I'm adding a new column (bigints, initial values all null), to a large (~5 billion rows) Postgres (9.6) table.
I want to add an index on the row (algorithm: concurrently), and I'm wondering if there's a way I can save it the time of scanning through all five billion null fields to add them to the index, since I'm adding the index when the column is brand new and has no data. Is there any kind of optimization for this, or should I just let it run? (Is it even theoretically possible for there to be an optimization, or does it need to add something to the index even for null fields?)