DSXchange

Well... they are "associated" with tables in the sense that they are generally (as in a best practice) named after the table that the sequence will be used with, but you're right that they are standalone objects and can be named anything you like.

I've got several that fall squarely into that camp, hence the question about actually trying it before judging it evil.

Then you'll need to ensure there is always at least one record delivered. If the job delivers 'dynamic table names' what table would you create when no rows are generated?

What is the 'flaw' here? What do you think needs to be fixed?

Clarification, in case I wasn't clear. For Windows you should have these as equivalents to the UNIX SH(ell) functions:

ExecDOS
ExecDOSSilent

From what I recall... no Windows servers here.

Well... for one thing you are actually only writing to the hashed file on the first call to the routine per hashed file name - i.e. when the name is not found in the list of opened files.

Welcome. :D You'll probably need to write your own version of the Execute Command as a custom routine to leverage the 'Silent' version of the appropriate Exec function. For example, on UNIX servers there is: ExecSH ExecSHSilent I'm assuming the same is true for a Windows server. If not, you could ea...

You are probably going to need to post your modified routine.

Without posting your derivation or the compiler errors, no-one is going to be able to provide much help.

For HP-UX on Itanium, the installer either doesn't install a link for an Oracle library or it puts it in the wrong place. Support sent me the command to create a softlink so that DataStage could find what it needed for Oracle connectivity, nothing Oracle related worked before that even though I had ...

No problem. The Merge stage is a frequent source of confusion as there are two stages with the same name that bear pretty much zero resemblance to each other.

Sorry, can't really help with your specific UDB issues as I have no experience with it. However, just wanted to comment on your "Both of these jobs were fine before the upgrade so I am at a loss to explain it." comment. You see things like this from an upgrade for two different reasons. The first is...

Welcome. :D I don't believe so as DEFFUN doesn't return a status. Or at least not easily. I suppose one could query the repository to verify it exists but then what? Abort if it can't be found? End result seems pretty much the same to me. And I would wager if you checked deeper in your logs the miss...

Your problem, of course, is that you now need to do 10 lookups for each record rather than one. Have you tried putting all 10M records into one hashed file, even if you need to make it 64BIT to do so? Build both and constrast the speeds. If you can, what I'd do is partition the input data in the sam...

And be careful where you poke your... browser.