Re: CEE* FC handling (was RE: date math in CL)

[Scott Klement <klemscot@xxxxxxxxxxxx>]

Hello Barbara,

> Besides that, in general, I'd rather monitor for exceptions than check
> if an error occurred.  If you use exception-based code, you don't have
> to worry about forgetting an opportunity to check for errors.  If you
> forget to check the feedback, you can have silent errors; forgetting to
> code an exception handler gives you a nice noisy error.
>
> So most of the time I code *OMIT for the feedback and bytesProvided = 0
> for the errcode.

I absolutely agree with you.  I think one of the things that makes the
iSeries more stable than other platforms is the exception handling.  The
fact that errors default to a program "crash" instead of the program
having to check the return code and signal it's own error if something
failed.

The problem that this creates on other platforms (most notably Windows) is
that if there's a particular error that doesn't occur to a programmer, the
error can go unnoticed and the code and continue with strange results.
This leads to errors in production that are very difficult to trace back
to the source.

Java also uses this "you must handle it or pass it on" type of exception
method, which is great.

One thing, though...  if your program EXPECTS certain errors to occur
it should remove the error message from the job log so that it does not
confuse someone who is trying to troubleshoot something unrelated.   For
example, if your program asks a user for a date, calls a CEE* API to parse
that date, and expects the possibility of an "invalid date" error, it
should handle the error AND remove the error from the job log.

Too many times programmers will leave this stuff in the job log, and
that only creates confusion.

Just my opinion of course.