Re: Canonical Indentation

Benjamin Teuber <beteub@[EMAIL PROTECTED]
> writes:

> On 18 Jul., 03:23, Joost Diepenmaat <jo...@[EMAIL PROTECTED]
> wrote:
> 
> > Isn't that why &body is for?
> 
> Right, but I'm not sure whether this is enough - maybe there happens
> to be a macro with two bodies or some other reason for different
> indentation. Question to the experienced guys: Did you ever stumble
> across a macro where &body alone wasn't enough to specify a good
> indentation?

It's extremely rare and usually not an issue, not because it can't be
wanted but because Lisp syntax doesn't really allow for it.

The problem comes that if you wanted multiple bodies, there would be 
no obvious cue about how to divide them.  So usually what you do is
like what COND does where you do:

  (defmacro my-cond (&body clauses)
    ...
    (dolist (clause clauses)
       ... (destructuring-bind (test &body forms) clause ...) ...)
    ...)

Or else you just take one as a form and make the user use progn, as in:

  (defmacro my-multiple-value-bind (bindings value-form &body forms) ...)

where in practice one often occasionally wishes for 

  (defmacro my-multiple-value-bind (bindings &body value-forms 
                                             &more-body other-forms)
    ...)

The fact is that &body is not the source of any problem, it just reminds
you
that there are inherent limitations with positionally supplied semantics.
That problem is a problem of the universe, not even of Lisp. e.g., it
comes
up with regular expressions, for example, when you try to juxtapose to * 
wildcards and either combinatoric answers come up or the string doesn't
compile, depending on the tool.  It's also the reason that if you're
wanting
something scanf-like, as in (setf (format nil "~D" x) "3"), you can't 
allow cases like (setf (format nil "~D~D" x y) "357") to set multiple
values
because you can't know what the partitioning is. &body is just another
case
of this broad class of issues.