remove unused import

meilisearch-types/src/error.rs

@@ -236,10 +236,13 @@ InvalidSearchHighlightPreTag          , InvalidRequest       , BAD_REQUEST ;
 InvalidSearchHitsPerPage              , InvalidRequest       , BAD_REQUEST ;
 InvalidSearchLimit                    , InvalidRequest       , BAD_REQUEST ;
 InvalidSearchMatchingStrategy         , InvalidRequest       , BAD_REQUEST ;
+InvalidMultiSearchMergeStrategy       , InvalidRequest       , BAD_REQUEST ;
 InvalidSearchOffset                   , InvalidRequest       , BAD_REQUEST ;
 InvalidSearchPage                     , InvalidRequest       , BAD_REQUEST ;
 InvalidSearchQ                        , InvalidRequest       , BAD_REQUEST ;
 InvalidSearchShowMatchesPosition      , InvalidRequest       , BAD_REQUEST ;
+InvalidSearchShowRankingScore         , InvalidRequest       , BAD_REQUEST ;
+InvalidSearchShowRankingScoreDetails  , InvalidRequest       , BAD_REQUEST ;
 InvalidSearchSort                     , InvalidRequest       , BAD_REQUEST ;
 InvalidSettingsDisplayedAttributes    , InvalidRequest       , BAD_REQUEST ;
 InvalidSettingsDistinctAttribute      , InvalidRequest       , BAD_REQUEST ;

meilisearch/src/routes/indexes/search.rs

@@ -56,6 +56,10 @@ pub struct SearchQueryGet {
     sort: Option<String>,
     #[deserr(default, error = DeserrQueryParamError<InvalidSearchShowMatchesPosition>)]
     show_matches_position: Param<bool>,
+    #[deserr(default, error = DeserrQueryParamError<InvalidSearchShowRankingScore>)]
+    show_ranking_score: Param<bool>,
+    #[deserr(default, error = DeserrQueryParamError<InvalidSearchShowRankingScoreDetails>)]
+    show_ranking_score_details: Param<bool>,
     #[deserr(default, error = DeserrQueryParamError<InvalidSearchFacets>)]
     facets: Option<CS<String>>,
     #[deserr( default = DEFAULT_HIGHLIGHT_PRE_TAG(), error = DeserrQueryParamError<InvalidSearchHighlightPreTag>)]
@@ -91,6 +95,8 @@ impl From<SearchQueryGet> for SearchQuery {
             filter,
             sort: other.sort.map(|attr| fix_sort_query_parameters(&attr)),
             show_matches_position: other.show_matches_position.0,
+            show_ranking_score: other.show_ranking_score.0,
+            show_ranking_score_details: other.show_ranking_score_details.0,
             facets: other.facets.map(|o| o.into_iter().collect()),
             highlight_pre_tag: other.highlight_pre_tag,
             highlight_post_tag: other.highlight_post_tag,

meilisearch/src/routes/multi_search.rs

@@ -1,20 +1,26 @@
+use std::collections::HashMap;
+
 use actix_http::StatusCode;
 use actix_web::web::{self, Data};
 use actix_web::{HttpRequest, HttpResponse};
 use deserr::actix_web::AwebJson;
+use deserr::Deserr;
 use index_scheduler::IndexScheduler;
 use log::debug;
 use meilisearch_types::deserr::DeserrJsonError;
+use meilisearch_types::error::deserr_codes::InvalidMultiSearchMergeStrategy;
 use meilisearch_types::error::ResponseError;
 use meilisearch_types::keys::actions;
+use meilisearch_types::milli::score_details::NotComparable;
 use serde::Serialize;
 
 use crate::analytics::{Analytics, MultiSearchAggregator};
 use crate::extractors::authentication::policies::ActionPolicy;
 use crate::extractors::authentication::{AuthenticationError, GuardedData};
 use crate::extractors::sequential_extractor::SeqHandler;
+use crate::milli::score_details::ScoreDetails;
 use crate::search::{
-    add_search_rules, perform_search, SearchQueryWithIndex, SearchResultWithIndex,
+    add_search_rules, perform_search, SearchHit, SearchQueryWithIndex, SearchResultWithIndex,
 };
 
 pub fn configure(cfg: &mut web::ServiceConfig) {
@@ -23,13 +29,34 @@ pub fn configure(cfg: &mut web::ServiceConfig) {
 
 #[derive(Serialize)]
 struct SearchResults {
+    #[serde(skip_serializing_if = "Option::is_none")]
+    aggregate_hits: Option<Vec<SearchHitWithIndex>>,
     results: Vec<SearchResultWithIndex>,
 }
 
+#[derive(Serialize, Debug, Clone, PartialEq)]
+#[serde(rename_all = "camelCase")]
+struct SearchHitWithIndex {
+    pub index_uid: String,
+    #[serde(flatten)]
+    pub hit: SearchHit,
+}
+
 #[derive(Debug, deserr::Deserr)]
 #[deserr(error = DeserrJsonError, rename_all = camelCase, deny_unknown_fields)]
 pub struct SearchQueries {
     queries: Vec<SearchQueryWithIndex>,
+    #[deserr(default, error = DeserrJsonError<InvalidMultiSearchMergeStrategy>, default)]
+    merge_strategy: MergeStrategy,
+}
+
+#[derive(Debug, Clone, PartialEq, Eq, Deserr, Default)]
+#[deserr(rename_all = camelCase)]
+pub enum MergeStrategy {
+    #[default]
+    None,
+    ByNormalizedScore,
+    ByScoreDetails,
 }
 
 pub async fn multi_search_with_post(
@@ -38,7 +65,13 @@ pub async fn multi_search_with_post(
     req: HttpRequest,
     analytics: web::Data<dyn Analytics>,
 ) -> Result<HttpResponse, ResponseError> {
-    let queries = params.into_inner().queries;
+    let SearchQueries { queries, merge_strategy } = params.into_inner();
+    // FIXME: REMOVE UNWRAP
+    let max_hits = queries
+        .iter()
+        .map(|SearchQueryWithIndex { limit, hits_per_page, .. }| hits_per_page.unwrap_or(*limit))
+        .max()
+        .unwrap();
 
     let mut multi_aggregate = MultiSearchAggregator::from_queries(&queries, &req);
 
@@ -104,7 +137,117 @@ pub async fn multi_search_with_post(
 
     debug!("returns: {:?}", search_results);
 
-    Ok(HttpResponse::Ok().json(SearchResults { results: search_results }))
+    let aggregate_hits = match merge_strategy {
+        MergeStrategy::None => None,
+        MergeStrategy::ByScoreDetails => Some(merge_by_score_details(&search_results, max_hits)),
+        MergeStrategy::ByNormalizedScore => {
+            Some(merge_by_normalized_score(&search_results, max_hits))
+        }
+    };
+
+    Ok(HttpResponse::Ok().json(SearchResults { aggregate_hits, results: search_results }))
+}
+
+fn merge_by_score_details(
+    search_results: &[SearchResultWithIndex],
+    max_hits: usize,
+) -> Vec<SearchHitWithIndex> {
+    let mut iterators: Vec<_> = search_results
+        .iter()
+        .filter_map(|SearchResultWithIndex { index_uid, result }| {
+            let mut it = result.hits.iter();
+            let next = it.next()?;
+            Some((index_uid, it, next))
+        })
+        .collect();
+
+    let mut hits = Vec::with_capacity(max_hits);
+
+    let mut inconsistent_indexes = HashMap::new();
+
+    for _ in 0..max_hits {
+        iterators.sort_by(|(left_uid, _, left_hit), (right_uid, _, right_hit)| {
+            let error = match ScoreDetails::partial_cmp_iter(
+                left_hit.ranking_score_raw.iter(),
+                right_hit.ranking_score_raw.iter(),
+            ) {
+                Ok(ord) => return ord,
+                Err(NotComparable(incomparable_index)) => incomparable_index,
+            };
+            inconsistent_indexes.entry((left_uid.to_owned(), right_uid.to_owned())).or_insert_with(
+                || {
+                    format!(
+                        "Detailed score {:?} is not comparable with {:?}: (left: {:#?}, right: {:#?})",
+                        left_hit.ranking_score_raw.get(error),
+                        right_hit.ranking_score_raw.get(error),
+                        left_hit.ranking_score_raw,
+                        right_hit.ranking_score_raw
+                    )
+                },
+            );
+            std::cmp::Ordering::Less
+        });
+        if !inconsistent_indexes.is_empty() {
+            let mut s = String::new();
+            for ((left_uid, right_uid), error) in &inconsistent_indexes {
+                use std::fmt::Write;
+                writeln!(s, "Indexes {} and {} are inconsistent: {}", left_uid, right_uid, error)
+                    .unwrap();
+            }
+            // Replace panic with proper error
+            panic!("{}", s);
+        }
+
+        let Some((index_uid, it, next)) = iterators.last_mut()
+        else {
+            break;
+        };
+
+        let hit = SearchHitWithIndex { index_uid: index_uid.clone(), hit: next.clone() };
+        if let Some(next_hit) = it.next() {
+            *next = next_hit;
+        } else {
+            iterators.pop();
+        }
+        hits.push(hit);
+    }
+    hits
+}
+
+fn merge_by_normalized_score(
+    search_results: &[SearchResultWithIndex],
+    max_hits: usize,
+) -> Vec<SearchHitWithIndex> {
+    let mut iterators: Vec<_> = search_results
+        .iter()
+        .filter_map(|SearchResultWithIndex { index_uid, result }| {
+            let mut it = result.hits.iter();
+            let next = it.next()?;
+            Some((index_uid, it, next))
+        })
+        .collect();
+
+    let mut hits = Vec::with_capacity(max_hits);
+
+    for _ in 0..max_hits {
+        iterators.sort_by_key(|(_, _, hit)| {
+            ScoreDetails::global_score_linear_scale(hit.ranking_score_raw.iter())
+        });
+
+        let Some((index_uid, it, next)) = iterators.last_mut()
+        else {
+            break;
+        };
+
+        let hit = SearchHitWithIndex { index_uid: index_uid.clone(), hit: next.clone() };
+        if let Some(next_hit) = it.next() {
+            *next = next_hit;
+        } else {
+            iterators.pop();
+        }
+        hits.push(hit);
+    }
+    hits
 }
 
 /// Local `Result` extension trait to avoid `map_err` boilerplate.

