pyright

Enhancement: A rewritten implementation of the code flow and type constraint engine provides better handling of complex code flow like nested loops, try/except/finally statements, etc.

Enhancement: In VS Code extension, changed the logic for how to handle opened files that don't fall within any workspace. Previously, all such files were analyzed in a default workspace using default settings. The new logic uses the current workspace if there is only one. The "default" workspace fallback is still used for multi-root workspaces.

Bug Fix: Fixed bugs in logic that determines when to allow importing a non-stub from a stub file. It's not allowed for third-party files.

Bug Fix: Changed import logic to search for imported symbols in the target module before looking for submodules of the same name.

Bug Fix: Unreachable code is displayed even for files where errors or warnings are suppressed.

Bug Fix: Fixed TypeVar matching logic to avoid adding widened types if the TypeVar was already matched against subclasses of the type that is being added.

Bug Fix: Union types that include Literal[True] and Literal[False] are now collapsed into a 'bool' type.

Bug Fix: Fixed type completions for TypedDict classes when TypedDict is imported from typings (versus typing_extensions).

Bug Fix: Fixed bug that affected proper type evaluation for CTypes like c_uint64.

Bug Fix: Changed server to create service on demand, when a file is opened that doesn't match any existing workspace.

Bug Fix: Fixed bug that caused list[x] and List[x] not to be combined into the same union type.

Bug Fix: Added support for proper type inference for assignments that target lists (e.g. [a, b] = [1, 2]).

Enhancement: Added missing methods and attribute to memoryview class in builtins.pyi.

Enhancement: Command-line version now exits with a new exit code (3) if the pyrightconfig.json is malformed or cannot be read. Previously, it proceeded to perform the analysis with default settings.

Enhancement: Added support for interleaved for and if clauses within a list comprehension statement. Previously, the analyzer assumed that there would be only one if clause and it would be after all for clauses.

Enhancement: the built-in function issubclass is now supported by the analyzer in the same way as isinstance. It can be used as a type constraint in conditional code flow statements (if, while, etc.), and the reportUnnecessaryIsInstance setting applies to it.

Enhancement: Added support for the @abc.abstractproperty function decorator.

Bug Fix: Fixed a reported bug relating to the handling of the @overload function decorator. This was due to a recent regression.

Enhancement: Added code to interpret the python.path VS Code setting in the same manner as the Python VS Code extension. In particular, it now handles the special-case value of "python" to mean "use the default python interpreter" rather than treating it as a relative path.

Enhancement: Added special-case handling of __import__ built-in call so it always returns Any type.

Bug Fix: Fixed bug in code that determines whether code is reachable.

Bug Fix: Fixed static expression evaluation for expressions that involve python platform when no platform is specified.

Bug Fix: Fixed bug in alias resolution logic that caused the definition provider not to take the user to the right location in cases where there were multiple alias hops.

Bug Fix: Changed implicit module symbol __path__ to be properly typed as Iterable[str] rather than str.

Bug Fix: Fixed regression introduced in 1.0.77 that caused imports of third-party libraries to generate "not a known member of module" error messages.

This version includes a big change in the way pyright tracks variable declarations. This change should not be visible to users. It was done in preparation for an upcoming change in the analysis engine.

New Feature: Signature help provider now extracts parameter-specific comments and presents them in the help text when typing function arguments.

Enhancement: Added knowledge of __doc__ symbol to all modules.

Bug Fix: Fixed crash that resulted when launching the production VS Code extension in debug mode.

Bug Fix: Fixed recent regression that broke type completions within import statements.

Bug Fix: Fixed issue that caused analysis to fail after repeated edits while in "watch" mode.

This version includes a big change in the way pyright processes import statements. This change should not be visible to users. It was done in preparation for an upcoming change in the type analysis engine.

Bug Fix: Fixed Windows-specific path processing bug.

Bug Fix: Fixed bug that caused "import cycle" errors to persist even after they were corrected.

Enhancement: Added support for bidirectional type inference for assignments where the LHS is an indexed expression.

Bug Fix: Switched to a new file system watching mechanism so "--watch" works correctly on Linux.

Bug Fix: Fixed recent regression that caused errors to be reported incorrectly after editing a dependent source file.

New Feature: Added support for Python 3.8 assignment expressions (aka walrus operator).

Enhancement: Bumped latest stable build of python from 3.7 to 3.8, so the latter is now the default setting for pyright.

Enhancement: Added verbose logging for file system watchers.

Enhancement: Improved robustness of code that invokes the local python interpreter to query for paths. Thanks to Rami Chowdhury for the contribution.

Enhancement: Exposed WindowError from builtins.pyi file.

Enhancement: Added overload definitions for built-in "open" function so it distinguishes between Io[str] and Io[bytes] if the provided mode argument is a string literal.

Bug Fix: Added support for metaclasses when using magic function lookups. This allows classes, for example, to be used in binary operation expressions.

Bug Fix: Fixed recent regression in "Organize Imports" feature that caused incorrect ordering of imports.

Bug Fix: Fixed issue that caused sporadic type checker errors to be reported when editing or opening certain files within a workspace.

New Feature: Added support for Python 3.8 "position-only parameters".

