btmit

I often find myself needing to operate on a subset of a sparse array, typically a few adjacent columns. Creating a new `SparseMatrix` is relatively expensive, so being able to...

`nonzeros` appears to be giving the wrong answer for this case. See example: > julia> a = sparse([1 0 2; 0 3 0]) 2×3 SparseMatrixCSC{Int64, Int64} with 3 stored entries:...

`eigen` and `svd` with `Symmetric{Float16, ...}`

2

`eigen` and `svd` of `Symmetric{Float16, ...}` returns values with element type `Float32`. This is not the case when the input is not wrapped in `Symmetric`. `cholesky` decompositions don't have this...

Inverse Triangular methods missing

2

The inverse of a Triangular `SMatrix` does not preserve type. The resulting triangular is based on `Matrix` not `SMatrix` because the [fallback method](https://github.com/JuliaLang/julia/blob/147bdf428cd14c979202678127d1618e425912d6/stdlib/LinearAlgebra/src/triangular.jl#L803) is hard-coded to `Matrix`. It seems like...

`*` and `UniformScaling`

2

I'm seeing some strange typing when multiplying `StaticArrays` and `UniformScaling`. Is all of this expected? ```julia N = 2 typeof(rand(SMatrix{N,N}) * I) # SMatrix{N, N} typeof(rand(SMatrix{N,N}) * I(N)) # Matrix...

Currently passing this argument is causing dispatch back to the original `LinearAlgebra` methods. Here's a first cut at replacing the existing method: ```julia function LinearAlgebra.cholesky(B::BlockDiagonal{T,M}; check=true)::Cholesky where {T,M} N =...