redo snippet creation

Author: JorisLambrechts

themes/default/view/search/resource.phtml

@@ -123,55 +123,25 @@ endif;
                 $media = $resource->media();
 
                 if(isset($media[1])):
-                  echo $media[1]->mediaType();
-                  $text = $media[1]->render();
-                  //echo $text;
-                  
-                  $text = str_replace(array("-\r", "-\n"), '', $text);
-                  $text = str_replace(array("\r", "\n"), ' ', $text);
-                  $text = str_replace("  ", ' ', $text);
-                  $text = str_replace(" : ", ' ', $text);
-                  $text = str_replace(" — ", '', $text);
-                  $text = str_replace("|", '', $text);
-                  $text = strip_tags($text);       
-                  
+                  $text = normalize_ocr_text($media[1]->render());                  
                 endif;  
 
-                if($text):
-                  foreach($_GET["filter"] as $key => $value):
-                    if($value["join"] != "not" && $value["value"]):
-                      $search_text = htmlspecialchars(strip_tags($value["value"]));
-                      //$texts = explode($search_text,str_ireplace($search_text,$search_text,$text));
-                      $search_text = str_replace("-"," ",$search_text);
-                      $search_text_escaped = preg_quote($search_text, '/');
-                      $texts = preg_split("/\b".$search_text_escaped."\b/i", $text);
-                     
-                      if(sizeof($texts)>1):
-                        //$first = explode(" ",$texts[0]);
-                        $first = substr($texts[0],-45);
-                        //$last = explode(" ",$texts[1]);
-                        $last = substr($texts[1],0,45);
-                        echo "<div class='snippet'>";
-                        echo "...".$first."<span class='hit'>".$search_text."</span>".$last.'...';
-                        echo "</div>";
-                        break;
-                      endif;  
-                      $search_text = str_replace(" ","-",$search_text);
-                      $search_text_escaped = preg_quote($search_text, '/');
-                      $texts = preg_split("/\b".$search_text_escaped."\b/i", $text);
-                      if(sizeof($texts)>1):
-                        //$first = explode(" ",$texts[0]);
-                        $first = substr($texts[0], -45);
-                        //$last = explode(" ",$texts[1]);
-                        $last = substr($texts[1],0,45);
-                        echo "<div class='snippet'>";
-                        echo "...".$first."<span class='hit'>".$search_text."</span>".$last.'...';
-                        echo "</div>";
-                        break;
-                      endif;
-                    endif;  
-                  endforeach;
-                endif;
+                if ($text) {
+                  foreach ($_GET["filter"] as $key => $value) {
+
+                      if ($value["join"] != "not" && $value["value"]) {
+
+                          $search_text = mb_strtolower(strip_tags($value["value"]));
+
+                          $snippet = make_snippet_flexible($text, $search_text);
+
+                          if ($snippet) {
+                              echo "<div class='snippet'>{$snippet}</div>";
+                              break;
+                          }
+                      }
+                  }
+              }
               endif;  
             ?>
 
@@ -244,3 +214,60 @@ endif;
 
 </div></div>
 
+<?php 
+function normalize_ocr_text($text) {
+    // unify case
+    $text = mb_strtolower($text, 'UTF-8');
+
+    // fix word splits: "to-\nmaat" → "tomaat"
+    $text = preg_replace('/(\w)-\s*\n\s*(\w)/u', '$1$2', $text);
+
+    // remove remaining line-breaks
+    $text = preg_replace('/\s+/', ' ', $text);
+
+    // remove punctuation noise commonly introduced by OCR
+    $text = preg_replace('/[^\p{L}\p{N}\s]/u', '', $text);
+
+    $text = preg_replace('/[-‐-‒–—―]/u', ' ', $text); // all dash types → space
+
+    // normalize double spaces
+    $text = preg_replace('/\s+/', ' ', $text);
+
+    return trim($text);
+}
+
+function make_snippet_flexible($text, $search, $radius = 45) {
+
+    // normalize search term for consistent behavior
+    $search = mb_strtolower(trim($search));
+
+    // create variants
+    $search_space = preg_replace('/[-‐-‒–—―]/u', ' ', $search); // hyphens → spaces
+    $search_dash  = preg_replace('/\s+/', '-', $search_space);   // spaces → hyphens
+
+    // build regex (matches either version)
+    $pattern = 
+        '/(?<!\p{L})('
+        . preg_quote($search_space, '/') 
+        . '|'
+        . preg_quote($search_dash, '/')
+        . ')(?!\p{L})/iu';
+
+    if (!preg_match($pattern, $text, $m, PREG_OFFSET_CAPTURE)) {
+        return null;
+    }
+
+    $pos = $m[0][1];
+    $len = mb_strlen($m[0][0]);
+
+    // build snippet
+    $start = max(0, $pos - $radius);
+    $end   = $pos + $len + $radius;
+    $snippet = mb_substr($text, $start, $end - $start);
+
+    // highlight match
+    $snippet = preg_replace($pattern, '<span class="hit">$0</span>', $snippet);
+
+    return "…" . $snippet . "…";
+}
+?>