Thanks for the comments on what you ended up changing. It helps point out the kind of things I should be looking for. I'll try to let less slip through in the future.
Mike
__________________________________________________________________________________ Mike Blackwell | Technical Analyst, Distribution Services/Rollout Management | RR Donnelley 1750 Wallace Ave | St Charles, IL 60174-3401 Office: 630.313.7818 Mike.Blackwell@rrd.com http://www.rrdonnelley.com
On Mon, Jan 19, 2015 at 10:09 AM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
"Timmer, Marius" <marius.timmer@uni-muenster.de> writes: > We think, you wanted to switch to DESC behavior > (print out NULLS FIRST) in cases, where > „USING“ uses an operator which is considered to be > a DESC operator.
Right, because that's how addTargetToSortList() would parse it. > But get_equality_op_for_ordering_op is called in > cases, where reverse is false, but > the part > if (reverse) > *reverse = (strategy == BTGreaterStrategyNumber); > never changes this to true?
Sorry, not following? It's true that what I added to explain.c doesn't worry too much about the possibility of get_ordering_op_properties() failing --- that really shouldn't happen for something that was previously accepted as a sorting operator. But if it does, "reverse" will just be left as false, so the behavior will anyway be unsurprising I think. We could alternatively make it throw a "cache lookup failed" error but I'm not sure how that makes anyone's life better.