DSXchange

Are you doing a 'COUNT' in your SQL? What is your SQL in the ODBC stage?

Joins want sorted data, so it is best to put in your own explicit sort stages and, when not necessary, mark sort key columns as "don't sort, previously sorted". Partitioning and repartitioning must be done on the join keys.

I just realized that I cut-and-pasted an incorrect line 2. The first line is a stage variable that is an integer which gives the starting position of the substring. The second line should be: IF svKeyPosition THEN Xfm_DuplicateCAAPS_IN.Chinese_text[1,svKeyPosition-1]:Xfm_DuplicateCAAPS_IN.PORDNUMB:X...

This is a very basic error message, it means that your file contents do not match up with your DataStage file definition. Specifically, with a fixed width file, your column widths don't match up. In addition, usually fixed width files don't have record delimiters.

I vaguely recall having seen this before as well, and it was a compiler-related issue. I think I changed the compiler optimization switches around to get rid of the error. If you google your platform, compiler and the error message you might be able to get more information.

Is it reproduceable if you do a multiple-job compile with just that sequence, does it also generate the error message? If so, you should submit it to your support provider.

The convert function works on a single character, not a string and won't work for you in this case. You might use this:

kollurianu wrote:... [IIS-CONN-ORA-003005] A data conversion error was encountered in bulk load mode for row 3, column 10...

What is the data type and value of row 3, column 10?

Nope, the documentation states that the divisor and remainder are integers.

What about in the NLS tab of the Oracle enterprise stage, could you set that to use project default?

The BASIC transform stage is not visible in the palette. If you go to the Repository tree, select Stage Types -> Parallel -> Processing -> BASIC Transformer and drag that to the canvas.

If you are using upsert method, then no error would be generated, since the first record does an insert, the second an update.

When calling a job from a job sequence you choose what parameters to pass, so doing what you are asking is a simple task.

Would it be possible to post the generated OSH code to the thread?

There are so many possible sources for this error and so little information to work with that it is impossible to even guess as to the cause. Take one job as an example, then see if the read or write to or from the database is taking longer - do this by making a copy of the job and replacing the out...