如何在 Python 中同时迭代和运行 AsyncGenerator

Question

A bit new to Python, not sure if this question is too naive. Trying to grasp the concurrency model.

Third party function (from the library) connects to multiple hosts through ssh and perform some bash command. This function returns AsyncGenerator.

Then my code iterates through this AsyncGenerator using async for :

        # some code before
        asyncgenerator_var = # ...
        async for (host, exit_code, stdout, stderr) in asyncgenerator_var:
            if exit_code != 0:
                self.err(f"[{host}]: {''.join(decode_strings(stderr))}")
                continue
            self.out(f"[{host}]: {''.join(decode_strings(stdout))}")
            self.err(f"[{host}]: {''.join(decode_strings(stderr))}")
        # some code after

And then code calls await on this function. But if runs one after another. Not concurrently.

Code someone explain why is that? And what should be done to make it run concurrently.

Answer 1

In async model, no functions are run simultaneously. The event loop may switch functions if the current function is await -ing other functions / futures.

The async for statement essentially means the event loop may run other scheduled callbacks/tasks between iterations.

The async for body still run in the order yielded by the async generator.

To run the body in arbitrary order, wrap it inside an async function. Create a separate task for each input, and finally gather the results of all tasks.

# some code before

asyncgenerator_var =  # ...

async def task(host, exit_code, stdout, stderr):
    if exit_code != 0:
        self.err(f"[{host}]: {''.join(decode_strings(stderr))}")
        return 
    self.out(f"[{host}]: {''.join(decode_strings(stdout))}")
    self.err(f"[{host}]: {''.join(decode_strings(stderr))}")
    
tasks = []
async for (host, exit_code, stdout, stderr) in asyncgenerator_var:
    tasks.append(asyncio.create_task(task, host, exit_code, stdout, stderr))

await asyncio.gather(*tasks)
    
# some code after