improvements: display links, images style, remove break after header dots.

2 years ago · ba9085d5bd
2 changed files with 41 additions and 12 deletions
--- a/app/Ingest/PDFConvertor.php
+++ b/app/Ingest/PDFConvertor.php
@ -2,7 +2,6 @@

 namespace App\Ingest;

-use League\HTMLToMarkdown\HtmlConverter;
 use Symfony\Component\Process\Exception\ProcessFailedException;
 use Symfony\Component\Process\Process;

@ -98,7 +97,7 @@ class PDFConvertor extends AbstractConvertor
                foreach ($page as $items) {
                    $continuousP = '';

-                    foreach ($items as $p) {
+                    foreach ($items as $key => $p) {
                        if ($p->getName() == 'image') {
                            $basePath = $this->storage->path('');
                            $imageFilePath = str_replace($basePath, '', $p['src']);
@ -138,7 +137,16 @@ class PDFConvertor extends AbstractConvertor
                        }

                        if ($p->getName() == 'text') {
-                            $continuousP = $continuousP . $this->handleText($p, $fonts);
+                            if($p == '·') {
+                                continue;
+                            }
+
+                            $addition = null;
+                            if(isset($items[$key-1]) && $items[$key-1] == '·') {
+                                $addition = '· ';
+                            }
+
+                            $continuousP = $continuousP . $this->handleText($p, $fonts, $addition);

                            $hasText =  true;
                        }
@ -179,17 +187,17 @@ class PDFConvertor extends AbstractConvertor
    {
        $html = '';

-        $src = './' . pathinfo($p['src'], PATHINFO_BASENAME);
+        $src = './contracts-images/' . pathinfo($this->directoryPath, PATHINFO_BASENAME) . '/' . pathinfo($p['src'], PATHINFO_BASENAME);

        $html = $html . '<br>';
-        $html = $html . '<img style="position: absolute; top: ' . $p['top'] . 'px; left: ' . $p['left'] . 'px;" width="' . $p['width'] . '" height="' . $p['height'] . '" src="' . $src . '" alt="' . $caption . '" title="' . $caption . '">';
+        $html = $html . '<img style="position: relative; width:' . $p['width'] . 'px; height:' . $p['height'] . 'px;" src="' . $src . '" alt="' . $caption . '" title="' . $caption . '">';
        $html = $html . '<br>';
        $html = $html . '<br>';

        return $html;
    }

-    protected function handleText($p, $fonts)
+    protected function handleText($p, $fonts, $addition = null)
    {
        $id = (int) $p['font'];
        $font_size = $fonts[$id]['size'];
@ -197,7 +205,7 @@ class PDFConvertor extends AbstractConvertor
        $font_family = $fonts[$id]['family'];

        $style = '';
-        $style = $style . 'position: absolute;';
+        $style = $style . 'position: relative;';
        $style = $style . "color: $font_color;";
        $style = $style . "font-family: $font_family;";
        $style = $style . "font-weight: 900;";
@ -211,10 +219,16 @@ class PDFConvertor extends AbstractConvertor
            $content = '<i>' . $p->i . '</i>';
        } else if ($p->b) {
            $content = '<b>' . $p->b . '</b>';
+        } else if ($p->a) {
+            $content = $p . '<a>' . $p->a . '</a>';
        } else {
            $content = $p;
        }

+        if($addition) {
+            $content = $addition . $content;
+        }
+
        $tag = $this->getTag($font_size);

        return '<' . $tag . ' style="' . $style . '">' . $content . '</' . $tag . '>';
@ -236,7 +250,7 @@ class PDFConvertor extends AbstractConvertor
            return 'h3';
        }

-        return 'span';
+        return 'p';
    }

    protected function applyOCR($path)
@ -249,11 +263,27 @@ class PDFConvertor extends AbstractConvertor
    protected function convertHtmlToMD($contents)
    {
        $html = '<html><head><title></title></head><body>' . $contents . '</body></html>';
+        $filepath = $this->storage->path($this->directoryPath);
+
+        file_put_contents($filepath . '/document.html', $html);

-        $converter = new HtmlConverter();
-        $converter->getConfig()->setOption('strip_tags', true);
+        $process = new Process([
+            'pandoc',
+            '-f',
+            'html',
+            $filepath . '/document.html',
+            '-t',
+            'markdown_strict',
+        ]);
+
+        $process->run();

-        return $converter->convert($html);
+        
+        if (!$process->isSuccessful()) {
+            throw new ProcessFailedException($process);
+        }
+        
+        return $process->getOutput();
    }

    protected function prepareForConvertPDF()
--- a/composer.json
+++ b/composer.json
@ -13,7 +13,6 @@
        "fideloper/proxy": "^4.0",
        "laravel/framework": "^6.2",
        "laravel/tinker": "^2.0",
-        "league/html-to-markdown": "^5.0",
        "phpoffice/phpword": "^0.17.0",
        "predis/predis": "^1.1",
        "spatie/laravel-webhook-server": "^1.13",