METHOD OF DML OPERATION OPTIMIZATION IN SALESFORCE BASED ON THE UNIT OF WORK PATTERN: AN ARCHITECTURAL APPROACH TO TRANSACTIONAL CHANGE CONSOLIDATION

Abstract

The article presents a method for optimizing Data Manipulation Language (DML) operations on the Salesforce platform based on the Unit of Work architectural pattern. The multi-tenant architecture of the Salesforce platform imposes strict constraints on the number of DML operations within a single transaction: a maximum of 150 statements for the synchronous execution context. As business logic complexity and trigger handler count grow, each handler independently executes its own insert, update, and delete operations, causing linear growth of DML resource consumption and increasing the risk of transaction failures. The paper formalizes a mathematical model of DML operation consumption describing the dependency of statement count on the number of active handlers, the number of object types, and the transactional coupling coefficient. The proposed method introduces a centralized change management layer that accumulates all pending data modifications from business handlers in a change registry and executes them in consolidated form at the end of the transaction while respecting dependency ordering between objects. It is proven that under the Unit of Work pattern, the DML statement count is determined solely by the number of unique combinations of object type and operation kind and does not depend on the number of business handlers, which provides mathematically guaranteed compliance with platform constraints. Experimental validation was conducted in a Salesforce Developer Edition environment on a test bench with five business handlers during bulk processing of 200 records. Experimental results confirmed a 73% reduction in DML statement count compared to the traditional approach, a 42% reduction in transaction execution time, and a 28% reduction in CPU time consumption. The proposed method, combined with the previously developed data access centralization methodology, forms a complete approach to transactional resource management covering both data read and data write operations.