From: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
---|---|
To: | Florian Pflug <fgp(at)phlo(dot)org> |
Cc: | Greg Stark <stark(at)mit(dot)edu>, pgsql-hackers(at)postgresql(dot)org, Hitoshi Harada <umi(dot)tanuki(at)gmail(dot)com> |
Subject: | Re: Underspecified window queries in regression tests |
Date: | 2011-10-16 22:14:03 |
Message-ID: | 3161.1318803243@sss.pgh.pa.us |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
Florian Pflug <fgp(at)phlo(dot)org> writes:
> But some frame clauses (row 1 preceding, for example) have an effect despite
> there being no ORDER BY, like here:
Yeah, why did you expect differently? Without ORDER BY, all rows are
peers in the frame ordering, so there's no way for a RANGE spec to
select less than the whole partition. But with ROWS, you can select
less than that.
In general it's not that hard to create nondeterministic window-function
queries, since the SQL standard doesn't require you to specify a unique
ordering for the window function's input rows. Even in RANGE mode there
are plenty of functions that are sensitive to the exact ordering, eg
first_value/last_value. I guess the committee felt that locking this
down would restrict the feature too much.
regards, tom lane
From | Date | Subject | |
---|---|---|---|
Next Message | Florian Pflug | 2011-10-16 22:51:22 | Re: Underspecified window queries in regression tests |
Previous Message | Jeff Davis | 2011-10-16 22:09:13 | Re: Range Types - typo + NULL string constructor |