[Distutils] namespace packages

[P.J. Eby]

At 04:36 PM 4/22/2010 +0900, David Cournapeau wrote:
>On Thu, Apr 22, 2010 at 1:10 PM, P.J. Eby <pje at telecommunity.com> wrote:
> > At 10:18 AM 4/22/2010 +0900, David Cournapeau wrote:
> >>
> >> One problem with the setuptools implementation is
> >> that several packages sharing the same namespace have files in common,
> >
> > If that were actually true (it isn't), then it would be considered a bug in
> > setuptools.
>
>I am confused, then, by the setuptools doc:
>
>Note, by the way, that your project's source tree must include the
>namespace packages' __init__.py files (and the __init__.py  of any
>parent packages), in a normal Python package layout. These __init__.py
>files must contain the line:

If you read two paragraphs further in that doc, you'll see where it 
goes on to say:

"You must NOT include any other code and data in a namespace 
package's __init__.py. Even though it may appear to work during 
development, or when projects are installed as .egg files, it will 
not work when the projects are installed using "system" packaging 
tools -- in such cases the __init__.py files **will not be 
installed**, let alone executed. "  (emphasis added)


>and because that's in the rationale of Pep 382. If this is not true,
>we need to update our own instructions because we advise to write such
>__init__.py in scikits namespace in the scipy community.

You still need to write them, for egg-based installs.  But when 
packaged for *system* installation, they aren't going to be there.


> >> (setuptools namespace implementation depends on pkg_resources, which is a
> >> complicated and quite slow piece of code).
> >
> > Slow at doing what, precisely, and slower compared to what alternative?
>
>Merely importing it is already quite slow,

I'd be interested in knowing under what conditions that's the case 
(e.g. how many .egg zipfiles vs. how many .egg subdirs); it might be 
possible to improve it.