Dan Sugalski <[EMAIL PROTECTED]> wrote:Since Sam Ruby's been poking at it, and I'm waiting on the register coloring function to churn, I figured I'd poke at the python converter in CVS a bit. So I did. More ops are done, and things get a bit further on his big list 'o tests than they did before, which is nice.
pie-thon.pl was a short-term hack to get Python bytecode running. Its broken in several areas, mostly name handling and function calls. And I still think that going via Python AST is the better way.
If we want to do it right we should just go from source -- the language isn't that grammatically challenging.
I do realize this is a sub-optimal way to do it. Still, it's workable, can be done now without fiddling with parsers, and people have gotten interested enough in it to poke at it again. I don't see any reason not to do the simple stuff that's outstanding if people find it, or to apply patches to the system as it stands.
> ... My translator handles a separate set of the tests, but stillfails on some of them too. (I've not given it a full shakeout yet)
Running "make test" in languages/python is already using ast2past.py, but most of the tests are just turned off. It would be rather simple, though, to turn on either Python -> Parrot generation in lib/Parrot/Test/Python.pm.
Having one set of tests in one place would be really good.
But before continuining seriously, we need the Python PMCs working - after 0.1.1.
Absolutely, as well as any builtins that might need to be added. -- Dan
--------------------------------------it's like this-------------------
Dan Sugalski even samurai
[EMAIL PROTECTED] have teddy bears and even
teddy bears get drunk
