Display and index hyphenated words as normal words (#1009)

Co-authored-by: Sebastian Meyer <[email protected]>
kitodo · Sep 28, 2023 · 72e9635 · 72e9635
1 parent 9583e92
commit 72e9635
Show file tree

Hide file tree

Showing 3 changed files with 50 additions and 14 deletions.
diff --git a/Classes/Format/Alto.php b/Classes/Format/Alto.php
@@ -38,8 +38,20 @@ public function getRawText(\SimpleXMLElement $xml)
         $rawText = '';
         $xml->registerXPathNamespace('alto', 'http://www.loc.gov/standards/alto/ns-v2#');
         // Get all (presumed) words of the text.
-        $words = $xml->xpath('./alto:Layout/alto:Page/alto:PrintSpace//alto:TextBlock/alto:TextLine/alto:String/@CONTENT');
-        if (!empty($words)) {
+        $strings = $xml->xpath('./alto:Layout/alto:Page/alto:PrintSpace//alto:TextBlock/alto:TextLine/alto:String');
+        $words = [];
+        if (!empty($strings)) {
+            for ($i = 0; $i < count($strings); $i++) {
+                $attributes = $strings[$i]->attributes();
+                if (isset($attributes['SUBS_TYPE'])) {
+                    if ($attributes['SUBS_TYPE'] == 'HypPart1') {
+                        $i++;
+                        $words[] = $attributes['SUBS_CONTENT'];
+                    }
+                } else {
+                    $words[] = $attributes['CONTENT'];
+                }
+            }
             $rawText = implode(' ', $words);
         }
         return $rawText;
@@ -101,6 +113,12 @@ public function getTextAsMiniOcr(\SimpleXMLElement $xml)
      */
     private function getWord($attributes)
     {
+        if (!empty($attributes['SUBS_CONTENT'])) {
+            if ($attributes['SUBS_TYPE'] == 'HypPart1') {
+                return htmlspecialchars((string) $attributes['SUBS_CONTENT']);
+            }
+            return ' ';
+        }
         return htmlspecialchars((string) $attributes['CONTENT']) . ' ';
     }
 

diff --git a/Resources/Public/JavaScript/PageView/AltoParser.js b/Resources/Public/JavaScript/PageView/AltoParser.js
@@ -323,36 +323,51 @@ dlfAltoParser.prototype.parseTextLineFeatures_ = function(node) {
  * @private
  */
 dlfAltoParser.prototype.parseContentFeatures_ = function(node) {
-    var textlineContentElements = $(node).children(),
-        textlineContentFeatures = [];
+    var textLineContentElements = $(node).children(),
+        textLineContentFeatures = [];
 
-    for (var i = 0; i < textlineContentElements.length; i++) {
-        var feature = this.parseFeatureWithGeometry_(textlineContentElements[i]),
+    for (var i = 0; i < textLineContentElements.length; i++) {
+        var feature = this.parseFeatureWithGeometry_(textLineContentElements[i]),
             fulltext = '';
 
-        // parse fulltexts
-        switch (textlineContentElements[i].nodeName.toLowerCase()) {
+        // parse full texts
+        switch (textLineContentElements[i].nodeName.toLowerCase()) {
             case 'string':
-                fulltext = textlineContentElements[i].getAttribute('CONTENT');
+                fulltext = this.parseString_(textLineContentElements[i]);
                 break;
             case 'sp':
                 fulltext = ' ';
                 break;
             case 'hyp':
-                fulltext = '-';
+                fulltext = '';
                 break;
             default:
                 fulltext = '';
         };
         feature.setProperties({fulltext});
 
-        textlineContentFeatures.push(feature);
+        textLineContentFeatures.push(feature);
     };
 
-    return textlineContentFeatures;
+    return textLineContentFeatures;
 };
 
-
+/**
+ *
+ * @param {Element}
+ * @return {string}
+ * @private
+ */
+dlfAltoParser.prototype.parseString_ = function(textLineContentElement) {
+    var hyphen = textLineContentElement.getAttribute('SUBS_TYPE')
+    if (typeof(hyphen) != 'undefined' && hyphen != null) {
+        if (hyphen == 'HypPart1') {
+            return textLineContentElement.getAttribute('SUBS_CONTENT');
+        }
+        return '';
+    };
+    return textLineContentElement.getAttribute('CONTENT');
+};
 
 /**
  *

diff --git a/Resources/Public/JavaScript/PageView/FulltextControl.js b/Resources/Public/JavaScript/PageView/FulltextControl.js
@@ -595,7 +595,10 @@ dlfViewerFullTextControl.prototype.getTextLineSpan = function(textLine) {
         textLineSpan.append(this.getItemForTextLineSpan(item));
     }
 
-    textLineSpan.append(dlfTmplFulltext.space.cloneNode());
+    // clone space only if last element is not a hyphen
+    if (content[content.length - 1].get('type') != 'hyp') {
+        textLineSpan.append(dlfTmplFulltext.space.cloneNode());
+    }
 
     return textLineSpan;
 };