Decoding Large JSON Objects: A Summary

Published August 22, 2016 · 6 Minute Read · ∞ Permalink


There have been several recent questions on the elm-discuss mailing list about decoding large JSON objects. The problem: Elm’s decoders provide for decoding objects with up to 8 fields, but what happens when you need more? The solution here is, unfortunately, not super obvious.

The Problem

Let’s flesh out the problem a little bit more. Say we’re working with Users from the GitHub API. We have a bunch of fields here:

{
  "login": "octocat",
  "id": 1,
  "avatar_url": "https://github.com/images/error/octocat_happy.gif",
  "gravatar_id": "",
  "url": "https://api.github.com/users/octocat",
  "html_url": "https://github.com/octocat",
  "followers_url": "https://api.github.com/users/octocat/followers",
  "following_url": "https://api.github.com/users/octocat/following{/other_user}",
  "gists_url": "https://api.github.com/users/octocat/gists{/gist_id}",
  "starred_url": "https://api.github.com/users/octocat/starred{/owner}{/repo}",
  "subscriptions_url": "https://api.github.com/users/octocat/subscriptions",
  "organizations_url": "https://api.github.com/users/octocat/orgs",
  "repos_url": "https://api.github.com/users/octocat/repos",
  "events_url": "https://api.github.com/users/octocat/events{/privacy}",
  "received_events_url": "https://api.github.com/users/octocat/received_events",
  "type": "User",
  "site_admin": false,
  "name": "monalisa octocat",
  "company": "GitHub",
  "blog": "https://github.com/blog",
  "location": "San Francisco",
  "email": "octocat@github.com",
  "hireable": false,
  "bio": "There once was...",
  "public_repos": 2,
  "public_gists": 1,
  "followers": 20,
  "following": 0,
  "created_at": "2008-01-14T04:33:35Z",
  "updated_at": "2008-01-14T04:33:35Z"
}

If this object had fewer fields, we could use like this:

import Json.Decode as Decode

type alias User =
    { login : String
    , id : Int,
    , name : String
    }

user : Decode.Decoder User
user =
    Decoder.object3 User
        ("login" := Decode.string)
        ("id" := Decode.int)
        ("name" := Decode.string)

But since this record has 30 fields, what do we do? We only have object* up to object8. Even if we did have object30 would we want to use that := syntax to keep everything in order?

Three Solutions

There are a couple of solutions to this. We’ll go over them in order of how easy the solutions are. These examples are going to use only a couple fields of the object to aid readability. This leads to our first suggestion:

Use Fewer Fields

The object above is really comprehensive. I’ve implemented several apps against GitHub’s API and have never had to use more than a handful of the fields returned by any of their responses. Elm’s decoders don’t need to decode every field in the object. They don’t even need to decode them in the order from the JSON. (See in the example above how “name” is the third field in our record, but the 18th in the JSON.) This will also cut your API surface area and make your code easier to maintain.

This tractor worked too many fields, and just look what happened.

This tractor worked too many fields, and just look what happened. Photo by Xavi Moll

So try hard to use only what you need. But if you find yourself stuck needing an object with more than 8 fields:

Json.Decode.Extra from elm-community/json-extra

We have to make a pitstop here to look at incremental decoding. When you define a record, you also get a record constructor for free! It will have the the same thing as the type, so in our User example above: User : String -> Int -> String -> User.

Remember our discussion on currying? We know from there that if we call User with just a String, we’ll have Int -> String -> User left over. We can use that to our advantage to make a pipeline using apply from Json.Decode.Extra! In this example, we’ll use (|:), which is the infix version of apply.

import Json.Decode as Decode exposing ((:=))
import Json.Decode.Extra exposing ((|:))

user : Decode.Decoder User
user =
    Decode.succeed User
        |: ("login" := Decode.string)
        |: ("id" := Decode.int)
        |: ("name" := Decode.string)

If you want to dive deeper into how this works, the documentation for (|:) has a great explanation. But we’re not quite done here yet.

We’re now importing two custom infix operators. But that’s two more than I’d like. They’re not self-documenting, so they make your app harder to maintain. We also have a mysterious call to Decode.succeed in the top of the function. succeed is just turning our record constructor into a JSON decoder, but that’s not super obvious. All this means one more step!

Json.Decode.Pipeline

NoRedInk’s elm-style-guide recommends using their Json.Decode.Pipeline. It’s a good recommendation, since it will clean up our code that last little bit. So how do we use it?

When we think of transforming values, |> comes to mind. It’s a built-in operator function, so there are no extra imports to worry about. If you’re not familiar with |>, you can think of it like a pipe that values can flow through. Let’s use this with Json.Decode.Pipeline to improve our decoder:

import Json.Decode as Decode
import Json.Decode.Pipeline exposing (required, decode)

user : Decode.Decoder User
user =
    decode User
        |> required "login" Decode.string
        |> required "id" Decode.int
        |> required "name" Decode.string

This is much clearer, at least to my eye. I can see at first glance that:

Using this approach lets you have optional fields as well.

Summarized!

There you have it! Three ways to decode JSON objects with more than 8 fields. To summarize our summary, you can:

  1. Use only the fields that you need. This means your problem can just go away with a mindset change.
  2. Use apply or (|:) from Json.Decode.Extra for a quick fix.
  3. Use Json.Decode.Pipeline to make a readable decoding pipeline to make future maintenance easier.

Want More?

Want to get hot, fresh Elm help and tips in your inbox? Slap your email in the box below and I'll send you new articles!

    We won't send you spam. Unsubscribe at any time.


    Edit August 23, 2016: Corrected the type signatures of the User constructor. It didn’t have User as the return value. Thanks to Martin Janiczek for the catch!