Several months ago we had some detailed discussions about whether to use separate date columns to indicate a date range, or to use the daterange data type. We opted for the latter because this type is specifically designed for this use case - a table that has a range of valid dates for the data it contains. It also has some great operators and functions.
But I recently discovered that daterange is not supported in any way as a partition key because it depends on an expression. I was excited about this possibility in pg11 with unique constraints on the parent table, but now it appears it may have instead been to our advantage if we had two separate date columns instead, so that we could use UPSERT transparently for date-ranged tables.
Is there any possibility of this feature coming for range types, or, if we really want to partition using daterange, should we look instead at two separate date columns?
Thanks,
Jeremy