Sheehan Olver
Sheehan Olver
No because having more bands than the dimension is permitted I would recommend not using random bandwidths in the unit tests
Note if we want to support many operators we should add support for vector arguments eg ApplyArray(*, [A1,…,An]) but removing the need for ApplyStyle seems like a first step
That probably won’t be the actual constructor. Note args is usually a ` Tuple ` but I think will accept a ` Vector ` already. Reduce doesn’t return an array...
In that case you want to support ` ApplyArray(reduce, +, A) `
No because I don't agree with the proposed change to ArrayLayouts.jl so this shouldn't be needed
We should probably restructure support for block arrays as an extension in ContinuumArrays.jl
Ive come across the issue of LinearAlgebra imposing different components have the same type many times. My solution has been to copy the code with an extra templated variable, see...
Probably we just need to overload `findblock(::BlockedOneTo, ::RealInfinity)` to return `Block(ℵ₀)` and change the `Int` call to `Integer`
At the moment it would go in InfiniteLinearAlgebra.jl. In the future that package will be broken up into extensions in InfiniteArrays.jl at which point it would be in InfiniteArraysBlockArraysExt.jl
Nice catch! A PR to fix this would be greatly appreciated.