Haskell - Encoding and Decoding JSON with Aeson
piq9117
•May 1, 2024
(Source/Credits: https://dev.to/piq9117/haskell-encoding-and-decoding-json-with-aeson-5d7n)
Techniques I've learned from tutorials that I use daily
title: Haskell - Encoding and Decoding JSON with Aeson published: true description: Techniques I've learned from tutorials that I use daily tags: haskell, functional, json
There are a lot of blog posts and tutorial about encoding/decoding JSON with aeson. Even the library is pretty good at teaching how to do this. The tutorial I always go back to is Artyom Kazak's tutorial. They talk about lots of different techniques on how to decode and encode json on different cases.
Let's start off with the basics by deriving instances of FromJSON and ToJSON manually.
```Haskell data Book = Book { bookTitle :: Text , bookISBN :: Text , bookPublisher :: Text , bookLanguage :: Text } deriving Show
instance FromJSON Book where parseJSON = withObject "Book" $ \b -> Book <$> b .: "title" <> b .: "ISBN" <> b .: "publisher" <*> b .: "language"
instance ToJSON Book where toJSON Book {..} = object [ "title" .= bookTitle , "ISBN" .= bookISBN , "publisher" .= bookPublisher , "language" .= bookLanguage ] ```
With FromJSON and ToJSON instances we can now consume json of this shape:
json
{ "title": ".."
, "ISBN": ".."
, "publisher": ".."
, "language": ".."
}
As you can see when the type has more fields that also means that we have to type out all those fields. Your fingers are going to fall off by the time you are done with your app.
One solution to minimize the boilerplate is by using Generics
```Haskell {-# LANGUAGE DeriveGeneric #-}
import GHC.Generics
data Book = Book { bookTitle :: Text , bookISBN :: Text , bookPublisher :: Text , bookLanguage :: Text } deriving (Generic, Show)
instance FromJSON Book
instance ToJSON Book
If we use `{-# LANGUAGE DeriveAnyClass #-}` pragma we can do this.Haskell
{-# LANGUAGE DeriveAnyClass #-}
{-# LANGUAGE DeriveGeneric #-}
data Book = Book { bookTitle :: Text , bookISBN :: Text , bookPublisher :: Text , bookLanguage :: Text } deriving (Generic, Show, FromJSON, ToJSON) ```
DeriveAnyClass and GenerlizedNewTypeDeriving
If we have both of these language extensions enabled, ghc will complain about derivation being ambigious. To get around this use {-# LANGUAGE DerivingStrategies #-} language extension.
```Haskell
{-# LANGUAGE DerivingStrategies #-}
{-# LANGUAGE DeriveGeneric #-}
data Book = Book { bookTitle :: Text , bookISBN :: Text , bookPublisher :: Text , bookLanguage :: Text } deriving (Generic, Show) deriving anyclass (FromJSON, ToJSON) ```
If we don't need to do anything with the field name this will suffice. However, our field name is title not bookTitle so we have to do a little modification to the field names by doing the following
```Haskell
import Text.Casing (camel)
import Data.Aeson
instance FromJSON Book where parseJSON = genericParseJSON defaultOptions { fieldLabelModifier = camel . drop 4 }
instance ToJSON Book where
toJSON = genericToJSON
defaultOptions { fieldLabelModifier = camel . drop 4 }
``
Here's a reference to [defaultOptions][aeson-default-options]. In the code above we're doing a record update. That means it's gonnadrop` 4 characters from the beginning and then camel case it.
Nullable Fields
When it comes to nullable fields, Generics} will automatically use this operator (.:?) on fields that are Maybes, which will use Nothing if
the field is null or missing.
Optional Fields
For optional fields we have to go back to manually deriving ToJSON
and FromJSON manually.
```Haskell
data Book = Book
{ bookTitle :: Text
, bookISBN :: Text
, bookPublisher :: Text
, bookLanguage :: Text
, bookPrice :: Maybe (Fixed E2)
} deriving Show
instance FromJSON Book where parseJSON = withObject "Book" $ \b -> Book <$> b .: "title" <> b .: "ISBN" <> b .: "publisher" <> b .: "language" <> optional (b .: "price") ```
Sum Types
```Haskell {-# LANGUAGE RecordWildCards #-}
data BookFormat = Ebook { price :: Fixed E2 } | PhysicalBook { price :: Fixed E2, coverType :: Text } deriving Show
instance FromJSON BookFormat where parseJSON = withObject "BookFormat" $ \b -> asum [ Ebook <$> b .: "price" , PhysicalBook <$> b .: "price" <*> b .: "coverType" ]
instance ToJSON BookFormat where toJSON = \case Ebook {..} -> object [ "price" .= price ] PhysicalBook {..} -> object [ "price" .= price, "coverType" .= coverType]
or if we we can decide based on the `format`Haskell
instance FromJSON BookFormat where
parseJSON = withObject "BookFormat" $ \b -> do
format <- b .: "format"
case (format :: Text) of
"ebook" -> Ebook <$> b .: "price"
"physicalBook" ->
PhysicalBook <$> b .: "price"
<*> b .: "coverType"
The same with product types we can also use `Generics` and some
language extensions to derive `FromJSON` and `ToJSON` instanceHaskell
{-# LANGUAGE DeriveGeneric #-}
{-# LANGUAGE DeriveAnyClass #-}
import GHC.Generics
data BookFormat = Ebook { price :: Fixed E2 } | PhysicalBook { price :: Fixed E2, coverType :: Text } deriving (Show, Generic, ToJSON, FromJSON) ``` These are the usual day to day techniques of encoding/decoding json data that I use.
Comments section