assume all annotation can be matched in regex search without a prefix

Author: thomaskrause

CHANGELOG.md

@@ -27,6 +27,8 @@ exist.
 
 - Less frequent corpus cache status updates in log. Before, every corpus access
 could trigger an entry into the log which is not desired under heavy load.
+- Improve query execution planning by assuming all annotations can be matched in
+regular expressions without a prefix.
 
 ## [3.7.1] - 2025-04-14
 

core/src/annostorage/inmemory.rs

@@ -752,6 +752,10 @@ where
     fn guess_max_count_regex(&self, ns: Option<&str>, name: &str, pattern: &str) -> Result<usize> {
         let full_match_pattern = util::regex_full_match(pattern);
 
+        // Get the total number of annotations with the namespace/name. We
+        // can't get larger than this number
+        let total = self.number_of_annotations_by_name(ns, name)?;
+
         // Try to parse the regular expression
         let parsed = regex_syntax::Parser::new().parse(&full_match_pattern);
         if let Ok(parsed) = parsed {
@@ -770,11 +774,10 @@ where
                         guessed_count += self.guess_max_count(ns, name, lower_val, &upper_val)?;
                     }
                 }
+            } else {
+                guessed_count = total;
             }
 
-            // Get the total number of annotations with the namespace/name. We
-            // can't get larger than this number
-            let total = self.number_of_annotations_by_name(ns, name)?;
             Ok(guessed_count.min(total))
         } else {
             Ok(0)

core/src/annostorage/ondisk.rs

@@ -895,6 +895,10 @@ where
     fn guess_max_count_regex(&self, ns: Option<&str>, name: &str, pattern: &str) -> Result<usize> {
         let full_match_pattern = util::regex_full_match(pattern);
 
+        // Get the total number of annotations with the namespace/name. We
+        // can't get larger than this number
+        let total = self.number_of_annotations_by_name(ns, name)?;
+
         // Try to parse the regular expression
         let parsed = Parser::new().parse(&full_match_pattern);
         if let Ok(parsed) = parsed {
@@ -913,11 +917,10 @@ where
                         guessed_count += self.guess_max_count(ns, name, lower_val, &upper_val)?;
                     }
                 }
+            } else {
+                guessed_count = total;
             }
 
-            // Get the total number of annotations with the namespace/name. We
-            // can't get larger than this number
-            let total = self.number_of_annotations_by_name(ns, name)?;
             Ok(guessed_count.min(total))
         } else {
             Ok(0)

graphannis/src/annis/db/aql/conjunction.rs

@@ -119,6 +119,7 @@ fn get_cost_estimates<'a>(
     }
 }
 
+/// Returns true if it is estimated to switch the operands in a join.
 fn should_switch_operand_order(
     op_spec: &BinaryOperatorSpecEntry,
     node2cost: &BTreeMap<usize, CostEstimate>,

graphannis/src/annis/db/exec/mod.rs

@@ -16,8 +16,7 @@ pub struct CostEstimate {
     pub output: usize,
     /// Sum of all processed tuples including the ones of the sub-steps.
     pub intermediate_sum: usize,
-    /// The estimated number of tuples that are processed in a join in this
-    /// execution step.
+    /// Simplistic estimated number of tuples that are processed in a join.
     pub processed_in_step: usize,
 }
 
@@ -58,13 +57,13 @@ impl ExecutionNodeDesc {
     pub fn empty_with_fragment(
         node_nr: usize,
         query_fragment: String,
-        est_size: Option<usize>,
+        estimated_output: usize,
     ) -> ExecutionNodeDesc {
         let mut node_pos = BTreeMap::new();
         node_pos.insert(node_nr, 0);
 
-        let cost = est_size.map(|output| CostEstimate {
-            output,
+        let cost = Some(CostEstimate {
+            output: estimated_output,
             intermediate_sum: 0,
             processed_in_step: 0,
         });

graphannis/src/annis/db/exec/nodesearch.rs

@@ -569,7 +569,7 @@ impl<'a> NodeSearch<'a> {
                     desc: Some(ExecutionNodeDesc::empty_with_fragment(
                         common_args.node_nr,
                         common_args.query_fragment,
-                        Some(est_output),
+                        est_output,
                     )),
                     node_search_desc: Arc::new(NodeSearchDesc {
                         qname: (
@@ -676,7 +676,7 @@ impl<'a> NodeSearch<'a> {
             desc: Some(ExecutionNodeDesc::empty_with_fragment(
                 common_args.node_nr,
                 common_args.query_fragment.clone(),
-                Some(est_output),
+                est_output,
             )),
             node_search_desc: Arc::new(NodeSearchDesc {
                 qname: (qname.0, Some(qname.1)),
@@ -772,7 +772,7 @@ impl<'a> NodeSearch<'a> {
             desc: Some(ExecutionNodeDesc::empty_with_fragment(
                 common_args.node_nr,
                 common_args.query_fragment,
-                Some(est_output),
+                est_output,
             )),
             node_search_desc: Arc::new(NodeSearchDesc {
                 qname: (qname.0, Some(qname.1)),
@@ -920,7 +920,7 @@ impl<'a> NodeSearch<'a> {
             desc: Some(ExecutionNodeDesc::empty_with_fragment(
                 common_args.node_nr,
                 common_args.query_fragment.clone(),
-                Some(est_output),
+                est_output,
             )),
             node_search_desc: Arc::new(NodeSearchDesc {
                 qname: (
@@ -978,7 +978,7 @@ impl<'a> NodeSearch<'a> {
             desc: Some(ExecutionNodeDesc::empty_with_fragment(
                 common_args.node_nr,
                 common_args.query_fragment.clone(),
-                Some(est_output),
+                est_output,
             )),
             node_search_desc: Arc::new(NodeSearchDesc {
                 qname: (