Document how import chains are affected by sys.setrecursionlimit

[seddonym]

While working on a very large Python codebase, I ran into some undocumented behaviour.

A chain of imports (i.e. one module imports another, which imports another, which imports another) will cause a RecursionError to be raised if the Python recursion limit is not high enough. (Note, there is no recursion involved.)

Longer chains of imports can be enabled by increasing the recursion limit. But the limit has no effect beyond a certain point (around 10,000). (I believe this is due to Py_C_RECURSION_LIMIT, mentioned in by Serhiy Storchaka.)

A workaround is available: you can import the modules in a different order.

I've created a repository with more analysis and scripts to reproduce the problem. I also started a Discuss topic on the subject.

From what I understand, this is a complex problem that is unlikely to be fixed. But I think it would be helpful to document this, probably in the docs for sys.setrecursionlimit.

I would be happy to try to do a pull request, assuming the core team agree that it should be added.

[pochmann3]

Import chains are just one case. It's a more general issue, also see these:

#112215

#112282

[seddonym]

Thanks for linking. I've asked on that second issue whether it would make sense to fold this issue into that one.