Remove Postgres Schema before Release

[chrismeller]

Since we don't have a resident Postgres guru and the schema hasn't been updated or tested to work with the ACL changes, Postgres should be removed from maakanga before release.

Simply deleting the pgsql directory under system/schema/ should be all that's needed.

[rickc]

If this is done, removing/commenting out references to Postgres in the installer should be done, too, so people don't try to install it.

[mnacos]

a few changes here and there to fix postgresql warnings

[mnacos]

Replying to chrismeller:

Since we don't have a resident Postgres guru and the schema hasn't been updated or tested to work with the ACL changes, Postgres should be removed from maakanga before release.

Simply deleting the pgsql directory under system/schema/ should be all that's needed.

Please let me be your Postgresql guru :-) Postgres now works for me with a few harmless changes. Please see the attached patch file, check the patch and mark this ticket as worksforme :-)

[gsnedders]

We've fixed PostgreSQL, so I guess this should be notabug.

[chrismeller]

I've got a question related to the patch, which ringmaster added in r3530, so hopefully someone will answer here - if not I'll post to -dev.

In both of our PDO::prepare() calls (lines 265 and 340 in the patch) we're now passing the PDO::ATTR_EMULATE_PREPARES attribute. In the MySQLConnection class, we set this on connection.

My question is really 3-part:

First, does Postgres need this? I thought Postgres supported prepared statements natively, which would mean we don't need PDO to emulate them (like you do for MySQL).

Second, is there a need or some added value in passing the argument to each prepare() call vs. once on connection? The PDO documentation can be very confusion / contradictory / spotty / etc., so I honestly have no clue when or where it should be passed.

Finally, why is this in DatabaseConnection? If it's PG-specific we need to abstract out the method (or come up with some other manner of doing it, rather than duplicating the entire method), since we're now emulating prepared statements on all platforms - even those that natively support it. Right now (assuming PG does not) that isn't any, but adding Oracle support would change that.

[michaeltwofish]

See the  discussion on -dev

[chrismeller]

Fixed in r4223.