2020-12-24 19:58:34 +08:00
|
|
|
use std::collections::HashSet;
|
|
|
|
use std::mem;
|
|
|
|
use std::time::Instant;
|
2020-12-12 20:32:06 +08:00
|
|
|
|
2021-02-11 00:08:37 +08:00
|
|
|
use anyhow::{bail, Context};
|
2021-02-02 02:51:47 +08:00
|
|
|
use meilisearch_tokenizer::{Analyzer, AnalyzerConfig};
|
|
|
|
use milli::{Index, obkv_to_json, FacetCondition};
|
|
|
|
use serde::{Deserialize, Serialize};
|
|
|
|
use serde_json::{Value, Map};
|
2020-12-12 20:32:06 +08:00
|
|
|
|
2021-01-28 21:12:34 +08:00
|
|
|
use crate::index_controller::IndexController;
|
2020-12-29 18:11:06 +08:00
|
|
|
use super::Data;
|
2020-12-12 20:32:06 +08:00
|
|
|
|
2021-02-16 22:54:07 +08:00
|
|
|
pub const DEFAULT_SEARCH_LIMIT: usize = 20;
|
2020-12-24 19:58:34 +08:00
|
|
|
|
2021-01-14 00:50:36 +08:00
|
|
|
const fn default_search_limit() -> usize { DEFAULT_SEARCH_LIMIT }
|
|
|
|
|
2020-12-24 19:58:34 +08:00
|
|
|
#[derive(Deserialize)]
|
|
|
|
#[serde(rename_all = "camelCase", deny_unknown_fields)]
|
2020-12-29 18:11:06 +08:00
|
|
|
#[allow(dead_code)]
|
2020-12-24 19:58:34 +08:00
|
|
|
pub struct SearchQuery {
|
2021-01-14 18:27:07 +08:00
|
|
|
pub q: Option<String>,
|
|
|
|
pub offset: Option<usize>,
|
2021-01-14 00:50:36 +08:00
|
|
|
#[serde(default = "default_search_limit")]
|
2021-01-14 18:27:07 +08:00
|
|
|
pub limit: usize,
|
|
|
|
pub attributes_to_retrieve: Option<Vec<String>>,
|
|
|
|
pub attributes_to_crop: Option<Vec<String>>,
|
|
|
|
pub crop_length: Option<usize>,
|
2021-01-28 21:12:34 +08:00
|
|
|
pub attributes_to_highlight: Option<HashSet<String>>,
|
2021-01-14 18:27:07 +08:00
|
|
|
pub filters: Option<String>,
|
|
|
|
pub matches: Option<bool>,
|
|
|
|
pub facet_filters: Option<Value>,
|
|
|
|
pub facets_distribution: Option<Vec<String>>,
|
2021-01-28 21:12:34 +08:00
|
|
|
pub facet_condition: Option<String>,
|
|
|
|
}
|
|
|
|
|
|
|
|
impl SearchQuery {
|
|
|
|
pub fn perform(&self, index: impl AsRef<Index>) -> anyhow::Result<SearchResult>{
|
|
|
|
let index = index.as_ref();
|
|
|
|
let before_search = Instant::now();
|
|
|
|
let rtxn = index.read_txn().unwrap();
|
|
|
|
|
|
|
|
let mut search = index.search(&rtxn);
|
|
|
|
|
|
|
|
if let Some(ref query) = self.q {
|
|
|
|
search.query(query);
|
|
|
|
}
|
|
|
|
|
2021-02-02 02:51:47 +08:00
|
|
|
search.limit(self.limit);
|
|
|
|
search.offset(self.offset.unwrap_or_default());
|
|
|
|
|
2021-01-28 21:12:34 +08:00
|
|
|
if let Some(ref condition) = self.facet_condition {
|
|
|
|
if !condition.trim().is_empty() {
|
2021-02-04 20:21:15 +08:00
|
|
|
let condition = FacetCondition::from_str(&rtxn, &index, &condition)?;
|
2021-01-28 21:12:34 +08:00
|
|
|
search.facet_condition(condition);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2021-02-02 02:51:47 +08:00
