What is the more performant option in Firestore, an array of objects or a subcollection?

Question

I'm thinking about the best data model of my app using Firestore. I don't know which option is more performant/scalable? This one:

[Collection] Estimates -> 
    [Document] Estimate1 = items: [{name: 'Design', price: 200}, {name: 'Development', price: 200}]

Or this one:

[Collection] Estimates -> 
    [Document] Estimate1 ->
        [Collection] Items ->
            [Document] Item 1 = {name: 'Design', price: 200}
            [Document] Item 2 = {name: 'Development', price: 200}

On the other hand, other than perfomance issues, I have another consideration before I choose a data model. That is that I'm going to need to copy all contents of the 'estimate' node to an 'invoice' node and I don't know if I can do this in both data models.

Answer 1

This documentation will show you some of the limits of the FireStore Document model.

https://firebase.google.com/docs/firestore/quotas

In answer to your question(s):

It depends! It depends on how many items you will have in your estimate. A Document can hold 1 MB of data, so it is limited in scalability.

[Collection] Estimates -> 
    [Document] Estimate1 ->
        [Collection] Items ->
            [Document] Item 1 = {name: 'Design', price: 200}
            [Document] Item 2 = {name: 'Development', price: 200}

If you go with the above option, it will definitely be more scalable - you can have practically as many documents in the collection as anyone would ever want.

Both options will work, and you can copy the data to an invoice either way.

It may, however, be cheaper to just use one document, because firestore charges you based on each write and read. So for example, if you write all the data at the same time, in one write, to one document, that will be cheaper than writing all the data to several different documents, even if you use the WriteBatch to write them all at the same time.

Based on the model and guessing on what it is you are building, I would venture to guess each estimate won't have more than 100 items in it, so I would go with one doc for the estimate.