Move crates under a sub folder to clean up the code

crates/meilisearch/src/search/federated.rs

@@ -0,0 +1,910 @@
+use std::cmp::Ordering;
+use std::collections::BTreeMap;
+use std::fmt;
+use std::iter::Zip;
+use std::rc::Rc;
+use std::str::FromStr as _;
+use std::time::Duration;
+use std::vec::{IntoIter, Vec};
+
+use actix_http::StatusCode;
+use index_scheduler::{IndexScheduler, RoFeatures};
+use indexmap::IndexMap;
+use meilisearch_types::deserr::DeserrJsonError;
+use meilisearch_types::error::deserr_codes::{
+    InvalidMultiSearchFacetsByIndex, InvalidMultiSearchMaxValuesPerFacet,
+    InvalidMultiSearchMergeFacets, InvalidMultiSearchWeight, InvalidSearchLimit,
+    InvalidSearchOffset,
+};
+use meilisearch_types::error::ResponseError;
+use meilisearch_types::index_uid::IndexUid;
+use meilisearch_types::milli::score_details::{ScoreDetails, ScoreValue};
+use meilisearch_types::milli::{self, DocumentId, OrderBy, TimeBudget};
+use roaring::RoaringBitmap;
+use serde::Serialize;
+
+use super::ranking_rules::{self, RankingRules};
+use super::{
+    compute_facet_distribution_stats, prepare_search, AttributesFormat, ComputedFacets, FacetStats,
+    HitMaker, HitsInfo, RetrieveVectors, SearchHit, SearchKind, SearchQuery, SearchQueryWithIndex,
+};
+use crate::error::MeilisearchHttpError;
+use crate::routes::indexes::search::search_kind;
+
+pub const DEFAULT_FEDERATED_WEIGHT: f64 = 1.0;
+
+#[derive(Debug, Default, Clone, Copy, PartialEq, deserr::Deserr)]
+#[deserr(error = DeserrJsonError, rename_all = camelCase, deny_unknown_fields)]
+pub struct FederationOptions {
+    #[deserr(default, error = DeserrJsonError<InvalidMultiSearchWeight>)]
+    pub weight: Weight,
+}
+
+#[derive(Debug, Clone, Copy, PartialEq, deserr::Deserr)]
+#[deserr(try_from(f64) = TryFrom::try_from -> InvalidMultiSearchWeight)]
+pub struct Weight(f64);
+
+impl Default for Weight {
+    fn default() -> Self {
+        Weight(DEFAULT_FEDERATED_WEIGHT)
+    }
+}
+
+impl std::convert::TryFrom<f64> for Weight {
+    type Error = InvalidMultiSearchWeight;
+
+    fn try_from(f: f64) -> Result<Self, Self::Error> {
+        if f < 0.0 {
+            Err(InvalidMultiSearchWeight)
+        } else {
+            Ok(Weight(f))
+        }
+    }
+}
+
+impl std::ops::Deref for Weight {
+    type Target = f64;
+
+    fn deref(&self) -> &Self::Target {
+        &self.0
+    }
+}
+
+#[derive(Debug, deserr::Deserr)]
+#[deserr(error = DeserrJsonError, rename_all = camelCase, deny_unknown_fields)]
+pub struct Federation {
+    #[deserr(default = super::DEFAULT_SEARCH_LIMIT(), error = DeserrJsonError<InvalidSearchLimit>)]
+    pub limit: usize,
+    #[deserr(default = super::DEFAULT_SEARCH_OFFSET(), error = DeserrJsonError<InvalidSearchOffset>)]
+    pub offset: usize,
+    #[deserr(default, error = DeserrJsonError<InvalidMultiSearchFacetsByIndex>)]
+    pub facets_by_index: BTreeMap<IndexUid, Option<Vec<String>>>,
+    #[deserr(default, error = DeserrJsonError<InvalidMultiSearchMergeFacets>)]
+    pub merge_facets: Option<MergeFacets>,
+}
+
+#[derive(Copy, Clone, Debug, deserr::Deserr, Default)]
+#[deserr(error = DeserrJsonError<InvalidMultiSearchMergeFacets>, rename_all = camelCase, deny_unknown_fields)]
+pub struct MergeFacets {
+    #[deserr(default, error = DeserrJsonError<InvalidMultiSearchMaxValuesPerFacet>)]
+    pub max_values_per_facet: Option<usize>,
+}
+
+#[derive(Debug, deserr::Deserr)]
+#[deserr(error = DeserrJsonError, rename_all = camelCase, deny_unknown_fields)]
+pub struct FederatedSearch {
+    pub queries: Vec<SearchQueryWithIndex>,
+    #[deserr(default)]
+    pub federation: Option<Federation>,
+}
+#[derive(Serialize, Clone)]
+#[serde(rename_all = "camelCase")]
+pub struct FederatedSearchResult {
+    pub hits: Vec<SearchHit>,
+    pub processing_time_ms: u128,
+    #[serde(flatten)]
+    pub hits_info: HitsInfo,
+
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub semantic_hit_count: Option<u32>,
+
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub facet_distribution: Option<BTreeMap<String, IndexMap<String, u64>>>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub facet_stats: Option<BTreeMap<String, FacetStats>>,
+    #[serde(skip_serializing_if = "FederatedFacets::is_empty")]
+    pub facets_by_index: FederatedFacets,
+
+    // These fields are only used for analytics purposes
+    #[serde(skip)]
+    pub degraded: bool,
+    #[serde(skip)]
+    pub used_negative_operator: bool,
+}
+
+impl fmt::Debug for FederatedSearchResult {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        let FederatedSearchResult {
+            hits,
+            processing_time_ms,
+            hits_info,
+            semantic_hit_count,
+            degraded,
+            used_negative_operator,
+            facet_distribution,
+            facet_stats,
+            facets_by_index,
+        } = self;
+
+        let mut debug = f.debug_struct("SearchResult");
+        // The most important thing when looking at a search result is the time it took to process
+        debug.field("processing_time_ms", &processing_time_ms);
+        debug.field("hits", &format!("[{} hits returned]", hits.len()));
+        debug.field("hits_info", &hits_info);
+        if *used_negative_operator {
+            debug.field("used_negative_operator", used_negative_operator);
+        }
+        if *degraded {
+            debug.field("degraded", degraded);
+        }
+        if let Some(facet_distribution) = facet_distribution {
+            debug.field("facet_distribution", &facet_distribution);
+        }
+        if let Some(facet_stats) = facet_stats {
+            debug.field("facet_stats", &facet_stats);
+        }
+        if let Some(semantic_hit_count) = semantic_hit_count {
+            debug.field("semantic_hit_count", &semantic_hit_count);
+        }
+        if !facets_by_index.is_empty() {
+            debug.field("facets_by_index", &facets_by_index);
+        }
+
+        debug.finish()
+    }
+}
+
+struct WeightedScore<'a> {
+    details: &'a [ScoreDetails],
+    weight: f64,
+}
+
+impl<'a> WeightedScore<'a> {
+    pub fn new(details: &'a [ScoreDetails], weight: f64) -> Self {
+        Self { details, weight }
+    }
+
+    pub fn weighted_global_score(&self) -> f64 {
+        ScoreDetails::global_score(self.details.iter()) * self.weight
+    }
+
+    pub fn compare_weighted_global_scores(&self, other: &Self) -> Ordering {
+        self.weighted_global_score()
+            .partial_cmp(&other.weighted_global_score())
+            // both are numbers, possibly infinite
+            .unwrap()
+    }
+
+    pub fn compare(&self, other: &Self) -> Ordering {
+        let mut left_it = ScoreDetails::score_values(self.details.iter());
+        let mut right_it = ScoreDetails::score_values(other.details.iter());
+
+        loop {
+            let left = left_it.next();
+            let right = right_it.next();
+
+            match (left, right) {
+                (None, None) => return Ordering::Equal,
+                (None, Some(_)) => return Ordering::Less,
+                (Some(_), None) => return Ordering::Greater,
+                (Some(ScoreValue::Score(left)), Some(ScoreValue::Score(right))) => {
+                    let left = left * self.weight;
+                    let right = right * other.weight;
+                    if (left - right).abs() <= f64::EPSILON {
+                        continue;
+                    }
+                    return left.partial_cmp(&right).unwrap();
+                }
+                (Some(ScoreValue::Sort(left)), Some(ScoreValue::Sort(right))) => {
+                    match left.partial_cmp(right) {
+                        Some(Ordering::Equal) => continue,
+                        Some(order) => return order,
+                        None => return self.compare_weighted_global_scores(other),
+                    }
+                }
+                (Some(ScoreValue::GeoSort(left)), Some(ScoreValue::GeoSort(right))) => {
+                    match left.partial_cmp(right) {
+                        Some(Ordering::Equal) => continue,
+                        Some(order) => return order,
+                        None => {
+                            return self.compare_weighted_global_scores(other);
+                        }
+                    }
+                }
+                // not comparable details, use global
+                (Some(ScoreValue::Score(_)), Some(_))
+                | (Some(_), Some(ScoreValue::Score(_)))
+                | (Some(ScoreValue::GeoSort(_)), Some(ScoreValue::Sort(_)))
+                | (Some(ScoreValue::Sort(_)), Some(ScoreValue::GeoSort(_))) => {
+                    let left_count = left_it.count();
+                    let right_count = right_it.count();
+                    // compare how many remaining groups of rules each side has.
