From aacac198a8c8a3561b21c0b271024e3457a8bb40 Mon Sep 17 00:00:00 2001
From: Aaron Parecki <aaron@parecki.com>
Date: Sun, 23 Feb 2020 12:59:15 -0600
Subject: [PATCH] fix checking for empty post content

XRay now looks for images inside the HTML and does not consider those empty posts
---
 lib/XRay/Formats/Format.php                   | 33 +++++++++
 lib/XRay/Formats/Mf2.php                      | 12 ++-
 tests/SanitizeTest.php                        | 73 +++++++++++++++++--
 .../sanitize.example/content-is-only-video    | 11 +++
 .../entry-with-img-no-implied-photo           |  1 +
 .../photo-in-content-no-p-with-alt            | 11 +++
 .../photo-in-content-no-p-with-url-photo      | 12 +++
 .../photo-in-content-with-alt-no-text         | 11 +++
 .../photo-in-content-with-p-no-alt            | 11 +++
 9 files changed, 165 insertions(+), 10 deletions(-)
 create mode 100644 tests/data/sanitize.example/content-is-only-video
 create mode 100644 tests/data/sanitize.example/photo-in-content-no-p-with-alt
 create mode 100644 tests/data/sanitize.example/photo-in-content-no-p-with-url-photo
 create mode 100644 tests/data/sanitize.example/photo-in-content-with-alt-no-text
 create mode 100644 tests/data/sanitize.example/photo-in-content-with-p-no-alt

diff --git a/lib/XRay/Formats/Format.php b/lib/XRay/Formats/Format.php
index 6a58288..0042f40 100644
--- a/lib/XRay/Formats/Format.php
+++ b/lib/XRay/Formats/Format.php
@@ -75,6 +75,8 @@ abstract class Format implements iFormat {
     }
 
     $def = $config->getHTMLDefinition(true);
+
+    // add HTML <time> element
     $def->addElement(
       'time',
       'Inline',
@@ -84,6 +86,37 @@ abstract class Format implements iFormat {
         'datetime' => 'Text'
       ]
     );
+
+    /*
+    // This isn't working right now, not sure why
+    // http://developers.whatwg.org/the-video-element.html#the-video-element
+    $def->addElement(
+      'video',
+      'Block',
+      'Optional: (source, Flow) | (Flow, source) | Flow',
+      'Common',
+      [
+        'src' => 'URI',
+        'type' => 'Text',
+        'width' => 'Length',
+        'height' => 'Length',
+        'poster' => 'URI',
+        'preload' => 'Enum#auto,metadata,none',
+        'controls' => 'Bool',
+      ]
+    );
+    $def->addElement(
+      'source',
+      'Block',
+      'Flow',
+      'Common',
+      [
+        'src' => 'URI',
+        'type' => 'Text',
+      ]
+    );
+    */
+
     // Override the allowed classes to only support Microformats2 classes
     $def->manager->attrTypes->set('Class', new HTMLPurifier_AttrDef_HTML_Microformats2());
     $purifier = new HTMLPurifier($config);
diff --git a/lib/XRay/Formats/Mf2.php b/lib/XRay/Formats/Mf2.php
index 4f61459..bd700a8 100644
--- a/lib/XRay/Formats/Mf2.php
+++ b/lib/XRay/Formats/Mf2.php
@@ -278,7 +278,15 @@ class Mf2 extends Format {
         $data['html'] = $htmlContent;
       }
 
-      if(!$data['text'])
+      // Also add HTML content if it contains images, video or audio
+      // TODO: allow video and audio tags in content, then uncomment this
+      if(strpos($htmlContent, '<img') !== false
+        /* || strpos($htmlContent, '<video') !== false */
+        /* || strpos($htmlContent, '<audio') !== false */) {
+        $data['html'] = $htmlContent;
+      }
+
+      if(!$data['text'] && !isset($data['html']))
         return null;
 
       return $data;
@@ -419,7 +427,7 @@ class Mf2 extends Format {
     // If there is content, always return the plaintext content, and return HTML content if it's different
     if($content) {
       $content = self::parseHTMLValue('content', $item);
-      if($content['text']) {
+      if($content['text'] || $content['html']) {
         $data['content']['text'] = $content['text'];
         if(isset($content['html']))
           $data['content']['html'] = $content['html'];
diff --git a/tests/SanitizeTest.php b/tests/SanitizeTest.php
index f549d16..318f7fd 100644
--- a/tests/SanitizeTest.php
+++ b/tests/SanitizeTest.php
@@ -240,7 +240,7 @@ class SanitizeTest extends PHPUnit_Framework_TestCase {
     $this->assertEquals('This is a photo post with an <code>img</code> tag inside the content.', $data->data->content->html);
   }
 
-  public function testPhotoInContentWithNoText() {
+  public function testPhotoInContentWithNameAndNoText() {
     $url = 'http://sanitize.example/cleverdevil';
     $response = $this->parse(['url' => $url]);
 
