meilisearch/meilisearch-lib/src/index/search.rs

748 lines
25 KiB
Rust
Raw Normal View History

use std::cmp::min;
2021-06-17 22:59:01 +08:00
use std::collections::{BTreeMap, BTreeSet, HashSet};
2021-08-24 18:31:35 +08:00
use std::str::FromStr;
2021-03-16 01:11:10 +08:00
use std::time::Instant;
2021-03-04 18:56:32 +08:00
2021-03-16 01:11:10 +08:00
use either::Either;
2022-06-06 21:53:28 +08:00
use milli::tokenizer::TokenizerBuilder;
2022-05-18 16:26:52 +08:00
use milli::{
AscDesc, FieldId, FieldsIdsMap, Filter, FormatOptions, MatchBounds, MatcherBuilder, SortError,
TermsMatchingStrategy, DEFAULT_VALUES_PER_FACET,
2022-05-18 16:26:52 +08:00
};
2021-09-27 21:41:14 +08:00
use regex::Regex;
2021-03-16 01:11:10 +08:00
use serde::{Deserialize, Serialize};
2021-09-27 21:41:14 +08:00
use serde_json::{json, Value};
2021-03-04 18:56:32 +08:00
use crate::index::error::FacetError;
2021-10-06 19:01:02 +08:00
use super::error::{IndexError, Result};
2021-10-04 18:15:21 +08:00
use super::index::Index;
2021-03-04 18:56:32 +08:00
pub type Document = serde_json::Map<String, Value>;
type MatchesPosition = BTreeMap<String, Vec<MatchBounds>>;
2021-04-19 22:22:41 +08:00
2022-10-20 23:41:13 +08:00
pub const DEFAULT_SEARCH_OFFSET: fn() -> usize = || 0;
2022-06-02 16:48:02 +08:00
pub const DEFAULT_SEARCH_LIMIT: fn() -> usize = || 20;
pub const DEFAULT_CROP_LENGTH: fn() -> usize = || 10;
pub const DEFAULT_CROP_MARKER: fn() -> String = || "".to_string();
pub const DEFAULT_HIGHLIGHT_PRE_TAG: fn() -> String = || "<em>".to_string();
pub const DEFAULT_HIGHLIGHT_POST_TAG: fn() -> String = || "</em>".to_string();
/// The maximum number of results that the engine
/// will be able to return in one search call.
pub const DEFAULT_PAGINATION_MAX_TOTAL_HITS: usize = 1000;
2022-08-12 20:18:27 +08:00
#[derive(Deserialize, Debug, Clone, PartialEq, Eq)]
2021-03-04 18:56:32 +08:00
#[serde(rename_all = "camelCase", deny_unknown_fields)]
pub struct SearchQuery {
pub q: Option<String>,
2022-10-20 23:41:13 +08:00
#[serde(default = "DEFAULT_SEARCH_OFFSET")]
pub offset: usize,
#[serde(default = "DEFAULT_SEARCH_LIMIT")]
pub limit: usize,
pub page: Option<usize>,
pub hits_per_page: Option<usize>,
2021-06-16 22:18:55 +08:00
pub attributes_to_retrieve: Option<BTreeSet<String>>,
pub attributes_to_crop: Option<Vec<String>>,
2022-06-02 16:48:02 +08:00
#[serde(default = "DEFAULT_CROP_LENGTH")]
pub crop_length: usize,
2021-03-04 18:56:32 +08:00
pub attributes_to_highlight: Option<HashSet<String>>,
2021-06-22 05:38:59 +08:00
// Default to false
#[serde(default = "Default::default")]
pub show_matches_position: bool,
pub filter: Option<Value>,
2021-08-24 18:31:35 +08:00
pub sort: Option<Vec<String>>,
pub facets: Option<Vec<String>>,
2022-06-02 16:48:02 +08:00
#[serde(default = "DEFAULT_HIGHLIGHT_PRE_TAG")]
pub highlight_pre_tag: String,
2022-06-02 16:48:02 +08:00
