A lot of learning projects involve writing games: people have written clones of Tetris, Asteroids, Space Invaders, and even first person shooters (Frag) in Haskell. As I'm far less clever than these people, I thought I'd start with something a bit simpler: Dope Wars.

Dope Wars is basically a trading game. In 30 turns, you move from one location to another, buying and selling, er, drugs on the streets of New York. It's a fairly simple concept, but one which includes elements like:

• Input and Output
• Game state
• Random numbers

all of which seem like a good way to learn Monads and the other building blocks that you need to actually do anything useful in a functional programming language like Haskell.

So I had the idea, wrote up a couple of datatypes and some functions, and then forgot about it. Then, when Greg McCarroll mentioned that he'd accept talks about other languages for the London Perl Workshop on December 1st, I thought it would be a great opportunity to push myself to do it by proposing a lightning talk.

Only problem: I now have to actually write the game in order to talk about it (*). So... here goes. In this post, I'm going to show a first draft for the game prompt.

State

OK, people often find it most convenient to do State using “Monads”. I think I'm going to leave that for now, and just thread state explicitly. Mainly because I haven't yet got around to learning how to use the State Monad. Hopefully this will eventually become annoying enough that it will give me impetus to learn the monadic version.

Anyway, the idea here is that we'd have some function playTurn that will look a bit like this:

 > playTurn :: GameState -> IO ()
 > playTurn gs = let gs' = doSomething gs
 >               in  playTurn gs'

(There is an interesting post on haskell-cafe about a more sophisticated monadic representation of the prompt).

So we just need to work out how to represent this GameState object. To start off with, we'll want to store information like

• Which turn it is
• What our score is (how much money we have)
• Where we are on the game map

We could create a normal tuple:

 > data GameState = GameState Integer Integer Location
 >                         -- turn    score   location

and then pattern match on this, but it's going to get horrible if we add any fields later on! In Perl I'd just use a hash, but remember that Haskell Data.Map objects map from one type to another, and we might well have values of various types.

When I asked on #haskell, dons and firefly told me about the record syntax:

 > data GameState = GameState {
 >        turn     :: Integer,
 >        score    :: Integer,
 >        location :: Location
 >    } deriving Show
 > 
 > -- for now:
 > type Location = Integer

Defining the original state is easy:

 > startState = GameState {
 >    turn     = 1,
 >    score    = 0,
 >    location = 0
 >  }

And to “modify” it (or rather to clone it, overriding certain fields) there is a convenient syntax that just lets us declare those fields which have changed:

 > movedToBronx = gs { location = bronx }

Setting a field to a value is easy, but we might want to define some mutators to change the field relative to its current value:

 > nextTurn :: GameState -> GameState
 > nextTurn gs = gs { turn = succ $ turn gs }
 >
 > modScore :: Integer -> GameState -> GameState
 > modScore d gs = gs { score = score gs + d }

The record syntax will work even when we inevitably add new fields later. Yay!

Prompt

Now, the game cycle is a bit more complicated than the version I suggested above, as it will allow IO actions in it. Something perhaps like this:

 > playTurn gs = do showStatus gs
 >                  putStr prompt
 >                  s <- getLine
 >                  let f = parseLine s
 >                  let gs' = f gs
 >                  if isEnd gs' 
 >                     then endGame  gs'
 >                     else playTurn gs'

For now we'll just stub some of the declarations we need. showStatus can just show the GameState record (which is why we derived the Show class).

 > showStatus gs = putStrLn $ show gs

We may as well set the prompt to the dollar sign, appropriately:

 > prompt   = "$ "

Though we'll need to parse the line read from standard input to one of the commands like “Buy 4 X” or “Goto location 3“, right now, we'll just stub in a function that increments the score and the turn counter:

 > parseLine s =   nextTurn 
 >               . modScore 10 

We need to know if we're at the end of the game, and take action appropriately.

 > isEnd gs = turn gs > maxTurns
 > 
 > maxTurns = 3
 > 
 > endGame gs = do putStrLn   "Game over!"
 >                 putStrLn $ "Your score was " ++ (show $ score gs)
 >                 return ()

And here's a transcript

 *Main> playTurn startState
 GameState {turn = 1, score = 10, location = 0}
 > buy  2 foo
 GameState {turn = 2, score = 20, location = 0}
 > sell 4 bar
 GameState {turn = 3, score = 30, location = 0}
 > goto quuxtown
 Game over!
 Your score was 40

(*) to be fair, it's only a 5 minute “Lightning Talk”, so I could probably get away without even writing the whole game, but I'll feel better if I actually know what I'm talking about...