[Scheme-reports] editorial awkwardness for syntax-rules pattern-matching

[Per Bothner]

If there is a revision or successor to r7rs, I suggest changing the really
awkward specification of syntax-rules pattern-matching with ellipsis in 4.3.2:

For example:

P is of the form (P_1 . . . P_k P_e ellipsis P_m+1 ... P_n)
where E is a proper list of n elements, the first
k of which match P_1 through P_k , respectively, whose
next m−k elements each match P_e, whose remaining
n−m elements match P_m+1 through P_n

The phrase 'of the form (P_1 ... P_k P_e ellipsis P_m+1 ... P_n)'
doesn't really make sense since m and n come out of nowhere.

Much clearer to write:

P is of the form (P_1 . . . P_k P_e ellipsis P_k+1 ... P_k+l)
where E is a proper list of n elements, the first
k of which match P_1 through P_k , respectively, whose
next n−k-l elements each match P_e, whose remaining
l elements match P_k+1 through P_k+l

Similarly for other items.

[John Cowan]

Post by Per Bothner
P is of the form (P_1 . . . P_k P_e ellipsis P_k+1 ... P_k+l)
where E is a proper list of n elements, the first
k of which match P_1 through P_k , respectively, whose
next n−k-l elements each match P_e, whose remaining
l elements match P_k+1 through P_k+l

I don't see where the aditional clarity comes in; also, having both 1 and l
in a formula is asking for trouble.

[Per Bothner]

Post by John Cowan

Post by Per Bothner
P is of the form (P_1 . . . P_k P_e ellipsis P_k+1 ... P_k+l)
where E is a proper list of n elements, the first
k of which match P_1 through P_k , respectively, whose
next n−k-l elements each match P_e, whose remaining
l elements match P_k+1 through P_k+l

I don't see where the aditional clarity comes in;

Well, in my opinion 'P is of the form (P_1 . . . P_k P_e ellipsis P_m+1 ... P_n)'
is somewhat nonsensical. Where do m and n come from? n is the length of E, but m
is complicated to describe as it depends on both n and the number of patterns
in P following the ellipsis. This seems a rather unusual way to enumerate items in
a list.

Using l (or m) for the number of sub-patterns following the ellipsis is much
easier to understand: l is just the number of sub-patterns following the ellipsis,
and this is obvious from the form P_k+1 ... P_k+l.

Post by John Cowan
also, having both 1 and l in a formula is asking for trouble.

That occurred to me. Using m in place of l is one option.

Even better if we're willing to use other letters beside P for sub-patterns:

P is of the form (P_1 . . . P_k Q ellipsis R_1 ... R_m)
where E is a proper list of n elements, the first
k of which match P_1 through P_k , respectively, whose
next n−k-m elements each match Q, whose remaining
m elements match R_1 through R_m

[John Cowan]

Post by Per Bothner
P is of the form (P_1 . . . P_k P_e ellipsis P_k+1 ... P_k+l)
where E is a proper list of n elements, the first
k of which match P_1 through P_k , respectively, whose
next n−k-l elements each match P_e, whose remaining
l elements match P_k+1 through P_k+l

I don't see where the aditional clarity comes in; also, having both 1 and l
in a formula is asking for trouble.

[Vassil Nikolov | Васил Николов]

Post by Per Bothner
...
P is of the form (P_1 . . . P_k P_e ellipsis P_m+1 ... P_n)
where E is a proper list of n elements, the first
k of which match P_1 through P_k , respectively, whose
next m−k elements each match P_e, whose remaining
n−m elements match P_m+1 through P_n
The phrase 'of the form (P_1 ... P_k P_e ellipsis P_m+1 ... P_n)'
doesn't really make sense since m and n come out of nowhere.

In fact, m and n are defined by that
phrase (with an implicit "where m and n
are such that"); pretty usual practice in
my view.

---Vassil.