Re: MONITOR performance

[Barbara Morris <bmorris@xxxxxxxxxx>]

Buck wrote:
> ...
> Neither MONITOR nor the *PSSR have an impact on the code when an exception
> does not occur.  There is possible performance impact related to the program
> status data structure and fields beyond column 80.
> ...

That's true about the *PSSR, but MONITOR unfortunately has to have some
executable code - not too much, but some.  (Hope I didn't ever say that
MONITOR had no performance impact ... )

I would only use MONITOR where I wanted to recover from an error and
continue processing, and let the *PSSR catch everything else.

With existing code, if the change you're making would benefit from
MONITOR, you have to consider how you would handle it without MONITOR. 
If your ON-ERROR would just do a RETURN, then you're better off with the
*PSSR.  If your ON-ERROR would do something to enable the program to
continue, you'd have to do some of your own setup so you could get back
from the *PSSR - that setup would probably cost you as much as MONITOR.

In subprocedures that have I/O, you can only handle I/O errors with
error indicators, (E) extender or MONITOR.  Using one MONITOR around the
entire code will allow you to handle I/O errors the same as you would
handle program errors in a *PSSR.  In a subprocedure where you want to
handle exeptions, the alternative to MONITOR is to code (E) or an error
indicator on every I/O opcode, plus the code to check for errors after
every one.