Using the client in a Web Application

Morning, We have a standard web application (in our case Django, but the example holds for most cases including Flask, CherryPy, etc.,)…

I’d have expected this to work inside a function in my web app

        with vt.Client(api.api_key) as client:
            file_info = client.get_object(
                '/files/44d88612fea8a8f36de82e1278abb02f')
            return file_info.sha256

But it doesn’t unfortunately. This is due to the _make_sync call which tries to get the current event loop and run the awaitable to completion. I believe this will cause a fair bit of consternation for most web applications.

Any suggestions on how the library is intended to be used inside web applications? Here is the traceback

Traceback (most recent call last):
  File "/..../actions/vt_get_file.py", line 47, in process
    file_info = client.get_object(
  File "/.../lib/python3.8/site-packages/vt/client.py", line 416, in get_object
    return _make_sync(self.get_object_async(path, *path_args, params=params))
  File "/.../lib/python3.8/site-packages/vt/client.py", line 47, in _make_sync
    return asyncio.get_event_loop().run_until_complete(future)
  File "/usr/local/opt/python@3.8/Frameworks/Python.framework/Versions/3.8/lib/python3.8/asyncio/events.py", line 639, in get_event_loop
    raise RuntimeError('There is no current event loop in thread %r.'

[chexca]

vt-py’s version 0.5.4 has been released, which should fix this issue. Could you please update it and let me know if everything goes as it should?

[chexca]

Your understanding of how an async library should work is not incorrect but a different approach: typically, libraries are meant to be executed synchronously but have the capability of being executed in an asynchronous way. Our approach is the opposite as you could see by reading the code: vt-py is meant to be executed asynchronously, but we also provide a way to use it in a synchronous way.

The issue with Django (and Celery, and CherryPy) is that WSGI applications doesn’t define an event loop by themselves on the threads they spawn when handling web requests, as this is a term introduced by the internal Async IO module, which is a relatively new approach of parallel programming with Python, and those WSGI environments were originally developed under the “old” mindset (old = before Python 3.4).

This is less of a problem if we want to synchronously execute a piece of code that uses Async IO module, as we only have to define an event loop where the code will run, then wait for it to end. With the specific case of vt-py, instead of trying to get the event loop and calling the VT API if it is not present, you can do this:

    try:
        event_loop = get_event_loop()
    except RuntimeError:
        event_loop = new_event_loop()
        set_event_loop(event_loop)

‘new_event_loop’ and ‘set_event_loop’ are internal functions available in asyncio module. This will generate a new process and will set it as the main event loop where vt-py can run tasks. Then you can work with vt-py Client as usual, as it will do the rest of the stuff needed for waiting asynchronous tasks to be completed before continuing executing your code. However, this is a quick hack for you to have a workaround while we release vt-py 0.5.4 (which will do this same thing behind the scenes, so you don’t have to worry about if your WSGI application, whatever it is, has set a main event loop or not).