fix: ignore '&' tags in discovery and relation aggregation (#57)

- Add WHERE tag NOT LIKE '%&%' in getTagForDiscovery
- Skip '&' tags in suggestion parsing and source picking
- Exclude them from relation aggregation and updates
- Early-return in processDiscoveredTag to prevent propagation
- Provide SQL to delete existing '&' tag rows across related tables

Closes #56

ZerGo0 Bot

Co-authored-by: ZerGo0 <ZerGo0@users.noreply.github.com>

Author: ZerGo0

backend-go/handlers/tag_handler.go

@@ -260,28 +260,28 @@ func (h *TagHandler) GetTags(c *fiber.Ctx) error {
 					ChangePercent:   0,
 				}
 
-        // Calculate change from previous point based on view count
-        if i < len(history)-1 {
-            previousPoint := history[i+1]
-            // Calculate view count change as primary metric
-            viewChange := point.ViewCount - previousPoint.ViewCount
-            historyPoints[i].Change = viewChange
-
-            // Still track post count change for reference
-            postChange := point.PostCount - previousPoint.PostCount
-            historyPoints[i].PostCountChange = postChange
-            if previousPoint.ViewCount > 0 {
-                historyPoints[i].ChangePercent = float64(viewChange) / float64(previousPoint.ViewCount) * 100
-            }
-        }
-        }
-
-        // Calculate total change based on view count
-        if len(history) > 0 {
-            newest := history[0].ViewCount
-            oldest := history[len(history)-1].ViewCount
-            tagWithHist.TotalChange = newest - oldest
-        }
+				// Calculate change from previous point based on view count
+				if i < len(history)-1 {
+					previousPoint := history[i+1]
+					// Calculate view count change as primary metric
+					viewChange := point.ViewCount - previousPoint.ViewCount
+					historyPoints[i].Change = viewChange
+
+					// Still track post count change for reference
+					postChange := point.PostCount - previousPoint.PostCount
+					historyPoints[i].PostCountChange = postChange
+					if previousPoint.ViewCount > 0 {
+						historyPoints[i].ChangePercent = float64(viewChange) / float64(previousPoint.ViewCount) * 100
+					}
+				}
+			}
+
+			// Calculate total change based on view count
+			if len(history) > 0 {
+				newest := history[0].ViewCount
+				oldest := history[len(history)-1].ViewCount
+				tagWithHist.TotalChange = newest - oldest
+			}
 
 			if includeHistory {
 				tagWithHist.History = historyPoints

backend-go/workers/statistics_calculator.go

@@ -76,15 +76,15 @@ func (w *StatisticsCalculatorWorker) calculateTagStatistics(ctx context.Context)
 		return fmt.Errorf("failed to fetch previous statistics: %w", err)
 	}
 
-    // Calculate 24-hour changes based on view count
-    var change24h int64
-    var changePercent24h float64
-    if previousStats.ID != 0 {
-        change24h = totalViewCount - previousStats.TotalViewCount
-        if previousStats.TotalViewCount > 0 {
-            changePercent24h = (float64(change24h) / float64(previousStats.TotalViewCount)) * 100
-        }
-    }
+	// Calculate 24-hour changes based on view count
+	var change24h int64
+	var changePercent24h float64
+	if previousStats.ID != 0 {
+		change24h = totalViewCount - previousStats.TotalViewCount
+		if previousStats.TotalViewCount > 0 {
+			changePercent24h = (float64(change24h) / float64(previousStats.TotalViewCount)) * 100
+		}
+	}
 
 	// Create new statistics record
 	newStats := models.TagStatistics{

backend-go/workers/tag_discovery.go

@@ -150,19 +150,20 @@ func (w *TagDiscoveryWorker) Run(ctx context.Context) error {
 }
 