meilisearch/src/search.rs

@@ -9,6 +9,7 @@ use meilisearch_auth::IndexSearchRules;
 use meilisearch_types::deserr::DeserrJsonError;
 use meilisearch_types::error::deserr_codes::*;
 use meilisearch_types::index_uid::IndexUid;
+use meilisearch_types::milli::score_details::ScoreDetails;
 use meilisearch_types::settings::DEFAULT_PAGINATION_MAX_TOTAL_HITS;
 use meilisearch_types::{milli, Document};
 use milli::tokenizer::TokenizerBuilder;
@@ -54,6 +55,10 @@ pub struct SearchQuery {
     pub attributes_to_highlight: Option<HashSet<String>>,
     #[deserr(default, error = DeserrJsonError<InvalidSearchShowMatchesPosition>, default)]
     pub show_matches_position: bool,
+    #[deserr(default, error = DeserrJsonError<InvalidSearchShowRankingScore>, default)]
+    pub show_ranking_score: bool,
+    #[deserr(default, error = DeserrJsonError<InvalidSearchShowRankingScoreDetails>, default)]
+    pub show_ranking_score_details: bool,
     #[deserr(default, error = DeserrJsonError<InvalidSearchFilter>)]
     pub filter: Option<Value>,
     #[deserr(default, error = DeserrJsonError<InvalidSearchSort>)]
@@ -103,6 +108,10 @@ pub struct SearchQueryWithIndex {
     pub crop_length: usize,
     #[deserr(default, error = DeserrJsonError<InvalidSearchAttributesToHighlight>)]
     pub attributes_to_highlight: Option<HashSet<String>>,
+    #[deserr(default, error = DeserrJsonError<InvalidSearchShowRankingScore>, default)]
+    pub show_ranking_score: bool,
+    #[deserr(default, error = DeserrJsonError<InvalidSearchShowRankingScoreDetails>, default)]
+    pub show_ranking_score_details: bool,
     #[deserr(default, error = DeserrJsonError<InvalidSearchShowMatchesPosition>, default)]
     pub show_matches_position: bool,
     #[deserr(default, error = DeserrJsonError<InvalidSearchFilter>)]
@@ -134,6 +143,8 @@ impl SearchQueryWithIndex {
             attributes_to_crop,
             crop_length,
             attributes_to_highlight,
+            show_ranking_score,
+            show_ranking_score_details,
             show_matches_position,
             filter,
             sort,
@@ -155,6 +166,8 @@ impl SearchQueryWithIndex {
                 attributes_to_crop,
                 crop_length,
                 attributes_to_highlight,
+                show_ranking_score,
+                show_ranking_score_details,
                 show_matches_position,
                 filter,
                 sort,
@@ -194,7 +207,7 @@ impl From<MatchingStrategy> for TermsMatchingStrategy {
     }
 }
 
-#[derive(Debug, Clone, Serialize, PartialEq, Eq)]
+#[derive(Debug, Clone, Serialize, PartialEq)]
 pub struct SearchHit {
     #[serde(flatten)]
     pub document: Document,
@@ -202,6 +215,12 @@ pub struct SearchHit {
     pub formatted: Document,
     #[serde(rename = "_matchesPosition", skip_serializing_if = "Option::is_none")]
     pub matches_position: Option<MatchesPosition>,
+    #[serde(rename = "_rankingScore", skip_serializing_if = "Option::is_none")]
+    pub ranking_score: Option<u64>,
+    #[serde(rename = "_rankingScoreDetails", skip_serializing_if = "Option::is_none")]
+    pub ranking_score_details: Option<serde_json::Map<String, serde_json::Value>>,
+    #[serde(skip)]
+    pub ranking_score_raw: Vec<ScoreDetails>,
 }
 
 #[derive(Serialize, Debug, Clone, PartialEq)]
@@ -320,7 +339,8 @@ pub fn perform_search(
         search.sort_criteria(sort);
     }
 
-    let milli::SearchResult { documents_ids, matching_words, candidates, .. } = search.execute()?;
+    let milli::SearchResult { documents_ids, matching_words, candidates, document_scores, .. } =
+        search.execute()?;
 
     let fields_ids_map = index.fields_ids_map(&rtxn).unwrap();
 
@@ -392,7 +412,7 @@ pub fn perform_search(
 
     let documents_iter = index.documents(&rtxn, documents_ids)?;
 
-    for (_id, obkv) in documents_iter {
+    for ((_id, obkv), score) in documents_iter.into_iter().zip(document_scores.into_iter()) {
         // First generate a document with all the displayed fields
         let displayed_document = make_document(&displayed_ids, &fields_ids_map, obkv)?;
 
@@ -416,7 +436,19 @@ pub fn perform_search(
             insert_geo_distance(sort, &mut document);
         }
 
-        let hit = SearchHit { document, formatted, matches_position };
+        let ranking_score =
+            query.show_ranking_score.then(|| ScoreDetails::global_score_linear_scale(score.iter()));
+        let ranking_score_details =
+            query.show_ranking_score_details.then(|| ScoreDetails::to_json_map(score.iter()));
+
+        let hit = SearchHit {
+            document,
+            formatted,
+            matches_position,
+            ranking_score_details,
+            ranking_score,
+            ranking_score_raw: score,
+        };
         documents.push(hit);
     }
 

meilisearch/tests/search/formatted.rs

@@ -1,3 +1,4 @@
+use insta::{allow_duplicates, assert_json_snapshot};
 use serde_json::json;
 
 use super::*;
@@ -18,30 +19,45 @@ async fn formatted_contain_wildcard() {
         |response, code|
         {
             assert_eq!(code, 200, "{}", response);
-            assert_eq!(
+            allow_duplicates! {
-                response["hits"][0],
+              assert_json_snapshot!(response["hits"][0],
-                json!({
+                    { "._rankingScore" => "[score]" },
-                    "_formatted": {
+                    @r###"
-                        "id": "852",
+              {
-                        "cattos": "<em>pésti</em>",
+                "_formatted": {
-                    },
+                  "id": "852",
-                    "_matchesPosition": {"cattos": [{"start": 0, "length": 5}]},
+                  "cattos": "<em>pésti</em>"
-                })
+                },
-            );
+                "_matchesPosition": {
-        }
+                  "cattos": [
+                    {
+                      "start": 0,
+                      "length": 5
+                    }
+                  ]
+                },
+                "_rankingScore": "[score]"
+              }
+              "###);
+            }
+    }
     )
     .await;
 
     index
         .search(json!({ "q": "pésti", "attributesToRetrieve": ["*"] }), |response, code| {
             assert_eq!(code, 200, "{}", response);
-            assert_eq!(
+            allow_duplicates! {
-                response["hits"][0],
+                assert_json_snapshot!(response["hits"][0],
-                json!({
+                { "._rankingScore" => "[score]" },
-                    "id": 852,
+                @r###"
-                    "cattos": "pésti",
+                {
-                })
+                  "id": 852,
-            );
+                  "cattos": "pésti",
+                  "_rankingScore": "[score]"
+                }
+                "###)
+            }
         })
         .await;
 
@@ -50,20 +66,30 @@ async fn formatted_contain_wildcard() {
             json!({ "q": "pésti", "attributesToRetrieve": ["*"], "attributesToHighlight": ["id"], "showMatchesPosition": true }),
             |response, code| {
                 assert_eq!(code, 200, "{}", response);
-                assert_eq!(
+                allow_duplicates! {
-                    response["hits"][0],
+                  assert_json_snapshot!(response["hits"][0],
-                    json!({
+                 { "._rankingScore" => "[score]" },
-                        "id": 852,
+                 @r###"
-                        "cattos": "pésti",
+                  {
-                        "_formatted": {
+                    "id": 852,
-                            "id": "852",
+                    "cattos": "pésti",
-                            "cattos": "pésti",
+                    "_formatted": {
-                        },
+                      "id": "852",
-                        "_matchesPosition": {"cattos": [{"start": 0, "length": 5}]},
+                      "cattos": "pésti"
-                    })
+                    },
-                );
+                    "_matchesPosition": {
-            }
+                      "cattos": [
-        )
+                        {
+                          "start": 0,
+                          "length": 5
+                        }
+                      ]
+                    },
+                    "_rankingScore": "[score]"
+                  }
+                  "###)
+             }
+        })
         .await;
 
     index
@@ -71,17 +97,21 @@ async fn formatted_contain_wildcard() {
             json!({ "q": "pésti", "attributesToRetrieve": ["*"], "attributesToCrop": ["*"] }),
             |response, code| {
                 assert_eq!(code, 200, "{}", response);
-                assert_eq!(
+                allow_duplicates! {
-                    response["hits"][0],
+                    assert_json_snapshot!(response["hits"][0],
-                    json!({
+                    { "._rankingScore" => "[score]" },
-                        "id": 852,
+                    @r###"
-                        "cattos": "pésti",
+                    {
-                        "_formatted": {
+                      "id": 852,
-                            "id": "852",
+                      "cattos": "pésti",
-                            "cattos": "pésti",
+                      "_formatted": {
-                        }
+                        "id": "852",
-                    })
+                        "cattos": "pésti"
-                );
+                      },
+                      "_rankingScore": "[score]"
+                    }
+                    "###);
+                }
             },
         )
         .await;
@@ -89,17 +119,21 @@ async fn formatted_contain_wildcard() {
     index
         .search(json!({ "q": "pésti", "attributesToCrop": ["*"] }), |response, code| {
             assert_eq!(code, 200, "{}", response);
-            assert_eq!(
+            allow_duplicates! {
-                response["hits"][0],
+                assert_json_snapshot!(response["hits"][0],
-                json!({
+                { "._rankingScore" => "[score]" },
-                    "id": 852,
+                @r###"
-                    "cattos": "pésti",
+                {
-                    "_formatted": {
+                  "id": 852,
-                        "id": "852",
+                  "cattos": "pésti",
-                        "cattos": "pésti",
+                  "_formatted": {
-                    }
+                    "id": "852",
-                })
+                    "cattos": "pésti"
-            );
+                  },
+                  "_rankingScore": "[score]"
+                }
+                "###)
+            }
         })
         .await;
 }
@@ -116,21 +150,25 @@ async fn format_nested() {
     index
         .search(json!({ "q": "pésti", "attributesToRetrieve": ["doggos"] }), |response, code| {
             assert_eq!(code, 200, "{}", response);
-            assert_eq!(
+            allow_duplicates! {
-                response["hits"][0],
+                assert_json_snapshot!(response["hits"][0],
-                json!({
+                { "._rankingScore" => "[score]" },
-                    "doggos": [
+                @r###"
-                        {
+                {
-                            "name": "bobby",
+                  "doggos": [
-                            "age": 2,
+                    {
-                        },
+                      "name": "bobby",
-                        {
+                      "age": 2
-                            "name": "buddy",
+                    },
-                            "age": 4,
+                    {
-                        },
+                      "name": "buddy",
-                    ],
+                      "age": 4
-                })
+                    }
-            );
+                  ],
+                  "_rankingScore": "[score]"
+                }
+                "###)
+            }
         })
         .await;
 
@@ -139,19 +177,23 @@ async fn format_nested() {
             json!({ "q": "pésti", "attributesToRetrieve": ["doggos.name"] }),
             |response, code| {
                 assert_eq!(code, 200, "{}", response);
-                assert_eq!(
+                allow_duplicates! {
-                    response["hits"][0],
+                    assert_json_snapshot!(response["hits"][0],
-                    json!({
+                    { "._rankingScore" => "[score]" },
-                        "doggos": [
+                    @r###"
-                            {
+                    {
-                                "name": "bobby",
+                      "doggos": [
-                            },
+                        {
-                            {
+                          "name": "bobby"
-                                "name": "buddy",
+                        },
-                            },
+                        {
-                        ],
+                          "name": "buddy"
-                    })
+                        }
-                );
+                      ],
+                      "_rankingScore": "[score]"
+                    }
+                    "###)
+                }
             },
         )
         .await;
@@ -161,20 +203,31 @@ async fn format_nested() {
             json!({ "q": "bobby", "attributesToRetrieve": ["doggos.name"], "showMatchesPosition": true }),
             |response, code| {
                 assert_eq!(code, 200, "{}", response);
-                assert_eq!(
+                allow_duplicates! {
-                    response["hits"][0],
+                    assert_json_snapshot!(response["hits"][0],
-                    json!({
+                    { "._rankingScore" => "[score]" },
-                        "doggos": [
+                    @r###"
-                            {
+                    {
-                                "name": "bobby",
+                      "doggos": [
-                            },
+                        {
-                            {
+                          "name": "bobby"
-                                "name": "buddy",
+                        },
-                            },
+                        {
-                        ],
+                          "name": "buddy"
-                        "_matchesPosition": {"doggos.name": [{"start": 0, "length": 5}]},
+                        }
-                    })
+                      ],
-                );
+                      "_matchesPosition": {
+                        "doggos.name": [
+                          {
+                            "start": 0,
+                            "length": 5
+                          }
+                        ]
+                      },
+                      "_rankingScore": "[score]"
+                    }
+                    "###)
+                }
             }
         )
         .await;
@@ -183,21 +236,25 @@ async fn format_nested() {
         .search(json!({ "q": "pésti", "attributesToRetrieve": [], "attributesToHighlight": ["doggos.name"] }),
         |response, code| {
             assert_eq!(code, 200, "{}", response);
-            assert_eq!(
+            allow_duplicates! {
-                response["hits"][0],
+                assert_json_snapshot!(response["hits"][0],
-                json!({
+                { "._rankingScore" => "[score]" },
-                    "_formatted": {
+                @r###"
-                        "doggos": [
+                {
-                            {
+                  "_formatted": {
-                                "name": "bobby",
+                    "doggos": [
-                            },
+                      {
-                            {
+                        "name": "bobby"
-                                "name": "buddy",
+                      },
-                            },
+                      {
-                        ],
+                        "name": "buddy"
-                    },
+                      }
-                })
+                    ]
-            );
+                  },
+                  "_rankingScore": "[score]"
+                }
+                "###)
+            }
         })
         .await;
 
