fix(php): parse class constants named after reserved keywords

pyproject.toml

@@ -37,7 +37,8 @@ dependencies = [
     "tree-sitter-c-sharp>=0.23,<0.25",
     "tree-sitter-kotlin>=1.0,<2.0",
     "tree-sitter-scala>=0.23,<0.27",
-    "tree-sitter-php>=0.23,<0.25",
+    # Fork commit until tree-sitter/tree-sitter-php#307 ships: keyword-named class constants (#17).
+    "tree-sitter-php @ git+https://github.com/alexwrite/tree-sitter-php@9678351433dfbcbd43bddfc614927f3dd7e33501",
     "tree-sitter-swift>=0.7,<0.9",
     "tree-sitter-lua>=0.2,<0.6",
     "tree-sitter-zig>=1.0,<2.0",

tests/test_php_keyword_constant_names.py

@@ -0,0 +1,85 @@
+"""PHP class constants named after reserved keywords must parse (#17).
+
+PHP allows any keyword as a class constant name (``public const string DEFAULT``).
+tree-sitter-php up to 0.25.1 rejects most of them once the constant is typed
+(PHP 8.3), and a handful (``VOID``, ``TRUE``...) even untyped
+(tree-sitter/tree-sitter-php#295). Error recovery then swallows the whole class
+declaration: the class vanishes from the graph and every ``Foo::DEFAULT`` access
+loses its ``references_constant`` edge.
+"""
+from __future__ import annotations
+
+from pathlib import Path
+
+import pytest
+
+from graphify.extract import extract, extract_php
+
+_CORPUS = """<?php
+
+declare(strict_types=1);
+
+namespace App\\Http;
+
+final class HttpAgent
+{
+    public const string DEFAULT = 'Fetcher/1.0 (+https://example.org; ops@example.org)';
+
+    public const string CRAWLER = 'Fetcher/1.0 (Crawler; +https://example.org; ops@example.org)';
+}
+
+final class Client
+{
+    public function agent(): string
+    {
+        return HttpAgent::DEFAULT;
+    }
+}
+"""
+
+
+def _write(path: Path, text: str) -> Path:
+    path.parent.mkdir(parents=True, exist_ok=True)
+    path.write_text(text, encoding="utf-8")
+    return path
+
+
+def _node_id(result: dict, label: str) -> str | None:
+    return next(
+        (n["id"] for n in result["nodes"] if n.get("label") == label and n.get("source_file")),
+        None,
+    )
+
+
+@pytest.mark.parametrize(
+    "declaration",
+    [
+        "public const string DEFAULT = 'a';",
+        "final protected const int PRIVATE = 1;",
+        "const ?string CLASS = null;",
+        "public const string|int FUNCTION = 2;",
+        "public const string NEW = 'b';",
+        "const VOID = 3;",
+        "const TRUE = true;",
+    ],
+)
+def test_keyword_named_constant_parses_cleanly(tmp_path: Path, declaration: str) -> None:
+    src = _write(tmp_path / "Holder.php", f"<?php\nfinal class Holder\n{{\n    {declaration}\n}}\n")
+
+    result = extract_php(src)
+
+    assert "parse_errors" not in result
+    assert any(n.get("label") == "Holder" for n in result["nodes"])
+
+
+def test_constants_only_class_keeps_its_node_and_fan_in(tmp_path: Path) -> None:
+    src = _write(tmp_path / "src/Http/HttpAgent.php", _CORPUS)
+
+    result = extract([src], cache_root=tmp_path / "graphify-out")
+
+    agent_id = _node_id(result, "HttpAgent")
+    assert agent_id is not None, "HttpAgent declaration node is missing"
+    assert any(
+        e.get("relation") == "references_constant" and e["target"] == agent_id
+        for e in result["edges"]
+    ), "Client::agent() lost its references_constant edge to HttpAgent"