+                    // the group with the most remaining groups wins.
+                    return left_count
+                        .cmp(&right_count)
+                        // breaks ties with the global ranking score
+                        .then_with(|| self.compare_weighted_global_scores(other));
+                }
+            }
+        }
+    }
+}
+
+struct QueryByIndex {
+    query: SearchQuery,
+    federation_options: FederationOptions,
+    query_index: usize,
+}
+
+struct SearchResultByQuery<'a> {
+    documents_ids: Vec<DocumentId>,
+    document_scores: Vec<Vec<ScoreDetails>>,
+    federation_options: FederationOptions,
+    hit_maker: HitMaker<'a>,
+    query_index: usize,
+}
+
+struct SearchResultByQueryIter<'a> {
+    it: Zip<IntoIter<DocumentId>, IntoIter<Vec<ScoreDetails>>>,
+    federation_options: FederationOptions,
+    hit_maker: Rc<HitMaker<'a>>,
+    query_index: usize,
+}
+
+impl<'a> SearchResultByQueryIter<'a> {
+    fn new(
+        SearchResultByQuery {
+            documents_ids,
+            document_scores,
+            federation_options,
+            hit_maker,
+            query_index,
+        }: SearchResultByQuery<'a>,
+    ) -> Self {
+        let it = documents_ids.into_iter().zip(document_scores);
+        Self { it, federation_options, hit_maker: Rc::new(hit_maker), query_index }
+    }
+}
+
+struct SearchResultByQueryIterItem<'a> {
+    docid: DocumentId,
+    score: Vec<ScoreDetails>,
+    federation_options: FederationOptions,
+    hit_maker: Rc<HitMaker<'a>>,
+    query_index: usize,
+}
+
+fn merge_index_local_results(
+    results_by_query: Vec<SearchResultByQuery<'_>>,
+) -> impl Iterator<Item = SearchResultByQueryIterItem> + '_ {
+    itertools::kmerge_by(
+        results_by_query.into_iter().map(SearchResultByQueryIter::new),
+        |left: &SearchResultByQueryIterItem, right: &SearchResultByQueryIterItem| {
+            let left_score = WeightedScore::new(&left.score, *left.federation_options.weight);
+            let right_score = WeightedScore::new(&right.score, *right.federation_options.weight);
+
+            match left_score.compare(&right_score) {
+                // the biggest score goes first
+                Ordering::Greater => true,
+                // break ties using query index
+                Ordering::Equal => left.query_index < right.query_index,
+                Ordering::Less => false,
+            }
+        },
+    )
+}
+
+fn merge_index_global_results(
+    results_by_index: Vec<SearchResultByIndex>,
+) -> impl Iterator<Item = SearchHitByIndex> {
+    itertools::kmerge_by(
+        results_by_index.into_iter().map(|result_by_index| result_by_index.hits.into_iter()),
+        |left: &SearchHitByIndex, right: &SearchHitByIndex| {
+            let left_score = WeightedScore::new(&left.score, *left.federation_options.weight);
+            let right_score = WeightedScore::new(&right.score, *right.federation_options.weight);
+
+            match left_score.compare(&right_score) {
+                // the biggest score goes first
+                Ordering::Greater => true,
+                // break ties using query index
+                Ordering::Equal => left.query_index < right.query_index,
+                Ordering::Less => false,
+            }
+        },
+    )
+}
+
+impl<'a> Iterator for SearchResultByQueryIter<'a> {
+    type Item = SearchResultByQueryIterItem<'a>;
+
+    fn next(&mut self) -> Option<Self::Item> {
+        let (docid, score) = self.it.next()?;
+        Some(SearchResultByQueryIterItem {
+            docid,
+            score,
+            federation_options: self.federation_options,
+            hit_maker: Rc::clone(&self.hit_maker),
+            query_index: self.query_index,
+        })
+    }
+}
+
+struct SearchHitByIndex {
+    hit: SearchHit,
+    score: Vec<ScoreDetails>,
+    federation_options: FederationOptions,
+    query_index: usize,
+}
+
+struct SearchResultByIndex {
+    index: String,
+    hits: Vec<SearchHitByIndex>,
+    estimated_total_hits: usize,
+    degraded: bool,
+    used_negative_operator: bool,
+    facets: Option<ComputedFacets>,
+}
+
+#[derive(Debug, Clone, Default, Serialize)]
+pub struct FederatedFacets(pub BTreeMap<String, ComputedFacets>);
+
+impl FederatedFacets {
+    pub fn insert(&mut self, index: String, facets: Option<ComputedFacets>) {
+        if let Some(facets) = facets {
+            self.0.insert(index, facets);
+        }
+    }
+
+    pub fn is_empty(&self) -> bool {
+        self.0.is_empty()
+    }
+
+    pub fn merge(
+        self,
+        MergeFacets { max_values_per_facet }: MergeFacets,
+        facet_order: BTreeMap<String, (String, OrderBy)>,
+    ) -> Option<ComputedFacets> {
+        if self.is_empty() {
+            return None;
+        }
+
+        let mut distribution: BTreeMap<String, _> = Default::default();
+        let mut stats: BTreeMap<String, FacetStats> = Default::default();
+
+        for facets_by_index in self.0.into_values() {
+            for (facet, index_distribution) in facets_by_index.distribution {
+                match distribution.entry(facet) {
+                    std::collections::btree_map::Entry::Vacant(entry) => {
+                        entry.insert(index_distribution);
+                    }
+                    std::collections::btree_map::Entry::Occupied(mut entry) => {
+                        let distribution = entry.get_mut();
+
+                        for (value, index_count) in index_distribution {
+                            distribution
+                                .entry(value)
+                                .and_modify(|count| *count += index_count)
+                                .or_insert(index_count);
+                        }
+                    }
+                }
+            }
+
+            for (facet, index_stats) in facets_by_index.stats {
+                match stats.entry(facet) {
+                    std::collections::btree_map::Entry::Vacant(entry) => {
+                        entry.insert(index_stats);
+                    }
+                    std::collections::btree_map::Entry::Occupied(mut entry) => {
+                        let stats = entry.get_mut();
+
+                        stats.min = f64::min(stats.min, index_stats.min);
+                        stats.max = f64::max(stats.max, index_stats.max);
+                    }
+                }
+            }
+        }
+
+        // fixup order
+        for (facet, values) in &mut distribution {
+            let order_by = facet_order.get(facet).map(|(_, order)| *order).unwrap_or_default();
+
+            match order_by {
+                OrderBy::Lexicographic => {
+                    values.sort_unstable_by(|left, _, right, _| left.cmp(right))
+                }
+                OrderBy::Count => {
+                    values.sort_unstable_by(|_, left, _, right| {
+                        left.cmp(right)
+                            // biggest first
+                            .reverse()
+                    })
+                }
+            }
+
+            if let Some(max_values_per_facet) = max_values_per_facet {
+                values.truncate(max_values_per_facet)
+            };
+        }
+
+        Some(ComputedFacets { distribution, stats })
+    }
+}
+
+pub fn perform_federated_search(
+    index_scheduler: &IndexScheduler,
+    queries: Vec<SearchQueryWithIndex>,
+    mut federation: Federation,
+    features: RoFeatures,
+) -> Result<FederatedSearchResult, ResponseError> {
+    let before_search = std::time::Instant::now();
+
+    // this implementation partition the queries by index to guarantee an important property:
+    // - all the queries to a particular index use the same read transaction.
