Preserve urlencoded data in the rewritten path

packages/playground/data-liberation/phpunit.xml

@@ -7,6 +7,7 @@
       <file>tests/WPBlockMarkupProcessorTests.php</file>
       <file>tests/WPBlockMarkupUrlProcessorTests.php</file>
       <file>tests/URLParserWHATWGComplianceTests.php</file>
+      <file>tests/UrldecodeNTests.php</file> 
     </testsuite>
   </testsuites>
 </phpunit>

packages/playground/data-liberation/src/functions.php

@@ -47,17 +47,39 @@ function wp_rewrite_urls( $options ) {
 		$parsed_matched_url           = $p->get_parsed_url();
 		$parsed_matched_url->protocol = $new_site_url->protocol;
 		$parsed_matched_url->hostname = $new_site_url->hostname;
-		$decoded_matched_pathname     = urldecode( $parsed_matched_url->pathname );
 
 		// Update the pathname if needed.
 		if ( '/' !== $current_site_url->pathname ) {
-			// The matched URL starts with $current_site_name->pathname.
+			/**
+			 * The matched URL starts with $current_site_name->pathname.
+			 * 
+			 * We want to retain the portion of the pathname that comes
+			 * after $current_site_name->pathname. This is not a simple
+			 * substring operation because the matched URL may have
+			 * urlencoded bytes at the beginning. We need to decode
+			 * them before taking the substring.
+			 * 
+			 * However, we can't just decode the entire pathname because
+			 * the part after $current_site_name->pathname may itself
+			 * contain urlencoded bytes. If we decode them here, it
+			 * may change a path such as `/%2561/foo`, which decodes
+			 * as `/61/foo`, to `/a/foo`.
+			 * 
+			 * Therefore, we're going to decode a part of the string. We'll
+			 * start at the beginning and keep going until we've found
+			 * enough decoded bytes to skip over $current_site_name->pathname.
+			 * Then we'll take the remaining, still encoded bytes as the new pathname.
+			 */
+			$decoded_matched_pathname = urldecode_n(
+				$parsed_matched_url->pathname,
+				strlen( $current_site_pathname_with_trailing_slash )
+			);
 			$parsed_matched_url->pathname =
 				$new_site_pathname_with_trailing_slash .
-				substr(
-					$decoded_matched_pathname,
-					strlen( urldecode( $current_site_pathname_with_trailing_slash ) )
-				);
+					substr(
+						$decoded_matched_pathname,
+						strlen( $current_site_pathname_with_trailing_slash )
+					);
 		}
 
 		/*
@@ -129,3 +151,51 @@ function url_matches( URL $subject, string $current_site_url_no_trailing_slash )
 		str_starts_with( $matched_pathname_decoded, $current_pathname_no_trailing_slash . '/' )
 	);
 }
+
+/**
+ * Decodes the first n **encoded bytes** a URL-encoded string.
+ *
+ * For example, `urldecode_n( '%22is 6 %3C 6?%22 – asked Achilles', 1 )` returns
+ * '"is 6 %3C 6?%22 – asked Achilles' because only the first encoded byte is decoded.
+ *
+ * @param string $string The string to decode.
+ * @param int $target_length The maximum length of the resulting string.
+ * @return string The decoded string.
+ */
+function urldecode_n( $string, $target_length ) {
+	$result = '';
+	$at = 0;
+	while(true) {
+		if($at + 3 > strlen($string)) {
+			break;
+		}
+
+		$last_at = $at;
+		$at += strcspn( $string, '%', $at );
+		// Consume bytes except for the percent sign.
+		$result .= substr( $string, $last_at, $at - $last_at );
+
+		// If we've already decoded the requested number of bytes, stop.
+		if(strlen($result) >= $target_length) {
+			break;
+		}
+
+		++$at;
+		$decodable_length = strspn(
+			$string,
+			'0123456789ABCDEFabcdef',
+			$at,
+			2
+		);
+		if($decodable_length === 2) {
+			// Decode the hex sequence.
+			$result .= chr(hexdec($string[$at] . $string[$at + 1]));
+			$at += 2;
+		} else {
+			// Consume the percent sign and move on.
+			$result .= '%';
+		}
+	}
+	$result .= substr($string, $at);
+	return $result;
+}

packages/playground/data-liberation/tests/UrldecodeNTests.php

@@ -0,0 +1,49 @@
+<?php
+
+use PHPUnit\Framework\TestCase;
+
+class UrldecodeNTests extends TestCase {
+
+	/**
+	 *
+	 * @dataProvider provider_test_urldecode_n
+	 */
+	public function test_urldecode_n(
+		$original_string,
+		$decode_length,
+		$expected_string,
+	) {
+		$result = urldecode_n( $original_string, $decode_length );
+		$this->assertEquals( $expected_string, $result, 'Failed to decode the first n bytes of the string' );
+	}
+
+	static public function provider_test_urldecode_n() {
+		return [
+			'Encoded path segment with no encoded bytes later on' => [
+				'original_string' => '/%73/%63/image.png',
+				'decode_length' => 4,
+				'expected_string' => '/s/c/image.png',
+			],
+			'Encoded path segment with encoded bytes later on' => [
+				'original_string' => '/%73/%63/%73%63ience.png',
+				'decode_length' => 4,
+				'expected_string' => '/s/c/%73%63ience.png',
+			],
+			'Decode past the encoded path segment' => [
+				'original_string' => '/%73/%63/science.png',
+				'decode_length' => 10,
+				'expected_string' => '/s/c/science.png',
+			],
+			'Double percent sign – decode it' => [
+				'original_string' => '/%%73cience.png',
+				'decode_length' => 3,
+				'expected_string' => '/%science.png',
+			],
+			'Double percent sign – finish decoding after the first percent sign' => [
+				'original_string' => '/%%73cience.png',
+				'decode_length' => 2,
+				'expected_string' => '/%%73cience.png',
+			],
+		];
+	}
+}

packages/playground/data-liberation/tests/WPRewriteUrlsTests.php

@@ -64,6 +64,30 @@ static public function provider_test_wp_rewrite_urls() {
 				'https://legacy-blog.com/~jappleseed/1997.10.1/',
 				'https://modern-webstore.org/blog/'
 			],
+			/**
+			 * The urlencoded data needs to stay urlencoded. If it's decoded, the
+			 * resulting path will be wrongly rewritten as
+			 * 
+			 * "/blog/%65-reasons-to-migrate-data/"
+			 * 
+			 * Which decodes to:
+			 * 
+			 * "/blog/a-reasons-to-migrate-data/"
+			 * 
+			 * But if the path is not decoded, it correctly becomes
+			 * 
+			 * "/blog/%2565-reasons-to-migrate-data/"
+			 * 
+			 * Which decodes to:
+			 * 
+			 * "/blog/%65-reasons-to-migrate-data/"
+			 */
+			'Path in an HTML attribute with URLencoded data' => [ 
+				'<a href="/~jappleseed/1997.10.1/%2561-reasons-to-migrate-data/">61 reasons to migrate data</a>',
+				'<a href="/blog/%2561-reasons-to-migrate-data/">61 reasons to migrate data</a>',
+				'https://legacy-blog.com/~jappleseed/1997.10.1/',
+				'https://modern-webstore.org/blog/'
+			],
 			'Domain in an HTML attribute – encoded using HTML entities' => [ 
 				'<a href="&#104;&#116;tps://&#108;&#101;g&#97;&#99;&#121;&#45;&#98;&#108;&#111;&#103;.&#99;&#111;&#109;/pages/contact-us">Contact us</a>',
 				'<a href="https://modern-webstore.org/pages/contact-us">Contact us</a>',