CombineBond should exist as non-member function

[manuschneider]

It is a bit counterintuitive that combineBond is a member function:

bond1 = cytnx.Bond(10)
bond2 = cytnx.Bond(2)
bond12 = bond1.combineBond(bond2)

In this case there should be no difference between bond1 and bond2. I would suggest to have a non-member function instead:

bond1 = cytnx.Bond(10)
bond2 = cytnx.Bond(2)
bond12 = cytnx.CombineBonds([bond1, bond2])

[IvanaGyro]

What does no difference between bond1 and bond2 mean?

[pcchen]

We want to combine bond1 and bond2 into a new bond=bond12. In terms of physics, bond1 is not special. So, why bond1.combineBond(bond2) and not bond2.combineBond(bond1).

I would say the only different is memery layout.

bond12 = cytnx.CombineBonds([bond1, bond2]) bond21 = cytnx.CombineBonds([bond2, bond1])

differ only in memory layout.

I also tend to use bond12 = cytnx.CombineBonds([bond1, bond2]). And the order in the bond list naturally corresonds to diffetnt memory latyout.

[kaihsin]

The reason they stay for member function is because you might need different implementation depending on the impl of UniTensor (e.g. Fermion/Dense/Sparse)