|
|
|
let milli::SearchResult { documents_ids, found_words, candidates } = search.execute()?;
|
2021-01-28 21:12:34 +08:00
|
|
|
|
|
|
|
let mut documents = Vec::new();
|
|
|
|
let fields_ids_map = index.fields_ids_map(&rtxn).unwrap();
|
|
|
|
|
|
|
|
let displayed_fields = match index.displayed_fields_ids(&rtxn).unwrap() {
|
|
|
|
Some(fields) => fields,
|
|
|
|
None => fields_ids_map.iter().map(|(id, _)| id).collect(),
|
|
|
|
};
|
|
|
|
|
|
|
|
let stop_words = fst::Set::default();
|
|
|
|
let highlighter = Highlighter::new(&stop_words);
|
|
|
|
|
|
|
|
for (_id, obkv) in index.documents(&rtxn, documents_ids).unwrap() {
|
2021-02-11 00:08:37 +08:00
|
|
|
let mut object = obkv_to_json(&displayed_fields, &fields_ids_map, obkv)?;
|
2021-01-28 21:12:34 +08:00
|
|
|
if let Some(ref attributes_to_highlight) = self.attributes_to_highlight {
|
|
|
|
highlighter.highlight_record(&mut object, &found_words, attributes_to_highlight);
|
|
|
|
}
|
|
|
|
documents.push(object);
|
|
|
|
}
|
|
|
|
|
|
|
|
Ok(SearchResult {
|
|
|
|
hits: documents,
|
2021-02-02 02:51:47 +08:00
|
|
|
nb_hits: candidates.len(),
|
2021-01-28 21:12:34 +08:00
|
|
|
query: self.q.clone().unwrap_or_default(),
|
2021-02-02 02:51:47 +08:00
|
|
|
limit: self.limit,
|
2021-01-28 21:12:34 +08:00
|
|
|
offset: self.offset.unwrap_or_default(),
|
|
|
|
processing_time_ms: before_search.elapsed().as_millis(),
|
|
|
|
})
|
|
|
|
}
|
2020-12-24 19:58:34 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
#[derive(Serialize)]
|
|
|
|
#[serde(rename_all = "camelCase")]
|
|
|
|
pub struct SearchResult {
|
|
|
|
hits: Vec<Map<String, Value>>,
|
2021-02-02 02:51:47 +08:00
|
|
|
nb_hits: u64,
|
2020-12-24 19:58:34 +08:00
|
|
|
query: String,
|
|
|
|
limit: usize,
|
|
|
|
offset: usize,
|
|
|
|
processing_time_ms: u128,
|
|
|
|
}
|
|
|
|
|
2020-12-29 18:11:06 +08:00
|
|
|
struct Highlighter<'a, A> {
|
|
|
|
analyzer: Analyzer<'a, A>,
|
2020-12-12 20:32:06 +08:00
|
|
|
}
|
|
|
|
|
2020-12-29 18:11:06 +08:00
|
|
|
impl<'a, A: AsRef<[u8]>> Highlighter<'a, A> {
|
|
|
|
fn new(stop_words: &'a fst::Set<A>) -> Self {
|
|
|
|
let analyzer = Analyzer::new(AnalyzerConfig::default_with_stopwords(stop_words));
|
|
|
|
Self { analyzer }
|
2020-12-12 20:32:06 +08:00
|
|
|
}
|
|
|
|
|
2020-12-29 18:11:06 +08:00
|
|
|
fn highlight_value(&self, value: Value, words_to_highlight: &HashSet<String>) -> Value {
|
|
|
|
match value {
|
|
|
|
Value::Null => Value::Null,
|
|
|
|
Value::Bool(boolean) => Value::Bool(boolean),
|
|
|
|
Value::Number(number) => Value::Number(number),
|
|
|
|
Value::String(old_string) => {
|
|
|
|
let mut string = String::new();
|
|
|
|
let analyzed = self.analyzer.analyze(&old_string);
|
|
|
|
