Speed up graph based ranking rule when a lot of different costs exist

milli/src/search/new/graph_based_ranking_rule.rs

@@ -309,11 +309,6 @@ impl<'ctx, G: RankingRuleGraphTrait> RankingRule<'ctx, QueryGraph> for GraphBase
                 Ok(ControlFlow::Continue(()))
             }
         })?;
-        // if at_least_one {
-        //     unsafe {
-        //         println!("\n===== {id}  COST: {cost} ====  PATHS: {COUNT_PATHS} ==== NODES: {COUNT_VISITED_NODES} ===== UNIVERSE: {universe}", id=self.id, universe=universe.len());
-        //     }
-        // }
         logger.log_internal_state(graph);
         logger.log_internal_state(&good_paths);
 
@@ -337,8 +332,14 @@ impl<'ctx, G: RankingRuleGraphTrait> RankingRule<'ctx, QueryGraph> for GraphBase
 
         let next_query_graph = QueryGraph::build_from_paths(paths);
 
-        if !nodes_with_removed_outgoing_conditions.is_empty() {
+        #[allow(clippy::comparison_chain)]
-            graph.update_all_costs_before_nodes(&nodes_with_removed_outgoing_conditions, all_costs);
+        if nodes_with_removed_outgoing_conditions.len() == 1 {
+            graph.update_all_costs_before_node(
+                *nodes_with_removed_outgoing_conditions.first().unwrap(),
+                all_costs,
+            );
+        } else if nodes_with_removed_outgoing_conditions.len() > 1 {
+            *all_costs = graph.find_all_costs_to_end();
         }
 
         self.state = Some(state);

milli/src/search/new/query_graph.rs

@@ -8,7 +8,6 @@ use crate::search::new::interner::Interner;
 use crate::Result;
 use fxhash::{FxHashMap, FxHasher};
 use std::cmp::Ordering;
-use std::collections::hash_map::Entry;
 use std::collections::BTreeMap;
 use std::hash::{Hash, Hasher};
 
@@ -364,8 +363,6 @@ impl QueryGraph {
            └──│ b2 │──│ c2 │───│ d  │───│ e2 │
               └────┘  └────┘   └────┘   └────┘
     ```
-    But we accept the first representation as it reduces the size
-    of the graph and shouldn't cause much problems.
     */
     pub fn build_from_paths(
         paths: Vec<Vec<(Option<LocatedQueryTermSubset>, LocatedQueryTermSubset)>>,

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

@@ -1,8 +1,11 @@
 #![allow(clippy::too_many_arguments)]
 
 use std::collections::{BTreeSet, VecDeque};
+use std::iter::FromIterator;
 use std::ops::ControlFlow;
 
+use fxhash::FxHashSet;
+
 use super::{DeadEndsCache, RankingRuleGraph, RankingRuleGraphTrait};
 use crate::search::new::interner::{Interned, MappedInterner};
 use crate::search::new::query_graph::QueryNode;
@@ -112,9 +115,6 @@ impl<G: RankingRuleGraphTrait> VisitorState<G> {
                 }
             }
         }
-        // if there wasn't any valid path from this node to the end node, then
-        // this node is a dead end **for this specific cost**.
-        // we could encode this in the dead-ends cache
 
         Ok(ControlFlow::Continue(any_valid))
     }
@@ -126,11 +126,11 @@ impl<G: RankingRuleGraphTrait> VisitorState<G> {
         visit: VisitFn<G>,
         ctx: &mut VisitorContext<G>,
     ) -> Result<ControlFlow<(), bool>> {
-        if ctx
+        if !ctx
             .all_costs_from_node
             .get(dest_node)
             .iter()
-            .all(|next_cost| *next_cost != self.remaining_cost)
+            .any(|next_cost| *next_cost == self.remaining_cost)
         {
             return Ok(ControlFlow::Continue(false));
         }
@@ -158,14 +158,12 @@ impl<G: RankingRuleGraphTrait> VisitorState<G> {
     ) -> Result<ControlFlow<(), bool>> {
         assert!(dest_node != ctx.graph.query_graph.end_node);
 
-        if self.forbidden_conditions_to_nodes.contains(dest_node)
+        if self.forbidden_conditions.contains(condition)
+            || self.forbidden_conditions_to_nodes.contains(dest_node)
             || edge_new_nodes_to_skip.intersects(&self.visited_nodes)
         {
             return Ok(ControlFlow::Continue(false));
         }
-        if self.forbidden_conditions.contains(condition) {
-            return Ok(ControlFlow::Continue(false));
-        }
 
         // Checking that from the destination node, there is at least
         // one cost that we can visit that corresponds to our remaining budget.
@@ -244,48 +242,41 @@ impl<G: RankingRuleGraphTrait> RankingRuleGraph<G> {
         costs_to_end
     }
 
-    pub fn update_all_costs_before_nodes(
+    pub fn update_all_costs_before_node(
         &self,
-        removed_nodes: &BTreeSet<Interned<QueryNode>>,
+        node_with_removed_outgoing_conditions: Interned<QueryNode>,
         costs: &mut MappedInterner<QueryNode, Vec<u64>>,
     ) {
-        // unsafe {
-        //     FIND_ALL_COSTS_INC_COUNT += 1;
-        //     println!(
-        //         "update_all_costs_after_removing_edge incrementally count: {}",
-        //         FIND_ALL_COSTS_INC_COUNT
-        //     );
-        // }
-
         let mut enqueued = SmallBitmap::new(self.query_graph.nodes.len());
         let mut node_stack = VecDeque::new();
 
-        for node in removed_nodes.iter() {
+        enqueued.insert(node_with_removed_outgoing_conditions);
-            enqueued.insert(*node);
+        node_stack.push_back(node_with_removed_outgoing_conditions);
-            node_stack.push_back(*node);
-        }
 
-        while let Some(cur_node) = node_stack.pop_front() {
+        'main_loop: while let Some(cur_node) = node_stack.pop_front() {
-            let mut self_costs = BTreeSet::<u64>::new();
+            let mut costs_to_remove = FxHashSet::default();
+            for c in costs.get(cur_node) {
+                costs_to_remove.insert(*c);
+            }
 
             let cur_node_edges = &self.edges_of_node.get(cur_node);
             for edge_idx in cur_node_edges.iter() {
                 let edge = self.edges_store.get(edge_idx).as_ref().unwrap();
-                let succ_node = edge.dest_node;
+                for cost in costs.get(edge.dest_node).iter() {
-                let succ_costs = costs.get(succ_node);
+                    costs_to_remove.remove(&(*cost + edge.cost as u64));
-                for succ_distance in succ_costs {
+                    if costs_to_remove.is_empty() {
-                    self_costs.insert(edge.cost as u64 + succ_distance);
+                        continue 'main_loop;
+                    }
                 }
             }
-            let costs_to_end_cur_node = costs.get_mut(cur_node);
+            if costs_to_remove.is_empty() {
-            for cost in self_costs.iter() {
+                continue 'main_loop;
-                costs_to_end_cur_node.push(*cost);
             }
-            let self_costs = self_costs.into_iter().collect::<Vec<_>>();
+            let mut new_costs = BTreeSet::from_iter(costs.get(cur_node).iter().copied());
-            if &self_costs == costs.get(cur_node) {
+            for c in costs_to_remove {
-                continue;
+                new_costs.remove(&c);
             }
-            *costs.get_mut(cur_node) = self_costs;
+            *costs.get_mut(cur_node) = new_costs.into_iter().collect();
 
             for prev_node in self.query_graph.nodes.get(cur_node).predecessors.iter() {
                 if !enqueued.contains(prev_node) {