dotJS 2012 - Axel Rauschmayer - Multiple dispatch

Filmed in Paris on Nov 30th, 2012. More talks on http://dotconferences.eu
Slides: http://goo.gl/EIR40

Hello, so my talk is going to be on multiple dispatch.

First for starters what is method dispatch?

If you have a polymorphic method

that mean if you have multiple implementations

and we'll soon see an example to make it clear what I mean

if you have those multiple implementations, the dispatch determines

which of those implementations will be called when you

invoke the method.

JavaScript has simple dispatch, so a single value is used for dispatch

we will then look at something new

and that is called multiple dispatch.

You can either dispatch via an object or

via a function. And if you dispatch via an

object this is what it looks like.

You have a so-called polymorphic method

and the type of jane determines which of the two functions

will be called if you make that method call up here.

So it depends on whether jane is a person or an employee

If on the other hand you have a generic function

then things work differently

jane is now just a normal argument and

the function internally checks the type of the argument

and then depending on the type of the argument

chooses between two different implementations.

The same principle can be used if you have more than one parameter

and naturally you then have to check the type of each

of those two parameters.

So if you implemented multiple dispatch as a library in JavaScript

this is what it would look like.

You would first create a generic function that's called plus

then you would add two methods

and finally you would invoke it as you would with any other function

and then the types of the parameters

determine which of the two implementations are invoked when you make the call

as you can see it's like a table that maps function types

or parameters types to methods.

What are the benefits? Why do you do that?

What's good about it?

So what's actually happens is

you get more symmetry

Normally, the implicit parameter "this" is more powerful

than the other explicit parameters when you have

single dispatch like in JavaScript

So that's one of the properties of multiple dispatch

And then functions also become object-oriented

they become aware of the type hierachy of objects

The benefits are that whenever you have a polymorphic algorithm

that applies to several objects, then you can put that into a generic function

and here you have the generic function foo

it works with three objects and then it's obvious that foo

doesn't belong to a single object

it belongs to all of the objects at the same time.

So generic functions allow you to do that

and you have the visitor pattern for

single dispatch languages and that is actually just

a generic function implemented as an object

an you use single dispatch in the host to trigger the right visitor method

and languages with multiple dispatch include

Common Lisp, Haskell, R, Groovy, Clojure and C#

That's it, thank you!