On Mon, 16 Apr 2012 15:42:12 +0200, Thomas Berg wrote:
>>
>> >Premature terminations (posit/quit/admit) can almost always be handled
>> with
>> >LEAVE-type statements or immediate return from a subroutine. Some
>> languages have
>> >SIGNAL, EXIT, etc. which can help provide structured premature
>> termination for
>> >larger routines without resorting to the dreaded GOTO.
>> >
>> Rexx is a nightmare here. The LEAVE statement, even with a specified
>> target, won't exit a (nest of) procedure(s). The SIGNAL statement
>> trashes the DO nesting at the target level. Shame on Rexx.
>
>I can see that as a feature. ;)
>That prohibits some obfuscate like coding.
>
As it stands, the only way to exit a nest of procedures is to set
a flag and RETURN, and test that flag and RETURN after each
intervening cal; hardly unobfuscated. I'd be delighted to be
able to code, instead
DO I =1
DO J = 1
DO K= 1
CALL P1
...
END K
END J
END I
P1:
CALL P2
P2: PROCEDURE EXPOSE J
LEAVE J
I forgot to mention that SIGNAL trashes the DO nesting
but does not exit _any_ procedures. I was once called on
to advise a novice colleague who had used SIGNAL, he
thought, to leave a procedure nest. Necessarily, he had
coded all his loops with GOTO instead of DO. It worked
fine on small data sets; he came to me for help when it
overflowed the stack on a large data set.
>(BTW, How do You imagine the behavior if the Signal DID NOT trash the DO
>nesting? I mean, of what use would that be? As You have the CALL stmt.)
>
It could be used as a local GOTO. But that would still be
of no use to me, because I never use SIGNAL as a GOTO,
only in SIGNAL ON NOVALUE/ERROR.
-- gil
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN
