I'm late to the thread, but I've felt the same pain and wrote a library to help deal with the cognitive load of working on "shapes" of data [1].

[1] https://github.com/escherize/tracks