+    // This is an important property, otherwise we cannot guarantee the self-consistency of the results.
+
+    // 1. partition queries by index
+    let mut queries_by_index: BTreeMap<String, Vec<QueryByIndex>> = Default::default();
+    for (query_index, federated_query) in queries.into_iter().enumerate() {
+        if let Some(pagination_field) = federated_query.has_pagination() {
+            return Err(MeilisearchHttpError::PaginationInFederatedQuery(
+                query_index,
+                pagination_field,
+            )
+            .into());
+        }
+
+        if let Some(facets) = federated_query.has_facets() {
+            let facets = facets.to_owned();
+            return Err(MeilisearchHttpError::FacetsInFederatedQuery(
+                query_index,
+                federated_query.index_uid.into_inner(),
+                facets,
+            )
+            .into());
+        }
+
+        let (index_uid, query, federation_options) = federated_query.into_index_query_federation();
+
+        queries_by_index.entry(index_uid.into_inner()).or_default().push(QueryByIndex {
+            query,
+            federation_options: federation_options.unwrap_or_default(),
+            query_index,
+        })
+    }
+
+    // 2. perform queries, merge and make hits index by index
+    let required_hit_count = federation.limit + federation.offset;
+
+    // In step (2), semantic_hit_count will be set to Some(0) if any search kind uses semantic
+    // Then in step (3), we'll update its value if there is any semantic search
+    let mut semantic_hit_count = None;
+    let mut results_by_index = Vec::with_capacity(queries_by_index.len());
+    let mut previous_query_data: Option<(RankingRules, usize, String)> = None;
+
+    // remember the order and name of first index for each facet when merging with index settings
+    // to detect if the order is inconsistent for a facet.
+    let mut facet_order: Option<BTreeMap<String, (String, OrderBy)>> = match federation.merge_facets
+    {
+        Some(MergeFacets { .. }) => Some(Default::default()),
+        _ => None,
+    };
+
+    for (index_uid, queries) in queries_by_index {
+        let first_query_index = queries.first().map(|query| query.query_index);
+
+        let index = match index_scheduler.index(&index_uid) {
+            Ok(index) => index,
+            Err(err) => {
+                let mut err = ResponseError::from(err);
+                // Patch the HTTP status code to 400 as it defaults to 404 for `index_not_found`, but
+                // here the resource not found is not part of the URL.
+                err.code = StatusCode::BAD_REQUEST;
+                if let Some(query_index) = first_query_index {
+                    err.message = format!("Inside `.queries[{}]`: {}", query_index, err.message);
+                }
+                return Err(err);
+            }
+        };
+
+        // Important: this is the only transaction we'll use for this index during this federated search
+        let rtxn = index.read_txn()?;
+
+        let criteria = index.criteria(&rtxn)?;
+
+        let dictionary = index.dictionary(&rtxn)?;
+        let dictionary: Option<Vec<_>> =
+            dictionary.as_ref().map(|x| x.iter().map(String::as_str).collect());
+        let separators = index.allowed_separators(&rtxn)?;
+        let separators: Option<Vec<_>> =
+            separators.as_ref().map(|x| x.iter().map(String::as_str).collect());
+
+        // each query gets its individual cutoff
+        let cutoff = index.search_cutoff(&rtxn)?;
+
+        let mut degraded = false;
+        let mut used_negative_operator = false;
+        let mut candidates = RoaringBitmap::new();
+
+        let facets_by_index = federation.facets_by_index.remove(&index_uid).flatten();
+
+        // TODO: recover the max size + facets_by_index as return value of this function so as not to ask it for all queries
+        if let Err(mut error) =
+            check_facet_order(&mut facet_order, &index_uid, &facets_by_index, &index, &rtxn)
+        {
+            error.message = format!(
+                "Inside `.federation.facetsByIndex.{index_uid}`: {error}{}",
+                if let Some(query_index) = first_query_index {
+                    format!("\n - Note: index `{index_uid}` used in `.queries[{query_index}]`")
+                } else {
+                    Default::default()
+                }
+            );
+            return Err(error);
+        }
+
+        // 2.1. Compute all candidates for each query in the index
+        let mut results_by_query = Vec::with_capacity(queries.len());
+
+        for QueryByIndex { query, federation_options, query_index } in queries {
+            // use an immediately invoked lambda to capture the result without returning from the function
+
+            let res: Result<(), ResponseError> = (|| {
+                let search_kind = search_kind(&query, index_scheduler, &index, features)?;
+
+                let canonicalization_kind = match (&search_kind, &query.q) {
+                    (SearchKind::SemanticOnly { .. }, _) => {
+                        ranking_rules::CanonicalizationKind::Vector
+                    }
+                    (_, Some(q)) if !q.is_empty() => ranking_rules::CanonicalizationKind::Keyword,
+                    _ => ranking_rules::CanonicalizationKind::Placeholder,
+                };
+
+                let sort = if let Some(sort) = &query.sort {
+                    let sorts: Vec<_> =
+                        match sort.iter().map(|s| milli::AscDesc::from_str(s)).collect() {
+                            Ok(sorts) => sorts,
+                            Err(asc_desc_error) => {
+                                return Err(milli::Error::from(milli::SortError::from(
+                                    asc_desc_error,
+                                ))
+                                .into())
+                            }
+                        };
+                    Some(sorts)
+                } else {
+                    None
+                };
+
+                let ranking_rules = ranking_rules::RankingRules::new(
+                    criteria.clone(),
+                    sort,
+                    query.matching_strategy.into(),
+                    canonicalization_kind,
+                );
+
+                if let Some((previous_ranking_rules, previous_query_index, previous_index_uid)) =
+                    previous_query_data.take()
+                {
+                    if let Err(error) = ranking_rules.is_compatible_with(&previous_ranking_rules) {
+                        return Err(error.to_response_error(
+                            &ranking_rules,
+                            &previous_ranking_rules,
+                            query_index,
+                            previous_query_index,
+                            &index_uid,
+                            &previous_index_uid,
+                        ));
+                    }
+                    previous_query_data = if previous_ranking_rules.constraint_count()
+                        > ranking_rules.constraint_count()
+                    {
+                        Some((previous_ranking_rules, previous_query_index, previous_index_uid))
+                    } else {
+                        Some((ranking_rules, query_index, index_uid.clone()))
+                    };
+                } else {
+                    previous_query_data = Some((ranking_rules, query_index, index_uid.clone()));
+                }
+
+                match search_kind {
+                    SearchKind::KeywordOnly => {}
+                    _ => semantic_hit_count = Some(0),
+                }
+
+                let retrieve_vectors = RetrieveVectors::new(query.retrieve_vectors, features)?;
+
+                let time_budget = match cutoff {
+                    Some(cutoff) => TimeBudget::new(Duration::from_millis(cutoff)),
+                    None => TimeBudget::default(),
+                };
+
+                let (mut search, _is_finite_pagination, _max_total_hits, _offset) =
+                    prepare_search(&index, &rtxn, &query, &search_kind, time_budget, features)?;
+
+                search.scoring_strategy(milli::score_details::ScoringStrategy::Detailed);
+                search.offset(0);
+                search.limit(required_hit_count);
+
+                let (result, _semantic_hit_count) = super::search_from_kind(search_kind, search)?;
+                let format = AttributesFormat {
+                    attributes_to_retrieve: query.attributes_to_retrieve,
+                    retrieve_vectors,
+                    attributes_to_highlight: query.attributes_to_highlight,
+                    attributes_to_crop: query.attributes_to_crop,
+                    crop_length: query.crop_length,
+                    crop_marker: query.crop_marker,
+                    highlight_pre_tag: query.highlight_pre_tag,
+                    highlight_post_tag: query.highlight_post_tag,
+                    show_matches_position: query.show_matches_position,
+                    sort: query.sort,
+                    show_ranking_score: query.show_ranking_score,
+                    show_ranking_score_details: query.show_ranking_score_details,
