Re: [HACKERS] failures with tuplesort and ordered set aggregates (due to 5cefbf5a6c44)

2015-03-09 Thread Robert Haas
On Wed, Mar 4, 2015 at 1:55 PM, Peter Geoghegan wrote: > On Wed, Mar 4, 2015 at 8:26 AM, Robert Haas wrote: >> I think we should commit my patch, and if a future patch needs >> sortKeys set in more places, it can make that change itself. There's >> no reason why it's needed with the code as it i

Re: [HACKERS] failures with tuplesort and ordered set aggregates (due to 5cefbf5a6c44)

2015-03-04 Thread Robert Haas
On Tue, Mar 3, 2015 at 7:14 PM, Peter Geoghegan wrote: > My patch actually matches Andrew Gierth's datumsort patch, in that it > also uses this convention, as I believe it should. For that reason, > I'd prefer to make the comment added in November true, rather than > changing the comment...I feel

Re: [HACKERS] failures with tuplesort and ordered set aggregates (due to 5cefbf5a6c44)

2015-03-04 Thread Peter Geoghegan
On Wed, Mar 4, 2015 at 8:26 AM, Robert Haas wrote: > I think we should commit my patch, and if a future patch needs > sortKeys set in more places, it can make that change itself. There's > no reason why it's needed with the code as it is today, and no reason > to let bits of future changes leak i

Re: [HACKERS] failures with tuplesort and ordered set aggregates (due to 5cefbf5a6c44)

2015-03-03 Thread Peter Geoghegan
On Tue, Mar 3, 2015 at 3:53 PM, Robert Haas wrote: > I find your statement that this is a pre-existing issue in > tuplesort_begin_datum() to be pretty misleading, unless what you mean > by it is "pre-existing since November, when an earlier patch by Peter > Geoghegan changed the comments without c

Re: [HACKERS] failures with tuplesort and ordered set aggregates (due to 5cefbf5a6c44)

2015-03-03 Thread Robert Haas
On Fri, Feb 20, 2015 at 4:01 PM, Peter Geoghegan wrote: > On Fri, Feb 20, 2015 at 11:58 AM, Tomas Vondra > wrote: >> This seems to happen because ordered_set_startup() calls >> tuplesort_begin_datum() when (use_tuples == true), which only sets >> 'onlyKey' and leaves (sortKeys == NULL). So 'merge

Re: [HACKERS] failures with tuplesort and ordered set aggregates (due to 5cefbf5a6c44)

2015-02-20 Thread Peter Geoghegan
On Fri, Feb 20, 2015 at 11:58 AM, Tomas Vondra wrote: > This seems to happen because ordered_set_startup() calls > tuplesort_begin_datum() when (use_tuples == true), which only sets > 'onlyKey' and leaves (sortKeys == NULL). So 'mergeruns' fails because it > does not expect that. Oops. You're rig

[HACKERS] failures with tuplesort and ordered set aggregates (due to 5cefbf5a6c44)

2015-02-20 Thread Tomas Vondra
Hi, while reviewing one of the 'abbreviated keys' patches, I noticed that the ordered set aggregates are broken when using tuplesort with multiple runs. ISTM this got broken by 5cefbf5a6c4466ac6b1cc2a4316b4eba9108c802: Don't use abbreviated keys for the final merge pass. When we write t