Nose plugin to show progress bar and tracebacks during tests
Project description
nose-progressive is a nose plugin which displays progress in a stationary progress bar, freeing the rest of the screen (as well as the scrollback buffer) for the compact display of test failures, which it formats beautifully and usefully. It displays failures and errors as soon as they occur and avoids scrolling them off the screen in favor of less useful output. It also offers a number of other human-centric features to speed the debugging process.
The governing philosophy of nose-progressive is to get useful information onto the screen as soon as possible and keep it there as long as possible while still indicating progress.
Features
Progress Bar
nose-progressive indicates progress in a stationary progress bar at the bottom of the screen. It supports a wide variety of terminal types and reacts to terminal resizing with all the grace it can muster. And unlike with the standard dot-strewing testrunner, you can always see what test is running.
Tracebacks: Prompt, Pretty, and Practical
nose typically waits until the bitter end to show error and failure tracebacks, which wastes a lot of time in large tests suites that take many minutes to complete. We show tracebacks as soon as they occur so you can start chasing them immediately, and we format them much better:
Judicious use of color and other formatting makes the traceback easy to scan. It’s especially easy to slide down the list of function names to keep your place while debugging.
Omitting the Traceback (most recent call last) line and using relative paths (optional), along with many other tweaks, fits much more in limited screen space.
Identifying failed tests in a format that can be fed back to nose makes it easy to re-run them:
FAIL: kitsune.apps.wiki.tests.test_parser:TestWikiVideo.test_video_english
To re-run the above, do this:
nosetests --with-progressive kitsune.apps.wiki.tests.test_parser:TestWikiVideo.test_video_english
The frame of the test itself always comes first; we skip any setup frames from test harnesses and such. This keeps your concentration where it counts. Also, like unittest itself, we hide any frames that descend into trivial comparison helpers like assertEquals() or assertRaises().
(We’re actually better at it than unittest. We don’t just start hiding frames at the first unittest one after the test; we snip off only the last contiguous run of unittest frames. This lets you wrap your tests in the decorators from the mock library, which masquerades as unittest, and still see your tracebacks.)
Editor shortcuts (see below) let you jump right to any problem line in your editor.
Editor Shortcuts
For each frame of a traceback, nose-progressive provides an editor shortcut. This is a combination of a filesystem path and line number in a format understood by vi, emacs, the BBEdit command-line tool, and a number of other editors:
vi +361 apps/notifications/tests.py # test_notification_completeness
Just triple-click (or what have you) to select the line, and copy and paste it onto the command line. You’ll land right at the offending line in your editor of choice. As a bonus, the editor shortcut is more compact than the stock traceback formatting.
You can set which editor to use by setting any of these, which nose-progressive checks in order:
The --progressive-editor commandline option
The NOSE_PROGRESSIVE_EDITOR environment variable
The $EDITOR environment variable
Custom Error Classes
nose-progressive fully supports custom error classes like Skip and Deprecated. We note the tests that raise them in realtime, just like normal errors and failures:
TODO: kitsune.apps.sumo.tests.test_readonly:ReadOnlyModeTest.test_login_error
However, when an error class is not considered a failure, we don’t show it unless the --progressive-advisories option is used, and, even in that case, we don’t show a traceback (since usually the important bit of information is that the test was skipped, not the line it was skipped on). This stems from our philosophy of prioritizing useful information.
Custom error classes are summarized in the counts after the run, along with failures and errors:
4 tests, 1 failure, 1 error, 1 skip in 0.0s
^^^^^^ Bold ^^^^^^
The non-zero counts of error classes that represent failures are bold to draw the eye and to correspond with the bold details up in the scrollback. Just follow the bold, and you’ll find your bugs.
Django Support
nose-progressive can run your Django tests via django-nose. Just install django-nose, then run your tests like so:
./manage.py test --with-progressive --logging-clear-handlers
Installation
pip install nose-progressive
Or, get the bleeding-edge, unreleased version:
pip install -e git://github.com/erikrose/nose-progressive.git#egg=nose-progressive
Upgrading
To upgrade from an older version of nose-progressive, assuming you didn’t install it from git:
pip install --upgrade nose-progressive
Use
The simple way:
nosetests --with-progressive
My favorite way, which suppresses any noisy log messages thrown by tests unless they fail:
nosetests --with-progressive --logging-clear-handlers
To use nose-progressive by default, add with-progressive=1 to .noserc.
