- Added nosetests setuptools command. Now you can run python setup.py
nosetests and have access to all nose features and plugins. Thanks to James
Casbon for the patch.
- Added make_decorator function to nose.tools. Used to construct decorator
functions that are well-behaved and preserve as much of the original
function's metadata as possible. Thanks to Antoine Pitrou for the patch.
- Added nose.twistedtools, contributed by Antoine Pitrou. This module adds
@deferred decorator that makes it simple to write deferred tests, with or
- Added monkeypatch to doctests that keeps doctest from stepping on coverage
when the two plugins are used together. Thanks to David Avraamides for the
- Added isolation plugin. Use this plugin to automatically restore sys.modules
after each test module or package. Thanks to Michal Kwiatkowski for the
- Fixed bug where -vvvv turned off verbose logging instead of making it even
more verbose. Thanks to Ian Bicking for the bug report.
- Fixed bug where assert inspection would fail when the trailing """ of a
docstring was one of the inspected lines. Thanks to cito at online dot de
for the bug report.
- Updated attrib plugin to allow selection of test methods by attributes of
the test case class. Thanks to Jason Hildebrand for the patch.
- Improved compatibility with python 2.2. Thanks to Chad Whitacre for the
- Fixed bug in handling of options from setup.cfg. Thanks to Kumar McMillan for
- Fixed bug in generator methods, where a generator method using an inline
funciton would result in an AttributeError. Thanks to Antoine Pitrou for the
- Updated coverage plugin to ignore lines tagged with #pragma: no cover,
matching the behavior of coverage.py on the command line. Thanks to Bill
Zingler for the bug report.
- Added a man page for nosetests. Thanks to Gustavo Noronha Silva for the
request and providing an example.
- New function nose.runmodule() finds and runs tests only in a
single module, which defaults to __main__ (like unittest.main() or
doctest.runmodule()). Thanks Greg Wilson for the suggestion.
- Multiple -w (--where) arguments can now be used in one command line,
to find and run tests in multiple locations. Thanks Titus Brown for
- Multiple --include and --exclude arguments are now accepted in one command
line. Thanks Michal Kwiatkowski for the feature request.
- Coverage will now include modules not imported by any test when
using the new --cover-inclusive switch. Thanks James Casbon for the
- module:TestClass test selections now properly select all tests in the test
- startTest and stopTest are now called in plugins at the beginning and end of
test suites, including test modules, as well as individual tests. Thanks
Michal Kwiatkowski for the suggestion.
- Fix bug in test selection when run as ``python setup.py test``: 'test' was
passing through and being used as the test name selection. Thanks Kumar
McMillan for the bug report.
- Fix bug in handling of -x/--stop option where the test run would stop on
skipped or deprecated tests. Thanks Kumar McMillan for the bug report.
- Fix bug in loading tests from projects with layouts that place modules in
/lib or /src dirs and tests in a parallel /tests dir.
- Fix bug in python version detection. Thanks Kevin Dangoor for the bug report
- Fix log message in selector that could raise IndexError. Thanks Kumar
McMillan for the bug report and patch.
- Fix bug in handling doctest extension arguments specified in environment and
on command line. Thanks Ian Bicking for the bug report.
- Fix bug in running fixtures (setup/teardown) that are not functions, and
report a better error message when a fixture is not callable. Thanks Ian
Bicking for the bug report.
- More unit tests and better test coverage. Numerous bugfixes deriving from
- Make --exe option do what it says, and turn it on by default on
Windows. Add --noexe option so windows users can turn if off.Thanks
richard at artsalliancemedia dot com for the bug reports.
- Handle a working directory that happens to be in the middle of a package
more gracefully. Thanks Max Ischenko for the bug report and test case.
- Fix bugs in test name comparison when a test module is specified whose name
overlaps that of a non-test module. Thanks Max Ischenko for the bug report
and test case.
- Fix warning spam when a non-existent test file is requested on the command
line. Thanks Max Ischenko for the bug report.
- Allow --debug to set any logger to DEBUG. Thanks to casbon at gmail dot com for
- Fix doctest help, which was missing notes about the environment variables
that it accepts. Thanks to Kumar McMillan for the patch.
- Restore sys.stdout after run() in nose.core. Thanks to Titus Brown for the
- Correct handling of trailing comma in attrib plugin args. Thanks Titus Brown
for the patch.
- Fix bug in handling of OR conditions in attrib plugin. Thanks to Titus
Brown for the bug report.
- Fix bug in nose.importer that would cause an attribute error when a local
module shadowed a builtin, or other object in sys.modules, without a
__file__ attribute. Thanks to casbon at gmail dot com for the bug report.
- Fix bug in nose.tools decorators that would cause decorated tests to appear
with incorrect names in result output.
- In TestLoader, use inspect's isfunction() and ismethod() to filter functions
and methods, instead of callable(). Thanks to Kumar McMillan for reporting
- Fix doctest plugin: return an empty iterable when no tests are found in a
directory instead of None. Thanks to Kumar McMillan for the bug report and
- Ignore executable python modules, unless run with --exe file. This is a
partial defense against nose causing trouble by loading python modules that
are not import-safe. The full defense: don't write modules that aren't
- Catch and warn about errors on plugin load instead of dying.
