From 1cce4e54794f1f907f4afaafadcf3eb50858e543 Mon Sep 17 00:00:00 2001
From: OpaqueRock <305760598+OpaqueRock@users.noreply.github.com>
Date: Wed, 5 Aug 2026 12:17:59 +0300
Subject: [PATCH] Skip runs of horizontal whitespace in one scan in
 Lexer::positionAfterWhitespace

---
 benchmarks/LexerBench.php | 31 +++++++++++++++++++++++++++++++
 src/Language/Lexer.php    | 17 ++++++++++++++---
 2 files changed, 45 insertions(+), 3 deletions(-)

diff --git a/benchmarks/LexerBench.php b/benchmarks/LexerBench.php
index e1f0ea5ff..b90018d7c 100644
--- a/benchmarks/LexerBench.php
+++ b/benchmarks/LexerBench.php
@@ -16,9 +16,24 @@ class LexerBench
 {
     private Source $introQuery;
 
+    private Source $deeplyIndentedQuery;
+
     public function setUp(): void
     {
         $this->introQuery = new Source(Introspection::getIntrospectionQuery());
+        $this->deeplyIndentedQuery = new Source($this->buildDeeplyIndentedQuery());
+    }
+
+    /** Query dominated by long runs of leading-tab indentation, to weigh the whitespace-skipping path. */
+    private function buildDeeplyIndentedQuery(): string
+    {
+        $indent = str_repeat("\t", 16);
+        $fields = [];
+        for ($i = 0; $i < 200; ++$i) {
+            $fields[] = "{$indent}field{$i}(arg: \"{$i}\")";
+        }
+
+        return "query DeepIndent {\n" . implode("\n", $fields) . "\n}\n";
     }
 
     /**
@@ -36,4 +51,20 @@ public function benchIntrospectionQuery(): void
             $token = $lexer->advance();
         } while ($token->kind !== Token::EOF);
     }
+
+    /**
+     * @Warmup(2)
+     *
+     * @Revs(100)
+     *
+     * @Iterations(5)
+     */
+    public function benchDeeplyIndentedQuery(): void
+    {
+        $lexer = new Lexer($this->deeplyIndentedQuery);
+
+        do {
+            $token = $lexer->advance();
+        } while ($token->kind !== Token::EOF);
+    }
 }
diff --git a/src/Language/Lexer.php b/src/Language/Lexer.php
index 928a49550..c5b8711a3 100644
--- a/src/Language/Lexer.php
+++ b/src/Language/Lexer.php
@@ -43,6 +43,9 @@ class Lexer
      */
     private const STRING_STOP_BYTES = "\"\\\x00\x01\x02\x03\x04\x05\x06\x07\x08\x0a\x0b\x0c\x0d\x0e\x0f\x10\x11\x12\x13\x14\x15\x16\x17\x18\x19\x1a\x1b\x1c\x1d\x1e\x1f";
 
+    /** Ignored single-byte characters that never start or end a line, so runs of them can be skipped in bulk. */
+    private const HORIZONTAL_WHITESPACE_BYTES = "\t ,";
+
     public Source $source;
 
     /** @phpstan-var ParserOptions */
@@ -640,12 +643,20 @@ private function assertValidBlockStringCharacterCode(int $code, int $position):
      */
     private function positionAfterWhitespace(): void
     {
+        $body = $this->source->body;
+
         while ($this->position < $this->source->length) {
+            // Tab, space, and comma runs carry no line/column bookkeeping, so
+            // an entire run can be consumed in one native scan instead of one
+            // readChar()/moveStringCursor() call per character.
+            $run = strspn($body, self::HORIZONTAL_WHITESPACE_BYTES, $this->byteStreamPosition);
+            if ($run > 0) {
+                $this->moveStringCursor($run, $run);
+            }
+
             [, $code, $bytes] = $this->readChar();
 
-            // Skip whitespace
-            // tab | space | comma | BOM
-            if (in_array($code, [9, 32, 44, 0xFEFF], true)) {
+            if ($code === 0xFEFF) { // BOM
                 $this->moveStringCursor(1, $bytes);
             } elseif ($code === 10) { // new line
                 $this->moveStringCursor(1, $bytes);