Clamp the output of estimate_hash_bucketsize() to a sane range;

per example from Bruno Wolff in which it produced a silly result.

diff --git a/src/backend/optimizer/path/costsize.c b/src/backend/optimizer/path/costsize.c
index 5feaed49158695e1c2224f13af563842353781c9..a19dd92c826bd2cff177cdea72a63079fa998ecc 100644 (file)
--- a/src/backend/optimizer/path/costsize.c
+++ b/src/backend/optimizer/path/costsize.c
@@ -42,7 +42,7 @@
  * Portions Copyright (c) 1994, Regents of the University of California
  *
  * IDENTIFICATION
- *       $Header: /cvsroot/pgsql/src/backend/optimizer/path/costsize.c,v 1.96 2002/12/14 00:17:55 tgl Exp $
+ *       $Header: /cvsroot/pgsql/src/backend/optimizer/path/costsize.c,v 1.97 2002/12/26 23:38:42 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -1164,6 +1164,16 @@ estimate_hash_bucketsize(Query *root, Var *var)
        if (avgfreq > 0.0 && mcvfreq > avgfreq)
                estfract *= mcvfreq / avgfreq;
 
+       /*
+        * Clamp bucketsize to sane range (the above adjustment could easily
+        * produce an out-of-range result).  We set the lower bound a little
+        * above zero, since zero isn't a very sane result.
+        */
+       if (estfract < 1.0e-6)
+               estfract = 1.0e-6;
+       else if (estfract > 1.0)
+               estfract = 1.0;
+
        ReleaseSysCache(tuple);
 
        return (Selectivity) estfract;