• Share
  • Email
  • Embed
  • Like
  • Save
  • Private Content
The other side of functional programming: Haskell for Erlang people
 

The other side of functional programming: Haskell for Erlang people

on

  • 8,006 views

These are the slides I presented in a talk at the Bay Area Erlang Factory, introducing Erlang programmers to some of the ideas behind Haskell.

These are the slides I presented in a talk at the Bay Area Erlang Factory, introducing Erlang programmers to some of the ideas behind Haskell.

Statistics

Views

Total Views
8,006
Views on SlideShare
7,496
Embed Views
510

Actions

Likes
9
Downloads
198
Comments
0

7 Embeds 510

http://www.serpentine.com 468
http://www.linkedin.com 16
http://www.slideshare.net 12
https://www.linkedin.com 11
http://www.feedage.com 1
http://www.mefeedia.com 1
http://blogeotime.wordpress.com 1
More...

Accessibility

Categories

Upload Details

Uploaded via as Adobe PDF

Usage Rights

© All Rights Reserved

Report content

Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
  • Full Name Full Name Comment goes here.
    Are you sure you want to
    Your message goes here
    Processing…
Post Comment
Edit your comment

    The other side of functional programming: Haskell for Erlang people The other side of functional programming: Haskell for Erlang people Presentation Transcript

