Description

I've noticed[1] a difference between DataSnapshots methods in this SDK compared to the one used when fetching data (when using the admin SDK).

When setting values that are (or contain) empty object/arrays and null values using the admin SDK, when fetching these values back, they're always null when calling .val().
However when creating a DataSnapshot with these values and calling .val() the results don't match.
I'm only raising this as an issue because when the functions are being called, I'd expect the DataSnapshot provided to the function handler to match the DataSnapshot when fetched from the same location - currently this isn't the case.

Code sample

Using the following data: { child: 'foo', nullChild: null, emptyChild: {}, emptyArray: [], child: { nullVal: null, key: 'bar' } }

Setting and then getting this value using the admin SDK [2] results in { child1: 'foo', child2: { key: 'bar } } when calling .val(), which is what you'd expect - all the empty/null nodes are dropped.

However using the firebase-functions database.DataSnapshot constructor the result includes these empty/null values { child: 'foo', nullChild: null, emptyChild: {}, emptyArray: {}, child: { nullVal: null, key: 'bar' } } when calling .val().

I think this is mostly due to the lack of this check the @firebase/database does when calling .val().

The discrepancies are mainly when using .val() but I've updated the other methods and their tests (.exists(), .numChildren(), .hasChildren(), etc.) for completeness...

[1] noticed when using firebase-functions-test makeDataSnapshot method (which calls this repos database.DataSnapshot constructor) for mocking parameters for testing a Firebase Function.

[2] By using:

const firebase = require('firebase-admin')
firebase.initializeApp({ databaseURL: 'https://<databaseName>.firebaseio.com' })

const setAndGet = async val => {
   const ref = firebase.database().ref('node')
   
   await ref.set(val)
   return await ref.get()
}