From 0fcac762040a2a77da9b9b9ab60a4c3c8bd76ca8 Mon Sep 17 00:00:00 2001
From: Radu Liviu Carjan <radu.carjan@gmail.com>
Date: Thu, 20 Oct 2022 18:15:33 +0300
Subject: [PATCH] Fix ingest for PDF documents

---
 app/Ingest/PDFConvertor.php | 83 ++++++++++++++++++++-----------------
 1 file changed, 45 insertions(+), 38 deletions(-)

diff --git a/app/Ingest/PDFConvertor.php b/app/Ingest/PDFConvertor.php
index 29a266f..ed19530 100644
--- a/app/Ingest/PDFConvertor.php
+++ b/app/Ingest/PDFConvertor.php
@@ -5,6 +5,10 @@ namespace App\Ingest;
 use Symfony\Component\Process\Exception\ProcessFailedException;
 use Symfony\Component\Process\Process;
 
+use Illuminate\Support\Facades\Log;
+
+use thiagoalessio\TesseractOCR\TesseractOcrException;
+
 class PDFConvertor extends AbstractConvertor
 {
     public function execute()
@@ -17,7 +21,7 @@ class PDFConvertor extends AbstractConvertor
             throw new \Exception('Cannot get pdf file contents.');
         }
 
-        $this->storage->put("$this->directoryPath/document.json", json_encode($contents));
+        $this->storage->put("$this->directoryPath/document.html", $contents);
     }
 
     protected function getFileContents()
@@ -85,10 +89,7 @@ class PDFConvertor extends AbstractConvertor
 
         $imagesCount = 0;
 
-        $mdContents = '';
         $htmlContents = '';
-        $json = [];
-        $i = 0;
 
         try {
             foreach ($orderedList as $page) {
@@ -98,26 +99,28 @@ class PDFConvertor extends AbstractConvertor
                 foreach ($page as $items) {
                     $continuousP = '';
                     $firstOfText = true;
-                    
+
                     foreach ($items as $key => $p) {
                         if ($p->getName() == 'image') {
                             $imageInFooter = false;
                             $basePath = $this->storage->path('');
                             $imageFilePath = str_replace($basePath, '', $p['src']);
 
-                            // $textContents = $this->applyOCR($imageFilePath); @@ uncomment
-                            $textContents = null; // remove this
+                            try {
+                                $textContents = $this->applyOCR($imageFilePath);
+                            } catch (TesseractOcrException $e) {
+                                # Could not get text content from image. This means the image doesn't have text.
+                                $textContents = '';
+                            }
 
-                            if (false) {
+                            if ($textContents) {
                                 $imageInFooter = true;
                                 if ($html) {
-                                    $mdContents = $mdContents . $this->convertHtmlToMD($html) . "\n\n";
                                     $htmlContents = $htmlContents . $html;
 
                                     $html = '';
                                 }
 
-                                $mdContents = $mdContents . $textContents . "\n\n";
                                 $htmlContents = $htmlContents . "<div>$textContents</div>";
 
                                 $this->storage->delete($imageFilePath);
@@ -129,17 +132,14 @@ class PDFConvertor extends AbstractConvertor
                                 $imagesCount += 1;
                                 $caption = "Fig. $imagesCount";
 
-                                $imageJSON = $this->handleImage($p, $caption);
+                                $imageHTML = $this->handleImage($p, $caption);
 
                                 if (!$imageInFooter) {
-                                    $json[$i]['tag'] = 'img';
-                                    $json[$i]['src'] = $imageJSON['src'];
-                                    $json[$i]['style'] = $imageJSON['style'];
-                                    $json[$i]['details'] = $caption;
+                                    $html = $html . $imageHTML;
                                 } else {
                                     $html = $html . "<p> $caption </p>";
 
-                                    $footerImages[] = $imageJSON;
+                                    $footerImages[] = $imageHTML;
                                 }
                             }
                         }
@@ -148,33 +148,37 @@ class PDFConvertor extends AbstractConvertor
                             if($p == '·') {
                                 continue;
                             }
-                            
+
                             $addition = null;
                             if(isset($items[$key-1]) && $items[$key-1] == '·') {
                                 $addition = '· ';
                             }
-                            
-                            $continuousP = $this->handleText($p, $fonts, $addition, $firstOfText, true);
-                            if($firstOfText) {
-                                $json[$i]['tag'] = $continuousP['tag'];
-                                $json[$i]['style'] = $continuousP['style'];
-                            }
 
-                            (isset($json[$i]['content'])) ? $json[$i]['content'] = $json[$i]['content'] . $continuousP['content'] : $json[$i]['content'] = $continuousP['content'];
+                            $continuousP = $continuousP . $this->handleText($p, $fonts, $addition, $firstOfText);
 
                             $firstOfText = false;
                             $hasText =  true;
                         }
                     }
-                    $i++;
+
+                    $html = $html . '<p>' . $continuousP . '</p>';
+                }
+
+                if (!empty($footerImages)) {
+                    foreach ($footerImages as $footerImage) {
+                        $html = $html . '<p>' . $footerImage . '</p>';
+                    }
                 }
+
+                $htmlContents = $htmlContents . "<html><head></head><body>$html</body></html>";
             }
         } catch (\Exception $exception) {
             $this->storage->deleteDirectory($this->directoryPath);
 
-            \Illuminate\Support\Facades\Log::info($exception->getTraceAsString());
+            Log::info($exception->getTraceAsString());
 
-            throw new \Exception('Something went wrong.');
+            // throw new \Exception('Something went wrong.');
+            throw $exception;
         }
 
         if ( ! $hasText && ! $hasImages) {
@@ -185,15 +189,22 @@ class PDFConvertor extends AbstractConvertor
             $this->storage->delete($xmlFilePath);
         }
 
-        return $json;
+        // return $mdContents;
+        return $htmlContents;
     }
 
-    protected function handleImage($p)
+    protected function handleImage($p, $caption)
     {
-        return [
-            'src' => pathinfo($this->directoryPath, PATHINFO_BASENAME) . '/' . pathinfo($p['src'], PATHINFO_BASENAME),
-            'style' => 'width="' . $p['width'] . 'px" ' . 'height="' . $p['height'] . 'px"'
-        ];
+        $html = '';
+
+        $src = './contracts-images/' . pathinfo($this->directoryPath, PATHINFO_BASENAME) . '/' . pathinfo($p['src'], PATHINFO_BASENAME);
+
+        $html = $html . '<br>';
+        $html = $html . '<img width=' . $p['width'] . ' ' . 'height=' . $p['height'] . ' src="' . $src . '" alt="' . $caption . '" title="' . $caption . '">';
+        $html = $html . '<br>';
+        $html = $html . '<br>';
+
+        return $html;
     }
 
     protected function handleText($p, $fonts, $addition = null, $firstOfText = false)
@@ -228,11 +239,7 @@ class PDFConvertor extends AbstractConvertor
 
         $tag = $this->getTag($font_size);
 
-        return [
-            'tag' => $tag,
-            'content' => (string) $content,
-            'style' => $style
-        ];
+        return '<' . $tag . ' style="' . $style . '">' . $content . '</' . $tag . '>';
     }
 
     protected function getTag($size)