So given the following schema:

users
    id (int)
    some_group_id (int)
    some_field (string)
some_group
    id (int)

I'd have the following relationship method in my User model:

public function someGroup()
{
    return $this->belongsTo('SomeGroup');
}

Normally I'd therefore access the various fields like this:

$user = user::find(1);
echo $user->some_field; // echoes the value of some_field
echo $user->someGroup->id; // echoes the value of the id field on the related SomeGroup

That's all well and good, but I found out today that I can access the relation using snake case:

echo $user->some_group->id; // works!

This is good for me because it means that my view code can now have more continuity. Instead of the disparity between $user->some_field and $user->someGroup I can use $user->some_field and $user->some_group.

Interestingly, the same doesn't work the other way - so I can't use $user->someField.

My question is can I rely on this behaviour? Is it defined somewhere? If I start to use it, can I rely on it to not change (with reasonable confidence of course, I'm aware even defined behaviour can change)?