This sounds like common sense until it becomes common nonsense, because you use Python or Javascript or Ruby or whatever language where you don't have an optimizing compiler and optimized immutable data structures, so what could have been be a single-pass low-memory scan over a big dataset in 10 minutes that could run on a toaster is implemented as an inefficient clusterfuck that takes 12 hours and 4 GB of memory.
It's absolutely important to treat mutability as either a side effect or a local optimization from a design perspective. But Python is not Clojure no matter how hard you try, and at some point you're going to want to mutate a dictionary.
Anyway, the point is that what might seem like "common sense" in some situations is not always obvious or unambiguous in general.
No comments yet.
