Re: [Gm2] Use of functions in expressions

[Scott Robinson]

Hi Andreas,

I apologize for not explaining the situation thoroughly so I'll do so
now.  I have some existing Modula-2 code that was built with an
internal version of a Modula-2 compiler (the compiler was never sold
as a separate product).   I would like to run that code through gm2
since it will allow me to port the code to a different
architecture...hopefully with minimal effort.  It will also be an
excellent test for the gm2 compiler.  The main problem is that this
internal Modula-2 compiler has some non-standard enhancements so
whenever I run into something that doesn't compile through gm2 I have
to figure out if it is a problem with the gm2 parser (which is a beta)
or if it is a problem with the original Modula-2 code being
non-standard.  Hopefully that explains the questions I am posting to
the list since I don't have access to either the ISO Modula-2 standard
or  other production Modula-2 compilers.

Now to answer your particular points:

I'm not particularly interested in arguing the pros and cons of the
ISO Modula-2 standard especially since I don't have access to it :-).
However, it is a standard so I am assuming that Gaius is at least
designing the gm2 parser to accept language constructs accepted by
that standard as opposed to restricting it to Wirth (1983).  Hopefully
if there are production compilers out there that accept non-standard
language constructs Gaius is at least considering inclusion of those
constructs otherwise there would be porting difficulties moving
existing Modula-2 code from those compilers to gm2.

In regards to my example, evidently it wasn't acceptable so I'll try a
different one with a function returning a pointer instead:

MODULE test1;

FROM InOut   IMPORT WriteString, WriteCard, WriteLn;

TYPE
  tItemPtr = POINTER TO tItemNode;
  tItemNode =
    RECORD
      key   : CARDINAL;
      value : CARDINAL
    END;

  tListPtr = POINTER TO tListNode;
  tListNode =
    RECORD
      item : tItemPtr;
      next : tListPtr
    END;

PROCEDURE FindItem(list : tListPtr; key : CARDINAL) : tItemPtr;
BEGIN
  (* Empty to make this shorter *)
  RETURN NIL
END FindItem;

VAR
  list : tListPtr;

BEGIN

  (* Code to build the list here *)

  (* Attempting to dereference a pointer '^' returned by a function within
     an expression results in a compile failure. *)
  WriteString("The value with key 3 is ");
  WriteCard(FindItem(list, 3)^.value, 1);
  WriteLn;

  (* Code to destroy the list here *)

END test1.

The gm2 compiler gives the following error when compiling the above example:

#gm2 -c test1.mod
test1.mod:36:3: error: syntax error, found `^'
test1.mod:36:3: error: syntax error, `)' missing
test1.mod:36:3: error: syntax error, `END' missing
test1.mod:36:3: error: inconsistant program module name
test1.mod:36:3: error: syntax error, `.' missing
test1.mod:36:3: error: syntax error, found `,'
test1.mod:42:1: error: compilation failed

Because of the above error, I would have to store the function result
into a temporary variable and then do the pointer dereference using
the temporary.  Thus a function returning a POINTER has inconsistent
behavior versus a function returning a REAL, INTEGER, or
CARDINAL...all of which can be used immediately in the expression
without storing into a temporary first.  My opinion is that that
inconsistency is a design flaw in the Modula-2 language definition as
specified in Wirth (1983) and I am wondering if the ISO Modula-2
standard fixed the inconsistency by allowing the above syntax?  I am
also interested if any other Modula-2 compilers accept it?

Your suggestions on how to change the code aren't relevant (to me
anyways) since I have no intention of going to that much effort to
port the code to gm2.

You asked "why do I want to have string expressions?"  Well, to get a
specific character out of it would be one reason.  Other languages
treat strings as first class data types and have quite a few
operations defined on them (length, replication, splicing,
concatenation, etc).  All of those make perfect sense to use on a
string in an expression.

Those BNF diagrams you are pointing me to are derived from Wirth
(1983) and not the ISO Modula-2 standard (see
http://cui.unige.ch/db-research/Enseignement/analyseinfo/Modula2/) so
they do not provide the information I am after.

Thanks,
Scott

Cited references:
--------------------
Wirth, N., 1983 (2 (corr.) ed.). Programming in Modula-2.
Springer-Verlag: Berlin a.o, 176 pp.

