await;
index
.search(
json!({ "q":"prince",
"attributesToCrop": ["title"],
"cropLength": 2,
"filter": "genre = adventure",
"attributesToHighlight": ["title"],
"attributesToRetrieve": ["title"]
}),
|response, code| {
assert_eq!(code, 200, "{}", response);
allow_duplicates! {
assert_json_snapshot!(response["hits"][0],
{ "._rankingScore" => "[score]" },
@r###"
{
"title": "Le Petit Prince",
"_formatted": {
"title": "…Petit <em>Prince</em>"
}
}
"###);
}
},
)
.await;
}
#[actix_rt::test]
async fn formatted_contain_wildcard() {
let server = Server::new_shared();
@ -208,7 +260,10 @@ async fn format_nested() {
"doggos.name": [
{
"start": 0,
"length": 5
"length": 5,
"indices": [
0
]
}
]
}

View File

@ -15,6 +15,7 @@ mod pagination;
mod restrict_searchable;
mod search_queue;
use meili_snap::{json_string, snapshot};
use meilisearch::Opt;
use tempfile::TempDir;
@ -62,6 +63,71 @@ async fn simple_search() {
.await;
}
#[actix_rt::test]
async fn search_with_stop_word() {
// related to https://github.com/meilisearch/meilisearch/issues/4984
let server = Server::new().await;
let index = server.index("test");
let (_, code) = index
.update_settings(json!({"stopWords": ["the", "The", "a", "an", "to", "in", "of"]}))
.await;
meili_snap::snapshot!(code, @"202 Accepted");
let documents = DOCUMENTS.clone();
index.add_documents(documents, None).await;
index.wait_task(1).await;
// prefix search
index
.search(json!({"q": "to the", "attributesToHighlight": ["title"], "attributesToRetrieve": ["title"] }), |response, code| {
assert_eq!(code, 200, "{}", response);
snapshot!(json_string!(response["hits"]), @"[]");
})
.await;
// non-prefix search
index
.search(json!({"q": "to the ", "attributesToHighlight": ["title"], "attributesToRetrieve": ["title"] }), |response, code| {
assert_eq!(code, 200, "{}", response);
snapshot!(json_string!(response["hits"]), @r###"
[
{
"title": "Shazam!",
"_formatted": {
"title": "Shazam!"
}
},
{
"title": "Captain Marvel",
"_formatted": {
"title": "Captain Marvel"
}
},
{
"title": "Escape Room",
"_formatted": {
"title": "Escape Room"
}
},
{
"title": "How to Train Your Dragon: The Hidden World",
"_formatted": {
"title": "How to Train Your Dragon: The Hidden World"
}
},
{
"title": "Gläss",
"_formatted": {
"title": "Gläss"
}
}
]
"###);
})
.await;
}
#[actix_rt::test]
async fn phrase_search_with_stop_word() {
// related to https://github.com/meilisearch/meilisearch/issues/3521

View File

@ -367,3 +367,50 @@ async fn search_on_exact_field() {
})
.await;
}
#[actix_rt::test]
async fn phrase_search_on_title() {
let server = Server::new().await;
let documents = json!([
{ "id": 8, "desc": "Document Review", "title": "Document Review Specialist II" },
{ "id": 5, "desc": "Document Review", "title": "Document Review Attorney" },
{ "id": 4, "desc": "Document Review", "title": "Document Review Manager - Cyber Incident Response (Remote)" },
{ "id": 3, "desc": "Document Review", "title": "Document Review Paralegal" },
{ "id": 2, "desc": "Document Review", "title": "Document Controller (Saudi National)" },
{ "id": 1, "desc": "Document Review", "title": "Document Reviewer" },
{ "id": 7, "desc": "Document Review", "title": "Document Review Specialist II" },
{ "id": 6, "desc": "Document Review", "title": "Document Review (Entry Level)" }
]);
let index = index_with_documents(&server, &documents).await;
index
.search(
json!({"q": "\"Document Review\"", "attributesToSearchOn": ["title"], "attributesToRetrieve": ["title"]}),
|response, code| {
snapshot!(code, @"200 OK");
snapshot!(json_string!(response["hits"]), @r###"
[
{
"title": "Document Review Specialist II"
},
{
"title": "Document Review Attorney"
},
{
"title": "Document Review Manager - Cyber Incident Response (Remote)"
},
{
"title": "Document Review Paralegal"
},
{
"title": "Document Review Specialist II"
},
{
"title": "Document Review (Entry Level)"
}
]
"###);
},
)
.await;
}

