Support for sparse vectors/matrices
It'd be great if support for sparse vectors/matrices could be made to work.
I've wanted this a lot recently while working with sparse document-term matrices.
My use case exactly! I'll poke around with it, but it seemed there were view implementations missing, amongst other things, so not just a matter of defining a method or two.
PRs are welcomed. But I don't have much time working on this in near term.
Is there any specific reason that At_mul_b for two sparse matrices multiplication not implemented in base/linalg/sparse.jl?
I had to bypass At_mul_b via spmatmul to leverage pairwise work with sparse matrices, though the performance is not as blazing fast as using dense matrices.
For @quinnj 's view problem, Tim Holy pointed out SubArrays support views of sparse matrices.
With the two inconsistent and ugly modifications above on pairwise, one can merely make it work, whereas further work is still needed.
Until type coverage for in-place in-place functions isn't decided, how about using InplaceOps for abstracting in-place functions?
Fixing this issue might help in implementing this.
I made a PR #235 to add a good bit of support for SparseVector, although it doesn't cover everything. In particular I did not support or even look into matrices.