report warn-redundant-annotation (similar to warn-redundant-cast)
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 20.6k
- Forks
- 3.3k
- PR merge metrics
- PR metrics pending
Description
Feature
I frequently see people new to python typing over typing everything -- think x: int = function_that_obviously_returns_int(). I would like this to be an optional error
Pitch
in most cases it's somewhat harmless -- it adds a little bit of extra reading and occasionally a slight amount of runtime overhead. but in some cases it makes refactoring and typing a burdensome task (especially if the manual annotation was wrong or subtly wrong, or slightly incorrect, or correct at the time but inconsistent after a refactor).
I would like to suggest a new feature which allows mypy to report this as an optional diagnostic when someone adds an inline annotation which is unnecessary (when mypy already knows the rvalue's type given its expression).
I initially attempted to write something like this as a mypy plugin but I don't think the plugin interface actually gives enough interface points to make this diagnostic possible so I instead looked at how much work it would be to implement this in mypy directly. at a first pass it seems to be not that difficult!
diff --git a/mypy/checker.py b/mypy/checker.py
index 3d0f40283..9229c197c 100644
--- a/mypy/checker.py
+++ b/mypy/checker.py
@@ -2980,6 +2980,9 @@ class TypeChecker(NodeVisitor[None], CheckerPluginInterface):
Handle all kinds of assignment statements (simple, indexed, multiple).
"""
+ if self.expr_checker.accept(s.rvalue) == s.type:
+ self.msg.redundant_annotation(s.type, s.rvalue)
+
# Avoid type checking type aliases in stubs to avoid false
# positives about modern type syntax available in stubs such
# as X | Y.
diff --git a/mypy/errorcodes.py b/mypy/errorcodes.py
index 8f650aa30..5eeb7139c 100644
--- a/mypy/errorcodes.py
+++ b/mypy/errorcodes.py
@@ -164,6 +164,9 @@ NO_UNTYPED_CALL: Final = ErrorCode(
"Disallow calling functions without type annotations from annotated functions",
"General",
)
+REDUNDANT_ANNOTATION: Final = ErrorCode(
+ "redundant-annotation", "Check that the annotation is necessary or can be omitted", "General"
+)
REDUNDANT_CAST: Final = ErrorCode(
"redundant-cast", "Check that cast changes type of expression", "General"
)
diff --git a/mypy/messages.py b/mypy/messages.py
index b63310825..8e9648ed0 100644
--- a/mypy/messages.py
+++ b/mypy/messages.py
@@ -1768,6 +1768,13 @@ class MessageBuilder:
f'Cannot instantiate type "Type[{format_type_bare(item, self.options)}]"', context
)
+ def redundant_annotation(self, typ: Type, context: Context) -> None:
+ self.fail(
+ f"Annotation {format_type(typ, self.options)} is redundant (inferred type is the same)",
+ context,
+ code=codes.REDUNDANT_ANNOTATION,
+ )
+
def redundant_cast(self, typ: Type, context: Context) -> None:
self.fail(
f"Redundant cast to {format_type(typ, self.options)}",
with an example file like this:
def f() -> int: ...
def g():
x: int = f() # stop doing this please!
...
my hacked version of mypy produces the desired diagnostic (though I didn't do the part where it's disabled by default and opted into but I just wanted to show a proof of concept!):
$ rm -rf .mypy_cache/ && mypy t.py --show-traceback
t.py:13: error: Annotation "int" is redundant (inferred type is the same) [redundant-annotation]
Found 1 error in 1 file (checked 1 source file)
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start in mypy/checker.py at TypeChecker.visit_assignment_stmt, then read mypy/errorcodes.py and mypy/messages.py alongside the proposed diff. Use the example file with f() and g() as the initial reproduction. Done means mypy can report unnecessary inline annotations through an optional diagnostic while remaining disabled by default.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- compilers
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Mostly clear
- Newbie friendliness
- 30/100