@@ -285,6 +285,67 @@ class SanitizeTest extends PHPUnit_Framework_TestCase {
     $this->assertEquals('http://sanitize.example/photo.jpg', $data->data->photo[0]);
   }
 
+  public function testPhotoInContentWithNoText() {
+    $url = 'http://sanitize.example/photo-in-content-with-alt-no-text';
+    $response = $this->parse(['url' => $url]);
+
+    $body = $response->getContent();
+    $this->assertEquals(200, $response->getStatusCode());
+    $data = json_decode($body, true);
+
+    $this->assertEquals('<p><img src="http://sanitize.example/photo.jpg" alt="test" /></p>', $data['data']['content']['html']);
+    $this->assertEquals('', $data['data']['content']['text']);
+  }
+
+  public function testPhotoInContentWithPNoAlt() {
+    $url = 'http://sanitize.example/photo-in-content-with-p-no-alt';
+    $response = $this->parse(['url' => $url]);
+
+    $body = $response->getContent();
+    $this->assertEquals(200, $response->getStatusCode());
+    $data = json_decode($body, true);
+
+    $this->assertEquals('<p><img src="http://sanitize.example/photo.jpg" alt="photo.jpg" /></p>', $data['data']['content']['html']);
+    $this->assertEquals('', $data['data']['content']['text']);
+  }
+
+  public function testPhotoInContentNoPWithURLPhoto() {
+    $url = 'http://sanitize.example/photo-in-content-no-p-with-url-photo';
+    $response = $this->parse(['url' => $url]);
+
+    $body = $response->getContent();
+    $this->assertEquals(200, $response->getStatusCode());
+    $data = json_decode($body, true);
+
+    $this->assertEquals('<img src="http://sanitize.example/photo.jpg" alt="test" />', $data['data']['content']['html']);
+    $this->assertEquals('', $data['data']['content']['text']);
+  }
+
+  public function testPhotoInContentNoPWithAlt() {
+    // This h-entry has no u-url so has an implied u-photo. we don't actually care what happens with it because
+    // this should never happen in the wild
+    $url = 'http://sanitize.example/photo-in-content-no-p-with-alt';
+    $response = $this->parse(['url' => $url]);
+
+    $body = $response->getContent();
+    $this->assertEquals(200, $response->getStatusCode());
+    $data = json_decode($body, true);
+  }
+
+  /*
+  // TODO: add support for embedded video and audio tags in html content
+  public function testContentIsOnlyVideo() {
+    $url = 'http://sanitize.example/content-is-only-video';
+    $response = $this->parse(['url' => $url]);
+
+    $body = $response->getContent();
+    $this->assertEquals(200, $response->getStatusCode());
+    $data = json_decode($body, true);
+
+    print_r($data);
+  }
+  */
+
   public function testPhotosWithAlt() {
     // https://github.com/microformats/microformats2-parsing/issues/16
 
@@ -295,17 +356,15 @@ class SanitizeTest extends PHPUnit_Framework_TestCase {
     $this->assertEquals(200, $response->getStatusCode());
     $data = json_decode($body);
 
-    #print_r($data->data);
-
     $this->assertEquals('🌆 Made it to the first #NPSF #earlygang of the year, did in-betweeners abs, and 6:30 workout with a brutal burnout that was really its own workout. But wow pretty sunrise. Plus 50+ deg F? I’ll take it. #100PDPD'."\n\n".'#justshowup #darknesstodawn #wakeupthesun #fromwhereirun #NovemberProject #sunrise #latergram #nofilter', $data->data->content->text);
     $this->assertObjectNotHasAttribute('name', $data->data);
     $this->assertEquals('https://igx.4sqi.net/img/general/original/476_g7yruXflacsGr7PyVmECefyTBMB_R99zmPQxW7pftzA.jpg', $data->data->photo[0]);
     $this->assertEquals('https://igx.4sqi.net/img/general/original/476_zM3UgU9JHNhom907Ac_1WCEcUhGOJZaNWGlRmev86YA.jpg', $data->data->photo[1]);
   }
 
-  /*
-  // Commented out until #55 is resolved
+  // Ignoring this issue for now. This should not happen in the wild.
   // https://github.com/aaronpk/XRay/issues/55
+  // Skipping the implied photo check because in the wild, h-entrys should not exist withou a u-url, which stops implied parsing.
   public function testEntryWithImgNoImpliedPhoto() {
     // See https://github.com/microformats/microformats2-parsing/issues/6#issuecomment-357286985
     // and https://github.com/aaronpk/XRay/issues/52#issuecomment-357269683
@@ -319,11 +378,9 @@ class SanitizeTest extends PHPUnit_Framework_TestCase {
 