View File

@ -105,6 +105,8 @@ impl FormatOptions {
pub struct MatchBounds {
pub start: usize,
pub length: usize,
#[serde(skip_serializing_if = "Option::is_none")]
pub indices: Option<Vec<usize>>,
}
/// Structure used to analyze a string, compute words that match,
@ -220,15 +222,20 @@ impl<'t, 'tokenizer> Matcher<'t, 'tokenizer, '_, '_> {
}
/// Returns boundaries of the words that match the query.
pub fn matches(&mut self) -> Vec<MatchBounds> {
pub fn matches(&mut self, array_indices: &[usize]) -> Vec<MatchBounds> {
match &self.matches {
None => self.compute_matches().matches(),
None => self.compute_matches().matches(array_indices),
Some((tokens, matches)) => matches
.iter()
.map(|m| MatchBounds {
start: tokens[m.get_first_token_pos()].byte_start,
// TODO: Why is this in chars, while start is in bytes?
length: m.char_count,
indices: if array_indices.is_empty() {
None
} else {
Some(array_indices.to_owned())
},
})
.collect(),
}
@ -268,7 +275,7 @@ impl<'t, 'tokenizer> Matcher<'t, 'tokenizer, '_, '_> {
last_match_last_token_position_plus_one
} else {
// we have matched the end of possible tokens, there's nothing to advance
tokens.len() - 1
tokens.len()
}
};

View File

@ -193,15 +193,23 @@ pub fn compute_phrase_docids(
if words.is_empty() {
return Ok(RoaringBitmap::new());
}
let mut candidates = RoaringBitmap::new();
let mut candidates = None;
for word in words.iter().flatten().copied() {
if let Some(word_docids) = ctx.word_docids(None, Word::Original(word))? {
candidates |= word_docids;
if let Some(candidates) = candidates.as_mut() {
*candidates &= word_docids;
} else {
candidates = Some(word_docids);
}
} else {
return Ok(RoaringBitmap::new());
}
}
let Some(mut candidates) = candidates else {
return Ok(RoaringBitmap::new());
};
let winsize = words.len().min(3);
for win in words.windows(winsize) {

View File

@ -57,9 +57,9 @@ pub fn extract_docid_word_positions<R: io::Read + io::Seek>(
.map(|s| s.iter().map(String::as_str).collect());
let old_dictionary: Option<Vec<_>> =
settings_diff.old.dictionary.as_ref().map(|s| s.iter().map(String::as_str).collect());
let del_builder =
let mut del_builder =
tokenizer_builder(old_stop_words, old_separators.as_deref(), old_dictionary.as_deref());
let del_tokenizer = del_builder.into_tokenizer();
let del_tokenizer = del_builder.build();
let new_stop_words = settings_diff.new.stop_words.as_ref();
let new_separators: Option<Vec<_>> = settings_diff
@ -69,9 +69,9 @@ pub fn extract_docid_word_positions<R: io::Read + io::Seek>(
.map(|s| s.iter().map(String::as_str).collect());
let new_dictionary: Option<Vec<_>> =
settings_diff.new.dictionary.as_ref().map(|s| s.iter().map(String::as_str).collect());
let add_builder =
let mut add_builder =
tokenizer_builder(new_stop_words, new_separators.as_deref(), new_dictionary.as_deref());
let add_tokenizer = add_builder.into_tokenizer();
let add_tokenizer = add_builder.build();
// iterate over documents.
let mut cursor = obkv_documents.into_cursor()?;

