[Python-Dev] PEP 269 will live again.

[Guido van Rossum]

> I move to move PEP 269 (pgen module for Python) to zombie monster status
> and refer all interested parties to my post today in the parser-sig for
> more details.  Apparently I am only interested in parser generators in the
> month of August (PEP 269 was drafted in Aug.2001).

I suppose you're referring to this message:

http://mail.python.org/pipermail/parser-sig/2002-August/000010.html

I have not retired your PEP and am glad you're interested in this
subject again.  Let's try to reach a conclusion before August is over.

I don't think you should try to tell the Jython folks what to do.  A
pgen module that only works in CPython is still valuable.  If you want
to port pgen to Jython and support it as a module, that's fine, but I
don't think you should try to get the Jython developers to use pgen as
their parser.  After all, Jython's *implementation* is *supposed* to
be Javaesque.

Are you interested in implementing PEP 269 as it currently stands?
Then fine, let's do it and get it into Python 2.3.

If you want to expand the scope, I predict that it'll never happen, so
then let's retire the PEP.  It's up to you.

Note that Jeremy has a new Python compiler package (Lib/python/ in the
Python 2.3 CVS tree), which currently uses parse trees as produced by
the old 'parser' module as input, and then restructures them into more
abstract syntax trees.  This compiler is easily retargetable to other
input and output structures though -- I believe Finn Bock already has
a Jython version of it.  I don't know what it generates, I doubt it
generates CPython bytecode, maybe it generates Java source or JVM
assembler; I believe it takes the same parse tree that Jython uses as
input.

I think it would be useful if you use the same form of abstract syntax
trees as Jeremy's parser uses (not the parser module output, but the
restructured abstract syntax trees); I think they are quite flexible
and useful.

If you don't want to do this, you'll have to motivate why your
alternative is better, and also show how Jeremy's compiler package can
be easily adapted to use your form of parse trees.

--Guido van Rossum (home page: http://www.python.org/~guido/)