Thanks Tom; this is at a client site, so I have limited access, but it looks like a REINDEX resolves the issue.
Doug
-----Original Message-----
From: Tom Lane [mailto:tgl@sss.pgh.pa.us]
Sent: March 16, 2012 6:33 AM
To: Doug Gorley
Cc: pgsql-general@postgresql.org
Subject: Re: [GENERAL] Zero-length character breaking query?
Doug Gorley <dgorley@aihs.ca> writes:
> The table is called tdt_unsent. The field is str_name_l. For demonstration purposes, the value is "SMITH".
> "select * from tdt_unsent where str_name_l = 'SMITH'" returns 0 rows.
> "select * from tdt_unsent where str_name_l ~ '^SMITH'" returns 3 rows.
> "select * from tdt_unsent where str_name_l ~ '^SMITH$'" returns 0 rows.
> "select length(str_name_l) from tdt_unsent where str_name_l ~ '^SMITH'" returns "5".
I'd check EXPLAIN (with the actual problematic string, not SMITH).
The planner is probably trying to build an index range condition from the regex pattern --- is it doing the right thing
givenyour locale?
If the plan looks okay, maybe you need to reindex whatever index it's using.
regards, tom lane