mhg_cython in python 3.9 and above

[zhengrongbin]

Hello, thank you very much for developing xlmhg! I found a problem when using xlmhg in python above 3.8. The mhg_cython seems likely can be installed in python 3.9 but everything looks perfect in python 3.8.

Under python 3.9, I got warning:
Warning (xlmhg): Failed to import "mhg_cython" C extension.
Warning (xlmhg): Failed to import the "mhg_cython" C extension.Falling back to the pure Python implementation, which is very slow.

and I got error report if I run xlmhg testing:
File ~/miniconda3/envs/mebocost/lib/python3.9/site-packages/xlmhg/test.py:207, in get_xlmhg_test_result(N, indices, X, L, exact_pval, pval_thresh, escore_pval_thresh, table, use_alg1, tol)
201 raise ValueError('Supplied array for dynamic programming table not'
202 'large enough. It is: %d x %d, but must be at least '
203 '%d x %d ((K+1) x (W+1)).'
204 % (table.shape[0], table.shape[1], K+1, W+1))
206 ### Step 1: Calculate XL-mHG test statistic.
--> 207 stat, cutoff = mhg_cython.get_xlmhg_stat(indices, N, K, X, L, tol)
208 assert 0.0 <= stat <= 1.0
210 # check for special cases

TypeError: get_xlmhg_stat() takes from 3 to 4 positional arguments but 6 were given

Do you have any solution to this? Thank you very much!

[semenko]

I'm working on patching this so mebocost can run -- looks like the pure python implementation is broken in Py3.8 as well, and you'll get the same error if you forcably use it, e.g. by adjusting the test runner at:
https://github.com/flo-compbio/xlmhg/blob/master/xlmhg/test.py#L14

[GuyTeichman]

@zhengrongbin If you still require xlmhg, I have recently released a fork of this repository that supports Python>=3.8 and addresses some of the issues with the pure Python implementation of the algorithm.
https://github.com/GuyTeichman/xlmhglite
https://pypi.org/project/xlmhglite

[zhengrongbin]

Thank you! I am incorporating xlmhglite into MEBOCOST updates to replace xlmhg. I will keep you posted if I met anything. Appreciate your help.