Initial xentral_oss_20.3.c9ffacf

classes/Modules/HocrParser/Service/HocrDataExtractor.php

@@ -0,0 +1,71 @@
+<?php
+
+declare(strict_types=1);
+
+namespace Xentral\Modules\HocrParser\Service;
+
+use Xentral\Modules\HocrParser\Data\BoundingBoxCollection;
+use Xentral\Modules\HocrParser\Finder\BoundingBoxFinder;
+use Xentral\Modules\HocrParser\Finder\CurrencyCodeFinderFacet;
+use Xentral\Modules\HocrParser\Finder\PatternMatcher;
+use Xentral\Modules\HocrParser\Finder\RelativePositionFinderFacet;
+
+class HocrDataExtractor
+{
+    /** @var array|string[] $validCurrencies */
+    private $validCurrencies;
+
+    /** @var HocrParser $parser */
+    private $parser;
+
+    public function __construct(HocrParser $parser, array $validCurrencies)
+    {
+        $this->validCurrencies = $validCurrencies;
+        $this->parser = $parser;
+    }
+
+    /**
+     * @param string $document HOCR-Dokument
+     * @param array  $settings Einstellungen (pro Lieferant)
+     *
+     * @return array
+     */
+    public function extractLiabilityDataFromHocrDocument(string $document, array $settings): array
+    {
+        // Suchkriterien zusammenstellen
+        $criteria = [];
+        if(!empty($settings['invoice_number']['term']) && !empty($settings['invoice_number']['direction'])){
+            $criteria['invoice_number'] =
+                new RelativePositionFinderFacet(
+                    $settings['invoice_number']['term'],
+                    $settings['invoice_number']['direction'],
+                    PatternMatcher::PATTERN_DOCUMENT_NUMBER
+                );
+        }
+        if(!empty($settings['invoice_date']['term']) && !empty($settings['invoice_date']['direction'])){
+            $criteria['invoice_date'] =
+                new RelativePositionFinderFacet(
+                    $settings['invoice_date']['term'],
+                    $settings['invoice_date']['direction'],
+                    PatternMatcher::PATTERN_DATE
+                );
+        }
+        if(!empty($settings['total_gross']['term']) && !empty($settings['total_gross']['direction'])){
+            $criteria['total_gross'] =
+                new RelativePositionFinderFacet(
+                    $settings['total_gross']['term'],
+                    $settings['total_gross']['direction'],
+                    PatternMatcher::PATTERN_MONEY
+                );
+        }
+        // Immer nach Währungen suchen
+        $waehrungen = $this->validCurrencies;
+        $criteria['currency'] = new CurrencyCodeFinderFacet($waehrungen);
+
+        // BoundingBoxes aus HOCR extrahieren + Suche anhand der Suchkriterien durchführen
+        $boxes = new BoundingBoxCollection($this->parser->parse($document));
+        $finder = new BoundingBoxFinder($boxes, $criteria);
+
+        return $finder->Find();
+    }
+}

classes/Modules/HocrParser/Service/HocrParser.php

@@ -0,0 +1,67 @@
+<?php
+
+declare(strict_types=1);
+
+namespace Xentral\Modules\HocrParser\Service;
+
+use DOMDocument;
+use Xentral\Modules\HocrParser\Data\BoundingBox;
+
+final class HocrParser
+{
+    /**
+     * @param string $content HOCR-Dokument
+     *
+     * @return array|BoundingBox[]
+     */
+    public function parse(string $content): array
+    {
+        $boxes = [];
+
+        $dom = new DOMDocument;
+        $dom->loadXML($content);
+        $spans = $dom->getElementsByTagName('span');
+
+        /** @var DOMElement $span */
+        foreach ($spans as $span) {
+            if ($span->getAttribute('class') === 'ocrx_word') {
+                $title = $span->getAttribute('title');
+                $coords = $this->extractCoordinates($title);
+                $text = trim($span->nodeValue);
+
+                // Boxen ohne Text auslassen; kann vorkommen bei Barcode-Fonts
+                if (empty($text)) {
+                    continue;
+                }
+
+                // Boxen mit Text sammeln
+                $boxes[] = new BoundingBox(
+                    $coords['tlx'], $coords['tly'], $coords['brx'], $coords['bry'], ['text' => $text]
+                );
+            }
+        }
+
+        return $boxes;
+    }
+
+    /**
+     * @param string $text
+     *
+     * @return array|bool
+     */
+    private function extractCoordinates(string $text)
+    {
+        // bbox 599 2737 743 2758; x_wconf 96
+        $parts = explode(' ', $text);
+        if ($parts[0] === 'bbox') {
+            return [
+                'tlx' => (int)$parts[1], // Top left X-Coordinate
+                'tly' => (int)$parts[2], // Top left Y
+                'brx' => (int)$parts[3], // Bottom right X
+                'bry' => (int)$parts[4], // Bottom right Y
+            ];
+        }
+
+        return false;
+    }
+}