Overlap Helper Module about iodata HOT 7 CLOSED

We should. I put the harmonics.py code into the gbasis repo I think, but I didn't think about the overlap code in IOData.

from iodata.

I have been working on iodata all day today trying to improve it, and I think its overlap code is over-complicated and even out of place to some extent. If I am not wrong, it's only used in the molden module to check normalization, so we don't even need the fastest overlap computation. I think we can write something much simpler for iodata and have a better/faster implementation in gbasis.

from iodata.

Yep, it's only used to check the normalization and normalize if needed. We don't need to have the full overlap matrix implementation in theory.

What did you have in mind?

from iodata.

I am implementing a simple and easy to understand compute_overlap function based on @PaulWAyers notes. It seems to be pretty efficient, so it may not even need to be Cythonized considering it's limited use in iodata module. I will share it with you this week.

from iodata.

Sure, sounds good!

from iodata.

You need the entire overlap matrix to check the normalization of the orbitals. We're not (only) checking the normalization of the basis functions.

A better name for this module would be good: transform_cart_pure.py or something along those lines.

We could compute the transformation matrices when needed and cached them. The harmonics.py in the tools directory of HORTON uses Sympy, which should be avoided. It is a tedious thing to implement in straight Python but good practice when your intuition for recursion relations is fading.

from iodata.

@FarnazH This seems to have been finished (a long time ago) but never closed.

from iodata.