Update API to fit the proto needs

This commit is contained in:
ManyTheFish 2022-07-11 17:27:07 +02:00
parent 96acbf815d
commit b1bf6722e8
4 changed files with 88 additions and 21 deletions

View File

@ -10,7 +10,7 @@ use http::header::CONTENT_TYPE;
use meilisearch_auth::SearchRules; use meilisearch_auth::SearchRules;
use meilisearch_lib::index::{ use meilisearch_lib::index::{
SearchQuery, SearchResult, DEFAULT_CROP_LENGTH, DEFAULT_CROP_MARKER, SearchQuery, SearchResult, DEFAULT_CROP_LENGTH, DEFAULT_CROP_MARKER,
DEFAULT_HIGHLIGHT_POST_TAG, DEFAULT_HIGHLIGHT_PRE_TAG, DEFAULT_HIGHLIGHT_POST_TAG, DEFAULT_HIGHLIGHT_PRE_TAG, DEFAULT_SEARCH_LIMIT,
}; };
use meilisearch_lib::index_controller::Stats; use meilisearch_lib::index_controller::Stats;
use meilisearch_lib::MeiliSearch; use meilisearch_lib::MeiliSearch;
@ -373,6 +373,7 @@ pub struct SearchAggregator {
// pagination // pagination
max_limit: usize, max_limit: usize,
max_offset: usize, max_offset: usize,
finite_pagination: bool,
// formatting // formatting
highlight_pre_tag: bool, highlight_pre_tag: bool,
@ -427,12 +428,19 @@ impl SearchAggregator {
ret.max_terms_number = q.split_whitespace().count(); ret.max_terms_number = q.split_whitespace().count();
} }
if query.limit.is_none() && query.offset.is_none() {
ret.max_limit = query.hits_per_page;
ret.max_offset = query.page.saturating_sub(1) * query.hits_per_page;
ret.finite_pagination = true;
} else {
ret.max_limit = query.limit.unwrap_or_else(DEFAULT_SEARCH_LIMIT);
ret.max_offset = query.offset.unwrap_or_default();
ret.finite_pagination = false;
}
ret.matching_strategy ret.matching_strategy
.insert(format!("{:?}", query.matching_strategy), 1); .insert(format!("{:?}", query.matching_strategy), 1);
ret.max_limit = query.limit;
ret.max_offset = query.offset.unwrap_or_default();
ret.highlight_pre_tag = query.highlight_pre_tag != DEFAULT_HIGHLIGHT_PRE_TAG(); ret.highlight_pre_tag = query.highlight_pre_tag != DEFAULT_HIGHLIGHT_PRE_TAG();
ret.highlight_post_tag = query.highlight_post_tag != DEFAULT_HIGHLIGHT_POST_TAG(); ret.highlight_post_tag = query.highlight_post_tag != DEFAULT_HIGHLIGHT_POST_TAG();
ret.crop_marker = query.crop_marker != DEFAULT_CROP_MARKER(); ret.crop_marker = query.crop_marker != DEFAULT_CROP_MARKER();

View File

@ -2,8 +2,8 @@ use actix_web::{web, HttpRequest, HttpResponse};
use log::debug; use log::debug;
use meilisearch_auth::IndexSearchRules; use meilisearch_auth::IndexSearchRules;
use meilisearch_lib::index::{ use meilisearch_lib::index::{
MatchingStrategy, SearchQuery, DEFAULT_CROP_LENGTH, DEFAULT_CROP_MARKER, SearchQuery, DEFAULT_CROP_LENGTH, DEFAULT_CROP_MARKER, DEFAULT_HIGHLIGHT_POST_TAG,
DEFAULT_HIGHLIGHT_POST_TAG, DEFAULT_HIGHLIGHT_PRE_TAG, DEFAULT_SEARCH_LIMIT, DEFAULT_HIGHLIGHT_PRE_TAG, DEFAULT_HIT_PER_PAGE, DEFAULT_PAGE, MatchingStrategy
}; };
use meilisearch_lib::MeiliSearch; use meilisearch_lib::MeiliSearch;
use meilisearch_types::error::ResponseError; use meilisearch_types::error::ResponseError;
@ -29,6 +29,10 @@ pub struct SearchQueryGet {
q: Option<String>, q: Option<String>,
offset: Option<usize>, offset: Option<usize>,
limit: Option<usize>, limit: Option<usize>,
#[serde(default = "DEFAULT_PAGE")]
page: usize,
#[serde(default = "DEFAULT_HIT_PER_PAGE")]
hits_per_page: usize,
attributes_to_retrieve: Option<CS<String>>, attributes_to_retrieve: Option<CS<String>>,
attributes_to_crop: Option<CS<String>>, attributes_to_crop: Option<CS<String>>,
#[serde(default = "DEFAULT_CROP_LENGTH")] #[serde(default = "DEFAULT_CROP_LENGTH")]
@ -62,7 +66,9 @@ impl From<SearchQueryGet> for SearchQuery {
Self { Self {
q: other.q, q: other.q,
offset: other.offset, offset: other.offset,
limit: other.limit.unwrap_or_else(DEFAULT_SEARCH_LIMIT), limit: other.limit,
page: other.page,
hits_per_page: other.hits_per_page,
attributes_to_retrieve: other attributes_to_retrieve: other
.attributes_to_retrieve .attributes_to_retrieve
.map(|o| o.into_iter().collect()), .map(|o| o.into_iter().collect()),

