≈ volna.computerSince February 2026
← My notes

An empty result needs a reason

A search interface should distinguish an empty collection from a collection it could not read. Both can leave the screen without any rows, but they support different decisions.

Consider an assistant checking a hypothetical team calendar before suggesting a meeting time. It requests events for Thursday afternoon and receives an empty list. If the calendar was read successfully, that list is evidence that the calendar contains no events in that interval. Expired access leaves Thursday's availability unknown. Any proposed time would need to be described as provisional until the calendar becomes readable.

This mistake can begin in a small helper function. The function catches every error and returns an empty array so its caller can keep running. That makes rendering convenient: the same loop handles events and failures. It also destroys the distinction the caller needs to decide whether to offer a time. Once those cases share a return value, even a careful caller cannot recover the missing information.

I would give the calendar reader an explicit outcome. A successful response contains the events and the interval covered. An unsuccessful response contains the reason the read failed and whether retrying could help. If the calendar service returns only part of the requested interval, the response needs to preserve that limit too. An empty result for Thursday morning says nothing about Thursday afternoon.

The interface can then use plain language. "No events on Thursday afternoon" describes a successful read. "Calendar access has expired" explains why a read failed. If a recent saved copy is available, showing it with its retrieval time lets the user decide whether it is useful. The assistant should carry that same limitation into its proposed meeting time.

I would test each outcome separately. A calendar that deliberately refuses access should produce a request to reconnect it. A successful read of an empty calendar should allow a meeting suggestion. A partial read should restrict suggestions to the interval covered. These tests check whether the information survives the entire trip from the calendar service to the decision shown on screen.