Improve DFXP parsing

mantas-done · Jun 15, 2023 · 6bbdbaa · 6bbdbaa
1 parent c9a5258
commit 6bbdbaa
Show file tree

Hide file tree

Showing 4 changed files with 40 additions and 76 deletions.
diff --git a/src/Code/Converters/DfxpConverter.php b/src/Code/Converters/DfxpConverter.php
@@ -14,19 +14,7 @@ public function canParseFileContent($file_content)
 
     public function fileContentToInternalFormat($file_content)
     {
-        preg_match_all('/<p.+begin="(?<start>[^"]+).*end="(?<end>[^"]+)[^>]*>(?<text>(?!<\/p>).+)<\/p>/', $file_content, $matches, PREG_SET_ORDER);
-
-        $internal_format = [];
-        foreach ($matches as $block) {
-            $block['text'] = preg_replace('/<br\s*\/?>/', '<br>', $block['text']); // normalize <br>
-            $internal_format[] = [
-                'start' => static::dfxpTimeToInternal($block['start']),
-                'end' => static::dfxpTimeToInternal($block['end']),
-                'lines' => explode('<br>', $block['text']),
-            ];
-        }
-
-        return $internal_format;
+        return (new TtmlConverter())->fileContentToInternalFormat($file_content);
     }
 
     public function internalFormatToFileContent(array $internal_format)
@@ -74,22 +62,4 @@ protected static function internalTimeToDfxp($internal_time)
     {
         return ($internal_time * 10000000) . 't';
     }
-
-    protected static function dfxpTimeToInternal($dfxp_time)
-    {
-        if (substr($dfxp_time, -1) === 't') { // if last symbol is "t"
-            // parses 340400000t
-            return substr($dfxp_time, 0, -1) / 10000000;
-        } else {
-            // parses 00:00:34,040
-            $parts = explode(',', $dfxp_time);
-
-            $only_seconds = strtotime("1970-01-01 {$parts[0]} UTC");
-            $milliseconds = (float)('0.' . $parts[1]);
-
-            $time = $only_seconds + $milliseconds;
-
-            return $time;
-        }
-    }
 }
diff --git a/src/Code/Converters/TtmlConverter.php b/src/Code/Converters/TtmlConverter.php
@@ -17,36 +17,38 @@ public function fileContentToInternalFormat($file_content)
         $array = array();
 
         $body = $dom->getElementsByTagName('body')->item(0);
-        if ($body) {
-            $div = $body->getElementsByTagName('div')->item(0);
-            if ($div) {
-                $pElements = $div->getElementsByTagName('p');
-                foreach ($pElements as $p) {
-                    $begin = $p->getAttribute('begin');
-                    $end = $p->getAttribute('end');
-                    $lines = '';
-
-                    $textNodes = $p->childNodes;
-                    foreach ($textNodes as $node) {
-                        if ($node->nodeType === XML_TEXT_NODE) {
-                            $lines .= $node->nodeValue;
-                        } else {
-                            $lines .= $dom->saveXML($node); // Preserve HTML tags
-                        }
-                    }
-
-                    $lines = preg_replace('/<br\s*\/?>/', '<br>', $lines); // normalize <br>*/
-                    $lines = explode('<br>', $lines);
-                    $lines = array_map('strip_tags', $lines);
-                    $lines = array_map('trim', $lines);
-
-                    $array[] = array(
-                        'start' => static::ttmlTimeToInternal($begin),
-                        'end' => static::ttmlTimeToInternal($end),
-                        'lines' => $lines,
-                    );
+        if (!$body) {
+            throw new \Exception('no body');
+        }
+        $div = $body->getElementsByTagName('div')->item(0);
+        if (!$div) {
+            throw new \Exception('no div');
+        }
+        $pElements = $div->getElementsByTagName('p');
+        foreach ($pElements as $p) {
+            $begin = $p->getAttribute('begin');
+            $end = $p->getAttribute('end');
+            $lines = '';
+
+            $textNodes = $p->childNodes;
+            foreach ($textNodes as $node) {
+                if ($node->nodeType === XML_TEXT_NODE) {
+                    $lines .= $node->nodeValue;
+                } else {
+                    $lines .= $dom->saveXML($node); // Preserve HTML tags
                 }
             }
+
+            $lines = preg_replace('/<br\s*\/?>/', '<br>', $lines); // normalize <br>*/
+            $lines = explode('<br>', $lines);
+            $lines = array_map('strip_tags', $lines);
+            $lines = array_map('trim', $lines);
+
+            $array[] = array(
+                'start' => static::ttmlTimeToInternal($begin),
+                'end' => static::ttmlTimeToInternal($end),
+                'lines' => $lines,
+            );
         }
 
