Develop Reference

How can I lazily import a module in python?

I have classes which require dependencies in order to be instantiated but are otherwise optional. I'd like to lazily import the dependencies and fail to instantiate the class if they aren't available. Note that these dependencies are not required at the package level (otherwise they'd be mandatory via setuptools). I currently have something like this:
class Foo:
def __init__(self):
try:
import module
except ImportError:
raise ModuleNotFoundError("...")
def foo(self):
import module
Because this try/except pattern is common, I'd like to abstract it into a lazy importer. Ideally if module is available, I won't need to import it again in Foo.foo so I'd like module to be available once it's been imported in __init__. I've tried the following, which populates globals() and fails to instantiate the class if numpy isn't available, but it pollutes the global namespace.
def lazy_import(name, as_=None):
# Doesn't handle error_msg well yet
import importlib
mod = importlib.import_module(name)
if as_ is not None:
name = as_
# yuck...
globals()[name] = mod
class NeedsNumpyFoo:
def __init__(self):
lazy_import("numpy", as_="np")
def foo(self):
return np.array([1,2,])
I could instantiate the module outside the class and point to the imported module if import doesn't fail, but that is the same as the globals() approach. Alternatively lazy_import could return the mod and I could call it whenever the module is needed, but this is tantamount to just importing it everywhere as before.
Is there a better way to handle this?

Pandas actually has a function import_optional_dependency which may make a good example (link GitHub) as used in SQLAlchemyEngine (link GitHub)
However, this is only used during class __init__ to get a meaningful error (raise ImportError(...) by default!) or warn about absence or old dependencies (which is likely a more practical use of it, as older or newer dependencies may import correctly anywhere if they exist, but not work or be explicitly tested against or even be an accidental local import)
I'd consider doing similarly, and either not bother to have special handling or only do it in the __init__ (and then perhaps only for a few cases where you're interested in the version, etc.) and otherwise simply import where needed
class Foo():
def __init__(self, ...):
import bar # only tests for existence
def usebar(self, value):
import bar
bar.baz(value)
Plausibly you could assign to a property of the class, but this may cause some trouble or confusion (as the import should already be available in globals once imported)
class Foo():
def __init__(self, ...):
import bar
self.bar = bar
def usebar(self, value):
self.bar.baz(value)

Gave it a quick test with a wrapper, seems to work fine:
def requires_math(fn):
def wrapper(*args, **kwargs):
global math
try:
math
except NameError:
import math
return fn(*args, **kwargs)
return wrapper
#requires_math
def func():
return math.ceil(5.5)
print(func())
Edit: More advanced one that works with any module, and ensures it is a module in case it's been set to something else.
from types import ModuleType
def requires_import(*mods):
def decorator(fn):
def wrapper(*args, **kwargs):
for mod in mods:
if mod not in globals() or not isinstance(globals()[mod], ModuleType):
globals()[mod] = __import__(mod)
return fn(*args, **kwargs)
return wrapper
return decorator
#requires_import('math', 'random')
def func():
return math.ceil(random.uniform(0, 10))
print(func())

Testing with Optional Import in Python

I have a library that can use two modules; one is fast but available only on Linux and macOS, the other is slower but it's multi-platform. My solution was to make the library compatible with both and have something like the following:
try:
import fastmodule
except ImportError:
import slowmodule
Now I want to compare the timing of the library when using either module. Is there any way of masking the fastmodule without changing the source code (i.e. within a Jupyter Notebook), in an environment where both modules are installed, so that the slowmodule is used?

This is a bit of a hack, but here it goes:
You can write your own importer and register it (note that this is Python 3-specific, Python 2 had another API for this):
import sut
import functools
import importlib
import sys
def use_slow(f):
#functools.wraps(f)
def wrapped(*args, **kwargs):
ImportRaiser.use_slow = True
if 'fastmodule' in sys.modules:
del sys.modules['fastmodule'] # otherwise it will remain cached
importlib.reload(sut)
f(*args, **kwargs)
return wrapped
def use_fast(f):
#functools.wraps(f)
def wrapped(*args, **kwargs):
ImportRaiser.use_slow = False
importlib.reload(sut)
f(*args, **kwargs)
return wrapped
class ImportRaiser:
use_slow = False
def find_spec(self, fullname, path, target=None):
if fullname == 'fastmodule':
if self.use_slow:
raise ImportError()
sys.meta_path.insert(0, ImportRaiser())
#use_fast
def test_fast():
# test code
#use_slow
def test_slow():
# test code
Here, sut is your module under test, which you have to reload in order to change the behavior. I added decorators for readabilty, but this can be done by some function or in the test setup, of course.
If you use the slow version, fastmodule will raise ImportError on import and slowmodule will be used instead. In "fast" case, all works as usual.

