Kornel

This commit performs a significant **refactoring** by **replacing `HashSet` with a custom `DenseHashSet`** (a dense bitindex) for storing various match IDs throughout the system. This change primarily impacts the **`selectors_vm`** module, where `AstNode`, `ExecutionCtx`, and `MatchInfo` are updated to use this more memory-efficient representation for selector matches. Additionally, the **`rewriter`** module's `ElementDescriptor` and `handlers_dispatcher` are adjusted to utilize `DenseHashSet` for content handler IDs. The nature of this work is to **improve memory efficiency** and optimize the storage and processing of matched identifiers, leading to a more compact and potentially faster execution.

This commit introduces a significant **refactoring** of the **selector matching mechanism**, replacing the generic `MatchPayload` with a concrete `MatchId` type, which serves as a simple index. This change primarily affects the **`selectors_vm`** module, including its AST, compiler, program structure, and execution stack, along with the **`rewriter`** module's handler dispatching and element descriptor management. The **refactoring** aims to simplify the internal representation and handling of matched selectors. By standardizing on an indexed `MatchId`, the system benefits from improved clarity, type safety, and potentially enhanced performance for identifying and dispatching content handlers.

This commit **adds a new test** to the `src/selectors_vm/mod.rs` file, specifically targeting the **`selectors_vm` module**. The test ensures that the **`match handler`** is called exactly once for each matched payload, even when the VM navigates multiple instruction paths. This **enhances the quality assurance** and **correctness** of the `selectors_vm`'s matching logic, preventing potential issues with duplicate or missed handler invocations.

This commit performs a significant **refactoring** within the **`selectors_vm` module** to alter its match handling logic. It **defers the reporting of matched payloads** by introducing a new **flush mechanism**, moving away from eager match processing. This required extensive modifications across various execution paths, including `RecoveryPointHandler` and functions like `exec_without_attrs` and `try_exec_jumps_without_attrs`, to integrate the new deferred reporting. The change aims to decouple the matching event from its immediate reporting, potentially enhancing performance or flexibility in how matches are processed within the VM.

This commit **enhances the project's testing infrastructure** by integrating `cargo check` commands for both the **C API** and **JS API** directories into the `scripts/test.sh` script. This **maintenance improvement** ensures that these critical components are now explicitly type-checked and validated as part of the standard test suite. The change helps to **proactively identify compilation and type-related issues** in these language bindings, improving overall code quality and stability.

This commit **optimizes attribute access** within the **rewritable units** by preventing unnecessary materialization of attributes during reading. It introduces new, efficient `get_attribute` and `has_attribute` methods directly on the `StartTag` and `attributes` structures for streamlined lookup. Existing attribute access methods in `Element` are refactored to delegate to these new, optimized implementations, improving performance. Additionally, a centralized `eq_case_insensitive` utility function is added to the `base` module and adopted by the `tree_builder_simulator` for consistent comparisons. This **refactoring and feature addition** significantly enhances the **efficiency of attribute handling** during parsing and rewriting operations.

This commit introduces a **performance refactoring** to the **`attributes` module** within `rewritable_units/tokens`, specifically optimizing how attributes are searched and manipulated. It **eliminates unnecessary string allocations** in functions like `set_attribute` and `remove_attribute` by removing a `try_from` helper and implementing direct case-insensitive comparisons. This change significantly reduces memory overhead and improves the efficiency of attribute handling, particularly when setting or removing attributes in `src/rewritable_units/tokens/attributes.rs`.

This commit **refactors** the handler removal logic within the **`rewriter` module's `handlers_dispatcher` component**. The change specifically improves the **efficiency and correctness** of processing and removing "tail handlers" by leveraging the `drain` method. The responsible function has been renamed to `do_for_each_active_and_remove_tail` to better reflect its updated behavior. This **maintenance refactoring** ensures more robust and performant handler management within the rewriter's dispatching mechanism.

This commit introduces a **new test case**, `many_end_tags`, to `src/rewritable_units/element.rs` to specifically verify the correct behavior of **multiple end tag handlers**. This **enhances test coverage** for the `rewritable_units` module, ensuring the robustness of element parsing and rewriting logic when encountering complex sequences of closing tags. This **maintenance** work helps guarantee the stability and correctness of the system's core parsing capabilities.

This commit introduces a **new capability** by adding an `on_end_tag` convenience method to the `Element` struct within the `rewritable_units` module. This method simplifies the process of adding end tag handlers, offering improved error handling for these operations. To further enhance developer experience, a new `end_tag!` macro is introduced in `src/rewriter/settings.rs` to provide a type hint for the closure argument of the `on_end_tag` method. This enhancement streamlines the definition of end tag processing logic, making the API more user-friendly and robust.

This commit introduces a **refactoring** to optimize memory usage within the **rewriter** subsystem. It replaces the existing handler indexing mechanism with a new, smaller `Locator` struct, leveraging `NonZero<u32>` for improved efficiency. This change primarily affects the **`handlers_dispatcher`** module, where handler indexing logic is updated, and the **`rewrite_controller`** module, which now uses the optimized `Locator` within its `ElementDescriptor` struct. The overall goal is to reduce the memory footprint associated with tracking content handlers, particularly within `ElementDescriptor` instances.

