Overview

We've seen the power of specific and expressive queries that let us retrieve exactly the data we're looking for, all at once. But querying in GraphQL is just one part of the equation.

When we want to actually change, insert, or delete data, we need to reach for a new tool: GraphQL mutations.

In this lesson, we will:

• Explore mutation syntax and write an operation to add tracks to a playlist
• Learn about GraphQL input types
• Learn about mutation response best practices

Mutations in Spotify

On to the next feature in our MusicMatcher project: adding tracks to an existing playlist.

Let's take a look at the corresponding REST API method that enables this feature: POST /playlists/{playlist_id}/tracks

From the documentation, we need the following parameters:

• playlist_id - The ID of the playlist, as a string
• position - An integer, zero-indexed, where we want to insert the track(s)
• uris - A comma-separated string of uri values corresponding to the tracks we want to add

The method then returns an object with a snapshot_id property that represents the state of the playlist at that point in time.

All right, now how do we enable this functionality in GraphQL?

Designing mutations

Much like the Query type, the Mutation type serves as an entry point to our schema. It follows the same syntax as the schema definition language, or SDL, that we've been using so far.

We declare the Mutation type using the type keyword, then the name Mutation. Inside the curly braces, we have our entry points, the fields we'll be using to mutate our data.

Let's open up schema.graphql and add a new Mutation type.

For the fields of the Mutation, we recommend starting with a verb that describes the specific action of our update operation (such as add, delete, or create), followed by whatever data the mutation acts on.

We'll explore how we can add items to a playlist, so we'll call this mutation addItemsToPlaylist.

For the return type of the addItemsToPlaylist mutation, we could return the Playlist type; it's the object type we want the mutation to act upon. However, we recommend following a consistent Response type for mutation responses. Let's see what this looks like in a new type.

The Mutation response type

Return types for Mutation fields usually end with the word Payload or Response.

Following convention, we'll combine the name of our mutation (addItemsToPlaylist) with Payload. (Don't forget to capitalize!)

We should return the object type that we're mutating (Playlist, in our case), so that clients have access to the updated object.

Notice that playlist can be null, because our mutation might fail.

To account for any partial errors that might occur and return helpful information to the client, there are a few additional fields we can include in a response type.

• code: an Int that refers to the status of the response, similar to an HTTP status code.

• success: a Boolean flag that indicates whether all the updates the mutation was responsible for succeeded.

• message: a String to display information about the result of the mutation on the client side. This is particularly useful if the mutation was only partially successful and a generic error message can't tell the whole story.

Let's also add comments for each of these fields so that it makes our GraphQL API documentation more useful.

Lastly, we can set the return type of our mutation to this new AddItemsToPlaylistPayload type, and make it non-nullable. Here's what the addItemsToPlaylist mutation should look like now:

The Mutation input

To make any changes to a particular playlist, our mutation needs to receive some input.

Let's think about the kind of input this addItemsToPlaylist mutation would expect. We're potentially adding many new tracks to a singular playlist. This means that whoever is sending this query should be able to provide the specific playlist, along with the item(s) to be added. Furthermore, we could let them pass additional customization, specifying where in the playlist the items should be inserted.

We've used a GraphQL argument before in the Query.playlist field: we passed in a single argument called id.

But addItemsToPlaylist takes more than one argument. One way we could tackle this is to add each argument, one-by-one, to our addItemsToPlaylist mutation. But this approach can become unwieldy and hard to understand. Instead, it's a good practice to use GraphQL input types as arguments for a field.

Exploring the input type

The input type in a GraphQL schema is a special object type that groups a set of arguments together, and can then be used as an argument to another field. Using input types helps us group and understand arguments, especially for mutations.

To define an input type, use the input keyword followed by the name and curly braces ({}). Inside the curly braces, we list the fields and types as usual. Note that fields of an input type can be only a scalar, an enum, or another input type.

Next, we'll add properties. Remember, we need the ID of the playlist and a list of URIs, at the very minimum. We could also specify the position in the playlist these items get added to, but it's not required for the REST API. By default, tracks will be appended to the end of the playlist, so we're safe to omit it from our GraphQL schema. Remember, your GraphQL API does not need to match your REST API exactly!

Using the input

To use an input type in the schema, we can set it as the type of a field argument. Let's update the addItemsToPlaylist mutation to use the AddItemsToPlaylistInput type.

Notice that the AddItemsToPlaylistInput is non-nullable. To run this mutation, we actually need to require some input!

Building the SpotifyAPI method

As we've done for other requests, we'll add a method to the SpotifyAPI class to manage this call to update data.