+                    locales: query.locales.map(|l| l.iter().copied().map(Into::into).collect()),
+                };
+
+                let milli::SearchResult {
+                    matching_words,
+                    candidates: query_candidates,
+                    documents_ids,
+                    document_scores,
+                    degraded: query_degraded,
+                    used_negative_operator: query_used_negative_operator,
+                } = result;
+
+                candidates |= query_candidates;
+                degraded |= query_degraded;
+                used_negative_operator |= query_used_negative_operator;
+
+                let tokenizer = HitMaker::tokenizer(dictionary.as_deref(), separators.as_deref());
+
+                let formatter_builder = HitMaker::formatter_builder(matching_words, tokenizer);
+
+                let hit_maker = HitMaker::new(&index, &rtxn, format, formatter_builder)?;
+
+                results_by_query.push(SearchResultByQuery {
+                    federation_options,
+                    hit_maker,
+                    query_index,
+                    documents_ids,
+                    document_scores,
+                });
+                Ok(())
+            })();
+
+            if let Err(mut error) = res {
+                error.message = format!("Inside `.queries[{query_index}]`: {}", error.message);
+                return Err(error);
+            }
+        }
+        // 2.2. merge inside index
+        let mut documents_seen = RoaringBitmap::new();
+        let merged_result: Result<Vec<_>, ResponseError> =
+            merge_index_local_results(results_by_query)
+                // skip documents we've already seen & mark that we saw the current document
+                .filter(|SearchResultByQueryIterItem { docid, .. }| documents_seen.insert(*docid))
+                .take(required_hit_count)
+                // 2.3 make hits
+                .map(
+                    |SearchResultByQueryIterItem {
+                         docid,
+                         score,
+                         federation_options,
+                         hit_maker,
+                         query_index,
+                     }| {
+                        let mut hit = hit_maker.make_hit(docid, &score)?;
+                        let weighted_score =
+                            ScoreDetails::global_score(score.iter()) * (*federation_options.weight);
+
+                        let _federation = serde_json::json!(
+                            {
+                                "indexUid": index_uid,
+                                "queriesPosition": query_index,
+                                "weightedRankingScore": weighted_score,
+                            }
+                        );
+                        hit.document.insert("_federation".to_string(), _federation);
+                        Ok(SearchHitByIndex { hit, score, federation_options, query_index })
+                    },
+                )
+                .collect();
+
+        let merged_result = merged_result?;
+
+        let estimated_total_hits = candidates.len() as usize;
+
+        let facets = facets_by_index
+            .map(|facets_by_index| {
+                compute_facet_distribution_stats(
+                    &facets_by_index,
+                    &index,
+                    &rtxn,
+                    candidates,
+                    super::Route::MultiSearch,
+                )
+            })
+            .transpose()
+            .map_err(|mut error| {
+                error.message = format!(
+                    "Inside `.federation.facetsByIndex.{index_uid}`: {}{}",
+                    error.message,
+                    if let Some(query_index) = first_query_index {
+                        format!("\n - Note: index `{index_uid}` used in `.queries[{query_index}]`")
+                    } else {
+                        Default::default()
+                    }
+                );
+                error
+            })?;
+
+        results_by_index.push(SearchResultByIndex {
+            index: index_uid,
+            hits: merged_result,
+            estimated_total_hits,
+            degraded,
+            used_negative_operator,
+            facets,
+        });
+    }
+
+    // bonus step, make sure to return an error if an index wants a non-faceted field, even if no query actually uses that index.
+    for (index_uid, facets) in federation.facets_by_index {
+        let index = match index_scheduler.index(&index_uid) {
+            Ok(index) => index,
+            Err(err) => {
+                let mut err = ResponseError::from(err);
+                // Patch the HTTP status code to 400 as it defaults to 404 for `index_not_found`, but
+                // here the resource not found is not part of the URL.
+                err.code = StatusCode::BAD_REQUEST;
+                err.message = format!(
+                    "Inside `.federation.facetsByIndex.{index_uid}`: {}\n - Note: index `{index_uid}` is not used in queries",
+                    err.message
+                );
+                return Err(err);
+            }
+        };
+
+        // Important: this is the only transaction we'll use for this index during this federated search
+        let rtxn = index.read_txn()?;
+
+        if let Err(mut error) =
+            check_facet_order(&mut facet_order, &index_uid, &facets, &index, &rtxn)
+        {
+            error.message = format!(
+                "Inside `.federation.facetsByIndex.{index_uid}`: {error}\n - Note: index `{index_uid}` is not used in queries",
+            );
+            return Err(error);
+        }
+
+        if let Some(facets) = facets {
+            if let Err(mut error) = compute_facet_distribution_stats(
+                &facets,
+                &index,
+                &rtxn,
+                Default::default(),
+                super::Route::MultiSearch,
+            ) {
+                error.message =
+                    format!("Inside `.federation.facetsByIndex.{index_uid}`: {}\n - Note: index `{index_uid}` is not used in queries", error.message);
+                return Err(error);
+            }
+        }
+    }
+
+    // 3. merge hits and metadata across indexes
+    // 3.1 merge metadata
+    let (estimated_total_hits, degraded, used_negative_operator, facets) = {
+        let mut estimated_total_hits = 0;
+        let mut degraded = false;
+        let mut used_negative_operator = false;
+
+        let mut facets: FederatedFacets = FederatedFacets::default();
+
+        for SearchResultByIndex {
+            index,
+            hits: _,
+            estimated_total_hits: estimated_total_hits_by_index,
+            facets: facets_by_index,
+            degraded: degraded_by_index,
+            used_negative_operator: used_negative_operator_by_index,
+        } in &mut results_by_index
+        {
+            estimated_total_hits += *estimated_total_hits_by_index;
+            degraded |= *degraded_by_index;
+            used_negative_operator |= *used_negative_operator_by_index;
+
+            let facets_by_index = std::mem::take(facets_by_index);
+            let index = std::mem::take(index);
+
+            facets.insert(index, facets_by_index);
+        }
+
+        (estimated_total_hits, degraded, used_negative_operator, facets)
+    };
+
+    // 3.2 merge hits
+    let merged_hits: Vec<_> = merge_index_global_results(results_by_index)
+        .skip(federation.offset)
+        .take(federation.limit)
+        .inspect(|hit| {
+            if let Some(semantic_hit_count) = &mut semantic_hit_count {
+                if hit.score.iter().any(|score| matches!(&score, ScoreDetails::Vector(_))) {
+                    *semantic_hit_count += 1;
+                }
+            }
+        })
+        .map(|hit| hit.hit)
+        .collect();
+
+    let (facet_distribution, facet_stats, facets_by_index) =
+        match federation.merge_facets.zip(facet_order) {
+            Some((merge_facets, facet_order)) => {
+                let facets = facets.merge(merge_facets, facet_order);
+
+                let (facet_distribution, facet_stats) = facets
+                    .map(|ComputedFacets { distribution, stats }| (distribution, stats))
+                    .unzip();
+
+                (facet_distribution, facet_stats, FederatedFacets::default())
+            }
+            None => (None, None, facets),
+        };
+
+    let search_result = FederatedSearchResult {
+        hits: merged_hits,
+        processing_time_ms: before_search.elapsed().as_millis(),
+        hits_info: HitsInfo::OffsetLimit {
+            limit: federation.limit,
+            offset: federation.offset,
+            estimated_total_hits,
+        },
+        semantic_hit_count,
+        degraded,
+        used_negative_operator,
+        facet_distribution,
+        facet_stats,
+        facets_by_index,
+    };
+
+    Ok(search_result)
+}
+
+fn check_facet_order(
+    facet_order: &mut Option<BTreeMap<String, (String, OrderBy)>>,
+    current_index: &str,
+    facets_by_index: &Option<Vec<String>>,
+    index: &milli::Index,
+    rtxn: &milli::heed::RoTxn<'_>,
+) -> Result<(), ResponseError> {
+    if let (Some(facet_order), Some(facets_by_index)) = (facet_order, facets_by_index) {
+        let index_facet_order = index.sort_facet_values_by(rtxn)?;
