Jim Fulton wrote at 2009-2-10 14:01 -0500:
On Feb 10, 2009, at 1:49 PM, Dieter Maurer wrote:
Jim Fulton wrote at 2009-2-8 13:00 -0500:
... IMO, introducing an extra is like introducing a new package and in a rather complicated way.
I agree with the first part of your sentence -- but cannot follow you with the second part:
How can "'extra' : <sequence of required distributions>" be more complicated than creating, maintaining and distributing a complete package?
Because you have to remember to test each valid permutation of the package. I bet no one does.
But that is the same when each extra is represented by an individual package. Then, too, you have to test each valid combination -- and few will do it.
Also, users have to be aware of the extras. PyPI doesn't advertise extras
I am interested in extras only when I am interested in the package itself. Of course, the package should document in some way relevant extras.
nor are there standard ways to document them.
I recently looked at documentation for a few "standard" packages on PyPI -- and apart from all using "rest" and a bit of classifiction, I could not detect much "standard" for them, too.
In general, it makes an already complicated packaging system more complicated.
That's something I do not get... An extra is a shorthand for a separate package with a few extra dependancies. Sure, shorthands are not strictly necessary. Sure, complexity increases slightly. Nevertheless, they are often valuable. -- Dieter