Softacom Transformed a Delphi XE4 System for MSSQL, PostgreSQL, and Linux

The company was seeking to modernize its database access layer, business logic, and runtime, as well as reduce the technical debt. After analyzing their request, we offered solutions to address the client’s challenges. We designed a modernization plan.

The Softacom team assessed the entire codebase of the Delphi XE4 app, consisting of 2510409 lines, along with its database and infrastructure. After the analysis, the following results on the application statistics were received:

• pas-files (incl. 3rd-party components): 3698
• pas-files (excl. 3rd-party components): 3395
• dfm-files (incl. 3rd-party components): 666
• dfm-files (excl. 3rd-party components): 666

Statistics of existing DLL:

• Total units: 353;
• Total lines: 1 416 608;
• pas-files (incl. 3rd-party components): 353;
• pas-files (excl. 3rd-party components): 117;
• dfm-files (incl. 3rd-party components): 33;
• dfm-files (excl. 3rd-party components): 33;

Statistics of DLL with code related to data only. Code that needs to be migrated to Linux:

• Total units: 235;
• Total lines: 769434;
• pas-files (incl. 3rd-party components): 235;
• pas-files (excl. 3rd-party components): 59;
• dfm-files (incl. 3rd-party components): 20;
• dfm-files (excl. 3rd-party components): 20.

The application was used for day-to-day payroll & HR activities. 

As we worked on the project, it became clear that the client had limited familiarity with certain aspects of their application. This required our team to conduct extensive research to fully understand the system’s architecture and functionality.

Our team identified several complexities in the modernization process:

• Delphi XE4 intricacies – to better understand how its specific implementations affected performance and maintainability. 
• Database migration risks – to transition from the legacy database access layer to UniDAC for MSSQL and PostgreSQL. It required evaluating differences in SQL dialects. For example, string functions – LEN() in MSSQL and LENGTH() in PostgreSQL, or auto-increment columns – IDENTITY in MSSQL and SERIAL in PostgreSQL. 
• Concurrency handling – PostgreSQL manages transactions differently than MSSQL. So, when moving to it, poorly optimized transactions could cause performance bottlenecks. Our team needed to conduct a careful compatibility assessment of queries, functions, and procedures.  

To bring the software to modern standards:

• The Softacom team of developers replaced the outdated database access technology with UniDAC. This replacement helped make the library compatible with MSSQL and PostgreSQL databases and addressed varying database behaviors. 
• Converting large datasets and keeping the app operational at the same time is challenging. To balance these processes, we used several approaches, such as Change Data Capture (CDC) and Dual-Write Mechanism. 

“CDC allows for real-time synchronization and reduces downtime. It also works well with large datasets because it doesn’t require bulk transfers. The dual-write mechanism ensures data consistency, minimizing loss and enabling a gradual migration.”
– Project Manager

• The Softacom team transformed the software into a multi-platform solution by adding support for Linux. It required an evaluation of the existing codebase to determine which parts could be migrated and which needed to be re-engineered. Since Delphi XE4 doesn’t provide support for Linux, our team refactored the Windows-specific parts of the app to make it compatible with Linux. That included rewriting 769434 lines of code. The Linux support ensured adaptability across diverse operating systems. 
• To enhance code optimization and streamline debugging, we developed a unit testing framework, which would maintain development and mitigate risks in future iterations.
• We also automated part of the migration with our in-house database conversion tool. It accelerated 30% of the process, which allowed us to finish the project 3 months earlier. 

The whole project took 9 months to complete.