The paper you posted is a good example. The conclusion is:

    sudoku = map (map head) . search . prune . choice 

Note how the "code" never mentions the data to which it is applied, namely the Board object. This is why I can't agree with the GP's argument that FP is closer to "code" than anything else -- I think it's as far from "code" as you can get. There are no "instructions" here, just an equation representing something complex ("the solution to sudoku"), in terms of slightly simpler parts. The simpler parts are, earlier in the paper, the complex parts; and are of course described there by simpler parts.

And, this is not just something that's useful for esoteric mathematics papers; my "enterprise" code in Haskell looks very similar.