bug#60953: The :match predicate with large regexp in tree-sitter font-lock seems inefficient

[Dmitry Gutov <dgutov@yandex.ru>]

[-- Attachment #1: Type: text/plain, Size: 596 bytes --]

On 26/01/2023 20:07, Dmitry Gutov wrote:
> One could hope to avoid recreating the list of predicates on every 
> match, but that seems to be a limitation of the TS API: 
> ts_query_predicates_for_pattern requires a second argument, 
> match.pattern_index. Maybe we could memoize that, though?

Speaking of memoization, here is a POC patch.

It's a definite improvement: with the attached :match almost reaches the 
performance of :pred. Not sure why it's still not faster, though.

(I also tried a more comprehensive memoization using a hash table, the 
resulting performance was slightly worse.)

[-- Attachment #2: memoize_simple.diff --]
[-- Type: text/x-patch, Size: 1162 bytes --]

diff --git a/src/treesit.c b/src/treesit.c
index 917db582676..69f54976509 100644
--- a/src/treesit.c
+++ b/src/treesit.c
@@ -2722,6 +2722,7 @@ DEFUN ("treesit-query-capture",
      bottleneck (98.4% of the running time spent on nconc).  */
   Lisp_Object result = Qnil;
   Lisp_Object prev_result = result;
+  Lisp_Object predicates_for_0 = NULL;
   while (ts_query_cursor_next_match (cursor, &match))
     {
       /* Record the checkpoint that we may roll back to.  */
@@ -2750,9 +2751,18 @@ DEFUN ("treesit-query-capture",
 	  result = Fcons (cap, result);
 	}
       /* Get predicates.  */
-      Lisp_Object predicates
-	= treesit_predicates_for_pattern (treesit_query,
-					  match.pattern_index);
+      Lisp_Object predicates;
+      if (match.pattern_index == 0)
+	{
+	  if (predicates_for_0 == NULL)
+	    predicates_for_0 = treesit_predicates_for_pattern (treesit_query, 0);
+
+	  predicates = predicates_for_0;
+	}
+      else
+	{
+	  predicates = treesit_predicates_for_pattern (treesit_query, match.pattern_index);
+	}
 
       /* captures_lisp = Fnreverse (captures_lisp); */
       struct capture_range captures_range = { result, prev_result };