`VisibleDeprecationWarning` in llh.py

[mlincett]

Trying to test a stacking analysis with a box-like time PDF, I get the following warning:

kwargs["SoB_spacetime_cache"] = np.array(SoB_spacetime)
/lustre/fs20/group/icecube/mlincett/software/flarestack/flarestack/core/llh.py:912: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray.

Not sure whether this is due to something wrong in my inputs or it is inherent to the way the code currently works, any clue?

[robertdstein]

I suspect this is a numpy version change from when I coded it up N years ago. I think just follow the suggestion and do np.array(SoB_spacetime, dtype=object)?

[mlincett]

It seems this deprecation is in full force in python 3.10.2 numpy 1.24 so we should likely fix this ASAP.

[mlincett]

Fixing this in #266 . Just wanted add here that when this segment of the code throws an exception the program does not exit, hence the CI jobs were dangling for hours before getting killed.

[mlincett]

A minimal fix is now available in PR #266 however it requires to use astype(float) when accessing the loaded arrays.

I guess np.array is not really the best option to store a collection of non-homogeneous arrays and maybe we should just use pickle for this.

[mlincett]

After checking a bit better the code structure, it seems there's no real benefit from wrapping SoB_spacetime_cache in a numpy array.

Note that this code is only used the "standard" LLH.