ActiveRecord `schema_migrations` store a hash of the migration text

parsonsmatt · August 24, 2023, 11:22pm

We use ActiveRecord migrations for our application. I’m trying to optimize our test initialization time by avoiding running migrations when possible. A common pattern that happens is a developer will introduce a new migration, and then modify the file to iterate on the schema while working on their feature. Rails sees that the migration timestamp is present in the schema_migrations table, and won’t re-run the migration.

If ActiveRecord also stored a hash of the migration text (ie all the code that Rails sends to the database server), then it could be determined whether or not a migration has changed from when Rails initially ran the migration. This would be helpful for our cache avoidance - we’d check each migration’s presence in the version and then if the hashes match. If they match, then we have a good cache, and can avoid redoing work.

rafaelfranca · August 24, 2023, 11:50pm

Maybe I’m missing something but Rails by default doesn’t check the schema_migrations table when maintaining the test schema.

Instead, it checks if the hash of the schema file is the same as the one that is in its internal metadata table.

github.com

rails/rails/blob/c35b45b3f41c5f3bf519c151f4ed18dc510047bd/activerecord/lib/active_record/tasks/database_tasks.rb#L402


      
          
            file ||= schema_dump_path(db_config)
          
            return true unless file && File.exist?(file)
          
            ActiveRecord::Base.establish_connection(db_config)
          
            return false unless ActiveRecord::InternalMetadata.enabled?
            return false unless ActiveRecord::InternalMetadata.table_exists?
          
            ActiveRecord::InternalMetadata[:schema_sha1] == schema_sha1(file)
          end
          
          def reconstruct_from_schema(db_config, format = ActiveRecord.schema_format, file = nil) # :nodoc:
            file ||= schema_dump_path(db_config, format)
          
            check_schema_file(file) if file
          
            ActiveRecord::Base.establish_connection(db_config)
          
            if schema_up_to_date?(db_config, format, file)

Isn’t that cache already sufficient? If you changed the schema of your database Rails will load the new schema. If you did, Rails will not run any migration.

github.com

rails/rails/blob/c35b45b3f41c5f3bf519c151f4ed18dc510047bd/activerecord/lib/active_record/migration.rb#L665-L686


      
          def load_schema_if_pending!
            current_db_config = Base.connection_db_config
            all_configs = ActiveRecord::Base.configurations.configs_for(env_name: Rails.env)
          
            needs_update = !all_configs.all? do |db_config|
              Tasks::DatabaseTasks.schema_up_to_date?(db_config, ActiveRecord.schema_format)
            end
          
            if needs_update
              # Roundtrip to Rake to allow plugins to hook into database initialization.
              root = defined?(ENGINE_ROOT) ? ENGINE_ROOT : Rails.root
              FileUtils.cd(root) do
                Base.clear_all_connections!
                system("bin/rails db:test:prepare")
              end
            end
          
            # Establish a new connection, the old database may be gone (db:test:prepare uses purge)
            Base.establish_connection(current_db_config)

This file has been truncated. show original

Or are you seeing a different behavior?