It was nice to meet up with Sterling and Thies at Franfurt, after seeing a few mentions in their blog about PHP in Parrot, I finally go to see the work.
Having got so far with my PHPSharp, toy, It was interesting to see how similar the two project codebases where, (apart from PINT is written C, and PHPSharp is in PHP). But I guess the biggest supprise was IMC, the kind of bytecode for dummies that they where generating. (If I'd seen this earlier I perhaps would not have bothered so much with a .NET target).
IMC is kind of hard to dig up on
parrotcode.org, but it's a richer IL language, that can be compiled down into parrot opcodes, so you dont have to bother understanding about register based stuff, (as it's alot simpler to work with stack based bytecodes)
I download parrot, and tryed to get the examples imc code to work, but found it segfaulted all the time, however code from PINT didnt segfault until after it had run. so Obviously the examples may not have been updated.
For the last day at the hotel (as I got stuck in Frankfurt for one extra day after the conference finished), I started changing the code in PHPSharp to parse PHP, using the PHP_Parser code I'd done before, on the premise that I could add a file Nodes.imcEmit.tc which would implement emiting IMC bytecodes from the tree that it builds..
Well I got quite a way on changing the parser to create Nodes for PHP, but didnt quite finish it.. (and I really need to change the parser to use a more classic $tokenizer->advance() approach.
But perhaps with a few more hours I might be able to get Pint's mandleprot compiled as well. :).. Although that still leaves the doors open to the two fundimental problems of both projects
a) It takes alot of time to do this compiler stuff..
b) There are lots of funky issues to sort out, like type guessing, how to implement all the existing 'PHP' functions.... and on and on..
