(I've just finished the pretty printing part in Pugs, so I'll use actual command line transcripts below. The leading "?" does not denote boolean context -- it's just telling pugs to do a big-step evaluation. Also, boolean literals are written in their Scheme forms.)
In S06, the meaning of chaining comparison operators is defined as a
derived form:
(a < b < c) ==> (a < b) and (b < c)
With the note that "b" must be evaluated at most once. However, if
taken literally, it gives this rather weird result:
pugs> ? 2 < (0 | 3) < 4
(#t|#t)
My intuition is that it should be equivalent to this:
pugs> ? (2 < 0 < 4) | (2 < 3 < 4)
(#f|#t)
That is, the autothreading should operate on the whole comparison chain,
treating it as a large variadic function with short-circuiting semantics.
Is this perhaps "saner" than the blind rewrite suggested in the spec?
Also, consider this:
pugs> ? 1|2 => 3|4
(((1 => 3)|(1 => 4))|((2 => 3)|(2 => 4)))
Since junctions are documented to only "flatten" on boolean context,
here the pair-building arrow has been autothreaded. Is it the intended
semantic? What about the list-building semicolon?
Thanks,
/Autrijus/
pgp876hJTXK2E.pgp
Description: PGP signature
