Re: [Distutils] conventions or best practice to choose package names?

Le 29/06/2012 17:59, Paul Moore a écrit :

...

I agree that "foo.ext.bar" seems nice... but it uses 3 namespace levels. It breaks the "avoid deep nesting" rule, which recommends not more than two levels. So the PEP can't recommend this pattern, or we will also have to reconsider the "avoid deep nesting" rule. ??? I thought the PEP was about distribution naming (i.e., things

On 29 June 2012 14:05, Benoît Bryon wrote: published on PyPI.

It is about naming, limited to packaging, but not limited to things published on PyPI.

If the package "foo" is published under that name on PyPI, I see no reason it can't (good taste permitting) use multi-level names like foo.bar.baz internally. That's a completely different question than whether a distribution named foo.bar.baz should be available for download from PyPI (presumably alongside foo.bar.bip and foo.bop.boop...)

I understood the foo.ext.bar example as: * foo is a namespace package * foo.ext is a namespace package * foo project invites users to publish contributions as foo.ext.* If foo is a project without namespace packages, then, yes, the PEP proposal doesn't cover its internal code organization.

Maybe I'm misunderstanding the conversation here. We need to be *extremely* precise about the differences between distributions (things published to somewhere like PyPI, which can be installed independently, and have a version number, etc) and packages (which are things you import in Python). I've been just as vague as everyone else in this thread, so I'm not trying to blame anyone. But things are definitely getting muddled.

The terminology reference used in the PEP proposal is https://bitbucket.org/benoitbryon/cpython/src/doc-package-names/Doc/packagin... (which is http://docs.python.org/dev/packaging/introduction.html + some changes related to the PEP proposal). The PEP proposal deals with : * the "name" argument passed to packaging.core.setup() or setuptools.setup(), i.e. project name * names of packages or modules being distributed, as a consequence of the "use a single name" rule Benoit