Re: [HACKERS] JDBC pg_description update needed for CVS - Mailing list pgsql-jdbc

From Jeroen van Vianen
Subject Re: [HACKERS] JDBC pg_description update needed for CVS
Date
Msg-id 5.1.0.14.2.20010907163902.00a864c8@ams010.satama.com
Whole thread Raw
In response to Re: [HACKERS] JDBC pg_description update needed for CVS tip  (Bruce Momjian <pgman@candle.pha.pa.us>)
List pgsql-jdbc
At 00:30 9/7/2001 -0400, Bruce Momjian wrote:
>Can someone tackles this and supply a patch?

This has been addressed in the patch that was recently committed for JDBC's
broken getColumn() support. As I'm using outer joins and was unable to come
up with SQL syntax that would correctly use an outer join with a function
returning a single row (col_description), I used the actual function
definition for col_description for >= 7.2 servers. For details see my mail
at http://fts.postgresql.org/db/mw/msg.html?mid=1032468

OTOH, I haven't touched JDBC's getTable() code.

> > Would some JDBC hacker develop a patch for the following issue?  The
> > change is just barely large enough that I don't want to commit untested
> > code for it --- but not having a Java development environment at hand,
> > I can't test the updated code.
> >
> > The problem is in DatabaseMetaData.java (same code in both jdbc1 and
> > jdbc2, looks like).  It does direct access to pg_description that isn't
> > right anymore.  In getTables, instead of
> >
> >       java.sql.ResultSet dr = connection.ExecSQL("select description
> from pg_description where objoid="+r.getInt(2));
> >
> > it should be
> >
> >       java.sql.ResultSet dr = connection.ExecSQL("select
> obj_description("+r.getInt(2)+",'pg_class')");
> >
> > In getColumns, the change is a little more involved, because
> > pg_attribute doesn't have an OID column anymore.  The initial query
> > can't fetch a.oid, but should fetch a.attrelid instead, and then the
> > pg_description query should become
> >
> >       java.sql.ResultSet dr = connection.ExecSQL("select
> col_description("+r.getInt(1)+","+r.getInt(5)+")");
> >
> > (col_description takes the table OID and the column's attnum).
> >
> > The reason this is more than a 3-line change is that it should be done
> > either the old way or the new way depending on whether server version >=
> > 7.2 or not, for backwards-compatibility of the driver.
> >
> > It's possible there are other similar changes needed that I missed in a
> > quick lookover.
> >
> > So, would some enterprising person fix the JDBC code to work with CVS
> > tip, and submit a patch?
> >
> >               thanks, tom lane
>
>--
>   Bruce Momjian                        |  http://candle.pha.pa.us
>   pgman@candle.pha.pa.us               |  (610) 853-3000
>   +  If your life is a hard drive,     |  830 Blythe Avenue
>   +  Christ can be your backup.        |  Drexel Hill, Pennsylvania 19026

Cheers,


Jeroen


pgsql-jdbc by date:

Previous
From: Bruce Momjian
Date:
Subject: Re: JDBC patch procedures (Re: [PATCHES] Patch for jdbc2 ResultSet.java)
Next
From: "Rene Pijlman"
Date:
Subject: Fix set/get transaction isolation level test in ConnectionTest