Automatically backoff when rate limited

[iloveitaly]

Enhancement description

Add option to automatically use a backoff when you've hit a rate limit.

The problem it solves

Intermittent failures due to rate limiting.

Alternatives

Re-running the script.

Use case / screenshots

I've hit 429s on this this project: https://github.com/iloveitaly/todoist-scheduler/commits/master

Additional information

[iloveitaly]

Ran into this again, would be great if this was built in.

[iloveitaly]

Here's how I did it:

# backoff 5xx errors
def patch_todoist_api():
    import todoist_api_python.http_requests
    import backoff
    import requests

    patch_targets = ["delete", "get", "json", "post"]
    for target in patch_targets:
        original_function = getattr(todoist_api_python.http_requests, target)

        setattr(
            todoist_api_python.http_requests,
            f"original_{target}",
            original_function,
        )

        patched_function = backoff.on_exception(
            backoff.expo, requests.exceptions.HTTPError
        )(original_function)

        setattr(
            todoist_api_python.http_requests,
            target,
            patched_function,
        )

[goncalossilva]

Thanks for the report, and sorry for the (very) slow turnaround time here!

You can pass a session when initializing the Todoist API. Which means you can create an HTTPAdapter with a retry strategy, and mount it on the session.

It would be great if the SDK could handle this automatically, but since rate limiting incurs additional wait time (often, much longer than a typical network request), there is no way to handle this transparently given the abnormal wait times. I think I'd rather SDK users have tools (as described above) to be in control of and customize this if they need to.