#[serde(default = "DEFAULT_HIGHLIGHT_POST_TAG")]
pub highlight_post_tag: String,
2022-06-02 16:48:02 +08:00
#[serde(default = "DEFAULT_CROP_MARKER")]
pub crop_marker: String,
#[serde(default)]
pub matching_strategy: MatchingStrategy,
}
impl SearchQuery {
pub fn is_finite_pagination(&self) -> bool {
self.page.or(self.hits_per_page).is_some()
}
}
#[derive(Deserialize, Debug, Clone, PartialEq, Eq)]
#[serde(rename_all = "camelCase")]
pub enum MatchingStrategy {
/// Remove query words from last to first
Last,
/// All query words are mandatory
All,
}
impl Default for MatchingStrategy {
fn default() -> Self {
Self::Last
}
}
impl From<MatchingStrategy> for TermsMatchingStrategy {
fn from(other: MatchingStrategy) -> Self {
match other {
MatchingStrategy::Last => Self::Last,
MatchingStrategy::All => Self::All,
}
}
2021-03-04 18:56:32 +08:00
}
2021-10-06 19:01:02 +08:00
#[derive(Debug, Clone, Serialize, PartialEq)]
2021-04-19 16:13:13 +08:00
pub struct SearchHit {
#[serde(flatten)]
2021-04-19 22:22:41 +08:00
pub document: Document,
2021-04-20 19:10:50 +08:00
#[serde(rename = "_formatted", skip_serializing_if = "Document::is_empty")]
2021-04-19 22:22:41 +08:00
pub formatted: Document,
#[serde(rename = "_matchesPosition", skip_serializing_if = "Option::is_none")]
pub matches_position: Option<MatchesPosition>,
2021-04-19 16:13:13 +08:00
}
2021-10-06 19:01:02 +08:00
#[derive(Serialize, Debug, Clone, PartialEq)]
2021-03-04 18:56:32 +08:00
#[serde(rename_all = "camelCase")]
pub struct SearchResult {
2021-04-19 16:13:13 +08:00
pub hits: Vec<SearchHit>,
2021-03-04 18:56:32 +08:00
pub query: String,
pub processing_time_ms: u128,
2022-07-11 23:27:07 +08:00
#[serde(flatten)]
pub hits_info: HitsInfo,
2021-03-04 18:56:32 +08:00
#[serde(skip_serializing_if = "Option::is_none")]
pub facet_distribution: Option<BTreeMap<String, BTreeMap<String, u64>>>,
2021-03-04 18:56:32 +08:00
}
2022-10-24 19:57:26 +08:00
#[derive(Serialize, Debug, Clone, PartialEq, Eq)]
2022-07-11 23:27:07 +08:00
#[serde(untagged)]
pub enum HitsInfo {
2022-07-12 23:11:59 +08:00
#[serde(rename_all = "camelCase")]
2022-07-11 23:27:07 +08:00
Pagination {
hits_per_page: usize,
page: usize,
total_pages: usize,
total_hits: usize,
2022-07-11 23:27:07 +08:00
},
2022-07-12 23:11:59 +08:00
#[serde(rename_all = "camelCase")]
2022-07-11 23:27:07 +08:00
OffsetLimit {
limit: usize,
offset: usize,
estimated_total_hits: usize,
},
}
2021-03-04 18:56:32 +08:00
impl Index {
pub fn perform_search(&self, query: SearchQuery) -> Result<SearchResult> {
2021-03-04 18:56:32 +08:00
let before_search = Instant::now();
let rtxn = self.read_txn()?;
let mut search = self.search(&rtxn);
if let Some(ref query) = query.q {
search.query(query);
}
let is_finite_pagination = query.is_finite_pagination();
search.terms_matching_strategy(query.matching_strategy.into());
let max_total_hits = self
.pagination_max_total_hits(&rtxn)?
