feat: improve office document discovery

Signed-off-by: Sebastian Krupinski <krupinski01@gmail.com>

shared/lib/Blob/Signature.php

@@ -23,7 +23,10 @@ use finfo;
 class Signature {
 
 	/** Minimum bytes needed for reliable detection */
-	public const HEADER_SIZE = 256;
+	public const SAMPLE_SIZE = 65536;
+
+	/** Cached finfo instance */
+	private static ?finfo $finfo = null;
 
 	/**
 	 * Fallback magic byte signatures for when finfo is unavailable
@@ -41,16 +44,32 @@ class Signature {
 		['offset' => 0, 'bytes' => '52494646', 'format' => 'riff'], // WAV/AVI/WEBP
 	];
 
-	/** Cached finfo instance */
+	/**
-	private static ?finfo $finfo = null;
+	 * Zip-container marker strings used to distinguish OOXML/ODF/EPUB documents
+	 * from a generic ZIP archive. Filenames inside a ZIP's local file headers
+	 * (and ODF's mandatory uncompressed "mimetype" entry content) are stored
+	 * as plain text, so a simple substring search reliably identifies these
+	 * formats without needing to parse the archive structure.
+	 */
+	private const ZIP_CONTAINER_MARKERS = [
+		'word/document.xml' => ['application/vnd.openxmlformats-officedocument.wordprocessingml.document', 'docx'],
+		'xl/workbook.xml' => ['application/vnd.openxmlformats-officedocument.spreadsheetml.sheet', 'xlsx'],
+		'ppt/presentation.xml' => ['application/vnd.openxmlformats-officedocument.presentationml.presentation', 'pptx'],
+		'application/vnd.oasis.opendocument.text' => ['application/vnd.oasis.opendocument.text', 'odt'],
+		'application/vnd.oasis.opendocument.spreadsheet' => ['application/vnd.oasis.opendocument.spreadsheet', 'ods'],
+		'application/vnd.oasis.opendocument.presentation' => ['application/vnd.oasis.opendocument.presentation', 'odp'],
+		'application/epub+zip' => ['application/epub+zip', 'epub'],
+	];
 
 	/**
 	 * Detect both MIME type and format from content bytes in a single operation
 	 *
-	 * @param string $headerBytes First bytes of the file content (256 recommended)
+	 * @param string $headerBytes First bytes of the file content
+	 * @param string|null $content Full (or larger) content, when available, used to
+	 *   distinguish OOXML/ODF/EPUB documents from a generic ZIP archive
 	 * @return array{mime: string, format: string} Array with 'mime' and 'format' keys
 	 */
-	public static function detect(string $headerBytes): array {
+	public static function detect(string $headerBytes, ?string $content = null): array {
 		if (strlen($headerBytes) === 0) {
 			return ['mime' => MimeTypes::MIME_BINARY, 'format' => MimeTypes::FORMAT_BINARY];
 		}
@@ -75,6 +94,16 @@ class Signature {
 			$format = self::detectFromMagicBytes($headerBytes);
 		}
 
+		// A bare "zip" result is a generic container; when more of the file is
+		// available, check for OOXML/ODF/EPUB markers rather than reporting the
+		// misleadingly generic zip mime/format for what is really a document.
+		if ($format === 'zip' && $content !== null) {
+			$container = self::detectZipContainer($content);
+			if ($container !== null) {
+				[$mime, $format] = $container;
+			}
+		}
+
 		// Ensure MIME type is set
 		if ($mime === null || $mime === MimeTypes::MIME_BINARY) {
 			$mime = MimeTypes::toMime($format) ?? MimeTypes::MIME_BINARY;
@@ -83,6 +112,32 @@ class Signature {
 		return ['mime' => $mime, 'format' => $format];
 	}
 
+	/**
+	 * Look for known OOXML/ODF/EPUB entry markers within ZIP content
+	 *
+	 * Only a fixed-size window from the head and tail of the content is scanned,
+	 * regardless of total length, so detection cost does not grow with the size
+	 * of large archive uploads. Every entry name is mirrored in full in the ZIP
+	 * central directory near the end of the archive, so sampling the head and
+	 * tail is enough without scanning the whole file.
+	 *
+	 * @param string $data ZIP content to scan (filenames/mimetype entry are stored uncompressed)
+	 * @return array{0: string, 1: string}|null [mime, format] pair, or null if no marker matched
+	 */
+	private static function detectZipContainer(string $data): ?array {
+		$length = strlen($data);
+		$sample = $length <= self::SAMPLE_SIZE * 2
+			? $data
+			: substr($data, 0, self::SAMPLE_SIZE) . substr($data, -self::SAMPLE_SIZE);
+
+		foreach (self::ZIP_CONTAINER_MARKERS as $marker => $result) {
+			if (str_contains($sample, $marker)) {
+				return $result;
+			}
+		}
+		return null;
+	}
+
 	/**
 	 * Detect both MIME type and format from a stream in a single operation
 	 *
@@ -91,7 +146,7 @@ class Signature {
 	 */
 	public static function detectFromStream($stream): array {
 		$position = ftell($stream);
-		$headerBytes = fread($stream, self::HEADER_SIZE);
+		$headerBytes = fread($stream, self::SAMPLE_SIZE);
 		fseek($stream, $position);
 
 		if ($headerBytes === false || $headerBytes === '') {
@@ -140,7 +195,7 @@ class Signature {
 	 */
 	public static function detectFormatFromStream($stream): string {
 		$position = ftell($stream);
-		$headerBytes = fread($stream, self::HEADER_SIZE);
+		$headerBytes = fread($stream, self::SAMPLE_SIZE);
 		fseek($stream, $position);
 
 		if ($headerBytes === false || $headerBytes === '') {
@@ -158,7 +213,7 @@ class Signature {
 	 */
 	public static function detectMimeTypeFromStream($stream): ?string {
 		$position = ftell($stream);
-		$headerBytes = fread($stream, self::HEADER_SIZE);
+		$headerBytes = fread($stream, self::SAMPLE_SIZE);
 		fseek($stream, $position);
 
 		if ($headerBytes === false || $headerBytes === '') {