Options
General Options
- --progressive-editor
The editor to use for the shortcuts in tracebacks. Defaults to the value of $EDITOR and then “vi”. Equivalent environment variable: NOSE_PROGRESSIVE_EDITOR.
- --progressive-abs
Display paths in traceback as absolute, rather than relative to the current working directory. This lets you copy and paste it to a shell in a different cwd or to another program entirely. Equivalent environment variable: NOSE_PROGRESSIVE_ABSOLUTE_PATHS.
- --progressive-advisories
Show even non-failure custom errors, like Skip and Deprecated, during test runs. Equivalent environment variable: NOSE_PROGRESSIVE_ADVISORIES.
- --progressive-with-styling
nose-progressive automatically omits bold and color formatting when its output is directed to a non- terminal. Specifying --progressive-with-styling forces such styling to be output regardless. Equivalent environment variable: NOSE_PROGRESSIVE_WITH_STYLING.
- --progressive-with-bar
nose-progressive automatically omits the progress bar when its output is directed to a non-terminal. Specifying --progressive-with-bar forces the bar to be output regardless. This option implies --progressive-with-styling. Equivalent environment variable: NOSE_PROGRESSIVE_WITH_BAR.
Color Options
Each of these takes an ANSI color expressed as a number from 0 to 15.
- --progressive-function-color=<0..15>
Color of function names in tracebacks. Equivalent environment variable: NOSE_PROGRESSIVE_FUNCTION_COLOR.
- --progressive-dim-color=<0..15>
Color of de-emphasized text (like editor shortcuts) in tracebacks. Equivalent environment variable: NOSE_PROGRESSIVE_DIM_COLOR.
- --progressive-bar-filled=<0..15>
Color of the progress bar’s filled portion. Equivalent environment variable: NOSE_PROGRESSIVE_BAR_FILLED_COLOR.
- --progressive-bar-empty=<0..15>
Color of the progress bar’s empty portion. Equivalent environment variable: NOSE_PROGRESSIVE_BAR_EMPTY_COLOR.
Caveats and Known Bugs
Makes a cosmetic mess when used with ipdb. Consider pdbpp instead.
Some logging handlers will smear bits of the progress bar upward if they don’t print complete lines. I hope to fix this with some monkeypatching, but in the meantime, passing --logging-clear-handlers works around this.
Requires Python 2.5 or greater and doesn’t support Python 3 yet.
Having trouble? Pop over to the issue tracker.
Kudos
Thanks to Kumar McMillan for his nose-nicedots plugin, which provided inspiration and starting points for the path formatting. Thanks to my support.mozilla.com teammates for writing so many tests that this became necessary. Thanks to Jeff Balogh for django-nose, without which I would have had little motivation to write this.
License
GPL
Version History
- 1.4.2
Clear the TestLoader’s path cache (new in nose 1.3.0) after counting the tests. This solves the problem of finding 0 tests to run under nose 1.3.0.
Make progress bar tests less brittle so they don’t falsely fail on OS X 10.8 or other platforms where the terminfo isn’t exactly what I wrote the test under.
- 1.4.1
Fix the “AttributeError: ‘dict’ object has no attribute ‘raw_input’” error that sometimes occurred at pdb breakpoints. Thanks to David Baumgold for finding the cause!
- 1.4
Make the final “OK!” green and bold. This helps me pick it out faster.
Warn when using --with-id and --verbosity=2 or higher. (Jason Ward)
Add experimental Python 3 support. Functionality might work, but tests need to be ported to pass.
Allow other nose plugins to process the test loader. (Ratnadeep Debnath)
Show parameter values in the names of generated tests. (Bruno Binet)
Tolerate a corner case in skipped tests without crashing. (Will Kahn-Greene)
Swallow chars that don’t decode with UTF-8 when printing tracebacks: both in filenames and source code. (Thanks to Bruno Binet for some patches inspiring a rethink here.)