Bug Fix: Enum classes are now properly treated as iterable types.

Bug Fix: Fixed several bugs in typestub generation feature, some of which resulted in crashes or out-of-memory errors.

Enhancement: Changed TypeVar matching to perform type expansion. Removed hack init overload for dict class that was required because we weren't performing type expansion.

Enhancement: Used aggressive bundling and minification to reduce the VS Code download size by over 50%. Shaved 200ms (80%) off its startup time.

Bug Fix: Fixed a case where a Union type alias (e.g. A = Union[T, B[T]]) was not being handled properly.

Bug Fix: Fixed issue where non-type-stub files were imported and analyzed if they were within a directory with an init.pyi file.

Bug Fix: Added code to suppress all analysis and errors of third-party code files.

Bug Fix: Fixed issue where certain types that were already specialized were re-specialized, overwriting the previous type arguments.

New Feature: Added config switch for reportUnknownLambdaType.

Enhancement: Implemented parameter type inference for lambdas (uses bidirectional type inference where possible).

Enhancement: Extended bidirectional type inference for types that are nested within dictionaries and tuples.

Enhancement: Implemented two-pass argument for parameters that require type variable matching. This allows for better type checking of arguments and provides better compliance with PEP 484 standard.

Enhancement: Improved type completion within lambda expressions.

Bug Fix: Fixed bug that caused type checker to mistakenly analyze third-party source files that are imported by third-party stub files. The analyzer now stops when it encounters a third-party source file, even if it's included by a stub file.

Bug Fix: Fixed bug that caused some unioned types to be dropped because the deduping logic was mistakenly detecting duplicates.

Bug Fix: Fixed handling of augmented expressions that are defined by the special "i" forms of operator magic methods.

Bug Fix: Fixed handling of tuple expressions that contain unpack operations.

Bug Fix: Fixed regression introduced in previous release relating to yield statements with no expression.

New Feature: Added support for Python 3.8 ability to use unpack operators within return or yield statements.

New Feature: Added support for Python 3.8 f-string expressions that end in "=".

Bug Fix: Reduced size of pyright extension by 70%+ by deleting large "LSIF" files that have started to be included in some of the dependent modules.

Bug Fix: Fixed analysis of the second (optional) parameter of assert statements.

Bug Fix: Fixed issue that caused Callable type not to be assignable to generic Type[_T] parameter.

Bug Fix: Fixed bug where unpack operation was not allowed in right-hand side of simple assignments.

New Feature: Added support for "# type: import" with a config switch to disable it ("enableTypeIgnoreComments").

Enhancement: Added support for the implicit "class" symbol that is added to all methods.

Bug Fix: Fixed bug in type checker's handling of assignments to indexed expressions. It was not properly validating that the assigned type was appropriate.

Bug Fix: Fixed path parsing bugs that affect Windows platform.

Bug Fix: Fixed bug in handling of enums when using index operator.

Bug Fix: Fixed recent regression (interaction between "expected type" logic for list/set/dict literals and type var matching) that caused the type checker to miss some errors.

Bug Fix: Added support for comparison fallbacks (e.g. if < isn't supported, try its complement >).

Bug Fix: Improved PEP 484 compliance with respect to handling of int, float and complex types.

Enhancement: Improved type completion in cases where an index (subscript) construct is used to the left of a member access construct.

New Feature: Added support for PEP 589 (TypedDict).

Enhancement: Updated all typestub files to the latest from the typeshed repo.

Bug Fix: Fixed issue that caused public member variables with annotated Literal types to not accept a literal assigned value.

Bug FIx: Fixed several bugs in type checker's logic that caused some type violations not to be reported.

Bug Fix: Changed Tuple (with no type arguments) to default to Tuple[...] rather than Tuple[].

Bug Fix: Fixed issue with Literal type arguments. String literals were being parsed as forward-declared types, and some strings generated tokenizer and parser errors.

Bug Fix: Fixed type checker bug where functions and classes were flagged as not being assignable to 'object'.

Enhancement: Added check within type checker to emit an error when code attempts to assign a value to a generic type (e.g. "Dict[str, str] = {}").

Enhancement: Implemented support in type checker for Python's "comparison chaining" feature (e.g. "3 < a < 4").

Enhancement: Augmented static expression evaluation to handle "and" and "or" expressions.

Bug Fix: Fixed a bunch of misspellings in error and warning text.

Bug Fix: Fixed a bug in the auto-complete feature that caused new import symbols to be inserted in the wrong place if the previous symbol (alphabetically) had an "as" clause.

Bug Fix: Fixed a bug in the auto-complete feature that caused bad performance for really large code bases.

Enhancement: Added support for constraining the type of the RHS of an assignment statement based on the expected type of the LHS.

Enhancement: Suppressed return type check for functions whose implementations consist only of "...".

Enhancement: Changed heuristic about whether to strip literal values from inferred class or instance member type. If it's not a constant or private member, I now use the more general type (not the literal).

Bug Fix: Changed tuple type expression evaluation to retain literals rather than strip them.

Bug Fix: Fixed recursion check that caused stack overflow during type evaluation.

Bug Fix: Removed assumption in type stub generation logic that __all__ assignment would appear after all imports in the file.