.unwrap_or(DEFAULT_PAGINATION_MAX_TOTAL_HITS);
2022-06-09 16:17:55 +08:00
search.exhaustive_number_hits(is_finite_pagination);
2022-10-20 23:41:13 +08:00
// compute the offset on the limit depending on the pagination mode.
2022-07-11 23:27:07 +08:00
let (offset, limit) = if is_finite_pagination {
2022-10-20 23:41:13 +08:00
let limit = query.hits_per_page.unwrap_or_else(DEFAULT_SEARCH_LIMIT);
let page = query.page.unwrap_or(1);
2022-10-20 23:41:13 +08:00
// page 0 gives a limit of 0 forcing Meilisearch to return no document.
page.checked_sub(1).map_or((0, 0), |p| (limit * p, limit))
2022-07-11 23:27:07 +08:00
} else {
2022-10-20 23:41:13 +08:00
(query.offset, query.limit)
2022-07-11 23:27:07 +08:00
};
2022-03-31 01:06:15 +08:00
2022-10-20 23:41:13 +08:00
// Make sure that a user can't get more documents than the hard limit,
// we align that on the offset too.
let offset = min(offset, max_total_hits);
let limit = min(limit, max_total_hits.saturating_sub(offset));
2022-03-31 01:06:15 +08:00
search.offset(offset);
search.limit(limit);
2021-03-04 18:56:32 +08:00
if let Some(ref filter) = query.filter {
if let Some(facets) = parse_filter(filter)? {
2021-06-04 01:36:25 +08:00
search.filter(facets);
2021-03-04 18:56:32 +08:00
}
}
2021-08-24 18:31:35 +08:00
if let Some(ref sort) = query.sort {
let sort = match sort.iter().map(|s| AscDesc::from_str(s)).collect() {
Ok(sorts) => sorts,
2021-09-28 20:49:13 +08:00
Err(asc_desc_error) => {
return Err(IndexError::Milli(SortError::from(asc_desc_error).into()))
2021-09-27 21:41:14 +08:00
}
2021-08-24 18:31:35 +08:00
};
search.sort_criteria(sort);
}
2021-03-04 18:56:32 +08:00
let milli::SearchResult {
documents_ids,
2021-03-12 02:40:18 +08:00
matching_words,
2021-03-04 18:56:32 +08:00
candidates,
..
} = search.execute()?;
2021-06-17 20:36:32 +08:00
2021-03-04 18:56:32 +08:00
let fields_ids_map = self.fields_ids_map(&rtxn).unwrap();
2021-05-31 22:03:39 +08:00
let displayed_ids = self
.displayed_fields_ids(&rtxn)?
2021-06-16 22:18:55 +08:00
.map(|fields| fields.into_iter().collect::<BTreeSet<_>>())
2021-04-20 22:21:30 +08:00
.unwrap_or_else(|| fields_ids_map.iter().map(|(id, _)| id).collect());
2021-04-20 19:10:50 +08:00
2021-06-16 22:18:55 +08:00
let fids = |attrs: &BTreeSet<String>| {
let mut ids = BTreeSet::new();
2021-04-20 19:10:50 +08:00
for attr in attrs {
if attr == "*" {
ids = displayed_ids.clone();
break;
}
if let Some(id) = fields_ids_map.id(attr) {
ids.insert(id);
}
}
ids
2021-04-19 22:22:41 +08:00
};
2021-06-15 22:21:41 +08:00
// The attributes to retrieve are the ones explicitly marked as to retrieve (all by default),
// but these attributes must be also be present
// - in the fields_ids_map
// - in the the displayed attributes
2021-06-16 22:18:55 +08:00
let to_retrieve_ids: BTreeSet<_> = query
2021-04-19 22:22:41 +08:00
.attributes_to_retrieve
.as_ref()
.map(fids)
2021-06-15 22:21:41 +08:00
.unwrap_or_else(|| displayed_ids.clone())
2021-06-14 05:51:33 +08:00
