os.path.join with the import statement

Question

I would like to simplify as much as possible the syntax and the number of lines, so I was wondering why the following statement is an erro:

from os.path.join(path1,path2,filename) import *

the os.path.join seems not to run with the import statement.

Is there any workaround?

That's not how import statements work. You need to import the method, and then run it. In this case, that could look like `from os import path` then `path.join(path1,path2,filename)` — Shadow, Jul 22 '19 at 22:41
Python doesn’t import from paths. It imports from module names. — deceze, Jul 22 '19 at 22:47
The question makes sense, and should not be downvoted as it is being. — jsbueno, Jul 22 '19 at 23:28

score 2 · Answer 1 · edited Jul 22 '19 at 23:28

Python's import statement needs a "module name", not a file path to work. The way to import using a variable is to use the importlib.import_module built-in function. You have, of course, to use your filesystem "path" directory names to package names, and the top most of which must be "visible" in the PythonPath. The [:-3] part is just there to strip the .py part, but you can just use the modulename without the ".py" to startwith.

from importlib import import_module

module = importmodule('.'.join([path1, path2, filename[:-3]])
for attr in dir(module):
    globals()[attr] = getattr(module, attr)

However, as you see, you will populate your namespace with names you don't know which they are, and thus, you can't have program code using those names - what makes it kind of pointless. It may be more useful to simply create a new dictionary with the attributes of your imported module:

namespace = {}

for attr in dir(module):
    namespace[attr] = getattr(module, attr)

so you can introspect what was imported and call/use those values.

Now, if your "path1", and "path2" are not a package, and you really want to just import module from an arbitrary location, what is needed is to have the location inserted in the Pythonpath bedore calling import_module. The Pythonpath, however, is a simple list located at sys.path - just insert the full-folder path there (and pop it later if you want):


def import_path(path1, path2, module_name, namespace=None):
    import sys
    from importlib import import_module
    from pathlib import Path # new code should use this instead of "os.path" stuff
    path = Path(path1).joinpath(path2)
    name = module_name[-3:] if module_name.endswith(".py") else module_name
    try:
        sys.path.insert(0, str(path))
        module = import_module(name)
    finally:
        sys.path.remove(str(path))
    if namespace is None:
        namespace = sys._getframe().f_back.f_globals()
    for name in dir(module):
        namespace[name] = getattr(module, name)

(One can also use directly the __import__ function without needing to to use importlib.import_module, but its use is not advised in the documentation. Among other things when using the single argument form __import__("path1.path2.module") will return path1 not module.)

os.path.join with the import statement

1 Answers1