DSXchange

Which part of "undocumented" was unclear?

You can not.

Search the forums!

There is no other stage that can perform the functionality. Why do you want to replace the Transformer stage? That is it relatively inefficient compared to other stage types is no longer true (in versions 7.5 and later).

What format does Teradata expect? Perhaps 01/01/0001 ? Is there a lower limit on valid dates in Teradata?

Definitely NO.

You can get timeouts on the server itself, on the domain controller, and/or in the network, all caused by too great a load.

In a server job before/after subroutines must be in written in DataStage BASIC. However within these you have DSExecute, which you can use to execute any command and capture its output and exit status.

Prevention is always better than cure.

If you can't manage that, then the duplicates should be discarded into the sqlldr "bad" file and mentioned in the "log" file.

For source files that are text files it does not matter what the column names are. The data types have to be correct.
Other than that, I don't believe you can do what this requirement specifies.

Resist stupid requirements.

You don't use any of them when you set up a project. As for maintaining a project, the main commands you use (the DS.TOOLS menu and the DS.CHECKER command) are undocumented.

Analogy: Is there any other way to kill people, as if it is necessary to kill ?

It is never necessary to kill DS processes if jobs are well-designed.

Not possible as shell variables, as these are local to the shell in which the script runs. However you can echo the pathnames to stdout and pick those up as the command output ($CommandOutput activity variable) of the Execute Command activity that you use to run the shell script from a job sequence ...

You might be better off tracking down a UNIX-based variant of osql.