for (word, token) in analyzed.reconstruct() {
|
|
|
|
if token.is_word() {
|
|
|
|
let to_highlight = words_to_highlight.contains(token.text());
|
|
|
|
if to_highlight { string.push_str("<mark>") }
|
|
|
|
string.push_str(word);
|
|
|
|
if to_highlight { string.push_str("</mark>") }
|
|
|
|
} else {
|
|
|
|
string.push_str(word);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
Value::String(string)
|
|
|
|
},
|
|
|
|
Value::Array(values) => {
|
|
|
|
Value::Array(values.into_iter()
|
|
|
|
.map(|v| self.highlight_value(v, words_to_highlight))
|
|
|
|
.collect())
|
|
|
|
},
|
|
|
|
Value::Object(object) => {
|
|
|
|
Value::Object(object.into_iter()
|
|
|
|
.map(|(k, v)| (k, self.highlight_value(v, words_to_highlight)))
|
|
|
|
.collect())
|
|
|
|
},
|
2020-12-12 20:32:06 +08:00
|
|
|
}
|
|
|
|
}
|
2020-12-23 00:53:13 +08:00
|
|
|
|
2020-12-29 18:11:06 +08:00
|
|
|
fn highlight_record(
|
2020-12-23 20:52:28 +08:00
|
|
|
&self,
|
2020-12-29 18:11:06 +08:00
|
|
|
object: &mut Map<String, Value>,
|
|
|
|
words_to_highlight: &HashSet<String>,
|
|
|
|
attributes_to_highlight: &HashSet<String>,
|
|
|
|
) {
|
|
|
|
// TODO do we need to create a string for element that are not and needs to be highlight?
|
|
|
|
for (key, value) in object.iter_mut() {
|
|
|
|
if attributes_to_highlight.contains(key) {
|
|
|
|
let old_value = mem::take(value);
|
|
|
|
*value = self.highlight_value(old_value, words_to_highlight);
|
|
|
|
}
|
2020-12-23 20:52:28 +08:00
|
|
|
}
|
|
|
|
}
|
2020-12-29 18:11:06 +08:00
|
|
|
}
|
2020-12-23 20:52:28 +08:00
|
|
|
|
2020-12-29 18:11:06 +08:00
|
|
|
impl Data {
|
2021-01-14 00:50:36 +08:00
|
|
|
pub fn search<S: AsRef<str>>(&self, index: S, search_query: SearchQuery) -> anyhow::Result<SearchResult> {
|
2021-01-28 21:12:34 +08:00
|
|
|
match self.index_controller.index(&index)? {
|
|
|
|
Some(index) => Ok(search_query.perform(index)?),
|
|
|
|
None => bail!("index {:?} doesn't exists", index.as_ref()),
|
2020-12-24 19:58:34 +08:00
|
|
|
}
|
|
|
|
}
|
2021-02-11 00:08:37 +08:00
|
|
|
|
2021-02-16 06:02:20 +08:00
|
|
|
pub async fn retrieve_documents<S>(
|
2021-02-11 00:08:37 +08:00
|
|
|
&self,
|
2021-02-16 06:02:20 +08:00
|
|
|
index: impl AsRef<str> + Send + Sync + 'static,
|
2021-02-11 00:08:37 +08:00
|
|
|
offset: usize,
|
2021-02-11 17:59:23 +08:00
|
|
|
limit: usize,
|
2021-02-16 06:02:20 +08:00
|
|
|
attributes_to_retrieve: Option<Vec<S>>,
|
|
|
|
) -> anyhow::Result<Vec<Map<String, Value>>>
|
|
|
|
where
|
|
|
|
S: AsRef<str> + Send + Sync + 'static
|
|
|
|
{
|
|
|
|
let index_controller = self.index_controller.clone();
|
|
|
|
let documents: anyhow::Result<_> = tokio::task::spawn_blocking(move || {
|
|
|
|
let index = index_controller
|
|
|
|
.index(&index)?