View File

@ -45,7 +45,7 @@ fn contained_in(selector: &str, key: &str) -> bool {
/// map_leaf_values(
/// value.as_object_mut().unwrap(),
/// ["jean.race.name"],
/// |key, value| match (value, key) {
/// |key, _array_indices, value| match (value, key) {
/// (Value::String(name), "jean.race.name") => *name = "patou".to_string(),
/// _ => unreachable!(),
/// },
@ -66,17 +66,18 @@ fn contained_in(selector: &str, key: &str) -> bool {
pub fn map_leaf_values<'a>(
value: &mut Map<String, Value>,
selectors: impl IntoIterator<Item = &'a str>,
mut mapper: impl FnMut(&str, &mut Value),
mut mapper: impl FnMut(&str, &[usize], &mut Value),
) {
let selectors: Vec<_> = selectors.into_iter().collect();
map_leaf_values_in_object(value, &selectors, "", &mut mapper);
map_leaf_values_in_object(value, &selectors, "", &[], &mut mapper);
}
pub fn map_leaf_values_in_object(
value: &mut Map<String, Value>,
selectors: &[&str],
base_key: &str,
mapper: &mut impl FnMut(&str, &mut Value),
array_indices: &[usize],
mapper: &mut impl FnMut(&str, &[usize], &mut Value),
) {
for (key, value) in value.iter_mut() {
let base_key = if base_key.is_empty() {
@ -94,12 +95,12 @@ pub fn map_leaf_values_in_object(
if should_continue {
match value {
Value::Object(object) => {
map_leaf_values_in_object(object, selectors, &base_key, mapper)
map_leaf_values_in_object(object, selectors, &base_key, array_indices, mapper)
}
Value::Array(array) => {
map_leaf_values_in_array(array, selectors, &base_key, mapper)
map_leaf_values_in_array(array, selectors, &base_key, array_indices, mapper)
}
value => mapper(&base_key, value),
value => mapper(&base_key, array_indices, value),
}
}
}
@ -109,13 +110,24 @@ pub fn map_leaf_values_in_array(
values: &mut [Value],
selectors: &[&str],
base_key: &str,
mapper: &mut impl FnMut(&str, &mut Value),
base_array_indices: &[usize],
mapper: &mut impl FnMut(&str, &[usize], &mut Value),
) {
for value in values.iter_mut() {
// This avoids allocating twice
let mut array_indices = Vec::with_capacity(base_array_indices.len() + 1);
array_indices.extend_from_slice(base_array_indices);
array_indices.push(0);
for (i, value) in values.iter_mut().enumerate() {
*array_indices.last_mut().unwrap() = i;
match value {
Value::Object(object) => map_leaf_values_in_object(object, selectors, base_key, mapper),
Value::Array(array) => map_leaf_values_in_array(array, selectors, base_key, mapper),
value => mapper(base_key, value),
Value::Object(object) => {
map_leaf_values_in_object(object, selectors, base_key, &array_indices, mapper)
}
Value::Array(array) => {
map_leaf_values_in_array(array, selectors, base_key, &array_indices, mapper)
}
value => mapper(base_key, &array_indices, value),
}
}
}
@ -743,12 +755,14 @@ mod tests {
}
});
map_leaf_values(value.as_object_mut().unwrap(), ["jean.race.name"], |key, value| {
match (value, key) {
map_leaf_values(
value.as_object_mut().unwrap(),
["jean.race.name"],
|key, _, value| match (value, key) {
(Value::String(name), "jean.race.name") => *name = S("patou"),
_ => unreachable!(),
}
});
},
);
assert_eq!(
value,
@ -775,7 +789,7 @@ mod tests {
});
let mut calls = 0;
map_leaf_values(value.as_object_mut().unwrap(), ["jean"], |key, value| {
map_leaf_values(value.as_object_mut().unwrap(), ["jean"], |key, _, value| {
calls += 1;
match (value, key) {
(Value::String(name), "jean.race.name") => *name = S("patou"),
@ -798,4 +812,52 @@ mod tests {
})
);
}
#[test]
fn map_array() {
let mut value: Value = json!({
"no_array": "peter",
"simple": ["foo", "bar"],
"nested": [
{
"a": [
["cat", "dog"],
["fox", "bear"],
],
"b": "hi",
},
{
"a": ["green", "blue"],
},
],
});
map_leaf_values(
value.as_object_mut().unwrap(),
["no_array", "simple", "nested"],
|_key, array_indices, value| {
*value = format!("{array_indices:?}").into();
},
);
assert_eq!(
value,
json!({
"no_array": "[]",
"simple": ["[0]", "[1]"],
"nested": [
{
"a": [
["[0, 0, 0]", "[0, 0, 1]"],
["[0, 1, 0]", "[0, 1, 1]"],
],
"b": "[0]",
},
{
"a": ["[1, 0]", "[1, 1]"],
},
],
})
);
}
}