I've configured my VS Code project to use the Pylance extension. I know that Pylance is using Pyright internally. We have a "no-warning" policy. I want to enforce this by running Pyright (in pre-commit and in the CI pipeline). Since Pyright is used internally by Pylance, I figured I should be able to run an equivalent analysis with a standalone Pyright. My problem is that I don't know how to replicate the Pylance analysis in Pyright. They seem to disagree even if I use the same value for the typeCheckingMode parameter (i.e. strict). For this code: from typing import Any a: str = 12345 b: dict[str, Any] = {} c = b.get("a", {}).get("b", []) Pylance only spots the first problem: Expression of type "Literal[12345]" cannot be assigned to declared type "str" "Literal[12345]" is incompatible with "str" However, running Pyright on the command-line, I get an additional error: test.py test.py:3:10 - error: Expression of type "Literal[12345]" cannot be assigned to declared type "str" "Literal[12345]" is incompatible with "str" (reportGeneralTypeIssues) test.py:6:1 - error: Type of "c" is partially unknown Type of "c" is "Any | list[Unknown]" (reportUnknownVariableType) 2 errors, 0 warnings, 0 informations I'm wondering: Am I missing a Pyright parameter? Is it due to a mismatch between my Pyright version and the one embedded in Pylance? Is this a bug in Pylance? But the real question is: how can I validate, outside of VS Code, that we don't have any type error in the code that are reported by Pylance? Continue reading...
