API improvements

[SergioGasquez]

• Created non_exhaustinve FlashData and FlashSettings structs to reduce number of input arguments in the following functions:
  • save_elf_as_image
  • flash_elf_image
  • load_elf_to_flash
  • get_flash_image for all targets
• Using FlashData avoids duplicating code of parsing the bootloader and partition table file that was present in several places.

Not completely convinced on the FlashData name.

[jessebraham]

Thanks for taking care of this! I took a quick look and I think it looks pretty good overall, but I have a couple features I'd like to continue developing to see how they affect this.

Will try to get this taken care of sooner than later and update here when able.

[SergioGasquez]

Rebased on top of #516! But now Clippy is back at screaming that a function has too many args (8/7) :(

[SergioGasquez]

Rebased on top of https://github.com/esp-rs/espflash/pull/525! Clippy is now screaming louder that a function has too many args (9/7)

[bugadani]

I've spent a few minuted touching the related code in probe-rs, here are my initial impressions:

I find #[non_exhaustive] structs with all-public fields somewhat weird if I don't have a shorthand to create them. If I have to basically list all fields while calling new, it's just a worse version of non-#[non_exhaustive]. That FlashData does some actual login in its new somewhat mitigates this point, but:

I think a builder for FlashData would help greatly in its usability. Many optional parameters in new doesn't really help usability compared to the previous state. If the API looked like Builder::new(required_params).with_optional_x(x).with_optional_y(y).build() there would be no question of what the individual None-s mean and which order they come in.

Also I find it somewhat weird that I have to specify a chip revision for FlashData, then I need a minimum revision (a pair, which is... unclear from the signature) in get_flash_image.

FlashData::new() also does not return espflash::error::Error. I'm not yet sure if espflash::error::Error was split up or changed further, but if not, then this fact complicates error handling a bit. miettealso does not recommend Report:

It’s a best practice to wrap any “external” error types in your error enum instead of using something like Report in a library.