Optimize database queries to only pull out fields actually requested

Can use the info parameter passed to the resolver to pick out only the fields.

Eg, given this schema:

type User {
  id: ID!
  name: String
  avatar: Blob
  longVideo: Blob
}

When we do a query like:

User(id: "foobar") {
  id
  name
}

We don’t want to pay the cost of pulling the avatar and longVideo out of the database.

In the resolver, we can use the info parameter to see what fields are actually requested:

const resolvers = {
  User: (_, { id }, context, info) => {
    const nodeQueryInfo = info.fieldNodes[0]; // 0th is always the currently resolving node
    const fields = nodeQueryInfo.selectionSet.selections.map(({ name }) => name.value);
    return db.findById(id, fields); // (or `SELECT ${fields.join()} FROM ...`, etc)
  },
};

[jesstelford]

From https://github.com/keystonejs/keystone-5/pull/665#discussion_r

Will also need to think about it in the context of the hooks framework, and whether we need to perform this operation anyway to provide the full object to the hooks.

🤔 The test could be

if (hasHooks) {
  returnEverything();
} else {
  limitResults();
}

[timleslie]

@timleslie was any of this optimised with the relationship restructure?

Nope, this issue is still in the category of nice to have, but tricky to implement.