Re: Windowing Function Patch Review -> Standard Conformance

Hitoshi Harada wrote:
> I'm glad to hear that. Actually thanks to git it is quite easy for me
> to merge my own repository with the HEAD. It tells me which lines are
> new coming and which lines I modified are newer than else in CVS. This
> is my first project where I use git (and I am not guru of cvs either
> -:) but I love it now.
> So feel free to tell me to recreate a new patch against head. Not so
> heavy work as making on-the-fly script.
>

CUME_DIST and COUNT(*) seem to be working as expected now with the original
tests. I'll continue to test more and also test the other windowing
functions that I've not got to yet.

I did receive a small warning from the compiler when compiling nodeWindow.c.
It was complaining about a possible un-initialised f_shrinking. Just to keep
it quiet I changed the break's to return's after the elog() call.

*** nodeWindow.c 2008-11-09 10:54:50.000000000 +0000
--- nodeWindow.c 2008-11-09 10:39:24.000000000 +0000
***************
*** 818,824 ****
case FRAME_CURRENT_RANGE:
/* UNSUPPORTED */
elog(ERROR, "unknown preceding type %d",
node->preceding_type);
! break;
case FRAME_VALUE_ROWS:
if (node->preceding_rows > 0)
f_shrinking = 0;
--- 818,824 ----
case FRAME_CURRENT_RANGE:
/* UNSUPPORTED */
elog(ERROR, "unknown preceding type %d",
node->preceding_type);
! return; /* keep compiler quiet */
case FRAME_VALUE_ROWS:
if (node->preceding_rows > 0)
f_shrinking = 0;
***************
*** 922,928 ****
case FRAME_CURRENT_RANGE:
/* UNSUPPORTED */
elog(ERROR, "unknown preceding type %d",
node->preceding_type);
! break;
case FRAME_VALUE_ROWS:
if (node->preceding_rows <=
winobj->p_currentpos + 1)
f_shrinking = 1;
--- 922,928 ----
case FRAME_CURRENT_RANGE:
/* UNSUPPORTED */
elog(ERROR, "unknown preceding type %d",
node->preceding_type);
! return; /* keep compiler quiet */
case FRAME_VALUE_ROWS:
if (node->preceding_rows <=
winobj->p_currentpos + 1)
f_shrinking = 1;