Store the type for assignment expr targets - #9479

cc @gvanrossum since you brought this ticket to my attention. Someone should review to see if this is the right solution. At the very least it seems to fix the example presented by the OP.

This is a good start! From the CI failure we learn that there's exactly one test that fails: testWalrusPartialTypes (at the end of test-data/unit/check-python38.test). This tests that the type of a variable initialized with an empty list is guessed if there's a following operation that appends or inserts into the list. (Same as when the initialization occurs in a regular assignment.) Apparently this code is disturbed by your small change. Maybe you can investigate a bit more? (I can give you more tips if you want to.)

Okay, this is green now for both the testWalrusPartialTypes test plus the new testWalrusExpr test that replicates the OP's issue.

Thank you so much! I'm swamped but will review over the weekend (unless @hauntsaninja beats me to it :-).

Thanks for fixing this! I was a little bit curious about the has_uninhabited_component... it felt a little ad hoc.

If you're willing to accept a small suggestion, I think it might be nice to fix by moving self.chk.store_type(e.target, value) over here, in visit_assignment_expr:

diff --git a/mypy/checkexpr.py b/mypy/checkexpr.py
index 0cc2f13e..2224d5e1 100644
--- a/mypy/checkexpr.py
+++ b/mypy/checkexpr.py
@@ -2815,6 +2815,7 @@ class ExpressionChecker(ExpressionVisitor[Type]):
         value = self.accept(e.value)
         self.chk.check_assignment(e.target, e.value)
         self.chk.check_final(e)
+        self.chk.store_type(e.target, value)
         self.find_partial_type_ref_fast_path(e.target)
         return value

It looks like find_partial_type_ref_fast_path makes its own call to self.chk.store_type, which is what we overwrote here before we added the has_uninhabited_component check, hence previous test failure.

Sorry I didn't wait for you @hauntsaninja.

Not at all, thank you for being proactive!