Back in datasources/spotify-api.ts, let's add a new method called addItemsToPlaylist. We'll expect this method to receive an input parameter, which is an object containing playlistId (a string type) and uris (an array of string types).

This time, because the endpoint uses the POST method, we'll use the this.post helper method from RESTDataSource instead of this.get.

Next, we'll pass our endpoint to this method. We'll replace the {playlist_id} query param with ${playlistId} so we can interpolate it from this method's arguments.

We need to send a parameter with our call to this endpoint: namely, the uris that we wish to add to the specified playlist. We can pass a second argument to the this.post method, which is an object with a params key. Inside of the params object, we'll add our uris. Referring to our REST API, we know that uris needs to be a single string of comma-separated uri values, so we'll use the join method to transform our array of strings into a single string.

To give this method its correct return type, let's consider the two possible outcomes from running the mutation.

Glancing back at our REST API documentation, we can see that when a playlist is successfully updated, we get back an object with a "snapshot_id" key that matches the ID of the playlist we updated.

And if something goes wrong—like if we try to update a playlist that doesn't exist—we should instead get back an object with an "error" key.

Back in models.ts, let's create a type specifically for these possible mutation responses. It will have two optional properties, both string types: snapshot_id and error.

We'll first import this type into spotify-api.ts.

Then we can give our addItemsToPlaylist method the appropriate return type of Promise<SnapshotOrError>.

That's it for the call to the endpoint—let's add a resolver function for our Mutation.addItemsToPlaylist field, and make sure it's calling this method in SpotifyAPI!

Connecting the dots in the resolvers

Jump back into the resolvers.ts file. We'll add a new entry to our resolvers object called Mutation.

Let's add our addItemsToPlaylist resolver.

We won't need the parent or info parameters here, so we'll replace parent with _ and remove info altogether.

We know from our Mutation.addItemsToPlaylist schema field that this resolver will receive an argument called input. We'll destructure args for this property. And while we're here, we'll also destructure contextValue for the dataSources property.

Now we'll call the addItemsToPlaylist method from the spotifyAPI data source, passing it our input.

We're not returning anything from this method just yet, so we might see some errors. In our schema we specified that the Mutation.addItemsToPlaylist field should return a AddItemsToPlaylistPayload type.

This means that the object we return from our resolver needs to match this shape. Let's start by setting up the object that we'll return in the event that our mutation operation succeeds.

Checking the response

Now, let's take a look at the response from our API call. We'll need to make our entire resolver async in order to await the results of calling dataSources.spotifyAPI.addItemsToPlaylist().

In Sandbox, let's put together a mutation operation.

And add the following to the Variables panel.

When we run the operation, we'll see that this mutation actually works as expected! We can clearly see the values we set for code, success, and message in our happy path.

And in the terminal of our running server, we'll see the response from the API logged out.

The snapshot_id, we can see, is the playlistId of the playlist we added tracks to!

Let's build out the sad path—when our mutation doesn't go as expected.

Handling the sad path

Returning to our Mutation.addItemsToPlaylist, we'll account for the error state.

We'll start by wrapping everything in a try/catch block.

Next, we'll contain the object we return in an if block. We'll check for the existence of the snapshot_id in the response from the REST API. If it doesn't exist, we'll throw an Error.

Inside of the catch block, we'll set some different properties.

Here's how your entire resolver function should look.

Let's try out a mutation that we know won't work. Back in Sandbox, replace the values in the Variables panel with the input below.

When we run the operation, we'll see the values we expect: the mutation failed, and our error message comes through loud and clear. Something went wrong: 404 Not Found!

Great! We have both the happy path and the sad path working, but we'll notice that there's one piece missing: we're still not returning any useful data for the AddItemsToPlaylistPayload.playlist field.

This means if we include subfields from the Playlist type (as shown in the mutation operation below), we won't get actual data—we don't have a Playlist object to pull subfields from, much less its tracks! We clearly can't yet take advantage of GraphQL's ability to traverse from object type to object type.

So how can we use the value we do have—the ID for the playlist we modified—to build out the remainder of our AddItemsToPlaylistPayload response?

We'll tackle this in the next lesson!

Practice

Key takeaways

• Mutations are write operations used to modify data.
• Naming mutations usually starts with a verb that describes the action, such as "add," "delete," or "create."
• It's a common convention to create a consistent response type for mutation responses.
• Mutations in GraphQL often require multiple arguments to perform actions. To group arguments together, we use a GraphQL input type for clarity and maintainability.

Up next

In our final lesson, let's complete our mutation response—and return Playlist data!