@@ -205,21 +262,25 @@ async fn format_nested() {
         .search(json!({ "q": "pésti", "attributesToRetrieve": [], "attributesToCrop": ["doggos.name"] }),
         |response, code| {
             assert_eq!(code, 200, "{}", response);
-            assert_eq!(
+            allow_duplicates! {
-                response["hits"][0],
+                assert_json_snapshot!(response["hits"][0],
-                json!({
+                { "._rankingScore" => "[score]" },
-                    "_formatted": {
+                @r###"
-                        "doggos": [
+                {
-                            {
+                  "_formatted": {
-                                "name": "bobby",
+                    "doggos": [
-                            },
+                      {
-                            {
+                        "name": "bobby"
-                                "name": "buddy",
+                      },
-                            },
+                      {
-                        ],
+                        "name": "buddy"
-                    },
+                      }
-                })
+                    ]
-            );
+                  },
+                  "_rankingScore": "[score]"
+                }
+                "###)
+            }
         })
         .await;
 
@@ -227,55 +288,63 @@ async fn format_nested() {
         .search(json!({ "q": "pésti", "attributesToRetrieve": ["doggos.name"], "attributesToHighlight": ["doggos.age"] }),
         |response, code| {
     assert_eq!(code, 200, "{}", response);
-    assert_eq!(
+    allow_duplicates! {
-        response["hits"][0],
+        assert_json_snapshot!(response["hits"][0],
-        json!({
+        { "._rankingScore" => "[score]" },
-            "doggos": [
+        @r###"
-                {
+        {
-                    "name": "bobby",
+          "doggos": [
-                },
+            {
-                {
+              "name": "bobby"
-                    "name": "buddy",
-                },
-            ],
-            "_formatted": {
-                "doggos": [
-                    {
-                        "name": "bobby",
-                        "age": "2",
-                    },
-                    {
-                        "name": "buddy",
-                        "age": "4",
-                    },
-                ],
             },
-        })
+            {
-    );
+              "name": "buddy"
-        })
+            }
+          ],
+          "_formatted": {
+            "doggos": [
+              {
+                "name": "bobby",
+                "age": "2"
+              },
+              {
+                "name": "buddy",
+                "age": "4"
+              }
+            ]
+          },
+          "_rankingScore": "[score]"
+        }
+        "###)
+    }
+    })
         .await;
 
     index
         .search(json!({ "q": "pésti", "attributesToRetrieve": [], "attributesToHighlight": ["doggos.age"], "attributesToCrop": ["doggos.name"] }),
         |response, code| {
                 assert_eq!(code, 200, "{}", response);
-    assert_eq!(
+                allow_duplicates! {
-        response["hits"][0],
+                    assert_json_snapshot!(response["hits"][0],
-        json!({
+                    { "._rankingScore" => "[score]" },
-            "_formatted": {
+                    @r###"
-                "doggos": [
                     {
-                        "name": "bobby",
+                      "_formatted": {
-                        "age": "2",
+                        "doggos": [
-                    },
+                          {
-                    {
+                            "name": "bobby",
-                        "name": "buddy",
+                            "age": "2"
-                        "age": "4",
+                          },
-                    },
+                          {
-                ],
+                            "name": "buddy",
-            },
+                            "age": "4"
-        })
+                          }
-    );
+                        ]
+                      },
+                      "_rankingScore": "[score]"
+                    }
+                    "###)
+                }
             }
         )
         .await;
@@ -297,54 +366,70 @@ async fn displayedattr_2_smol() {
         .search(json!({ "attributesToRetrieve": ["father", "id"], "attributesToHighlight": ["mother"], "attributesToCrop": ["cattos"] }),
         |response, code| {
     assert_eq!(code, 200, "{}", response);
-    assert_eq!(
+    allow_duplicates! {
-        response["hits"][0],
+        assert_json_snapshot!(response["hits"][0],
-        json!({
+        { "._rankingScore" => "[score]" },
-            "id": 852,
+        @r###"
-        })
+        {
-    );
+          "id": 852,
+          "_rankingScore": "[score]"
+        }
+        "###)
+    }
         })
         .await;
 
     index
         .search(json!({ "attributesToRetrieve": ["id"] }), |response, code| {
             assert_eq!(code, 200, "{}", response);
-            assert_eq!(
+            allow_duplicates! {
-                response["hits"][0],
+                assert_json_snapshot!(response["hits"][0],
-                json!({
+                { "._rankingScore" => "[score]" },
-                    "id": 852,
+                @r###"
-                })
+                {
-            );
+                  "id": 852,
+                  "_rankingScore": "[score]"
+                }
+                "###)
+            }
         })
         .await;
 
     index
         .search(json!({ "attributesToHighlight": ["id"] }), |response, code| {
             assert_eq!(code, 200, "{}", response);
-            assert_eq!(
+            allow_duplicates! {
-                response["hits"][0],
+                assert_json_snapshot!(response["hits"][0],
-                json!({
+                { "._rankingScore" => "[score]" },
-                    "id": 852,
+                @r###"
-                    "_formatted": {
+                {
-                        "id": "852",
+                  "id": 852,
-                    }
+                  "_formatted": {
-                })
+                    "id": "852"
-            );
+                  },
+                  "_rankingScore": "[score]"
+                }
+                "###)
+            }
         })
         .await;
 
     index
         .search(json!({ "attributesToCrop": ["id"] }), |response, code| {
             assert_eq!(code, 200, "{}", response);
-            assert_eq!(
+            allow_duplicates! {
-                response["hits"][0],
+                assert_json_snapshot!(response["hits"][0],
-                json!({
+                { "._rankingScore" => "[score]" },
-                    "id": 852,
+                @r###"
-                    "_formatted": {
+                {
-                        "id": "852",
+                  "id": 852,
-                    }
+                  "_formatted": {
-                })
+                    "id": "852"
-            );
+                  },
+                  "_rankingScore": "[score]"
+                }
+                "###)
+            }
         })
         .await;
 
@@ -353,15 +438,19 @@ async fn displayedattr_2_smol() {
             json!({ "attributesToHighlight": ["id"], "attributesToCrop": ["id"] }),
             |response, code| {
                 assert_eq!(code, 200, "{}", response);
-                assert_eq!(
+                allow_duplicates! {
-                    response["hits"][0],
+                    assert_json_snapshot!(response["hits"][0],
-                    json!({
+                    { "._rankingScore" => "[score]" },
-                        "id": 852,
+                    @r###"
-                        "_formatted": {
+                    {
-                            "id": "852",
+                      "id": 852,
-                        }
+                      "_formatted": {
-                    })
+                        "id": "852"
-                );
+                      },
+                      "_rankingScore": "[score]"
+                    }
+                    "###)
+                }
             },
         )
         .await;
@@ -369,31 +458,47 @@ async fn displayedattr_2_smol() {
     index
         .search(json!({ "attributesToHighlight": ["cattos"] }), |response, code| {
             assert_eq!(code, 200, "{}", response);
-            assert_eq!(
+            allow_duplicates! {
-                response["hits"][0],
+                assert_json_snapshot!(response["hits"][0],
-                json!({
+                { "._rankingScore" => "[score]" },
-                    "id": 852,
+                @r###"
-                })
+                {
-            );
+                  "id": 852,
+                  "_rankingScore": "[score]"
+                }
+                "###)
+            }
         })
         .await;
 
     index
         .search(json!({ "attributesToCrop": ["cattos"] }), |response, code| {
             assert_eq!(code, 200, "{}", response);
-            assert_eq!(
+            allow_duplicates! {
-                response["hits"][0],
+                assert_json_snapshot!(response["hits"][0],
-                json!({
+                { "._rankingScore" => "[score]" },
-                    "id": 852,
+                @r###"
-                })
+                {
-            );
+                  "id": 852,
+                  "_rankingScore": "[score]"
+                }
+                "###)
+            }
         })
         .await;
 
     index
         .search(json!({ "attributesToRetrieve": ["cattos"] }), |response, code| {
             assert_eq!(code, 200, "{}", response);
-            assert_eq!(response["hits"][0], json!({}));
+            allow_duplicates! {
+                assert_json_snapshot!(response["hits"][0],
+                { "._rankingScore" => "[score]" },
+                @r###"
+                {
+                  "_rankingScore": "[score]"
+                }
+                "###)
+            }
         })
         .await;
 
@@ -402,7 +507,15 @@ async fn displayedattr_2_smol() {
             json!({ "attributesToRetrieve": ["cattos"], "attributesToHighlight": ["cattos"], "attributesToCrop": ["cattos"] }),
             |response, code| {
     assert_eq!(code, 200, "{}", response);
-    assert_eq!(response["hits"][0], json!({}));
+    allow_duplicates! {
+        assert_json_snapshot!(response["hits"][0],
+        { "._rankingScore" => "[score]" },
+        @r###"
+        {
+          "_rankingScore": "[score]"
+        }
+        "###)
+    }
 