.intersection(&displayed_ids)
.cloned()
2021-06-15 22:21:41 +08:00
.collect();
2021-04-19 22:22:41 +08:00
2021-06-22 05:38:59 +08:00
let attr_to_highlight = query.attributes_to_highlight.unwrap_or_default();
2021-04-19 22:22:41 +08:00
2021-06-22 05:38:59 +08:00
let attr_to_crop = query.attributes_to_crop.unwrap_or_default();
2021-04-19 22:22:41 +08:00
// Attributes in `formatted_options` correspond to the attributes that will be in `_formatted`
// These attributes are:
// - the attributes asked to be highlighted or cropped (with `attributesToCrop` or `attributesToHighlight`)
// - the attributes asked to be retrieved: these attributes will not be highlighted/cropped
// But these attributes must be also present in displayed attributes
2021-06-16 20:23:08 +08:00
let formatted_options = compute_formatted_options(
&attr_to_highlight,
&attr_to_crop,
query.crop_length,
&to_retrieve_ids,
&fields_ids_map,
&displayed_ids,
);
2021-03-04 18:56:32 +08:00
2022-06-06 21:53:28 +08:00
let tokenizer = TokenizerBuilder::default().build();
2022-06-06 21:53:28 +08:00
let mut formatter_builder = MatcherBuilder::new(matching_words, tokenizer);
2022-05-18 16:26:52 +08:00
formatter_builder.crop_marker(query.crop_marker);
formatter_builder.highlight_prefix(query.highlight_pre_tag);
formatter_builder.highlight_suffix(query.highlight_post_tag);
2021-03-04 18:56:32 +08:00
let mut documents = Vec::new();
2021-06-17 20:36:32 +08:00
let documents_iter = self.documents(&rtxn, documents_ids)?;
2021-06-21 18:09:59 +08:00
for (_id, obkv) in documents_iter {
2022-04-19 22:49:38 +08:00
// First generate a document with all the displayed fields
let displayed_document = make_document(&displayed_ids, &fields_ids_map, obkv)?;
// select the attributes to retrieve
let attributes_to_retrieve = to_retrieve_ids
.iter()
.map(|&fid| fields_ids_map.name(fid).expect("Missing field name"));
let mut document =
permissive_json_pointer::select_values(&displayed_document, attributes_to_retrieve);
2021-06-22 05:38:59 +08:00
let (matches_position, formatted) = format_fields(
2022-04-19 22:49:38 +08:00
&displayed_document,
2021-04-19 22:22:41 +08:00
&fields_ids_map,
2022-05-18 16:26:52 +08:00
&formatter_builder,
2021-06-14 05:51:33 +08:00
&formatted_options,
query.show_matches_position,
2022-05-18 16:26:52 +08:00
&displayed_ids,
2021-04-19 22:22:41 +08:00
)?;
2021-09-27 21:41:14 +08:00
if let Some(sort) = query.sort.as_ref() {
insert_geo_distance(sort, &mut document);
}
2021-04-19 16:13:13 +08:00
let hit = SearchHit {
2021-04-19 22:22:41 +08:00
document,
formatted,
matches_position,
2021-04-19 16:13:13 +08:00
};
documents.push(hit);
2021-03-04 18:56:32 +08:00
}
2022-07-11 23:27:07 +08:00
let number_of_hits = min(candidates.len() as usize, max_total_hits);
let hits_info = if is_finite_pagination {
2022-10-20 23:41:13 +08:00
let hits_per_page = query.hits_per_page.unwrap_or_else(DEFAULT_SEARCH_LIMIT);
// If hit_per_page is 0, then pages can't be computed and so we respond 0.