View File

@ -1,6 +1,7 @@
pub use search::{ pub use search::{
MatchingStrategy, SearchQuery, SearchResult, DEFAULT_CROP_LENGTH, DEFAULT_CROP_MARKER, SearchQuery, SearchResult, DEFAULT_CROP_LENGTH, DEFAULT_CROP_MARKER,
DEFAULT_HIGHLIGHT_POST_TAG, DEFAULT_HIGHLIGHT_PRE_TAG, DEFAULT_SEARCH_LIMIT, DEFAULT_HIGHLIGHT_POST_TAG, DEFAULT_HIGHLIGHT_PRE_TAG, DEFAULT_HIT_PER_PAGE, DEFAULT_PAGE,
DEFAULT_SEARCH_LIMIT, MatchingStrategy
}; };
pub use updates::{apply_settings_to_builder, Checked, Facets, Settings, Unchecked}; pub use updates::{apply_settings_to_builder, Checked, Facets, Settings, Unchecked};

View File

@ -1,4 +1,4 @@
use std::cmp::min; use std::cmp::{max, min};
use std::collections::{BTreeMap, BTreeSet, HashSet}; use std::collections::{BTreeMap, BTreeSet, HashSet};
use std::str::FromStr; use std::str::FromStr;
use std::time::Instant; use std::time::Instant;
@ -26,6 +26,8 @@ pub const DEFAULT_CROP_LENGTH: fn() -> usize = || 10;
pub const DEFAULT_CROP_MARKER: fn() -> String = || "".to_string(); pub const DEFAULT_CROP_MARKER: fn() -> String = || "".to_string();
pub const DEFAULT_HIGHLIGHT_PRE_TAG: fn() -> String = || "<em>".to_string(); pub const DEFAULT_HIGHLIGHT_PRE_TAG: fn() -> String = || "<em>".to_string();
pub const DEFAULT_HIGHLIGHT_POST_TAG: fn() -> String = || "</em>".to_string(); pub const DEFAULT_HIGHLIGHT_POST_TAG: fn() -> String = || "</em>".to_string();
pub const DEFAULT_PAGE: fn() -> usize = || 1;
pub const DEFAULT_HIT_PER_PAGE: fn() -> usize = || 20;
/// The maximum number of results that the engine /// The maximum number of results that the engine
/// will be able to return in one search call. /// will be able to return in one search call.
@ -36,8 +38,11 @@ pub const DEFAULT_PAGINATION_MAX_TOTAL_HITS: usize = 1000;
pub struct SearchQuery { pub struct SearchQuery {
pub q: Option<String>, pub q: Option<String>,
pub offset: Option<usize>, pub offset: Option<usize>,
#[serde(default = "DEFAULT_SEARCH_LIMIT")] pub limit: Option<usize>,
pub limit: usize, #[serde(default = "DEFAULT_PAGE")]
pub page: usize,
#[serde(default = "DEFAULT_HIT_PER_PAGE")]
pub hits_per_page: usize,
pub attributes_to_retrieve: Option<BTreeSet<String>>, pub attributes_to_retrieve: Option<BTreeSet<String>>,
pub attributes_to_crop: Option<Vec<String>>, pub attributes_to_crop: Option<Vec<String>>,
#[serde(default = "DEFAULT_CROP_LENGTH")] #[serde(default = "DEFAULT_CROP_LENGTH")]
@ -97,15 +102,30 @@ pub struct SearchHit {
#[serde(rename_all = "camelCase")] #[serde(rename_all = "camelCase")]
pub struct SearchResult { pub struct SearchResult {
pub hits: Vec<SearchHit>, pub hits: Vec<SearchHit>,
pub estimated_total_hits: u64,
pub query: String, pub query: String,
pub limit: usize,
pub offset: usize,
pub processing_time_ms: u128, pub processing_time_ms: u128,
#[serde(flatten)]
pub hits_info: HitsInfo,
#[serde(skip_serializing_if = "Option::is_none")] #[serde(skip_serializing_if = "Option::is_none")]