             }
         )
@@ -413,14 +526,18 @@ async fn displayedattr_2_smol() {
             json!({ "attributesToRetrieve": ["cattos"], "attributesToHighlight": ["id"] }),
             |response, code| {
                 assert_eq!(code, 200, "{}", response);
-                assert_eq!(
+                allow_duplicates! {
-                    response["hits"][0],
+                    assert_json_snapshot!(response["hits"][0],
-                    json!({
+                    { "._rankingScore" => "[score]" },
-                        "_formatted": {
+                    @r###"
-                            "id": "852",
+                    {
-                        }
+                      "_formatted": {
-                    })
+                        "id": "852"
-                );
+                      },
+                      "_rankingScore": "[score]"
+                    }
+                    "###)
+                }
             },
         )
         .await;
@@ -430,14 +547,18 @@ async fn displayedattr_2_smol() {
             json!({ "attributesToRetrieve": ["cattos"], "attributesToCrop": ["id"] }),
             |response, code| {
                 assert_eq!(code, 200, "{}", response);
-                assert_eq!(
+                allow_duplicates! {
-                    response["hits"][0],
+                    assert_json_snapshot!(response["hits"][0],
-                    json!({
+                    { "._rankingScore" => "[score]" },
-                        "_formatted": {
+                    @r###"
-                            "id": "852",
+                    {
-                        }
+                      "_formatted": {
-                    })
+                        "id": "852"
-                );
+                      },
+                      "_rankingScore": "[score]"
+                    }
+                    "###)
+                }
             },
         )
         .await;

meilisearch/tests/search/multi.rs

@@ -65,14 +65,15 @@ async fn simple_search_single_index() {
         ]}))
         .await;
     snapshot!(code, @"200 OK");
-    insta::assert_json_snapshot!(response["results"], { "[].processingTimeMs" => "[time]" }, @r###"
+    insta::assert_json_snapshot!(response["results"], { "[].processingTimeMs" => "[time]", ".**._rankingScore" => "[score]" }, @r###"
     [
       {
         "indexUid": "test",
         "hits": [
           {
             "title": "Gläss",
-            "id": "450465"
+            "id": "450465",
+            "_rankingScore": "[score]"
           }
         ],
         "query": "glass",
@@ -86,7 +87,8 @@ async fn simple_search_single_index() {
         "hits": [
           {
             "title": "Captain Marvel",
-            "id": "299537"
+            "id": "299537",
+            "_rankingScore": "[score]"
           }
         ],
         "query": "captain",
@@ -170,14 +172,15 @@ async fn simple_search_two_indexes() {
         ]}))
         .await;
     snapshot!(code, @"200 OK");
-    insta::assert_json_snapshot!(response["results"], { "[].processingTimeMs" => "[time]" }, @r###"
+    insta::assert_json_snapshot!(response["results"], { "[].processingTimeMs" => "[time]", ".**._rankingScore" => "[score]" }, @r###"
     [
       {
         "indexUid": "test",
         "hits": [
           {
             "title": "Gläss",
-            "id": "450465"
+            "id": "450465",
+            "_rankingScore": "[score]"
           }
         ],
         "query": "glass",
@@ -203,7 +206,8 @@ async fn simple_search_two_indexes() {
                 "age": 4
               }
             ],
-            "cattos": "pésti"
+            "cattos": "pésti",
+            "_rankingScore": "[score]"
           },
           {
             "id": 654,
@@ -218,7 +222,8 @@ async fn simple_search_two_indexes() {
             "cattos": [
               "simba",
               "pestiféré"
-            ]
+            ],
+            "_rankingScore": "[score]"
           }
         ],
         "query": "pésti",

milli/src/index.rs

@@ -2488,8 +2488,12 @@ pub(crate) mod tests {
 
         let rtxn = index.read_txn().unwrap();
         let search = Search::new(&rtxn, &index);
-        let SearchResult { matching_words: _, candidates: _, mut documents_ids } =
+        let SearchResult {
-            search.execute().unwrap();
+            matching_words: _,
+            candidates: _,
+            document_scores: _,
+            mut documents_ids,
+        } = search.execute().unwrap();
         let primary_key_id = index.fields_ids_map(&rtxn).unwrap().id("primary_key").unwrap();
         documents_ids.sort_unstable();
         let docs = index.documents(&rtxn, documents_ids).unwrap();

milli/src/lib.rs

@@ -17,6 +17,7 @@ mod fields_ids_map;
 pub mod heed_codec;
 pub mod index;
 pub mod proximity;
+pub mod score_details;
 mod search;
 pub mod update;
 

