I have a Go application, which uses Firestore to store data.

When updating a document, I’m passing a map of data to Firestore, but it behaves differently, depending on whether I’m using firestore.MergeAll, or firestore.Merge(...).

I have the following struct:

type User struct {
    Tokens Tokens
}
type Tokens struct {
    Sessions map[string]Token
}
type Token struct {
    Token string
    Expires int
}

In certain situation, I’d like to clear the Sessions property of any values, and I do it like so:

dataMap := map[string]interface{}{
    "Tokens": map[string]interface{}{
            "Sessions": map[string]interface{}{},
    },
}
_, err = c.ref.Doc(id).Set(ctx, dataMap, firestore.MergeAll)
// dataMap looks like so: map[Tokens:map[Sessions:map[]]]

When I do this, I get a Firestore error: firestore: missing WriteResult.

However, if I update it like below, it works as expected, and Sessions becomes an empty map inside Firestore…

dataMap := map[string]interface{}{
    "Tokens": map[string]interface{}{
            "Sessions": map[string]interface{}{},
    },
}
_, err = c.ref.Doc(id).Set(ctx, dataMap, firestore.Merge(firevault.FieldPath{"Tokens", "Sessions"}))
// dataMap looks like so: map[Tokens:map[Sessions:map[]]]

What is the difference between the two? Is there a way to achieve the same functionality with firestore.MergeAll, so that I don’t always have to pass a FieldPath?