Maggie Moss

This commit performs a **maintenance update** by incrementing the project's **version number**. Specifically, it updates the `VERSION` constant within the `version.bzl` file from `0.57.0` to `0.58.0`. This change signifies a new development milestone or release cycle for the project, impacting any build systems or external integrations that depend on this version identifier.

This commit **fixes two critical bugs** in the **`pyrefly infer` command** that caused it to produce **invalid Python code** when generating type annotations. It corrects the import insertion logic to ensure new imports are placed *after* `__future__` statements, preventing syntax errors. Additionally, it prevents invalid `Overload[...]` annotation syntax by filtering out `Overload` types during hint formatting, aligning with existing filters for unrepresentable types. This significantly improves the correctness and usability of `pyrefly`'s inferred type annotations.

This commit **updates the configuration** for **Pyrefly's `buck_check` command** to facilitate its rollout within fbcode. It specifically configures `buck_check` to **infer return types only for annotated functions**, disables checking unannotated definitions, and reduces the severity of the `UnusedIgnore` error. This **refactoring** of the type checking settings, primarily in `pyrefly/lib/commands/buck_check.rs`, enables a more controlled and gradual integration of Pyrefly's type inference capabilities across the fbcode monorepo. The changes allow for a smoother adoption by focusing on existing annotations and adjusting error handling.

This commit **extends the `pyrefly suppress --remove-unused` command** to identify and remove **unused Pyre-style suppression comments**, including `# pyre-fixme`, `# pyre-ignore`, and `# pyre: ignore`. This **new capability** is activated when `Tool::Pyre` is enabled in the configuration, updating the comment matching regex and removal logic within the `suppress` module. The `errors` module was updated to correctly collect these unused Pyre comments based on the module's `enabled_ignores` configuration. This **enhances code hygiene** by automatically cleaning up more types of dead suppression comments, improving code clarity and maintainability for projects using Pyre.

This commit **refactors Pyre's type-checking configuration** by replacing the monolithic `untyped-def-behavior` option with two more granular and orthogonal settings: `--check-unannotated-defs` (boolean) and `--infer-return-types` (a 3-value enum). This **enhancement** provides **finer control** over how **unannotated function bodies are type-checked** and when **return types are inferred**, addressing limitations of the previous design which could not express all desired configuration states. The **configuration system** (`crates/pyrefly_config`) is updated to introduce these new fields and handle **backwards compatibility** for the deprecated option, while the **core type-checking logic** (`pyrefly/lib/binding/function.rs`) and various commands are adapted to utilize the new settings. This significantly improves the **flexibility of Pyre's analysis behavior**, allowing users to precisely configure type-checking and inference for Python functions.

This commit **optimizes the performance** of the **`pyrefly infer` command** and related LSP features. It **refactors** the `infer.rs` module to use `Require::Exports` for dependency modules and parallelizes handle processing, significantly speeding up the core inference logic. Additionally, the **LSP `inlay_hints`** functionality is improved by skipping inference for functions where all parameters are already annotated, reducing redundant computations. These targeted **performance optimizations** collectively result in **faster inference times** for users of `pyrefly`.

This commit **refactors** the **`pyrefly infer` command** to resolve imports more efficiently and accurately. It **moves import resolution from a heuristic-based Phase 2 to a type-based Phase 1**, directly collecting `QName` information from inferred types to determine necessary imports. This change **fixes a duplicate import bug**, eliminates an expensive second type-checking pass, and makes import resolution more robust by leveraging precise type information instead of fragile heuristics. The `pyrefly/lib/commands/infer.rs` module is primarily affected, resulting in improved performance and correctness for type annotation generation.

This commit introduces a **feature enhancement** to the **PyreFly static analysis tool**, specifically improving the `infer` command's capabilities. It **forces the `infer` command** to utilize the more comprehensive `CheckAndInferReturnType` strategy for untyped definitions by default, unless explicitly overridden by the user. This change, primarily affecting `pyrefly/lib/commands/infer.rs` and supported by a new configuration function in `crates/pyrefly_config/src/args.rs`, significantly **improves the accuracy and completeness of inferred type annotations**. Consequently, users will now receive **substantially better type inference results** for their Python code, leading to more robust static analysis.

This commit implements a **bug fix** within the **`pyrefly` build system** to prevent incorrect typeshed stubs from being processed. It **broadens the filter in `crates/pyrefly_build/src/source_db/buck_check.rs`** to correctly exclude Pyre-versioned typeshed stubs, including those provided via user overrides, from `pyrefly`'s dependency index. This ensures that `pyrefly` falls back to its own bundled typeshed, **resolving an issue where incorrect stubs caused a massive explosion of type errors** and improving the accuracy of type checking.

This commit introduces a **new capability** to the **`pyrefly`** type checker, enabling explicit control over the reporting of "unused ignore" error codes. It configures and collects `UnusedIgnore` errors within the `pyrefly/lib/commands/buck_check.rs` module, specifically impacting the `compute_errors` function and enhancing the overall error reporting. This **feature enhancement** mimics Pyre's behavior of reporting but filtering these errors by default, which is crucial for simplifying the future implementation of a `remove-unused-ignores` flag without complex flag threading through BXL layers. The change provides more flexible error handling for `pyrefly` users.

