Add .docx reader

mantas-done · Apr 9, 2024 · b1058a7 · b1058a7
1 parent 7e75387
commit b1058a7
Show file tree

Hide file tree

Showing 6 changed files with 222 additions and 0 deletions.
diff --git a/README.md b/README.md
@@ -25,6 +25,7 @@
 | [SAMI](https://en.wikipedia.org/wiki/SAMI)                                                            | .smi      | smi                  |
 | QuickTime                                                                                             | .qt.txt   | txt_quicktime        |
 | Rich text format (only reader)                                                                        | .rtf      | rtf                  |
+| DOCX (only reader)                                                                                    | .docx  | docx                 |
 | [LyRiCs](https://en.wikipedia.org/wiki/LRC_(file_format))                                             | .lrc      | lrc                  |
 | Comma separated values                                                                                | .csv      | csv                  |
 | Plaintext                                                                                             | .txt      | txt                  |

diff --git a/src/Code/Converters/DocxReader.php b/src/Code/Converters/DocxReader.php
@@ -0,0 +1,27 @@
+<?php
+
+namespace Done\Subtitles\Code\Converters;
+
+use Done\Subtitles\Code\Other\DocxToText;
+use Done\Subtitles\Subtitles;
+
+class DocxReader implements ConverterContract
+{
+    public function canParseFileContent($file_content)
+    {
+        return strpos($file_content, 'PK') === 0 && strpos($file_content, '[Content_Types].xml') !== false;
+    }
+
+    public function fileContentToInternalFormat($file_content, $original_file_content)
+    {
+        $text = DocxToText::text($original_file_content);
+        return Subtitles::loadFromString($text)->getInternalFormat();
+    }
+
+    public function internalFormatToFileContent(array $internal_format, array $options)
+    {
+        throw new \Exception('not implemented');
+    }
+
+
+}
diff --git a/src/Code/Other/DocxToText.php b/src/Code/Other/DocxToText.php
@@ -0,0 +1,168 @@
+<?php
+
+namespace Done\Subtitles\Code\Other;
+
+class DocxToText
+{
+    // text might not be correctly ordered
+    public static function text($path): string
+    {
+        $that = new self($path);
+
+        $content = '';
+//        $content .= $that->getAllHeadersXml();
+        $content .= $that->getFileContent('word/document.xml');
+//        $content .= $that->getFileContent('word/footnotes.xml');
+//        $content .= $that->getAllFootersXml();
+
+
+//        $content = str_replace('</w:r></w:p></w:tc><w:tc>', " ", $content); // original code poster had this
+        $content = str_replace('<w:tab/>', "    ", $content); // tab
+        $content = str_replace('<w:pStyle w:val="ListParagraph"/>', '1. ', $content); // numbering but not correct, jus for word count
+        $content = preg_replace('/<w:drawing>.*<\/w:drawing>/Um', '', $content);
+        $content = preg_replace('/<w:instrText.*<\/w:instrText>/Um', '', $content);
+        $content = str_replace('</w:r></w:p>', "\r\n", $content);
+        $striped_content = strip_tags($content);
+        $striped_content = html_entity_decode($striped_content, ENT_QUOTES | ENT_SUBSTITUTE | ENT_XML1, 'UTF-8');
+
+        return $striped_content;
+    }
+
+    private \ZipArchive $zip;
+
+    private function __construct($file_content)
+    {
+        $tmp_file = tempnam(sys_get_temp_dir(), 'prefix_');
+        file_put_contents($tmp_file, $file_content);
+
+        $zip = new \ZipArchive();
+        $opened = $zip->open($tmp_file, \ZipArchive::RDONLY); // zip archive can only open real file
+        if ($opened !== true) {
+            throw new \Exception();
+        }
+        $this->zip = $zip;
+        $this->tmp_path = $tmp_file;
+    }
+
+    public function __destruct()
+    {
+        $this->zip->close();
+        unlink($this->tmp_path);
+    }
+
+    private function pageCount(): int
+    {
+        $content = $this->getFileContent('docProps/app.xml');
+        if ($content === '') {
+            return 1; // file docProps/app.xml doesn't exist, guess that we have 1 page
+        }
+        $xml = new \SimpleXMLElement($content);
+        return (int)$xml->Pages;
+    }
+
+    private function getFileContent($internal_path): string
+    {
+        return $this->zip->getFromName($internal_path);
+    }
+
+    private function getAllHeadersXml(): string
+    {
+        $string = $this->getFileContent('word/document.xml');
+        preg_match_all('/<w:headerReference w:type="(?<type>.+)" r:id="(?<id>.+)"\/>/U', $string, $matches, PREG_SET_ORDER);
+        $headers = [];
+        foreach ($matches as $match) {
+            $internal_filename = $this->idToFileName($match['id']);
+            $headers[$match['type']] = $this->getFileContent($internal_filename);
+        }
+
+        $page_count = $this->pageCount();
+        $pages = [];
+        for ($i = 1; $i <= $page_count; $i++) {
+            $pages[$i] = '';
+        }
+
+        if (isset($headers['default'])) {
