I disagree, the reason almost all ‘functional’ programming can not be said to be referentially transparent is because of the level of abstract the designers take as a base. The work Greg Morrisette did (which include Robert Harper for some branches) on Typed and Dependently-Typed Assembly could be used as the basis for a ‘C-level’ programming language that is ‘functional’.
At a slightly higher level, if memory is explicit, and that would include input and output buffers and the like, then it is possible to make functions exist inside a ‘memory monad’, similar to Haskell’s state monad. Then any and all operations involving the memory used can be made referentially transparent.
Now, the real question is would anyone want to program in such a style? I know I wouldn’t mind, it’s while I broke down and designed a personal programming language for my daily use. But it’s a matter of taste and tolerance.
At a slightly higher level, if memory is explicit, and that would include input and output buffers and the like, then it is possible to make functions exist inside a ‘memory monad’, similar to Haskell’s state monad. Then any and all operations involving the memory used can be made referentially transparent.
Now, the real question is would anyone want to program in such a style? I know I wouldn’t mind, it’s while I broke down and designed a personal programming language for my daily use. But it’s a matter of taste and tolerance.