[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Scheme-reports] equivalent rewrite of a sequence of (syntax) definitions?

On Fri, Jan 4, 2013 at 12:03 PM, Daniel Villeneuve <dvilleneuve@x> wrote:
Without syntax definitions, a sequence of definitions is
equivalent to a letrec* form (Section 5.3.2).

With syntax definitions, since this is explicitly not covered by Section
the wording of Section 5.4 should be used.  However, it is not said:
- what happens with too mutually recursive syntactic definitions;
- how to split subsequences of regular and syntactic definitions.
Mutually recursive syntactic definitions work as they always
have.  I'll see if I can clarify the language, but basically this
works just like the top-level - the variables and syntax all refer
to each other.  Redefinitions are forbidden, and obviously
actually using a value before it is lexically occurs is an error.

There's no need or desire to split syntax and non-syntax.

I would propose the following rewrite:

Given a sequence of definitions and syntax definitions at the start of
a <body>, let r_i be the identifier bound by the i^{th} definition v_i
and s_i be the identifier bound by the i^{th} syntax definition d_i.
Then, the entire sequence of definitions and syntax definitions
is equivalent to

(letrec-syntax ((s_i d_i) ...)
   (letrec* ((r_i v_i) ...)

Is this a plausible rewrite?

Unfortunately this rewrite wouldn't work - it's important that the
syntax definitions refer to the non-syntax definitions, thus they
must all be in the same lexical contour.
I know it conflicts with the wording at the end of Section 5.4 which says

 > Any use of a syntax keyword before its corresponding definition is
 > an error.

but if the above rewrite is "correct", this error condition could be

Note this is not necessarily conflicting - it is still an error
to _use_ a syntax before its corresponding definition,
even in letrec-syntax.  It is not an error to write a macro
which expands into a reference to some syntax, so
long as that macro is never expanded before the syntax
is defined.


Scheme-reports mailing list