- Renamed builtin profile module from nose.plugins.profile to
nose.plugins.prof to avoid shadowing stdlib profile.py module.
- Add support for plugins, with hooks for selecting, loading and reporting on
tests. Doctest and coverage are now plugins.
- Add builtin plugins for profiling with hotshot, selecting tests by
attribute (contributed by Mika Eloranta), and warning of missed tests
specified on command line.
- Change command line test selection syntax to match unittest. Thanks to Titus
Brown for the suggestion.
- Option to drop into pdb on error or failure.
- Option to stop running on first error or failure. Thanks to Kevin Dangoor
for the suggestion.
- Support for doctests in files other than python modules (python 2.4 only)
- Reimplement base test selection as single self-contained class.
- Reimplement test loading as unittest-compatible TestLoader class.
- Remove all monkeypatching.
- Reimplement output capture and assert introspection support in
unittest-compatible Result class.
- Better support for multiline constructions in assert introspection.
- More context output with assert introspections.
- Refactor setuptools test command support to use proxied result, which
enables output capture and assert introspection support without
monkeypatching. Thanks to Philip J. Eby for the suggestion and skeleton
- Add support for generators in test classes. Thanks to Jay Parlar for the
suggestion and patch.
- Add nose.tools package with some helpful test-composition functions and
decorators, including @raises, contributed by Scot Doyle.
- Reimplement nose.main (TestProgram) to have unittest-compatible signature.
- All-new import path handling. You can even turn it off! (If you don't,
nose will ensure that all directories from which it imports anything are on
sys.path before the import.)
- Logging package used for verbose logging.
- Support for skipped and deprecated tests.
- Configuration is no longer global.
- Add support for py.test-style test generators. Thanks to Jay Parlar for
- Fix bug in doctest discovery. Thanks to Richard Cooper for the bug report.
- Fix bug in output capture being appended to later exceptions. Thanks to
Titus Brown for the patch that uncovered the bug.
- Fix bug(?) in Exception patch that caused masked hasattr/__getattr__ loops
to either become actual infinite loops, or at least take so long to finally
error out that they might as well be infinite.
- Add -m option to restrict test running to only tests in a particular package
or module. Like the -f option, -m does not restrict test *loading*, only
- When loading and running a test module, ensure that the module's path is in
sys.path for the duration of the run, not just while importing the module.
- Add id() method to all callable test classes, for greater unittest
- Fix bug with coverage output when sys.modules contains entries without
- Added -p (--cover-packages) switch that may be used to restrict coverage
report to modules in the indicated package(s)
- Output capture and verbose assertion errors now work when run like
'python setup.py test', as advertised.
- Code coverage improvements: now coverage will be output for all modules
imported by any means that were not in sys.modules at the start of the test
run. By default, test modules will be excluded from the coverage report, but
you can include them with the -t (--cover-tests) option.
- Fix bugs in handling of setup/teardown fixtures that could cause TypeError
exceptions in fixtures to be silently ignored, or multiple fixtures of the
same type to run. Thanks to Titus Brown for the bug report.
- Add -V (--version) switch to nosetests
- Fix bug where sys.path would not be set up correctly when running some
tests, producing spurious import errors (Thanks to Titus Brown and Mike
Thomson for the bug reports)
- For test classses not derived from unittest.TestCase, output (module.Class)
"doc string" as test description, when method doc string is available
(Thanks to David Keeney for the suggestion, even if this isn't quite what he
- Revise import to bypass sys.path and manipulate sys.modules more
intelligently, ensuring that the test module we think we are loading is the
module we actually load, and that modules loaded by other imports are not
reloaded without cause
- Allow test directories inside of packages. Formerly directories matching
testMatch but lacking an __init__.py would cause an ImportError when located
inside of packages
- Fix bugs in different handling of -f switch in combination with -w and -o
- Fix bug in main() that resulted in incorrect exit status for nosetests
script when tests fail
- Add missing test files to MANIFEST.in
- Miscellaneous pylint cleanups
- Add doctest support
- Add optional code coverage support, using Ned Batchelder's coverage.py;
activate with --coverage switch or NOSE_COVERAGE environment variable
- More informative error message on import error
- Fix bug where module setup could be called twice and teardown skipped
for certain setup method names.
- main() returns success value, does not exit. run_exit() added to support
old behavior; nosetests script now calls nose.run_exit()
- Fix bus error on exit
- Discover tests inside of non-TestCase classes that match testMatch
- Reorganize selftest: now selftest tests the output of a full nose run
- Add test_with_setup.py contributed by Kumar McMillan
- Refactor and correct bugs in discovery and test loading
- Reorganize and expand documentation
- Add -f (run this test file only) switch
- Bugfix release: test files in root of working directory were not being
stripped of file extension before import.
- Change license to LGPL
- Major rework of output capture and assert introspection
- Improve test discovery: now finds tests in packages
- Replace -n switch ('no cwd') with -w switch ('look here')
- New nosetests script
- Allow specification of names on command line that are loadable but not
directly loadable as modules (eg nosetests -o path/to/tests.py)
- Add optional py.test-like assert introspection. Thanks to Kevin Dangoor
for the suggestion.
- Improvements to selftest
- Increased compatibility with python 2.3 (and maybe earlier)
- Increased compatibility with tests written for py.test: now calls
module.setup_module(module) if module.setup_module() fails