> I understand that COUNT queries are expensive. So I'm looking for advice on
> displaying paginated query results.
>
> I display my query results like this:
>
> Displaying 1 to 50 of 2905.
> 1-50 | 51-100 | 101-150 | etc.
>
> I do this by executing two queries. One is of the form:
>
> SELECT <select list> FROM <view/table list> WHERE <filter> LIMIT m OFFSET n
>
> The other is identical except that I replace the select list with COUNT(*).
>
> I'm looking for suggestions to replace that COUNT query. I cannot use the
> method of storing the number of records in a separate table because my queries
> (a) involve joins, and (b) have a WHERE clause.
Well, on all my sites, I do what you do and just live with it :P You
can investigate using cursors however (DECLARE, MOVE & FETCH)
> And an unrelated question:
> I'm running PG 7.2.2 and want to upgrade to 7.4.1. I've never upgraded PG
> before and I'm nervous. Can I simply run pg_dumpall, install 7.4.1, and then
> feed the dump into psql? I'm planning to use pg_dumpall rather than pg_dump
> because I want to preserve the users I've defined. My database is the only one
> on the system.
I recommend something like this:
-- disable access to your database to make sure you have a complete dump
-- run dump as database owner account
su pgsql (or whatever your postgres user is)
-- do compressed dump
pg_dumpall > backup.sql
-- backup old data dir
mv /usr/local/pgsql/data /usr/local/pgsql/data.7.2
-- remove old postgres, install new
-- run NEW initdb. replace latin1 with your encoding
-- -W specifies a superuser password
initdb -D /usr/local/pgsql/data -E LATIN1 -W
-- restore dump, watching output VERY CAREFULLY:
-- (run as pgsql user again)
psql template1 < backup.sql > log.txt
-- Watch stderr very carefully to check any errors that might occur.
-- If restore fails, re-initdb and re-restore
Chris
