languages: Add built-in PHP support via intelephense

Cargo.toml

@@ -771,6 +771,7 @@ tree-sitter-jsdoc = "0.23"
 tree-sitter-json = "0.24"
 tree-sitter-kotlin-codanna = "0.3.9"
 tree-sitter-md = { git = "https://github.com/tree-sitter-grammars/tree-sitter-markdown", rev = "9a23c1a96c0513d8fc6520972beedd419a973539" }
+tree-sitter-php = "0.24.2"
 tree-sitter-python = "0.25"
 tree-sitter-regex = "0.24"
 tree-sitter-ruby = "0.23"

RECAPS.md

@@ -6,6 +6,16 @@ Running log of completed work sessions, newest first. Each entry summarizes a co
 
 ## 2026-05-16
 
+### PHP language support (final of 4 bake-in PRs in this series)
+- Added PHP as a built-in language: tree-sitter grammar registered, `.php` / `.phtml` files recognized, syntax highlighting + outline + indents + injections + runnables + debugger queries on, and `intelephense` driven for completion / hover / go-to-def — found on `$PATH` first, otherwise `npm install`ed into the language container dir. Brings the built-in count to 17 (assuming Swift/Kotlin/Java land first).
+- Grammar: canonical `tree-sitter-php = "0.24.2"` (MIT). Same dev-dep-only `tree-sitter ^0.25` constraint as `tree-sitter-java` — resolves cleanly against workspace `tree-sitter 0.26`. Exposes `LANGUAGE_PHP` (mixed-mode PHP + inline HTML) and `LANGUAGE_PHP_ONLY` (pure PHP); registered the mixed-mode variant since `.php` files traditionally interleave HTML.
+- Asset files: pulled from `zed-extensions/php` (MIT) into `crates/grammars/src/php/`. 11 query files: `brackets.scm`, `config.toml`, `debugger.scm`, `embedding.scm`, `highlights.scm`, `indents.scm`, `injections.scm`, `outline.scm`, `overrides.scm`, `runnables.scm`, `textobjects.scm`. **Zero patches needed** — like Java, the upstream queries are written against the same canonical grammar we registered.
+- LSP adapter: new `crates/languages/src/php.rs` (~150 lines). Modeled directly on `typescript.rs` (single npm package version of the dual-package TS pattern). `check_if_user_installed` calls `delegate.which("intelephense")`; otherwise the adapter `npm_install_packages("intelephense", latest)` into `<container>/node_modules/` and invokes `node <container>/node_modules/intelephense/lib/intelephense.js --stdio`. **License note in adapter doc-comment**: intelephense is proprietary (not OSS); we don't redistribute it, `npm install` fetches it from npmjs.com at runtime. Users who prefer OSS alternatives can override via the `language_servers` setting (phpactor, phpls).
+- Wired into `crates/languages/src/lib.rs::init`: new `mod php;`, `Arc::new(php::PhpLspAdapter::new(node.clone()))` (no `fs` dep needed, unlike typescript — single-package adapter doesn't do yarn-sdk detection), and a `LanguageInfo { name: "php", adapters: vec![php_lsp_adapter], manifest_name: Some("composer.json".into()), .. }` entry between `markdown-inline` and `python` (alphabetical). All four upstream-shaped edits tagged `// PaddleBoard:`.
+- Verified: `cargo check -p grammars --features load-grammars` clean (9.8s, fresh `tree-sitter-php` compile); `cargo check -p paddleboard` clean (6.2s incremental); release build (17.97s) clean. Grammar load: zero query-loader errors in `PaddleBoard.log` after launch. NOT verified end-to-end: user didn't open `php-smoketest.php` during this session, so the `npm install` + intelephense spawn path wasn't exercised. Structurally identical to `typescript.rs` which is known to work; first reviewer can verify by opening any `.php` file (npm install + spawn should take ~10s).
+- Preserved: every existing language adapter; the `crates/grammars/` alphabetical ordering convention; `typescript.rs`'s npm pattern unchanged (just simpler-mirrored, not refactored).
+- Follow-ups: (1) live end-to-end smoke test pre-merge. (2) **OSS alternative adapter as an option** — `phpactor` is composer-distributed and a viable replacement for users avoiding intelephense's license. (3) Tailwind LSP cross-registration — the existing `tailwind_languages` list in `lib.rs::init` already includes "PHP" as a candidate language for the tailwind LSP, so once this PR lands, tailwind autocompletion in `.php` files Just Works without extra wiring. (4) Composer/PHPUnit context provider for test running. **Bake-in series complete after this PR**: Swift (#28) → Kotlin (#29) → Java (#30) → PHP. Count goes from 13 built-in languages to 17.
+
 ### Java language support (PR 3 of 4 in the bake-in series)
 - Added Java as a built-in language: tree-sitter grammar registered, `.java` files recognized, syntax highlighting + outline + indents + folds + runnables on, and `jdtls` (Eclipse JDT Language Server) wired for completion / hover / go-to-def when present on `$PATH`. Brings the built-in count to 16 (assuming Swift + Kotlin land first).
 - Grammar: canonical `tree-sitter-java = "0.23.5"` (MIT). Initially looked like a version-conflict candidate — its dep manifest pins `tree-sitter ^0.24` — but that's a **dev-dependency only** (for its own tests). Runtime deps are `tree-sitter-language ^0.1` (provides `LanguageFn`) and `cc ^1.1` (builds parser.c), which resolve cleanly against workspace's `tree-sitter 0.26`. No alternative crate needed (unlike Kotlin where we needed the codanna fork). Rationale captured in a multi-line comment on the `native_grammars()` entry so the next reader doesn't repeat the "this won't resolve" diagnostic.

