From b65f73e3483d49ea09163d5d0255693ade135976 Mon Sep 17 00:00:00 2001
From: Michael <heluecht@pirati.ca>
Date: Wed, 14 Apr 2021 19:12:01 +0000
Subject: [PATCH] Improved mimeType detection and setting of the "type" field

---
 src/Model/Item.php                     |   7 +-
 src/Model/Post/Media.php               | 119 ++++++++++++++++++++-----
 src/Protocol/ActivityPub/Processor.php |  16 +---
 3 files changed, 104 insertions(+), 38 deletions(-)

diff --git a/src/Model/Item.php b/src/Model/Item.php
index 1d50dba2e7..863d89790e 100644
--- a/src/Model/Item.php
+++ b/src/Model/Item.php
@@ -2684,7 +2684,12 @@ class Item
 	{
 		$leading = '';
 		$trailing = '';
-		foreach (Post\Media::getByURIId($item['uri-id'], [Post\Media::DOCUMENT, Post\Media::TORRENT, Post\Media::UNKNOWN]) as $attachment) {
+		foreach (Post\Media::getByURIId($item['uri-id'], [Post\Media::AUDIO, Post\Media::VIDEO,
+			Post\Media::DOCUMENT, Post\Media::TORRENT, Post\Media::UNKNOWN]) as $attachment) {
+			if (in_array($attachment['type'], [Post\Media::AUDIO, Post\Media::VIDEO]) && strpos($item['body'], $attachment['url'])) {
+				continue;
+			}
+
 			$mime = $attachment['mimetype'];
 
 			$author = ['uid' => 0, 'id' => $item['author-id'],
diff --git a/src/Model/Post/Media.php b/src/Model/Post/Media.php
index 5dc5a0daa2..8601be02a3 100644
--- a/src/Model/Post/Media.php
+++ b/src/Model/Post/Media.php
@@ -36,12 +36,17 @@ use Friendica\Util\Images;
  */
 class Media
 {
-	const UNKNOWN  = 0;
-	const IMAGE    = 1;
-	const VIDEO    = 2;
-	const AUDIO    = 3;
-	const TORRENT  = 16;
-	const DOCUMENT = 128;
+	const UNKNOWN     = 0;
+	const IMAGE       = 1;
+	const VIDEO       = 2;
+	const AUDIO       = 3;
+	const TEXT        = 4;
+	const APPLICATION = 5;
+	const TORRENT     = 16;
+	const HTML        = 17;
+	const XML         = 18;
+	const PLAIN       = 19;
+	const DOCUMENT    = 128;
 
 	/**
 	 * Insert a post-media record
@@ -51,7 +56,7 @@ class Media
 	 */
 	public static function insert(array $media, bool $force = false)
 	{
-		if (empty($media['url']) || empty($media['uri-id']) || empty($media['type'])) {
+		if (empty($media['url']) || empty($media['uri-id']) || !isset($media['type'])) {
 			Logger::warning('Incomplete media data', ['media' => $media]);
 			return;
 		}
@@ -64,12 +69,7 @@ class Media
 			return;
 		}
 
-		$fields = ['mimetype', 'height', 'width', 'size', 'preview', 'preview-height', 'preview-width', 'description'];
-		foreach ($fields as $field) {
-			if (empty($media[$field])) {
-				unset($media[$field]);
-			}
-		}
+		$media = self::unsetEmptyFields($media);
 
 		// We are storing as fast as possible to avoid duplicated network requests
 		// when fetching additional information for pictures and other content.
@@ -78,6 +78,7 @@ class Media
 		$stored = $media;
 
 		$media = self::fetchAdditionalData($media);
+		$media = self::unsetEmptyFields($media);
 
 		if (array_diff_assoc($media, $stored)) {
 			$result = DBA::insert('post-media', $media, Database::INSERT_UPDATE);
@@ -87,6 +88,23 @@ class Media
 		}
 	}
 
+	/**
+	 * Remove empty media fields
+	 *
+	 * @param array $media
+	 * @return array cleaned media array
+	 */
+	private static function unsetEmptyFields(array $media)
+	{
+		$fields = ['mimetype', 'height', 'width', 'size', 'preview', 'preview-height', 'preview-width', 'description'];
+		foreach ($fields as $field) {
+			if (empty($media[$field])) {
+				unset($media[$field]);
+			}
+		}
+		return $media;
+	}
+
 	/**
 	 * Copy attachments from one uri-id to another
 	 *
@@ -130,23 +148,22 @@ class Media
 	public static function fetchAdditionalData(array $media)
 	{
 		// Fetch the mimetype or size if missing.
-		// We don't do it for torrent links since they need special treatment.
-		// We don't do this for images, since we are fetching their details some lines later anyway.
-		if (!in_array($media['type'], [self::TORRENT, self::IMAGE]) && (empty($media['mimetype']) || empty($media['size']))) {
+		if (empty($media['mimetype']) || empty($media['size'])) {
 			$timeout = DI::config()->get('system', 'xrd_timeout');
 			$curlResult = DI::httpRequest()->head($media['url'], ['timeout' => $timeout]);
 			if ($curlResult->isSuccess()) {
-				$header = $curlResult->getHeaderArray();
-				if (empty($media['mimetype']) && !empty($header['content-type'])) {
-					$media['mimetype'] = $header['content-type'];
+				if (empty($media['mimetype'])) {
+					$media['mimetype'] = $curlResult->getHeader('Content-Type');
 				}
-				if (empty($media['size']) && !empty($header['content-length'])) {
-					$media['size'] = $header['content-length'];
+				if (empty($media['size'])) {
+					$media['size'] = (int)$curlResult->getHeader('Content-Length');
 				}
+			} else {
+				Logger::notice('Could not fetch head', ['media' => $media]);
 			}
 		}
 