This commit introduces a significant **refactoring** to optimize string handling when creating **HTML local names** and attribute names. It modifies the `from_str_without_replacements` function in `src/base/bytes.rs` and `src/html/local_name.rs` to accept `impl Into<Cow>`, allowing for more flexible and efficient reuse of string data and reducing unnecessary allocations. Consequently, the **selector VM** and **rewritable units** modules are updated to align with these new signatures, ensuring that attribute names can now directly accept `String` ownership. This change improves overall performance and memory efficiency by minimizing string cloning and copying throughout the parsing and rewriting pipeline.

This commit introduces a **bug fix** to the **rewriter** component, specifically within the `src/rewriter/rewrite_controller.rs` module. It adds **explicit error handling** and a **debug assertion** to the `handle_start_tag_with_aux_info` function. This change addresses a scenario where `selector_matching_vm` could unexpectedly be `None` during start tag processing, preventing potential runtime panics. The fix improves the overall **robustness** of the rewriter's tag handling logic by ensuring the system correctly identifies and handles the absence of an expected VM, leading to more stable operation.

This commit introduces a **refactoring** within the **`selectors_vm` module** to enhance code clarity and consistency. Specifically, it removes the redundant `ns` parameter from the `exec_after_immediate_aux_info_request` function in `src/selectors_vm/mod.rs`. The function now consistently retrieves the namespace from `ctx.ns`, eliminating an unnecessary argument and simplifying the function signature. This change is a purely internal code quality improvement, with no functional impact on the system's behavior or external APIs.

This commit introduces a **bug fix** to the **CSS selector parser**, specifically within the `selectors_vm` module, by explicitly **rejecting CSS4 negated combinators** when found inside `:not()` pseudo-classes. The **selector validation logic** in `src/selectors_vm/parser.rs` was updated, introducing an `inside_any_negation` flag to functions like `validate_component` and `validate_selectors` to enforce this restriction. This ensures the parser correctly identifies and disallows unsupported advanced selector patterns, preventing potential parsing errors or misinterpretations. New **test cases** were added to `src/selectors_vm/ast.rs` to verify the proper rejection of these invalid selector constructs.

This commit performs a **dependency update** across the project, primarily affecting the **core parsing and selection logic** and the **JavaScript API**. It bumps `selectors`, `html5ever`, `markup5ever_rcdom`, and `rand` in the main `Cargo.toml`, and updates the `lol-html` crate to version 2.8.0 within `js-api/Cargo.toml`. This **maintenance** task ensures the project leverages the latest features and bug fixes from its underlying libraries. A necessary adjustment was made in `tests/harness/input.rs` to accommodate an API change in the updated `rand` crate, switching from `Rng` to `RngExt` for random number generation.

This commit introduces **support for nested `:not()` selectors** within the `selectors_vm` module, significantly enhancing its parsing capabilities. It achieves this by implementing a new method in `src/selectors_vm/ast.rs` to recursively add selector components, correctly processing complex negation structures. A minor **refactoring** in `src/selectors_vm/parser.rs` improves error handling for unsupported combinators, and **new test cases** in `src/selectors_vm/mod.rs` validate the correct matching behavior. This **new capability** expands the expressiveness of the selector engine, allowing for more sophisticated and precise element selection.

This commit officially marks the **release of version 2.7.2**, primarily focusing on **improving application stability** and user experience. It introduces a significant **enhancement to error handling** across the project by replacing previous panicking assertions with gracefully reported errors, thereby **reducing unexpected crashes**. This **maintenance release** updates the project version from 2.7.1 to 2.7.2 in `Cargo.toml` and documents these critical changes, including the improved error reporting, within `CHANGELOG.md`. The overall impact is a more robust and reliable application due to better handling of exceptional conditions.

This commit primarily focuses on **code style and maintenance improvements** across various modules, largely driven by Clippy linter suggestions. It **refactors** the initialization of the `LAST_ERROR` static variable within the **C-API error handling** module (`c-api/src/errors.rs`) to use a `const` block for `RefCell::new(None)`. Additionally, a `#[allow(clippy::too_many_arguments)]` attribute is added to the `lol_html_rewriter_build_inner` function in the **C-API rewriter** module (`c-api/src/rewriter.rs`) to suppress a specific linter warning. Finally, a minor **style adjustment** is made in the **memory arena** module (`src/memory/arena.rs`), changing an unused closure argument from `_` to `()` in the `Arena::new` method. These changes are non-functional and aim to improve code quality and silence linter warnings.

This commit introduces a significant **refactoring** to the **tokenization layer**, primarily focusing on **simplifying attribute serialization**. The `Attributes` struct in `src/rewritable_units/tokens/attributes.rs` has been refactored by removing its `Deref` implementation and adding explicit helper methods like `is_empty` and `to_slice` for better control and clarity. Consequently, the `StartTag` struct in `src/rewritable_units/tokens/start_tag.rs` is updated to utilize these new methods and align its own serialization process with the streamlined `Attributes` serialization. This internal improvement enhances the maintainability and efficiency of how attributes are managed and serialized within the **rewriting units**.