POC React-Scenes in SM

[ckbedwell]

Problem

In SM we find Scenes quite burdensome to write and often blocks our ability to innovate and iterate on new ideas we have for features. The API is often difficult to reason with as your mental model for how React works does not align very well with the way that scenes is currently written.

Acceptance criteria

Before exploring alternative ideas (such as potentially dropping Scenes altogether... 🙈 ), we want to see if the proposed React-Scenes feature is a viable solution addressing our developer concerns for managing a scenes-heavy application.

• Create a POC branch rewriting a single dashboard (for the POC do not do multiple dashboards unless it feels warranted).
• Write a short report of the findings (e.g. pros and cons) when converting a dashboard from the current Scenes API to the React-based one.

The outcome of this POC should be an honest evaluation of if we should continue to invest in Scenes or explore alternative approaches which align better with our goals.

A running list of some of the challenges we face using Scenes right now (please update as more issues are found):

• Change the query based on the time range selected (e.g. under 768 hours use a Mimir instant query, above 768 hours use a Mimir range query)
• Query management.
  • e.g. right now I (@ckbedwell) am exploring a custom logs visualisation and keep thinking how the Loki query could be 'lifted' and provide additional context to all of the other panels on the dashboard. I know that if I did this in react-query the query provider automatically provides it to any other component that would wish to use it. I don't know what that workflow looks like in Scenes? Lifting the QueryRunner? Should all the QueryRunners be lifted to the outermost parent?