I'm designing an API, that I would like to make asynchronous. On the other hand I would still like to keep a simple to use synchronous version of it. Is there a simple way to achieve this?
I imagine something like a decorator that auto generates blocking wrappers for my async def methods.
It's possible to make a decorator that just creates an event loop every time you want to run an async function, but that adds a lot of overhead and confusion.
Another common practice is to create a client that can accept a transport, that way you can just choose to make the transport sync or async, for example, this project https://github.com/aio-libs/aioelasticsearch, which takes the sync elasticsearch pkg, but adds an async transport that you can use with it.
Or elasticsearch's official solution to using async with their pkg https://elasticsearch-py.readthedocs.io/en/master/async.html
Anyways, if you ask me I think its better to stick to one thing, if your project benefits from async, just keep the client async.
Running an async function requires an event loop that takes control over the execution.
Just consider every async function you want to convert as a separate program. What you are asking for is then running an async program.
# async version:
await asyncfunc(arg1, arg2, ...)
# sync version:
asyncio.run(asyncfunc(arg1, arg2, ...))
Writing such decorator should be is easy.
Of course, the overhead is big.
