Pyright

Bug Fix: Changed logic for reportMissingModuleSource diagnostic rule so it isn't reported for stub files.

Enhancement: Added support for typing.OrderedDict, which is a generic alias for collections.OrderedDict.

Enhancement: Added support for new Python extension callback so Pyright extension is notified when pythonPath is modified.

Bug Fix: Fixed bug in docstring trimming code that resulted in some docstrings (those consisting of two lines where the second line was empty) not appearing when hovering over functions.

Bug Fix: Fixed bug in type checker that resulted in incorrect error when creating a generic type alias with a compatible TypeVar as one of the type arguments.

Bug Fix: Fixed bug that caused value expressions for default parameter values in lambdas to be evaluated within the wrong scope resulting in errors if the lambda scope had a same-named symbol.

Bug Fix: Fixed bugs in handling of wildcard imports. First, it was not properly handling the implicit introduction of symbol A in the statement "from .A import *". Second, it was implicitly including submodules as part of the wildcard, and it shouldn't.

Bug Fix: Fixed bug that resulted in incorrect error when using an unpack operator in an argument expression that corresponds to a *varg parameter in the callee.

Bug Fix: Fixed recent regression that caused 'isinstance' check to emit a bad error when 'self.__class__' was passed as a second argument.

Enhancement: Added json schema for mspythonconfig.json (in addition to pyrightconfig.json).

Enhancement: Updated config file watcher logic so it can detect when a new config file is added to a workspace.

Bug Fix: "Find all references" should not return references to a symbol within library code unless that library source file is currently open in the editor.

Bug Fix: Fixed bug in type checker that caused a crash when analyzing an abstract class with a constructor that contained two or more parameters, all of which are unannotated.

Bug Fix: Fixed pyrightconfig.json JSON schema to accept "information" as a valid diagnostic severity setting.

Enhancement: Updated log levels for messages logged by the Pyright service. Some log levels were "info" but should have been "warning" or "error".

Bug Fix: Fixed bug that caused incorrect type evaluation for *args or **kwargs parameters if no type annotation was present. This bug also affected completion suggestions for these parameters.

Bug Fix: Fixed a bug that resulted in Pyright attempting to parse and analyze binaries (native libraries) like ".pyd" and ".so" files.

Bug Fix: Fixed bug in argument/parameter matching when an unpack operator is used in the argument and the parameter is a *varg type.

Enhancement: Renamed setting "pyright.useLibraryCodeForTypes" to "python.analysis.useLibraryCodeForTypes" for compatibility with Pylance. The older setting name is still supported but will be removed in the future.

Enhancement: Added code to handle the case where a class is assigned to a type described by a callable protocol object. In this case, the class constructor's signature should be compared against the `__call__` signature defined in the protocol.

Bug Fix: Fixed bug in import resolver that caused imports that referred to local namespace packages not to resolve.

Bug Fix: Fixed bug that caused enum names that were not uppercase to be handled incorrectly.

Bug Fix: Fixed bug that caused incorrect type analysis when a package `__init__.py` imported and re-exported a symbol that matched the submodule it was being imported from, e.g. `from .foo import foo`.

Bug Fix: Fixed bug in type analyzer where default value expressions for lambda parameters were not being evaluated. This meant that errors related to these expressions were not reported, and symbols referenced within them were marked as unreferenced.

Bug Fix: Fixed bug in parser where it was emitting a spurious error about function return type annotations being a tuple when it was simply enclosed in parentheses.

Bug Fix: Fixed a bug that caused completion suggestions not to work for the LHS of a member access expression (e.g. the "a" in "a.b").

Bug Fix: Fixed diagnostic message for "partially unknown" types (used in strict mode). It was incorrectly using the "unknown" message rather than "partially unknown", which could lead to confusion.

Enhancement: Changed type printing logic to emit "Unknown" annotations when in strict mode to make it clearer to the user which part of the type is unknown.

Bug Fix: Fixed bug that caused extension to report empty diagnostics for all tracked files when unnecessary (in particular, when the diagnostic mode is set to openFilesOnly and the file is not open).

Enhancement: Added partial support for mypy-supported variant of " type: ignore" comment where specific error codes in square brackets after the "ignore". Pyright doesn't honor the specific error codes but now treats it as a normal type: ignore" comment.

Bug Fix: Fixed bug that caused the arguments of a call to remain unanalyzed if the LHS of the call was not callable. This resulted in omitted errors and spurious unreferenced symbols.

Bug Fix: Changed diagnostic for second argument to "Enum" call to be dependent on the reportGenalTypeIssues diagnostic rule rather than unconditional.

Bug Fix: Fixed recent regression relating to "isinstance" type narrowing when the type of the target is a constrained TypeVar.

Bug Fix: Fixed bug in the handling of the NewType function introduced in PEP 484. The previous code was not synthesizing a constructor (`__init__` method) as specified in the PEP.

Enhancement: Changed fallback mechanism for detecting the configured python interpreter to use the shell command "python3" first and then "python" if that fails. This is preferable on Linux and MacOS because "python" typically points to a Python 2.7 interpreter.