         return $array;
@@ -112,7 +114,10 @@ protected static function internalTimeToTtml($internal_time)
 
     protected static function ttmlTimeToInternal($ttml_time)
     {
-        if (substr($ttml_time, -1) === 's') {
+        if (substr($ttml_time, -1) === 't') { // if last symbol is "t"
+            // parses 340400000t
+            return substr($ttml_time, 0, -1) / 10000000;
+        } elseif (substr($ttml_time, -1) === 's') {
             return rtrim($ttml_time, 's');
         } else {
             $timeParts = explode(':', $ttml_time);

diff --git a/tests/files/dfxp_with_different_br.dfxp b/tests/files/dfxp_with_different_br.dfxp
@@ -14,7 +14,7 @@
   </head>
   <body>
     <div style="s1" xml:id="d1">
-    <p xml:id="p1" begin="00:00:00,000" end="00:00:01,000" region="bottomCenter">one<br/>two<br>three<br />four</p>
+    <p xml:id="p1" begin="00:00:00,000" end="00:00:01,000" region="bottomCenter">one<br/>two<br />three</p>
   </div>
   </body>
 </tt>
diff --git a/tests/formats/DfxpTest.php b/tests/formats/DfxpTest.php
@@ -39,31 +39,20 @@ public function testConvertFromDfxpToSrt()
         $srt_path = './tests/files/srt.srt';
         $dfxp_path = './tests/files/dfxp.dfxp';
 
-        // stl to srt
         $dfxp_object = Subtitles::loadFromFile($dfxp_path);
-        $stl_internal_format = $dfxp_object->getInternalFormat();
+        $actual = $dfxp_object->getInternalFormat();
 
         $srt_object = Subtitles::loadFromFile($srt_path);
-        $srt_internal_format = $srt_object->getInternalFormat();
-        // compare both internal formats
-        foreach ($srt_internal_format as $block_key => $srt_block) {
-            $start_time_diff = abs($srt_block['start'] - $stl_internal_format[$block_key]['start']);
-            $this->assertLessThan(0.1, $start_time_diff);
-
-            $end_time_diff = abs($srt_block['end'] - $stl_internal_format[$block_key]['end']);
-            $this->assertLessThan(0.1, $end_time_diff);
+        $expected = $srt_object->getInternalFormat();
 
-            foreach ($srt_block['lines'] as $line_key => $srt_line) {
-                $this->assertEquals($srt_line, $stl_internal_format[$block_key]['lines'][$line_key]);
-            }
-        }
+        $this->assertInternalFormatsEqual($expected, $actual);
     }
 
     public function testParsesDifferentBr()
     {
         $dfxp_object = Subtitles::loadFromFile('./tests/files/dfxp_with_different_br.dfxp');
         $actual = $dfxp_object->getInternalFormat();
-        $expected = (new Subtitles())->add(0, 1, ['one', 'two', 'three', 'four'])->getInternalFormat();
+        $expected = (new Subtitles())->add(0, 1, ['one', 'two', 'three'])->getInternalFormat();
         $this->assertInternalFormatsEqual($expected, $actual);
     }