skip to navigation
skip to content

testresources 0.2.7

Testresources, a pyunit extension for managing expensive test resources

testresources: extensions to python unittest to allow declarative use of resources by test cases.

Copyright (C) 2005-2013 Robert Collins <>

Licensed under either the Apache License, Version 2.0 or the BSD 3-clause license at the users choice. A copy of both licenses are available in the project source as Apache-2.0 and BSD. You may not use this file except in compliance with one of these two licences.

Unless required by applicable law or agreed to in writing, software distributed under these licenses is distributed on an “AS IS” BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the license you chose for the specific language governing permissions and limitations under that license.

See the COPYING file for full details on the licensing of Testresources.


testresources extends unittest with a clean and simple api to provide test optimisation where expensive common resources are needed for test cases - for example sample working trees for VCS systems, reference databases for enterprise applications, or web servers … let imagination run wild.

Dependencies to build/selftest

Dependencies to use testresources

  • Python 2.4+ (or 3.2+)

How testresources Works

The basic idea of testresources is:

  • Tests declare the resources they need in a resources attribute.
  • When the test is run, the required resource objects are allocated (either newly constructed, or reused), and assigned to attributes of the TestCase.

testresources distinguishes a ‘resource manager’ (a subclass of TestResourceManager) which acts as a kind of factory, and a ‘resource’ which can be any kind of object returned from the manager class’s getResource method.

Resources are either clean or dirty. Being clean means they have same state in all important ways as a newly constructed instance and they can therefore be safely reused.

Main Classes


By extending or mixing-in this class, tests can have necessary resources automatically allocated and disposed or recycled.

ResourceTestCase can be used as a base class for tests, and when that is done tests will have their resources attribute automatically checked for resources by both OptimisingTestSuite and their own setUp() and tearDown() methods. (This allows tests to remain functional without needing this specific TestSuite as a container). Alternatively, you can call setUpResources(self, resources, test_result) and tearDownResources(self, resources, test_result) from your own classes setUp and tearDown and the same behaviour will be activated.

To declare the use of a resource, set the resources attribute to a list of tuples of (attribute_name, resource_manager).

During setUp, for each declared requirement, the test gains an attribute pointing to an allocated resource, which is the result of calling resource_manager.getResource(). finishedWith will be called on each resource during tearDown().

For example:

class TestLog(testresources.ResourcedTestCase):

    resources = [('branch', BzrPopulatedBranch())]

    def test_log(self):
        show_log(self.branch, ...)


A TestResourceManager is an object that tests can use to create resources. It can be overridden to manage different types of resources. Normally test code doesn’t need to call any methods on it, as this will be arranged by the testresources machinery.

When implementing a new TestResourceManager subclass you should consider overriding these methods:


Must be overridden in every concrete subclass.

Returns a new instance of the resource object (the actual resource, not the TestResourceManager). Doesn’t need to worry about reuse, which is taken care of separately. This method is only called when a new resource is definitely needed.

make is called by getResource; you should not normally need to override the latter.

Cleans up an existing resource instance, eg by deleting a directory or closing a network connection. By default this does nothing, which may be appropriate for resources that are automatically garbage collected.
Reset a no-longer-used dirty resource to a clean state. By default this just discards it and creates a new one, but for some resources there may be a faster way to reset them.
Check whether an existing resource is dirty. By default this just reports whether TestResourceManager.dirtied has been called or any of the dependency resources are dirty.

For instance:

class TemporaryDirectoryResource(TestResourceManager):

    def clean(self, resource):

    def make(self):
        return tempfile.mkdtemp()

    def isDirty(self, resource):
        # Can't detect when the directory is written to, so assume it
        # can never be reused.  We could list the directory, but that might
        # not catch it being open as a cwd etc.
        return True

The resources list on the TestResourceManager object is used to declare dependencies. For instance, a DataBaseResource that needs a TemporaryDirectory might be declared with a resources list:

class DataBaseResource(TestResourceManager):

    resources = [("scratchdir", TemporaryDirectoryResource())]

Most importantly, two getResources to the same TestResourceManager with no finishedWith call in the middle, will return the same object as long as it is not dirty.

When a Test has a dependency and that dependency successfully completes but returns None, the framework does not consider this an error: be sure to always return a valid resource, or raise an error. Error handling hasn’t been heavily exercised, but any bugs in this area will be promptly dealt with.

A sample TestResourceManager can be found in the doc/ folder.

See pydoc testresources.TestResourceManager for details.


Glue to adapt testresources to an existing resource-like class.


Glue to adapt testresources to the simpler fixtures.Fixture API. Long term testresources is likely to consolidate on that simpler API as the recommended method of writing resources.


This TestSuite will introspect all the test cases it holds directly and if they declare needed resources, will run the tests in an order that attempts to minimise the number of setup and tear downs required. It attempts to achieve this by callling getResource() and finishedWith() around the sequence of tests that use a specific resource.

Tests are added to an OptimisingTestSuite as normal. Any standard library TestSuite objects will be flattened, while any custom TestSuite subclasses will be distributed across their member tests. This means that any custom logic in test suites should be preserved, at the price of some level of optimisation.

Because the test suite does the optimisation, you can control the amount of optimising that takes place by adding more or fewer tests to a single OptimisingTestSuite. You could add everything to a single OptimisingTestSuite, getting global optimisation or you could use several smaller OptimisingTestSuites.


This is a trivial TestLoader that creates OptimisingTestSuites by default.


testresources will log activity about resource creation and destruction to the result object tests are run with. 6 extension methods are looked for: startCleanResource, stopCleanResource, startMakeResource, stopMakeResource, startResetResource and finally stopResetResource. testresources.tests.ResultWithResourceExtensions is an example of a TestResult with these methods present.

Controlling Resource Reuse

When or how do I mark the resource dirtied?

The simplest approach is to have TestResourceManager.make call self.dirtied: the resource is always immediately dirty and will never be reused without first being reset. This is appropriate when the underlying resource is cheap to reset or recreate, or when it’s hard to detect whether it’s been dirtied or to trap operations that change it.

Alternatively, override TestResourceManager.isDirty and inspect the resource to see if it is safe to reuse.

Finally, you can arrange for the returned resource to always call back to TestResourceManager.dirtied on the first operation that mutates it.


  • Can I dynamically request resources inside a test method?

    Generally, no, you shouldn’t do this. The idea is that the resources are declared statically, so that testresources can “smooth” resource usage across several tests.

    But, you may be able to find some object that is statically declared and reusable to act as the resource, which can then provide methods to generate sub-elements of itself during a test.

  • If the resource is held inside the TestResourceManager object, and the TestResourceManager is typically constructed inline in the test case resources attribute, how can they be shared across different test classes?

    Good question.

    I guess you should arrange for a single instance to be held in an appropriate module scope, then referenced by the test classes that want to share it.

File Type Py Version Uploaded on Size
testresources-0.2.7.tar.gz (md5, pgp) Source 2013-01-20 31KB
  • Downloads (All Versions):
  • 1603 downloads in the last day
  • 16143 downloads in the last week
  • 61402 downloads in the last month