- 1.3
Redo progress bar. Now it is made of beautiful terminal magic instead of equal signs. It looks best when your terminal supports at least 16 colors, but there’s a monochrome fallback for fewer. Or, you can customize the colors using several new command-line options.
Fix a Unicode encoding error that happened when non-ASCII chars appeared in traceback text. (Naoya INADA)
- 1.2.1
Tolerate empty tracebacks in the formatter. This avoids exacerbating crashes that occur before any test frames.
- 1.2
Fix Python 2.5 support. (David Warde-Farley)
Fix display of skipped tests in Python 2.7.
Require nose 0.11.0 or greater. Before that, test counting didn’t work sometimes when test generators were involved. (David Warde-Farley)
Hide the progress bar by default when not outputting to a terminal. This lets you redirect nose-progressive’s output to a file or another process and get a nice list of tracebacks.
Add an option for forcing the display of terminal formatting, even when redirecting the output to a non-terminal.
Factor out the terminal formatting library into its own package.
Start using tox for testing under multiple versions of Python.
- 1.1.1
Fix a bug that would cause the formatter to crash on many SyntaxErrors. This also improves the heuristics for identifying the test frame when there’s a SyntaxError: we can now find it as long as the error happens at a frame below that of the test.
- 1.1
You can now set the editor nose-progressive uses separately from the $EDITOR shell variable.
- 1.0
Every stack frame is now an editor shortcut. Not only does this make it easier to navigate, but it’s shorter in both height and width.
Reformat tracebacks for great justice. Subtle coloring guides the eye down the list of function names.
Hide unittest-internal and other pre-test stack frames when printing tracebacks. Fewer frames = less noise onscreen = less thinking = win!
Add an option to use absolute paths in tracebacks.
- 0.7
Pick the correct stack frame for editor shortcuts to syntax errors. Had to handle syntax errors specially, since they don’t make it into the traceback proper.
Show the actual value of the $EDITOR env var rather than just “$EDITOR”. I’m hoping it makes it a little more obvious what to do with it, plus it gives a working default if $EDITOR is not set. Plus plus it doesn’t explode if you have flags in your $EDITOR, e.g. bbedit -w.
- 0.6.1
Fix a crash triggered by a test having no defined module. –failed should always work now.
- 0.6
Major refactoring. nose-progressive now has its own testrunner and test result class. This makes it fully compatible with the capture plugin and other plugins that make output.
Fully support custom error classes, like Skips and Deprecations. They are printed during the test run, bolded if they represent failure, and summarized in the counts after the run.
Tests which write directly to stderr or stdout no longer smear the progress bar.
Add $EDITOR to editor shortcut: no more typing!
Work with tests that don’t have an address() method.
Work with tests that return a null filename from test_address().
Don’t pave over pdb prompts (anymore?).
Don’t obscure the traceback when the @with_setup decorator on a test generator fails.
- 0.5.1
Fix a crash on error when file of a stack frame or function of a test are None.
- 0.5
Guess the frame of the test, and spit that out as the editor shortcut. No more pointers to eq_()!
More reliably determine the editor shortcut pathname, e.g. when running tests from an egg distribution directory.
Embolden bits of the summary that indicate errors or failures.
- 0.4
Add time elapsed to the final summary.
Print “OK!” if no tests went ill. I seem to need this explicit affirmation in order to avoid thinking after a test run.
In the test failure output, switch the order of the line number and file name. This makes it work with the BBEdit command-line tool in addition to emacs and vi.
- 0.3.1
Cowboy attempt to fix a crasher on error by changing the entry_point to nose.plugin.0.10
- 0.3
Progress bar now works with plain old nosetests, not just django-nose. Sorry about that!
Stop printing the test name twice in the progress bar.
Add basic terminal resizing (SIGWINCH) support. Expanding is great, but contracting is still a little ugly. Suggestions welcome.
- 0.2
Real progress bar!
Don’t crash at the end when --no-skips is passed.
Print the exception, not just the traceback. That’s kind of important. :-)
Don’t crash when a requested test doesn’t exist.
- 0.1.2
More documentation tweaks. Package long_description now contains README.
- 0.1.1
Add instructions for installing without git.
Change package name in readme to the hypenated one. No behavior changes.
- 0.1
Initial release
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
