Python – 'import' or pass modules as parameters?

Question

I was wondering if passing module objects to a class or object which requires it rather than using import might not be a bad idea, as it avoids hard dependencies on those modules.

I was wondering if someone more aware of the Python Zen than I might be able to explain why this is a terrible/not terrible idea?

Asked By: Magnus

||

Source

Answer 1

What you’re talking about is called dependency injection and is considered a good practice for making your code testable. I don’t think there’s anything about Python that would make it unPythonic or a bad practice.

There are other ways you could do it in Python, for example by importing different modules depending on some kind of flag you pass in:

class Foo(object):
    def __init__(self, testing=False):
        if testing:
            import module_test as module
        else:
            import module
        self.module = module

But passing a reference to the module you wish to use is more flexible, separates concerns better, and is no less Pythonic than passing a reference to a class or instance (or string or integer) you wish to use.

For the ordinary (non-test) use case, you can use a default argument value:

class Foo(object):
    def __init__(self, module=None):
        if not module:
            import module
        self.module = module

Answered By: kindall

Answer 2

If the class in question can use one of several different modules, then it could make sense to pass in the one you want it to use.

In every situation I have encountered so far, however, the modules needed are certain, so using import would be preferred.

Answered By: Ethan Furman

Answer 3

I’m late for the game, but I have just learned a way to do this and wanted to share:

chosen_config='moduleName' #string, relative path. example: 'configurations.local.logistic'. You would take this from input, or the command line, or something like that
import importlib
config=importlib.import_module(chosen_config,package='myPackage')
#...do stuff...
data=pd.read_csv(config.dataFile)
#...more stuff...
save(config.outputFile)

Is equivalent to:

import moduleName as config

But I let the user (aka Future Me or someone I trust) decide which module to import.

Use case (my situation): You want to run a series of experiments with a sequence of different settings that will affect several scripts. You have stored the settings for each experiment in a dedicated config file. You want to run all the tasks without modifying your code.

Answered By: alex

Python – 'import' or pass modules as parameters?

Question:

Answers: