[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: reference implementation; multiple patterns after ellipsis




[ I've lost count of how many times I've accidentally sent a mail to
  one person instead of the list.  Grrrr. ]

On Sunday, Oct 19, 2003, at 21:10 US/Eastern, Alexpander Petrofsky wrote:

The draft says:

  No implementation is provided because this is SRFI is merely a
  request to add two basic features to the SYNTAX-RULES pattern
  language, and so the implementation of it would be a modification of
  an implementation of SYNTAX-RULES.

I think that's bogus.  It's true that RFI stands for "Request For
Implementation", but that should be taken as "Request for others to
implement in the same way what I have found to work well" rather than
"Request for someone to implement my idea that's never been tried,
even by me".

Well, it hasn't!  (except in your macro transformer, and some
hypothetical macros that I've written involving making Andre van
Tonder's monadic CPS macros even cooler)

Even though it's impossible to provide a simple implementation that
everyone can just drop into their system, I think it's still valuable
to have something that at least proves the idea is implementable, and
that people can look to for guidance if the English specification is
later found to be self-contradictory.

(see two sections below)

One advantage of the original (... ...) idea over
choose-your-own-ellipsis is that it is already widely implemented, and
you could use the free somewhat-portable syntax-case expander from
chez scheme as the reference implementation.

If you opt for the more pioneering CYOE concept, then you could
perhaps provide a modified version of the chez scheme expander.  An
advantage of the chez-scheme expander is that it is used by several
systems (like SISC and Chicken, IIRC), and patches to it could
actually be used by those systems' implementors.

Eek.  That would require that I somehow comprehend psyntax.  If
Scott and Felix can only _barely_ comprehend it, I have grave doubts
about _me_ comprehending it...(but I suppose I can try to, and maybe
try to comprehend yours as well, although the more I think about it
the more I dislike the idea of writing a macro transformer purely for
SYNTAX-RULES and _directly_ for SYNTAX-RULES; Scheme48's macro
expander, for instance, totals fewer than one thousand lines of code,
_including_ comments, and it's a SYNTAX-RULES->explicit renaming
transformer _as_well_as_ a macro expander for explicit renaming;
unfortunately, the explicit renaming bit uses Scheme48's internal AST
stuff, so it's not portable at all, and not many Schemes support
explicit renaming)

Another option, more in the "so at least people have something to look
at" category, would be to use the macro expander I have been working
on occasionally, available at petrofsky.org/src/alexpander.scm, which
I just updated to include support for CYOE and tail patterns.  (I also
added a few pages of commentary, but it doesn't rise to the level of
comprehensibility yet.)

I guess I could put a link to your expander in the implementation
section.  Would that be OK?

I just noticed that the draft only allows one extra pattern after an
ellipsis.  Don't you want to allow any number, as in Chez scheme?
That's the way I implemented it.

Yes, I realized this a few days ago, but I won't change it until I do
some more drastic stuff to the document.

-al