Consider refactoring CommandPalette + Switcher => LiveFilteringListView

[DHowett]

There's a good core of an implementation here for a live filtering list view, and I believe that having it serve two leaders (command palette and tab switcher) may unnecessarily complicate its implementation.

Is there something we can do about that?

[Don-Vito]

@DHowett - I guess I am causing a lot of headache for you with tons of small PRs (realized how distracting this can be after reading https://www.wired.com/story/open-source-coders-few-tired/).

I cannot work on big tasks, especially the ones that are prioritized (very hard to combine it with a family and 12h/day work), especially the ones that require a deep understanding of the architecture and the product. Hence I am hunting for quick wins (up to one night of coding) that are close to surface (I have no clue how rendering works, why VT, etc., etc.)

So probably instead I can work on a stuff like this

• it will take me a while
• the interface is simple
• it is quite standalone (a lot of coding that can be merged, and then a short integration)
• it is not on the roadmap

Actually I would like to do even more with the Command Palette, which is IMHO much more complex than it should be. This is at least how I would like to see it:

• We need a simple ephemeral host control with no business logic (it should open nicely and close on time - something like a flyout)
• Inside the host we will have
• a control for Tab Switching (with the tab search capabilities)
• a control for Actions (with nested commands capabilities)
• a control for Command Lines
• a logic that switches between their visibility (can be fully Xaml)
• Terminal page will work with these controls directly (updating data via model-viewmodel binding, and registering to business domain events like "CommandSelectedForExecution", "TabSelectedForPreview", "TabSelectedForSwitch")
• LiveFilteringListView will implement a significant part of the common logic (filtered command has already extracted the filtering and sorting, but we still need to extract the list rendering and the navigation)

But this one I am afraid to work on, as it will take me few months to implement and even months to merge - I will probably die while merging conflicts.

Long story short - let me know how I can be of a greater use (and a lesser pita). What would you prefer: polishes like today, stuff like the above or something else (given the limitations I mentioned)?

@zadjii-msft - FYI.

[zadjii-msft]

I know you originally asked Dustin, but these are my thoughts on the matter:

Absolutely don't worry about sending out a ton of small PRs. I kinda like it, they're more easily digestible and reviewable IMO. And at this rate, you've probably committed more code to the Command Palette than I have 😆 It's been nice to have someone working on UX improvements while I've been so heads-down on architectural pieces.

This does seem like a task that's relatively right up your alley, but if it's not something that works with your own contribution schedule, don't worry about it. I think the team is happy to work on more of the big-picture architectural work and let the community come in with the smaller polish fixes and features. I think at the end of the day, you should work on whatever you're passionate about and and think you can ship. I have no doubt that you'd be able to get this one done, but if you want to work on 10 other smaller fixes instead, that's great too.

That being said, I don't think I'd worry too much about merge conflicts till the end of the year - probably a lot less code flow than usual with all the holidays coming up 😄

[Don-Vito]

I know you originally asked Dustin, but these are my thoughts on the matter:

@zadjii-msft - I really value your opinion as well, as you are mostly the one working with me! 💟

I asked Dustin while assuming he suffers most: usually it's an area expert + Dustin who review the code (in addition to all other context switches he has being a lead).

[DHowett]

@Don-Vito i love smaller PRs. They're easy to cherry-pick and when you look back at the history there's a bunch of individual pure changes -- easily revertable, easily blameable, etc.

I'm always asking folks on the team to please break this up into multiple PRs. Sometimes it works ;) and sometimes it doesn't!

[DHowett]

That having been said: this is a pretty intense refactoring, so I would totally not expect you to (if you choose to do it) break it into smaller chunks. Sometimes you need to rip the band-aid off and come in and wrecking-ball something. miniksa has done that more than once 😁