- Testing code that generates output can be tedious to maintain
pytest-datarecorderaims to simplify this task.
Table of Contents:
To install pytest-datarecorder:
pip install pytest-datarecorder
To make it easy to use in with py.test the DataRecorder object can be injected into a test function by using the datarecorder fixture.
Example:
def test_this_function(datarecorder):
datarecorder.record_data(
data={'a':1, 'b':2}, recording_file="test/data/recording.json")
The data passed will be serialized to JSON since the recording file
has a .json extension. If data changes, we will get an exception
containing a diff with what changed. If we want to accept the changes
we simply delete the recording and run the code again.
Edit NEWS.rst and wscript (set correct VERSION)
Run
./waf upload
The main functionality is found in src/datarecorder.py and the
corresponding unit test is in test/test_datarecorder.py if you
want to play/modify/fix the code this would, in most cases, be the place
to start.
We try to make our projects as independent as possible of a local system setup. For example with our native code (C/C++) we compile as much as possible from source, since this makes us independent of what is currently installed (libraries etc.) on a specific machine.
To "fetch" sources we use Waf (https://waf.io/) augmented with dependency resolution capabilities: https://github.com/steinwurf/waf
The goal is to enable a work-flow where running:
./waf configure ./waf build --run_tests
Configures, builds and runs any available tests for a given project, such that you as a developer can start hacking at the code.
For Python project this is a bit unconventional, but we think it works well.
The tests will run automatically by passing --run_tests to waf:
./waf --run_tests
This follows what seems to be "best practice" advise, namely to install the package in editable mode in a virtualenv.
- Why use an
srcfolder (https://hynek.me/articles/testing-packaging/). tl;dr you should run your tests in the same environment as your users would run your code. So by placing the source files in a non-importable folder you avoid accidentally having access to resources not added to the Python package your users will install... - Python packaging guide: https://packaging.python.org/distributing/