Sheehan Olver
Sheehan Olver
I thought Gauss–Kronrod tends to be low order eg `n = 15`. So does point (a) really matter?
Sounds more like a research question (not just a coding one) in the first instance. There would have to be a very strong application for this to be pursued though...
I don't think we are not religious about `O(n)` so it could live here. I see your point about error estimate.
I thought the plan was to move to more of an interface than an abstract type? Why not use this as an opportunity to just delete `Domain{T}` once and for...
One could argue that `matrix \ vector` for least squares is also confusing and that there should be a separate syntax when `matrix` is not-invertible, e.g., `matrix \\ vector`, when...
I use `/` all the time... e.g. `1 / 2`.... I like the principle of `/` and `\` being just left and right versions of the same operation. Anything else...
Have you run a debugger through your tests to double check they are being called?
Ah yes you are right, it is better to return a `Partials`. It's fixed now
Given that nothing is currently calling `partials(::Complex)` I don't see how this change can result in that happening, but yes there is a question about what `Partials` can represent.
I was confused by this: ```julia julia> Q = qr(randn(10,5)).Q; julia> size(collect(Q)) (10, 10) julia> size(Matrix(Q)) (10, 5) ``` At the very least, the behaviour should be documented. (The doc...