Merge #4313

4313: Fix document formatting performances r=Kerollmops a=ManyTheFish

reduce the formatted option list to the attributes that should be formatted,
instead of all the attributes to display.
The time to compute the `format` list scales with the number of fields to format;
cumulated with `map_leaf_values` that iterates over all the nested fields, it gives a quadratic complexity:
`d*f` where `d` is the total number of fields to display and `f` is the total number of fields to format.

Co-authored-by: ManyTheFish <many@meilisearch.com>

meilisearch/src/search.rs

@@ -900,6 +900,14 @@ fn format_fields<'a>(
     let mut matches_position = compute_matches.then(BTreeMap::new);
     let mut document = document.clone();
 
+    // reduce the formatted option list to the attributes that should be formatted,
+    // instead of all the attributes to display.
+    let formatting_fields_options: Vec<_> = formatted_options
+        .iter()
+        .filter(|(_, option)| option.should_format())
+        .map(|(fid, option)| (field_ids_map.name(*fid).unwrap(), option))
+        .collect();
+
     // select the attributes to retrieve
     let displayable_names =
         displayable_ids.iter().map(|&fid| field_ids_map.name(fid).expect("Missing field name"));
@@ -908,13 +916,15 @@ fn format_fields<'a>(
         // to the value and merge them together. eg. If a user said he wanted to highlight `doggo`
         // and crop `doggo.name`. `doggo.name` needs to be highlighted + cropped while `doggo.age` is only
         // highlighted.
-        let format = formatted_options
+        // Warn: The time to compute the format list scales with the number of fields to format;
+        // cumulated with map_leaf_values that iterates over all the nested fields, it gives a quadratic complexity:
+        // d*f where d is the total number of fields to display and f is the total number of fields to format.
+        let format = formatting_fields_options
             .iter()
-            .filter(|(field, _option)| {
-                let name = field_ids_map.name(**field).unwrap();
+            .filter(|(name, _option)| {
                 milli::is_faceted_by(name, key) || milli::is_faceted_by(key, name)
             })
-            .map(|(_, option)| *option)
+            .map(|(_, option)| **option)
             .reduce(|acc, option| acc.merge(option));
         let mut infos = Vec::new();
 
@@ -1011,7 +1021,7 @@ fn format_value<'a>(
                     let value = matcher.format(format_options);
                     Value::String(value.into_owned())
                 }
-                None => Value::Number(number),
+                None => Value::String(s),
             }
         }
         value => value,

milli/src/search/new/bucket_sort.rs

@@ -15,6 +15,7 @@ pub struct BucketSortOutput {
 
 // TODO: would probably be good to regroup some of these inside of a struct?
 #[allow(clippy::too_many_arguments)]
+#[logging_timer::time]
 pub fn bucket_sort<'ctx, Q: RankingRuleQueryTrait>(
     ctx: &mut SearchContext<'ctx>,
     mut ranking_rules: Vec<BoxRankingRule<'ctx, Q>>,

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

@@ -72,7 +72,7 @@ impl<'m> MatcherBuilder<'m> {
     }
 }
 
-#[derive(Copy, Clone, Default)]
+#[derive(Copy, Clone, Default, Debug)]
 pub struct FormatOptions {
     pub highlight: bool,
     pub crop: Option<usize>,
@@ -82,6 +82,10 @@ impl FormatOptions {
     pub fn merge(self, other: Self) -> Self {
         Self { highlight: self.highlight || other.highlight, crop: self.crop.or(other.crop) }
     }
+
+    pub fn should_format(&self) -> bool {
+        self.highlight || self.crop.is_some()
+    }
 }
 
 #[derive(Clone, Debug)]

milli/src/search/new/mod.rs

@@ -191,6 +191,7 @@ fn resolve_maximally_reduced_query_graph(
     Ok(docids)
 }
 
+#[logging_timer::time]
 fn resolve_universe(
     ctx: &mut SearchContext,
     initial_universe: &RoaringBitmap,
@@ -556,6 +557,7 @@ pub fn execute_vector_search(
 }
 
 #[allow(clippy::too_many_arguments)]
+#[logging_timer::time]
 pub fn execute_search(
     ctx: &mut SearchContext,
     query: Option<&str>,

milli/src/search/new/query_term/parse_query.rs

@@ -5,6 +5,7 @@ use super::*;
 use crate::{Result, SearchContext, MAX_WORD_LENGTH};
 
 /// Convert the tokenised search query into a list of located query terms.
+#[logging_timer::time]
 pub fn located_query_terms_from_tokens(
     ctx: &mut SearchContext,
     query: NormalizedTokenIter,