Minor corrections

[MikhailRyazanov]

1. I've noticed that Python 3.12 started to complain (correctly) about an invalid escape sequence in setup.py because a regex string wasn't prefixed with r.
2. Supported Python versions are updated to 3.12. We can't test it automatically yet (at least it's not available for Windows in Appveyor), but it works locally, and in fact the conda packages for it have been built 2 weeks ago (without any complains so far).
3. The README had a confusing mistake: abel.center.center_image instead of abel.tools.center.center_image (perhaps danging from some old times).
4. I think, it would be useful to direct questions to "discussions" rather than "issues", in line with Dan's comment, so I've revised the "Support" section accordingly. Any objections?

[stggh]

All looks good @MikhailRyazanov, thanks.

Some additional comments:

• Item 3. abel.tools.center.center_image is correct, but abel.center_image() also works because of the content of abel/__init__.py:

 (py312)[Gibbo:.+MR-tmp/examples]$ ipython
      
 In [1]: import abel

 In [2]: abel.center_image?

I don't know whether this is a good or bad "feature".

• Not due to any changes in this PR. I ran examples/example_all_O2.py and it failed with the error:

  File "/home/stg121/anaconda3/envs/py312/lib/python3.12/site-packages/abel/tools/polar.py", line 71, in reproject_image_into_polar
    nr = np.int(np.ceil((r.max() - r.min()) / dr))
         ^^^^^^
  File "/home/stg121/anaconda3/envs/py312/lib/python3.12/site-packages/numpy/__init__.py", line 324, in __getattr__
    raise AttributeError(__former_attrs__[attr])
AttributeError: module 'numpy' has no attribute 'int'.
`np.int` was a deprecated alias for the builtin `int`.

(py312)[Gibbo:.+-tmp/abel/tools]$ grep 'np.int' *.py
polar.py:    nr = np.int(np.ceil((r.max() - r.min()) / dr))
polar.py:        nt = np.int(np.ceil((theta.max() - theta.min()) / dt))

A convenient point to correct the above file. (Edit: Only the one file I think)

[MikhailRyazanov]

Thanks, @stggh!
It seems that @DanHickstein has imported center_image directly to abel in a4f46f9, together with transform (which later became Transform), probably because these are the most important functions. I guess, it would then make sense to use abel.center_image() in the README and mention this alias in the center_image() docstring. (Another possibility would be to deprecate abel.center_image()...)

Regarding np.int, I've corrected them already in e5c2f86. Are you using some old versions?

[stggh]

Regarding np.int, I've corrected them already in e5c2f86. Are you using some old versions?

Hmm, it appears so. I re-cloned your tmp and now all is well. Sorry, about that!

[stggh]

Another possibility would be to deprecate abel.center_image()...

I agree with this.

[DanHickstein]

Another possibility would be to deprecate abel.center_image()...

If I recall correctly, our thinking at the time was that center_image is such a commonly used function that, while burying it in abel.tools.center.center_image() makes sense from the viewpoint of code-organization, it was too annoying for users. Currently we can access the function via both abel.center_image and abel.tools.center.center_image, correct? In my view this is the desirable behavior. I think adding a statement to the docstring to indicate the shortcut is a good idea.

But, my feelings on the issue are not too strong, so I could be convinced otherwise.

And directing people to the Discussions feature seems like a reasonable idea. If we don't like it, it's easy enough to switch back to Issues.

Thanks again for keeping PyAbel up-to-date @MikhailRyazanov!

[MikhailRyazanov]

A difficult decision... :–)
In principle, I agree with Dan that allowing abel.center_image() is more user-friendly than requiring abel.tools.center.center_image() (and centering an image before transforming it might be useful for sanity checks, although for this purpose, the centered input image is also available as Transform.IM). On the other hand, Steve, as a real user, doesn't think that the shorthand is necessary; and I don't know whether anybody at all uses it, since it wasn't documented anywhere.
We can flip a coin, if nobody has strong arguments either way. :–D

[stggh]

Steve, as a real user

Not so much these days. No new data from my lab, as it has been closed down, and the spectrometer dismantled :-(

abel.tools.center.center_image()

I prefer that the explicit call provides the directory location of the function (and API)

Not a big deal. I am happy with any call.

[MikhailRyazanov]

OK, I'll merge this as is (using the full path in the README), and let's discuss in #379 what should be done with abel.center_image().