Rebinding 0 (was Re: "?:", "a and b or c" or "iif")

[Tim Peters]

[Tim]
> If the revolting "(x and [a] ..." got popular, a peephole optimizer
> could get rid of the list ops -- but only at the cost of frustrating
> Michael and Neel's attempts to change what the generated code maps
> "0" to <0.9 wink>.

[Neel Krishnaswami]
> How *would* you rebind 0 without hacking the parser?

The answer is 42, right?

def f():
    print 0 + 0

c = f.func_code
c2 = []
for attr in ('co_argcount',
             'co_nlocals',
             'co_stacksize',
             'co_flags',
             'co_code',
             'co_consts',
             'co_names',
             'co_varnames',
             'co_filename',
             'co_name',
             'co_firstlineno',
             'co_lnotab'):
    c2.append(getattr(c, attr))

c2[5] = (None, 21)

import new
f.func_code = apply(new.code, tuple(c2))

f()  # guess what this prints

Download Michael's bytecode hacks and this kind of fiddling is much easier.

> ...
> Now, I'm not asking this simply to prove my bad character, but because
> I am thinking that this might make it simpler to implement a macro
> system for Python (in Python, preferably). By that I don't mean a
> simple text preprocessor like C's, but an honest-to-god hygienic macro
> facility, because I could create symbols and manipulate symbols that
> aren't legal Python identifiers.

Michael has gone a bit in that direction already.  It's reprehensible, but I
encourage it <wink>.

> (And then all I need are continuations, and I'll be happy. Honest.)

Hmm.  Guido has all but promised he'd get a sex-change operation before he'd
add user-visible continuations to Python -- and I suspect most Python users
would find the former change less confusing.

if-python-were-scheme-there'd-be-no-need-for-either-ly y'rs  - tim