+            foreach ($pages as $key => &$page) {
+                $page = $headers['default'];
+            }
+        }
+        if (isset($headers['even'])) {
+            foreach ($pages as $key => &$page) {
+                if ($key % 2 === 0) {
+                    $page = $headers['even'];
+                }
+            }
+        }
+        if (isset($headers['odd'])) {
+            foreach ($pages as $key => &$page) {
+                if ($key % 2 === 1) {
+                    $page = $headers['odd'];
+                }
+            }
+        }
+        if (isset($headers['first'])) {
+            $pages[1] = $headers['first'];
+        }
+
+        return implode("\n", $pages);
+    }
+
+    private function getAllFootersXml(): string
+    {
+        $string = $this->getFileContent('word/document.xml');
+        preg_match_all('/<w:footerReference w:type="(?<type>.+)" r:id="(?<id>.+)"\/>/U', $string, $matches, PREG_SET_ORDER);
+        $footers = [];
+        foreach ($matches as $match) {
+            $internal_filename = $this->idToFileName($match['id']);
+            $footers[$match['type']] = $this->getFileContent($internal_filename);
+        }
+
+        $page_count = $this->pageCount();
+        $pages = [];
+        for ($i = 1; $i <= $page_count; $i++) {
+            $pages[$i] = '';
+        }
+
+        if (isset($footers['default'])) {
+            foreach ($pages as $key => &$page) {
+                $page = $footers['default'];
+            }
+        }
+        if (isset($footers['even'])) {
+            foreach ($pages as $key => &$page) {
+                if ($key % 2 === 0) {
+                    $page = $footers['even'];
+                }
+            }
+        }
+        if (isset($footers['odd'])) {
+            foreach ($pages as $key => &$page) {
+                if ($key % 2 === 1) {
+                    $page = $footers['odd'];
+                }
+            }
+        }
+        if (isset($footers['first'])) {
+            $pages[1] = $footers['first'];
+        }
+
+        $xml = implode("\n", $pages);
+
+        // remove page numbers from the footer
+        $xml = preg_replace('/<w:fldChar w:fldCharType="begin">.*w:fldCharType="end".*<\/w:r>/Um', '', $xml);
+
+        return $xml;
+    }
+
+    private function idToFileName(string $xml_id): string
+    {
+        $content = $this->getFileContent('word/_rels/document.xml.rels');
+        preg_match_all('/Relationship.+Id="(?<id>.+)".+Target="(?<filename>.+)"/U', $content, $matches, PREG_SET_ORDER);
+        foreach ($matches as $match) {
+            if ($match['id'] === $xml_id) {
+                return 'word/' . $match['filename'];
+            }
+        }
+        throw new \Exception("Can't find: " . $xml_id);
+    }
+}
diff --git a/src/Subtitles.php b/src/Subtitles.php
@@ -6,6 +6,7 @@
 use Done\Subtitles\Code\Converters\BinaryFinder;
 use Done\Subtitles\Code\Converters\CsvConverter;
 use Done\Subtitles\Code\Converters\DfxpConverter;
+use Done\Subtitles\Code\Converters\DocxReader;
 use Done\Subtitles\Code\Converters\EbuStlReader;
 use Done\Subtitles\Code\Converters\LrcConverter;
 use Done\Subtitles\Code\Converters\RtfReader;
@@ -49,6 +50,7 @@ class Subtitles
         ['extension' => 'txt',  'format' => 'txt_quicktime',    'name' => 'Quick Time Text',            'class' => TxtQuickTimeConverter::class],
         ['extension' => 'scc',  'format' => 'scc',              'name' => 'Scenarist',                  'class' => SccConverter::class],
         ['extension' => 'lrc',  'format' => 'lrc',              'name' => 'LyRiCs',                     'class' => LrcConverter::class],
+        ['extension' => 'docx', 'format' => 'docx',             'name' => 'DOCX',                       'class' => DocxReader::class],
         ['extension' => 'rtf',  'format' => 'rtf',              'name' => 'Rich text format',           'class' => RtfReader::class], // libraryies eather throws exception, not parses, or takes long to parse 2h file
         ['extension' => 'csv',  'format' => 'csv',              'name' => 'Coma Separated Values',      'class' => CsvConverter::class], // must be last from bottom
         ['extension' => 'bin',  'format' => 'bin',              'name' => 'Binary',                     'class' => BinaryFinder::class],

diff --git a/tests/files/docx.docx b/tests/files/docx.docx
diff --git a/tests/formats/DocxTest.php b/tests/formats/DocxTest.php
@@ -0,0 +1,24 @@
+<?php
+
+namespace Formats;
+
+use Done\Subtitles\Subtitles;
+use PHPUnit\Framework\TestCase;
+use Helpers\AdditionalAssertionsTrait;
+
+class DocxTest extends TestCase
+{
+
+    use AdditionalAssertionsTrait;
+
+    public function testParsesDocxFile()
+    {
+        $content = file_get_contents('./tests/files/docx.docx');
+        $actual = Subtitles::loadFromString($content)->getInternalFormat();
+        $expected = (new Subtitles())
+            ->add(137.4, 140.4, ["Senator, we're making", 'our final approach into Coruscant.'])
+            ->add(3740.5, 3742.5, ['Very good, Lieutenant.'])
+            ->getInternalFormat();
+        $this->assertInternalFormatsEqual($expected, $actual);
+    }
+}