FIX: improved category merging in discourse_merger. Use full paths to look for uniqueness instead of category names.

script/bulk_import/discourse_merger.rb

@@ -177,9 +177,25 @@ class BulkImport::DiscourseMerger < BulkImport::Base
 
     sql = "COPY categories (#{columns.map { |c| "\"#{c}\"" }.join(', ')}) FROM STDIN"
     @raw_connection.copy_data(sql, @encoder) do
-      source_raw_connection.exec("SELECT #{columns.map { |c| "\"#{c}\"" }.join(', ')} FROM categories").each do |row|
+      source_raw_connection.exec(
+          "SELECT concat('/c/', x.parent_slug, '/', x.slug) as path,
+                  #{columns.map { |c| "c.\"#{c}\"" }.join(', ')}
+             FROM categories c
+       INNER JOIN (
+              SELECT c1.id AS id,
+                     c2.slug AS parent_slug,
+                     c1.slug AS slug
+                FROM categories c1
+     LEFT OUTER JOIN categories c2 ON c1.parent_category_id = c2.id
+                  ) x ON c.id = x.id"
+        ).each do |row|
 
-        if existing = Category.where(name: row['name']).first
+        source_category_path = row.delete('path')&.squeeze('/')
+
+        existing = Category.where(slug: row['slug']).first
+        parent_slug = existing&.parent_category&.slug
+        if existing &&
+            source_category_path == "/c/#{parent_slug}/#{existing.slug}".squeeze('/')
           @categories[row['id']] = existing.id
           next
         end