Can't Pickle Wrapper Class Object

All, I'm trying to parallelize some code using multiprocessing, and I've stripped my code down such that commenting the line of my main class
self.obs = Observer(self.guess)
causes the system to run error free. If I write
obs = Observer(self.guess)
It works, but if I write self.obs, I get
TypeError: can't pickle Observer objects
Here is the entire class I'm trying to import. It is just a wrapper for ephem.Observer, which had the same error when importing.
import numpy as np
from req import SETTINGS
from req.helpers import load_db, pack_into_vector, create_observer
class Observer:
def __init__(self, beta=np.zeros((2,))):
self.observer = create_observer(beta)
return
def __getstate__(self):
return {'observer': self.observer}
The error occurs on p.start() where
p = Process(target=selector,args=(first_guess, recording_queue, guess_queue))

I actually solved this. I guess the issue was with all of the self.___ methods/attributes of my selector class. I created an additional class, selector_wrapper, with an init method that created and ran the selector class. This worked perfectly.
In summary, creating a wrapper class with no methods other than init, and with no attributes, fixed the problem!

how to break circular import in python

I have a circular import problem in my python code. The decorator from one module need to call a function in another module. And this function is using the decorator which is imported from previous module. The error message shows can't import name xxxx
my code structure:
__init__.py
worker1.py
worker2.py
...
and there is a decorator in init.py
def defer():
def wrapper(func):
def wrapper_f(*args, **kargs):
some logic here...
new_func = eval("%s.apply_async" % func.func_name) # I use celery to do async call here
new_func(args = args, kwargs = kargs)
wrapper_f.__name__ = func.func_name # This is to allow celery to load different task based on function name
return wrapper_f
return wrapper
In worker1.py
#celery.task
#defer
def task1(*args, **kargs):
some logic
It will have circular import in __init__.py and worker1.py, since eval("%s.apply_async" % func.func_name) need to load function in worker1, and worker1 need to load decorator in __init__. How can I solve this circular import problem?
Btw, I can't move the decorator function into worker1.py, bcz it is needed by other workers in the same folder.

How can I log all imports that are executed in a block of code?

I'm writing a test suite, and the code I'm testing makes excessive use of delayed module imports. So it's possible that with 5 different inputs to the same method, this may end up importing 5 additional modules. What I'd like to be able to do is set up tests so that I can assert that running the method with one input causes one import, and doesn't cause the other 4.
I had a few ideas of how to start on this, but none so far have been successful. I already have a custom importer, and I can put the logging code in the importer. But this doesn't work, because the import statements only run once. I need the log statement to be executed regardless of if the module has been previously imported. Just running del sys.modules['modname'] also doesn't work, because that runs in the test code, and I can't reload the module in the code being tested.
The next thing I tried was subclassing dict to do the monitoring, and replace sys.modules with this subclass. This subclass has a reimplemented __getitem__ method, but calling import module doesn't seem to trigger the __getitem__ call in the subclass. I also can't assign directly to sys.modules.__getitem__, because it is read-only.
Is what I'm trying to do even possible?
UPDATE
nneonneo's answer seems to only work if the implementation of logImports() is in the same module as where it is used. If I make a base test class containing this functionality, it has problems. The first is that it can't find just __import__, erroring with:
# old_import = __import__
# UnboundLocalError: local variable '__import__' referenced before assignment
When I change that to __builtin__.__import__, I another error:
myunittest.py:
import unittest
class TestCase(unittest.TestCase):
def logImports(self):
old_import = __builtins__.__import__
def __import__(*args, **kwargs):
print args, kwargs
return old_import(*args, **kwargs)
__builtins__.__import__ = __import__
test.py:
import myunittest
import unittest
class RealTest(myunittest.TestCase):
def setUp(self):
self.logImports()
def testSomething(self):
import unittest
self.assertTrue(True)
unittest.main()
# old_import = __builtins__.__import__
# AttributeError: 'dict' object has no attribute '__import__'

Try
old_import = __import__
def __import__(*args, **kwargs):
print args, kwargs
return old_import(*args, **kwargs)
__builtins__.__import__ = __import__
This overrides __import__ completely, allowing you to monitor every invocation of import.

Building on the previous answer, in Python 3, I've had success with the following.
import builtins
old_import = __import__
def importWithLog(*args, **kwargs):
print(args[0]) # This is the module name
print(args, kwargs)
return old_import(*args, **kwargs)
builtins.__import__ = importWithLog
# Rest of the code goes here.
import time
import myModule
...