pyheatmy: Potential NaN values in core.py

[pierre-guillou]

There seems to be a stability issue with pyheatmy/core.py around line 1022 in np.random.choice(ncr, p=pcr[l]), pcr[l] sometimes contains NaNs that make np.random.choice raise ValueError: probabilities contain NaN.

This issue has been noticed when running demoPyheatmy.ipynb under the GitHub Actions workflow (c.f. #80).

An additional check should be added somewhere in the function to ensure that NaN value are properly handled: should we replace them with 0? should we skip the corresponding for loop iteration?

[flipoyo]

I added those line in core.py if not ncr: raise ValueError("ncr est vide.") if l >= len(pcr): raise IndexError("Index l est en dehors des limites de pcr.") if np.isnan(pcr[l]).any(): raise ValueError("pcr[l] contient des valeurs NaN.") if not np.isclose(sum(pcr[l]), 1): raise ValueError("Les probabilités dans pcr[l] ne sont pas normalisées.")

I have python 3.11.4 and don't have any errors raised when executing demoPyheatmy.ipynb @pierre-guillou Maybe you can test it again with the CI (I add an issue) so we can have more explainations. I suspect this is linked to DREAMS. I open 2 new issues one for the CI and one for DREAM

[flipoyo]

lines added in core.py 1021-1028