Conditional import based on python version check doesn't work

[Dr-Irv]

Describe the bug
If you have a conditional import based on the python version, pyright still says the import is missing.

This comes up in pandas-stubs where we have some dependent libraries that are not ready for python 3.11 yet, so we want to conditionally import them.

To Reproduce

import sys

PY310 = sys.version_info >= (3, 10)
reveal_type(PY310)

if not PY310:
    import foobar

pyright reports:

pyvertst.py:7:12 - error: Import "foobar" could not be resolved (reportMissingImports)
pyvertst.py:4:13 - information: Type of "PY310" is "Literal[True]"

Expected behavior
pyright should not try to do the import based on the version test.

VS Code extension or command-line
command line pyright version 1.1.278
Try the above in a python 3.10 environment.

[erictraut]

This is expected. You cannot use a variable for the conditional. You need to use the sys.version_info >= (3, 10) directly in the if statement.

Pyright (and other type checkers) look for very specific conditional expression forms in the first stage of analysis (the "binder"), which occurs long before the types of any symbols are evaluated.

PEP 484 indicates "type checkers are expected to understand simple version and platform checks" and then gives a few examples. Pyright and all other type checkers support these expression forms.

You'll need to change the above code to the following for it to work in pyright (and other type checkers):

import sys

if not sys.version_info < (3, 10):
    import foobar