pytest-regtest is not a widely used pytest plugin, and its implementation of snapshot testing is fairly non-standard (although it does seem to handle NumPy/Pandas/Polars...)
I worry that teaching regression testing with matplotlib could lead to bad habits forming, e.g. testing the plot of some calculation rather than testing the results themselves, which could be dangerous.
I would lean towards teaching syrupy as the standard solution, and dropping the matplotlib part. Textual data would be easiest here. For NumPy etc, we can point people towards pytest-regtest, snaptol, or just handling it manually at the end.
pytest-regtestis not a widely used pytest plugin, and its implementation of snapshot testing is fairly non-standard (although it does seem to handle NumPy/Pandas/Polars...)I worry that teaching regression testing with matplotlib could lead to bad habits forming, e.g. testing the plot of some calculation rather than testing the results themselves, which could be dangerous.
I would lean towards teaching syrupy as the standard solution, and dropping the matplotlib part. Textual data would be easiest here. For NumPy etc, we can point people towards pytest-regtest, snaptol, or just handling it manually at the end.