[Issue 1394] New: Regex does not support forgetfull matching (?:.......

[d-bugmail at puremagic.com]

http://d.puremagic.com/issues/show_bug.cgi?id=1394

           Summary: Regex does not support forgetfull matching (?:.......
           Product: D
           Version: 1.021
          Platform: All
        OS/Version: All
            Status: NEW
          Severity: normal
          Priority: P2
         Component: Phobos
        AssignedTo: bugzilla at digitalmars.com
        ReportedBy: alan at akbkhome.com


Excuse the slight CS issues - but this fixes it.

testcase: 
auto g = new
Regex("/^(?:br|frame|hr|img|input|link|meta|range|spacer|wbr|area|param|col)$/i;");


--- dmd/src/phobos/std/regexp.d 2007-02-27 18:45:28.000000000 +0800
+++ /usr/src/dmd/src/phobos/std/regexp.d        2007-08-02 12:47:16.000000000
+0800
@@ -747,6 +747,7 @@
     REbol,             // beginning of line
     REeol,             // end of line
     REparen,           // parenthesized subexpression
+    REparenforget,     // forgetfull parenthesized subexpression (?:
     REgoto,            // goto offset

     REwordboundary,
@@ -1372,11 +1397,13 @@
                break;

            case REparen:
+           case REparenforget:
+
                // len, n, ()
                puint = cast(uint *)&prog[pc + 1];
                len = puint[0];
                n = puint[1];
-               printf("\tREparen len=%d n=%d, pc=>%d\n", len, n, pc + 1 +
uint.sizeof * 2 + len);
+               printf("\t%s len=%d n=%d, pc=>%d\n", (prog[pc] == REparen ?
"REparen" : "REparenforget") , len, n, pc + 1 + uint.sizeof * 2 + len);
                pc += 1 + uint.sizeof * 2;
                break;

@@ -1835,6 +1862,7 @@
                pc = pop + len;
                break;

+           case REparenforget:
            case REparen:
                // len, ()
                debug(regexp) printf("\tREparen\n");
@@ -1845,8 +1873,10 @@
                ss = src;
                if (!trymatch(pop, pop + len))
                    goto Lnomatch;
+               if (program[pc] == REparen) { 
                pmatch[n + 1].rm_so = ss;
                pmatch[n + 1].rm_eo = src;
+               }
                pc = pop + len;
                break;
@@ -2150,16 +2180,25 @@

            case '(':
                p++;
+
+               if (
+                       ((p + 2) < pattern.length) &&
+                       (pattern[p] == '?') && 
+                       (pattern[p+1] == ':')
+               ) {
+                       p+=2;
+                       buf.write(REparenforget);
+               } else {
                buf.write(REparen);
+               }
                offset = buf.offset;
                buf.write(cast(uint)0);         // reserve space for length
-               buf.write(re_nsub);
+               buf.write(re_nsub);             // match number..
                re_nsub++;
-               parseRegexp();
+               parseRegexp();                  // get all the sub pieces...
                *cast(uint *)&buf.data[offset] =
                    buf.offset - (offset + uint.sizeof * 2);
-               if (p == pattern.length || pattern[p] != ')')
-               {
+               if (p == pattern.length || pattern[p] != ')') {
                    error("')' expected");
                    return 0;
                }
@@ -2738,6 +2777,7 @@
            case REnm:
            case REnmq:
            case REparen:
+           case REparenforget:
            case REgoto:
            {
                auto bitbuf = new OutBuffer;
@@ -2880,6 +2920,7 @@
                i += 1 + uint.sizeof * 3 + len;
                break;

+           case REparenforget:
            case REparen:
                // len, ()
                len = (cast(uint *)&prog[i + 1])[0];


--