Skip to main content
Join the official 2019 Python Developers SurveyStart the survey!

Utilities for spying on function calls in unit tests.

Project description

Ever deal with a large test suite before, monkey patching functions to figure out whether it was called as expected? It’s a dirty job. If you’re not careful, you can make a mess of things. Leave behind evidence.

kgb’s spies will take care of that little problem for you.

What are spies?

Spies intercept and record calls to functions. They can report on how many times a function was called and with what arguments. They can allow the function call to go through as normal, to block it, or to reroute it to another function.

Spies are awesome.

(If you’ve used Jasmine, you know this.)

Where is kgb used?

We use kgb at Beanbag for our Review Board and RBCommons products.

If you use kgb, let us know and we’ll add you to a shiny new list on this page.

Installing kgb

Before you can use kgb, you need to install it. You can do this by typing:

$ pip install kgb


$ easy_install kgb

kgb supports Python 2.5 through 2.7 and 3.4 through 3.6.

Spying for fun and profit

Spying is really easy. There are three ways to initiate a spy.

1. Creating a SpyAgency

A SpyAgency manages all your spies. You can create as many or as few as you want. Generally, you’ll create one per unit test run. Then you’ll call spy_on(), passing in the function you want.

from kgb import SpyAgency

class TopSecretTests(unittest.TestCase):
    def test_mind_control_device(self):
        mcd = MindControlDevice()
        agency = SpyAgency()
        agency.spy_on(mcd.assassinate, call_fake=give_hugs)

2. Mixing a SpyAgency into your tests

A SpyAgency can be mixed into your test suite, making it super easy to spy all over the place, discretely, without resorting to a separate agency. (We call this the “inside job.”)

from kgb import SpyAgency

class TopSecretTests(SpyAgency, unittest.TestCase):
    def test_weather_control(self):
        weather = WeatherControlDevice()

3. Using a context manager

If you just want a spy for a quick job, without all that hassle of a full agency, just use the spy_on context manager, like so:

from kgb import spy_on

class TopSecretTests(unittest.TestCase):
    def test_the_bomb(self):
        bomb = Bomb()

        with spy_on(bomb.explode, call_original=False):
            # This won't explode. Phew.

A spy’s abilities

A spy can do many things. The first thing you need to do is figure out how you want to use the spy.

Creating a spy that calls the original function


When your spy is called, the original function will be called as well. It won’t even know you were there.

Creating a spy that blocks the function call

agency.spy_on(obj.function, call_original=False)

Useful if you want to know that a function was called, but don’t want the original function to actually get the call.

Creating a spy that reroutes to a fake function

agency.spy_on(obj.function, call_fake=my_fake_function)

Fake return values or operations without anybody knowing.

Stopping a spy operation


Do your job and get out.

Check the call history

for call in obj.function.calls:
    print(calls.args, calls.kwargs)

See how many times your spy’s intercepted a function call, and what was passed.

Check a specific call

# Check the latest call...
print obj.function.last_call.args
print obj.function.last_call.kwargs
print obj.function.last_call.return_value
print obj.function.last_call.exception

# For an older call...
print obj.function.calls[0].args
print obj.function.calls[0].kwargs
print obj.function.calls[0].return_value
print obj.function.calls[0].exception

Also a good way of knowing whether it’s even been called. last_call will be None if nobody’s called yet.

Check if the function was ever called


If the function was ever called at all, this will let you know.

Check if the function was ever called with certain arguments

# Check if it was ever called with these arguments...
self.assertTrue(obj.function.called_with('foo', bar='baz'))

# Check a specific call...
self.assertTrue(obj.function.calls[0].called_with('foo', bar='baz'))

# Check the last call...
self.assertTrue(obj.function.last_called_with('foo', bar='baz'))

The whole call history will be searched. You can provide the entirety of the arguments passed to the function, or you can provide a subset. You can pass positional arguments as-is, or pass them by name using keyword arguments.

Recorded calls always follow the function’s original signature, so even if a keyword argument was passed a positional value, it will be recorded as a keyword argument.

Check if the function ever returned a certain value

# Check if the function ever returned a certain value...

# Check a specific call...

# Check the last call...

Handy for checking if some function ever returned what you expected it to, when you’re not calling that function yourself.

Check if a function ever raised a certain type of exception

# Check if the function ever raised a certain exception...

# Check a specific call...

# Check the last call...

You can also go a step further by checking the exception’s message.

# Check if the function ever raised an exception with a given message...
    "'type' object is not iterable"))

# Check a specific call...
    "'type' object is not iterable"))

# Check the last call...
    "'type' object is not iterable"))

Reset all the calls


Wipe away the call history. Nobody will know.

Call the original function

result = obj.function.call_original('foo', bar='baz')

Super, super useful if you want to use call_fake= to wrap a function and track or influence some part of it, but still want the original function to do its thing. For instance:

stored_results = []

def my_fake_function(*args, **kwargs):
    kwargs['bar'] = 'baz'
    result = obj.function.call_original(*args, **kwargs)

    return result

agency.spy_on(obj.function, call_fake=my_fake_function)


Doesn’t this just do what mock does?

kgb’s spies and mock’s patching are very different from each other. When patching using mock, you’re simply replacing a method on a class with something that looks like a method, and that works great except you’re limited to methods on classes. You can’t override something a top-level function, like urllib2.urlopen.

kgb spies leave the function or method where it is. What it does do is replace the bytecode of the function, intercepting calls on a very low level, recording everything about it, and then passing on the call to the original function or your replacement function. It’s pretty powerful, and allows you to listen to or override calls you normally would have no control over.

What?! There’s no way that’s stable.

It is! It really is! We’ve been using it for years across a wide variety of codebases. It’s pretty amazing.

Python actually allows this. We’re not scanning your RAM and doing terrible things with it, or something like that. Every function or method in Python has a func_code (Python 2) or __code__ (Python 3) attribute, which is mutable. We can go in and replace the bytecode with something compatible with the original function.

How we actually do that, well, that’s complicated, and you may not want to know.

Does this work with PyPy?

I’m going to level with you, I was going to say “hell no!”, and then decided to give it a try.

Hell yes! (But only accidentally. YMMV… We’ll try to officially support this later.)

What else do you build?

Lots of things. Check out some of our other open source projects.

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Files for kgb, version 4.0
Filename, size File type Python version Upload date Hashes
Filename, size kgb-4.0-py2.7.egg (59.6 kB) File type Egg Python version 2.7 Upload date Hashes View hashes
Filename, size kgb-4.0-py2.py3-none-any.whl (27.5 kB) File type Wheel Python version py2.py3 Upload date Hashes View hashes
Filename, size kgb-4.0-py3.4.egg (63.3 kB) File type Egg Python version 3.4 Upload date Hashes View hashes
Filename, size kgb-4.0-py3.5.egg (63.2 kB) File type Egg Python version 3.5 Upload date Hashes View hashes
Filename, size kgb-4.0-py3.6.egg (62.1 kB) File type Egg Python version 3.6 Upload date Hashes View hashes
Filename, size kgb-4.0-py3.7.egg (62.1 kB) File type Egg Python version 3.7 Upload date Hashes View hashes
Filename, size kgb-4.0.tar.gz (33.5 kB) File type Source Python version None Upload date Hashes View hashes

Supported by

Elastic Elastic Search Pingdom Pingdom Monitoring Google Google BigQuery Sentry Sentry Error logging AWS AWS Cloud computing DataDog DataDog Monitoring Fastly Fastly CDN SignalFx SignalFx Supporter DigiCert DigiCert EV certificate StatusPage StatusPage Status page