perl6-language-objects Report 31 Aug 2000

For those not on perl6-announce or who thread their mailers...

=head1 VERSION

   Date: 31 Aug 2000
   Number: 1
   Mailing List: [email protected]
   Chair: Nathan Wiger <[email protected]>

=head1 SUMMARY

The main point which most discussions are currently centering around is
the idea of fundamental embedded objects in Perl 6. With this concept, a
simple scalar would instead be an object with properties and methods.
Developing special datatypes (for example, int) would be as simple as
creating a subclass.

Under the current thinking, a statement like this:

   my int $x = 5;

Would do something like this behind the scenes:

   $x = int->CREATE;
   $x->STORE(5);

Then 'int' would simply be a subclass of SCALAR which might look
something like this (pardon the extreme pseudocode):

   package int;

   # inherit all methods from SCALAR except:

   sub STORE {
       if ( is_not_an_int($_[0]) ) {
           return ERROR_BAD_DATA_TYPE;
       } else {
           internally_store($_[0]);
       }
   }

Concepts such as optimization of mathematic ops, promotion of datatypes,
operator overloading, specialized stringification, and much more could
all be handled by simply overriding class methods. 

Several RFC's - 159, 161, and 171 - work together to make this possible.
These RFC's have comprised the majority of the discussion for the past
two weeks. Overall the response is positive, although there are many
details to be worked out.

=head1 RFC DISCUSSIONS

=head2 RFC 124 (v1): Sort order for any hash (David Nicol)

Many people liked the idea, which is a shortcut to automatically sorting
hashes. Many had very different suggestions for its implementation,
however.

=head2 RFC 137 (v1): Perl OO should I<not> be fundamentally changed
(Damian Conway)

This RFC explored several extensions to current Perl OO in order to make
it more powerful, without sacrificing its current simplicity or
flexibility. Unfortunately, I think most people interpreted this as a
simple philosophy RFC because of the title, when in fact it was a very
thorough document, so little discussion was recorded. I recommend
renaming it.

=head2 RFC 152 (v1): Replace $self in @_ with self() builtin (Nathan
Wiger)

This sparked a semi-religious debate which resulted in no clear
resolution. Most everyone agreed that the invocant being in $_[0] is a
PITA. But some people claimed that the invocant was worthy of a special
pragma to allow it to be named whatever you want. Others maintained that
this was extraneous and confusing, and that a single keyword or variable
should be selected.

=head2 RFC 159 (v1): True Polymorphic Objects (Nathan Wiger)

This RFC proposes a set of default methods, similar to those used by
tie, which could allow object-based operator, context, and data
overloading. Some discussion insued of whether || and && should be
overloadable, and additional methods were proposed. Overall the proposal
was well-received.

=head2 RFC 161 (v2): OO Integration/Migration Path (Matt Youell)

This proposal suggests the fundamental integration of OO into Perl, with
the key being that it should be behind the scenes and invisible. Many
agreed it was a good concept, assuming that it was both truly invisible
and did not cause Perl to become dog slow.

=head2 RFC 163 (v1): Automatic accessors for hash-based objects (James
Mastros)

There was dislike of the complexity and implementation of this RFC,
which allows for hash keys to masquerade as subs transparently. Several
suggestions for simplifying it were proposed. The idea itself was
relatively well-received.

=head2 RFC 171 (v1): my Dog $spot should call a constructor implicitly
(Michael Fowler)

This proposed a C++ like constructor mechanism whereby the declaration
of a variable implicitly creates an object. This mechanism would allow
for declarations of datatypes to cause class inheritance. Some concerns
over confusion and implementation details were raised, which are still
being addressed.

=head2 Protoproposal for new Backslash Operator (David Nicol)

This generated a lukewarm reception. Most people objected to the idea,
noting that it can be done other ways and didn't appear to add enough
value for the confusion it caused.

=head1 REFERENCES

If you're interested in these topics, here are some highlights of the
mail threads for several discussions. I encourage people to read them
and join into the discussion on -objects if they're interested:

http://www.mail-archive.com/perl6-language-objects%40perl.org/msg00068.html
http://www.mail-archive.com/perl6-language-objects%40perl.org/msg00131.html
http://www.mail-archive.com/perl6-language-objects%40perl.org/msg00148.html
http://www.mail-archive.com/perl6-language-objects%40perl.org/msg00169.html
http://www.mail-archive.com/perl6-language-objects%40perl.org/msg00152.html
http://www.mail-archive.com/perl6-language-objects%40perl.org/msg00109.html
http://www.mail-archive.com/perl6-language-objects%40perl.org/msg00093.html
http://www.mail-archive.com/perl6-language-objects%40perl.org/msg00006.html
http://www.mail-archive.com/perl6-language-objects%40perl.org/msg00037.html