Enhancement: Added parser error for relative imports of the form "import .abc". This is treated as a syntax error by the Python interpreter and should be flagged as such.

Bug Fix: Fixed bug with "from . import a" form of import. Diagnostic was not logged when "a" could not be resolved.

Bug Fix: Fixed escaping of literal strings when printing Literal string types.

Enhancement: Improved completion suggestions related to member access expressions (e.g. obj.method) by binding the method to the object when appropriate.

Enhancement: When hovering over class name that is used in a constructor call, display the `__init__` method signature rather than the class.

Bug Fix: Fixed recent regression in unreachable code reporting at the module level.

Bug Fix: Removed error message for unexpected dynamic argument types to `type` initializer.

Bug Fix: Fixed a bug in the code that validates an exception type in an "except" clause. It was not properly handling the case where the type of the exception was specified as a Type[X] object.

Bug Fix: Reverted part of a previous change where constrained type vars were specialized as a union of the constrained types. Changed logic to use first constrained type only.

Bug Fix: Fixed bug in logic that detects assignment compatibility for function types. It wasn't properly handling generic parameters, including synthesized TypeVar types used for "self" parameters.

Bug Fix: Added diagnostic for TypeVar or generic class with type args being used as a second argument for isinstance or issubclass. These will raise a TypeError exception at runtime.

Enhancement: Changed Pyright import resolution order to match that described in PEP 561. In particular, stubs in stubPath are now searched prior to user code, and third-party typeshed stubs are searched only after installed packages are searched for stub packages and inline stubs. There is one place where Pyright's import resolution still differs from a strict interpretation of PEP 561: it searches stdlib typeshed stubs first (unless typeshedPath is defined, in which case it searches there). This is more consistent with the way the Python interpreter resolves stdlib types.

Bug Fix: Fixed bug in handling of constructor that uses a specialized class (e.g. `MyClass[int]()`). The previous code was inappropriate overriding the provided type arguments as part of bidirectional inference logic.

Bug Fix: Fixed bug that caused spurious errors when assigning a specialized object/class to a variable whose type is a specialized base class of the specialized object/class.

New Feature: Added document highlight provider. When you click on a symbol in the editor, all other symbols within the same file that have the same name and have the same semantic meaning are also highlighted.

Enhancement: If reportGeneralTypeIssues rule is disabled, don't replace assigned type with declared type because it will lead to additional errors that will confuse users.

Enhancement: Added type narrowing support for "in" operator when RHS is a specialized list, set, frozenset, or deque.

Enhancement: Added logic to validate that RHS operand used with "in" and "not in" operators support the `__contains__` magic method.

Bug Fix: Fixed bug where "field" initialization of dataclass member didn't take into account "default" or "default_factory" parameters when determining whether the field had a default value.

Bug Fix: Added code to deal with the special case where a method declared with a "def" statement is later overwritten with a callable instance variable.

Bug Fix: Fixed bug whereby a TypeVar type was not treated the same when it was alone versus within a union leading to some subtle differences in error reporting. Also changed specialization of constrained TypeVars to be a union of constrained types rather than Unknown if the TypeVar is not used as a type argument.

Bug Fix: Fixed bug in diagnostic message for constrained TypeVar type mismatch. The wrong type was being printed leading to confusing errors.

Bug Fix: Fixed a bug that caused incorrect linearization of classes during MRO calculation.

Bug Fix: Fixed bug in synthesized version of `get` method for `TypedDict` class. It should provide an overload that allows for any str key and return an "Unknown" type.

Bug Fix: Fixed regression in completion provider when retrieving suggestions for "self.". Added test to cover this case.

Enhancement: Changed "x is not iterable" diagnostic to be part of the "reportGeneralTypeIssues" rule so it doesn't get reported if typeCheckingMode is "off".

Bug Fix: Fixed bug that caused incorrect behavior when a symbol was imported multiple times in the same file.

Bug Fix: Fixed bug that caused Callable instance variables to be treated as though they needed to be "bound" to the object at the time they were accessed. This resulted in spurious errors about parameter count because an implicit "self" parameter was assumed.

Enhancement: Improved type analysis performance by 5-10% on typical code and by significantly more on certain code sequences that involve many if statements within a loop. This optimization uses code flow caching to determine when incomplete types (those that haven't been fully resolved) are potentially stale.

Bug Fix: Fixed recent regression related to imports of the form "from .x import y" within an `__init__.py(i)` file.

Enhancement: Changed type analyzer to use module-level `__getattr__` for types only if the file is a stub.

Enhancement: Added code to prevent "variable possibly unbound" error from propagating to other variables. It should be reported only once.

Enhancement: Switched "pyright.typeCheckingMode" to "python.analysis.typeCheckingMode" for compatibility with Pylance.

Enhancement: Moved a few parameter-related diagnostics to the "reportGeneralTypeIssues" diagnostic rule rather than being unconditional errors.

Bug Fix: Fixed bug that resulted in incorrect type inference for a member variable that is not assigned within a class but is assigned within an ancestor class.

Enhancement: Added type narrowing support for "is" and "is not" operator where RHS is an enum literal value.