DSXchange

Empty string ("") is a perfectly valid value for a non-nullable column. Why not add cycles of possible values to the schema? record ( firstname string[max=16] {cycle= 'John' 'Paul' 'George' 'Ringo'} ... ) (I may not have this syntax exactly right - I'm in an airport lounge (again!) and t...

The best way to solve this kind of problem is to figure out how you would do it if you only had SQL. In your case this will convince you that it can't be done in one pass. In SQL you would need a nested query. So too in DataStage. The "nested query" populates a lookup of some kind with the maximum v...

Try using two scratch disks per processing node. It's OK that they are shared, or you can create separate subdirectories if you really must keep them separate.

You need to draw a second output link, then right click on that link and choose Convert to Reject in order to make it a rejects link.

Very fast.

And it's hashed file, not hash file.

Yes, DataStage is built on UniObjects (which is now called DSObjects since the split, but Ascential never published a manual for it). The AccountName property of a Session object is the project. After that, however, there are no sub-objects called job, stage, etc. Rather, you need to use File and Se...

The second parameter of RESIZE is interpreted as a MINIMUM.MODULUS setting; the third parametr is interpreted as GROUP.SIZE 2 only if the parameter value is 8. RESIZE is primarily used by dynamic hashed files to reclaim free space within the file structure. As Arnd noted, there probably isn't any in...

or
(the alias becomes a virtual column name).

My guess is that it's exactly what it says. Can you add extra scratch disks on each processing node? Ideally these should be on separate spindles but you have not control over that in a SAN.

Possibly an index corruption somewhere in the Repository. Search the forum for DS.REINDEX.

That's completely unrelated to the topic. Please start a new thread.

Look (search) in the Parallel Job Developer's Guide for detailed information; the term is defined on page 21-9 and in other location. In short it means performing a lookup directly against a database table rather than pre-loading the lookup table into a virtual Data Set.

To quote Deep Thought: "Hmm. Tricky, I'll need to think about it."

Care to share?

Why not just build the appropriate difference query as SQL and execute that?
Otherwise, perform a reference lookup and constrain the output to be that the lookup failed (perhaps use reflink.NOTFOUND link variable)