I will remove them.

Please check carefully, what you are removing. There are some unused imports that can actually be removed, however, there are others which are intentionally in place (e.g. for making code elements accessible to the user from another namespace).

Thank you for the warnings. In that case retract from claiming this for myself and I leave this up for someone with more intimate knowledge of the project.

When some eager person would like to take this on, which approach would make you confident that no necessary imports are removed?

I've removed some obviously unused imports.

As a rule of thumb:

• don't change imports in pylab
• it's mostly ok to remove module imports
• removing imported methods or classes require careful consideration (in particular in widely used namespaces like matplotlib or pyplot). For other namespaces it depends if this could be considered part of the public API.

Thanks for explaining the guidelines and removing unused imports.

To add to @timhoffm 's list,

• removing standard library / third-party imports is usually ok, removing internal imports usually is not.

__all__ is just a very mild limitation. Basically, it

• excludes non-listed elements from a wildcard import
• defines the intention of a public interface (but essentially only visible to the reader of the source code)

As a result, most people won't notice the effect of __all__ anyway (you shouldn't do wildcard imports anyway and a library user does usually not check the source code).

In particular __all__ does neither prevent an explicit import nor does it exclude elements from dir() and thus completion.

For an explanation see e.g. http://xion.io/post/code/python-all-wild-imports.html