+        for facet in facets_by_index {
+            let index_facet_order = index_facet_order.get(facet);
+            let (previous_index, previous_facet_order) = facet_order
+                .entry(facet.to_owned())
+                .or_insert_with(|| (current_index.to_owned(), index_facet_order));
+            if previous_facet_order != &index_facet_order {
+                return Err(MeilisearchHttpError::InconsistentFacetOrder {
+                    facet: facet.clone(),
+                    previous_facet_order: *previous_facet_order,
+                    previous_uid: previous_index.clone(),
+                    current_uid: current_index.to_owned(),
+                    index_facet_order,
+                }
+                .into());
+            }
+        }
+    };
+    Ok(())
+}

crates/meilisearch/src/search/ranking_rules.rs

@@ -0,0 +1,823 @@
+use std::collections::HashMap;
+use std::fmt::Write;
+
+use itertools::Itertools as _;
+use meilisearch_types::error::{Code, ResponseError};
+use meilisearch_types::milli::{AscDesc, Criterion, Member, TermsMatchingStrategy};
+
+pub struct RankingRules {
+    canonical_criteria: Vec<Criterion>,
+    canonical_sort: Option<Vec<AscDesc>>,
+    canonicalization_actions: Vec<CanonicalizationAction>,
+    source_criteria: Vec<Criterion>,
+    source_sort: Option<Vec<AscDesc>>,
+}
+
+pub enum CanonicalizationAction {
+    PrependedWords {
+        prepended_index: RankingRuleSource,
+    },
+    RemovedDuplicate {
+        earlier_occurrence: RankingRuleSource,
+        removed_occurrence: RankingRuleSource,
+    },
+    RemovedWords {
+        reason: RemoveWords,
+        removed_occurrence: RankingRuleSource,
+    },
+    RemovedPlaceholder {
+        removed_occurrence: RankingRuleSource,
+    },
+    TruncatedVector {
+        vector_rule: RankingRuleSource,
+        truncated_from: RankingRuleSource,
+    },
+    RemovedVector {
+        vector_rule: RankingRuleSource,
+        removed_occurrence: RankingRuleSource,
+    },
+    RemovedSort {
+        removed_occurrence: RankingRuleSource,
+    },
+}
+
+pub enum RemoveWords {
+    WasPrepended,
+    MatchingStrategyAll,
+}
+
+impl std::fmt::Display for RemoveWords {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        let reason = match self {
+            RemoveWords::WasPrepended => "it was previously prepended",
+            RemoveWords::MatchingStrategyAll => "`query.matchingWords` is set to `all`",
+        };
+        f.write_str(reason)
+    }
+}
+
+pub enum CanonicalizationKind {
+    Placeholder,
+    Keyword,
+    Vector,
+}
+
+pub struct CompatibilityError {
+    previous: RankingRule,
+    current: RankingRule,
+}
+impl CompatibilityError {
+    pub(crate) fn to_response_error(
+        &self,
+        ranking_rules: &RankingRules,
+        previous_ranking_rules: &RankingRules,
+        query_index: usize,
+        previous_query_index: usize,
+        index_uid: &str,
+        previous_index_uid: &str,
+    ) -> meilisearch_types::error::ResponseError {
+        let rule = self.current.as_string(
+            &ranking_rules.canonical_criteria,
+            &ranking_rules.canonical_sort,
+            query_index,
+            index_uid,
+        );
+        let previous_rule = self.previous.as_string(
+            &previous_ranking_rules.canonical_criteria,
+            &previous_ranking_rules.canonical_sort,
+            previous_query_index,
+            previous_index_uid,
+        );
+
+        let canonicalization_actions = ranking_rules.canonicalization_notes();
+        let previous_canonicalization_actions = previous_ranking_rules.canonicalization_notes();
+
+        let mut msg = String::new();
+        let reason = self.reason();
+        let _ = writeln!(
+            &mut msg,
+            "The results of queries #{previous_query_index} and #{query_index} are incompatible: "
+        );
+        let _ = writeln!(&mut msg, "  1. {previous_rule}");
+        let _ = writeln!(&mut msg, "  2. {rule}");
+        let _ = writeln!(&mut msg, "  - {reason}");
+
+        if !previous_canonicalization_actions.is_empty() {
+            let _ = write!(&mut msg, "  - note: The ranking rules of query #{previous_query_index} were modified during canonicalization:\n{previous_canonicalization_actions}");
+        }
+
+        if !canonicalization_actions.is_empty() {
+            let _ = write!(&mut msg, "  - note: The ranking rules of query #{query_index} were modified during canonicalization:\n{canonicalization_actions}");
+        }
+
+        ResponseError::from_msg(msg, Code::InvalidMultiSearchQueryRankingRules)
+    }
+    pub fn reason(&self) -> &'static str {
+        match (self.previous.kind, self.current.kind) {
+            (RankingRuleKind::Relevancy, RankingRuleKind::AscendingSort)
+            | (RankingRuleKind::Relevancy, RankingRuleKind::DescendingSort)
+            | (RankingRuleKind::AscendingSort, RankingRuleKind::Relevancy)
+            | (RankingRuleKind::DescendingSort, RankingRuleKind::Relevancy) => {
+                "cannot compare a relevancy rule with a sort rule"
+            }
+
+            (RankingRuleKind::Relevancy, RankingRuleKind::AscendingGeoSort)
+            | (RankingRuleKind::Relevancy, RankingRuleKind::DescendingGeoSort)
+            | (RankingRuleKind::AscendingGeoSort, RankingRuleKind::Relevancy)
+            | (RankingRuleKind::DescendingGeoSort, RankingRuleKind::Relevancy) => {
+                "cannot compare a relevancy rule with a geosort rule"
+            }
+
+            (RankingRuleKind::AscendingSort, RankingRuleKind::DescendingSort)
+            | (RankingRuleKind::DescendingSort, RankingRuleKind::AscendingSort) => {
+                "cannot compare two sort rules in opposite directions"
+            }
+
+            (RankingRuleKind::AscendingSort, RankingRuleKind::AscendingGeoSort)
+            | (RankingRuleKind::AscendingSort, RankingRuleKind::DescendingGeoSort)
+            | (RankingRuleKind::DescendingSort, RankingRuleKind::AscendingGeoSort)
+            | (RankingRuleKind::DescendingSort, RankingRuleKind::DescendingGeoSort)
+            | (RankingRuleKind::AscendingGeoSort, RankingRuleKind::AscendingSort)
+            | (RankingRuleKind::AscendingGeoSort, RankingRuleKind::DescendingSort)
+            | (RankingRuleKind::DescendingGeoSort, RankingRuleKind::AscendingSort)
+            | (RankingRuleKind::DescendingGeoSort, RankingRuleKind::DescendingSort) => {
+                "cannot compare a sort rule with a geosort rule"
+            }
+
+            (RankingRuleKind::AscendingGeoSort, RankingRuleKind::DescendingGeoSort)
+            | (RankingRuleKind::DescendingGeoSort, RankingRuleKind::AscendingGeoSort) => {
+                "cannot compare two geosort rules in opposite directions"
+            }
+            (RankingRuleKind::Relevancy, RankingRuleKind::Relevancy)
+            | (RankingRuleKind::AscendingSort, RankingRuleKind::AscendingSort)
+            | (RankingRuleKind::DescendingSort, RankingRuleKind::DescendingSort)
+            | (RankingRuleKind::AscendingGeoSort, RankingRuleKind::AscendingGeoSort)
+            | (RankingRuleKind::DescendingGeoSort, RankingRuleKind::DescendingGeoSort) => {
+                "internal error, comparison should be possible"
+            }
+        }
+    }
+}
+
+impl RankingRules {
+    pub fn new(
+        criteria: Vec<Criterion>,
+        sort: Option<Vec<AscDesc>>,
+        terms_matching_strategy: TermsMatchingStrategy,
+        canonicalization_kind: CanonicalizationKind,
+    ) -> Self {
+        let (canonical_criteria, canonical_sort, canonicalization_actions) =
+            Self::canonicalize(&criteria, &sort, terms_matching_strategy, canonicalization_kind);
+        Self {
+            canonical_criteria,
+            canonical_sort,
+            canonicalization_actions,
+            source_criteria: criteria,
+            source_sort: sort,
+        }
+    }
+
+    fn canonicalize(
+        criteria: &[Criterion],
+        sort: &Option<Vec<AscDesc>>,
+        terms_matching_strategy: TermsMatchingStrategy,
+        canonicalization_kind: CanonicalizationKind,
+    ) -> (Vec<Criterion>, Option<Vec<AscDesc>>, Vec<CanonicalizationAction>) {
+        match canonicalization_kind {
+            CanonicalizationKind::Placeholder => Self::canonicalize_placeholder(criteria, sort),
+            CanonicalizationKind::Keyword => {
+                Self::canonicalize_keyword(criteria, sort, terms_matching_strategy)
+            }
+            CanonicalizationKind::Vector => Self::canonicalize_vector(criteria, sort),
+        }
+    }
+
+    fn canonicalize_placeholder(
+        criteria: &[Criterion],
+        sort_query: &Option<Vec<AscDesc>>,
+    ) -> (Vec<Criterion>, Option<Vec<AscDesc>>, Vec<CanonicalizationAction>) {
+        let mut sort = None;
+
+        let mut sorted_fields = HashMap::new();
+        let mut canonicalization_actions = Vec::new();
+        let mut canonical_criteria = Vec::new();
+        let mut canonical_sort = None;
+
+        for (criterion_index, criterion) in criteria.iter().enumerate() {
+            match criterion.clone() {
+                Criterion::Words
+                | Criterion::Typo
+                | Criterion::Proximity
+                | Criterion::Attribute
