elm-street

Crossing the road between Haskell and Elm.

What is this library about?

elm-street allows you to automatically generate definitions of Elm data types and compatible JSON encoders and decoders from Haskell types. This helps to avoid writing and maintaining huge chunk of boilerplate code when developing full-stack applications.

Getting started

In order to use elm-street features, you need to perform the following steps:

1. Add elm-street to the dependencies of your Haskell package.

2. Derive the Elm typeclass for relevant data types. You also need to derive JSON instances according to elm-street naming scheme. This can be done like this:

   import Elm (Elm, elmStreetParseJson, elmStreetToJson)
   
   data User = User
       { userName :: Text
       , userAge  :: Int
       } deriving (Generic)
         deriving anyclass (Elm)
   
   instance ToJSON   User where toJSON = elmStreetToJson
   instance FromJSON User where parseJSON = elmStreetParseJson
   

   NOTE: This requires extensions -XDerivingStrategies, -XDeriveGeneric, -XDeriveAnyClass.

   Alternatively you can use -XDerivingVia to remove some boilerplate (available since GHC 8.6.1):

   import Elm (Elm, ElmStreet (..))
   
   data User = User
       { userName :: Text
       , userAge  :: Int
       } deriving (Generic)
         deriving (Elm, ToJSON, FromJSON) via ElmStreet User
   

3. Create list of all types you want to expose to Elm:

   type Types =
      '[ User
       , Status
       ]
   

   NOTE: This requires extension -XDataKinds.

4. Use generateElm function to output definitions to specified directory under specified module prefix.

   main :: IO ()
   main = generateElm @Types $ defaultSettings "frontend/src" ["Core", "Generated"]
   

   NOTE: This requires extension -XTypeApplications.

   When executed, the above program generates the following files:

   • frontend/src/Core/Generated/Types.elm: Core.Generated.Types module with the definitions for the types, as well as show*, read*, un*, and universe* functions as specified in src/Elm/Print/Types.hs
   • frontend/src/Core/Generated/Encoder.elm: Core.Generated.Encoder module with the JSON encoders for the types
   • frontend/src/Core/Generated/Decoder.elm: Core.Generated.Decoder module with the JSON decoders for the types
   • frontend/src/Core/Generated/ElmStreet.elm: Core.Generated.ElmStreet module with bundled helper functions

Elm-side preparations

If you want to use capabilities provided by elm-street in your Elm application, you need to have several Elm packages preinstalled in the project. You can install them with the following commands:

elm install elm/time
elm install elm/json
elm install NoRedInk/elm-json-decode-pipeline
elm install rtfeldman/elm-iso8601-date-strings

Library restrictions

Elm-street is not trying to be as general as possible and support every use-case. The library is opinionated in some decisions and contains several limitations, specifically:

1. Record fields must be prefixed with the type name or its abbreviation.

   data UserStatus = UserStatus
       { userStatusId      :: Id
       , userStatusRemarks :: Text
       }
   
   data HealthReading = HealthReading
       { hrUser   :: User
       , hrDate   :: UTCTime
       , hrWeight :: Double
       }
   

2. Data types with type variables are not supported (see issue #45 for more details). Though, if type variables are phantom, you can still implement Elm instance which will generate valid Elm defintions. Here is how you can create Elm instance for newtypes with phantom type variables:

   newtype Id a = Id { unId :: Text }
   
   instance Elm (Id a) where
       toElmDefinition _ = elmNewtype @Text "Id" "unId"
   

3. Sum types with records are not supported (because it's a bad practice to have records in sum types).

   -- - Not supported
   data Address
       = Post { postCode :: Text }
       | Full { fullStreet :: Text, fullHouse :: Int }
   
   

4. Sum types with more than 8 fields in at least one constructor are not supported.

   -- - Not supported
   data Foo
       = Bar Int Text
       | Baz Int Int Text Text Double Double Bool Bool Char
   

5. Records with fields that reference the type itself are not supported. This limitation is due to the fact that elm-street generates type alias for record data type. So the generated Elm type for the following Haskell data type won't compile in Elm:

   data User = User
       { userName      :: Text
       , userFollowers :: [User]
       }
   

