On 9/19/07, Andrew Dunstan <[EMAIL PROTECTED]> wrote: > It's at least good to see that the LIKE case has some useful speedup in > 8.3.
It can be due to your patch or to the varlena header patch. Seqscan is a bit faster too. > Can you run the same set of tests in a single byte encoding like latin1? As discussed on IRC, I'm loading the data in a LATIN1 database for 8.1, 8.2 and 8.3. I'll let you know when I have the results. > We might have to look at doing on-demand lowering, but in a case like > yours it looks like we'd still end up lowering almost every character > anyway, so I'm not quite sure what to do. Note that the 8.2 change was a > bug fix, so we can't just revert it. Maybe we need to look closely at > the efficiency of lower(). Yes, I know it's a bug fix but the performance decrease is far from being negligible in our case. -- Guillaume ---------------------------(end of broadcast)--------------------------- TIP 9: In versions below 8.0, the planner will ignore your desire to choose an index scan if your joining column's datatypes do not match