-		$filetype = !empty($media['mimetype']) ? strtolower(substr($media['mimetype'], 0, strpos($media['mimetype'], '/'))) : '';
+		$filetype = !empty($media['mimetype']) ? strtolower(current(explode('/', $media['mimetype']))) : '';
 
 		if (($media['type'] == self::IMAGE) || ($filetype == 'image')) {
 			$imagedata = Images::getInfoFromURLCached($media['url']);
@@ -155,6 +172,8 @@ class Media
 				$media['size'] = $imagedata['size'];
 				$media['width'] = $imagedata[0];
 				$media['height'] = $imagedata[1];
+			} else {
+				Logger::notice('No image data', ['media' => $media]);
 			}
 			if (!empty($media['preview'])) {
 				$imagedata = Images::getInfoFromURLCached($media['preview']);
@@ -164,9 +183,65 @@ class Media
 				}
 			}
 		}
+
+		if ($media['type'] != self::DOCUMENT) {
+			$media = self::addType($media);
+		}
+
 		return $media;
 	}
 
+	/**
+	 * Add the detected type to the media array
+	 *
+	 * @param array $data 
+	 * @return array data array with the detected type
+	 */
+	public static function addType(array $data)
+	{
+		if (empty($data['mimetype'])) {
+			Logger::info('No MimeType provided', ['media' => $data]);
+			return $data;
+		}
+
+		$type = explode('/', current(explode(';', $data['mimetype'])));
+		if (count($type) < 2) {
+			Logger::info('Unknown MimeType', ['type' => $type, 'media' => $data]);
+			$data['type'] = self::UNKNOWN;
+			return $data;
+		}
+
+		$filetype = strtolower($type[0]);
+		$subtype = strtolower($type[1]);
+
+		if ($filetype == 'image') {
+			$data['type'] = self::IMAGE;
+		} elseif ($filetype == 'video') {
+			$data['type'] = self::VIDEO;
+		} elseif ($filetype == 'audio') {
+			$data['type'] = self::AUDIO;
+		} elseif (($filetype == 'text') && ($subtype == 'html')) {
+			$data['type'] = self::HTML;
+		} elseif (($filetype == 'text') && ($subtype == 'xml')) {
+			$data['type'] = self::XML;
+		} elseif (($filetype == 'text') && ($subtype == 'plain')) {
+			$data['type'] = self::PLAIN;
+		} elseif ($filetype == 'text') {
+			$data['type'] = self::TEXT;
+		} elseif (($filetype == 'application') && ($subtype == 'x-bittorrent')) {
+			$data['type'] = self::TORRENT;
+		} elseif ($filetype == 'application') {
+			$data['type'] = self::APPLICATION;
+		} else {
+			$data['type'] = self::UNKNOWN;
+			Logger::info('Unknown type', ['filetype' => $filetype, 'subtype' => $subtype, 'media' => $data]);
+			return $data;
+		}
+
+		Logger::debug('Detected type', ['filetype' => $filetype, 'subtype' => $subtype, 'media' => $data]);
+		return $data;
+	}
+
 	/**
 	 * Tests for path patterns that are usef for picture links in Friendica
 	 *
diff --git a/src/Protocol/ActivityPub/Processor.php b/src/Protocol/ActivityPub/Processor.php
index 66aedf65b0..fdb97337fd 100644
--- a/src/Protocol/ActivityPub/Processor.php
+++ b/src/Protocol/ActivityPub/Processor.php
@@ -123,21 +123,7 @@ class Processor
 		}
 
 		$data = ['uri-id' => $uriid];
-
-		$filetype = strtolower(substr($attachment['mediaType'], 0, strpos($attachment['mediaType'], '/')));
-		if ($filetype == 'image') {
-			$data['type'] = Post\Media::IMAGE;
-		} elseif ($filetype == 'video') {
-			$data['type'] = Post\Media::VIDEO;
-		} elseif ($filetype == 'audio') {
-			$data['type'] = Post\Media::AUDIO;
-		} elseif (in_array($attachment['mediaType'], ['application/x-bittorrent', 'application/x-bittorrent;x-scheme-handler/magnet'])) {
-			$data['type'] = Post\Media::TORRENT;
-		} else {
-			Logger::info('Unknown type', ['attachment' => $attachment]);
-			return;
-		}
-
+		$data['type'] = Post\Media::UNKNOWN;
 		$data['url'] = $attachment['url'];
 		$data['mimetype'] = $attachment['mediaType'];
 		$data['height'] = $attachment['height'] ?? null;