 func (w *TagDiscoveryWorker) getTagForDiscovery() (string, error) {
-	// First, try to get a tag from the database that hasn't been used recently
-	hoursAgo := time.Now().Add(-3 * time.Hour)
-
-	var tags []models.Tag
-	// Get multiple tags that haven't been used recently and aren't deleted, ordered by rank
-	if err := w.db.Where("(last_used_for_discovery IS NULL OR last_used_for_discovery < ?) AND is_deleted = ?", hoursAgo, false).
-		Order("rank ASC").
-		Limit(10).
-		Find(&tags).Error; err == nil && len(tags) > 0 {
-		// Select a random tag from the top 10 to avoid always picking the same one
-		idx := time.Now().Unix() % int64(len(tags))
-		return tags[idx].Tag, nil
-	}
+    // First, try to get a tag from the database that hasn't been used recently
+    hoursAgo := time.Now().Add(-3 * time.Hour)
+
+    var tags []models.Tag
+    // Get multiple tags that haven't been used recently and aren't deleted, ordered by rank
+    if err := w.db.Where("(last_used_for_discovery IS NULL OR last_used_for_discovery < ?) AND is_deleted = ?", hoursAgo, false).
+        Where("tag NOT LIKE ?", "%&%").
+        Order("rank ASC").
+        Limit(10).
+        Find(&tags).Error; err == nil && len(tags) > 0 {
+        // Select a random tag from the top 10 to avoid always picking the same one
+        idx := time.Now().Unix() % int64(len(tags))
+        return tags[idx].Tag, nil
+    }
 
 	// If no suitable database tag, use a random seed tag
 	if len(w.seedTags) > 0 {
@@ -176,17 +177,18 @@ func (w *TagDiscoveryWorker) getTagForDiscovery() (string, error) {
 
 // extractTagsFromSuggestions extracts unique tags from media offer suggestions
 func (w *TagDiscoveryWorker) extractTagsFromSuggestions(suggestions []fansly.MediaOfferSuggestion) []fansly.FanslyTag {
-	tagMap := make(map[string]fansly.FanslyTag)
-
-	for _, suggestion := range suggestions {
-		for _, tag := range suggestion.PostTags {
-			// Use the tag name as key to ensure uniqueness
-			tagName := strings.ToLower(strings.TrimSpace(tag.Tag))
-			if tagName != "" {
-				tagMap[tagName] = tag
-			}
-		}
-	}
+    tagMap := make(map[string]fansly.FanslyTag)
+
+    for _, suggestion := range suggestions {
+        for _, tag := range suggestion.PostTags {
+            // Use the tag name as key to ensure uniqueness
+            tagName := strings.ToLower(strings.TrimSpace(tag.Tag))
+            // Ignore tags that contain '&' as Fansly truncates them before '&'
+            if tagName != "" && !strings.Contains(tagName, "&") {
+                tagMap[tagName] = tag
+            }
+        }
+    }
 
 	// Convert map to slice
 	tags := make([]fansly.FanslyTag, 0, len(tagMap))
@@ -198,11 +200,15 @@ func (w *TagDiscoveryWorker) extractTagsFromSuggestions(suggestions []fansly.Med
 }
 
 func (w *TagDiscoveryWorker) processDiscoveredTag(tag fansly.FanslyTag) error {
-	// Check if tag already exists
-	var existingTag models.Tag
-	if err := w.db.Where("tag = ?", tag.Tag).First(&existingTag).Error; err == nil {
-		// Tag already exists
-		return nil
+    // Ignore tags that contain '&' as Fansly truncates them before '&'
+    if strings.Contains(tag.Tag, "&") {
+        return nil
+    }
+    // Check if tag already exists
+    var existingTag models.Tag
+    if err := w.db.Where("tag = ?", tag.Tag).First(&existingTag).Error; err == nil {
+        // Tag already exists
+        return nil
 	}
 
 	// Create new tag using the data we already have
@@ -243,14 +249,18 @@ func (w *TagDiscoveryWorker) updateTagRelationsFromSuggestions(ctx context.Conte
 	counts := make(map[key]int64)
 
 	for _, s := range suggestions {
-		// Build a unique set of tag IDs observed in this suggestion
-		seen := make(map[string]struct{})
-		for _, t := range s.PostTags {
-			id := strings.TrimSpace(t.ID)
-			if id == "" {
-				continue
-			}
-			seen[id] = struct{}{}
+        // Build a unique set of tag IDs observed in this suggestion
+        seen := make(map[string]struct{})
+        for _, t := range s.PostTags {
+            // Skip tags that contain '&' to avoid polluted relations
+            if strings.Contains(strings.ToLower(strings.TrimSpace(t.Tag)), "&") {
+                continue
+            }
+            id := strings.TrimSpace(t.ID)
+            if id == "" {
+                continue
+            }
+            seen[id] = struct{}{}
 		}
 		// For all unique tags on this suggestion, generate directed pairs A->B (A != B)
 		if len(seen) == 0 {