Raw resources are integral part for most 4x games. Here’s one way of modeling them in Haskell. I wanted a system that is easy to use, doesn’t require too much typing and is type safe.

RawResource is basic building block:

newtype RawResource a = RawResource { unRawResource :: Int }
    deriving (Show, Read, Eq)

It can be parametrised with anything, but I’m using three different types:

data Biological = Biological
data Mechanical = Mechanical
data Chemical = Chemical

Example of defining harvest being 100 units of biological raw resources:

  harvest :: RawResource Biological
  harvest = RawResource 100

Raw resources are often manipulated (added and subtracted mostly). Defining Num instance allows us to use them as numbers:

instance Num (RawResource t) where
    (+) (RawResource a) (RawResource b) = RawResource $ a + b
    (-) (RawResource a) (RawResource b) = RawResource $ a - b
    (*) (RawResource a) (RawResource b) = RawResource $ a * b
    abs (RawResource a) = RawResource $ abs a
    signum (RawResource a) = RawResource $ signum a
    fromInteger a = RawResource $ fromInteger a

For example, adding harvest to stock pile:

  stock :: RawResource Biological
  stock = RawResource 1000

  harvest :: RawResource Biological
  harvest = RawResource 100

  newStock = stock + harvest

Comparing size of two resource piles is common operation. Ord instance has methods we need for comparing:

instance Ord (RawResource t) where
    (<=) (RawResource a) (RawResource b) = a <= b

One function is enough, as rest is defined in terms of it. Sometimes (usually for reasons of optimization), one might want to define other functions too.

Another way to add bunch of resources of same type together is defining Monoid instance:

instance Semigroup (RawResource t) where
    (<>) a b = a + b

instance Monoid (RawResource t) where
    mempty = RawResource 0

For example, combining harvests of many fields can be achieved as:

  harvests :: [RawResource Biological]
  harvests = [RawResource 20, RawResource 50, RawResource 25]

  total :: RawResource Biological
  total = mappend harvests

All these functions keep track of type of resources being manipulated. Compiler will emit an error if two different types of resources are being mixed together.

Raw resources are often grouped together for specific purpose. This again uses phantom types to keep track the intended usage:

data RawResources a = RawResources
    { ccdMechanicalCost :: RawResource Mechanical
    , ccdBiologicalCost :: RawResource Biological
    , ccdChemicalCost :: RawResource Chemical
    } deriving (Show, Read, Eq)

data ResourceCost = ResourceCost
data ConstructionSpeed = ConstructionSpeed
data ConstructionLeft = ConstructionLeft
data ConstructionDone = ConstructionDone
data ResourcesAvailable = ResourcesAvailable

And in order to be able to combine piles of RawResources, we’ll define Semigroup and Monoid instances. Notice how both instances make use of Semigroup and Monoid instances of RawResource:

instance Semigroup (RawResources t) where
    (<>) a b = RawResources
        { ccdMechanicalCost = ccdMechanicalCost a <> ccdMechanicalCost b
        , ccdBiologicalCost = ccdBiologicalCost a <> ccdBiologicalCost b
        , ccdChemicalCost = ccdChemicalCost a <> ccdChemicalCost b
        }

instance Monoid (RawResources t) where
    mempty = RawResources
        { ccdMechanicalCost = mempty
        , ccdBiologicalCost = mempty
        , ccdChemicalCost = mempty
        }

For those interested seeing some code, source is available at https://github.com/Tuula/deep-sky/ (https://github.com/Tuula/deep-sky/tree/baa0807dd36b61fd02174b17c10013862af4ec18 is situation before lots of Elm related changes that I mentioned in passing in the previous episode)