This commit performs a **maintenance** task by **bumping the Pyrefly project version** from `0.51.0` to `0.51.1` within the `version.bzl` file. This essential **release preparation** step is specifically undertaken to facilitate a new **PyPi release** of the Pyrefly library. The version increment ensures proper tracking and precedes upcoming, more substantial development work, including a significant PyTorch upgrade.

This commit provides a **bug fix** for the **PyreFly error suppression mechanism**, specifically addressing issues in `pyrefly/lib/error/suppress.rs`. It introduces a **new regular expression** to accurately parse and identify ignore comments, even when other comments are present on the same line. Functions like `update_ignore_comment_with_used_codes` and `remove_unused_ignores_from_serialized` are updated to leverage this improved parsing logic. This ensures more reliable and precise **removal of unused error suppression directives**, preventing incorrect modifications or retention of ignores within the codebase.

This commit introduces a **new feature** to the **`pyrefly` static analysis tool**, enhancing the **`suppress` command** with the ability to remove unused ignore comments. A `remove_unused` flag is added to the command-line arguments in `pyrefly/lib/commands/suppress.rs`, allowing users to specify this cleanup action. The underlying logic for identifying and removing these unused ignores is implemented in `pyrefly/lib/error/suppress.rs`, including a new public function `remove_unused_ignores_from_serialized` for processing serialized errors. This improvement helps maintain cleaner codebases by automatically eliminating obsolete suppression directives, thereby improving the accuracy and relevance of static analysis results.

This commit **refactors** the `pyrefly` library's **error handling system** by renaming the `SuppressableError` struct to `SerializedError` to better reflect its role in serializing various error types. This change, primarily in `pyrefly/lib/error/suppress.rs`, also enhances the `SerializedError` with a `message` field and introduces an `is_unused_ignore` method for improved error introspection. As a result, the `pyrefly/lib/commands/check.rs` and `pyrefly/lib/commands/suppress.rs` modules are updated to leverage the new `SerializedError` type. This ensures `UnusedIgnore` errors are consistently filtered and handled during both error checking and suppression processes, improving clarity and maintainability within the **error reporting and suppression logic**.

This commit introduces a **new `UnusedIgnore` error kind** to the `pyrefly_config` module, setting its default severity to `Ignore`. It enhances the **error reporting system** by adding methods to `pyrefly/lib/state/errors.rs` for collecting and filtering these new errors. The `pyrefly check` command is updated to integrate the display of these unused ignore errors, providing users with more comprehensive feedback. This **feature addition** also includes **documentation** for the new error kind and lays the groundwork for improved flexibility in managing and potentially surfacing unused ignores within the project and future IDE integrations.

This commit **refactors** the **unused suppression removal logic** within the `pyrefly` tool to enhance code sharing and prepare for future integrations. The `remove_unused_ignores` function in `pyrefly/lib/error/suppress.rs` has been significantly rewritten to accept a list of unused errors directly, streamlining its internal identification and removal process. This **refactoring** also updates the `check` command in `pyrefly/lib/commands/check.rs` to reflect these changes, removing an 'all' option and adjusting the call to the refactored function. The primary goal is to enable the system to eventually process JSON blobs of errors for silencing, which is crucial for an upcoming `arc pyre` integration.

This commit **fixes a critical bug** in the `pyrefly` library's **error suppression mechanism** by preventing it from suppressing `UnusedIgnore` errors. The `is_ignored` function in `pyrefly/lib/error/error.rs` now includes a check to avoid suppressing these specific errors, which previously could lead to an infinite loop within the system. Concurrently, the `from_error` function in `pyrefly/lib/error/suppress.rs` is modified to explicitly return `None` when encountering an `UnusedIgnore` error, ensuring it is never suppressed. This change guarantees the **stability and correctness of the error reporting system** by ensuring `UnusedIgnore` errors are always visible and not hidden by the suppression logic itself.

This commit introduces a **new feature** to the **`pyrefly` tool's `suppress` command**, enabling it to consume error data directly from a JSON file. A new `--json` argument is added to the `suppress` command, allowing users to specify a JSON file containing errors, thereby bypassing the need to run a full type check. To facilitate this, the `SuppressableError` struct within `pyrefly/lib/error/suppress.rs` has been updated to derive `Deserialize`, making it compatible with JSON input. This enhancement provides a more flexible and potentially faster method for error suppression, particularly useful for integrating with external error reporting or pre-analyzed error lists.

This commit introduces a **new `SuppressableError` abstraction** to unify error handling within the **`pyrefly` error suppression system**. It **refactors** the core `pyrefly/lib/error/suppress.rs` module, adapting functions like `dedup_errors` and `suppress_errors` to operate on this common error representation. Both the `pyrefly check` command (`pyrefly/lib/commands/check.rs`) and the `pyrefly suppress` command (`pyrefly/lib/commands/suppress.rs`) are updated to convert their errors into `SuppressableError` objects. This **refactoring** enables **reuse of existing suppression logic** across different error types, enhancing consistency and maintainability.

This commit **implements the core logic** for the new **`pyrefly suppress` command**, enabling it to **run checks and automatically apply suppressions** to identified errors. This **new feature** introduces a critical capability within the **`pyrefly` tool's command-line interface**, specifically affecting the `pyrefly/lib/commands/suppress.rs` module. By allowing the command to resolve files, execute checks, and then apply suppressions, it significantly streamlines the process of managing code quality issues. This enhancement provides users with a powerful mechanism to programmatically handle diagnostics, improving developer workflow and code maintenance.