🖋IO is Your Command Pattern

In this post, I’ll try to give some intuition behind IO in Haskell for people who know what command pattern is, without explaining what M-word or do-notation is.

First, let’s revisit command pattern. According to Wikipedia:

In object-oriented programming, the command pattern is a behavioral design pattern in which object is used to encapsulate all information needed to perform an action or trigger an event at a later time. This information includes the method name, the object that owns the method and values for the method parameters.

Let’s implement the pattern in Java:

interface Command {
    void execute();
}

Command is an interface for implementing different commands. It has method execute that triggers command execution. Pretty simple.

But let’s extended the interface and allow return values.

interface Command<T> {
    T execute();
}

This allows to define commands that return values, such as GetLine (remember getLine :: IO String in Haskell).

class GetLine implements Command<String> {
    public String execute() {
        Scanner scanner = new Scanner(System.in);
        return scanner.nextLine();
    }
}

Now you should see how IO String in Haskell is similar to Command<String>.

Let’s implement PutStrLn now. The Haskell analog is putStrLn :: String -> IO (), that is a function that takes a String and returns a command that prints that string to the console. The proper Java interface for such a thing is Function<String, Command<Void>> (from java.util.function).

class PutStrLn implements Function<String, Command<Void>> {
    public Command<Void> apply(String str) {
        return () -> { System.out.print(str); return null; };
    }
}

We can use it like this:

new PutStrLn().apply("Hello, world!").execute();

(you see how execution stage is separated from command creation?)

Now the fun begins. Let’s add a couple of methods for combining our =Command=s:

public interface Command<T> {
    T execute();

    static <T> Command<T> pure(T value) {
        return () -> value;
    }

    default <U> Command<U> then(Command<U> next) {
        return () -> {
            this.execute();
            return next.execute();
        };
    }

    default <U> Command<U> bind(Function<T, Command<U>> that) {
        return () -> that.apply(this.execute()).execute();
    }
}

pure is a simple method that creates new Command which, when executed, returns given value. This method is the same as Haskell’s return :: a -> IO a function.

then takes additional command and returns a new command which, when executed, executes this command and then executes the second command and return its result. This method is Haskell’s (>>) :: IO a -> IO b -> IO b.

bind takes a function from T to a new command and returns a command which, when executed, executes the this command and passes its result to the function and executes the final command. This function is Haskell’s (>>) :: IO a -> (a -> IO b) -> IO b=.

Let’s write a “real-world” application with that.

class Main {
    static Function<String, Command<String>> ask = (String prompt) ->
        new PutStrLn().apply(prompt).then(
        new GetLine());

    static Command<Integer> cmd =
        ask.apply("What's your name? ").bind((String name) ->
        ask.apply("What's your birth year? ").bind((String year) ->
        new PutStrLn().apply("Hello, " + name + ". You should be " +
            (2015 - Integer.parseInt(year)) + " years old now.").then(
        Command.pure(2015 - Integer.parseInt(year)))));

    public static void main(String[] args) {
        Integer age = cmd.execute();
    }
}

ask is a helper function (command) that asks the user for some info and returns user’s answer.

The same program in Haskell:

ask :: String -> IO String
ask prompt = putStrLn prompt >> getLine

main :: IO Integer
main =
    ask "What's your name?" >>= \name ->
    ask "What's your birth year?" >>= \year ->
    putStrLn ("Hello, " ++ name ++ ". You should be " ++
        show (2015 - read year) ++ " years old now.") >>
    return (2015 - read year)

But Haskell can do better!

ask :: String -> IO String
ask prompt = do
    putStrLn prompt
    getLine

main :: IO Integer
main = do
    name <- ask "What's your name?"
    year <- ask "What's your birth year?"
    putStrLn ("Hello, " ++ name ++ ". You should be " ++
        show (2015 - read year) ++ " years old now.")
    return (2015 - read year)

Haskell has a special do-notation. In fact, it’s just syntactic sugar for >>, >>= and lambdas. So the last program is the same as previous one.

Now you should notice that the whole Haskell program is one big command and there is no execute in Haskell---all programs should be composed of primitive commands provided by the base library.

That’s how Haskell keeps things pure. Think! Your whole program is just a command---a recipe that describes what program should do. putStrLn is a pure(!) function that returns command (which in turn knows what to print on screen). getLine is a constant(!) value---a command; you’re not performing I/O when you type name <- getLine.

Conclusion

• IO is a free command pattern in Haskell; you can pass IO values around like any other values.

• All functions in Haskell are pure;

• do is not a hack for doing I/O; it’s just syntactic sugar;

P.S. Note that >>, >>=, return and do-notation are not unique to IO and I lied about their types. They all work for any Monad, not just IO.