Merge pull request #8 from daita/feature/noid/ocr-on-pdf

convert pdf and ocr the image
nextcloud · Feb 17, 2019 · 676a166 · 676a166
2 parents ba302d7 + 51fadf7
commit 676a166
Show file tree

Hide file tree

Showing 8 changed files with 146 additions and 8 deletions.
diff --git a/composer.json b/composer.json
@@ -10,6 +10,7 @@
     }
   ],
   "require": {
-    "thiagoalessio/tesseract_ocr": "2.4.0"
+    "thiagoalessio/tesseract_ocr": "2.4.0",
+    "spatie/pdf-to-image": "1.8.1"
   }
 }
diff --git a/composer.lock b/composer.lock
diff --git a/js/admin.elements.js b/js/admin.elements.js
@@ -34,16 +34,19 @@ var fts_tesseract_elements = {
 	tesseract_ocr: null,
 	tesseract_psm: null,
 	tesseract_lang: null,
+	tesseract_pdf: null,
 
 	init: function () {
 		fts_tesseract_elements.tesseract_div = $('#files_ocr-tesseract');
 		fts_tesseract_elements.tesseract_psm = $('#tesseract_psm');
 		fts_tesseract_elements.tesseract_lang = $('#tesseract_lang');
 		fts_tesseract_elements.tesseract_ocr = $('#tesseract_ocr');
+		fts_tesseract_elements.tesseract_pdf = $('#tesseract_pdf');
 
 		fts_tesseract_elements.tesseract_ocr.on('change', fts_tesseract_elements.updateSettings);
 		fts_tesseract_elements.tesseract_psm.on('change', fts_tesseract_elements.updateSettings);
 		fts_tesseract_elements.tesseract_lang.on('change', fts_tesseract_elements.updateSettings);
+		fts_tesseract_elements.tesseract_pdf.on('change', fts_tesseract_elements.updateSettings);
 	},
 
 

diff --git a/js/admin.settings.js b/js/admin.settings.js
@@ -49,6 +49,7 @@ var fts_tesseract_settings = {
 		fts_tesseract_elements.tesseract_ocr.prop('checked', (result.tesseract_enabled === '1'));
 		fts_tesseract_elements.tesseract_psm.val(result.tesseract_psm);
 		fts_tesseract_elements.tesseract_lang.val(result.tesseract_lang);
+		fts_tesseract_elements.tesseract_pdf.prop('checked', (result.tesseract_pdf === '1'));
 
 		fts_admin_settings.tagSettingsAsSaved(fts_tesseract_elements.tesseract_div);
 
@@ -69,7 +70,8 @@ var fts_tesseract_settings = {
 		var data = {
 			tesseract_enabled: (fts_tesseract_elements.tesseract_ocr.is(':checked')) ? 1 : 0,
 			tesseract_psm: fts_tesseract_elements.tesseract_psm.val(),
-			tesseract_lang: fts_tesseract_elements.tesseract_lang.val()
+			tesseract_lang: fts_tesseract_elements.tesseract_lang.val(),
+			tesseract_pdf: (fts_tesseract_elements.tesseract_pdf.is(':checked')) ? 1 : 0
 		};
 
 		$.ajax({

diff --git a/lib/AppInfo/Application.php b/lib/AppInfo/Application.php
@@ -82,5 +82,6 @@ function(GenericEvent $e) {
 			}
 		);
 	}
+
 }
 
diff --git a/lib/Service/ConfigService.php b/lib/Service/ConfigService.php
@@ -46,11 +46,13 @@ class ConfigService {
 	const TESSERACT_ENABLED = 'tesseract_enabled';
 	const TESSERACT_PSM = 'tesseract_psm';
 	const TESSERACT_LANG = 'tesseract_lang';
+	const TESSERACT_PDF = 'tesseract_pdf';
 
 	public $defaults = [
 		self::TESSERACT_ENABLED => '0',
 		self::TESSERACT_PSM     => '4',
-		self::TESSERACT_LANG    => 'eng'
+		self::TESSERACT_LANG    => 'eng',
+		self::TESSERACT_PDF     => '0'
 	];
 
 

diff --git a/lib/Service/TesseractService.php b/lib/Service/TesseractService.php
@@ -38,6 +38,9 @@
 use OCP\Files\NotFoundException;
 use OCP\Files_FullTextSearch\Model\AFilesDocument;
 use OCP\FullTextSearch\Model\IndexDocument;
+use OCP\FullTextSearch\Model\ISearchRequest;
+use Spatie\PdfToImage\Exceptions\PageDoesNotExist;
+use Spatie\PdfToImage\Pdf;
 use Symfony\Component\EventDispatcher\GenericEvent;
 use thiagoalessio\TesseractOCR\TesseractOCR;
 