6. Generated JSON encoders and decoders are consistent with default behavior of derived ToJSON/FromJSON instances from the aeson library except you need to strip record field prefixes. Fortunately, this also can be done generically. You can use functions from Elm.Aeson module to derive JSON instances from the aeson package.
7. Only UTCTime Haskell data type is supported and it's translated to Posix type in Elm.
8. Some words in Elm are considered reserved and naming a record field with one of these words (prefixed with the type name, see 1) will result in the generated Elm files to not compile. So, the following words should not be used as field names:
   • if
   • then
   • else
   • case
   • of
   • let
   • in
   • type
   • module
   • where
   • import
   • exposing
   • as
   • port
   • tag (reserved for constructor name due to aeson options)
9. For newtypes FromJSON and ToJSON instances should be derived using newtype strategy. And Elm should be derived using anyclass strategy:

   newtype Newtype = Newtype Int
       deriving newtype (FromJSON, ToJSON)
       deriving anyclass (Elm)
   

Play with frontend example

The frontend directory contains example of minimal Elm project that shows how generated types are used. To play with this project, do:

1. Build and execute the generate-elm binary:

   cabal new-run generate-elm
   

2. Run Haskell backend:

   cabal new-run run-backend
   

3. In separate terminal tab go to the frontend folder:

   cd frontend
   

4. Run the frontend:

   elm-app start
   

Generated examples

Below you can see some examples of how Haskell data types are converted to Elm types with JSON encoders and decoders using the elm-street library.

Records

Haskell

data User = User
    { userName :: Text
    , userAge  :: Int
    } deriving (Generic)
      deriving (Elm, ToJSON, FromJSON) via ElmStreet User

Elm

type alias User =
    { name : String
    , age : Int
    }

encodeUser : User -> Value
encodeUser x = E.object
    [ ("name", E.string x.name)
    , ("age", E.int x.age)
    ]

decodeUser : Decoder User
decodeUser = D.succeed User
    |> required "name" D.string
    |> required "age" D.int

Enums

Haskell

data RequestStatus
    = Approved
    | Rejected
    | Reviewing
    deriving (Generic)
    deriving (Elm, ToJSON, FromJSON) via ElmStreet RequestStatus

Elm

type RequestStatus
    = Approved
    | Rejected
    | Reviewing

showRequestStatus : RequestStatus -> String
showRequestStatus x = case x of
    Approved -> "Approved"
    Rejected -> "Rejected"
    Reviewing -> "Reviewing"

readRequestStatus : String -> Maybe RequestStatus
readRequestStatus x = case x of
    "Approved" -> Just Approved
    "Rejected" -> Just Rejected
    "Reviewing" -> Just Reviewing
    _ -> Nothing

universeRequestStatus : List RequestStatus
universeRequestStatus = [Approved, Rejected, Reviewing]

encodeRequestStatus : RequestStatus -> Value
encodeRequestStatus = E.string << showRequestStatus

decodeRequestStatus : Decoder RequestStatus
decodeRequestStatus = elmStreetDecodeEnum readRequestStatus

Newtypes

Haskell

newtype Age = Age
    { unAge :: Int
    } deriving (Generic)
      deriving newtype (FromJSON, ToJSON)
      deriving anyclass (Elm)

Elm

type alias Age =
    { age : Int
    }

encodeAge : Age -> Value
encodeAge x = E.int x.age

decodeAge : Decoder Age
decodeAge = D.map Age D.int

Newtypes with phantom types

Haskell

newtype Id a = Id
    { unId :: Text
    } deriving (Generic)
      deriving newtype (FromJSON, ToJSON)

instance Elm (Id a) where
    toElmDefinition _ = elmNewtype @Text "Id" "unId"

Elm

type alias Id =
    { unId : String
    }

encodeId : Id -> Value
encodeId x = E.string x.unId

decodeId : Decoder Id
decodeId = D.map Id D.string

Sum types

Haskell

data Guest
    = Regular Text Int
    | Visitor Text
    | Blocked
    deriving (Generic)
    deriving (Elm, ToJSON, FromJSON) via ElmStreet Guest

Elm

type Guest
    = Regular String Int
    | Visitor String
    | Blocked

encodeGuest : Guest -> Value
encodeGuest x = E.object <| case x of
    Regular x1 x2 -> [("tag", E.string "Regular"), ("contents", E.list identity [E.string x1, E.int x2])]
    Visitor x1 -> [("tag", E.string "Visitor"), ("contents", E.string x1)]
    Blocked  -> [("tag", E.string "Blocked"), ("contents", E.list identity [])]

decodeGuest : Decoder Guest
decodeGuest =
    let decide : String -> Decoder Guest
        decide x = case x of
            "Regular" -> D.field "contents" <| D.map2 Regular (D.index 0 D.string) (D.index 1 D.int)
            "Visitor" -> D.field "contents" <| D.map Visitor D.string
            "Blocked" -> D.succeed Blocked
            c -> D.fail <| "Guest doesn't have such constructor: " ++ c
    in D.andThen decide (D.field "tag" D.string)