Fix similar_escape() so that SIMILAR TO works properly for patterns involving

alternatives ("|" symbol).  The original coding allowed the added ^ and $
constraints to be absorbed into the first and last alternatives, producing
a pattern that would match more than it should.  Per report from Eric Noriega.

I also changed the pattern to add an ARE director ("***:"), ensuring that
SIMILAR TO patterns do not change behavior if regex_flavor is changed.  This
is necessary to make the non-capturing parentheses work, and seems like a
good idea on general principles.

Back-patched as far as 7.4.  7.3 also has the bug, but a fix seems impractical
because that version's regex engine doesn't have non-capturing parens.

diff --git a/src/backend/utils/adt/regexp.c b/src/backend/utils/adt/regexp.c
index d8b038a09a8e7079655c8a1951fd4fe6f0bbff43..cff81b7475ea940ff3ea670d93aed854aa41d468 100644 (file)
--- a/src/backend/utils/adt/regexp.c
+++ b/src/backend/utils/adt/regexp.c
@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *       $PostgreSQL: pgsql/src/backend/utils/adt/regexp.c,v 1.56 2004/12/31 22:01:22 pgsql Exp $
+ *       $PostgreSQL: pgsql/src/backend/utils/adt/regexp.c,v 1.56.4.1 2006/04/13 18:01:45 tgl Exp $
  *
  *             Alistair Crooks added the code for the regex caching
  *             agc - cached the regular expressions used - there's a good chance
@@ -481,11 +481,36 @@ similar_escape(PG_FUNCTION_ARGS)
                          errhint("Escape string must be empty or one character.")));
        }
 
-       /* We need room for ^, $, and up to 2 output bytes per input byte */
-       result = (text *) palloc(VARHDRSZ + 2 + 2 * plen);
+       /*----------
+        * We surround the transformed input string with
+        *                      ***:^(?: ... )$
+        * which is bizarre enough to require some explanation.  "***:" is a
+        * director prefix to force the regex to be treated as an ARE regardless
+        * of the current regex_flavor setting.  We need "^" and "$" to force
+        * the pattern to match the entire input string as per SQL99 spec.  The
+        * "(?:" and ")" are a non-capturing set of parens; we have to have
+        * parens in case the string contains "|", else the "^" and "$" will
+        * be bound into the first and last alternatives which is not what we
+        * want, and the parens must be non capturing because we don't want them
+        * to count when selecting output for SUBSTRING.
+        *----------
+        */
+
+       /*
+        * We need room for the prefix/postfix plus as many as 2 output bytes per
+        * input byte
+        */
+       result = (text *) palloc(VARHDRSZ + 10 + 2 * plen);
        r = VARDATA(result);
 
+       *r++ = '*';
+       *r++ = '*';
+       *r++ = '*';
+       *r++ = ':';
        *r++ = '^';
+       *r++ = '(';
+       *r++ = '?';
+       *r++ = ':';
 
        while (plen > 0)
        {
@@ -525,6 +550,7 @@ similar_escape(PG_FUNCTION_ARGS)
                p++, plen--;
        }
 
+       *r++ = ')';
        *r++ = '$';
 
        VARATT_SIZEP(result) = r - ((unsigned char *) result);