On May 11, 2012, at 11:35 AM, Eduardo Ochs wrote:
> Hello list,
>
> I am trying to reimplement in LPeg a recursive-descent parser that I
> wrote by hand, and - for some reasons that I'll explain in a next
> message if it turns out that my doubts are not too dumb - I had to use
> lpeg.Cmt(f) in several of my sub-patterns, and I realized that my work
> would be much easier if lpeg.Cp accepted a pattern as an argument,
> i.e., if instead of
>
> lpeg.C(patt) the match for patt plus all captures made by patt
> lpeg.Cp() the current position (matches the empty string)
>
> we had:
>
> lpeg.C(patt) the match for patt plus all captures made by patt
> lpeg.Cp(patt) if patt is nil, then the current position;
> if patt is a pattern, then the position after
> matching patt (if the match succeeds), plus all
> captures made by patt
>
> Anyone knows a way to simulate the behavior I'm suggesting for lpeg.Cp
> when its `patt' is not nil?
>
> Cheers and thanks in advance,
> Eduardo Ochs
>
eduardoochs@gmail.com
>
http://angg.twu.net/
>
>
> P.S.: I _think_ I know two ways for doing that: one is a very, very,
> _very_ ugly hack using the fact that `patt / f' is supposed to
> evaluates f as late as possible (modulo optimizations), and the other
> way involves running table.pack on the captures of patt*lpeg.Cp(),
> then returning the last value in that table plus the unpacking of the
> rest... But there must be better ways! =|
>