2022-10-20 23:41:13 +08:00
let total_pages = (number_of_hits + hits_per_page.saturating_sub(1))
.checked_div(hits_per_page)
.unwrap_or(0);
2022-07-11 23:27:07 +08:00
HitsInfo::Pagination {
2022-10-24 19:57:26 +08:00
hits_per_page,
2022-10-20 23:41:13 +08:00
page: query.page.unwrap_or(1),
total_pages,
total_hits: number_of_hits,
2022-07-11 23:27:07 +08:00
}
} else {
HitsInfo::OffsetLimit {
2022-10-20 23:41:13 +08:00
limit: query.limit,
2022-07-11 23:27:07 +08:00
offset,
estimated_total_hits: number_of_hits,
}
};
2021-03-04 18:56:32 +08:00
let facet_distribution = match query.facets {
2021-03-04 18:56:32 +08:00
Some(ref fields) => {
let mut facet_distribution = self.facets_distribution(&rtxn);
let max_values_by_facet = self
.max_values_per_facet(&rtxn)?
.unwrap_or(DEFAULT_VALUES_PER_FACET);
facet_distribution.max_values_per_facet(max_values_by_facet);
2021-03-04 18:56:32 +08:00
if fields.iter().all(|f| f != "*") {
facet_distribution.facets(fields);
2021-03-04 18:56:32 +08:00
}
let distribution = facet_distribution.candidates(candidates).execute()?;
Some(distribution)
2021-03-04 18:56:32 +08:00
}
None => None,
};
let result = SearchResult {
hits: documents,
2022-07-11 23:27:07 +08:00
hits_info,
2021-03-04 18:56:32 +08:00
query: query.q.clone().unwrap_or_default(),
processing_time_ms: before_search.elapsed().as_millis(),
facet_distribution,
2021-03-04 18:56:32 +08:00
};
Ok(result)
}
}
2021-09-27 21:41:14 +08:00
fn insert_geo_distance(sorts: &[String], document: &mut Document) {
lazy_static::lazy_static! {
static ref GEO_REGEX: Regex =
Regex::new(r"_geoPoint\(\s*([[:digit:].\-]+)\s*,\s*([[:digit:].\-]+)\s*\)").unwrap();
};
if let Some(capture_group) = sorts.iter().find_map(|sort| GEO_REGEX.captures(sort)) {
// TODO: TAMO: milli encountered an internal error, what do we want to do?
let base = [
capture_group[1].parse().unwrap(),
capture_group[2].parse().unwrap(),
];
let geo_point = &document.get("_geo").unwrap_or(&json!(null));
if let Some((lat, lng)) = geo_point["lat"].as_f64().zip(geo_point["lng"].as_f64()) {
let distance = milli::distance_between_two_points(&base, &[lat, lng]);
document.insert("_geoDistance".to_string(), json!(distance.round() as usize));
}
}
}
2021-06-16 20:23:08 +08:00
fn compute_formatted_options(
attr_to_highlight: &HashSet<String>,
attr_to_crop: &[String],
query_crop_length: usize,
to_retrieve_ids: &BTreeSet<FieldId>,
fields_ids_map: &FieldsIdsMap,
displayed_ids: &BTreeSet<FieldId>,
2021-06-22 05:38:59 +08:00
) -> BTreeMap<FieldId, FormatOptions> {
let mut formatted_options = BTreeMap::new();
add_highlight_to_formatted_options(
&mut formatted_options,
2021-06-16 23:13:21 +08:00
attr_to_highlight,
fields_ids_map,
displayed_ids,
);
add_crop_to_formatted_options(
&mut formatted_options,
2021-06-16 23:13:21 +08:00
attr_to_crop,
query_crop_length,
fields_ids_map,
displayed_ids,
);
// Should not return `_formatted` if no valid attributes to highlight/crop
if !formatted_options.is_empty() {
2021-06-22 05:38:59 +08:00
add_non_formatted_ids_to_formatted_options(&mut formatted_options, to_retrieve_ids);
}
2021-06-16 23:13:21 +08:00
formatted_options
}
fn add_highlight_to_formatted_options(
formatted_options: &mut BTreeMap<FieldId, FormatOptions>,
2021-06-16 23:13:21 +08:00
attr_to_highlight: &HashSet<String>,
fields_ids_map: &FieldsIdsMap,
displayed_ids: &BTreeSet<FieldId>,
) {
for attr in attr_to_highlight {
