plugin development best practices

[Paul Boddie]

On 22 Feb, 16:13, "Diez B. Roggisch" <d... at nospam.web.de> wrote:
>
> Darn. You're right of course - I just got the basic idea, and formed in my
> mind the "get the modules filename, thus the path, glob over it for *py,
> and thus get the subsequent module names"-pattern. Which is trivial of
> course, but not as trivial as just dir(module)

The __init__.py file of a plugins package could contain something like
this:

# Start
def init():
    import os
    global __all__
    this_dir = os.path.split(__file__)[0]
    py_suffix = os.path.extsep + "py"
    __all__ = []
    for filename in os.listdir(this_dir):
        if os.path.isdir(os.path.join(this_dir, filename)) and \
            os.path.exists(os.path.join(this_dir, filename, "__init__"
+ py_suffix)):
            __all__.append(filename)
        else:
            module, suffix = os.path.splitext(filename)
            if suffix == py_suffix and module != "__init__":
                __all__.append(module)
init()
del init
# End

This should populate the __all__ attribute of the package with then
names of any submodules or subpackages. Although that in itself won't
provide the names of the plugins via the dir function, the __all__
attribute is some kind of standard, and things like "from plugins
import *" will import all the known plugins. In fact, if you add such
an import statement to the end of the above code, you'll get all the
names of the plugins stored within the package (and thus returned by
the dir function) because the submodules and subpackages will actually
have been imported. Even reloading the plugins package will update the
__all__ attribute, although things like the unloading or removal of
plugins might be challenging in a solution where such things are
automatically imported.

Having a variation of the above function in the standard library could
be fairly useful, I suppose.

Paul