47

I've found the SOLID principles quite useful when thinking about object-oriented design.

Is there a similar / equivalent set of language-agnostic principles tailored for functional programming?

mikera
  • 20,617
  • 5
  • 75
  • 80
  • 12
    FWIW, this was discussed briefly on [SO](http://stackoverflow.com/questions/5577054/solid-for-functional-programming) a year ago – StuartLC Sep 19 '12 at 07:49
  • DI and HOF: http://programmers.stackexchange.com/questions/103508/how-is-dependency-inversion-related-to-higher-order-functions – paul Oct 26 '12 at 18:30
  • [This video](http://www.infoq.com/presentations/SOLID-Clojure) as well as [these slides](http://www.slideshare.net/mstine/functional-solid) present the SOLID principles applied to Functional Programming. They both use the Clojure language as an example, but the principles hold in other languages. – mascip Jul 14 '14 at 08:26
  • http://www.slideshare.net/RichardWarburton/twins-object-oriented-programming-and-functional-programming – Ray Tayek Feb 14 '15 at 11:24
  • see also: [Is functional programming a superset of object oriented?](http://programmers.stackexchange.com/q/163432/31260) – gnat Aug 25 '16 at 12:52

2 Answers2

57

SOLID turns out to be a good idea for the functional/imperative realms too.

SRP - 'Only do one thing' was taken from imperative programming in the first place. Having small, focused functions is good.

OCP - Allowing you to change behaviors without modifying code is good. Functional programming uses higher order functions more than inheritance, but the principle holds.

LSP - Abiding by some interface contract is just as good in functional programming as in object oriented. If a sort function takes a comparator, then you would expect the '0 is equals, less than provides negative results, greater than positive results' behavior.

ISP - Most functional languages still have structs. Specifying the smallest set of data required by a function is still good practice. Requiring the least specific interface to the data (why use Lists of ints when Enumerations of T work just as well?) is still good practice.

DIP - Specifying parameters to a function (or a higher order function to retrieve them) rather than hard coding the function to go get some value is just as good in functional programming as in object oriented.

And even when doing object oriented programming, many of these principles apply to the design of methods in the objects too.

Telastyn
  • 108,850
  • 29
  • 239
  • 365
18

It is a bit difficult to find equivalents but I can try:

  • S (SRP) in FP a function creates ALWAYS the same output for the same arguments this is called referential transparency
  • O (OCP) in FP there is a concept called algebraic data types, have a look how it relates to Class hierarchies and what problem both try to solve 1
  • L (LSP) Liskov Substitution Principle is Contravariance 2
  • D (DIP) in general functional programming achieves abstraction through function composition, there are also other mechanism with the help of category theory(for example monoid or functor), for "Dependency Injection" 3
gnat
  • 21,442
  • 29
  • 112
  • 288
AndreasScheinert
  • 454
  • 3
  • 7
  • 25
    I'm still thinking about how did you get from *Single Responsibility Principle* to *referential transparency*. Those two are unrelated. SRP is about a function having a single purpose. It may or may not be referentially transparent, with regard to that. – Goran Jovic Oct 26 '12 at 19:35
  • 4
    Ah, my bad - I get it now. These are equivalents in the sense of being principles and forming the same acronym, not in the sense of meaning the same or similar thing. Sorry for the downvote! – Goran Jovic Oct 27 '12 at 23:15
  • 1
    Right that's the intended way to read it. I tried to describe a mapping for those terms in context of fp. – AndreasScheinert Oct 28 '12 at 18:46
  • Man I hate that you can't edit a comment, in fact the things SHOULD mean at least something similar. – AndreasScheinert Oct 31 '12 at 20:05
  • Maybe higher-order functions can provide some kind of dependency injection: you inject a concrete function as a parameter into a generic (higher-order) function. – Giorgio Dec 07 '12 at 12:22
  • Yes Giorgio, see Here for example: http://marakana.com/s/post/1108/dependency_injection_in_scala – AndreasScheinert Dec 13 '12 at 22:34