Thread

1. psort fixed

   Bruce Momjian <maillist@candle.pha.pa.us> — 1998-02-01T22:41:45Z

   I have fixed the regression test problem.  The original psort() NULL
   handling patch is attached.  It made the while() into a for() which is
   good, but he did not remove the increment at the bottom or the for(), so
   nkey was being incremented twice in the loop.  Fixed it, and now
   regression looks good to me.
   
   I have updated the files in tools/RELEASE_CHANGES to say 6.3, so I think
   we are ready for some announcement on 6.3 beta, Marc.  I recommend you
   attach the HISTORY file for the release, so people know what is new.
   
   
   ---------------------------------------------------------------------------
   
   Index: psort.c
   ===================================================================
   RCS file: /usr/local/cvsroot/pgsql/src/backend/utils/sort/psort.c,v
   retrieving revision 1.33
   retrieving revision 1.34
   diff -c -r1.33 -r1.34
   *** psort.c	1998/01/25 05:14:49	1.33
   --- psort.c	1998/01/25 05:18:34	1.34
   ***************
   *** 7,13 ****
      *
      *
      * IDENTIFICATION
   !  *	  $Header: /usr/local/cvsroot/pgsql/src/backend/utils/sort/psort.c,v 1.33 1998/01/25 05:14:49 momjian Exp $
      *
      * NOTES
      *		Sorts the first relation into the second relation.
   --- 7,13 ----
      *
      *
      * IDENTIFICATION
   !  *	  $Header: /usr/local/cvsroot/pgsql/src/backend/utils/sort/psort.c,v 1.34 1998/01/25 05:18:34 scrappy Exp $
      *
      * NOTES
      *		Sorts the first relation into the second relation.
   ***************
   *** 1094,1100 ****
         int		result = 0;
         bool	isnull1, isnull2;
         
   !     while ( nkey < PsortNkeys && !result )
         {
     		lattr = heap_getattr(*ltup, InvalidBuffer,
     				     PsortKeys[nkey].sk_attno, 
   --- 1094,1100 ----
         int		result = 0;
         bool	isnull1, isnull2;
         
   !     for (nkey = 0; nkey < PsortNkeys && !result; nkey++ )
         {
     		lattr = heap_getattr(*ltup, InvalidBuffer,
     				     PsortKeys[nkey].sk_attno, 
   ***************
   *** 1106,1119 ****
     				     &isnull2);
     		if ( isnull1 )
     		{
   ! 	    	if ( isnull2 )
   ! 	    		return (0);
   ! 		    return(1);
     		}
     		else if ( isnull2 )
   ! 		    return (-1);
     		
   ! 		if (PsortKeys[nkey].sk_flags & SK_COMMUTE)
     		{
     	    	if (!(result = -(long) (*fmgr_faddr(&PsortKeys[nkey].sk_func)) (rattr, lattr)))
     			result = (long) (*fmgr_faddr(&PsortKeys[nkey].sk_func)) (lattr, rattr);
   --- 1106,1118 ----
     				     &isnull2);
     		if ( isnull1 )
     		{
   ! 			if ( !isnull2 )
   ! 				result = 1;
     		}
     		else if ( isnull2 )
   ! 		    result = -1;
     		
   ! 		else if (PsortKeys[nkey].sk_flags & SK_COMMUTE)
     		{
     	    	if (!(result = -(long) (*fmgr_faddr(&PsortKeys[nkey].sk_func)) (rattr, lattr)))
     			result = (long) (*fmgr_faddr(&PsortKeys[nkey].sk_func)) (lattr, rattr);
   
   -- 
   Bruce Momjian
   maillist@candle.pha.pa.us
   
   

2. Re: [HACKERS] psort fixed

   Thomas Lockhart <lockhart@alumni.caltech.edu> — 1998-02-02T01:43:02Z

   > I have fixed the regression test problem.  The original psort() NULL
   > handling patch is attached.  It made the while() into a for() which is
   > good, but he did not remove the increment at the bottom or the for(), so
   > nkey was being incremented twice in the loop.  Fixed it, and now
   > regression looks good to me.
   
   We be stylin' now... Nice.
   
                                               - Tom