milli/src/score_details.rs

@@ -0,0 +1,544 @@
+use std::cmp::Ordering;
+
+use serde::Serialize;
+
+use crate::distance_between_two_points;
+
+#[derive(Debug, Clone, PartialEq)]
+pub enum ScoreDetails {
+    Words(Words),
+    Typo(Typo),
+    Proximity(Rank),
+    Fid(Rank),
+    Position(Rank),
+    ExactAttribute(ExactAttribute),
+    Exactness(Rank),
+    Sort(Sort),
+    GeoSort(GeoSort),
+}
+
+impl PartialOrd for ScoreDetails {
+    fn partial_cmp(&self, other: &Self) -> Option<Ordering> {
+        use ScoreDetails::*;
+        match (self, other) {
+            // matching left and right hands => defer to sub impl
+            (Words(left), Words(right)) => left.partial_cmp(right),
+            (Typo(left), Typo(right)) => left.partial_cmp(right),
+            (Proximity(left), Proximity(right)) => left.partial_cmp(right),
+            (Fid(left), Fid(right)) => left.partial_cmp(right),
+            (Position(left), Position(right)) => left.partial_cmp(right),
+            (ExactAttribute(left), ExactAttribute(right)) => left.partial_cmp(right),
+            (Exactness(left), Exactness(right)) => left.partial_cmp(right),
+            (Sort(left), Sort(right)) => left.partial_cmp(right),
+            (GeoSort(left), GeoSort(right)) => left.partial_cmp(right),
+            // non matching left and right hands => None
+            // written this way rather than with a single `_` arm, so that adding a new variant
+            // still results in a compile error
+            (Words(_), _) => None,
+            (Typo(_), _) => None,
+            (Proximity(_), _) => None,
+            (Fid(_), _) => None,
+            (Position(_), _) => None,
+            (ExactAttribute(_), _) => None,
+            (Exactness(_), _) => None,
+            (Sort(_), _) => None,
+            (GeoSort(_), _) => None,
+        }
+    }
+}
+
+impl ScoreDetails {
+    pub fn local_score(&self) -> Option<f64> {
+        self.rank().map(Rank::local_score)
+    }
+
+    pub fn rank(&self) -> Option<Rank> {
+        match self {
+            ScoreDetails::Words(details) => Some(details.rank()),
+            ScoreDetails::Typo(details) => Some(details.rank()),
+            ScoreDetails::Proximity(details) => Some(*details),
+            ScoreDetails::Fid(details) => Some(*details),
+            ScoreDetails::Position(details) => Some(*details),
+            ScoreDetails::ExactAttribute(details) => Some(details.rank()),
+            ScoreDetails::Exactness(details) => Some(*details),
+            ScoreDetails::Sort(_) => None,
+            ScoreDetails::GeoSort(_) => None,
+        }
+    }
+
+    pub fn global_score<'a>(details: impl Iterator<Item = &'a Self>) -> f64 {
+        Rank::global_score(details.filter_map(Self::rank))
+    }
+
+    pub fn global_score_linear_scale<'a>(details: impl Iterator<Item = &'a Self>) -> u64 {
+        (Self::global_score(details) * LINEAR_SCALE_FACTOR).round() as u64
+    }
+
+    /// Panics
+    ///
+    /// - If Position is not preceded by Fid
+    /// - If Exactness is not preceded by ExactAttribute
+    /// - If a sort fid is not contained in the passed `fields_ids_map`.
+    pub fn to_json_map<'a>(
+        details: impl Iterator<Item = &'a Self>,
+    ) -> serde_json::Map<String, serde_json::Value> {
+        let mut order = 0;
+        let mut details_map = serde_json::Map::default();
+        for details in details {
+            match details {
+                ScoreDetails::Words(words) => {
+                    let words_details = serde_json::json!({
+                            "order": order,
+                            "matchingWords": words.matching_words,
+                            "maxMatchingWords": words.max_matching_words,
+                            "score": words.rank().local_score_linear_scale(),
+                    });
+                    details_map.insert("words".into(), words_details);
+                    order += 1;
+                }
+                ScoreDetails::Typo(typo) => {
+                    let typo_details = serde_json::json!({
+                        "order": order,
+                        "typoCount": typo.typo_count,
+                        "maxTypoCount": typo.max_typo_count,
+                        "score": typo.rank().local_score_linear_scale(),
+                    });
+                    details_map.insert("typo".into(), typo_details);
+                    order += 1;
+                }
+                ScoreDetails::Proximity(proximity) => {
+                    let proximity_details = serde_json::json!({
+                        "order": order,
+                        "score": proximity.local_score_linear_scale(),
+                    });
+                    details_map.insert("proximity".into(), proximity_details);
+                    order += 1;
+                }
+                ScoreDetails::Fid(fid) => {
+                    // For now, fid is a virtual rule always followed by the "position" rule
+                    let fid_details = serde_json::json!({
+                        "order": order,
+                        "attributes_ranking_order": fid.local_score_linear_scale(),
+                    });
+                    details_map.insert("attribute".into(), fid_details);
+                    order += 1;
+                }
+                ScoreDetails::Position(position) => {
+                    // For now, position is a virtual rule always preceded by the "fid" rule
+                    let attribute_details = details_map
+                        .get_mut("attribute")
+                        .expect("position not preceded by attribute");
+                    let attribute_details = attribute_details
+                        .as_object_mut()
+                        .expect("attribute details was not an object");
+                    attribute_details.insert(
+                        "attributes_query_word_order".into(),
+                        position.local_score_linear_scale().into(),
+                    );
+                    // do not update the order since this was already done by fid
+                }
+                ScoreDetails::ExactAttribute(exact_attribute) => {
+                    let exactness_details = serde_json::json!({
+                        "order": order,
+                        "exactIn": exact_attribute,
+                        "score": exact_attribute.rank().local_score_linear_scale(),
+                    });
+                    details_map.insert("exactness".into(), exactness_details);
+                    order += 1;
+                }
+                ScoreDetails::Exactness(details) => {
+                    // For now, exactness is a virtual rule always preceded by the "ExactAttribute" rule
+                    let exactness_details = details_map
+                        .get_mut("exactness")
+                        .expect("Exactness not preceded by exactAttribute");
+                    let exactness_details = exactness_details
+                        .as_object_mut()
+                        .expect("exactness details was not an object");
+                    if exactness_details.get("exactIn").expect("missing 'exactIn'")
+                        == &serde_json::json!(ExactAttribute::NoExactMatch)
+                    {
+                        let score = Rank::global_score_linear_scale(
+                            [ExactAttribute::NoExactMatch.rank(), *details].iter().copied(),
+                        );
+                        *exactness_details.get_mut("score").expect("missing score") = score.into();
+                    }
+                    // do not update the order since this was already done by exactAttribute
+                }
+                ScoreDetails::Sort(details) => {
+                    let sort = format!(
+                        "{}:{}",
+                        details.field_name,
+                        if details.ascending { "asc" } else { "desc" }
+                    );
+                    let sort_details = serde_json::json!({
+                        "order": order,
+                        "value": details.value,
+                    });
+                    details_map.insert(sort, sort_details);
+                    order += 1;
+                }
+                ScoreDetails::GeoSort(details) => {
+                    let sort = format!(
+                        "_geoPoint({}, {}):{}",
+                        details.target_point[0],
+                        details.target_point[1],
+                        if details.ascending { "asc" } else { "desc" }
+                    );
+                    let point = if let Some(value) = details.value {
+                        serde_json::json!({ "lat": value[0], "lng": value[1]})
+                    } else {
+                        serde_json::Value::Null
+                    };
+                    let sort_details = serde_json::json!({
+                        "order": order,
+                        "value": point,
+                        "distance": details.distance(),
+                    });
+                    details_map.insert(sort, sort_details);
+                    order += 1;
+                }
+            }
+        }
+        details_map
+    }
+
+    pub fn partial_cmp_iter<'a>(
+        mut left: impl Iterator<Item = &'a Self>,
+        mut right: impl Iterator<Item = &'a Self>,
+    ) -> Result<Ordering, NotComparable> {
+        let mut index = 0;
+        let mut order = match (left.next(), right.next()) {
+            (Some(left), Some(right)) => left.partial_cmp(right).incomparable(index)?,
+            _ => return Ok(Ordering::Equal),
+        };
+        for (left, right) in left.zip(right) {
+            if order != Ordering::Equal {
+                return Ok(order);
+            };
+
+            index += 1;
+            order = left.partial_cmp(right).incomparable(index)?;
+        }
+        Ok(order)
+    }
+}
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Hash)]
+pub struct NotComparable(pub usize);
+
+trait OptionToNotComparable<T> {
+    fn incomparable(self, index: usize) -> Result<T, NotComparable>;
+}
+
+impl<T> OptionToNotComparable<T> for Option<T> {
+    fn incomparable(self, index: usize) -> Result<T, NotComparable> {
+        match self {
+            Some(t) => Ok(t),
+            None => Err(NotComparable(index)),
+        }
+    }
+}
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq, Hash)]
+pub struct Words {
+    pub matching_words: u32,
+    pub max_matching_words: u32,
+}
+
+impl PartialOrd for Words {
+    fn partial_cmp(&self, other: &Self) -> Option<Ordering> {
+        (self.max_matching_words == other.max_matching_words)
+            .then(|| self.matching_words.cmp(&other.matching_words))
+    }
+}
+
+impl Words {
+    pub fn rank(&self) -> Rank {
+        Rank { rank: self.matching_words, max_rank: self.max_matching_words }
+    }
+
+    pub(crate) fn from_rank(rank: Rank) -> Words {
+        Words { matching_words: rank.rank, max_matching_words: rank.max_rank }
+    }
+}
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq, Hash)]
+pub struct Typo {
+    pub typo_count: u32,
+    pub max_typo_count: u32,
+}
+
+impl PartialOrd for Typo {
+    fn partial_cmp(&self, other: &Self) -> Option<Ordering> {
+        (self.max_typo_count == other.max_typo_count).then(|| {
+            // the order is reverted as having fewer typos gives a better score
+            self.typo_count.cmp(&other.typo_count).reverse()
+        })
+    }
+}
+
+impl Typo {
+    pub fn rank(&self) -> Rank {
+        Rank {
+            rank: self.max_typo_count - self.typo_count + 1,
+            max_rank: (self.max_typo_count + 1),
+        }
+    }
+
+    // max_rank = max_typo + 1
+    // max_typo = max_rank - 1
+    //
+    // rank = max_typo - typo + 1
+    // rank = max_rank - 1 - typo + 1
+    // rank + typo = max_rank
+    // typo = max_rank - rank
+    pub fn from_rank(rank: Rank) -> Typo {
+        Typo { typo_count: rank.max_rank - rank.rank, max_typo_count: rank.max_rank - 1 }
+    }
+}
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq, Hash)]
+pub struct Rank {
+    /// The ordinal rank, such that `max_rank` is the first rank, and 0 is the last rank.
+    ///
+    /// The higher the better. Documents with a rank of 0 have a score of 0 and are typically never returned
+    /// (they don't match the query).
+    pub rank: u32,
+    /// The maximum possible rank. Documents with this rank have a score of 1.
+    ///
+    /// The max rank should not be 0.
+    pub max_rank: u32,
+}
+
+impl PartialOrd for Rank {
+    fn partial_cmp(&self, other: &Self) -> Option<Ordering> {
+        (self.max_rank == other.max_rank).then(|| self.rank.cmp(&other.rank))
+    }
+}
+
+impl Rank {
+    pub fn local_score(self) -> f64 {
+        self.rank as f64 / self.max_rank as f64
+    }
+
+    pub fn local_score_linear_scale(self) -> u64 {
+        (self.local_score() * LINEAR_SCALE_FACTOR).round() as u64
+    }
+
+    pub fn global_score(details: impl Iterator<Item = Self>) -> f64 {
+        let mut rank = Rank { rank: 1, max_rank: 1 };
+        for inner_rank in details {
+            rank.rank -= 1;
+
+            rank.rank *= inner_rank.max_rank;
+            rank.max_rank *= inner_rank.max_rank;
+
+            rank.rank += inner_rank.rank;
+        }
+        rank.local_score()
+    }
+
+    pub fn global_score_linear_scale(details: impl Iterator<Item = Self>) -> u64 {
+        (Self::global_score(details) * LINEAR_SCALE_FACTOR).round() as u64
+    }
+}
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Hash, Serialize)]
+#[serde(rename_all = "camelCase")]
+pub enum ExactAttribute {
+    // Do not reorder as the order is significant, from least relevant to most relevant
+    NoExactMatch,
+    MatchesStart,
+    MatchesFull,
+}
+
+impl ExactAttribute {
+    pub fn rank(&self) -> Rank {
+        let rank = match self {
+            ExactAttribute::MatchesFull => 3,
+            ExactAttribute::MatchesStart => 2,
+            ExactAttribute::NoExactMatch => 1,
+        };
+        Rank { rank, max_rank: 3 }
+    }
+}
+
+#[derive(Debug, Clone, PartialEq)]
+pub struct Sort {
+    pub field_name: String,
+    pub ascending: bool,
+    pub value: serde_json::Value,
+}
+
+impl PartialOrd for Sort {
+    fn partial_cmp(&self, other: &Self) -> Option<Ordering> {
+        if self.field_name != other.field_name {
+            return None;
+        }
+        if self.ascending != other.ascending {
+            return None;
+        }
+        match (&self.value, &other.value) {
+            (serde_json::Value::Null, serde_json::Value::Null) => Some(Ordering::Equal),
+            (serde_json::Value::Null, _) => Some(Ordering::Less),
+            (_, serde_json::Value::Null) => Some(Ordering::Greater),
+            // numbers are always before strings
+            (serde_json::Value::Number(_), serde_json::Value::String(_)) => Some(Ordering::Greater),
+            (serde_json::Value::String(_), serde_json::Value::Number(_)) => Some(Ordering::Less),
+            (serde_json::Value::Number(left), serde_json::Value::Number(right)) => {
+                //FIXME: unwrap permitted here?
+                let order = left.as_f64().unwrap().partial_cmp(&right.as_f64().unwrap())?;
+                // always reverted, as bigger is better
+                Some(if self.ascending { order.reverse() } else { order })
+            }
+            (serde_json::Value::String(left), serde_json::Value::String(right)) => {
+                let order = left.cmp(right);
+                Some(if self.ascending { order.reverse() } else { order })
+            }
+            _ => None,
+        }
+    }
+}
+
+#[derive(Debug, Clone, Copy, PartialEq)]
+pub struct GeoSort {
+    pub target_point: [f64; 2],
+    pub ascending: bool,
+    pub value: Option<[f64; 2]>,
+}
+
+impl PartialOrd for GeoSort {
+    fn partial_cmp(&self, other: &Self) -> Option<Ordering> {
+        if self.target_point != other.target_point {
+            return None;
+        }
+        if self.ascending != other.ascending {
+            return None;
+        }
+        Some(match (self.distance(), other.distance()) {
+            (None, None) => Ordering::Equal,
+            (None, Some(_)) => Ordering::Less,
+            (Some(_), None) => Ordering::Greater,
+            (Some(left), Some(right)) => {
+                let order = left.partial_cmp(&right)?;
+                if self.ascending {
+                    // when ascending, the one with the smallest distance has the best score
+                    order.reverse()
+                } else {
+                    order
+                }
+            }
+        })
+    }
+}
+
+impl GeoSort {
+    pub fn distance(&self) -> Option<f64> {
+        self.value.map(|value| distance_between_two_points(&self.target_point, &value))
+    }
+}
+
+const LINEAR_SCALE_FACTOR: f64 = 1000.0;
+
+#[cfg(test)]
+mod test {
+    use super::*;
+    #[test]
+    fn compare() {
+        let left = [
+            ScoreDetails::Words(Words { matching_words: 3, max_matching_words: 4 }),
+            ScoreDetails::Sort(Sort {
+                field_name: "doggo".into(),
+                ascending: true,
+                value: "Intel the Beagle".into(),
+            }),
+        ];
+        let right = [
+            ScoreDetails::Words(Words { matching_words: 3, max_matching_words: 4 }),
+            ScoreDetails::Sort(Sort {
+                field_name: "doggo".into(),
+                ascending: true,
+                value: "Max the Labrador".into(),
+            }),
+        ];
+        assert_eq!(
+            Ok(Ordering::Greater),
+            ScoreDetails::partial_cmp_iter(left.iter(), right.iter())
+        );
+        // equal when all the common components are equal
+        assert_eq!(
+            Ok(Ordering::Equal),
+            ScoreDetails::partial_cmp_iter(left[0..1].iter(), right.iter())
+        );
+
+        let right = [
+            ScoreDetails::Words(Words { matching_words: 4, max_matching_words: 4 }),
+            ScoreDetails::Sort(Sort {
+                field_name: "doggo".into(),
+                ascending: true,
+                value: "Max the Labrador".into(),
+            }),
+        ];
+
+        assert_eq!(Ok(Ordering::Less), ScoreDetails::partial_cmp_iter(left.iter(), right.iter()));
+    }
+
+    #[test]
+    fn sort_not_comparable() {
+        let left = [
+            ScoreDetails::Words(Words { matching_words: 3, max_matching_words: 4 }),
+            ScoreDetails::Sort(Sort {
+                // not the same field name
+                field_name: "catto".into(),
+                ascending: true,
+                value: "Sylver the cat".into(),
+            }),
+        ];
+        let right = [
+            ScoreDetails::Words(Words { matching_words: 3, max_matching_words: 4 }),
+            ScoreDetails::Sort(Sort {
+                field_name: "doggo".into(),
+                ascending: true,
+                value: "Max the Labrador".into(),
+            }),
+        ];
+        assert_eq!(
+            Err(NotComparable(1)),
+            ScoreDetails::partial_cmp_iter(left.iter(), right.iter())
+        );
+        let left = [
+            ScoreDetails::Words(Words { matching_words: 3, max_matching_words: 4 }),
+            ScoreDetails::Sort(Sort {
+                field_name: "doggo".into(),
+                // Not the same order
+                ascending: false,
+                value: "Intel the Beagle".into(),
+            }),
+        ];
+        let right = [
+            ScoreDetails::Words(Words { matching_words: 3, max_matching_words: 4 }),
+            ScoreDetails::Sort(Sort {
+                field_name: "doggo".into(),
+                ascending: true,
+                value: "Max the Labrador".into(),
+            }),
+        ];
+        assert_eq!(
+            Err(NotComparable(1)),
+            ScoreDetails::partial_cmp_iter(left.iter(), right.iter())
+        );
+    }
+
+    #[test]
+    fn sort_behavior() {
+        let left = Sort { field_name: "price".into(), ascending: true, value: "5400".into() };
+        let right = Sort { field_name: "price".into(), ascending: true, value: 53.into() };
+        // number always better match than strings
+        assert_eq!(Some(Ordering::Less), left.partial_cmp(&right));
+
+        let left = Sort { field_name: "price".into(), ascending: false, value: "5400".into() };
+        let right = Sort { field_name: "price".into(), ascending: false, value: 53.into() };
+        // true regardless of the sort direction
+        assert_eq!(Some(Ordering::Less), left.partial_cmp(&right));
+    }
+}

milli/src/search/mod.rs

@@ -7,6 +7,7 @@ use roaring::bitmap::RoaringBitmap;
 pub use self::facet::{FacetDistribution, Filter, DEFAULT_VALUES_PER_FACET};
 pub use self::new::matches::{FormatOptions, MatchBounds, Matcher, MatcherBuilder, MatchingWords};
 use self::new::PartialSearchResult;
+use crate::score_details::ScoreDetails;
 use crate::{
     execute_search, AscDesc, DefaultSearchLogger, DocumentId, Index, Result, SearchContext,
 };
@@ -93,7 +94,7 @@ impl<'a> Search<'a> {
         self
     }
 
