one- and two-electron operators

[ilfreddy]

Is there any good (code-related) reason to keep 1- and 2-el operators separate in the code? As a matter of coding, this distinction seems to me rather arbitrary. I'm starting to code ZORA and I am implementing the ZORA "potential" (1-V/2c^2) and its inverse "kappa". In this first implementation I'm only considering the nuclear potential and therefore the ZORA potential is a 1-el operator. I have also implemented the corresponding kinetic energy operator (p kappa p)/2. As long as V is the nuclear potential all such operators will be 1-el, but as soon as I will add the coulomb and xc pieces, they will become 2-el operators and I will have to move them to the other folder. Does not make much sense to me.

[stigrj]

At some point in the past these were actually two different base classes OneElectronOperator and TwoElectronOperator. I think it might be a good idea to re-introduce this separation, so that a TwoElectronOperator takes an OrbitalVector as a second argument to setup, instead of to its constructor, i.e. instead of this (P is Poisson operator)

CoulombOperator J(P, Phi);
J.setup(prec);

we do this

CoulombOperator J(P);
J.setup(prec, Phi);

This way the operator is not tied to a particular Phi.

[stigrj]

But short answer: no it doesn't matter, put it where ever you want

[ilfreddy]

But then if you don't need Phi in the constructor, why would you want two base classes? Setup is anyway different for each derived class (or is it?). Or do you assume that each operator has a setup(prec)?

[stigrj]

All operators must have a setup(prec) function

[ilfreddy]

OK. I have no strong opinion about having the base classes. If you thin k it is a good idea, we can do it.