handling modules in packages

[Diez B. Roggisch]

Tommy Grav wrote:

> Hi,
> 
>    I am working on a package that contains a number of
> different modules:
> 
>  > ls pyAstro
> __init__.py
> constants.py
> conversion.py
> observation.py
> orbit.py
> transformation.py
> 
> however, I find that several of the modules have the
> same import statements:
> 
> orbit.py:
> 
> import numpy
> import constants
> import conversion
> import observations
> 
> observations.py:
> 
> import numpy
> import constants
> import conversions
> import transformations
> 
> The modules themselves are not overly large, but it bugs
> me to have to import numpy twice (or even more as the
> number of modules grow). Is there a way to import numpy
> once in the package (like in the __init__.py file) such that
> it is accessible to all the modules? Or is the multiple imports
> just something one has to live with?

Essentially, yes. That's the way it is, and it's better for understanding
how things work in the respective submodules.

However, you can stuff things into the __builtins__-module using setattr,
and thus make names known globally.

But it's a hack, and it means that you possibly create conflicts if
different modules have different ideas on what is supposed to live under
one key.

so - don't do it. And live with the imports. After all, that's only c'n'p,
and not of the bad kind.

Diez