stm32-hal VS unasync

Typestate-free HALs: This is in exchange for better ergonomics as the author claims. Only two HALs fall in this category right now which are the STM32-HAL & nRF-HAL.

As I worked with the stm32, as implied earlier, the HALs that I worked with were ones built around embedded-hal traits. Nevertheless, I came across a HAL at a certain point that adopted a different approach that felt more practical and easy to understand. This was the stm32-hal stm32-hal that I found to be more wholesome as it incroporated multiple families of the STM32 under a single HAL umbrella (my original expectation). The STM32-hal eliminates much of the trait confusion that I had encountered before. The thing is the stm32-hal does not seem to be mainstream yet. From what I understand, the HALs built with the mebedded-hal as a basis seem to be the ones mainly adopted by the embedded working group. Additionally, I am not sure if the stm32-hal has any equivalent counterparts for other manufacturer devices.

📝 At the time of writing this post, it came to my attention that there is an additional HAL that targets STM32 device families (the stm32-hal). From what I figure, right now there seem to be two approaches for developing HALs. The first approach is trait driven so to speak where the embedded-hal is used as a foundation. The second approach is more application-driven and provides a high-level API that targets several families of a device. However, this exists only for the stm32 through the stm32-hal. Right now, the first approach is what I found to be more widespread as it covers different microcontrollers and what this post is based on.

For STM32, check out the Peripheral Access Crates by the stm32-rs ream. For higher-level access, I wrote This HAL library for STM32. Works on most newer variants, and includes examples for specific peripherals, and simple applications.

Nice! This is similar to the solution here: https://github.com/python-trio/unasync

Lastly, I found another project name unasync that is pretty interesting and might works for me. Basically, you write the async version, you run unasync, it generate the sync version from the AST. This project is used by the official elastic search python client.

Fibers "allow blocking and non-blocking implementations to share the same API"

That's an interesting contrast to Python where the need to use "value = await fn()" v.s. "value = fn()" depending on whether or not that function is awaitable causes all kinds of API design complexity, all the way up to the existence of tools like https://github.com/python-trio/unasync which can code-generate the non-async version of a library from the async version.

Async Python has proven faster in my uses for IO and non-CPU-related stuff. But I think Python, either as a community or within the language, needs to solve the anti-pattern of maintaining separate sync and async versions of a library. I'm thinking specifically of aioredis and redis-py, both of which I've worked on.

Some people are looking at ways to solve this. I know urllib3, elasticsearch-py, and a few others use unasync (https://github.com/python-trio/unasync) to transform async code into sync code, leaving one codebase supporting both uses in different namespaces. This leaves you with some conditional logic (is_async_mode() -- https://github.com/python-trio/hip/blob/master/src/ahip/util...). I'm seriously considering this approach.