RETURNS length

Robin (or anyone else),
According to:

http://publibz.boulder.ibm.com/cgi-bin/bookmgr_OS390/BOOKS/ibm3lr30/6.15

"String lengths and area sizes must be specified by constants. The returned
value has the specified length or size."

I *hate* to sound like "you-know-whom" - but can you help me (better) understand
your statement (in another thread)

" A PL/I function can return a character string whose current length depends on
an input string length or anything else."

--
Bill Klein
wmklein <at> ix.netcom.com

I may not understand it all yet, but VARYING strings have a current and
a maximum length. The current length can change at anytime, the maximum
length needs to be at least known at allocation time.

-- glen

Thanks. That makes sense and is the same way that COBOL (the language that I
know best) works.

Of course, someone (either the CALLing/INFOKing/whatever) program or the CALLed
subprogram (function/whatever) needs to get at least enough storage to hold the
"current" data, but as long as one or the other does, this does work (in the
systems that I know).

--
Bill Klein
wmklein <at> ix.netcom.com

RETURNS( CHAR(whatever) VARYING ) will return a VARYING string with max
length "whatever" and current length determined by the data specified on
the RETURN statement, which sort of fits what the manual says, although
they could have been a little clearer. Multics PL/I used to allow
"RETURNS( CHAR(*) )" with or without varying, which is not allowed in
IBM compilers.

As do our compilers both for Unix and VMS. Odd that IBM never implemented,
but it can be tricky, if the stack isn't cleaned on exit.

There are C library routines that return a string value, or, in
C terms, a pointer to a character array. Returning a pointer to
a static array avoids any allocation/deallocation problems, but
causes strange results if you call the routine more than once
before using the value.

PL/I has, in general, been much happier to allocate temporary
arrays than Fortran. It seems that some of the Fortran 77 CHARACTER
variable rules were designed to avoid allocating temporaries.

Otherwise, one can use CONTROLLED variables (like Fortran ALLOCATABLE),
allocate them in a called subroutine. It might be that they need to
be arguments and not a function return value, but it should work.
That way it can be allocated to the appropriate size, and deallocated
by the calling routine.

-- glen

I may not understand it all yet, but VARYING strings have a current and
a maximum length. The current length can change at anytime, the maximum
length needs to be at least known at allocation time.

-- glen

Thanks. That makes sense and is the same way that COBOL (the language that I
know best) works.

Of course, someone (either the CALLing/INFOKing/whatever) program or the CALLed
subprogram (function/whatever) needs to get at least enough storage to hold the
"current" data, but as long as one or the other does, this does work (in the
systems that I know).

--
Bill Klein
wmklein <at> ix.netcom.com

RETURNS( CHAR(whatever) VARYING ) will return a VARYING string with max
length "whatever" and current length determined by the data specified on
the RETURN statement, which sort of fits what the manual says, although
they could have been a little clearer. Multics PL/I used to allow
"RETURNS( CHAR(*) )" with or without varying, which is not allowed in
IBM compilers.

As do our compilers both for Unix and VMS. Odd that IBM never implemented,
but it can be tricky, if the stack isn't cleaned on exit.

Thread Tools
Show Printable Version Email this Page
Display Modes
Linear Mode test 33Switch to Hybrid Mode Switch to Threaded Mode

2 24th October 08:44 glen herrmannsfeldt External User Posts: 1	RETURNS length I may not understand it all yet, but VARYING strings have a current and a maximum length. The current length can change at anytime, the maximum length needs to be at least known at allocation time. -- glen

4 24th October 08:44 peter flass External User Posts: 1	RETURNS length RETURNS( CHAR(whatever) VARYING ) will return a VARYING string with max length "whatever" and current length determined by the data specified on the RETURN statement, which sort of fits what the manual says, although they could have been a little clearer. Multics PL/I used to allow "RETURNS( CHAR(*) )" with or without varying, which is not allowed in IBM compilers.

5 24th October 08:45 tom linden External User Posts: 1	RETURNS length As do our compilers both for Unix and VMS. Odd that IBM never implemented, but it can be tricky, if the stack isn't cleaned on exit.

7 26th February 05:38 glen herrmannsfeldt External User Posts: 1	RETURNS "length" I may not understand it all yet, but VARYING strings have a current and a maximum length. The current length can change at anytime, the maximum length needs to be at least known at allocation time. -- glen

9 26th February 05:38 peter flass External User Posts: 1	RETURNS "length" RETURNS( CHAR(whatever) VARYING ) will return a VARYING string with max length "whatever" and current length determined by the data specified on the RETURN statement, which sort of fits what the manual says, although they could have been a little clearer. Multics PL/I used to allow "RETURNS( CHAR(*) )" with or without varying, which is not allowed in IBM compilers.

10 26th February 05:38 tom linden External User Posts: 1	RETURNS "length" As do our compilers both for Unix and VMS. Odd that IBM never implemented, but it can be tricky, if the stack isn't cleaned on exit.