Polytester is a simple, groovy multi-language test runner.
Polytester Just Works with any testing framework that runs in the shell (yo, that's pretty much everything on the planet), and ships with extra-smooth integration for lots of common frameworks including django, karma, protractor, and more. Polytester was built by Steven Skoczen at BuddyUp. If you're in school, and could use a study buddy, check us out!
pip install polytester
Create a tests.yml file. The following example shows django, protractor, and karma (jasmine) tests.
yml api: command: python manage.py test js: command: karma start karma.conf.js e2e: command: protractor
``` $ polytester Detecting... ✔ api detected as django tests. ✔ e2e detected as protractor tests. ✔ js detected as karma tests. Running tests... ✔ api - 35 tests passed. ✔ e2e - 17 tests passed. ✔ js - 23 tests passed.
✔ All tests passed. ```
Note that the status code returned is correct, so you can just dump this on your CI service, and be done.
That's it. There is no step 3.
Because everyone loves animated gifs
Here's the above - polytester installed, configured, and running in less than 30 seconds:
Any test framework that returns standard error codes will work out of the box. That's pretty much everything.
In addition, polytester progressively upgrades to extra-nice output for the frameworks it has parsers for. As of v1.0, the following parsers are built-in, and it's simple to write your own. More are very much welcome via PR, and as you can see below, writing them is easy!
But again, for extra clarity - if your test runner returns normal output codes, you can just drop it in and it'll work great.
Polytester will respond to
polytester. Both do the exact same thing.
There are a variety of options to make development simple.
polytester foo,barjust runs the test suite(s) specified.
--ciwatches all files specified in a
watch_glob, and immediately runs the relevant suite on file changes. Any running tests are killed.
--wipruns tests flagged as "work in progress" by running the
wip_commandfor all suites that specify it.
-vdumps all output to the shell. To prevent collisions, when run in this mode, test suites are run in serial, instead of the normal parallel execution.
--parallel n monly runs test chunk
m, for parallel build test environments.
--config foo.ymlspecifies a different location for the config file. Default is
If you want to get more out of polytester, there's more under the surface, including wip, autoreload, pass/fail counts, custom frameworks, and more.
Here's all the goodness.
Having your tests auto-run when files change is super-handy. With polytester, it's simple.
watch_glob(and optionally a
watch_dir) in your
yml python: command: nosetests watch_glob: "*.py" watch_dir: api
bash polytester --autoreload
Any time you change a file that matches the glob, polytester will immediately run the matching test suite. Any running tests for that suite will be immediately killed.
watch_diris not specified, it defaults to the current directory.
- To specify multiple file types, you can use standard unix globs, i.e.
- Running with
--autoreloadwill only run the tests that have a
watch_globin their config. Which makes sense once you think about it, but might suprise you at first glance.
Autoreload in action:
WIP (Work in Progress) tests
Being able to tag and run certain groups of tests becomes a huge develoment time-saver for larger codebases. Polytester makes it simple. Just specify a
wip_command, and run with
yml python: command: nosetests wip_command: nosetests -a wip
bash polytester --wip
If you're running on a ci platform with parallel builds, like CircleCI, the
--parallel option can save you some time.
Just set up your build config to use it, and pass in the appropriate shell variables.
For CircleCI, just set your
test: override: - polytester --parallel $CIRCLE_NODE_INDEX $CIRCLE_NODE_TOTAL: parallel: true
And you're all set. Your test suites will split out automatically, according to the number of build containers you have.
Specifying test frameworks
If you're using the default test command for any supported frameworks, polytester just detects the right one, and you're on your way. However, if you're using a custom runner, or something a bit special, you can easily just specify which parser polytester should use.
Let's say for reasons too complex to explain, I have a custom wrapper around my nose script. No problem. In my
tests.yml, I just tell polytester to expect nose output.
python: command: my_custom_nose_script.sh parser: polytester.NoseParser
Now, when you run, you get this output:
$ polytester Detecting... ✔ python specified as nose tests. Running tests... ✔ python: 47 tests passed. ✔ All tests passed.
Here's the full list of built-in parsers:
DefaultParser(Just listens to exit codes, no support for number of tests.)
If you need a parser not in this list, you can make it by writing a few simple functions. See Custom parsers below.
