To: dylan-partners@cambridge.apple.com
Subject: macro open issues
fcc: out-copy
Reply-To: ram+@cs.cmu.edu
--------

define method macros:
    It seems to us that users will often want macros that look like:
        define frobboz ?body end;

    Is there some syntactic reason why these aren't allowed?  If this is
    just due to the definitional complexity of potentially destructuring
    the parameter list, it would be better to have parameter lists match as
    a unit than to forego the natural syntax for code containing defining
    forms.


Reservedness of statement macro intermediate words:
    We do think it would be nice if intermediate words were only reserved
    within uses of that macro.  We don't think it should be done if it
    forbids LR(1) parsing.

    So far as we can tell, it would work to change the lexical class of the
    defining words dynamically during the parse.  This could be triggered (as
    already suggested by somebody) by seeing the statement keyword, or
    possibly by an early semantic action in parsing the statement.  This
    might have bad interactions with macro hygiene (see below.)  Obviously,
    we would need an implementation of this before we could choose local
    reserving.

    We don't think that it is all that bad to reserve intermediate words in
    all the modules where the macro is visible, but...


Renaming intermediate words:
    If intermediate words are module-reserved, then it would be very nice
    if they could be renamed via the module system.  This way, it would
    always be possible to work around conflicts that arise from using
    separately developed modules.

    Is there some reason why it doesn't make sense to make all intermediate
    words be handled (separately) by the module system in the same way as
    the initial word?


Reserving define words:
    We don't think that reserving the names of define macros is a problem.
    It seems like it might be desirably consistent for the "keywords" of
    all (non-function) macros to be reserved.  That is, if we decide to
    reserve statement intermediate words, then begin words and
    intermediate words are already consistent.  Why not make define words
    reserved too?  If define words are reserved, then the -definer suffix
    issue is moot.


Pattern variable constraint/name:
    We favor changing the "?constraint:name" syntax to "?name:constraint".
    Note that NAME still defaults to the constraint, so "?body" still is a
    constrained variable.

Constraint name abbreviation:
    We mildly favor keeping the abbreviations; they are quite clear, so the
    only possible confusion is when typing them in.

Constraint name digits:
    We oppose allowing "?var2"; if you really are that uncreative, you can
    always say "?var2:var".


Reserved word hygiene:
    Since macros can introduce new reserved words, we want some sort of
    hygiene for reserved words.  The intent isn't really clear from the
    existing discussion of hygiene.  For example, can I have a macro whose
    expansion freely references a normal variable which happens to be a
    reserved word in some modules which call that macro?