+                | Criterion::Exactness => {
+                    canonicalization_actions.push(CanonicalizationAction::RemovedPlaceholder {
+                        removed_occurrence: RankingRuleSource::Criterion(criterion_index),
+                    })
+                }
+
+                Criterion::Sort => {
+                    if let Some(previous_index) = sort {
+                        canonicalization_actions.push(CanonicalizationAction::RemovedDuplicate {
+                            earlier_occurrence: RankingRuleSource::Criterion(previous_index),
+                            removed_occurrence: RankingRuleSource::Criterion(criterion_index),
+                        });
+                    } else if let Some(sort_query) = sort_query {
+                        sort = Some(criterion_index);
+                        canonical_criteria.push(criterion.clone());
+                        canonical_sort = Some(canonicalize_sort(
+                            &mut sorted_fields,
+                            sort_query.as_slice(),
+                            criterion_index,
+                            &mut canonicalization_actions,
+                        ));
+                    } else {
+                        canonicalization_actions.push(CanonicalizationAction::RemovedSort {
+                            removed_occurrence: RankingRuleSource::Criterion(criterion_index),
+                        })
+                    }
+                }
+                Criterion::Asc(s) | Criterion::Desc(s) => match sorted_fields.entry(s) {
+                    std::collections::hash_map::Entry::Occupied(entry) => canonicalization_actions
+                        .push(CanonicalizationAction::RemovedDuplicate {
+                            earlier_occurrence: *entry.get(),
+                            removed_occurrence: RankingRuleSource::Criterion(criterion_index),
+                        }),
+                    std::collections::hash_map::Entry::Vacant(entry) => {
+                        entry.insert(RankingRuleSource::Criterion(criterion_index));
+                        canonical_criteria.push(criterion.clone())
+                    }
+                },
+            }
+        }
+
+        (canonical_criteria, canonical_sort, canonicalization_actions)
+    }
+
+    fn canonicalize_vector(
+        criteria: &[Criterion],
+        sort_query: &Option<Vec<AscDesc>>,
+    ) -> (Vec<Criterion>, Option<Vec<AscDesc>>, Vec<CanonicalizationAction>) {
+        let mut sort = None;
+
+        let mut sorted_fields = HashMap::new();
+        let mut canonicalization_actions = Vec::new();
+        let mut canonical_criteria = Vec::new();
+        let mut canonical_sort = None;
+
+        let mut vector = None;
+
+        'criteria: for (criterion_index, criterion) in criteria.iter().enumerate() {
+            match criterion.clone() {
+                Criterion::Words
+                | Criterion::Typo
+                | Criterion::Proximity
+                | Criterion::Attribute
+                | Criterion::Exactness => match vector {
+                    Some(previous_occurrence) => {
+                        if sorted_fields.is_empty() {
+                            canonicalization_actions.push(CanonicalizationAction::RemovedVector {
+                                vector_rule: RankingRuleSource::Criterion(previous_occurrence),
+                                removed_occurrence: RankingRuleSource::Criterion(criterion_index),
+                            });
+                        } else {
+                            canonicalization_actions.push(
+                                CanonicalizationAction::TruncatedVector {
+                                    vector_rule: RankingRuleSource::Criterion(previous_occurrence),
+                                    truncated_from: RankingRuleSource::Criterion(criterion_index),
+                                },
+                            );
+                            break 'criteria;
+                        }
+                    }
+                    None => {
+                        canonical_criteria.push(criterion.clone());
+                        vector = Some(criterion_index);
+                    }
+                },
+
+                Criterion::Sort => {
+                    if let Some(previous_index) = sort {
+                        canonicalization_actions.push(CanonicalizationAction::RemovedDuplicate {
+                            earlier_occurrence: RankingRuleSource::Criterion(previous_index),
+                            removed_occurrence: RankingRuleSource::Criterion(criterion_index),
+                        });
+                    } else if let Some(sort_query) = sort_query {
+                        sort = Some(criterion_index);
+                        canonical_criteria.push(criterion.clone());
+                        canonical_sort = Some(canonicalize_sort(
+                            &mut sorted_fields,
+                            sort_query.as_slice(),
+                            criterion_index,
+                            &mut canonicalization_actions,
+                        ));
+                    } else {
+                        canonicalization_actions.push(CanonicalizationAction::RemovedSort {
+                            removed_occurrence: RankingRuleSource::Criterion(criterion_index),
+                        })
+                    }
+                }
+                Criterion::Asc(s) | Criterion::Desc(s) => match sorted_fields.entry(s) {
+                    std::collections::hash_map::Entry::Occupied(entry) => canonicalization_actions
+                        .push(CanonicalizationAction::RemovedDuplicate {
+                            earlier_occurrence: *entry.get(),
+                            removed_occurrence: RankingRuleSource::Criterion(criterion_index),
+                        }),
+                    std::collections::hash_map::Entry::Vacant(entry) => {
+                        entry.insert(RankingRuleSource::Criterion(criterion_index));
+                        canonical_criteria.push(criterion.clone())
+                    }
+                },
+            }
+        }
+
+        (canonical_criteria, canonical_sort, canonicalization_actions)
+    }
+
+    fn canonicalize_keyword(
+        criteria: &[Criterion],
+        sort_query: &Option<Vec<AscDesc>>,
+        terms_matching_strategy: TermsMatchingStrategy,
+    ) -> (Vec<Criterion>, Option<Vec<AscDesc>>, Vec<CanonicalizationAction>) {
+        let mut words = None;
+        let mut typo = None;
+        let mut proximity = None;
+        let mut sort = None;
+        let mut attribute = None;
+        let mut exactness = None;
+        let mut sorted_fields = HashMap::new();
+
+        let mut canonical_criteria = Vec::new();
+        let mut canonical_sort = None;
+
+        let mut canonicalization_actions = Vec::new();
+
+        for (criterion_index, criterion) in criteria.iter().enumerate() {
+            let criterion = criterion.clone();
+            match criterion.clone() {
+                Criterion::Words => {
+                    if let TermsMatchingStrategy::All = terms_matching_strategy {
+                        canonicalization_actions.push(CanonicalizationAction::RemovedWords {
+                            reason: RemoveWords::MatchingStrategyAll,
+                            removed_occurrence: RankingRuleSource::Criterion(criterion_index),
+                        });
+                        continue;
+                    }
+                    if let Some(maybe_previous_index) = words {
+                        if let Some(previous_index) = maybe_previous_index {
+                            canonicalization_actions.push(
+                                CanonicalizationAction::RemovedDuplicate {
+                                    earlier_occurrence: RankingRuleSource::Criterion(
+                                        previous_index,
+                                    ),
+                                    removed_occurrence: RankingRuleSource::Criterion(
+                                        criterion_index,
+                                    ),
+                                },
+                            );
+                            continue;
+                        }
+                        canonicalization_actions.push(CanonicalizationAction::RemovedWords {
+                            reason: RemoveWords::WasPrepended,
+                            removed_occurrence: RankingRuleSource::Criterion(criterion_index),
+                        })
+                    }
+                    words = Some(Some(criterion_index));
+                    canonical_criteria.push(criterion);
+                }
+                Criterion::Typo => {
+                    canonicalize_criterion(
+                        criterion,
+                        criterion_index,
+                        terms_matching_strategy,
+                        &mut words,
+                        &mut canonicalization_actions,
+                        &mut canonical_criteria,
+                        &mut typo,
+                    );
+                }
+                Criterion::Proximity => {
+                    canonicalize_criterion(
+                        criterion,