Here's a yml file, with everything, just for easy reference.:
python: command: nosetests wip_command: nosetests -a wip watch_glob: "*.py;*.html" # Note you do need quotes because of the *. watch_dir: my_app/foo parser: my_parsers.MyNiftyCustomNoseParser
Writing a Custom Parser
Any test framework that returns standard error codes (0 for pass, non-zero for fail) will Just Work out of the box. However, if you want fancy test counts (and someday more), writing a custom parser is easy.
Just write a class that inherits
DefaultParser, stick it somewhere on your python path, put in in your
tests.yml file, and you're good to go. Here's an example for pep8.
Please note: if you're writing for a common framework/use case, please submit a pull request!
Write your own parser.
my_parsers.py ```python from polytester.parsers import DefaultParser
class MyCustomParser(DefaultParser): name = "my custom"
def tests_passed(self, result): # Required, the code below is the default in DefaultParser return result.return_code == 0 def num_failed(self, result): # Optional. m = re.match("(\d+) failed", result.output) return int(m.group(0)) def num_passed(self, result): # Optional. return self.num_total(result) - self.num_failed(result) def num_total(self, result): # Optional. m = re.match("(\d+) total", result.output) return int(m.group(0)) def command_matches(self, command_string): # Optional, used for trying to auto-detect the test framework. # Since this is totally custom, we just return false return False
resultsis an object with the following attributes:
output- The stdout and stderr, in the order produced while running.
output, but stripped of all ANSI colors and escape codes.
return_code- The return code.
passed- A boolean indicating if the tests have passed.
Noneuntil a definitive answer is known.
parser- An instance of the parser class. (i.e. you can call
Specify it in your test.yml file.
yml custom: command: run_tests.sh parser: my_parsers.MyCustomParser
Run your tests like normal!
```bash $ polytester Detecting... ✔ custom specified as my custom tests.
Running tests... ✔ custom: 18 tests passed.
✔ All tests passed. ```
When things go wrong.
When tests fail, polytester just falls back to the helpful output your test frameworks already give you:
$ polytester Detecting... ✔ api detected as django tests. ✔ e2e detected as karma tests. ✔ js detected as protractor tests. Running tests... ✘ api - 1 of 35 tests failed. ====================================================================== FAIL: test_addition (events.tests.SampleTest) ---------------------------------------------------------------------- Traceback (most recent call last): File "/Users/me/project/events/tests.py", line 6, in test_addition self.assertEquals(1+1, 3) AssertionError: 2 != 3 ---------------------------------------------------------------------- Ran 35 tests in 1.642s FAILED (failures=1) ✔ e2e - 17 tests passed. ✔ js - 23 tests passed. ✘ Tests failed
1.0.1 - Jan 27, 2015
- Bugfix on --parallel.
1.0 - Jan 27, 2015 - "Rick James"
- All RDD features in, we're using polytester in production.
0.1.3 - Jan 22, 2015
- Everything but
--failfast, and a couple parsers is in.
0.1.2 - Jan 21, 2015
- Works on python 3. First stable release!
0.1.1 - Jan 21, 2015
- Fixes inclusion of wsgiref, which breaks python 3 due to a print.
0.1 - Jan 21, 2015
- Initial release. Basic functionality works, only Django parser is in. Very much RDD.
Into the Future
As with all the open-source projects I run, I leave the future pretty open to what the people who use the project request, and the PRs that are sent.
But here's a short list of things that are rolling around in my head as future features:
- Test coverage. Yes, I appreciate the irony.
- Better parsing of test outputs, to just list failed test file names and line numbers or other fancy niceties.
- xUnit output
--failfast, for super-quick iteration.
- The ability for parsers to do better parallelization introspection (based on globs, etc)
- Whatever great stuff you bring to the table!
If you want to add support for a language or framework, those PRs are always welcome.
If you have a bigger idea, just pop open an issue, and we'll talk it through, so we don't cross wires when the PR comes!
Anyone is welcome to contribute to polytester, regardless of skill level or experience. To make polytester the best it can be, we have one big, overriding cultural principle:
Simple. Easy, right?
We've all been newbie coders, we've all had bad days, we've all been frustrated with libraries, we've all spoken a language we learned later in life. In discussions with other coders, PRs, and CRs, we just give each the benefit of the doubt, listen well, and assume best intentions. It's worked out fantastically.
This doesn't mean we don't have honest, spirited discussions about the direction to move polytester forward, or how to implement a feature. We do. We just respect one other while we do it. Not so bad, right? :)