let new_format = FormatOptions {
highlight: true,
crop: None,
};
if attr == "*" {
2021-06-16 00:44:56 +08:00
for id in displayed_ids {
formatted_options.insert(*id, new_format);
}
break;
}
2021-07-30 00:14:36 +08:00
if let Some(id) = fields_ids_map.id(attr) {
if displayed_ids.contains(&id) {
formatted_options.insert(id, new_format);
}
}
2021-06-16 00:44:56 +08:00
}
2021-06-16 23:13:21 +08:00
}
2021-06-16 23:13:21 +08:00
fn add_crop_to_formatted_options(
formatted_options: &mut BTreeMap<FieldId, FormatOptions>,
2021-06-16 23:13:21 +08:00
attr_to_crop: &[String],
crop_length: usize,
fields_ids_map: &FieldsIdsMap,
displayed_ids: &BTreeSet<FieldId>,
) {
for attr in attr_to_crop {
2021-06-17 22:59:01 +08:00
let mut split = attr.rsplitn(2, ':');
let (attr_name, attr_len) = match split.next().zip(split.next()) {
2021-06-16 00:44:56 +08:00
Some((len, name)) => {
2021-06-17 22:59:01 +08:00
let crop_len = len.parse::<usize>().unwrap_or(crop_length);
(name, crop_len)
2021-06-22 05:38:59 +08:00
}
2021-06-17 22:59:01 +08:00
None => (attr.as_str(), crop_length),
2021-06-16 00:44:56 +08:00
};
if attr_name == "*" {
2021-06-16 00:44:56 +08:00
for id in displayed_ids {
formatted_options
.entry(*id)
.and_modify(|f| f.crop = Some(attr_len))
.or_insert(FormatOptions {
highlight: false,
crop: Some(attr_len),
});
}
}
2021-07-30 00:14:36 +08:00
if let Some(id) = fields_ids_map.id(attr_name) {
if displayed_ids.contains(&id) {
formatted_options
.entry(id)
.and_modify(|f| f.crop = Some(attr_len))
.or_insert(FormatOptions {
highlight: false,
crop: Some(attr_len),
});
}
}
}
}
fn add_non_formatted_ids_to_formatted_options(
formatted_options: &mut BTreeMap<FieldId, FormatOptions>,
to_retrieve_ids: &BTreeSet<FieldId>,
) {
for id in to_retrieve_ids {
2021-06-22 05:38:59 +08:00
formatted_options.entry(*id).or_insert(FormatOptions {
highlight: false,
crop: None,
});
}
}
2021-04-21 03:19:37 +08:00
fn make_document(
2022-04-19 22:49:38 +08:00
displayed_attributes: &BTreeSet<FieldId>,
2021-04-21 03:19:37 +08:00
field_ids_map: &FieldsIdsMap,
2021-07-28 16:52:47 +08:00
obkv: obkv::KvReaderU16,
) -> Result<Document> {
let mut document = serde_json::Map::new();
2021-04-21 03:19:37 +08:00
// recreate the original json
for (key, value) in obkv.iter() {
let value = serde_json::from_slice(value)?;
let key = field_ids_map
.name(key)
.expect("Missing field name")
.to_string();
2021-04-21 03:19:37 +08:00
document.insert(key, value);
2021-04-21 03:19:37 +08:00
}
// select the attributes to retrieve
2022-04-19 22:49:38 +08:00
let displayed_attributes = displayed_attributes
.iter()
.map(|&fid| field_ids_map.name(fid).expect("Missing field name"));
2022-04-19 22:49:38 +08:00
let document = permissive_json_pointer::select_values(&document, displayed_attributes);
2021-04-21 03:19:37 +08:00
Ok(document)
}
2022-05-18 16:26:52 +08:00
fn format_fields<'a, A: AsRef<[u8]>>(
document: &Document,
2021-04-19 22:22:41 +08:00
field_ids_map: &FieldsIdsMap,
2022-06-06 21:53:28 +08:00
builder: &MatcherBuilder<'a, A>,
formatted_options: &BTreeMap<FieldId, FormatOptions>,
2022-05-18 16:26:52 +08:00
compute_matches: bool,
displayable_ids: &BTreeSet<FieldId>,
) -> Result<(Option<MatchesPosition>, Document)> {
let mut matches_position = compute_matches.then(BTreeMap::new);
2022-05-18 16:26:52 +08:00
let mut document = document.clone();
2022-05-18 16:26:52 +08:00
// 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.
