[m-dev.] My wishlist for 0.14

[doug.auclair at logicaltypes.com]

Now that 0.13 is nearly upon us (with my concern that I still
cannot build slice/ on powerpc.apple.darwin8.3 with
ROTD-2006-03-07), I figured I put forward my 'wishlist' for
release 0.14 ... in the spirit of facilitating program
development -- eliminating unnecessary obstacles and adding
useful new functionally; that kind of wishlist ... hey! one
can dream, can't one?

* I wish that every library call I make that is a function/N
  wouldn't cause the compiler to throw an ambiguity error as
  to whether I mean the function/N call or the equivalent
  predicate/N+1 call.  I hate this!  I especially hate it when
  there can be no ambiguity from the type and mode declarations
  that a curried pred is NOT the unification, but the returned
  value of the function IS.  So I wish that all predicates
  that have equivalent function declarations/definitions
  be removed from the standard library.
* I also wish that the compiler wouldn't throw ambiguity
  errors when I'm designating a pair in one situation and
  subtraction in another.  Each instance is clearly (eventually)
  typed and moded, but I still get the ambiguity error until
  I create these little one line auxilary predicates that
  only serve the purpose of casting the subtraction to a
  number and casting the pair to a pair.  Tiresome.
* I wish that list(io.poly_type) was an acceptable argument type
  to io.format/4.  I also hate it when the compiler tells me
  that io.format/4 needs string.poly_type within arg2 when
  I see clear as day that the library says io.format/4
  takes a list(io.poly_type) argument.
* I wish io.format/4 allowed something like Prolog's and
  Dylan's '~w' argument (for Mercury, it'd be '%w') where
  the corresponding argument would be a typed predicate:

  (pred(T::in, io::di, io::uo) is det)

  that tells io.format/4 how to print a T.  And, while
  we're at it, give string_streams (whatever they turn
  out to be) that power, too: any arbitrary T can be
  written out to an output stream, so it'd be nice if
  io.format/4 and string_streams could have 'put'
  functionality of the same strength.
* I wish std_util was 1/10th its current size: 'pair'
  is part of std_util?  Strange! And what are all these
  type_<foo> types and predicates doing hanging out
  here /and/ in module type_desc?  Also, 'maybe' a
  new module named 'monad' or something to handle
  foldl_maybe and its 45,000 variations in std_util?
  With all these huge sections of disparate protocol
  declarations, it looks as if std_util is a dumping
  ground, so nuggets of useful functionally are
  buried under this sea of pel-mel functionality.
* I wish module io was maybe 5 modules, or even 10,
  instead of this huge thing one needs scroll through
  with a bewildered look on one's face looking for 
  the signature of command_line_arguments/3.
* I wish, for the HTML library docs, that it has an
  index, sliced-and-diced by name, with a subsection
  for types and subsections for preds and funcs.  I
  wish each library module had a table of contents
  for types and preds/funcs.
* /soto voce/: op/3 declarations
* I wish io.state was a typeclass.  Sometimes I have
  a very different idea than the language runtime
  what the 'world' is.  The stream proposal seems to
  be a movement in this direction.  I have in mind
  something a little more radical.
* ... oh, yes, of course: 'Whirled Peas'.

Just something one thinks on while building a system
in one's favorite programming language ...

Sincerely,
Doug Auclair

--------------------------------------------------------------------------
mercury-developers mailing list
Post messages to:       mercury-developers at cs.mu.oz.au
Administrative Queries: owner-mercury-developers at cs.mu.oz.au
Subscriptions:          mercury-developers-request at cs.mu.oz.au
--------------------------------------------------------------------------