[Python-Dev] PEP 365 (Adding the pkg_resources module)
Phillip J. Eby
pje at telecommunity.com
Thu Mar 20 22:51:59 CET 2008
At 08:34 PM 3/20/2008 +0000, Paul Moore wrote:
>I then went on to say that putting dependency information in setup.exe
>and expecting users to use automatic dependency resolution encourages
>developers to omit dependency details from documentation (to an extent
>I can't quantify, but I believe is non-zero). That lack of
>documentation "forces" me to rely on the automatic process. THAT is
>the thing that removes my choice, not easy_install's ability to skip
>dependency checking.
Ah. Fair enough. So, if we get PyPI to display that information,
that should fix this problem for you?
>People are starting to omit distributing
>bdist_wininst installers in favour of eggs only.
You mean, they're shipping a .win32.egg, but not an .exe?
> And you cannot (to my
>knowledge) convert an egg into a bdist_wininst installer,
Not at the moment, no. It seems like it ought to be *possible*,
though, since the reverse translation can be done. Eggs are more
restrictive in what they can include, so the reverse step actually
ought to be relatively easy. Indeed, I would think that it could be
done by a standalone tool without even using setuptools. All that
really needs to happen (I believe) is that the zipfile directory
needs all its names prepended with PURELIB or PLATLIB, and then add
the appropriate prefix .exe and bdist_wininst extra data on the front
of the restructured zip file.
In fact, it should probably be possible to write such a tool by
subclassing the distutils bdist_wininst command and overriding the
run() and get_inidata() methods, using the existing create_exe()
method to do that part of the magic.
The other tool that would be handy to have, would be one that unpacks
eggs into standard distutils-style installation.
> > Personally, I'm not very thrilled with the number of complaints on
> > this thread that could be resolved by RTFMing.
>...
>Honestly, I'm trying to help improve (by my measure of improvement,
>certainly) setuptools. I've done as much (more!) homework as I feel is
>appropriate (no, I haven't studied the whole manual all the way
>through). Being treated as if it's my fault, and I haven't done
>enough, is both discouraging and to be honest, somewhat offensive.
My comment wasn't aimed specifically at you; you're only one of many
people today who have appeared to state that something or other
wasn't possible or documented, described optional behavior as
required, etc. Addressing each and every one point by point looks
petty, but then lumping them together like that makes it look like
I'm picking on you specifically. Sorry about that.
In any event, I'm not saying that anyone hasn't done enough or that
it's their fault. The fact that I'm not thrilled about some of the
things said in the thread doesn't somehow magically invalidate other
people's frustrations, nor was it my intent to accuse you (or anyone)
of making up their problems. I'm just expressing *my* frustration.
More information about the Python-Dev
mailing list