Re: bug report: slow getColumnTypeName

On 10/12/2012 04:39 PM, Luis Flores wrote:
> I agree, my only doubt is about the reasons behind the change, the driver was reporting int4, int8, and then was changed, why?

Yep, that I'd like to know. The change appears to be:

commit ddf8296bead68552a8e5de0f5bb20875273bb02d
Author: Kris Jurka <books(at)ejurka(dot)com>
Date: Thu Sep 30 07:58:11 2004 +0000

Return serial datatypes in both getTypeInfo and getColumns methods.

Jaroslaw J. Pysnzy

Also add a test case for this and fix my previous regression test
breakage on 7.2 servers. With serial columns before dependency
information dropping a table did not drop the sequences that went
with it. Explicitly drop them.

Kris Jurka

see git diff
ddf8296bead68552a8e5de0f5bb20875273bb02d..8c9d68ee7763851732de0bd0d14b2b51cdfe0622

That code is all kinds of wrong. Check this out:

if ( defval != null ) {
if ( pgType.equals("int4") ) {
if (defval.indexOf("nextval(") != -1)
tuple[5] = connection.encodeString("serial");
}
else if ( pgType.equals("int8") ) {
if (defval.indexOf("nextval(") != -1)
tuple[5] = connection.encodeString("bigserial");
}
}

*any* int4 or int8 with a DEFAULT nextval(... is reported as "serial" or
"bigserial" whether or not it is. See AbstractJdbc2DatabaseMetaData.java
line 2480.

--
Craig Ringer