Errors are shown despite files being in "exclude" in pyrightconfig.json

[bfelbo]

Describe the bug
I need to import a local version of psycopg2, which throws a lot of pyright errors (despite working fine). I want to ignore these errors and have created the following pyrightconfig.json in the root directory of the codebase:

{
    "reportMissingImports": false,
    "include": ["./*.py"],
    "exclude": ["./psycopg2/"]
}

However, the errors still show when I open any of the python files that imports the psycopg2 library. I've also tried "exclude": ["./psycopg2/*"] and "exclude": ["./psycopg2/**"] just in case that helped.

Expected behavior
Errors from analyzing files in "exclude" should be suppressed.

VS Code extension or command-line
VS Code Extension, newest version.

[erictraut]

The "exclude" section won't prevent pyright from analyzing files — or from reporting errors in those files — if they are referenced (imported by) files that are in the "include" section.

If you want to ignore any errors in this subdirectory, you should place it in an "ignore" section rather than "exclude".

[bfelbo]

Thanks. If I put it in "ignore", then those errors go away. However, I still get 'psycopg2' is unbound errors for that folder when viewing a file in root using that library. How do I make those be ignored?

[erictraut]

I'm not sure I understand. You can choose to ignore errors in any individual source file or any directory that you want. Of course, a type checker has no value if you ignore all errors, so it's better to address the errors rather than ignore them. For imported libraries, the best approach is to find (or create) a type stub file that defines the public interface for that library. I don't have enough context to tell you why you're seeing a "symbol is unbound" error.

[bfelbo]

I'm trying to use pyright for type checking, but do not have the time needed to find/create type stub files for imported libraries.

For instance, I would like to be able to use a file my_script.py that imports some_old_library. The some_old_library will not have a type stub file and may also include C bindings or similar that cause pyright to become unhappy. I can ignore the errors thrown by some_old_library by including it in "ignore" in pyrightconfig.json. However, I don't understand how to get pyright to ignore any errors inside my_script.py relating to some_old_library. I hope this makes things clearer.

Pyright is still really useful even if I have to disable it for some specific libraries and any lines in my script that interfaces with those libraries.

[erictraut]

If these are external imports, pyright will (by default) ignore the missing import and assume that all symbols associated with that import are typed as "Any". You can use the "reportMissingImports" configuration setting to change this default behavior.

It sounds like your situation is different though. If I understand you correctly, these are internal imports, and you want pyright to ignore them and act as though it couldn't resolve those imports even though it could.

Depending on the directory layout of your sources and the way you're importing, you might be able to work around this by specifying an "executionEnvironments" entry in the config file. Within each execution environment entry, you can specify a "root" and an "extraPaths" array. Pyright uses these paths to search for imports. By not including the paths where some_old_library is located, pyright will be unable to resolve the import.

The best solution is to create minimal type stub files for these libraries. It really doesn't take much time to implement them. I don't know how many classes or APIs you're dealing with, but it should take only a few minutes of time for each class. It's worth the investment, IMO.

[bfelbo]

Okay, thank you. I'll try that then :)