Page 1 of 1

Invalid character(s) codepage ibm-1051_P100-1995 to Unicode

Posted: Wed Feb 11, 2009 1:16 pm
by mouni
Hi

I have searched the forum for this issue.
My question is i read in the documents that the internal character representation used by datastage is UTF-16.

So , where is this ibm-1051 codepage coming to picture ?
Is the client server NLS mapping used causing the issue ?

The NLS_LANG for the database is set to WE8ISO8859P1 .
But i'm using ODBC enterprise stage so NLS_LANG is not used.
I believe NLS_LANG is for the Oracle Client.

So , please suggest what would be the real cause of the problem

Posted: Wed Feb 11, 2009 4:18 pm
by ray.wurlod
The NLS code page specifies how the data are encoded OUTSIDE of DataStage, for example in files and database tables. The NLS map translates from this encoding into the Unicode representation used internally by DataStage when reading. Or vice versa if writing.

Posted: Wed Feb 11, 2009 8:55 pm
by mouni
The project NLS is set to ISO-8859-1.
But i doubt , if the client server NLS plays any role here ?
In the admin clients even before going to the projects , there is a option of setting the NLS clientserver mapping .

If the project NLS had been used , it should have said Invalid characters found in ISO-8859-1 and then being converted to unicode .

Is that correct ?

Posted: Wed Feb 25, 2009 4:43 pm
by asorrell
Are you using 8.0.1? We saw some similar problems running PX jobs on 8.0.1 where it was reporting NLS conversion problems with various characters. The scary part was that we'd run the same job five times and it would report different bad characters each time, and might run cleanly the fifth time. We suspected some sort of memory corruption was occurring.

We tried the same jobs on 8.1 and the problem went away, so we're moving to 8.1 now.

Also - if you don't mind - I'd like to move this thread to the PX forum soon. NLS uses different mapping methodologies for Server and PX. Since you've referred to PX jobs this should probably be over there.