let format = formatted_options
.iter()
.filter(|(field, _option)| {
let name = field_ids_map.name(**field).unwrap();
milli::is_faceted_by(name, key) || milli::is_faceted_by(key, name)
})
2022-05-18 16:26:52 +08:00
.map(|(_, option)| *option)
.reduce(|acc, option| acc.merge(option));
let mut infos = Vec::new();
2022-05-18 16:26:52 +08:00
*value = format_value(
std::mem::take(value),
builder,
format,
&mut infos,
compute_matches,
);
2021-04-19 22:22:41 +08:00
if let Some(matches) = matches_position.as_mut() {
2022-05-18 16:26:52 +08:00
if !infos.is_empty() {
matches.insert(key.to_owned(), infos);
}
}
});
2021-04-20 01:03:53 +08:00
2022-05-18 16:26:52 +08:00
let selectors = formatted_options
.keys()
// This unwrap must be safe since we got the ids from the fields_ids_map just
// before.
.map(|&fid| field_ids_map.name(fid).unwrap());
let document = permissive_json_pointer::select_values(&document, selectors);
2021-04-20 01:03:53 +08:00
Ok((matches_position, document))
2021-04-20 01:03:53 +08:00
}
2022-05-18 16:26:52 +08:00
fn format_value<'a, A: AsRef<[u8]>>(
value: Value,
2022-06-06 21:53:28 +08:00
builder: &MatcherBuilder<'a, A>,
2022-05-18 16:26:52 +08:00
format_options: Option<FormatOptions>,
infos: &mut Vec<MatchBounds>,
compute_matches: bool,
) -> Value {
match value {
Value::String(old_string) => {
2022-06-06 21:53:28 +08:00
let mut matcher = builder.build(&old_string);
2022-05-18 16:26:52 +08:00
if compute_matches {
let matches = matcher.matches();
infos.extend_from_slice(&matches[..]);
}
2021-03-04 18:56:32 +08:00
2022-05-18 16:26:52 +08:00
match format_options {
Some(format_options) => {
let value = matcher.format(format_options);
Value::String(value.into_owned())
}
None => Value::String(old_string),
2021-03-04 18:56:32 +08:00
}
2022-05-18 16:26:52 +08:00
}
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,
)
})
.collect(),
),
Value::Object(object) => Value::Object(
object
.into_iter()
.map(|(k, v)| {
(
k,
format_value(
2021-06-22 05:38:59 +08:00
v,
2022-05-18 16:26:52 +08:00
builder,
format_options.map(|format_options| FormatOptions {
2021-06-22 05:38:59 +08:00
highlight: format_options.highlight,
crop: None,
2022-05-18 16:26:52 +08:00
}),
infos,
compute_matches,
),
)
})
.collect(),
),
Value::Number(number) => {
let s = number.to_string();
2022-06-06 21:53:28 +08:00
let mut matcher = builder.build(&s);
2022-05-18 16:26:52 +08:00
if compute_matches {
let matches = matcher.matches();
infos.extend_from_slice(&matches[..]);
2021-05-12 00:30:55 +08:00
}
2021-05-05 23:31:40 +08:00
2022-05-18 16:26:52 +08:00
match format_options {
Some(format_options) => {
let value = matcher.format(format_options);
Value::String(value.into_owned())
2021-05-12 00:30:55 +08:00
}
2022-05-18 16:26:52 +08:00
None => Value::Number(number),
2021-06-22 16:17:39 +08:00
}
}
2022-05-18 16:26:52 +08:00
value => value,
2021-05-05 23:31:40 +08:00
}
2021-03-04 18:56:32 +08:00
}
fn parse_filter(facets: &Value) -> Result<Option<Filter>> {
