Circular import when used in AWS lambda

[ytao-canva]

I am encountering a weird error with the package in AWS Lambda environment:

2025-01-20T04:50:06.284Z     File "/var/runtime/awslambdaric/lambda_runtime_client.py", line 68, in post_init_error
2025-01-20T04:50:06.284Z     import http
2025-01-20T04:50:06.284Z     File "/var/task/external/pypi_aiohttp/site-packages/aiohttp/http.py", line 2, in <module>
2025-01-20T04:50:06.284Z     from http import HTTPStatus
2025-01-20T04:50:06.284Z     ImportError: cannot import name 'HTTPStatus' from partially initialized module 'http' (most likely due to a circular import) (/var/task/external/pypi_aiohttp/site-packages/aiohttp/http.py)

Anyone encountered the same and how they get around this, if possible?

[Dreamsorcerer]

You appear to be importing http while presumably in the aiohttp/ directory or something (I'm not too familiar with Lambda env), so it's trying to import the aiohttp.http submodule as a top-level module, when you were probably trying to import the builtin http module (https://docs.python.org/3/library/http.html).

[ytao-canva]

Yeah I presume that was what happened - should we avoid the name http maybe?

[Dreamsorcerer]

No, you should avoid having the aiohttp/ directory on your path. The module is aiohttp.http, it shouldn't cause any conflicts because you shouldn't be looking inside aiohttp for your module imports.

[ytao-canva]

I don't think I have a control over that as it is either controlled by the package I imported or AWS lambda environment. Thanks for the help anyway.

[Dreamsorcerer]

I'm not familiar with lambda so much, but if you literally just install aiohttp and see that, you need to file a support request, there's no way the path should include the aiohttp/ directory by default.

[webknjaz]

@ytao-canva you can always hack sys.path very early in your process entry point, before anything else gets imported. Or you could probably nest the directory structure somehow. But yeah, it's a question to AWS, not to us.