An abundance of coverage data
Smother is a wrapper utility around coverage.py that measures code coverage separately for each test in a test suite. Its main features include:
- Fast and reliable coverage tracking using coverage.py.
- Ability to lookup which tests visit an arbitrary section of your application code.
- Ability to convert version control diffs into a subset of affected tests to rerun.
Smother contains plugins for nose and pytest, and behaves similarly to coverage.py:
py.test --smother=my_module test_suite/ or nosetests --with-smother --smother-package=my_module test_suite/
These commands create a .smother file that can be queried by the smother CLI
smother lookup foo.bar # which tests visited module foo.bar? smother lookup foo.bar:120-122 # or just some lines in that file? smother lookup foo.bar:baz # or just the `baz` function? smother diff # given local modifications to my repo, what tests might have broken? smother diff | xargs py.test # just run them! smother to_coverage # build a vanilla .coverage file from a .smother file smother csv test.csv # dump all (application, test) pairs to a file
Smother was designed to make it easier to work with legacy codebases. Such codebases often have several properties that make rapid iteration difficult:
- Long-running test suites. The initial codebase that smother was designed for took nearly 24 hours of CPU time to run its 11K tests. smother diff makes it easier to select a (hopefully much) smaller subset of tests to re-run to quickly identify possible regressions.
- Many different subsystems – most of which any single developer is unfamiliar with. smother lookup can be used to explore how and where particular modules are invoked in a test suite. These tests often reveal implicit contracts about code behavior that are not obvious from documentation alone.
- Scope creep. Over time the abstractions in codebases become leakier, and logic between different subsystems becomes more heavily coupled. smother csv catalogs the coupling between source code units (lines, functions, or classes) and tests. Exploring this data often yields insights about which subsystems are well-encapsulated, and which would benefit from refactoring.