Automated testing for I2C sensor driver module

[DavesCodeMusings]

TL;DR... https://github.com/DavesCodeMusings/SHT3x-MicroPython

It's a Sensiron SHT3x I2C driver. I know there are modules out there already and I mostly did this for the learning experience. The part I'm keen to show off is the automated testing used with this module. I don't know how many folks are using automated testing with their code, but I thought I'd share for those interested.

My testing up to this point had been wiring an actual SHT30 to an ESP32 board and running main.py to print temperature and humidity values. If it looked reasonable, I called it good.

Now I have run_tests.py that executes to ensure all the class functions are working correctly.

The test script is designed to run in CPython so I can automate the tests on a GitHub runner whenever I push code to the repository. And the GitHub runner has no concept of SHT3x sensors, machine.SoftI2C(), or machine.Pin(), or even micropython.const(). So I made some mock functions to act like an SHT30 responding to I2C commands. Most of this is in my machine.py script.

My run_tests.py is not fancy at all. There's no testing framework. It simply uses assert() to throw an exception when a test doesn't pass. This is enough to stop the pipeline job running on the GitHub runner and flag the build as unsuccessful. Other tests rely on the module itself throwing the exception to fail the test.

I'm not an automated testing expert by any stretch of the imagination, so I'm open to suggestions for improvement. Though as it stands, I think it's far better than my old method of running it on an ESP32 and calling it good. And, it runs automatically with each code push, so no manual effort from me.

[jonnor]

Automated testing is the way. Your approach seems to be simple and pragmatic. Thanks for documenting!

[sosi-deadeye]

Testing are always good.

We test our training devices (PLCs) manually as far as possible. I have automated tests for a large order (96 devices).

This gave me two advantages:

• Errors were detected immediately
• The automated test saved around 5 minutes per device.

With 32 digital inputs/outputs and 4 analog inputs/outputs, you can get it wrong if you test manually.

In the software world, testing has another advantage:

• You can change code without having to worry that nothing will work afterwards.

Sometimes errors slip through the tests and then the tests simply have to be improved.