Firestore: Subcollections vs Root-level collections

Question

I am working on a quiz game, where players have to answer questions of different categories (sports, music, .....). Right now, I have two different ideas how to store these categories /questions in Firestore:

Creating a new Root-level collection for each category. Every document in this collection would define a question of that category. Stcuture would look like this:
- user_collection
  - user_1_document
  - user_2_document
  - user_X_document
- music_questions_collection
  - music_questions_1_document
  - music_questions_2_document
  - music_questions_X_document
- sports_questions_collection
  - sports_questions_1_document
  - sports_questions_2_document
  - sports_questions_X_document
Creating ONE Root-level collection and create a document inside this collection for each category, where each document contains a subcollection of questions:

user_collection
- user_1_document
- user_2_document
- user_X_document
category_collection
- music_document
  - music_question_subcollection
    - music_question_document_1
    - music_question_document_2
    - music_question_document_3
- sports_document
  - sports_question_subcollection
    - sports_question_document_1
    - sports_question_document_2
    - sports_question_document_3

I have already read about some advantages and limitations of subcollections and root-collections, but have not found enough information regarding read-costs and speed for this specific use case. Can anyone give me some information about drawbacks and advantages of the two listed approaches.

There is a third option. Create one collection for all quiz questions and make fields like `questionLevel` `category` `questionNumber`. And just query them by fields: `query(collection(db, 'quizes'), where('category', '==', 'Sport'), where('quizLevel', '>=', '5'))` — Mises, Jul 24 '22 at 07:32
@Mises the problem I have with this approach is, that you first have to retrieve ALL questions from a specfic category (where('category', '==', 'Sport') and then retrieve a single question, which is not really cost efficient, because the category collection will contain a lot of documents(questions). — Clyon97, Jul 24 '22 at 08:21
In my opinion, you are not cost-efficient. Read about persistence. You can always download once documents and read them from browser cache. And if you make a separate collection for each category, you still have to query them all! — Mises, Jul 24 '22 at 08:28
Oh, and you are not able in Firestore to get more than 100 documents at once. And to get the certain number of documents you're using `limit(5)` function in query, and you're getting just 5 documents. Read about how to make paginated requests to Firestore. — Mises, Jul 24 '22 at 08:34
@Mises So let's say I have a collection called "Sports" with 10 000 documents in it. I also put an ID on these documents (1 ... 10 000). In my cloud functions, I would just create a random number between 1 and 10 000, and then query over sports collection with the condition: where('ID', '==', random_Number), which will result in only one document read. — Clyon97, Jul 24 '22 at 08:38
@Mises okay you are right, limit(x) would do the job and prevent read inefficiency — Clyon97, Jul 24 '22 at 08:39
Let us [continue this discussion in chat](https://chat.stackoverflow.com/rooms/246708/discussion-between-mises-and-clyon97). — Mises, Jul 24 '22 at 08:44
@Mises You mention that "you are not able in Firestore to get more than 100 documents at once". AFAIK there is not such a limit. Can you point out the reference for this limit? — Renaud Tarnec, Jul 24 '22 at 09:16
@RenaudTarnec Good to know. I don't have a clue why I was sure there is a limit to 100 documents. There is just maximum API request size 10 MiB, and it is about writes not reads. — Mises, Jul 24 '22 at 09:55

score 2 · Answer 1 · edited Jul 24 '22 at 08:15

I have not found enough information regarding read-costs and speed for this specific use case. Can anyone give me some information about the drawbacks and advantages of the two listed approaches?

In terms of "read-costs and speed", the two approaches are equivalent: using a root collection or a sub-collection is exactly the same.

As a matter of fact, from a technical perspective, a root collection and one of its sub-collections are not at all related to each other. They just share a part of their path but nothing else.

So, since you store the same documents in either a root collection or a sub-collection (i.e. same number of docs in each category collection and same field structure for these docs), there is no difference.

Okay I thought so; thank you very much!!! – Clyon97 Jul 24 '22 at 08:22 — Clyon97, Jul 24 '22 at 08:22

Firestore: Subcollections vs Root-level collections

1 Answers1