[Python-Dev] __import__ problems

[Guido van Rossum]

On Fri, Nov 28, 2008 at 6:56 PM, Nick Coghlan <ncoghlan at gmail.com> wrote:
> Steven D'Aprano wrote:
>> Should this be reported as a documentation bug? Given the new import
>> hooks, would it be fair to say that the main reason for __import__ is
>> to use it to import a module whose name is only known at runtime?
>
> Only known at runtime, and for some reason you want an actual module
> object rather than just the module's global namespace (since you can use
> runpy.run_module() if you only need the latter).
>
> At the very least, the __import__ docs should probably be updated to
> point to run_module() as an alternative approach, so a doc issue is
> probably a good idea.

This sounds wrong to me. run_module() runs the module each time it is
called. __import__ has all the semantics of the import statement (by
definition -- it is almost a tautology :-) in that it first tries to
see if the module is already imported. Pointing to run_module() as an
alternative just perpetuates the misunderstanding (alas fairly common
amongst casual users) about what exactly import does.

-- 
--Guido van Rossum (home page: http://www.python.org/~guido/)