Fix error with new numpy version

[BStoelzner]

np.product doesn't work with the new numpy, so I changed it to np.prod

[BStoelzner]

The tests seem to fail because of an incompatibility between numpy and pandas, see numpy/numpy#26710
A popular solution at the moment seems to be to just downgrade numpy or pandas.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 99.59%. Comparing base (60d6359) to head (064be05).

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #46      +/-   ##
==========================================
+ Coverage   97.97%   99.59%   +1.61%     
==========================================
  Files           7        7              
  Lines         742      738       -4     
==========================================
+ Hits          727      735       +8     
+ Misses         15        3      -12     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[hangqianjun]

Looks good to me, and thanks for adding the unit tests!

One thing I'm not sure about is the numpy version limit - I think a number of people argued against that as it might make co-updating rail and numpy harder in future versions. People also tend to use the newer versions of numpy, and reverting to older version in order to use rail may be slightly non-ideal.

Having said that, setting numpy<2.0.0 is an effective and short term solution, because the issue presented here seems to be unsolvable within rail, and needs actions from external packages.

Any thought from other reviewers?

[hangqianjun]

Do we want to set the numpy version limit?

[BStoelzner]

Another option would be to require pandas>=2.2.2, as this seems to be the first version that is compatible with numpy 2.0, see https://pandas.pydata.org/docs/whatsnew/v2.2.2.html#pandas-2-2-2-is-now-compatible-with-numpy-2-0

[BStoelzner]

I just tested setting pandas>=2.2.2, but this leads to a conflict with photerr because this requires pandas<2.0.0. @jfcrenshaw, is there a possibility to make photerr compatible with pandas 2.0?

[ztq1996]

Okay here is the thing:

photerr==1.1.0 requires pandas>1.4.3 <2.0.0
Now photerr==1.3.0 has loosen the requirement to pandas>1.4.3 only, but it only supports python>3.10, so python 3.9 still won't work.

[ztq1996]

ok