+                        criterion_index,
+                        terms_matching_strategy,
+                        &mut words,
+                        &mut canonicalization_actions,
+                        &mut canonical_criteria,
+                        &mut proximity,
+                    );
+                }
+                Criterion::Attribute => {
+                    canonicalize_criterion(
+                        criterion,
+                        criterion_index,
+                        terms_matching_strategy,
+                        &mut words,
+                        &mut canonicalization_actions,
+                        &mut canonical_criteria,
+                        &mut attribute,
+                    );
+                }
+                Criterion::Exactness => {
+                    canonicalize_criterion(
+                        criterion,
+                        criterion_index,
+                        terms_matching_strategy,
+                        &mut words,
+                        &mut canonicalization_actions,
+                        &mut canonical_criteria,
+                        &mut exactness,
+                    );
+                }
+
+                Criterion::Sort => {
+                    if let Some(previous_index) = sort {
+                        canonicalization_actions.push(CanonicalizationAction::RemovedDuplicate {
+                            earlier_occurrence: RankingRuleSource::Criterion(previous_index),
+                            removed_occurrence: RankingRuleSource::Criterion(criterion_index),
+                        });
+                    } else if let Some(sort_query) = sort_query {
+                        sort = Some(criterion_index);
+                        canonical_criteria.push(criterion);
+                        canonical_sort = Some(canonicalize_sort(
+                            &mut sorted_fields,
+                            sort_query.as_slice(),
+                            criterion_index,
+                            &mut canonicalization_actions,
+                        ));
+                    } else {
+                        canonicalization_actions.push(CanonicalizationAction::RemovedSort {
+                            removed_occurrence: RankingRuleSource::Criterion(criterion_index),
+                        })
+                    }
+                }
+                Criterion::Asc(s) | Criterion::Desc(s) => match sorted_fields.entry(s) {
+                    std::collections::hash_map::Entry::Occupied(entry) => canonicalization_actions
+                        .push(CanonicalizationAction::RemovedDuplicate {
+                            earlier_occurrence: *entry.get(),
+                            removed_occurrence: RankingRuleSource::Criterion(criterion_index),
+                        }),
+                    std::collections::hash_map::Entry::Vacant(entry) => {
+                        entry.insert(RankingRuleSource::Criterion(criterion_index));
+                        canonical_criteria.push(criterion)
+                    }
+                },
+            }
+        }
+
+        (canonical_criteria, canonical_sort, canonicalization_actions)
+    }
+
+    pub fn is_compatible_with(&self, previous: &Self) -> Result<(), CompatibilityError> {
+        for (current, previous) in self.coalesce_iterator().zip(previous.coalesce_iterator()) {
+            if current.kind != previous.kind {
+                return Err(CompatibilityError { current, previous });
+            }
+        }
+        Ok(())
+    }
+
+    pub fn constraint_count(&self) -> usize {
+        self.coalesce_iterator().count()
+    }
+
+    fn coalesce_iterator(&self) -> impl Iterator<Item = RankingRule> + '_ {
+        self.canonical_criteria
+            .iter()
+            .enumerate()
+            .flat_map(|(criterion_index, criterion)| {
+                RankingRule::from_criterion(criterion_index, criterion, &self.canonical_sort)
+            })
+            .coalesce(
+                |previous @ RankingRule { source: previous_source, kind: previous_kind },
+                 current @ RankingRule { source, kind }| {
+                    match (previous_kind, kind) {
+                        (RankingRuleKind::Relevancy, RankingRuleKind::Relevancy) => {
+                            let merged_source = match (previous_source, source) {
+                                (
+                                    RankingRuleSource::Criterion(previous),
+                                    RankingRuleSource::Criterion(current),
+                                ) => RankingRuleSource::CoalescedCriteria(previous, current),
+                                (
+                                    RankingRuleSource::CoalescedCriteria(begin, _end),
+                                    RankingRuleSource::Criterion(current),
+                                ) => RankingRuleSource::CoalescedCriteria(begin, current),
+                                (_previous, current) => current,
+                            };
+                            Ok(RankingRule { source: merged_source, kind })
+                        }
+                        _ => Err((previous, current)),
+                    }
+                },
+            )
+    }
+
+    fn canonicalization_notes(&self) -> String {
+        use CanonicalizationAction::*;
+        let mut notes = String::new();
+        for (index, action) in self.canonicalization_actions.iter().enumerate() {
+            let index = index + 1;
+            let _ = match action {
+                PrependedWords { prepended_index } => writeln!(
+                    &mut notes,
+                    "    {index}. Prepended rule `words` before first relevancy rule `{}` at position {}",
+                    prepended_index.rule_name(&self.source_criteria, &self.source_sort),
+                    prepended_index.rule_position()
+                ),
+                RemovedDuplicate { earlier_occurrence, removed_occurrence } => writeln!(
+                    &mut notes,
+                    "    {index}. Removed duplicate rule `{}` at position {} as it already appears at position {}",
+                    earlier_occurrence.rule_name(&self.source_criteria, &self.source_sort),
+                    removed_occurrence.rule_position(),
+                    earlier_occurrence.rule_position(),
+                ),
+                RemovedWords { reason, removed_occurrence } => writeln!(
+                    &mut notes,
+                    "    {index}. Removed rule `words` at position {} because {reason}",
+                    removed_occurrence.rule_position()
+                ),
+                RemovedPlaceholder { removed_occurrence } => writeln!(
+                    &mut notes,
+                    "    {index}. Removed relevancy rule `{}` at position {} because the query is a placeholder search (`q`: \"\")",
+                    removed_occurrence.rule_name(&self.source_criteria, &self.source_sort),
+                    removed_occurrence.rule_position()
+                ),
+                TruncatedVector { vector_rule, truncated_from } => writeln!(
+                    &mut notes,
+                    "    {index}. Truncated relevancy rule `{}` at position {} and later rules because the query is a vector search and `vector` was inserted at position {}",
+                    truncated_from.rule_name(&self.source_criteria, &self.source_sort),
+                    truncated_from.rule_position(),
+                    vector_rule.rule_position(),
+                ),
+                RemovedVector { vector_rule, removed_occurrence } => writeln!(
+                    &mut notes,
+                    "    {index}. Removed relevancy rule `{}` at position {} because the query is a vector search and `vector` was already inserted at position {}",
+                    removed_occurrence.rule_name(&self.source_criteria, &self.source_sort),
+                    removed_occurrence.rule_position(),
+                    vector_rule.rule_position(),
+                ),
+                RemovedSort { removed_occurrence } => writeln!(
+                    &mut notes,
+                    "   {index}. Removed rule `sort` at position {} because `query.sort` is empty",
+removed_occurrence.rule_position()
+                ),
+            };
+        }
+        notes
+    }
+}
+
+fn canonicalize_sort(
+    sorted_fields: &mut HashMap<String, RankingRuleSource>,
+    sort_query: &[AscDesc],
+    criterion_index: usize,
+    canonicalization_actions: &mut Vec<CanonicalizationAction>,
+) -> Vec<AscDesc> {
+    let mut geo_sorted = None;
+    let mut canonical_sort = Vec::new();
+    for (sort_index, asc_desc) in sort_query.iter().enumerate() {
+        let source = RankingRuleSource::Sort { criterion_index, sort_index };
+        let asc_desc = asc_desc.clone();
+        match asc_desc.clone() {
+            AscDesc::Asc(Member::Field(s)) | AscDesc::Desc(Member::Field(s)) => {
+                match sorted_fields.entry(s) {
+                    std::collections::hash_map::Entry::Occupied(entry) => canonicalization_actions
+                        .push(CanonicalizationAction::RemovedDuplicate {
+                            earlier_occurrence: *entry.get(),