@@ -80,7 +83,8 @@ public function parsedMimeType(string $mimeType, string $extension): bool {
 			'image/png',
 			'image/jpeg',
 			'image/tiff',
-			'image/vnd.djvu'
+			'image/vnd.djvu',
+			'application/pdf'
 		];
 
 		foreach ($ocrMimes as $mime) {
@@ -119,6 +123,9 @@ public function onFileIndexing(GenericEvent $e) {
 	 * @param GenericEvent $e
 	 */
 	public function onSearchRequest(GenericEvent $e) {
+		/** @var ISearchRequest $file */
+		$request = $e->getArgument('request');
+		$request->addPart('ocr');
 	}
 
 
@@ -134,13 +141,18 @@ private function extractContentUsingTesseractOCR(AFilesDocument &$document, File
 			}
 
 			$extension = pathinfo($document->getPath(), PATHINFO_EXTENSION);
+
 			if (!$this->parsedMimeType($document->getMimetype(), $extension)) {
 				return;
 			}
 
 			// TODO: How to set options so that the index can be reset if admin settings are changed
 			//	$this->configService->setDocumentIndexOption($document, ConfigService::FILES_OCR);
 
+			if ($this->ocrPdf($document, $file)) {
+				return;
+			}
+
 			$content = $this->ocrFile($file);
 		} catch (Exception $e) {
 			return;
@@ -161,9 +173,20 @@ private function ocrFile(File $file): string {
 		try {
 			$path = $this->getAbsolutePath($file);
 		} catch (Exception $e) {
-			throw new NotFoundException('file not found');
+			$this->miscService->log('Exception while ocr file: ' . $e->getMessage(), 1);
+			throw new NotFoundException();
 		}
 
+		return $this->ocrFileFromPath($path);
+	}
+
+
+	/**
+	 * @param string $path
+	 *
+	 * @return string
+	 */
+	private function ocrFileFromPath(string $path): string {
 		$ocr = new TesseractOCR($path);
 		$ocr->psm($this->configService->getAppValue(ConfigService::TESSERACT_PSM));
 		$lang = explode(',', $this->configService->getAppValue(ConfigService::TESSERACT_LANG));
@@ -175,6 +198,51 @@ private function ocrFile(File $file): string {
 	}
 
 
+	/**
+	 * @param AFilesDocument $document
+	 * @param File $file
+	 *
+	 * @return bool
+	 * @throws NotFoundException
+	 */
+	private function ocrPdf(AFilesDocument $document, File $file): bool {
+		if ($document->getMimetype() !== 'application/pdf') {
+			return false;
+		}
+
+		if ($this->configService->getAppValue(ConfigService::TESSERACT_PDF) !== '1') {
+			return true;
+		}
+
+		try {
+			$path = $this->getAbsolutePath($file);
+			$pdf = new Pdf($path);
+		} catch (Exception $e) {
+			$this->miscService->log('Exception while ocr pdf file: ' . $e->getMessage(), 1);
+			throw new NotFoundException();
+		}
+
+		$content = '';
+		for ($i = 1; $i <= $pdf->getNumberOfPages(); $i++) {
+			// we create a temp image file
+			$tmpFile = tmpfile();
+			$tmpPath = stream_get_meta_data($tmpFile)['uri'];
+
+			try {
+				$pdf->setPage($i);
+				$pdf->saveImage($tmpPath);
+
+				$content .= $this->ocrFileFromPath($tmpPath);
+			} catch (PageDoesNotExist $e) {
+			}
+		}
+
+		$document->addPart('ocr', $content);
+
+		return true;
+	}
+
+
 	/**
 	 * @param string $extension
 	 *

diff --git a/templates/settings.admin.php b/templates/settings.admin.php
@@ -75,6 +75,18 @@
 				<input type="text" class="big" id="tesseract_lang" value=""/>
 			</div>
 		</div>
+
+		<div class="div-table-row tesseract_ocr_enabled">
+			<div class="div-table-col div-table-col-left">
+				<span class="leftcol">PDF</span>
+				<br/>
+				<em>enable the OCR of PDF (heavy on resource)</em>
+			</div>
+			<div class="div-table-col">
+				<input type="checkbox" id="tesseract_pdf" value="1"/>
+			</div>
+		</div>
+
 	</div>