Re: Class Notes for a PL/I Course.

One minor error:

"that is, only parameters are passed."

should read:

"that is, only POINTERS are passed."

Sorry

Bernd


Am 03.10.2019 um 13:16 schrieb Bernd Oppolzer:
> Am 03.10.2019 um 12:42 schrieb Giuseppe Vitillaro 
> giuseppe-yuD/ahkh7LvrZ44/[email protected] [H390-MVS]:
>>
>> On Thu, 3 Oct 2019, Bernd Oppolzer berndoppolzer-/[email protected] [H390-MVS] 
>> wrote:
>> >
>> > Kind regards
>> >
>>
>> I guess declaring this way a variable in the main program:
>>
>> DECLARE ZAZA CHARACTER(10) STATIC EXTERNAL INITIAL ( 'GIGIO' );
>>
>> and using it in an external procedure in this way:
>>
>> ZZ: PROCEDURE;
>> DECLARE ZAZA CHARACTER(10) EXTERNAL;
>> PUT FILE(SYSPRINT) SKIP EDIT ('ZAZA=', ZAZA ) (2(A));
>> RETURN;
>> END ZZ;
>>
>> it is the best which can be done in PL/I(F) to get something
>> which looks like a global C variable?
>>
>
> The short answer is: yes, this is the best.
>
> The long answer is: this does not really help you, if you want to
> do some sort of dynamic linking, because you need the linker to
> resolve those external references.
>
> BTW: STATIC in the main program is not needed or even forbidded,
> because EXTERNAL variables are STATIC by default.
>
>
>>
>> Another thing I'm looking for is something like
>> the C typedef syntax.
>>
>> The "LIKE" syntax looks to do something similar,
>> but for an already "allocated" variable, if
>> the variable is not actually a pointer.
>>
>> Anything better can be done with PL/I (F)?
>>
>> Peppe.
>>
>>
>
> This is somehow related to the question above.
>
> If you want parameters to be passed between external functions,
> they may be structures - in the general case.
>
> The C way to do this would be
>
> - have a typedef of the struct in an include file
> - then use this include file on both sides
> - pass pointers of this struct between caller and called function
>
> This way the called function can read and write values from and
> to the parameter structure.
>
> In traditional PL/1, there is no typedef. There is simply no concept
> of type. See the funny assignment of structures (called aggregates),
> where it is only needed that the layout of the structures are
> congruent in a certain way, but the types of the components may differ.
>
> Or the second variant: BY NAME.
>
> This is all very different from what other languages do.
>
> So what is done "in real word" is often the following:
>
> we use %INCLUDE files, too. This works in PL/1, even if you don't
> use the preprocessor (which is another very interesting topic;
> the PL/1 preprocessor is much more powerful than the C macro processor,
> for example).
>
> in these %INCLUDE files, we define PL/1 structures (or aggregates)
> without the structure header, that is, without the DCL and the level 1
> definition. Only level 3 or 5 ...
>
> these structure bodies are used at both sides (caller and callee).
>
> But the caller declares:
>
> DCL 1 PARM_STRUCT AUTO,
> %INCLUDE (PARMINC);;
>
> and the callee declares:
>
> DCL 1 PARM_STRUCT BASED (PARM_PTR),
> %INCLUDE (PARMINC);;
>
> and the call looks like this:
>
> CALL CALLEE (ADDR (PARM_STRUCT));
>
> that is, only parameters are passed.
>
> It is possible to pass structures, but then you will get real trouble
> with structure aggregates.
>
> Passing pointers to structures will work well, when calling other
> languages, too. There will be some issues with component alignment
> (the PL/1 methods differ from that of other languages), but that
> can be solved.
>
> This is not typedef, but it works a little bit like typedef :-)
> == COBOL copybooks :-)
>
> HTH,
> kind regards
>
> Bernd
>
>