searchanddisplace-ingest/app/Ingest/PDFConvertor.php

<?php

namespace App\Ingest;

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

class PDFConvertor extends AbstractConvertor
{
    public function execute()
    {
        //        $this->prepareForConvertPDF();

        $result = $this->getFileContents();

        if ( ! $result['has_images'] && ! $result['has_text']) {
            throw new \Exception('Cannot get pdf file contents.');
        }

        if ($result['has_text']) {
            $mdContents = '';

            foreach ($result['htmls'] as $html) {
                $converter = new HtmlConverter();
                $converter->getConfig()->setOption('strip_tags', true);

                $contents = $converter->convert($html);

                $mdContents = $mdContents . "\n\n" . $contents;
            }

            $this->storage->put("$this->directoryPath/document.md", $mdContents);

            return;
        }

        // Only contains images.
        $imagesContent = '';
        $files = $this->storage->allFiles($this->path);

        foreach ($files as $file) {
            // Only get the image files from the directory, it may contain some empty html files too.

            // @TODO Only OCR images with text and delete them afterwards, the remaining ignore and keep.
            if (in_array(pathinfo($file, PATHINFO_EXTENSION), ['jpg', 'png'])) {
                $ocr = new OCR($this->storage->path($file));

                $imagesContent = $imagesContent . $ocr->execute();

                $this->storage->delete($file);
            }
        }

        $this->storage->put("$this->directoryPath/document.md", $imagesContent);
    }

    protected function getFileContents()
    {
        $outputPath = $this->storage->path("$this->directoryPath/html");

        $process = new Process([
            'pdftohtml',
            '-xml',
            $this->storage->path($this->path),
            $outputPath
        ]);

        $process->run();

        if (!$process->isSuccessful()) {
            throw new ProcessFailedException($process);
        }

        // Remove original document.
        $this->storage->delete($this->path);

        return $this->getDataFromXML();
    }

    protected function getDataFromXML()
    {
        $xmlFilePath = "$this->directoryPath/html.xml";

        $contents = $this->storage->get($xmlFilePath);

        $xml = simplexml_load_string($contents);

        $orderedList = [];
        $fonts = [];

        foreach ($xml->page as $page) {
            $pageNumber = (int) $page['number'][0];

            $orderedList[$pageNumber] = [];

            foreach ($page as $p) {
                if ($p->getName() === 'fontspec') {
                    $fonts[(int) $p['id']]['family'] = (string) $p['family'];
                    $fonts[(int) $p['id']]['size'] = (string) $p['size'];
                    $fonts[(int) $p['id']]['color'] = (string) $p['color'];
                }

                if (isset($p['top'])) {
                    $top = (int) $p['top'];

                    if ( ! array_key_exists($top, $orderedList[$pageNumber])) {
                        $orderedList[$pageNumber][$top] = [];
                    }

                    $orderedList[$pageNumber][$top][] = $p;
                }
            }

            ksort($orderedList[$pageNumber]);
        }

        $htmls = [];
        $hasImages = false;
        $hasText = false;

        $imagesCount = 0;
        $imagesInFooter = true;

        try {
            foreach ($orderedList as $page) {
                $html = '';
                $footerImages = [];

                foreach ($page as $items) {
                    $continuousP = '';

                    foreach ($items as $p) {
                        if ($p->getName() == 'image') {
                            $hasImages = true;

                            $imagesCount += 1;
                            $caption = "Fig. $imagesCount";

                            $imageHTML = $this->handleImage($p, $caption);

                            if ( ! $imagesInFooter) {
                                $html = $html . $imageHTML;
                            } else {
                                $html = $html . "<p> $caption </p>";

                                $footerImages[] = $imageHTML;
                            }
                        }

                        if ($p->getName() == 'text') {
                            $continuousP = $continuousP . $this->handleText($p, $fonts);

                            $hasText =  true;
                        }
                    }

                    $html = $html . '<p>' . $continuousP . '</p>';
                }

                if ($imagesInFooter) {
                    foreach ($footerImages as $index => $footerImage) {
                        $html = $html . '<p>' . $footerImage . '</p>';
//                        $html = $html . '<p> Fig. ' . ($index + 1) . '</p>';
                    }
                }

                $htmls[] = '<html><head><title></title></head><body>' . $html . '</body></html>';
            }
        } catch (\Exception $exception) {
            $this->storage->deleteDirectory($this->directoryPath);

            \Illuminate\Support\Facades\Log::info($exception->getTraceAsString());

            throw new \Exception('Something went wrong.');
        }

        if ( ! $hasText && ! $hasImages) {
            // Remove directory because we do not have any use for it anymore.
            $this->storage->deleteDirectory($this->directoryPath);
        } else {
            // Remove the unnecessary 'xml' file.
            $this->storage->delete($xmlFilePath);
        }

        return [
            'has_images' => $hasImages,
            'has_text' => $hasText,
            'htmls' => $htmls,
        ];
    }

    protected function handleImage($p, $caption)
    {
        $html = '';

        $src = './' . pathinfo($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 . '<br>';
        $html = $html . '<br>';

        return $html;
    }

    protected function handleText($p, $fonts)
    {
        $id = (int) $p['font'];
        $font_size = $fonts[$id]['size'];
        $font_color = $fonts[$id]['color'];
        $font_family = $fonts[$id]['family'];

        $style = '';
        $style = $style . 'position: absolute;';
        $style = $style . "color: $font_color;";
        $style = $style . "font-family: $font_family;";
        $style = $style . "font-weight: 900;";
        $style = $style . "width: " . $p['width'] . "px;";
        $style = $style . "height: " . $p['height'] . "px;";
        $style = $style . "top: " . $p['top'] . "px;";
        $style = $style . "left: " . $p['left'] . "px;";
        $style = $style . "font-size: $font_size" . "px;";

        if ($p->i) {
            $content = '<i>' . $p->i . '</i>';
        } else if ($p->b) {
            $content = '<b>' . $p->b . '</b>';
        } else {
            $content = $p;
        }

        $tag = $this->getTag($font_size);

        return '<' . $tag . ' style="' . $style . '">' . $content . '</' . $tag . '>';
    }

    protected function getTag($size)
    {
        if ($size > 24) {
            return 'h1';
        }

        if ($size > 18) {
            return 'h2';
        }

        if ($size > 16) {
            return 'h3';
        }

        return 'span';
    }

    protected function prepareForConvertPDF()
    {
        (new Process(['export HOME=' . env('USER_HOME_PATH')]))->run();

        $process = new Process([
            'pip3',
            'install',
            'pdftotext',
        ]);

        $process->run();

        if (!$process->isSuccessful()) {
            throw new ProcessFailedException($process);
        }
    }
}