-    /// Force the search to exhastivelly compute the number of candidates,
+    /// Forces the search to exhaustively compute the number of candidates,
     /// this will increase the search time but allows finite pagination.
     pub fn exhaustive_number_hits(&mut self, exhaustive_number_hits: bool) -> &mut Search<'a> {
         self.exhaustive_number_hits = exhaustive_number_hits;
@@ -102,7 +103,7 @@ impl<'a> Search<'a> {
 
     pub fn execute(&self) -> Result<SearchResult> {
         let mut ctx = SearchContext::new(self.index, self.rtxn);
-        let PartialSearchResult { located_query_terms, candidates, documents_ids } =
+        let PartialSearchResult { located_query_terms, candidates, documents_ids, document_scores } =
             execute_search(
                 &mut ctx,
                 &self.query,
@@ -124,7 +125,7 @@ impl<'a> Search<'a> {
             None => MatchingWords::default(),
         };
 
-        Ok(SearchResult { matching_words, candidates, documents_ids })
+        Ok(SearchResult { matching_words, candidates, document_scores, documents_ids })
     }
 }
 
@@ -160,8 +161,8 @@ impl fmt::Debug for Search<'_> {
 pub struct SearchResult {
     pub matching_words: MatchingWords,
     pub candidates: RoaringBitmap,
-    // TODO those documents ids should be associated with their criteria scores.
     pub documents_ids: Vec<DocumentId>,
+    pub document_scores: Vec<Vec<ScoreDetails>>,
 }
 
 #[derive(Debug, Clone, Copy, PartialEq, Eq)]

milli/src/search/new/bucket_sort.rs

@@ -3,11 +3,13 @@ use roaring::RoaringBitmap;
 use super::logger::SearchLogger;
 use super::ranking_rules::{BoxRankingRule, RankingRuleQueryTrait};
 use super::SearchContext;
+use crate::score_details::ScoreDetails;
 use crate::search::new::distinct::{apply_distinct_rule, distinct_single_docid, DistinctOutput};
 use crate::Result;
 
 pub struct BucketSortOutput {
     pub docids: Vec<u32>,
+    pub scores: Vec<Vec<ScoreDetails>>,
     pub all_candidates: RoaringBitmap,
 }
 
@@ -31,7 +33,11 @@ pub fn bucket_sort<'ctx, Q: RankingRuleQueryTrait>(
     };
 
     if universe.len() < from as u64 {
-        return Ok(BucketSortOutput { docids: vec![], all_candidates: universe.clone() });
+        return Ok(BucketSortOutput {
+            docids: vec![],
+            scores: vec![],
+            all_candidates: universe.clone(),
+        });
     }
     if ranking_rules.is_empty() {
         if let Some(distinct_fid) = distinct_fid {
@@ -49,22 +55,32 @@ pub fn bucket_sort<'ctx, Q: RankingRuleQueryTrait>(
             }
             let mut all_candidates = universe - excluded;
             all_candidates.extend(results.iter().copied());
-            return Ok(BucketSortOutput { docids: results, all_candidates });
+            return Ok(BucketSortOutput {
+                scores: vec![Default::default(); results.len()],
+                docids: results,
+                all_candidates,
+            });
         } else {
-            let docids = universe.iter().skip(from).take(length).collect();
+            let docids: Vec<u32> = universe.iter().skip(from).take(length).collect();
-            return Ok(BucketSortOutput { docids, all_candidates: universe.clone() });
+            return Ok(BucketSortOutput {
+                scores: vec![Default::default(); docids.len()],
+                docids,
+                all_candidates: universe.clone(),
+            });
         };
     }
 
     let ranking_rules_len = ranking_rules.len();
 
     logger.start_iteration_ranking_rule(0, ranking_rules[0].as_ref(), query, universe);
+
     ranking_rules[0].start_iteration(ctx, logger, universe, query)?;
 
+    let mut ranking_rule_scores: Vec<ScoreDetails> = vec![];
+
     let mut ranking_rule_universes: Vec<RoaringBitmap> =
         vec![RoaringBitmap::default(); ranking_rules_len];
     ranking_rule_universes[0] = universe.clone();
-
     let mut cur_ranking_rule_index = 0;
 
     /// Finish iterating over the current ranking rule, yielding
@@ -89,11 +105,16 @@ pub fn bucket_sort<'ctx, Q: RankingRuleQueryTrait>(
             } else {
                 cur_ranking_rule_index -= 1;
             }
+            // FIXME: check off by one
+            if ranking_rule_scores.len() > cur_ranking_rule_index {
+                ranking_rule_scores.pop();
+            }
         };
     }
 
     let mut all_candidates = universe.clone();
     let mut valid_docids = vec![];
+    let mut valid_scores = vec![];
     let mut cur_offset = 0usize;
 
     macro_rules! maybe_add_to_results {
@@ -104,23 +125,23 @@ pub fn bucket_sort<'ctx, Q: RankingRuleQueryTrait>(
                 length,
                 logger,
                 &mut valid_docids,
+                &mut valid_scores,
                 &mut all_candidates,
                 &mut ranking_rule_universes,
                 &mut ranking_rules,
                 cur_ranking_rule_index,
                 &mut cur_offset,
                 distinct_fid,
+                &ranking_rule_scores,
                 $candidates,
             )?;
         };
     }
 
     while valid_docids.len() < length {
-        // The universe for this bucket is zero or one element, so we don't need to sort
+        // The universe for this bucket is zero, so we don't need to sort
-        // anything, just extend the results and go back to the parent ranking rule.
+        // anything, just go back to the parent ranking rule.
-        if ranking_rule_universes[cur_ranking_rule_index].len() <= 1 {
+        if ranking_rule_universes[cur_ranking_rule_index].is_empty() {
-            let bucket = std::mem::take(&mut ranking_rule_universes[cur_ranking_rule_index]);
-            maybe_add_to_results!(bucket);
             back!();
             continue;
         }
@@ -130,6 +151,8 @@ pub fn bucket_sort<'ctx, Q: RankingRuleQueryTrait>(
             continue;
         };
 
+        ranking_rule_scores.push(next_bucket.score);
+
         logger.next_bucket_ranking_rule(
             cur_ranking_rule_index,
             ranking_rules[cur_ranking_rule_index].as_ref(),
@@ -143,10 +166,11 @@ pub fn bucket_sort<'ctx, Q: RankingRuleQueryTrait>(
         ranking_rule_universes[cur_ranking_rule_index] -= &next_bucket.candidates;
 
         if cur_ranking_rule_index == ranking_rules_len - 1
-            || next_bucket.candidates.len() <= 1
             || cur_offset + (next_bucket.candidates.len() as usize) < from
         {
             maybe_add_to_results!(next_bucket.candidates);
+            // FIXME: use index based logic like all the other rules so that you don't have to maintain the pop/push?
+            ranking_rule_scores.pop();
             continue;
         }
 
@@ -166,7 +190,7 @@ pub fn bucket_sort<'ctx, Q: RankingRuleQueryTrait>(
         )?;
     }
 
-    Ok(BucketSortOutput { docids: valid_docids, all_candidates })
+    Ok(BucketSortOutput { docids: valid_docids, scores: valid_scores, all_candidates })
 }
 
 /// Add the candidates to the results. Take `distinct`, `from`, `length`, and `cur_offset`
@@ -179,14 +203,18 @@ fn maybe_add_to_results<'ctx, Q: RankingRuleQueryTrait>(
     logger: &mut dyn SearchLogger<Q>,
 
     valid_docids: &mut Vec<u32>,
+    valid_scores: &mut Vec<Vec<ScoreDetails>>,
     all_candidates: &mut RoaringBitmap,
 
     ranking_rule_universes: &mut [RoaringBitmap],
     ranking_rules: &mut [BoxRankingRule<'ctx, Q>],
+
     cur_ranking_rule_index: usize,
 
     cur_offset: &mut usize,
+
     distinct_fid: Option<u16>,
+    ranking_rule_scores: &[ScoreDetails],
     candidates: RoaringBitmap,
 ) -> Result<()> {
     // First apply the distinct rule on the candidates, reducing the universes if necessary
@@ -231,13 +259,17 @@ fn maybe_add_to_results<'ctx, Q: RankingRuleQueryTrait>(
             let candidates =
                 candidates.iter().take(length - valid_docids.len()).copied().collect::<Vec<_>>();
             logger.add_to_results(&candidates);
-            valid_docids.extend(&candidates);
+            valid_docids.extend_from_slice(&candidates);
+            valid_scores
+                .extend(std::iter::repeat(ranking_rule_scores.to_owned()).take(candidates.len()));
         }
     } else {
         // if we have passed the offset already, add some of the documents (up to the limit)
         let candidates = candidates.iter().take(length - valid_docids.len()).collect::<Vec<u32>>();
         logger.add_to_results(&candidates);
-        valid_docids.extend(&candidates);
+        valid_docids.extend_from_slice(&candidates);
+        valid_scores
+            .extend(std::iter::repeat(ranking_rule_scores.to_owned()).take(candidates.len()));
     }
 
     *cur_offset += candidates.len() as usize;

milli/src/search/new/exact_attribute.rs

@@ -2,6 +2,7 @@ use roaring::{MultiOps, RoaringBitmap};
 
 use super::query_graph::QueryGraph;
 use super::ranking_rules::{RankingRule, RankingRuleOutput};
+use crate::score_details::{self, ScoreDetails};
 use crate::search::new::query_graph::QueryNodeData;
 use crate::search::new::query_term::ExactTerm;
 use crate::{Result, SearchContext, SearchLogger};
@@ -244,7 +245,13 @@ impl State {
                 candidates &= universe;
                 (
                     State::AttributeStarts(query_graph.clone(), candidates_per_attribute),
-                    Some(RankingRuleOutput { query: query_graph, candidates }),
+                    Some(RankingRuleOutput {
+                        query: query_graph,
+                        candidates,
+                        score: ScoreDetails::ExactAttribute(
+                            score_details::ExactAttribute::MatchesFull,
+                        ),
+                    }),
                 )
             }
             State::AttributeStarts(query_graph, candidates_per_attribute) => {
@@ -257,12 +264,24 @@ impl State {
                 candidates &= universe;
                 (
                     State::Empty(query_graph.clone()),
-                    Some(RankingRuleOutput { query: query_graph, candidates }),
+                    Some(RankingRuleOutput {
+                        query: query_graph,
+                        candidates,
+                        score: ScoreDetails::ExactAttribute(
+                            score_details::ExactAttribute::MatchesStart,
+                        ),
+                    }),
                 )
             }
             State::Empty(query_graph) => (
                 State::Empty(query_graph.clone()),
-                Some(RankingRuleOutput { query: query_graph, candidates: universe.clone() }),
+                Some(RankingRuleOutput {
+                    query: query_graph,
+                    candidates: universe.clone(),
+                    score: ScoreDetails::ExactAttribute(
+                        score_details::ExactAttribute::NoExactMatch,
+                    ),
+                }),
             ),
         };
         (state, output)

milli/src/search/new/geo_sort.rs

@@ -8,6 +8,7 @@ use rstar::RTree;
 
 use super::ranking_rules::{RankingRule, RankingRuleOutput, RankingRuleQueryTrait};
 use crate::heed_codec::facet::{FieldDocIdFacetCodec, OrderedF64Codec};
+use crate::score_details::{self, ScoreDetails};
 use crate::{
     distance_between_two_points, lat_lng_to_xyz, GeoPoint, Index, Result, SearchContext,
     SearchLogger,
@@ -80,7 +81,7 @@ pub struct GeoSort<Q: RankingRuleQueryTrait> {
     field_ids: Option<[u16; 2]>,
     rtree: Option<RTree<GeoPoint>>,
 
