David, Vladimir,
2008/11/5 Hitoshi Harada <umi.tanuki@gmail.com>:
> Now my assumption is broken. Let me take time to think about how to solve it...
I fixed the points we discussed a few days ago. The delta is here:
http://git.postgresql.org/?p=~davidfetter/window_functions/.git;a=commitdiff;h=bfbc039f68aa749b403c84a803d49a02b3d6c199;hp=bbba638f721a7e1d11cb3ee6af3bc1d7d3c11aa8
although attached is the whole (split) patch.
In addition to fixing cume_dist() and implicit frame definition, I
added two window function APIs, WinRowIsPeer() and WinIterFinish(). Up
to now I've avoided to touch ORDER BY clause comparisons deeply,
because I didn't see any abstraction in that except rank(). But now I
know the very important word "peers" appears so many times in the
spec, I'm inclined to implement some general mechanisms for those APIs
like IsPeer(). Also, as with this version part of RANGE is supported,
the road to the FRAME clause support got shorter than before.
Thanks for your feedback and continuing tests!
Regards,
--
Hitoshi Harada