Friday, December 7, 2012

Functional Programming in 2012

This was first written for josetteorama as publicity for CUFP 2012. but I thought I would share it on here as well …

Functional programming ideas have been around longer than computers: Church’s lambda-calculus was invented in the 1930s as a way of describing computations as functions around the same time that Turing was describing what an idealised (human) “computer” could do through Turing machines. The idea of using functions as the primary means of computing fed straight into Lisp and its (LAMBDA …), ideas of procedures in Algol68, but settled into what people now call functional languages in the late 1980s, with the definition of ML, Haskell and Erlang.

Erlang – which is really a language for high-concurrency, fault-tolerant, robust systems – is based on functional ideas because it’s so much easier to write concurrent systems when every assignment is a single assignment (check out Single Assignment C too). Haskell and ML (and also OCaml, SML, …) are general purpose languages where types do much of the heavy lifting for you: types are inferred from what you write, and if you have got a program that type checks then you’ve ironed out a whole class of errors before running a single instruction.

As I’ve said, these languages were first defined more than twenty years ago, so what are the modern functional programming languages? One answer is Java, and other mainstream languages, which are taking on more and more functional features. Using immutable data is inherited from the functional world, as were Java generics. Coming soon are closures – yes, that’s right, LAMBDA in Java! So you can do functional programming in Java (and O’Reilly have recently published Functional Programming for Java Developers: Tools for Better Concurrency, Abstraction, and Agility).

But what if you want to try the real thing? A first choice is F#, which puts a functional language into the .NET framework as a first-class citizen, with internationalization, intellisense, and full integration with C# and the .NET libraries. One of the ways of getting started with F# is to use it to explore some of the library APIs, executing calls interactively. Extending Java to give a fuller functional language, as well as Erlang-style concurrency, is Scala. From Scala too, there’s access to the Java libraries, as well as execution on the JVM. Finally, it’s worth taking a look at Haskell, which is functional first, but which provides controlled access to concurrency, IO, exceptions and a wealth of open source libraries.
But, why bother, you might say? One reason is sheer curiosity: you’ll find out what all the fuss is about, and see where ideas like map-reduce come naturally out of the programming style. It also give you a different way of attacking problems: even if you’re going to be programming in Java, then a functional style might well give you a different tool to solve the problem at hand.

Another reason is to get a job (really!). How are functional languages getting used? Erlang and Scala support high-throughput message handling and web systems, and are the “secret weapon” behind applications supporting Facebook, Amazon and other systems. Haskell gets used in building domain-specific languages in security, crypto and increasingly in the financial sector, where F# is getting traction too. You’ll see a steady stream of jobs calling for Erlang, Haskell, Scala, F#, as well as for people who are happy using one of these languages and C, or Java or whatever, because functional languages often come in as a solution to part of a problem, rather than solving the whole thing.

Added December 2012: Another sign of functional programming getting traction is Simon Marlow, one of the "two Simons" supporting GHC, moving to Facebook in the new year …

No comments:

Post a Comment