Pol Febrer Calabozo

Does `np.trace` already work? I don't know if it's a good idea to introduce the matrix multiplication on a simple `trace` method, because it feels like an arbitrary choice (you...

But it is also true that you can acheive that simply with `np.sum(H * dm)` :thinking: I wonder if having the explicit method helps, because I had not thinked that...

> Hmm, I think we shouldn't do complicated functions that are allready standardized, e.g. matmul and trace. Yes, but the point is that `trace(matmul(X))` doesn't actually require `matmul(X)` > Well,...

I think it is simpler to make something like `np.sum(H * H.S)` work. That doesn't necessarily involve separating the overlap in storage as I understand you propose in #770.

When we agree, I will change the docs that show how to combine plots.

> I think most people will use grouped plots to combine different types of plots (i.e. not just a variable). > > For instance, geometry plots + PDOS scaling of...

Another example: ```python from nodify import Batch, temporal_context with temporal_context(batch_iter="product"): H.plot.wavefunction(i=Batch(4, 5, 6, 9), represent=Batch("real", "imag")) ``` would generate 8 plots with 1 diagonalization and 4 grid projections.

> what does temporal_context mean? Isn't temporal implicit in contexts? I would think you should name this differently, nodify_context... or? Hmm interesting, so `context` here is referring to the nodes...

You mean after the comment? No I didn't write anything, it's just that if you zip those three batches you end up with the same as what you were proposing.

I'll try to check if everything is ok. I recall that the thing missing was documentation and tests