Michael F. Herbst

As requested by @lobpcg and @mohamed82008 I came up with another test case taken from our DFTK code as an example, see [this gist](https://gist.github.com/mfherbst/40df13f8406e49903eb330189ea7c3fc). I just ran it on my...

I amended the gist to allow for requesting a varying number of eigenvalues. For `nev==9`, where the number of requested eigenpairs cuts through the two-fold degenerate eigenvalue 0.851732, the success...

Just to confirm that in `DFTK` our usage of GenericLinearAlgebra is also the second one mentioned by @andreasnoack. I'll also mitigate by moving the dependency to the test dependencies. We...

I agree with you @JeffreySarnoff that this is one of the absolute strong points of Julia and in principle one should not give up on this easily. In the case...

Thanks for the reminder. I'll take a look!

Hmm. I can really see this potentially being very confusing for new users ... in particular if this is exposed to them via the `scfres`. The frequent constructs like `eltype(eltype(...

Ah now I get why componentarrays don't work: They encapsulate the wrong way round. But in that sense your basic "element" becomes not `T`, but some StaticArray type, right? So...

OTOH, I have to admit, it's pretty nice to see this only adds 200 lines of code so far. So our code structure in DFTK seems pretty decent in general.

BTW. I guess it might be good to schedule a call to discuss this after the long weekend.

It was really random. Like I had it once, but then removed the workaround later and it was ok, but had to reintroduce the workaround at some other point to...