On Mon, Feb 16, 2015 at 7:20 PM, Cameron Simpson <c...@zip.com.au> wrote:
> One might use exec() to use code that is valid in one python version but not
> another, when you need your program to run in both i.e. to get code that is
> syntacticly invalid in one version, but to use it (conditionally) in another
> version.
>
> I only have one use case for this presently: I have a use of exec() in my
> cs.py3 python2/3 compatability module:
>
> def raise3(exc_type, exc_value, exc_traceback):
> if sys.hexversion >= 0x03000000:
> raise exc_type(exc_value).with_traceback(exc_traceback)
> else:
> # subterfuge to let this pass a python3 parser; ugly
> exec('raise exc_type, exc_value, exc_traceback')
>
> I'm using exec() here because a Python 3 interpreter will reject the 3
> argument form of raise. Elsewhere in my code I just call cs.py3.raise3()
> with the requisite arguments. Note that the string passed to exec() is
> hardwired, not obtained from elsewhere in any form.
I'd try conditional imports, first:
if sys.hexversion >= ...:
from .compat_py3 import raise3
else:
from .compat_py2 import raise3
But maybe this breaks with the setuptools pre-compilation shenanigans?
At any rate, I didn't mean to make a general statement. Obviously,
sometimes exec/eval is necessary. If for no other reason than because
sometimes the requirement is to use exec (e.g. if you are implementing
something equivalent to python -i, etc.).
-- Devin
_______________________________________________
Tutor maillist - Tutor@python.org
To unsubscribe or change subscription options:
https://mail.python.org/mailman/listinfo/tutor
