Optimal way to backfill new field into historical data

influxdb
#1

What is the most efficient way of backfilling a new field into all the points within a measurement?
My current approach is to do a select * on progressively older time slices. And write back points with the same tag set, all original field removed, the new field added in and keeping the timestamp.
However this seems to be horribly inefficient. Do you have any better idea?