When are multimethods useful in practice?

Question

The Common Lisp Object System (CLOS) supports multiple dispatch (multimethods). When is this a useful feature in practice? I'm not just looking for an example of hypothetical functionality that would be easier to implement with multiple dispatch[1]. I'm looking for examples of where it's useful in real software, for any value of real that means it would get written for something other than just an example.

[1] In programming tutorials, are examples contrived more often than not?

Answer 1

• Anytime you use the visitor pattern.
• Anytime you have a binary operator that needs to dispatch on both params (eg. bullet hits rock vs bullet hits spaceship vs spaceship hits rock)
• Anytime you want specialization based on the parameter (eg. C++'s vector<bool> specialization, component systems where the component may or may not exist)

Answer 2

An easy to understand example would be arithmetic, where you might have definitions for

(plus integer integer) (plus float integer) (plus integer float)

and so on.

Used carefully, this can be a wonderful mechanism. Used carelessly..well, you get what you deserve.

Answer 3

You can use them for generalized setters.

Instead of writing

(defmethod set-[specific-property] ((obj obj-type)) ())

for each [specific-property] setter that you want, you can use a single generic function to dispatch:

(defmethod set- ((prop (eql :foreground)) (obj window)) ())

(defmethod set- ((prop (eql :background)) (obj t)) ())

etc...