Pyright won't use typing for a class even when directly typed, ends up as unknown

[C0rn3j]

This is the crux of my issue, I cannot for the love of me get Pyright to understand that the variable is of class Tauon.

The file and codebase in question is here - https://github.com/Taiko2k/Tauon/blob/7d41e99c41bdbc608d50db1e3c89f08a61a46905/src/tauon/t_modules/t_main.py#L40879-L40884

[erictraut]

Can you repro the problem with an minimal, self-contained code sample? Often, in the process of creating an MRE, the underlying cause of the problem will become clear.

[C0rn3j]

I created this branch and started deleting code https://github.com/C0rn3j/Tauon/tree/pyrightrepro

Eventually after deleting enough lines, it starts showing as Tauon.

So it seems that Pyright just refuses to keep going after a certain amount of lines in the function(?), should I report a bug?

Screencast_20250824_071325.webm

EDIT: Got it down to 606 lines, reducing it any more seems to almost always not work, so I suppose that is my minimal repro, I'll report it.

EDIT2: I do get the

	"message": "Code is too complex to analyze; reduce complexity by refactoring into subroutines or reducing conditional code paths",

Which seems to be what causes this, but it's odd it triggers even on this code.

[erictraut]

Ah, then we know what the problem is. The code is too complex.

Keep in mind that pyright is a static type analyzer. Unlike the runtime, which executes only one path through your code based on its inputs, a static type analyzer must consider all possible paths through your code. If the code flow graph is too complex, this becomes prohibitively expensive. Pyright calculates a metric called the cyclomatic complexity for each code flow graph. If the complexity metric exceeds a particular threshold, it will refuse to analyze that graph. Doing so would result in hangs and out-of-memory crashes.

If you want your code to be analyzed by pyright, you'll need to do what is suggested in the diagnostic message — simplify the code. Reduce the number of branching constructs (loops, if statements, try/catch) within a single function by refactoring your code.

[C0rn3j]

I can get the code to work if I I remove a simple logging.info("whatever") in top level of the function, that part does not seem to be related to complexity?

I will see about splitting it, but it still seems that Pyright is doing something odd, complexity metrics aside.