Please forgive me if these ideas have been discussed before. I don't
remember having read them elsewhere.
For specifying in-rule repetitions, why not use the rule modifer we
already have for specifying whole-rule repetitions; namely, C<:x>. Allow
:x inside rules like :i and :w, and we get something like this:
rx :w/ three m's\: [:3x m] /
rx :w/ three to five m's\: [:x(3..5) m] /
rx :w/ done at runtime\: [:x($m..$n) m] /
It seems straightforward to me, but I admit I don't know how difficult
that would be for parsing/compiling purposes. However, it seems too
different from :i and :w on some level, to be treated the same.
So instead of a match modifier, how about adding a special named
assertion akin to C<before> and C<after> called C<x>, that takes an
argument:
rx :w/ three m's\: <x(3) m> /
rx :w/ three to five m's: <x(3..5) m> /
rx :w/ done at runtime\: <x($m..$n) m> /
Repitition is a kind of assertion, after all, and it seems like it
should get to play in the same angle-bracket sandbox as the other
assertions.
There's already precedent for using C<x> for repetition as a normal
operator, and now we have the rule modifier as well, so why not extend
it into the rule innards?
I confess that I've purposely left out a small detail, and I'm sure
you've all noticed: non-greedy matching. I couldn't come up with a way
that looked all that nice. I don't know if putting the C<?> after the
argument would even work:
rx :w/ non-greedy m's\: <x(3..5)? m> /
And putting it before the argument (as part of the assertion name?) just
looks weird:
rx :w/ non-greedy m's\: <x?(3..5) m> /
So maybe this proposal is a complete no-go because of non-greediness,
but I thought I'd throw it out there anyway because it seemed to me to
fit in with assertions rather well, and I like the generalized assertion
idea.
Kurt