-    cached_sorted_docids: VecDeque<u32>,
+    cached_sorted_docids: VecDeque<(u32, [f64; 2])>,
     geo_candidates: RoaringBitmap,
 }
 
@@ -130,7 +131,7 @@ impl<Q: RankingRuleQueryTrait> GeoSort<Q> {
                 let point = lat_lng_to_xyz(&self.point);
                 for point in rtree.nearest_neighbor_iter(&point) {
                     if self.geo_candidates.contains(point.data.0) {
-                        self.cached_sorted_docids.push_back(point.data.0);
+                        self.cached_sorted_docids.push_back(point.data);
                         if self.cached_sorted_docids.len() >= cache_size {
                             break;
                         }
@@ -142,7 +143,7 @@ impl<Q: RankingRuleQueryTrait> GeoSort<Q> {
                 let point = lat_lng_to_xyz(&opposite_of(self.point));
                 for point in rtree.nearest_neighbor_iter(&point) {
                     if self.geo_candidates.contains(point.data.0) {
-                        self.cached_sorted_docids.push_front(point.data.0);
+                        self.cached_sorted_docids.push_front(point.data);
                         if self.cached_sorted_docids.len() >= cache_size {
                             break;
                         }
@@ -177,7 +178,7 @@ impl<Q: RankingRuleQueryTrait> GeoSort<Q> {
             // computing the distance between two points is expensive thus we cache the result
             documents
                 .sort_by_cached_key(|(_, p)| distance_between_two_points(&self.point, p) as usize);
-            self.cached_sorted_docids.extend(documents.into_iter().map(|(doc_id, _)| doc_id));
+            self.cached_sorted_docids.extend(documents.into_iter());
         };
 
         Ok(())
@@ -220,12 +221,19 @@ impl<'ctx, Q: RankingRuleQueryTrait> RankingRule<'ctx, Q> for GeoSort<Q> {
         logger: &mut dyn SearchLogger<Q>,
         universe: &RoaringBitmap,
     ) -> Result<Option<RankingRuleOutput<Q>>> {
-        assert!(universe.len() > 1);
         let query = self.query.as_ref().unwrap().clone();
         self.geo_candidates &= universe;
 
         if self.geo_candidates.is_empty() {
-            return Ok(Some(RankingRuleOutput { query, candidates: universe.clone() }));
+            return Ok(Some(RankingRuleOutput {
+                query,
+                candidates: universe.clone(),
+                score: ScoreDetails::GeoSort(score_details::GeoSort {
+                    target_point: self.point,
+                    ascending: self.ascending,
+                    value: None,
+                }),
+            }));
         }
 
         let ascending = self.ascending;
@@ -236,11 +244,16 @@ impl<'ctx, Q: RankingRuleQueryTrait> RankingRule<'ctx, Q> for GeoSort<Q> {
                 cache.pop_back()
             }
         };
-        while let Some(id) = next(&mut self.cached_sorted_docids) {
+        while let Some((id, point)) = next(&mut self.cached_sorted_docids) {
             if self.geo_candidates.contains(id) {
                 return Ok(Some(RankingRuleOutput {
                     query,
                     candidates: RoaringBitmap::from_iter([id]),
+                    score: ScoreDetails::GeoSort(score_details::GeoSort {
+                        target_point: self.point,
+                        ascending: self.ascending,
+                        value: Some(point),
+                    }),
                 }));
             }
         }

milli/src/search/new/graph_based_ranking_rule.rs

@@ -50,6 +50,7 @@ use super::ranking_rule_graph::{
 };
 use super::small_bitmap::SmallBitmap;
 use super::{QueryGraph, RankingRule, RankingRuleOutput, SearchContext};
+use crate::score_details::Rank;
 use crate::search::new::query_term::LocatedQueryTermSubset;
 use crate::search::new::ranking_rule_graph::PathVisitor;
 use crate::{Result, TermsMatchingStrategy};
@@ -118,6 +119,8 @@ pub struct GraphBasedRankingRuleState<G: RankingRuleGraphTrait> {
     all_costs: MappedInterner<QueryNode, Vec<u64>>,
     /// An index in the first element of `all_distances`, giving the cost of the next bucket
     cur_cost: u64,
+    /// One above the highest possible cost for this rule
+    next_max_cost: u64,
 }
 
 impl<'ctx, G: RankingRuleGraphTrait> RankingRule<'ctx, QueryGraph> for GraphBasedRankingRule<G> {
@@ -139,13 +142,12 @@ impl<'ctx, G: RankingRuleGraphTrait> RankingRule<'ctx, QueryGraph> for GraphBase
                     let mut forbidden_nodes =
                         SmallBitmap::for_interned_values_in(&query_graph.nodes);
                     let mut costs = query_graph.nodes.map(|_| None);
-                    let mut cost = 100;
+                    // FIXME: this works because only words uses termsmatchingstrategy at the moment.
                     for ns in removal_order {
                         for n in ns.iter() {
-                            *costs.get_mut(n) = Some((cost, forbidden_nodes.clone()));
+                            *costs.get_mut(n) = Some((1, forbidden_nodes.clone()));
                         }
                         forbidden_nodes.union(&ns);
-                        cost += 100;
                     }
                     costs
                 }
@@ -162,12 +164,16 @@ impl<'ctx, G: RankingRuleGraphTrait> RankingRule<'ctx, QueryGraph> for GraphBase
         // Then pre-compute the cost of all paths from each node to the end node
         let all_costs = graph.find_all_costs_to_end();
 
+        let next_max_cost =
+            all_costs.get(graph.query_graph.root_node).iter().copied().max().unwrap_or(0) + 1;
+
         let state = GraphBasedRankingRuleState {
             graph,
             conditions_cache: condition_docids_cache,
             dead_ends_cache,
             all_costs,
             cur_cost: 0,
+            next_max_cost,
         };
 
         self.state = Some(state);
@@ -181,17 +187,13 @@ impl<'ctx, G: RankingRuleGraphTrait> RankingRule<'ctx, QueryGraph> for GraphBase
         logger: &mut dyn SearchLogger<QueryGraph>,
         universe: &RoaringBitmap,
     ) -> Result<Option<RankingRuleOutput<QueryGraph>>> {
-        // If universe.len() <= 1, the bucket sort algorithm
-        // should not have called this function.
-        assert!(universe.len() > 1);
         // Will crash if `next_bucket` is called before `start_iteration` or after `end_iteration`,
         // should never happen
         let mut state = self.state.take().unwrap();
 
+        let all_costs = state.all_costs.get(state.graph.query_graph.root_node);
         // Retrieve the cost of the paths to compute
-        let Some(&cost) = state
+        let Some(&cost) = all_costs
-            .all_costs
-            .get(state.graph.query_graph.root_node)
             .iter()
             .find(|c| **c >= state.cur_cost) else {
                 self.state = None;
@@ -207,8 +209,12 @@ impl<'ctx, G: RankingRuleGraphTrait> RankingRule<'ctx, QueryGraph> for GraphBase
             dead_ends_cache,
             all_costs,
             cur_cost: _,
+            next_max_cost,
         } = &mut state;
 
+        let rank = *next_max_cost - cost;
+        let score = G::rank_to_score(Rank { rank: rank as u32, max_rank: *next_max_cost as u32 });
+
         let mut universe = universe.clone();
 
         let mut used_conditions = SmallBitmap::for_interned_values_in(&graph.conditions_interner);
@@ -325,7 +331,7 @@ impl<'ctx, G: RankingRuleGraphTrait> RankingRule<'ctx, QueryGraph> for GraphBase
 
         self.state = Some(state);
 
-        Ok(Some(RankingRuleOutput { query: next_query_graph, candidates: bucket }))
+        Ok(Some(RankingRuleOutput { query: next_query_graph, candidates: bucket, score }))
     }
 
     fn end_iteration(

milli/src/search/new/mod.rs

@@ -44,6 +44,7 @@ use self::geo_sort::GeoSort;
 pub use self::geo_sort::Strategy as GeoSortStrategy;
 use self::graph_based_ranking_rule::Words;
 use self::interner::Interned;
+use crate::score_details::ScoreDetails;
 use crate::search::new::distinct::apply_distinct_rule;
 use crate::{AscDesc, DocumentId, Filter, Index, Member, Result, TermsMatchingStrategy, UserError};
 
@@ -426,13 +427,15 @@ pub fn execute_search(
         )?
     };
 
-    let BucketSortOutput { docids, mut all_candidates } = bucket_sort_output;
+    let BucketSortOutput { docids, scores, mut all_candidates } = bucket_sort_output;
+
+    let fields_ids_map = ctx.index.fields_ids_map(ctx.txn)?;
 
     // The candidates is the universe unless the exhaustive number of hits
     // is requested and a distinct attribute is set.
     if exhaustive_number_hits {
         if let Some(f) = ctx.index.distinct_field(ctx.txn)? {
-            if let Some(distinct_fid) = ctx.index.fields_ids_map(ctx.txn)?.id(f) {
+            if let Some(distinct_fid) = fields_ids_map.id(f) {
                 all_candidates = apply_distinct_rule(ctx, distinct_fid, &all_candidates)?.remaining;
             }
         }
@@ -440,6 +443,7 @@ pub fn execute_search(
 
     Ok(PartialSearchResult {
         candidates: all_candidates,
+        document_scores: scores,
         documents_ids: docids,
         located_query_terms,
     })
@@ -491,4 +495,5 @@ pub struct PartialSearchResult {
     pub located_query_terms: Option<Vec<LocatedQueryTerm>>,
     pub candidates: RoaringBitmap,
     pub documents_ids: Vec<DocumentId>,
+    pub document_scores: Vec<Vec<ScoreDetails>>,
 }

milli/src/search/new/ranking_rule_graph/build.rs

@@ -49,10 +49,15 @@ impl<G: RankingRuleGraphTrait> RankingRuleGraph<G> {
                 if let Some((cost_of_ignoring, forbidden_nodes)) =
                     cost_of_ignoring_node.get(dest_idx)
                 {
+                    let dest = graph_nodes.get(dest_idx);
+                    let dest_size = match &dest.data {
+                        QueryNodeData::Term(term) => term.term_ids.len(),
+                        _ => panic!(),
+                    };
                     let new_edge_id = edges_store.insert(Some(Edge {
                         source_node: source_id,
                         dest_node: dest_idx,
-                        cost: *cost_of_ignoring,
+                        cost: *cost_of_ignoring * dest_size as u32,
                         condition: None,
                         nodes_to_skip: forbidden_nodes.clone(),
                     }));

milli/src/search/new/ranking_rule_graph/exactness/mod.rs

@@ -1,6 +1,7 @@
 use roaring::RoaringBitmap;
 
 use super::{ComputedCondition, RankingRuleGraphTrait};
+use crate::score_details::{Rank, ScoreDetails};
 use crate::search::new::interner::{DedupInterner, Interned};
 use crate::search::new::query_term::{ExactTerm, LocatedQueryTermSubset};
 use crate::search::new::resolve_query_graph::compute_query_term_subset_docids;
@@ -84,4 +85,8 @@ impl RankingRuleGraphTrait for ExactnessGraph {
 
