The Problem

Vocabustudy has a feature called custom collections. Users can supply Set IDs that are in the collection. The question arises: How do you fetch all of those sets in one network request?

This is how it currently works in custom-collections.js with the Firestore Lite JavaScript SDK:

import { collection, doc, documentId, getDoc, getDocs, query, where } from "firebase/firestore/lite";
let customCollection = await getDoc(doc(db, "collections", collectionId));
let collectionData = customCollection.data(); // the property `sets` contains a list of document IDs
let sets = await getDocs(query(collection(db, "meta_sets"), where(documentId(), "in", collectionData.sets)));

Tip: documentId() is a special key for querying on the, well, document ID

You might already be able to see some problems with this. Firestore has a (really annoying) limit for in, array-contains, etc. where you can only supply up to 10 items to search in. How am I supposed to allow users to put more than 10 sets in a collection?

Side note: Vocabustudy already implements pagination in groups of 10, so I could have just supplied a list of 10 IDs at a time. However, this wouldn't work if the pagination groups are larger then 10. Read on!

The Solution

Use the REST API directly instead! Under the hood, the SDKs just make calls to it. I've actually found many occasions where the REST API has more features than the SDK, including this example and document creation date/times.

While poking through the docs, I found the perfect method. It's called batchGet.

This is how I would use it:

let sets = await fetch("https://firestore.googleapis.com/v1/projects/PROJECT_ID/databases/(default)/documents:batchGet", {
    method: "POST",
    headers: {
        "Content-Type": "application/json"
    },
    body: JSON.stringify({
        documents: collectionData.sets.map(setId => `projects/PROJECT_ID/databases/(default)/documents/meta_sets/${setId}`)
    })
});

What's interesting is that rather than specifying a list of collection and document IDs, you specify the full path, including the project and database ID. The docs state that you have to fetch documents from the database specified request URL, but fetching across databases might be a future feature!

However, this does mean you can fetch documents across collections, in one request! That itself it a big benefit over my original solution.