My app has leases of units. A lease has a move in and a move out date. Depending upon those dates a lease is implicitly “past”, “active”, or “future”. I have composable queries like filter_past_leases and filter_future_leases and such. I would like to have an additional filter query that returns leases AND includes their status.

This could be done with a specialized query that returns leases in a list of maps with something like [%{lease: lease, status: "past"}, ...]. This is mostly fine to work with, but is not composable. I’d rather return a list of Lease structs and have the status included somehow and still be able to further filter the query as needed.

One way to do this is to use a virtual field. Basically, add a virtual status field on a Lease struct. Then a query uses select_merge to merge the calculated status into the Lease struct. This works really well and the query can then be further composed like any other ecto query.

Lease
|> preload_lease_unit()
# composable query in a pipeline that fills in "status" field
|> calculate_lease_statuses()
|> filter_leases_by_property(property)

I guess my only hesitancy is that it seems like virtual fields have a bit of a bad reputation. Perhaps that reputation has more to do with overusing virtual fields in cases where a changeset or different struct could be used to encapsulate the logic (e.g. setting a password).

Does using a virtual field and select_merge seem reasonable or should I consider alternate methods?

I should note that I’d like to avoid hard-coding a status because 1) it becomes something that needs to be updated in multiple scenarios and 2) I need to be able to calculate “past”, “active”, and “future” for any given day rather than just the current day (e.g. occupancy 30 days out).