Re: Identifier case folding notes

Andrew Dunstan wrote:
> I'm not sure if you've read all the archive history on this. Here are
> the pointers from the TODO list:
>
> http://archives.postgresql.org/pgsql-hackers/2004-04/msg00818.php
> http://archives.postgresql.org/pgsql-hackers/2006-10/msg01527.php
> http://archives.postgresql.org/pgsql-hackers/2008-03/msg00849.php
>
>
> The fact is that we have substantial groups of users who want
> different things:
> . current users who want no change so there is no breakage in existing
> apps
> . users on other DBs who want Spec compliance
> . users on yet other DBs who want case preservation
>
> The last group should not be lightly dismissed - it is quite common
> behaviour on MSSQL as well as on MySQL, so we have some incentive to
> make this possible to encourage migration.
>
> I'm strongly of the opinion therefore that this should be behaviour
> determined at initdb time (can't make it later because of shared
> catalogs). I suspect that we won't be able to do all this by simple
> transformations in the lexer, unlike what we do now. But I do think
> it's worth doing.
>
> cheers
>
> andrew
>
Hi,

as part of the
http://archives.postgresql.org/pgsql-hackers/2008-03/msg00849.php
thread, I did a reasonable amount of discovery work on making the 3
options andrew presents a reality. As I'm not skilled enough I never
got far enough to make them all work at once. I did however get lower
case and case preservation working.

To make those tow work the catalogs need no changes. Some of the
regressions tests expect case folding, so they did need changing to
operate correctly. I was unable to make the input files to initdb
correctly fold the case of system catalogs for the upper case version.
I'm sure somebody with more experience would not find it as difficult as
I did. Function names tended to be where all the gotchas were. Count()
vs count() vs COUNT() for example.

Once the db was up and running, the issue becomes all the supporting
tools. psql was made to autocomplete with case preservation, I was
going to make pg_dump just quote everything. I then got to the point of
adding a fixed GUC like LC_LOCALE that allows psql to read the case
folding situation and act according. That is where my progress ended.

Attached is what i had worked in. It's a patch against 8.3.1. I know
it's not CVS head, but it is what I was using at the time to experiment.

Regards

Russell