Sorry for not getting back to you on this more promptly!

What do you mean by "simplicial LL would require one to change Common->final_ll to 1 but I think it is not possible through scikits.sparse interface."? Anything that can be done in C can be done in Cython -- is there some change needed to cholmod.pyx?

Comments:

• can you rename the directory to cholmod_sparse_inverse or similar? We may e.g. add UMFPACK support at some point, but this code is clearly cholmod-specific.
• the spinv function has a different license than either CHOLMOD or the rest of scikits.sparse. We need to either document this somewhere more visibly (including the docstring), or, would it be possible to relicense it under GPLv2-or-later instead of GPLv3-or-later, so it matches the rest of CHOLMOD? (And can you confirm that the bit of code you're okay with licensing the bit of code added to cholmod.pyx under the BSD? That also makes it easier to keep track of, if we can keep the whole file under a single license...)
• Can you add a test that spinv does something sensible when called on a complex matrix? i.e., raises an appropriate exception (as opposed to returning nonsense or crashing).

• By saying that "it is not possible to change Common->final_ll to 1 through scikits.sparse interface", I mean that the users of scikits.sparse do not have access to it via the Python interface of scikits.sparse. Of course, in scikits.sparse's Cython code (cholmod.pyx) it would be possible to change this, but at the moment it is untouched.
• Yep, folder renamed to cholmod_sparse_inverse.
• The C-code (the files in cholmod_sparse_inverse folder) is now GPLv2-or-later. I confirm that my bit of code in cholmod.pyx, setup.py and test_cholmod.py are licensed under the BSD.
• Factor.spinv should now raise CholmodError exception for complex matrices. There is also a test for it in test_cholmod: test_spinv_complex.

Hi! This feature would be indeed very helpful. Is there any chance that it will make it to one of the future releases?

I'm also very interested in this.

It seems like Suitesparse has implemented this though confusingly in MATLAB_Tools:
https://github.com/DrTimothyAldenDavis/SuiteSparse/blob/stable/MATLAB_Tools/sparseinv/sparseinv.c