+                            removed_occurrence: source,
+                        }),
+                    std::collections::hash_map::Entry::Vacant(entry) => {
+                        entry.insert(source);
+                        canonical_sort.push(asc_desc);
+                    }
+                }
+            }
+            AscDesc::Asc(Member::Geo(_)) | AscDesc::Desc(Member::Geo(_)) => match geo_sorted {
+                Some(earlier_sort_index) => {
+                    canonicalization_actions.push(CanonicalizationAction::RemovedDuplicate {
+                        earlier_occurrence: RankingRuleSource::Sort {
+                            criterion_index,
+                            sort_index: earlier_sort_index,
+                        },
+                        removed_occurrence: source,
+                    })
+                }
+                None => {
+                    geo_sorted = Some(sort_index);
+                    canonical_sort.push(asc_desc);
+                }
+            },
+        }
+    }
+    canonical_sort
+}
+
+fn canonicalize_criterion(
+    criterion: Criterion,
+    criterion_index: usize,
+    terms_matching_strategy: TermsMatchingStrategy,
+    words: &mut Option<Option<usize>>,
+    canonicalization_actions: &mut Vec<CanonicalizationAction>,
+    canonical_criteria: &mut Vec<Criterion>,
+    rule: &mut Option<usize>,
+) {
+    *words = match (terms_matching_strategy, words.take()) {
+        (TermsMatchingStrategy::All, words) => words,
+        (_, None) => {
+            // inject words
+            canonicalization_actions.push(CanonicalizationAction::PrependedWords {
+                prepended_index: RankingRuleSource::Criterion(criterion_index),
+            });
+            canonical_criteria.push(Criterion::Words);
+            Some(None)
+        }
+        (_, words) => words,
+    };
+    if let Some(previous_index) = *rule {
+        canonicalization_actions.push(CanonicalizationAction::RemovedDuplicate {
+            earlier_occurrence: RankingRuleSource::Criterion(previous_index),
+            removed_occurrence: RankingRuleSource::Criterion(criterion_index),
+        });
+    } else {
+        *rule = Some(criterion_index);
+        canonical_criteria.push(criterion)
+    }
+}
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+enum RankingRuleKind {
+    Relevancy,
+    AscendingSort,
+    DescendingSort,
+    AscendingGeoSort,
+    DescendingGeoSort,
+}
+
+#[derive(Debug, Clone, Copy)]
+pub struct RankingRule {
+    source: RankingRuleSource,
+    kind: RankingRuleKind,
+}
+
+#[derive(Debug, Clone, Copy)]
+pub enum RankingRuleSource {
+    Criterion(usize),
+    CoalescedCriteria(usize, usize),
+    Sort { criterion_index: usize, sort_index: usize },
+}
+
+impl RankingRuleSource {
+    fn rule_name(&self, criteria: &[Criterion], sort: &Option<Vec<AscDesc>>) -> String {
+        match self {
+            RankingRuleSource::Criterion(criterion_index) => criteria
+                .get(*criterion_index)
+                .map(|c| c.to_string())
+                .unwrap_or_else(|| "unknown".into()),
+            RankingRuleSource::CoalescedCriteria(begin, end) => {
+                let rules: Vec<_> = criteria
+                    .get(*begin..=*end)
+                    .iter()
+                    .flat_map(|c| c.iter())
+                    .map(|c| c.to_string())
+                    .collect();
+                rules.join(", ")
+            }
+            RankingRuleSource::Sort { criterion_index: _, sort_index } => {
+                match sort.as_deref().and_then(|sort| sort.get(*sort_index)) {
+                    Some(sort) => match sort {
+                        AscDesc::Asc(Member::Field(field_name)) => format!("{field_name}:asc"),
+                        AscDesc::Desc(Member::Field(field_name)) => {
+                            format!("{field_name}:desc")
+                        }
+                        AscDesc::Asc(Member::Geo(_)) => "_geo(..):asc".to_string(),
+                        AscDesc::Desc(Member::Geo(_)) => "_geo(..):desc".to_string(),
+                    },
+                    None => "unknown".into(),
+                }
+            }
+        }
+    }
+
+    fn rule_position(&self) -> String {
+        match self {
+            RankingRuleSource::Criterion(criterion_index) => {
+                format!("#{criterion_index} in ranking rules")
+            }
+            RankingRuleSource::CoalescedCriteria(begin, end) => {
+                format!("#{begin} to #{end} in ranking rules")
+            }
+            RankingRuleSource::Sort { criterion_index, sort_index } => format!(
+                "#{sort_index} in `query.sort` (as `sort` is #{criterion_index} in ranking rules)"
+            ),
+        }
+    }
+}
+
+impl RankingRule {
+    fn from_criterion<'a>(
+        criterion_index: usize,
+        criterion: &'a Criterion,
+        sort: &'a Option<Vec<AscDesc>>,
+    ) -> impl Iterator<Item = Self> + 'a {
+        let kind = match criterion {
+            Criterion::Words
+            | Criterion::Typo
+            | Criterion::Proximity
+            | Criterion::Attribute
+            | Criterion::Exactness => RankingRuleKind::Relevancy,
+            Criterion::Asc(s) if s == "_geo" => RankingRuleKind::AscendingGeoSort,
+
+            Criterion::Asc(_) => RankingRuleKind::AscendingSort,
+            Criterion::Desc(s) if s == "_geo" => RankingRuleKind::DescendingGeoSort,
+
+            Criterion::Desc(_) => RankingRuleKind::DescendingSort,
+            Criterion::Sort => {
+                return either::Right(sort.iter().flatten().enumerate().map(
+                    move |(rule_index, asc_desc)| {
+                        Self::from_asc_desc(asc_desc, criterion_index, rule_index)
+                    },
+                ))
+            }
+        };
+
+        either::Left(std::iter::once(Self {
+            source: RankingRuleSource::Criterion(criterion_index),
+            kind,
+        }))
+    }
+
+    fn from_asc_desc(asc_desc: &AscDesc, sort_index: usize, rule_index_in_sort: usize) -> Self {
+        let kind = match asc_desc {
+            AscDesc::Asc(Member::Field(_)) => RankingRuleKind::AscendingSort,
+            AscDesc::Desc(Member::Field(_)) => RankingRuleKind::DescendingSort,
+            AscDesc::Asc(Member::Geo(_)) => RankingRuleKind::AscendingGeoSort,
+            AscDesc::Desc(Member::Geo(_)) => RankingRuleKind::DescendingGeoSort,
+        };
+        Self {
+            source: RankingRuleSource::Sort {
+                criterion_index: sort_index,
+                sort_index: rule_index_in_sort,
+            },
+            kind,
+        }
+    }
+
+    fn as_string(
+        &self,
+        canonical_criteria: &[Criterion],
+        canonical_sort: &Option<Vec<AscDesc>>,
+        query_index: usize,
+        index_uid: &str,
+    ) -> String {
+        let kind = match self.kind {
+            RankingRuleKind::Relevancy => "relevancy",
+            RankingRuleKind::AscendingSort => "ascending sort",
+            RankingRuleKind::DescendingSort => "descending sort",
+            RankingRuleKind::AscendingGeoSort => "ascending geo sort",
+            RankingRuleKind::DescendingGeoSort => "descending geo sort",
+        };
+        let rules = self.fetch_from_source(canonical_criteria, canonical_sort);
+
+        let source = match self.source {
+            RankingRuleSource::Criterion(criterion_index) => format!("`queries[{query_index}]`, `{index_uid}.rankingRules[{criterion_index}]`"),
+            RankingRuleSource::CoalescedCriteria(begin, end) => format!("`queries[{query_index}]`, `{index_uid}.rankingRules[{begin}..={end}]`"),
+            RankingRuleSource::Sort { criterion_index, sort_index } => format!("`queries[{query_index}].sort[{sort_index}]`, `{index_uid}.rankingRules[{criterion_index}]`"),
+        };
+
+        format!("{source}: {kind} {rules}")
+    }
+
+    fn fetch_from_source(
+        &self,
+        canonical_criteria: &[Criterion],
+        canonical_sort: &Option<Vec<AscDesc>>,
+    ) -> String {
+        let rule_name = match self.source {
+            RankingRuleSource::Criterion(index) => {
+                canonical_criteria.get(index).map(|criterion| criterion.to_string())
+            }
+            RankingRuleSource::CoalescedCriteria(begin, end) => {
+                let rules: Vec<String> = canonical_criteria
+                    .get(begin..=end)
+                    .into_iter()
+                    .flat_map(|criteria| criteria.iter())
+                    .map(|criterion| criterion.to_string())
+                    .collect();
+
+                (!rules.is_empty()).then_some(rules.join(", "))
+            }
+            RankingRuleSource::Sort { criterion_index: _, sort_index } => canonical_sort
+                .as_deref()
+                .and_then(|canonical_sort| canonical_sort.get(sort_index))
+                .and_then(|asc_desc: &AscDesc| match asc_desc {
+                    AscDesc::Asc(Member::Field(s)) | AscDesc::Desc(Member::Field(s)) => {
+                        Some(format!("on field `{s}`"))
+                    }
+                    _ => None,
+                }),
+        };
+
+        let rule_name = rule_name.unwrap_or_else(|| "default".into());
+
+        format!("rule(s) {rule_name}")
+    }
+}