Re: [ANN] LuLPeg v0.1

[Pierre-Yves Gérardy <pygy79@...>]

On Thu, Jun 13, 2013 at 2:43 PM, Philippe Lhoste <PhiLho@gmx.net> wrote:
> lpegj seems interesting too (thanks!). LuLPeg might be more compatible and
> up-to-date.
> It can be fun to compare their respective speeds (if speed is a concern, of
> course).

LPegJ targets LPeg 0.10, while LuLPeg targets v0.12.

The LuaJ metamethod limitation `p + 1` working, but `1 - p` not
working has been solved in the mean time, so I'd expect it to be,
compatible with the original.

It may be fun, for someone with the know-how, to write a backend for
LuLPeg that compiles patterns to Java bytecode.

I'm currently cleaning up the code, once I'm done, I'll be focusing on
the factorizer/analyzer part of the frontend, in order to provide the
simplest patterns possible to the backend.

>From there, the plan is to

1. implement tail calls, so that {1 * V(1) + 0 * V(1)} doesn't grow the stack.
2. refactor the capture code to reduce allocation (currently one table
per capture :-/)
3. rewrite the compiler.
4. UTF-8 support, via P8(), S8() and R8() functions that will live
alongside the byte-sized ones.

-- Pierre-Yves