Regarding ApproxFun.jl (likely more general): what is the rationale (guideline or rule of thumb) for setting or unsetting the space for the Derivative function?
1 Like
I think it’s analogous to the identity operator I in LinearAlgebra — when you act I*x on a vector (or matrix), it infers the correct dimension of the identity from x, analogous to UnsetSpace in ApproxFun. But occasionally it’s nice to have an explicit identity matrix, e.g. via I(n) to get an n \times n Diagonal matrix.
3 Likes
From a practical perspective, it’s almost always better to not specify the spaces for operators, and have these be inferred automatically.
Hurray for automatism.
Hurray for an explanation of what and how the automatism does.
Hurray for a switch allowing to turn off the automatism allowing user to experiment.
Thx.