pub facet_distribution: Option<BTreeMap<String, BTreeMap<String, u64>>>, pub facet_distribution: Option<BTreeMap<String, BTreeMap<String, u64>>>,
} }
#[derive(Serialize, Debug, Clone, PartialEq)]
#[serde(rename_all = "camelCase")]
#[serde(untagged)]
pub enum HitsInfo {
Pagination {
hits_per_page: usize,
page: usize,
total_pages: usize,
},
OffsetLimit {
limit: usize,
offset: usize,
estimated_total_hits: usize,
},
}
impl Index { impl Index {
pub fn perform_search(&self, query: SearchQuery) -> Result<SearchResult> { pub fn perform_search(&self, query: SearchQuery) -> Result<SearchResult> {
let before_search = Instant::now(); let before_search = Instant::now();
@ -125,8 +145,26 @@ impl Index {
// Make sure that a user can't get more documents than the hard limit, // Make sure that a user can't get more documents than the hard limit,
// we align that on the offset too. // we align that on the offset too.
let offset = min(query.offset.unwrap_or(0), max_total_hits); let is_finite_pagination = query.offset.is_none() && query.limit.is_none();
let limit = min(query.limit, max_total_hits.saturating_sub(offset));
let (offset, limit) = if is_finite_pagination {
// we start at least at page 1.
let page = max(query.page, 1);
// return at least 1 document.
let hits_per_page = max(query.hits_per_page, 1);
let offset = min(hits_per_page * (page - 1), max_total_hits);
let limit = min(hits_per_page, max_total_hits.saturating_sub(offset));
(offset, limit)
} else {
let offset = min(query.offset.unwrap_or(0), max_total_hits);
let limit = min(
query.limit.unwrap_or_else(DEFAULT_SEARCH_LIMIT),
max_total_hits.saturating_sub(offset),
);
(offset, limit)
};
search.offset(offset); search.offset(offset);
search.limit(limit); search.limit(limit);
@ -251,7 +289,23 @@ impl Index {
documents.push(hit); documents.push(hit);
} }
let estimated_total_hits = candidates.len(); let number_of_hits = min(candidates.len() as usize, max_total_hits);
let hits_info = if is_finite_pagination {
// return at least 1 document.
let hits_per_page = max(query.hits_per_page, 1);
HitsInfo::Pagination {
hits_per_page,
page: offset / hits_per_page + 1,
// TODO @many: estimation for now but we should ask milli to return an exact value
total_pages: (number_of_hits + hits_per_page - 1) / query.hits_per_page,
}
} else {
HitsInfo::OffsetLimit {
limit: query.limit.unwrap_or_else(DEFAULT_SEARCH_LIMIT),
offset,
estimated_total_hits: number_of_hits,
}
};
let facet_distribution = match query.facets { let facet_distribution = match query.facets {
Some(ref fields) => { Some(ref fields) => {
@ -274,10 +328,8 @@ impl Index {
let result = SearchResult { let result = SearchResult {
hits: documents, hits: documents,
estimated_total_hits, hits_info,
query: query.q.clone().unwrap_or_default(), query: query.q.clone().unwrap_or_default(),
limit: query.limit,
offset: query.offset.unwrap_or_default(),
processing_time_ms: before_search.elapsed().as_millis(), processing_time_ms: before_search.elapsed().as_millis(),
facet_distribution, facet_distribution,
}; };