Previous fix was wrong. Re-fixed

[scemama]

The normalization was not applied at the proper place in the previous commit. I fixed it.

[q-posev]

@scemama would it be possible to add a regression/unit test for this functionality? I feel like it's something that is easy to break and hard to catch.

[scemama]

@q-posev I detected it running trexio check-basis and trexio check-mos. So adding a test is not too difficult.

[q-posev]

@scemama we do run these functions on GAMESS->TREXIO output. Does this bug occur after spherical to cartesian conversion? We have PySCF and ORCA outputs that undergo that in the testing CI.

[scemama]

I just added a test.
Yes, it occured after the sphe->cart conversion. The check-mos was correct in spherical, but wrong after conversion. This explains problems when converting some PySCF spherical files and importing them into QP.

[q-posev]

@scemama I fixed the tests.

Should we include qmckl in the test environment to accelerate the check-mos executions? I can use the latest released tarball to build the qmckl python api

[scemama]

I am afraid that adding the qmckl dependency and compiling it will take as much time as the check-mos.... maybe we can just reduce the number of integration grid points. Even with 30 points, i think we can detect if someting is wrong

[q-posev]

Yes i also thought that the compilation will be slow. If only someone would make a Python package for qmckl... 😀

I actually increased the number of points in this MR and fixed the error thresholds for each file. This is the first attempt to address issue #12 but ideally it would be nice to store some converted trexio files as a ref somewhere and do a more rigorous check of converters using these ref files.