Zphota on my system does not exit if config is not present

[johannct]

https://github.com/lephare-photoz/lephare/blame/86fa1daae4fff37c47f6cf44cc0984da5ee55839/tests/lephare/test_zphota.py#L14

Indeed runner.py#L123 just prints a warning out. How come this passes CI unit tests?

[johannct]

Ok calling pytest from the root dir does not do the same as python -m pytest -s tests which is run in the CI

[johannct]

nope... Running pytest triggers the sequence of tests from each file (alphabetical order), and then fail on test_zphota.py, while running pytest tests/lephare/test_zphota.py does not fail....

[johannct]

What I am pretty sure about is that there is not system exit :

In [1]: from lephare import Zphota LEPHAREDIR is being set to the default cache directory: /home/cohen/.cache/lephare/data More than 1Gb may be written there. LEPHAREWORK is being set to the default cache directory: /home/cohen/.cache/lephare/work Default work cache is already linked. This is linked to the run directory: /home/cohen/.cache/lephare/runs/20240501T000334 In [2]: Zphota() WARNING: no config file provided! Out[2]: <lephare.zphota.Zphota at 0x725dd4b74140>

[raphaelshirley]

It throws a SystemExit: 2 for me

[johannct]

in an interactive session?

[raphaelshirley]

Yes in a notebook

[raphaelshirley]

[johannct]

ok we need to sort this difference of behavior out. Obviously it prints the WARNING, so the system exit occurs afterwards, so it has to occur when parsing the args....

[raphaelshirley]

Does this produce SystemExit: 2 on your machine:

import argparse
parser = argparse.ArgumentParser()
parser.add_argument('foo')
args = parser.parse_args()

[johannct]

yes. This does not :

import argparse parser = argparse.ArgumentParser() parser.add_argument('foo', action='store_true') args = parser.parse_args()

and this does not either :

parser.add_argument('--foo')

So it looks like the problem is in the formatting of some arguments?

[johannct]

By the way your system exit above is different : it says something about argument - f not allowed. Can you %tb after the exception occurred on the python session?

[raphaelshirley]

Can we just add raise SystemExit() after the printed warning?

[johannct]

I was expecting to be possible to run without a config file if all the necessary arguments are passed at the command line or as arguments in a python session. It would be better to try to understand what is going on. Especially because on your system it seems to be looking at a kernel configuration. Can you try outside of a notebook?

[raphaelshirley]

If I run it in a python script it prints the warning but doesn't throw a system exit but the test runs fine.

[raphaelshirley]

Bit of a long shot but is None parsed as a string "None" depending on system?

[johannct]

We do not depend on the system : I also have an exit error 2 when I run your snippet in a notebook. I do not know whether @OliviaLynn tried it on a notebook and then added this test in, but you and I are consistent, irrespective of the system : no exit error if run at the command line, exit error when run from a notebook.

What remains to understand is the difference in behavior of pytest depending on how one runs it.... But it is not a big deal. I suggest that we remove this test.

[raphaelshirley]

It's https://github.com/lephare-photoz/lephare/blob/86fa1daae4fff37c47f6cf44cc0984da5ee55839/src/lephare/runner.py#L47 that is breaking in the notebook

[johannct]

yes, but this is because jupyter notebook seems to handle argparse bizarrely. Not super surprising given the fact that it is really not a module meant for notebook, is it? L.47 is exactly where the code is when we run it as an executable

[raphaelshirley]

Ok as long as it is never needed to run in a notebook I am happy to just remove the test and accept that the argument parser has a few system dependent differences with regard to producing a SystemExit.

[johannct]

To go back to the start of this : we do not run unit tests on notebook, so the test of system exit is, no matter what, irrelevant.

[raphaelshirley]

I agree. We can close the issue after removing the test as far as I am concerned.

[raphaelshirley]

We could also replace it with a test that the Warning is printed:

https://pavolkutaj.medium.com/how-to-test-printed-output-in-python-with-pytest-and-its-capsys-fixture-161010cfc5ad

[johannct]

why not, thanks for the pointer, committed in c4cfe6c