Horton

Re: unaccent(text) fails depending on search_path (WAS: pg_upgrade fails saying function unaccent(text) doesn't exist)

Tom Lane <tgl@sss.pgh.pa.us>

View thread View on postgr.es
From: Tom Lane <tgl@sss.pgh.pa.us>
To: Bruce Momjian <bruce@momjian.us>
Cc: Gunnlaugur Thor Briem <gunnlaugur@gmail.com>, Adrian Klaver <adrian.klaver@aklaver.com>, pgsql-hackers@lists.postgresql.org
Date: 2018-09-07T23:05:39Z
Lists: pgsql-hackers, pgsql-general 
Bruce Momjian <bruce@momjian.us> writes:
> On Fri, Sep  7, 2018 at 06:43:52PM -0400, Tom Lane wrote:
>> AFAICS there are no internal-to-the-C-code search path dependencies
>> in earthdistance.c, so it's not the same problem.

> Uh, there is an SQL function that calls functions from the module that
> fail.  It would be a CREATE FUNCTION patch, I think, but I thought the
> issue was the same.

Not really.  You could either interpolate @extschema@ into the text
of the referencing function, or (though much inferior for performance)
have it SET SEARCH_PATH FROM CURRENT.  Either of those changes would
involve an extension version bump since they're changing the extension
script.  What's more of a problem is that we could no longer claim
the extension is relocatable.  My unaccent fix dodged that by looking
up the C function's current schema, but I don't think there's any
equivalent functionality available at SQL level.

			regards, tom lane

Commits

  1. Make contrib/unaccent's unaccent() function work when not in search path.

  2. Avoid unnecessary use of pg_strcasecmp for already-downcased identifiers.