70 %
Chris Biscardi

Sourcing Data from Multiple Locations in Gatsby

How can we construct a GraphQL interface such that we can source data from multiple locations? Check out the livestream here:

The old way (that still works) but is really low level and complicated is found here:

So the first thing we need is an interface, conceptually and literally. For this post we’ll keep it small and use an interface with a title and a body. Note that there are two additional features that are required for what we’re doing with Gatsby: @nodeInterface and the id field, which is required for satisfying the Gatsby Node interface.

GraphQL
interface BlogPost @nodeInterface {
id: ID!
title: String
body: String!
}`

This interface allows us to construct types that satisfy the interface which effectively means we are backing BlogPost with custom types. Being an interface means we can have multiple custom types that back the interface at the same time. So when we query allBlogPost, were really querying the collection of custom types that satisfy the interface.

Querying for Yaml

We'll start with something that uses the GraphQL SDL, using a Yaml node to satisfy the interface. Since gatsby-transformer-yaml names it's type after the directory name, we end up with a Yaml node that is of type ContentYaml. We can use the createTypes schema customization API to change the types that the fields resolve to by using the same name. Importantly, we specify that this type implements Node and BlogPost, as this will allow us to query the content through the interface. Here's the content of our yaml file:

yaml
subject: Some title
body: Some body content for the blog post

and the type we use to override it. Note that we're proxying the subject field to the title field to comply with the BlogPost interface.

JS
createTypes(`type ContentYaml implements Node & BlogPost {
id: ID!
title: String @proxy(from: "subject")
body: String!
}`);

And now we can query the Yaml content through allBlogPost.

Querying for Mdx

Querying for Mdx takes a bit more work. We have to solve an additional problem because some of the content (like body or excerpt) is only returned by resolvers and doesn't exist on the node itself. To achieve this, we'll take a subset of Mdx nodes and turn them into a new node time: MdxBlogPost. This new node type will be the type that implements our BlogPost interface. Here's the type definition for our new node. We use buildObjectType as that lets us pass resolvers through to the parent type (Mdx in this case).

JS
createTypes(
schema.buildObjectType({
name: `MdxBlogPost`,
fields: {
id: { type: `ID!` },
title: {
type: "String!"
},
body: {
type: "String!",
resolve(source, args, context, info) {
const type = info.schema.getType(`Mdx`);
const mdxNode = context.nodeModel.getNodeById({
id: source.parent
});
const resolver = type.getFields()["body"].resolve;
return resolver(mdxNode, {}, context, {
fieldName: "body"
});
}
}
},
interfaces: [`Node`, `BlogPost`]
})
);

Then we need to select the subset of Mdx nodes to turn into the new node type and create the new nodes.

JS
exports.onCreateNode = ({
node,
actions,
getNode,
createNodeId
}) => {
const {
createNodeField,
createNode,
createParentChildLink
} = actions;
if (node.internal.type === `Mdx`) {
const { frontmatter } = node;
const parent = getNode(node.parent);
if (parent.sourceInstanceName === "posts") {
const fieldData = {
title: node.frontmatter.title
};
createNode({
...fieldData,
// Required fields.
id: createNodeId(`${node.id} >>> MdxBlogPost`),
parent: node.id,
children: [],
internal: {
type: `MdxBlogPost`,
contentDigest: crypto
.createHash(`md5`)
.update(JSON.stringify(fieldData))
.digest(`hex`),
content: JSON.stringify(fieldData),
description: `Satisfies the BlogPost interface for Mdx`
}
});
createParentChildLink({
parent: parent,
child: node
});
}
}
};

That's it. We now have two types that implement the BlogPost interface that we can query at the same time through the interface. This sets the stage for a number of really cool opportunities such as sourcing BlogPost content from arbitrary sources, like different CMSs.

Check out the full gatsby-node.js file here:

JS
const crypto = require("crypto");
const fs = require("fs");
exports.sourceNodes = ({ actions, schema }) => {
const { createTypes } = actions;
createTypes(`interface BlogPost @nodeInterface {
id: ID!
title: String
body: String!
}`);
createTypes(`type ContentYaml implements Node & BlogPost {
id: ID!
title: String @proxy(from: "subject")
body: String!
}`);
createTypes(
schema.buildObjectType({
name: `MdxBlogPost`,
fields: {
id: { type: `ID!` },
title: {
type: "String!"
},
body: {
type: "String!",
resolve(source, args, context, info) {
const type = info.schema.getType(`Mdx`);
const mdxNode = context.nodeModel.getNodeById({
id: source.parent
});
const resolver = type.getFields()["body"]
.resolve;
return resolver(mdxNode, {}, context, {
fieldName: "body"
});
}
}
},
interfaces: [`Node`, `BlogPost`]
})
);
};
exports.onCreateNode = ({
node,
actions,
getNode,
createNodeId
}) => {
const {
createNodeField,
createNode,
createParentChildLink
} = actions;
if (node.internal.type === `Mdx`) {
const { frontmatter } = node;
const parent = getNode(node.parent);
if (parent.sourceInstanceName === "posts") {
const fieldData = {
title: node.frontmatter.title
};
createNode({
...fieldData,
// Required fields.
id: createNodeId(`${node.id} >>> MdxBlogPost`),
parent: node.id,
children: [],
internal: {
type: `MdxBlogPost`,
contentDigest: crypto
.createHash(`md5`)
.update(JSON.stringify(fieldData))
.digest(`hex`),
content: JSON.stringify(fieldData),
description: `Satisfies the BlogPost interface for Mdx`
}
});
createParentChildLink({
parent: parent,
child: node
});
}
}
};