    • The other side of functional programming Haskell: purity, types, and a damn good time Thursday, April 30, 2009
    • In the beginning Haskell was developed by academics to unify many streams of research Committee began work in 1987 Haskell Report 1.0 published April 1, 1990 Comparable in age with Erlang Thursday, April 30, 2009
    • Principal concerns Laziness Purity Strong, static types Thursday, April 30, 2009
    • Being lazy with style Thursday, April 30, 2009
    • Laziness The original unifying theme of the designers Evaluate an expression when its result is needed Evaluate only the minimum needed Thursday, April 30, 2009
    • A simple lazy example A simple Haskell function definition square x = x * x Define the name square Give it a free variable x The function body follows the = Thursday, April 30, 2009
    • Evaluate in a lazy world What is the result of square 3 ? The number 9 ? Or the unevaluated expression 3 * 3 ? (It’s the latter) Thursday, April 30, 2009
    • When do we evaluate? When does the expression 3 * 3 turn into something meaningful? For instance, when we need to print its result Evaluation is driven by need Often referred to as call-by-need Contrast with the more familiar call-by-value Thursday, April 30, 2009
    • Laziness as default Laziness is pervasive in Haskell code But sometimes it is not desirable Option: use strict evaluation when necessary Many strict languages provide optional laziness The apparent gulf isn’t so big after all Thursday, April 30, 2009
    • Purity is the new black Thursday, April 30, 2009
    • Purity Haskell data is immutable Functions are pure Only affected by their inputs Not subject to mutable global state (Again, these are defaults: mutability is available as an option) Thursday, April 30, 2009
    • Why purity? What’s a side effect? Mutating global state Performing I/O Remember laziness? Evaluation by need Laziness and side effects don’t mix! Thursday, April 30, 2009
    • Laziness needs purity Haskell chose laziness by default Therefore purity was inescapable This has big consequences Composability: glue functions together Safety: functions are black boxes Thursday, April 30, 2009
    • Purity as a big bet Haskell’s laziness led to its purity Purity is a bold step towards better software Arguably a more important choice than laziness! Thursday, April 30, 2009
    • Adventures with types Thursday, April 30, 2009
    • Strong static types Valid Haskell expressions are assigned types at compile time a :: String a = “some text” The :: says that a has the type String This is called a type annotation Thursday, April 30, 2009
    • Wait ... static types? Aren’t we supposed to hate static types? Didn’t types cause us RSI in Java and C++? Wasn’t that part of why we escaped to the dynamically typed languages? Crummy languages give static types a bad name Thursday, April 30, 2009
    • Yes, static types! A Haskell compiler infers the type of an expression It does this automatically The type annotations that you’ve seen are optional Handy for documentation, but superfluous Thursday, April 30, 2009
    • Simple use of types Any sensible language will reject stuff like this 1 + “3foo” (Notable exception: Perl) Dynamic languages barf at runtime Languages like Haskell reject at compile time Thursday, April 30, 2009
    • Pattern matching Here’s the classic way to calculate a list’s length length [] =0 length (x:xs) = 1 + length xs We’ve defined a function using two equations Choose which to use by input structure Thursday, April 30, 2009
    • Matching on structure If the input list is empty, the length is 0 length [] =0 If the input matches the list constructor :, bind the name xs to the list’s tail and recurse length (x:xs) = 1 + length xs Thursday, April 30, 2009
    • Typing a list What is the type of length? length :: [a] → Int The [a] above means “a list of values of some unknown type a” The → means “returns” In other words, we have a function that does not know or care about the elements of its input list Thursday, April 30, 2009
    • Why use static types? Static types are about more than catching mistakes They let the compiler make complex decisions about the program’s behaviour Thursday, April 30, 2009
    • User-defined containers Here’s a widely used Haskell type data Maybe a = Just a | Nothing We can pattern-match to inspect the structure of a user-defined type isJust (Just x) = True isJust Nothing = False Thursday, April 30, 2009
    • Algebraic data types data ClientError = BadRequest | Unauthorized | Forbidden | NotFound ... Thursday, April 30, 2009
    • What does this buy? If my function takes a HttpResponse The compiler guarantees that I’ll never be given a HttpRequest It guarantees that I’ll never see an unknown HttpResponse It warns me if a pattern match omits a valid response Thursday, April 30, 2009
    • Safety with types Static types give stronger guarantees than testing A simple example: “I know my function can never receive an argument of an invalid type” More ambitious: “This code can never perform I/O” Thursday, April 30, 2009
    • More serious type safety We can omit features that other languages bake in Ship them as libraries instead A recent example: Java-style checked exceptions as a library Throwable exceptions are inferred Thursday, April 30, 2009
    • More serious types We can model and enforce complex behaviour Examples: Information only flows from less secure to more secure code Communicating processes follow a well-defined messaging protocol Thursday, April 30, 2009
    • Real world concerns Thursday, April 30, 2009
    • Performance Haskell is ranked #3 on the Alioth Shootout Usually within 1x to 5x of C’s performance Great profiling tools help with tuning It’s easy to write fast, concise Haskell Community knowledge of how is a bit scattered Thursday, April 30, 2009
    • Going native Haskell has a beautiful FFI Call into and out of C code easily Nifty libraries for other languages Interop with .NET Act as an Erlang node Thursday, April 30, 2009
    • Concurrency Haskell has a fantastic concurrent runtime Works with multiple cores Millions of concurrent threads Advanced, but easy to use programming model The default choices of immutable data and pure functions really help to write correct, scalable code Thursday, April 30, 2009
    • Thread synchronisation Software Transactional Memory Database-like transactional concurrency to regular code Much safer than mutexes Strongly typed message channels Networked message support as a library Thursday, April 30, 2009
    • Parallel terminology Parallel and concurrent programming are different Parallel: how do I get one answer faster? Concurrent: how do I do 80,000 different things per second? Thursday, April 30, 2009
    • Parallel programming Mature support for making pure code parallel Development version of GHC scales well on modern multicore boxes Exciting research abounds Nested data parallel vector code GPU offload Thursday, April 30, 2009
    • Testing and assurance The famous QuickCheck library arose in Haskell Randomised property-based testing Beats the pants off unit tests when applicable Traditional unit testing libraries available too Excellent code coverage analysis tools Thursday, April 30, 2009
    • Libraries Over 1,000 libraries on http://hackage.haskell.org/ Game engines, bioinformatics, networking, database integration, music, compiler tools, ... Single-command install of any library and its dependencies Thursday, April 30, 2009
    • Community The best language community I know of Stellar researchers, informed OSS hackers Atmosphere is friendly, welcoming, and smart Notable absence of rock stars #haskell is 5th biggest channel on Freenode Many great online learning resources Thursday, April 30, 2009
    • Thank you for your time! Thursday, April 30, 2009