Coalesce sequences of regex literals

When we have `/st/`, the compiler frontend currently produces two RakuAST::Regex::Literal nodes. These were individually compiled also, which breaks matching against the ligature ﬆ (with the appropriate options set). Coalesce these as part of the compilation from RakuAST to QAST, meaning that the same job doesn't have to be done by all those synthetically constructing RakuAST in order to get correct matching.
rakudo · Jun 7, 2021 · f7244fc · f7244fc
1 parent 203a83f
commit f7244fc
Showing 1 changed file with 23 additions and 4 deletions.
diff --git a/src/Raku/ast/regex.rakumod b/src/Raku/ast/regex.rakumod
@@ -131,16 +131,35 @@ class RakuAST::Regex::Sequence is RakuAST::Regex {
 
     method IMPL-REGEX-QAST(RakuAST::IMPL::QASTContext $context, %mods) {
         my $concat := QAST::Regex.new(:rxtype<concat>);
+        my str $collect-literals := '';
         for $!terms {
-            my $qast := $_.IMPL-REGEX-QAST($context, %mods);
-            if $qast {
-                $qast.backtrack('r') if %mods<r> && !$qast.backtrack;
-                $concat.push($qast);
+            if nqp::istype($_, RakuAST::Regex::Literal) {
+                $collect-literals := $collect-literals ~ $_.text;
             }
+            else {
+                if $collect-literals ne '' {
+                    $concat.push(self.IMPL-LIT($collect-literals, %mods));
+                    $collect-literals := '';
+                }
+                my $qast := $_.IMPL-REGEX-QAST($context, %mods);
+                if $qast {
+                    $qast.backtrack('r') if %mods<r> && !$qast.backtrack;
+                    $concat.push($qast);
+                }
+            }
+        }
+        if $collect-literals ne '' {
+            $concat.push(self.IMPL-LIT($collect-literals, %mods));
         }
         $concat
     }
 
+    method IMPL-LIT(str $text, %mods) {
+        self.IMPL-APPLY-LITERAL-MODS:
+            QAST::Regex.new( :rxtype<literal>, $text ),
+            %mods
+    }
+
     method visit-children(Code $visitor) {
         for $!terms {
             $visitor($_);