AttributeError: module 'scipy.sparse' has no attribute 'coo_array'

[aafshinfard]

This is due to scipy and networkx conflicts.

Solution: As a workaround, I figured out that we should either have scipy and networkx recent versions, scipy v1.8 at least or downgrade networkx to networkx v2.6.3.

[aafshinfard]

We may try to somehow formulate this in our Conda recipes? doing simple scipy >= v1.8 is too constraining.

[lcoombe]

@aafshinfard - So if you have both most recent versions of scipy and networkx, you get that error? And on the conda side, they are compatible? (ie. would be installed at the same time, no constraints that would change the version of either?)

[aafshinfard]

If you have the most recent versions, you're good to run but scipy v1.8 requires python >= 3.9.sth and seems like pypy is not available for that and more recent versions ... so you will have to run with python instead of pypy.

I have not build a recipe with both pypy and scipy>1.8 but based on experiments I am guessing that it would not work at the same time. Maybe an incorrect guess, but will come back to check it soon. Just wanted to have a record of it here for ourselves and the users.

[aafshinfard]

currently, we have no constraints on scipy. When I tried, it installed scipy v1.7 for me with python 3.7 and I got that error running Physlr. I downgraded networkx and its working alright again.

[github-actions[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your interest in Physlr!

[aafshinfard]

Great news is that according to this post, pypy now works for python 3.9. Just double-checking if everything works fine with python v 3.9 and I will update conda recipes to require python>3.9 and avoid the issues mentioned above.