From 8bf199ad725c61c5c3a686cecc02c1f6c12ff4ca Mon Sep 17 00:00:00 2001
From: Andrew Dupont <github@andrewdupont.net>
Date: Sat, 18 May 2024 16:04:25 -0700
Subject: [PATCH] =?UTF-8?q?[language-php]=20Add=20a=20`firstLineRegex`?=
 =?UTF-8?q?=E2=80=A6?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

…so that the grammar-choosing logic picks the modern Tree-sitter grammar more consistently.
---
 .../grammars/modern-tree-sitter-php-html.cson            | 9 +++++++++
 1 file changed, 9 insertions(+)

diff --git a/packages/language-php/grammars/modern-tree-sitter-php-html.cson b/packages/language-php/grammars/modern-tree-sitter-php-html.cson
index 77b49260f..d7d2c658c 100644
--- a/packages/language-php/grammars/modern-tree-sitter-php-html.cson
+++ b/packages/language-php/grammars/modern-tree-sitter-php-html.cson
@@ -5,6 +5,15 @@ parser: 'tree-sitter-php'
 
 injectionRegex: 'php|PHP'
 
+# The TextMate grammar has an elaborate `firstLineMatch` pattern defined with
+# Oniguruma; it catches hashbang syntax and complex vim/emacs modelines along
+# with the typical `<?php` prefix.
+#
+# We can't compete with all that, but we can catch obvious instances of
+# shebangs and `<?php`s. In fact, we must, or else the TextMate grammar will be
+# ranked higher for certain files.
+firstLineRegex: '^\\s*<\\?(?:php|PHP|=|\\s|$)|^\\#!.*(?:\\s|\\/)php\\d?'
+
 treeSitter:
   parserSource: 'github:tree-sitter/tree-sitter-php#b569a5f2c0d592e67430520d1a0e1f765d83ceb0'
   grammar: 'tree-sitter/tree-sitter-php.wasm'