在 Cloud Firestore onCreate 触发器中运行批处理或事务

Question

Let's imagine that, in firestore, I have a collection of vendor documents :

vendors : {
  vendor1: {
    id: "vendor1",
    name: "John",
    shopId: "shop1"
  },
  vendor2: {
    id: "vendor2",
    name: "Mary",
    shopId: "shop2"
  }
}

And a collection of shop documents :

shops : {
  shop1: {
    id: "shop1",
    name: "My Super shop - City A",
    vendors : {
      vendor1: {
        id: "vendor1",
        name: "John"
  },
  shop2: {
    id: "shop2",
    name: "My Super shop - City B",
    vendors : {
      vendor2: {
        id: "vendor2",
        name: "Mary"
  }
}

For read performances, each shop document has a copy of its vendors. I copy only vendor data that are needed in my view (mobile app), and I update vendor datas on shop documents if an onUpdate trigger is launch on vendors collection.

Today, what I do is :

exports.updatesOnCreateVendor = functions
  .firestore.document("vendors/{vendorId}")
  .onCreate(async snapshot => {
    const vendor = snapshot.data();

    const { shopId } = vendor;

    const shopRef = db.collection("shops").doc(shopId);

    const shopAfter = {
      vendors: {}
    };

    shopAfter.vendors[vendorId] = { ...vendor };

    const batch = db.batch();

    batch.set(shopRef, shopAfter, { merge: true });

    return batch
           .commit()
           .then(console.log)
           .catch(console.error);
  });

I don't want to loose this copy if the cloud function failed (and I don't want to active retries on cloud functions). (same constraint onUpdate). After reading the documentation, I conclued that using batch or transaction will guaranty that it will retry if it hasn't succeeded after 25 tries.

So, I would like to know if it's a best practice to use batch / transaction like that on cloud function triggers ?

If I refactor this and replace the batch process by just using shopRef.set(shopAfter, { merge: true }) and if the cloud function failed, what's happening ? (I think I'll lose my copy :p )

Thank you for the time passed to help me better understand Firebase :)

Answer 1

The batch isn't going to change your situation at all here. A batch write of one document isn't any different than a normal write of one document. All a batch will do is ensure that all of the documents writes happen atomically, all taking effect at the exact same moment.