[r6rs-discuss] [Formal] The implicit cond in guard expressions is not a good idea

[Michael Sperber]

Abdulaziz Ghuloum <aghuloum at cs.indiana.edu> writes:

> The syntax is constructed on the assumption that conditional
> dispatch on the condition type is the default action that would be
> performed in the guard handler.  This is not necessarily the case.

It is not necessarily the case, but it is a very common one.  Witness
the exception-handling mechanisms in languages like ML, Java, or C++,
which all perform dispatch.  (Or even in the ancient Indiana proposal,
on which probably somebody at your site can provide information.)

> Examples from the draft:
>
> (guard (condition
>          (else
>            (display "condition: ")
>            (write condition)
>            (newline)
>            'exception))
>   (+ 1 (raise 'an-error)))

These examples are hardly representative, as they are taken verbatim
from SRFI 34, which did not assume a matching condition system.
Moreover, they are meant to illustrate the various cases distinguished
by the semantics rather than present common use cases.  Having said
that, I agree better examples would be nice.

> Suggestion:  Many ideas are plausible.  One possibility is to
> specify the syntax of guard expressions as:
>
>   (guard <variable> <handler-expression> <body-expression>)

I'm sympathetic to making any dispatch explicit, but note the
semantics of `guard' in the case where no clause matches.  One way to
do this would be to bind another variable to a procedure the handler
code can call to punt.

-- 
Cheers =8-} Mike
Friede, Völkerverständigung und überhaupt blabla