     $this->assertObjectNotHasAttribute('photo', $data->data);
     $this->assertObjectNotHasAttribute('name', $data->data);
-    $this->assertEquals('http://target.example.com/photo.jpg', $data->data->photo[0]);
     $this->assertEquals('This is a photo post with an img tag inside the content, which does not have a u-photo class so should not be removed.', $data->data->content->text);
-    $this->assertEquals('This is a photo post with an <code>img</code> tag inside the content, which does not have a u-photo class so should not be removed. <img src="http://target.example.com/photo.jpg" alt="a photo">', $data->data->content->html);
+    $this->assertEquals('This is a photo post with an <code>img</code> tag inside the content, which does not have a u-photo class so should not be removed. <img src="http://target.example.com/photo.jpg" alt="a photo" />', $data->data->content->html);
   }
-  */
 
   public function testWhitespaceWithBreakTags() {
     $url = 'http://sanitize.example/entry-with-br-tags';
diff --git a/tests/data/sanitize.example/content-is-only-video b/tests/data/sanitize.example/content-is-only-video
new file mode 100644
index 0000000..dec26fb
--- /dev/null
+++ b/tests/data/sanitize.example/content-is-only-video
@@ -0,0 +1,11 @@
+HTTP/1.1 200 OK
+Server: Apache
+Date: Wed, 09 Dec 2015 03:29:14 GMT
+Content-Type: text/html; charset=utf-8
+Connection: keep-alive
+
+<div class="h-entry">
+  <div class="e-content">
+    <video src="/video.mp4"></video>
+  </div>
+</div>
diff --git a/tests/data/sanitize.example/entry-with-img-no-implied-photo b/tests/data/sanitize.example/entry-with-img-no-implied-photo
index f788e57..9863694 100644
--- a/tests/data/sanitize.example/entry-with-img-no-implied-photo
+++ b/tests/data/sanitize.example/entry-with-img-no-implied-photo
@@ -10,5 +10,6 @@ Connection: keep-alive
   </head>
   <body class="h-entry">
     <p class="e-content p-name">This is a photo post with an <code>img</code> tag inside the content, which does not have a u-photo class so should not be removed. <img src="http://target.example.com/photo.jpg" alt="a photo"></p>
+    <a href="/permalink" class="u-url">permalink</a>
   </body>
 </html>
diff --git a/tests/data/sanitize.example/photo-in-content-no-p-with-alt b/tests/data/sanitize.example/photo-in-content-no-p-with-alt
new file mode 100644
index 0000000..b8e7cd0
--- /dev/null
+++ b/tests/data/sanitize.example/photo-in-content-no-p-with-alt
@@ -0,0 +1,11 @@
+HTTP/1.1 200 OK
+Server: Apache
+Date: Wed, 09 Dec 2015 03:29:14 GMT
+Content-Type: text/html; charset=utf-8
+Connection: keep-alive
+
+<div class="h-entry">
+  <div class="e-content">
+    <img src="/photo.jpg" alt="test">
+  </div>
+</div>
diff --git a/tests/data/sanitize.example/photo-in-content-no-p-with-url-photo b/tests/data/sanitize.example/photo-in-content-no-p-with-url-photo
new file mode 100644
index 0000000..b14bc77
--- /dev/null
+++ b/tests/data/sanitize.example/photo-in-content-no-p-with-url-photo
@@ -0,0 +1,12 @@
+HTTP/1.1 200 OK
+Server: Apache
+Date: Wed, 09 Dec 2015 03:29:14 GMT
+Content-Type: text/html; charset=utf-8
+Connection: keep-alive
+
+<div class="h-entry">
+  <div class="e-content">
+    <img src="/photo.jpg" alt="test">
+  </div>
+  <a href="/permalink" class="u-url">permalink</a>
+</div>
diff --git a/tests/data/sanitize.example/photo-in-content-with-alt-no-text b/tests/data/sanitize.example/photo-in-content-with-alt-no-text
new file mode 100644
index 0000000..43abbc6
--- /dev/null
+++ b/tests/data/sanitize.example/photo-in-content-with-alt-no-text
@@ -0,0 +1,11 @@
+HTTP/1.1 200 OK
+Server: Apache
+Date: Wed, 09 Dec 2015 03:29:14 GMT
+Content-Type: text/html; charset=utf-8
+Connection: keep-alive
+
+<div class="h-entry">
+  <div class="e-content">
+    <p><img src="/photo.jpg" alt="test"></p>
+  </div>
+</div>
diff --git a/tests/data/sanitize.example/photo-in-content-with-p-no-alt b/tests/data/sanitize.example/photo-in-content-with-p-no-alt
new file mode 100644
index 0000000..7eb3734
--- /dev/null
+++ b/tests/data/sanitize.example/photo-in-content-with-p-no-alt
@@ -0,0 +1,11 @@
+HTTP/1.1 200 OK
+Server: Apache
+Date: Wed, 09 Dec 2015 03:29:14 GMT
+Content-Type: text/html; charset=utf-8
+Connection: keep-alive
+
+<div class="h-entry">
+  <div class="e-content">
+    <p><img src="/photo.jpg"></p>
+  </div>
+</div>