On Sun, Nov 16, 2008 at 9:12 AM, Andreas Fischlin
<address@hidden> wrote:
> Hi Scott,
>
> Exactly because of such rather messy code did Wirth in his original language
> defintion NOT allow for such constructs. That some compilers accept such
> code is actually only due to the ISO standard of Modula-2. Prior to this
> your first formulation was not legal either and since I am not an adept with
> the ISO standard, I have no strong opinions about the pros and cons of this
> ISO extension of the original Modula-2. I'm wary it might complicate matters
> unnecessarily, something fundamentally at odds with Modual-2. Remember also
> that the ISO standard is not without difficulties, which makes people like
> me rather sceptical with respect to its benefits, except for settling on a
> standard library (but that has little to do with the language itself). In
> this context see also 10 little Modulans...
> http://www.scifac.ru.ac.za/cspt/sc22wg13.htm#Ten .
>
> More specifically:
>
> First, please note that I see very little reason to write in Modula-2 such
> code as you wrote. Function procedures should be used only if they return a
> well defined object, e.g. a real variable, if they have no side effects and
> if they can be used in place of a variable in an expression (e.g. think of
> it as a factor
> http://cui.unige.ch/db-research/Enseignement/analyseinfo/Modula2/factor.html).
> Therefore my first question to you whould be: Why do you want to have string
> expressions? Unless you would have some applications where you really want
> to multiply a string meaningfully with a number - or something similarly
> exotic - I see little sense in that. Modula-2 is helping you to stay away
> from such applications. Conclusion: use a procedure, NOT a function
> procedure.
>
> Second, when you comment some code by "this works", "this does not work"
> then it is not clear what you mean from that. You should in general always
> distinguish between, code that can be compiled, because the compiler accepts
> it as legal, or whether the code does fail during execution, typically
> causing a run-time error. From what you write further down I guess that you
> mean "can be compiled" vs. "can't be compiled".
>
> Thirdly,  Write(tester()[4]);  is a phantasy code clearly in conflict with
> any Modula-2 syntax I know of. It conflicts with the BNF for a procedure
> call (see
> http://cui.unige.ch/db-research/Enseignement/analyseinfo/Modula2/procedure_call.html).
> Therefore any Modula-2 compiler should reject it as syntactically wrong.
>
> A correct solution to what you try to do would be:
>
> PROCEDURE Tester(VAR s: ARRAY OF CHAR);
> BEGIN
>     Assign("testing",s);
> END Tester;
>
> BEGIN
>    Tester(temp); WriteString(temp);
>    ...
>
> That is all not very complicated and IMHO quite elegant code. Why trying
> something different that gives you and compiler builders only trouble?
>
> I recommend you give up on your idea of using a function procedure in
> relation with strings as illustrated above. Of course, there are some rather
> rare cases in which it may make sense to work with function procedures with
> respect to strings. E.g. if such a function returns an opaque string object
> or a single CHAR. See this module from the 'Dialog Machine' for such
> procedures:
>
>     http://se-server.ethz.ch/RAMSES/Objects/DM/DMStrings.html#DMStrings
>    (with Comments)
>
> http://se-server.ethz.ch/RAMSES/Objects/DM/DM_Quick_Reference.html#DMStrings
>    (without Comments)
>
> The case of a single CHAR may be interesting in a conversion function such
> as
>
>     PROCEDURE Upper(ch: CHAR): CHAR;
>     BEGIN
>         RETURN CAP(ch)  (*  should of course be something more meaningful *)
>     END Upper;
>
> and its application
>
>     PROCEDURE StringToUpper(VAR s: ARRAY OF CHAR);
>         VAR i,n: INTEGER;
>     BEGIN
>         n := HIGH(s); WHILE (i<=n) AND (s[i]<>0C) DO s[i] := Upper(s[i]);
> INC(i) END;
>     END StringToUpper;
>
> all not very useful, but I hope illustrative.
>
> In general, please note also, Modula-2 is NOT C and arrays, including
> strings, should be considered what they are, a structured data type that has
> first of all nothing to do with  dynamic data as passed by function
> procedures via the stack or pointers referencing data in the heap. An array
> is a static data structure with a finite size allocated by the compiler at
> compile time. It faciliates repetitive operation on its element, which, BTW,
> can be very complex, i.e. it can be an entire RECORD of a very complex type.
>
> Hope this helps to solve your problems.
>
> Sincerely yours,
> Andreas Fischlin
>
>
>
> Scott Robinson wrote:
>
> Hi all,
>
> I have another Modula-2 language question about function results and using
> them in expressions.  Specifically, it seems that the language allows you to
> return an array or a pointer but apparently (at least according to the GM2
> compiler) does not allow you to turn around and use an array subscript to
> access a particular element of the array within an expression.  The pointer
> dereference operator '^' also appears to be illegal after a call to a
> function.
>
> A specific example:
>
> MODULE test1;
>
> FROM InOut IMPORT WriteString, Write, WriteLn;
>
> TYPE
>  tArr = ARRAY [0..9] OF CHAR;
>
> VAR
>  temp : tArr;
>
> PROCEDURE tester() : tArr;
> BEGIN
>  RETURN "testing";
> END tester;
>
> BEGIN
>
>  (* This works *)
>  WriteString("The 5th character returned is ");
>  temp := tester();
>  Write(temp[4]);
>  WriteLn;
>
>  (* This does not work *)
>  WriteString("The 5th character returned is ");
>  Write(tester()[4]);
>  WriteLn;
>
> END test1.
>
> GM2 gives the following for the above:
>
> # gm2 -c test1.mod
> test1.mod:26:3: error: syntax error, found `['
> test1.mod:30:1: error: compilation failed
>
> I'm wondering if the Modula-2 ISO standard specifically disallows that
> syntax?  And, if anyone has access to other Modula-2 compilers, do those
> compilers accept the above?
>
> Thanks,
> Scott
>
>
>
> _______________________________________________
> gm2 mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/gm2
>
> --
>
> ________________________________________________________________________
> Dr. Andreas Fischlin, Ph.D., Group Director
>
> Terrestrial Systems Ecology
> Institute of Integrative Biology: Ecology, Evolution, Infectious Disease
> Department of Environmental Sciences, ETH Zurich
>
> Address:
> ETH Zurich, CHN E21.1
> 8092 Zurich, Switzerland
>
> Phone: +41 44 633-6090 / Fax: +41 44 633-1136
> http://www.sysecol.ethz.ch/Staff/af/
> http://www.sysecol.ethz.ch/
>
>      _/_/_/ _/_/_/ _/  _/
>     _/       _/   _/  _/   Eidgenoessische Technische Hochschule Zuerich
>    _/_/_/   _/   _/_/_/   Swiss Federal Institute of Technology Zurich
>   _/       _/   _/  _/   Ecole polytechnique federale de Zurich
>  _/_/_/   _/   _/  _/   Politecnico federale de Zurigo
>
>              Make it as simple as possible, but distrust it!
> ________________________________________________________________________
>
>