search

pydata / patsy

Describing statistical models in Python using symbolic formulas
Other
947 stars 103 forks source link

do not depend on nose #162

Closed pgajdos closed 4 years ago

pgajdos commented 4 years ago

Use pytest's raise instead of nose's assert_raise. nose is deprecated, will not work with future python versions.

bashtage commented 4 years ago

While I support this patch, you might want to see if formulaic might suit your needs. I think patsy is mostly in EOL mode.

pgajdos commented 4 years ago

It seems that few packages in openSUSE still require python-patsy: python-ggplot python-seaborn python-statsmodels

bashtage commented 4 years ago

seaborn needs it because of statsmodels IIRC. statsmodels is considering whether formulaic can become the formula parser.

ggplot is abandoned without a commit in 4 years. plotnine is the preferred grammar for graphics in Python.

bashtage commented 4 years ago

An alternative fix (and more standard) would be to pull all of the tests out of code files into test files so that nose isn't needed to run patsy.

pgajdos commented 4 years ago

seaborn needs it because of statsmodels IIRC. statsmodels is considering whether formulaic can become the formula parser.

ggplot is abandoned without a commit in 4 years. plotnine is the preferred grammar for graphics in Python.

Thanks for letting me know. I just go trough python modules that still use nose and try to fix them, without not much knowledge about the context.

pgajdos commented 4 years ago

An alternative fix (and more standard) would be to pull all of the tests out of code files into test files so that nose isn't needed to run patsy.

@bashtage, I am not sure I understand. Of course it would be better to have them tests in separate test files, but you use assert_raises, so you will have to youse nose anyway, no?

Anyway, it seems that travis does not like me today, so let's try later.

pgajdos commented 4 years ago

Closing for now.

bashtage commented 4 years ago

Normally projects have directories that look like

root\patsy root\patsy\tests root\setup.py

and then in root\patsy\tests you have files that are named test_<something meaningful>.py. Under normal execution files under root\patsy\test are not executed, and so are only picked up by the test runner, e.g., nose. This means that patsy will be importable under Python 3.9 even if nose it not. It won't be testable though.