Update

- Added "EPUB" file format support. The file format supportability added for testing purpose. Please report us if your found any error in "EPUB" formatted file.
- HTML purser now updated to improve the stability.

Author: niladrimcn

.php-cs-fixer.cache

@@ -27,6 +27,7 @@ check file content MIME type before execute.
 - **ODT**
 - **ODS**
 - **RTF**
+- **EPUB**
 
 <img src="./blobs/warning.png?raw=true" alt="Note" width="12">***PPT*** support is under development. 
 

README.md

@@ -39,7 +39,9 @@
         "phpoffice/phpword": "^0.18",
         "laravel/framework": "^8.0|^9.0",
         "thiagoalessio/tesseract_ocr": "^2.12",
-        "stechstudio/laravel-php-cs-fixer": "^3.1"
+        "stechstudio/laravel-php-cs-fixer": "^3.1",
+        "html2text/html2text": "^4.3",
+        "lywzx/php-epub": "^0.1.2"
     },
     "require-dev": {
         "friendsofphp/php-cs-fixer": "^v3.8.0",

composer.json

@@ -0,0 +1,45 @@
+<?php
+
+namespace Nilgems\PhpTextract\ExtractorService\Extractors;
+
+use Html2Text\Html2Text;
+use lywzx\epub\EpubParser;
+use Nilgems\PhpTextract\Exceptions\TextractException;
+use Nilgems\PhpTextract\ExtractorService\Contracts\AbstractTextExtractor;
+use PhpOffice\PhpWord\Shared\ZipArchive;
+
+class EpubExtractor extends AbstractTextExtractor
+{
+    protected array $supported_mime_types = [
+        'application/epub+zip'
+    ];
+
+    public array $supported_extension = ['epub'];
+
+    /**
+     * Get extracted text
+     * @throws \Nilgems\PhpTextract\Exceptions\TextractException
+     * @throws \Exception
+     */
+    protected function getExtractedText(): string
+    {
+        if ($zip_path = $this->utilsService->getFilePath()) {
+            try {
+                $zip = new ZipArchive();
+                $zip->open($zip_path);
+                $epub = (new EpubParser($this->utilsService->getFilePath()));
+                $epub->parse();
+                $data = [];
+                foreach ($epub->getTOC() as $chapter) {
+                    $data[] = trim((new Html2Text($zip->getFromName($chapter['file_name'])))->getText());
+                }
+                $zip->close();
+                return (string) implode("\n", $data);
+            } catch (\Exception $exception) {
+                report($exception);
+                throw new TextractException('The extractor unable to parse the \'epub\' file.');
+            }
+        }
+        return "";
+    }
+}

composer.lock

@@ -2,6 +2,7 @@
 
 namespace Nilgems\PhpTextract\ExtractorService\Extractors;
 
+use Html2Text\Html2Text;
 use Nilgems\PhpTextract\ExtractorService\Contracts\TextProcessorHaveFilter;
 use Nilgems\PhpTextract\ExtractorService\ExtractorCommonProcessors\TextProcessor;
 
@@ -21,7 +22,7 @@ class HtmlExtractor extends TextProcessor implements TextProcessorHaveFilter
     public function getFilteredText(string $output): string
     {
         if (!empty($output)) {
-            return strip_tags($output);
+            return (new Html2Text($output))->getText();
         }
         return "";
     }

src/ExtractorService/Extractors/EpubExtractor.php

@@ -3,6 +3,7 @@
 namespace Nilgems\PhpTextract\Providers;
 
 use Illuminate\Support\ServiceProvider as IlluminateServiceProvider;
+use Nilgems\PhpTextract\ExtractorService\Extractors\EpubExtractor;
 use Nilgems\PhpTextract\ExtractorService\Extractors\HtmlExtractor;
 use Nilgems\PhpTextract\ExtractorService\Extractors\ImageExtractor;
 use Nilgems\PhpTextract\ExtractorService\Extractors\MsOfficeDocExtractor;
@@ -56,7 +57,8 @@ protected function registerExtractors(): void
             OpenOfficeSpreadSheet::class,
             PdfExtractor::class,
             RtfExtractor::class,
-            TxtExtractor::class
+            TxtExtractor::class,
+            EpubExtractor::class
         ];
         foreach ($extractors as $extractor) {
             $this->app->bind($extractor);