coverage reporting and logging without a webservice

For quite a while now I wanted to have a cheap replacement for collecting and tracking coverage of codes I work on. Unfortunately, there is no quick solution I know of which does more or less what do. There is probably a Jenkins plugin to do it, or some heavy weight solution which needs a webservice, a database and maintenance.

Alas, we always lack time, and git offers teams a place to keep the history, so why not use it to track the coverage reporting?

Enter git commit --amend and some Makefile trickery. You can use git own powers to amend the last commit and add to the bottom of it the coverage report.

I do it with make coverage-record. This runs the tests, and saves the report in the body of the latest commit message.

You can add the following target to your project make file. If you feel more courageous, you can add this to your pre-push hook. This will have a nice bonus, which will by default, prevent people pushing broken builds (e.g. failing tests):

coverage-record: TMPFILE := $(shell mktemp)
coverage-record: coverage
    git show -s --format=%B HEAD > $(TMPFILE)
    coverage report >> $(TMPFILE)
    git commit --amend -F $(TMPFILE)
    rm $(TMPFILE)

Note that this target depends on coverage, which is your how you invoke your test suite, for example:

    py.test -vv --cov=blogit tests

To use this target in the pre-push hook, you will slightly have to change the Makefile syntax to shell syntax.

I would be happy to know of other solutions how you track the test coverage of your software projects!

This entry was tagged: python, testing, git

Share this post:


comments powered by Disqus