| From: | Bruce Momjian <bruce(at)momjian(dot)us> |
|---|---|
| To: | PostgreSQL-development <pgsql-hackers(at)postgreSQL(dot)org> |
| Subject: | pg_upgrade fixes, #99 ;-) |
| Date: | 2011-01-08 18:53:15 |
| Message-ID: | 201101081853.p08IrGX16678@momjian.us |
| Views: | Whole Thread | Raw Message | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
In my first attached, applied patch, I have found a way to speed
relations lookups in pg_upgrade. I knew there was a way to optimize
this but it was not clear until my major cleanups. Instead of doing
effectively a nested loop join on old/new relations, I now order them
and use a 1:1 mergejoin. This should speed up pg_upgrade for many
relations.
The second patch removes a hack for toast relations that is unnecessary
now that we always preserve pg_class.oid. The old code preserved
relfilenodes for non-toast tables and oids for toast relations, which
was obviously confusing and non-optimal.
Remember, this code was not originally written by me but someone at
EnteprriseDB, so I didn't fully understand it until now.
I love removing functions!
--
Bruce Momjian <bruce(at)momjian(dot)us> http://momjian.us
EnterpriseDB http://enterprisedb.com
+ It's impossible for everything to be true. +
| Attachment | Content-Type | Size |
|---|---|---|
| unknown_filename | text/plain | 12.9 KB |
| unknown_filename | text/plain | 2.4 KB |
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Jeff Davis | 2011-01-08 20:12:42 | Re: WIP: Range Types |
| Previous Message | Joel Jacobson | 2011-01-08 17:41:11 | Re: obj_unique_identifier(oid) |