From: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
---|---|
To: | Jan Wieck <JanWieck(at)yahoo(dot)com> |
Cc: | The Hermit Hacker <scrappy(at)hub(dot)org>, Bruce Momjian <pgman(at)candle(dot)pha(dot)pa(dot)us>, PostgreSQL HACKERS <pgsql-hackers(at)postgresql(dot)org> |
Subject: | Re: Applying TOAST to CURRENT |
Date: | 2000-05-31 03:53:18 |
Message-ID: | 4463.959745198@sss.pgh.pa.us |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
JanWieck(at)t-online(dot)de (Jan Wieck) writes:
> Hmmm - thinking about that it doesn't sound bad if we allways
> create a secondary relation at CREATE TABLE time, but NOT the
> index for it. And at VACUUM time we create the index if it
> doesn't exist AND there is external stored data.
Don't much like that --- what if the user doesn't run vacuum for
a good long while? Could be doing a lot of sequential scans over
a pretty large toast file...
If the 16K for an empty btree index really bothers you, let's
attack that head-on. I don't see why a freshly created index
couldn't be zero bytes, and the metadata page gets created on
first store into the index.
> The toaster needs to use sequential scans on the external
> table until the next vacuum run, but index usage allways
> depends on vacuum so that's not a real issue from my PoV.
What makes you say that? Indexes will be used on a never-vacuumed
table with the current planner.
regards, tom lane
From | Date | Subject | |
---|---|---|---|
Next Message | Andrew McMillan | 2000-05-31 04:01:15 | Re: Using BOOL in indexes |
Previous Message | Hiroshi Inoue | 2000-05-31 03:04:41 | RE: Using BOOL in indexes |