| From: | Gregory Stark <stark(at)enterprisedb(dot)com> |
|---|---|
| To: | pgsql-patches <pgsql-patches(at)postgresql(dot)org> |
| Subject: | Updated packed varlena patch |
| Date: | 2007-03-09 15:18:24 |
| Message-ID: | 877itq4f9b.fsf@stark.xeocode.com |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-patches |
I added a little-endian special-case using the low order bits as described by
Tom.
Also, I noticed that int2vector and oidvector are already flagged in pg_type
with attstorage == 'p'. Really any type with attstorage == 'p' should be
exempted from packed varlena handling since it isn't expecting to be toasted.
This lets me clean up heaptuple.c a bit so now it doesn't depend on pg_type.h
and doesn't have any special case exemption for oidvector and int2vector. It
also means any other type can opt out from packed varlena packing by setting
attstorage. I've left inet and cidr with toast support and changed them to
have attstorage 'm' instead.
http://community.enterprisedb.com/varlena/patch-varvarlena-17.patch.gz
--
Gregory Stark
EnterpriseDB http://www.enterprisedb.com
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Gregory Stark | 2007-03-09 17:37:55 | non-recursive WITH clause support |
| Previous Message | Mark Stosberg | 2007-03-09 14:31:27 | Re: suggestion for improving TMPDIR and "--format" docs for pg_dump |