[EMAIL PROTECTED] (Tom Lane) writes: > Chris Browne <[EMAIL PROTECTED]> writes: >> [EMAIL PROTECTED] (Bruce Momjian) writes: >>> I have seen no one do peroformance testing of this, so it seems it >>> will have to wait for 8.4. > >> I didn't have time... > >> (e.g. - we've got a case where dropping the threshold to ~900 bytes >> would give us a big win for certain databases and tables.) > > How do you know? Seems like you've got a readymade test case there.
I did some testing with Known Scenario, and found, indeed, that there was a significant gain to be had. I documented it at least partially on March 21... <http://www.nabble.com/Re%3A-TOASTing-smaller-things-p9602766.html> Unfortunately, the sample query that I used to validate usefulness isn't one I can share :-(. More importantly, it's only one test case, and is strongly influenced by some *very* strong regularity to the patterns of updates that take place to the table that I looked at. It's not nearly good enough to treat as a generalizable case. -- output = ("cbbrowne" "@" "linuxfinances.info") http://linuxdatabases.info/info/spiritual.html MICROS~1 is to quality software what MacDonalds is to gourmet cooking ---------------------------(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
