Tracking issue to reduce boilerplate

[bugadani]

The goal of this issue is to lower the amount of code the users need to write for no good reason.

• 1957

• 2012

• 1970

• 1999

• 2070

• 2133

• 2134

• 2136

• 2248

• 2260

• Type-erase peripheral instances where practical

  • 2334

  • 2359

  • 2361

  • 2367

  • 2381

  • ADC
  • Timers
  • Add to API-GUIDELINES to design new drivers with erasure in mind.

• 2339

• 2374

• 2373

• Resolve #2321
• Resolve #2389

• 2403

• 2532

• 2486

• 2508

Feel free to share your pain point if you have any!

[scottdalgliesh]

Would it make sense to also initialize and return the io driver by default in esp_hal::init? It seems like most use cases will require access to io, but I could certainly be overlooking/misunderstanding something.

Regardless, great work on these simplifications - love to see it.

[bugadani]

^ a good thing to keep in mind, although we deliberately don't initialize any peripherals at this point.

[MabezDev]

I think we should think about closing this.

It's been great to track the "obviously" good choices (the init changes etc), but I think at this point most of the decisions we need to make going forward are non-trivial and require discussion before we commit to the design. We should track those in individual issues.

[bugadani]

I suppose I'll mark the last DMA PR to close this :)