|
|
|
|
.with_context(|| format!("Index {:?} doesn't exist", index.as_ref()))?;
|
|
|
|
|
|
|
|
let txn = index.read_txn()?;
|
|
|
|
|
|
|
|
let fields_ids_map = index.fields_ids_map(&txn)?;
|
|
|
|
|
|
|
|
let attributes_to_retrieve_ids = match attributes_to_retrieve {
|
|
|
|
Some(attrs) => attrs
|
|
|
|
.iter()
|
|
|
|
.filter_map(|f| fields_ids_map.id(f.as_ref()))
|
|
|
|
.collect::<Vec<_>>(),
|
|
|
|
None => fields_ids_map.iter().map(|(id, _)| id).collect(),
|
|
|
|
};
|
|
|
|
|
|
|
|
let iter = index.documents.range(&txn, &(..))?
|
|
|
|
.skip(offset)
|
|
|
|
.take(limit);
|
|
|
|
|
|
|
|
let mut documents = Vec::new();
|
|
|
|
|
|
|
|
for entry in iter {
|
|
|
|
let (_id, obkv) = entry?;
|
|
|
|
let object = obkv_to_json(&attributes_to_retrieve_ids, &fields_ids_map, obkv)?;
|
|
|
|
documents.push(object);
|
|
|
|
}
|
2021-02-11 00:08:37 +08:00
|
|
|
|
2021-02-16 06:02:20 +08:00
|
|
|
Ok(documents)
|
|
|
|
}).await?;
|
|
|
|
documents
|
2021-02-11 00:08:37 +08:00
|
|
|
}
|
2021-02-11 17:59:23 +08:00
|
|
|
|
2021-02-16 06:02:20 +08:00
|
|
|
pub async fn retrieve_document<S>(
|
2021-02-11 17:59:23 +08:00
|
|
|
&self,
|
2021-02-16 06:02:20 +08:00
|
|
|
index: impl AsRef<str> + Sync + Send + 'static,
|
|
|
|
document_id: impl AsRef<str> + Sync + Send + 'static,
|
|
|
|
attributes_to_retrieve: Option<Vec<S>>,
|
|
|
|
) -> anyhow::Result<Map<String, Value>>
|
|
|
|
where
|
|
|
|
S: AsRef<str> + Sync + Send + 'static,
|
|
|
|
{
|
|
|
|
let index_controller = self.index_controller.clone();
|
|
|
|
let document: anyhow::Result<_> = tokio::task::spawn_blocking(move || {
|
|
|
|
let index = index_controller
|
|
|
|
.index(&index)?
|
|
|
|
.with_context(|| format!("Index {:?} doesn't exist", index.as_ref()))?;
|
|
|
|
let txn = index.read_txn()?;
|
|
|
|
|
|
|
|
let fields_ids_map = index.fields_ids_map(&txn)?;
|
|
|
|
|
|
|
|
let attributes_to_retrieve_ids = match attributes_to_retrieve {
|
|
|
|
Some(attrs) => attrs
|
|
|
|
.iter()
|
|
|
|
.filter_map(|f| fields_ids_map.id(f.as_ref()))
|
|
|
|
.collect::<Vec<_>>(),
|
|
|
|
None => fields_ids_map.iter().map(|(id, _)| id).collect(),
|
|
|
|
};
|
|
|
|
|
|
|
|
let internal_id = index
|
|
|
|
.external_documents_ids(&txn)?
|
|
|
|
.get(document_id.as_ref().as_bytes())
|
|
|
|
.with_context(|| format!("Document with id {} not found", document_id.as_ref()))?;
|
|
|
|
|
|
|
|
let document = index.documents(&txn, std::iter::once(internal_id))?
|
|
|
|
.into_iter()
|
|
|
|
.next()
|
|
|
|
.map(|(_, d)| d);
|
|
|
|
|
|
|
|
match document {
|
|
|
|
Some(document) => Ok(obkv_to_json(&attributes_to_retrieve_ids, &fields_ids_map, document)?),
|
|
|
|
None => bail!("Document with id {} not found", document_id.as_ref()),
|
|
|
|
}
|
|
|
|
}).await?;
|
|
|
|
document
|
2021-02-11 17:59:23 +08:00
|
|
|
}
|
2020-12-24 19:58:34 +08:00
|
|
|
}
|