Firestore Collection to map rather than list

[AaronNBrock]

I have the following binding from my Firestore collection to my state:

state: {
  assets: [],
},

actions: {
  bindAssets: firestoreAction(({ bindFirestoreRef }) => {
    return bindFirestoreRef('assets', db.collection('assets'))
  }),
},

However this binds them to assets as a list, I.e.

[
  {id: "id1" /* etc. */ },
  {id: "id2" /* etc. */ },
  {id: "id3" /* etc. */ },
]

Whereas I'd like it to be bound as a map like:

{
  "id1": { /* etc. */ },
  "id2": { /* etc. */ },
  "id3": { /* etc. */ },
}

How could I go about achieving this?

[posva]

You can create a computed that generates the object from the array. More information would be helpful to decide whether this feature is worth adding or not 🙂

[dannydulai]

I have a list of documents from a query, and I don't need the entire document in memory, just one of the ids

is there a way to use serialize option to solve this?

[rendomnet]

@posva
I also need this feature. It is more handy to store large data in object. So we can easily refer to specific document from collection like this todos[id]status rather than search array.

[ajgagnon]

I'm sure this would involve a major, breaking change since all the mutations are already based on arrays. But it's definitely better architecture for performance reasons too.

Based on @posva, you could do a computed property like this:

computed: {
  assets: {
      get(array) {
        return array.reduce((obj, item) => {
          obj[item.id] = item;
          return obj;
        }, {});
      },
      set(value) {
        // covert back to array
        this.$store.commit('SET_ASSETS', Object.values(value));
      },
    },
}

It doesn't solve the performance issue since you're actually doing MORE computations, but will ensure:

1. You have no documents with duplicate ids
2. You can access assets like this.assets[id]

[azeem-r00t]

Has there been any update to this?

It is not just about a computed property. By not storing the collection as a map, I feel like we are not being true to how firestore stores these collections. In firestore, collection is not a list of documents rather it is a dictionary of ids mapped to specific documents. Because, this ends up being a generic list, it creates some interesting challenges when subcollections are being used. In my case, I have got a subcollection that is being loaded as each component mapped to a document from the parent collection loads. I need to store these subcollection somewhere mapped by the original collection id in the state. It has proven to be an unnecessary hassle thus far (and I have had do a bunch gymnastics to get it right). It would be great if this was changed.

[charles-allen]

Based on @posva, you could do a computed property like this:

computed: {
  assets: {
      get(array) {
        return array.reduce((obj, item) => {
          obj[item.id] = item;
          return obj;
        }, {});
      },

FYI if you're using lodash:

getters {
  keyedAssets: (state) => keyBy(state.assets, 'id')
}

Though obviously building a bindMapRef() into the framework will be more efficient (since VueFire/VuexFire must already be enumerating the snapshot). +1 :)

[charles-allen]

I have a weaker case, but I'd like to also throw it out for discussion...

Consider I have another type that references assets, such as comment: { assetId: 123, ... } (i.e. many comments per asset). If I wanted to download comments related to multiple assets at once (e.g. a page showing multiple assets + their comments, or if I wanted to cache everything in Vuex), I'd probably want my comments stored like this:

{
  assetId1: [comments],
  assetId2: [comments],
  ...
}

Today I would do this in 2 steps:

1. Bind all relevant comments: bind('comments', db.collection(COMMENTS).where(...))
2. Use a getter to group them: get commentsByAsset() { return groupBy(this.comments, 'assetId') }

I think we could avoid the overhead of creating & storing the comments array if VueFire did the groupBy internally during snapshot parsing (similar to the OP's proposed change), eg: bindGroupBy('commentByAsset', ...). That said, an obvious argument against this is API-bloat!