Erratum requested (what is a BODY?)
Marc Nieper-WiÃkirchen 23 Oct 2020 14:00 UTC
In SRFI 2, the syntax of the and-let* special form is given as
(and-let* (CLAWS) BODY)
However, it doesn't exactly say what is meant by BODY. Given that
and-let* is advertised as a cross-breed of let* and given the expected
behavior of binding forms, it makes sense to assume that BODY is to be
understood in the sense of a <body> of the R5+RS, that is a sequence
of definitions followed by a non-empty sequence of expressions.
The formal denotational semantics does not help much to clarify the
issue. They contain a BEGIN, but it isn't specified whether BEGIN is
assumed to be the splicing or the expression version of R5+RS (or
something else). Incidentally, the R5RS and R7RS also don't handle
internal definitions in their denotational semantics but rely on
syntax expansion first, which rewrites internal defines into an
equivalent letrec* form.
It would look like an obvious design mistake if the BODY of an
and-let* weren't equivalent to the BODY of a let* form (and, moreover,
would be quite unconventional for a binding form), so I would like to
request an erratum or amendment of SRFI 2 specifying that BODY is a
<body> in the sense of R5+RS. (As, in particular, the use of BEGIN is
currently underspecified, this doesn't break conforming code.)
As the sample implementation doesn't handle internal defines, it will
have to be (trivially) updated.
Marc