         Ok(vec![(0, exact_condition), (dest_node.term_ids.len() as u32, skip_condition)])
     }
+
+    fn rank_to_score(rank: Rank) -> ScoreDetails {
+        ScoreDetails::Exactness(rank)
+    }
 }

milli/src/search/new/ranking_rule_graph/fid/mod.rs

@@ -2,6 +2,7 @@ use fxhash::FxHashSet;
 use roaring::RoaringBitmap;
 
 use super::{ComputedCondition, RankingRuleGraphTrait};
+use crate::score_details::{Rank, ScoreDetails};
 use crate::search::new::interner::{DedupInterner, Interned};
 use crate::search::new::query_term::LocatedQueryTermSubset;
 use crate::search::new::resolve_query_graph::compute_query_term_subset_docids_within_field_id;
@@ -68,7 +69,7 @@ impl RankingRuleGraphTrait for FidGraph {
         }
 
         let mut edges = vec![];
-        for fid in all_fields {
+        for fid in all_fields.iter().copied() {
             // TODO: We can improve performances and relevancy by storing
             //       the term subsets associated to each field ids fetched.
             edges.push((
@@ -80,6 +81,35 @@ impl RankingRuleGraphTrait for FidGraph {
             ));
         }
 
+        // always lookup the max_fid if we don't already and add an artificial condition for max scoring
+        let max_fid: Option<u16> = {
+            if let Some(max_fid) = ctx
+                .index
+                .searchable_fields_ids(ctx.txn)?
+                .map(|field_ids| field_ids.into_iter().max())
+            {
+                max_fid
+            } else {
+                ctx.index.fields_ids_map(ctx.txn)?.ids().max()
+            }
+        };
+
+        if let Some(max_fid) = max_fid {
+            if !all_fields.contains(&max_fid) {
+                edges.push((
+                    max_fid as u32 * term.term_ids.len() as u32, // TODO improve the fid score i.e. fid^10.
+                    conditions_interner.insert(FidCondition {
+                        term: term.clone(), // TODO remove this ugly clone
+                        fid: max_fid,
+                    }),
+                ));
+            }
+        }
+
         Ok(edges)
     }
+
+    fn rank_to_score(rank: Rank) -> ScoreDetails {
+        ScoreDetails::Fid(rank)
+    }
 }

milli/src/search/new/ranking_rule_graph/mod.rs

@@ -41,6 +41,7 @@ use super::interner::{DedupInterner, FixedSizeInterner, Interned, MappedInterner
 use super::query_term::LocatedQueryTermSubset;
 use super::small_bitmap::SmallBitmap;
 use super::{QueryGraph, QueryNode, SearchContext};
+use crate::score_details::{Rank, ScoreDetails};
 use crate::Result;
 
 pub struct ComputedCondition {
@@ -110,6 +111,9 @@ pub trait RankingRuleGraphTrait: Sized + 'static {
         source_node: Option<&LocatedQueryTermSubset>,
         dest_node: &LocatedQueryTermSubset,
     ) -> Result<Vec<(u32, Interned<Self::Condition>)>>;
+
+    /// Convert the rank of a path to its corresponding score for the ranking rule
+    fn rank_to_score(rank: Rank) -> ScoreDetails;
 }
 
 /// The graph used by graph-based ranking rules.

milli/src/search/new/ranking_rule_graph/position/mod.rs

@@ -2,6 +2,7 @@ use fxhash::{FxHashMap, FxHashSet};
 use roaring::RoaringBitmap;
 
 use super::{ComputedCondition, RankingRuleGraphTrait};
+use crate::score_details::{Rank, ScoreDetails};
 use crate::search::new::interner::{DedupInterner, Interned};
 use crate::search::new::query_term::LocatedQueryTermSubset;
 use crate::search::new::resolve_query_graph::compute_query_term_subset_docids_within_position;
@@ -105,8 +106,20 @@ impl RankingRuleGraphTrait for PositionGraph {
             ));
         }
 
+        // artificial empty condition for computing max cost
+        let max_cost = term.term_ids.len() as u32 * 10;
+        edges.push((
+            max_cost,
+            conditions_interner
+                .insert(PositionCondition { term: term.clone(), positions: Vec::default() }),
+        ));
+
         Ok(edges)
     }
+
+    fn rank_to_score(rank: Rank) -> ScoreDetails {
+        ScoreDetails::Position(rank)
+    }
 }
 
 fn cost_from_position(sum_positions: u32) -> u32 {

milli/src/search/new/ranking_rule_graph/proximity/mod.rs

@@ -4,6 +4,7 @@ pub mod compute_docids;
 use roaring::RoaringBitmap;
 
 use super::{ComputedCondition, RankingRuleGraphTrait};
+use crate::score_details::{Rank, ScoreDetails};
 use crate::search::new::interner::{DedupInterner, Interned};
 use crate::search::new::query_term::LocatedQueryTermSubset;
 use crate::search::new::SearchContext;
@@ -36,4 +37,8 @@ impl RankingRuleGraphTrait for ProximityGraph {
     ) -> Result<Vec<(u32, Interned<Self::Condition>)>> {
         build::build_edges(ctx, conditions_interner, source_term, dest_term)
     }
+
+    fn rank_to_score(rank: Rank) -> ScoreDetails {
+        ScoreDetails::Proximity(rank)
+    }
 }

milli/src/search/new/ranking_rule_graph/typo/mod.rs

@@ -1,6 +1,7 @@
 use roaring::RoaringBitmap;
 
 use super::{ComputedCondition, RankingRuleGraphTrait};
+use crate::score_details::{self, Rank, ScoreDetails};
 use crate::search::new::interner::{DedupInterner, Interned};
 use crate::search::new::query_term::LocatedQueryTermSubset;
 use crate::search::new::resolve_query_graph::compute_query_term_subset_docids;
@@ -75,4 +76,8 @@ impl RankingRuleGraphTrait for TypoGraph {
         }
         Ok(edges)
     }
+
+    fn rank_to_score(rank: Rank) -> ScoreDetails {
+        ScoreDetails::Typo(score_details::Typo::from_rank(rank))
+    }
 }

milli/src/search/new/ranking_rule_graph/words/mod.rs

@@ -1,6 +1,7 @@
 use roaring::RoaringBitmap;
 
 use super::{ComputedCondition, RankingRuleGraphTrait};
+use crate::score_details::{self, Rank, ScoreDetails};
 use crate::search::new::interner::{DedupInterner, Interned};
 use crate::search::new::query_term::LocatedQueryTermSubset;
 use crate::search::new::resolve_query_graph::compute_query_term_subset_docids;
@@ -41,9 +42,10 @@ impl RankingRuleGraphTrait for WordsGraph {
         _from: Option<&LocatedQueryTermSubset>,
         to_term: &LocatedQueryTermSubset,
     ) -> Result<Vec<(u32, Interned<Self::Condition>)>> {
-        Ok(vec![(
+        Ok(vec![(0, conditions_interner.insert(WordsCondition { term: to_term.clone() }))])
-            to_term.term_ids.len() as u32,
+    }
-            conditions_interner.insert(WordsCondition { term: to_term.clone() }),
+
-        )])
+    fn rank_to_score(rank: Rank) -> ScoreDetails {
+        ScoreDetails::Words(score_details::Words::from_rank(rank))
     }
 }

milli/src/search/new/ranking_rules.rs

@@ -2,6 +2,7 @@ use roaring::RoaringBitmap;
 
 use super::logger::SearchLogger;
 use super::{QueryGraph, SearchContext};
+use crate::score_details::ScoreDetails;
 use crate::Result;
 
 /// An internal trait implemented by only [`PlaceholderQuery`] and [`QueryGraph`]
@@ -66,4 +67,6 @@ pub struct RankingRuleOutput<Q> {
     pub query: Q,
     /// The allowed candidates for the child ranking rule
     pub candidates: RoaringBitmap,
+    /// The score for the candidates of the current bucket
+    pub score: ScoreDetails,
 }

milli/src/search/new/sort.rs

@@ -1,9 +1,11 @@
+use heed::BytesDecode;
 use roaring::RoaringBitmap;
 
 use super::logger::SearchLogger;
 use super::{RankingRule, RankingRuleOutput, RankingRuleQueryTrait, SearchContext};
-use crate::heed_codec::facet::FacetGroupKeyCodec;
+use crate::heed_codec::facet::{FacetGroupKeyCodec, OrderedF64Codec};
-use crate::heed_codec::ByteSliceRefCodec;
+use crate::heed_codec::{ByteSliceRefCodec, StrRefCodec};
+use crate::score_details::{self, ScoreDetails};
 use crate::search::facet::{ascending_facet_sort, descending_facet_sort};
 use crate::{FieldId, Index, Result};
 
@@ -67,7 +69,7 @@ impl<'ctx, Query> Sort<'ctx, Query> {
 impl<'ctx, Query: RankingRuleQueryTrait> RankingRule<'ctx, Query> for Sort<'ctx, Query> {
     fn id(&self) -> String {
         let Self { field_name, is_ascending, .. } = self;
-        format!("{field_name}:{}", if *is_ascending { "asc" } else { "desc " })
+        format!("{field_name}:{}", if *is_ascending { "asc" } else { "desc" })
     }
     fn start_iteration(
         &mut self,
@@ -118,12 +120,43 @@ impl<'ctx, Query: RankingRuleQueryTrait> RankingRule<'ctx, Query> for Sort<'ctx,
 
                     (itertools::Either::Right(number_iter), itertools::Either::Right(string_iter))
                 };
+                let number_iter = number_iter.map(|r| -> Result<_> {
+                    let (docids, bytes) = r?;
+                    Ok((
+                        docids,
+                        serde_json::Value::Number(
+                            serde_json::Number::from_f64(
+                                OrderedF64Codec::bytes_decode(bytes).expect("some number"),
+                            )
+                            .expect("too big float"),
+                        ),
+                    ))
+                });
+                let string_iter = string_iter.map(|r| -> Result<_> {
+                    let (docids, bytes) = r?;
+                    Ok((
+                        docids,
+                        serde_json::Value::String(
+                            StrRefCodec::bytes_decode(bytes).expect("some string").to_owned(),
+                        ),
+                    ))
+                });
 
                 let query_graph = parent_query.clone();
+                let ascending = self.is_ascending;
+                let field_name = self.field_name.clone();
                 RankingRuleOutputIterWrapper::new(Box::new(number_iter.chain(string_iter).map(
                     move |r| {
-                        let (docids, _) = r?;
+                        let (docids, value) = r?;
-                        Ok(RankingRuleOutput { query: query_graph.clone(), candidates: docids })
+                        Ok(RankingRuleOutput {
+                            query: query_graph.clone(),
+                            candidates: docids,
+                            score: ScoreDetails::Sort(score_details::Sort {
+                                field_name: field_name.clone(),
+                                ascending,
+                                value,
+                            }),
+                        })
                     },
                 )))
             }
@@ -150,7 +183,15 @@ impl<'ctx, Query: RankingRuleQueryTrait> RankingRule<'ctx, Query> for Sort<'ctx,
             Ok(Some(bucket))
         } else {
             let query = self.original_query.as_ref().unwrap().clone();
-            Ok(Some(RankingRuleOutput { query, candidates: universe.clone() }))
+            Ok(Some(RankingRuleOutput {
+                query,
+                candidates: universe.clone(),
+                score: ScoreDetails::Sort(score_details::Sort {
+                    field_name: self.field_name.clone(),
+                    ascending: self.is_ascending,
+                    value: serde_json::Value::Null,
+                }),
+            }))
         }
     }