2021-03-04 18:56:32 +08:00
match facets {
Value::String(expr) => {
let condition = Filter::from_str(expr)?;
2022-01-19 18:21:19 +08:00
Ok(condition)
}
Value::Array(arr) => parse_filter_array(arr),
2021-06-17 20:38:52 +08:00
v => Err(FacetError::InvalidExpression(&["Array"], v.clone()).into()),
2021-03-04 18:56:32 +08:00
}
}
2021-04-20 01:03:53 +08:00
fn parse_filter_array(arr: &[Value]) -> Result<Option<Filter>> {
2021-05-05 00:22:48 +08:00
let mut ands = Vec::new();
for value in arr {
match value {
Value::String(s) => ands.push(Either::Right(s.as_str())),
2021-05-05 00:22:48 +08:00
Value::Array(arr) => {
let mut ors = Vec::new();
for value in arr {
match value {
Value::String(s) => ors.push(s.as_str()),
v => {
return Err(FacetError::InvalidExpression(&["String"], v.clone()).into())
}
2021-05-05 00:22:48 +08:00
}
}
ands.push(Either::Left(ors));
}
v => {
return Err(
FacetError::InvalidExpression(&["String", "[String]"], v.clone()).into(),
)
}
2021-05-05 00:22:48 +08:00
}
}
Ok(Filter::from_array(ands)?)
2021-05-05 00:22:48 +08:00
}
2021-04-20 01:03:53 +08:00
#[cfg(test)]
mod test {
use super::*;
2021-09-27 21:41:14 +08:00
#[test]
fn test_insert_geo_distance() {
let value: Document = serde_json::from_str(
r#"{
2022-05-18 16:26:52 +08:00
"_geo": {
"lat": 50.629973371633746,
"lng": 3.0569447399419567
},
"city": "Lille",
"id": "1"
}"#,
2021-09-27 21:41:14 +08:00
)
.unwrap();
let sorters = &["_geoPoint(50.629973371633746,3.0569447399419567):desc".to_string()];
let mut document = value.clone();
insert_geo_distance(sorters, &mut document);
assert_eq!(document.get("_geoDistance"), Some(&json!(0)));
let sorters = &["_geoPoint(50.629973371633746, 3.0569447399419567):asc".to_string()];
let mut document = value.clone();
insert_geo_distance(sorters, &mut document);
assert_eq!(document.get("_geoDistance"), Some(&json!(0)));
let sorters =
&["_geoPoint( 50.629973371633746 , 3.0569447399419567 ):desc".to_string()];
let mut document = value.clone();
insert_geo_distance(sorters, &mut document);
assert_eq!(document.get("_geoDistance"), Some(&json!(0)));
let sorters = &[
"prix:asc",
"villeneuve:desc",
"_geoPoint(50.629973371633746, 3.0569447399419567):asc",
"ubu:asc",
]
.map(|s| s.to_string());
let mut document = value.clone();
insert_geo_distance(sorters, &mut document);
assert_eq!(document.get("_geoDistance"), Some(&json!(0)));
// only the first geoPoint is used to compute the distance
let sorters = &[
"chien:desc",
"_geoPoint(50.629973371633746, 3.0569447399419567):asc",
"pangolin:desc",
"_geoPoint(100.0, -80.0):asc",
"chat:asc",
]
.map(|s| s.to_string());
let mut document = value.clone();
insert_geo_distance(sorters, &mut document);
assert_eq!(document.get("_geoDistance"), Some(&json!(0)));
// there was no _geoPoint so nothing is inserted in the document
let sorters = &["chien:asc".to_string()];
let mut document = value;
insert_geo_distance(sorters, &mut document);
assert_eq!(document.get("_geoDistance"), None);
}
2021-04-20 01:03:53 +08:00
}