Project: Work on Theme extensibility

[khmakoto]

Problem

We have had multiple partner asks for an ability to extend our Theme definition with extra tokens of their own creation. We currently have no way of doing this and restrict everyone to only use the tokens we define.

Tracked also in #20900

Appetite

1-2 weeks for initial implementation of Shift's RFC and exploration of what is left to do after that.

Solution

Shift has written a great RFC in #20651 to remove functional style rules from makeStyles and replace them with imports from the theme/tokens itself that would solve most if not all of this problem. The goal of this project is to drive that RFC to completion and implement it. And then to investigate if there are any pieces left for Theme extensibility to work after this is done.

Risks (Rabbit holes)

If the RFC does not get enough traction, it might be hard to finish this work during the specified time-frame.

Out of scope (No-gos)

• Whatever work is left to do after the implementation of this initial RFC.
• Moving makeStyles out of @fluentui.

[khmakoto]

What was achieved?

Most work related to this project was completed and split among multiple PRs.

What did we miss?

Because of the number of PRs generated and how close to the holidays they were published, these PRs were not able to be merged during the December sprint.

What went well?

Not much changed from initial expectations aside from some babel-plugin issues that were fixed as part of this, which meant we could focus on the matter at hand.

What could have gone better?

The PRs had to remain in draft form for a long time because the RFC that preceded this work had not been approved yet, so we were left waiting to see if the approach was going to change.

What changed from initial expectations?

While not much changed in terms of expectations from beginning to end, one thing that was very time consuming was having to split the initial PR up into multiple distinct ones, which also probably increased the investment needed to review all of them (when added together, the time investment per individual PR did go down dramatically). This might be something worth discussing in the future to see if we should always follow that approach or not depending on the nature of the change.