Stefanos Carlström

I am particularly interested in the last two items in the list. For the custom metric, I implemented the Löwdin transform that symmetrically orthogonalizes a non-orthogonal basis, such that the...

Well, I implemented the Löwdin transform by eigendecomposing the overlap matrix (when it is not simply an `UniformScaling`) and caching the forward and backward transforms. I was more thinking if...

Yes, I found out that that works. Now I only need to figure out if I want `z^2 == z*z` or `conj(z)*z == abs2(z)`, but that's an issue for me...

I forgot that LoopVectorization does not work with complex numbers out-of-the box, but the same warning appears when changing to `Int`s in the example above.

> It might not be hard to always re-organise multi-line expressions like this. Meaning the last line of the block is the summand, basically?

@xtalax Thanks for the heads-up! As you see in the latest commit to this PR, I did not do anything big, just cleaned the concretization code. However, I did not...

Just FYI, CompactBases.jl has the lowest-order (tridiagonal multiply, tridiagonal solve) compact finite-differences scheme already implemented: https://juliaapproximation.github.io/CompactBases.jl/stable/fd_overview/. Not very good support for other boundary conditions than Dirichlet0 yet.

I came up with the following since I wanted both to pretty-print normal quantities and amounts of data: ```julia using Unitful using Formatting @dimension 𝐃 "𝐃" Data @refunit byte "B"...

I assume the intermittent CI failures (e.g. https://travis-ci.org/JuliaArrays/LazyArrays.jl/jobs/633648027#L305) are related to https://github.com/JuliaArrays/LazyArrays.jl/pull/90#issuecomment-568234141?

What do you think about the idea? @tkf any input?