Minor issues with bitmap - Mailing list pgsql-hackers

From Greg Sabino Mullane
Subject Minor issues with bitmap
Date
Msg-id 1d9ba2f3ddf8704cb14184fa6ffbeac8@biglumber.com
Whole thread Raw
List pgsql-hackers
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1


I've been playing with the new bitmap stuff, and so far, it seems
to be working fairly well. I have two very minor complaints, however.

First, some of my queries have a large number of "OR" conditions,
and this is where the bitmap gets used a lot. However, the explain
[analyze] plans are unnecessarily verbose: for each item in the
OR list, I get two lines telling me a Bitmap Index Scan is being used
on an index for that item, and a separate line indicating the Index
condition. It would be nice if all these lines could be somehow
combined or reduced, as my explain plans are now regularly
approaching 1000 lines, and the usefulness to me of all but the first
Bitmap Index Scan entry is limited.

Second, I wanted to compare the efficiency of bitmaps directly, by
running the same query with bitmapscan on and off, but when I set
it to "off" it ends up taking a very long time indeed, and is not
the same as running the query in a pre-bitmap (e.g. 8.0) environment.
This probably cannot be "fixed" but it does make the speed comparisons
a little more tricky as I have to end up comparing 8.0 queries with
8.1 queries, instead of 8.1 vs 8.1.

- --
Greg Sabino Mullane greg@turnstep.com
PGP Key: 0x14964AC8 200505211624
http://biglumber.com/x/web?pk=2529DF6AB8F79407E94445B4BC9B906714964AC8

-----BEGIN PGP SIGNATURE-----

iD8DBQFCj5mdvJuQZxSWSsgRAmfoAKD2CkXDVM5RQWuHaGcFAjy/KjaoZwCfcaOD
bgI5mcI12VtU8gPZWSLZ99A=
=XidF
-----END PGP SIGNATURE-----




pgsql-hackers by date:

Previous
From: Heikki Linnakangas
Date:
Subject: Re: Two-phase commit issues
Next
From: Bruce Momjian
Date:
Subject: Re: inet increment w/ int8