Skip to content


Subversion checkout URL

You can clone with
Download ZIP
A Random and Combinatorial Testing framweork, compatible with PyUnit, nose and py.test
Pull request Compare This branch is 55 commits ahead of dbravender:master.
Fetching latest commit...
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.




This framework does Random and Combinatorial Testing. It is a Python framework inspired by Haskell's QuickCheck and Scala's scalacheck (not a port to Python of those frameworks). The framework is not standalone, but works in your own favorite testing infrastructure: PyUnit (a.k.a. unittest) nose and py.test are all supported.

In Combinatorial Testing, qc provides a deterministic implementation of all-choices (simply picks all the combinatorial choices of parameters) and all-pairs algorithms. All-choices simply tests all possible choice of input parameters, so it's an exhaustive, but very slow technique when there are more than a few parameters with more than a handful of possible values each. The all-pairs algorithm is very clever and exponentially reduces the running time, while still guaranteeing that each pair of input parameter is tested for each possible combination of values.

In Random Testing, qc provides many convenient ways of generating test cases, and a very useful automatic test case reduction. The test case reduction uses binary search to find a small, failing test case, very quickly.

More on Random Testing

Since Random Testing is not popular, you may want to have a look at the following videos from Professor John Regehr, University of Utah (less than 15 minutes total) if you are not familiar with the technique:

As you can see, Random Testing is not just randomly feeding your software a random stream of bytes. It requires more thoughts. The Udacity course on testing has units 3 and 4 entirely dedicated to Random Testing, describing many things you need to know about Random Testing: how to create valid, good, random test cases (3.25-26 and 4.5-15), mutators (3.29), oracles (3.30-34), test case reduction (4.5-6), tradeoffs (4.18-20), and more. Random Testing is also mentioned elsewhere in the class (and expected to be known in the final exam). It is very worth watching (the introductory videos linked above are from this class).


The easy, system-wide way (requires administrative privileges):

sudo pip install -e git://

If you don't feel ready to commit for a whole system install of this library, or simply don't have root access on your machine (and don't want to use virtualenv), just copy the qc directory and its content (as seen in at the moment the content is a mere file) into the location of your choice. To make qc available to your programs you will have to set the PYTHONPATH environmental variable or have the qc directory as a subdirectory of the tree where you are running (for details, see )


These examples are more to be read than to be run, but of course you want to run them to see the framework in action (and of course all the failures are there on purpose...)

examples/ and examples/
These files provide a simple example (borrowed from scalacheck) on how to use this framework with Python native PyUnit framework (aka unittest module) and with the popular nose framework. Just run python examples/ or nosetests examples/ Py.test can run both the nose test or the unittest using py.test examples/ or py.test examples/

examples/ TBD

Simple example on how to have qc generate your own custom (random) objects and how to use them in practice. Run it with either nosetests examples/ or py.test examples/
A more elaborate example, showing the power of automatic shrinking and showing how to write your own shrinker. In this example you can see how qc automagically finds the root cause of the bug, compare the output with or without shrinking! Run it with either nosetests examples/ or py.test examples/

Known bugs

See for a list of known issues.

One common problem when using automatic shrinking is running out of stack space in the recursion process (the shrink algorithm call itself several times to produce a smaller test case). This may happen either if there is a bug in qc itself, or if there is a problem in your test code. You will see an error like:

RuntimeError: maximum recursion depth exceeded while calling a Python object

with a stack trace that shows the recursion tree of the shrinking method calling itself. To understand what is happening, it is usually useful to add the shrink=False option to the @forall decorator of the affected test method. In very rare cases it may be necessary to increase the stack depth with a call to sys.setrecursionlimit(NEWDEPTH), but do not do it until you understand that it is really the case for your test. More often than not, there will be a bug in your code (especially likely if you are writing your first shrinker) or in qc. Please report the latter to

Something went wrong with that request. Please try again.