DSXchange

Performance tip 3.

Preloading ValidChars at compile time will give a small gain in performance. Instead of use

Performance tip 2.

The Index() function will be faster than Count(), because it can stop at the first occurrence. Count will need to process the entire string on each iteration of the loop.

Performance tip.


will be faster than

and the difference is greater the longer the string in RetValue is.

Another common cause for abnormal termination of stage is that the stage is handling NULL values where it isn't expecting to.

Beware that the Sequential File stage, unless you change the Format, will map "" to NULL when reading the text file. This can introduce NULLs that you weren't expecting.

If you reset the job after it's aborted, you sometimes get an event logged "from previous run...". This can contain additional diagnostic information.

Delete anything you don't need from the disk?

In particular, clean up the file system referred to by the UVTEMP configuration parameters, anything old in &PH&, any scratch space your routines use. Purging log files of old entries can help, too.

But then you wouldn't need DataStage. You could, of course, use DataStage to trigger the bcp -out command if needed as part of something larger. But why not use the graphical design paradigm; SELECT from your SQL Server table and write to a file through a Sequential File stage. A Sequential File sta...

You're hoping to move to UNIX. SQL Server doesn't operate on UNIX. Does that make the choice easier? Oracle is WAY more expensive than SQL Server. Dare I suggest, if what you're doing is a star schema design (which all good DWs should be), that you investigate Red Brick Warehouse from IBM? Red Brick...

That's not how I understood the original problem. I thought you wanted the highest (or last) NEW for every OLD. That should work through an Aggregator. You have a more complex requirement, which would need to be solved differently that the way I did it. (Mine is on version 7.0; I'll need to replicat...

What are the data types of A and B?

And, if you do that, you can use UNIX commands to generate a file name that includes the sequence.
Yay!
It follows that you can use the same trick with dates and other bits of file names.

It is not possible to convert.
Under certain circumstances, which you will find documented in Parallel Job Developer's Guide, you can encapsulate server jobs in Shared Containers and include those in Parallel jobs.

It may not be an empty string; it may be a NULL. You'll have to look at the job that loads the hashed file to determine this. Remember that the default behaviour of a sequential file stage is to convert "" to NULL; you override this default on the Format tab on its Outputs tab. The representation of...

No.
When the server job is compiled, the definition of the Transform becomes inline code.
This can be verified by inspecting the code generated from the Transformer stage that uses the Transform, which is in the RT_BPnnn directory in your project (where nnn is the job number from DS_JOBS).

The Command object can execute any command that you could execute at TCL. To start a DataStage job from TCL the command is DSD.RUN. Search the Forum for examples of the use of the command. Look in a PORT.STATUS report for examples that are actually running. The Command object has few properties and ...