Select field improvements

[aerminas]

Hi!
I want to propose improvement for select fields across the Home Assistant.

In many places I stumble across this pattern - user need to select device or entity from list

• list of devices or events can be very lengthy
• if i missclick, content is not saved in the field (I need to write it again)
• it’s hard to find specific element, as similar name could be used in different things (room name can be used in device name, so even that I can filter by writing in the field, I would still get extra items)

Here’s prototype for problem statement:
https://www.figma.com/proto/AhcHPxlRHRxGJiqc9BOWqv/HomeAssistant?node-id=0-1&p=f&viewport=-11173%2C-389%2C0.6&t=5cjzkp8n9XcWlBBI-0&scaling=min-zoom&content-scaling=fixed&starting-point-node-id=2%3A23&show-proto-sidebar=1

My proposal is to modify the input field:

• as a error prevention we should keep the value provided by user in the field even if he close the select field (missclick)
• we should allow more advanced filter handling inside of the select field : there should be a way to select specific value to filter by (for example room name, type of device ect).

See the proposal prototype:
https://www.figma.com/proto/AhcHPxlRHRxGJiqc9BOWqv/HomeAssistant?node-id=0-1&p=f&viewport=-11173%2C-389%2C0.6&t=5cjzkp8n9XcWlBBI-0&scaling=min-zoom&content-scaling=fixed&starting-point-node-id=3%3A3435&show-proto-sidebar=1

Prototype should clearly explain how I see the solution for problem - UI details are something we can discuss, but overall the user journey and interactions are something I would like to get your opinion on.

Solution should be keyboard friendly.

What do you guys think?

[JonLaliberte]

as a error prevention we should keep the value provided by user in the field even if he close the select field (missclick)

Definitely. This happens to me all the time. I'd even like it to possibly keep the search after selecting an item for multi-select instances. Right now the UX is: type, click, type, click, type, click. Most of the time it could just be: type, click, click, click.

The rest of it would be nice to have, it just needs not be too aggressive at assuming intent (re: room selection).