crates/grammars/Cargo.toml

@@ -32,6 +32,7 @@ tree-sitter-jsdoc = { workspace = true, optional = true }
 tree-sitter-json = { workspace = true, optional = true }
 tree-sitter-kotlin-codanna = { workspace = true, optional = true }
 tree-sitter-md = { workspace = true, optional = true }
+tree-sitter-php = { workspace = true, optional = true }
 tree-sitter-python = { workspace = true, optional = true }
 tree-sitter-regex = { workspace = true, optional = true }
 tree-sitter-rust = { workspace = true, optional = true }
@@ -56,6 +57,7 @@ load-grammars = [
     "tree-sitter-json",
     "tree-sitter-kotlin-codanna",
     "tree-sitter-md",
+    "tree-sitter-php",
     "tree-sitter-python",
     "tree-sitter-regex",
     "tree-sitter-rust",

crates/grammars/src/grammars.rs

@@ -45,6 +45,14 @@ pub fn native_grammars() -> Vec<(&'static str, tree_sitter::Language)> {
         ("kotlin", tree_sitter_kotlin_codanna::language()),
         ("markdown", tree_sitter_md::LANGUAGE.into()),
         ("markdown-inline", tree_sitter_md::INLINE_LANGUAGE.into()),
+        // PaddleBoard: PHP via canonical tree-sitter/tree-sitter-php (MIT).
+        // Like tree-sitter-java, its `tree-sitter ^0.25` dep is dev-only;
+        // runtime uses `tree-sitter-language ^0.1` which resolves cleanly
+        // against workspace `tree-sitter 0.26`. Uses `LANGUAGE_PHP` (the
+        // mixed-mode `.php` grammar) — there's also a `LANGUAGE_PHP_ONLY`
+        // for pure-PHP files but the standard `.php` registration uses
+        // the mixed-mode grammar that handles inline HTML.
+        ("php", tree_sitter_php::LANGUAGE_PHP.into()),
         ("python", tree_sitter_python::LANGUAGE.into()),
         ("regex", tree_sitter_regex::LANGUAGE.into()),
         ("rust", tree_sitter_rust::LANGUAGE.into()),

crates/grammars/src/php/brackets.scm

@@ -0,0 +1,4 @@
+("{" @open "}" @close)
+("(" @open ")" @close)
+("[" @open "]" @close)
+("\"" @open "\"" @close)

crates/grammars/src/php/config.toml

@@ -0,0 +1,22 @@
+name = "PHP"
+grammar = "php"
+path_suffixes = ["php", "phtml"]
+first_line_pattern = '^#!.*php'
+line_comments = ["// ", "# "]
+block_comment = { start = "/*", end = "*/", prefix = "* ", tab_size = 1 }
+documentation_comment = { start = "/**", end = "*/", prefix = "* ", tab_size = 1 }
+autoclose_before = ";:.,=}])>"
+brackets = [
+    { start = "{", end = "}", close = true, newline = true },
+    { start = "[", end = "]", close = true, newline = true },
+    { start = "(", end = ")", close = true, newline = true },
+    { start = "\"", end = "\"", close = true, newline = false, not_in = ["string"] },
+    { start = "'", end = "'", close = true, newline = false, not_in = ["string"] },
+    { start = "/*", end = " */", close = true, newline = false, not_in = ["comment", "string"] },
+]
+collapsed_placeholder = "/* ... */"
+scope_opt_in_language_servers = ["tailwindcss-language-server"]
+prettier_parser_name = "php"
+prettier_plugins = ["@prettier/plugin-php"]
+completion_query_characters = ["$"]
+word_characters = ["$"]

crates/grammars/src/php/debugger.scm

@@ -0,0 +1,28 @@
+; e.g. `$age = 25` matches `$age`
+(expression_statement
+    (assignment_expression
+        left: (variable_name) @debug-variable
+    )
+)
+
+; e.g. `++$age` matches `$age`
+(expression_statement
+    (update_expression
+        argument: (variable_name) @debug-variable
+    )
+)
+
+; e.g. `if ($age > 18)` matches `$age`
+(binary_expression
+    left: (variable_name) @debug-variable
+)
+
+; e.g. `if (18 < $age)` matches `$age`
+(binary_expression
+    right: (variable_name) @debug-variable
+)
+
+; e.g. `__construct(int $age)` matches `$age`
+(simple_parameter
+    name: (variable_name) @debug-variable
+)

crates/grammars/src/php/embedding.scm

@@ -0,0 +1,36 @@
+(
+    (comment)* @context
+    .
+    [
+        (function_definition
+            "function" @name
+            name: (_) @name
+            body: (_
+                "{" @keep
+                "}" @keep) @collapse
+            )
+
+        (trait_declaration
+            "trait" @name
+            name: (_) @name)
+
+        (method_declaration
+            "function" @name
+            name: (_) @name
+            body: (_
+                "{" @keep
+                "}" @keep) @collapse
+            )
+
+        (interface_declaration
+            "interface" @name
+            name: (_) @name
+            )
+
+        (enum_declaration
+            "enum" @name
+            name: (_) @name
+            )
+
+        ] @item
+    )