At 06:04 PM 7/22/2009 -0700, Sridhar Ratnakumar wrote:
On Wed, 22 Jul 2009 12:14:11 -0700, P.J. Eby <[email protected]> wrote:
t 12:08 PM 7/22/2009 -0700, Sridhar Ratnakumar wrote:
On Tue, 21 Jul 2009 21:23:00 -0700, P.J. Eby <[email protected]>
wrote:
If I were trying to do what you're doing, I'd just run a bdist_egg on
the package and inspect it via the API. Or, alternately, I'd run the
egg_info command with an explicit target directory, then load it via the
regular API. (By "regular API" I mean such things as a
find_distributions() call aimed at the parent of the .egg-info
directory.)
And I'd do it that way because then I wouldn't need to care whether the
distribution was created with setuptools or not.
How can you run egg_info ('python setup.py egg_info') on a distribution
that does not use setuptools?
By calling setuptools.sandbox.run_setup() with the appropriate
arguments. (When setuptools is imported, it monkeypatches the
distutils, such that distutils is really setuptools. That's how
easy_install is able to build eggs from packages that don't use
setuptools.)
Ok, anyways .. running setup.py can be a pain (eg: running egg_info on
numpy, using the sandbox, actually tries to build the extensions!).
For now, I'll go with a custom parser (like pip) .. and wait for either
you or Tarek to split setuptools into reusable modules.
Well, all the lower-level tools you need to do the parsing are in
pkg_resources also; the code actually used is just:
dm = self.__dep_map = {None: []}
for name in 'requires.txt', 'depends.txt':
for extra,reqs in split_sections(self._get_metadata(name)):
if extra: extra = safe_extra(extra)
dm.setdefault(extra,[]).extend(parse_requirements(reqs))
return dm
Both split_sections() and parse_requirements() are public functions,
and you can replace the _get_metadata call with a text-mode open() of
the relevant file, if it exists.
_______________________________________________
Distutils-SIG maillist - [email protected]
http://mail.python.org/mailman/listinfo/distutils-sig
