Re:Ask HN: What is a monad?
by tel on Sunday December 12, 2010 @05:34PM
(#1998307)
There are already a number of excellent answers, but I'll try my hand. You don't know something until you can explain it 3 different ways.
Haskell excels as a functional language because it has excellent piping, Monads are just another form of piping with particularly great syntactic and psychological properties. They act as wrappers around a direct type and setting the return type of a function as a monadic type (such as IO () or [Int] or Maybe Bool) means that the primary type is () or Int or Bool, but there's something special being passed along as well.
In practice this saves typing since you don't explicitly pass the special monadically-hidden information. It also psychologically lets you think of simple types being wrapped in contexts which have certain properties. Some contexts can be unwrapped, some cannot. Some contexts can transform into other contexts directly, some cannot. Learning the behavior of each monad independently teaches you about these behaviors and then becomes important in advanced Haskell programming.
When learning, avoid them. Learn how to use IO by rote and ask in IRC about the confusing type errors that will arise. They usually translate as "you tried to combine two pipes which don't fit, so that monadic special sauce was lost". Part of learning Haskell is learning all of the joints and adaptors which help you to combine types effectively (including monads).
When you have a firm grasp of using compositions and maps and filters, start with the Maybe monad. Implement it and use all the general monad adaptors on them. Then learn the basic State monad because it makes the special sauce very explicit. Learn IO last because it's really a special case and should never be introduced as a monad.
Haskell excels as a functional language because it has excellent piping, Monads are just another form of piping with particularly great syntactic and psychological properties. They act as wrappers around a direct type and setting the return type of a function as a monadic type (such as IO () or [Int] or Maybe Bool) means that the primary type is () or Int or Bool, but there's something special being passed along as well.
In practice this saves typing since you don't explicitly pass the special monadically-hidden information. It also psychologically lets you think of simple types being wrapped in contexts which have certain properties. Some contexts can be unwrapped, some cannot. Some contexts can transform into other contexts directly, some cannot. Learning the behavior of each monad independently teaches you about these behaviors and then becomes important in advanced Haskell programming.
When learning, avoid them. Learn how to use IO by rote and ask in IRC about the confusing type errors that will arise. They usually translate as "you tried to combine two pipes which don't fit, so that monadic special sauce was lost". Part of learning Haskell is learning all of the joints and adaptors which help you to combine types effectively (including monads).
When you have a firm grasp of using compositions and maps and filters, start with the Maybe monad. Implement it and use all the general monad adaptors on them. Then learn the basic State monad because it makes the special sauce very explicit. Learn IO last because it's really a special case and should never be introduced as a monad.