feat(extraction): add Objective-C language support (#165)
Adds tree-sitter-objc extractor for `.m`/`.mm` files and `.h` files that content-sniff as Objective-C (`@interface`/`@implementation`/`@protocol`/`@synthesize`). Extraction covers: - `@interface` / `@implementation` (deduplicated into a single class node) - `@protocol` (as `protocol` nodes via new `interfaceKind` config) - Methods with full multi-part selectors (`doThing:with:`, not just `doThing`), including `+`/`-` static distinction - `@property` declarations - Inheritance (`extends`) and protocol conformance (`implements`) - C-style `function_definition` and `#import` (both `<system>` and `"local"` forms) - Call edges from both `call_expression` and `message_expression`, with `self`/`super` skipped on qualified callee names Two new generic hooks on `LanguageExtractor` (`resolveName`, `extractPropertyName`) handle the cases where the default name walk doesn't fit; usable by future languages with similar shape. Import resolver tries `.h`, `.m`, `.mm` for `objc` imports. Validated on AFNetworking (84 files, 100% file coverage), RestKit (282 files, 99.6%), and Texture (926 files, 100%, heavy `.mm` content) — multi-keyword selectors preserved up to 7 parts, no parse failures on ObjC++. Known limitations (disclosed in README): - Categories produce duplicate class nodes (one per category file) - Chained/nested message sends record only the innermost method - `[Class alloc]` patterns don't emit `instantiates` edges - `@protocol Foo <Bar>` refinement lists not yet wired to `implements` - Heavy C++ in `.mm` files may parse incompletely under the ObjC grammar
0
0x1306a94 committed
61153f96eebb6c49f36b546031873598b16b71a0
Parent: b48170e
Committed by GitHub <noreply@github.com>
on 5/26/2026, 5:31:43 AM