Re: getObject(<oid>) returns integer instead of LargeObject

Hi,

Is there some good reason for getObject(<some oid column>) to return the
bare oid integer instead of the actual LargeObject handle ?

I slightly modified the driver as below and it fixed my Blob issues
without any other side-effects. Would have some extensive testing
shown some problems I missed ? Sorry for not being very familiar with
Postgres' oids.

Thanks in advance.

Cheers,

Marc.

Index: postgresql/jdbc2/TypeInfoCache.java
===================================================================
RCS file: /usr/local/cvsroot/pgjdbc/pgjdbc/org/postgresql/jdbc2/TypeInfoCache.java,v
retrieving revision 1.1
diff -u -r1.1 TypeInfoCache.java
--- postgresql/jdbc2/TypeInfoCache.java 10 Apr 2005 21:54:16 -0000 1.1
+++ postgresql/jdbc2/TypeInfoCache.java 22 Jul 2005 13:45:35 -0000
@@ -51,7 +51,7 @@
private static Object types[][] = {
{"int2", new Integer(Oid.INT2), new Integer(Types.SMALLINT), "java.lang.Short"},
{"int4", new Integer(Oid.INT4), new Integer(Types.INTEGER), "java.lang.Integer"},
- {"oid", new Integer(Oid.OID), new Integer(Types.INTEGER), "java.lang.Integer"},
+ {"oid", new Integer(Oid.OID), new Integer(Types.BLOB), "java.sql.Blob"},
{"int8", new Integer(Oid.INT8), new Integer(Types.BIGINT), "java.lang.Long"},
{"money", new Integer(Oid.MONEY), new Integer(Types.DOUBLE), "java.lang.Double"},
{"numeric", new Integer(Oid.NUMERIC), new Integer(Types.NUMERIC), "java.math.BigDecimal"},

On Fri, 22 Jul 2005, Marc Herbert wrote:

> Is there some good reason for getObject(<some oid column>) to return the
> bare oid integer instead of the actual LargeObject handle ?

This makes the assumption that the only use of oids is for large objects.
What if someone wrote SELECT oid,relname FROM pg_class?

Kris Jurka

On Fri, Jul 22, 2005 at 01:41:19PM -0500, Kris Jurka wrote:
>
>
> On Fri, 22 Jul 2005, Marc Herbert wrote:
>
> > Is there some good reason for getObject(<some oid column>) to return the
> > bare oid integer instead of the actual LargeObject handle ?
>
> This makes the assumption that the only use of oids is for large objects.
> What if someone wrote SELECT oid,relname FROM pg_class?

Thanks for answering.

What would you recommend to unambiguously identify a Blob column in a
arbitrary ResultSet, using only JDBC?

For instance could you trust
Types.BLOB == rs.getMetaData.getColumnType() in every case?
Is there any other way?

Would it depend on the latest version of the driver/database?

Thanks in advance for you help!

PS: why do LOBs seem to use regular oids instead of some other less
ambiguous say, "loid" ?

Marc Herbert wrote:

> What would you recommend to unambiguously identify a Blob column in a
> arbitrary ResultSet, using only JDBC?

I don't think there is any way to do this in general, via JDBC or any
other interface; you need schema-specific knowledge to work out which
oid columns refer to LOs and which refer to a different sort of oid
somewhere else.

The actual LO interface itself just looks like "create LO" -> "ok,
here's an oid identifying it"; there's nothing that ties the column you
happen to use to store that oid to the LO itself. See
http://www.postgresql.org/docs/8.0/static/largeobjects.html for more
details.

For comparison, contrib/vacuumlo assumes that any (non-system-generated)
oid column in the system could potentially reference a LO (in GC-speak
it's a conservative collector).

> PS: why do LOBs seem to use regular oids instead of some other less
> ambiguous say, "loid" ?

It's just an artifact of the implementation AFAIK.

-O