ambiguity in definitions of purity and side effects about functional-programming-jargon HOT 8 OPEN

PRS are always welcome. Understandability is more important than accuracy in this document so if you can think of a good, simple wording I'm sure it'll get pulled

from functional-programming-jargon.

Feel free to assign this to me.

from functional-programming-jargon.

👍

from functional-programming-jargon.

@daira Thank you, please feel free to do a PR.

from functional-programming-jargon.

Untill now I understood that referencial transparency and purity were the same thing, but it turns out a function can be referentially transparent and have side effects, so it is not pure.

Any reason why Referencial Transparency is so far below purity and side-effects? I think the right order of jargons helps alot, and as I understand that purity = referential transparency + no side effects purity should be last shown of these three.

I am new to these concepts and haven't used any strictly functional language yet, so feel free to ignore me.

from functional-programming-jargon.

A side effect is an operation which is not pure. The purity implies the referential transparency (see the confluence of the lambda calculus). A function can also be observationaly pure, but with side effects inside. For example :

function sum(a, b) {
  let s = a;
  s = s + b;
  return s;
}

We do a mutation s = s + b, so a side effect. However, the function sum is observationaly pure as it always returns the same result given the same arguments, and does not modify its environment.

from functional-programming-jargon.

Interesting classification of effects done in this paper

but the basic effects defined in Koka are total, exn, div, ndet,
alloc<h>, read<h>, write<h>, and io. Of course total is not really an effect but signifies the absence of any effect and is assigned to pure mathematical functions. When a function can throw an exception, it gets the exn effect. Potential divergence or non-termination is signified by the div effect. Currently, Koka uses a simple termination analysis based on inductive data types to assign this effect to recursive functions. Non-deterministic functions get the ndet effect. The effects alloc<h>, read<h> and write<h> are used for stateful functions over a heap h. Finally io is used for functions that do any input/output operations.
alias total = <>
alias pure = <exn, div>
alias st<h> = <alloc<h>, read<h>, write<h>>
alias io = <st<ioheap>, pure, ndet>

from functional-programming-jargon.

Add totality definition http://alvinalexander.com/photos/totality-rule-functional-programming

from functional-programming-jargon.