Modernizing a Complex Legacy Codebase
Our client, a leading global management consulting firm, needed to modernize their legacy T-SQL codebase, which included numerous large and complex stored procedures. They were looking for a solution that would preserve the extensive business logic that had accumulated over the years, but they wanted to avoid a resource-intensive and time-consuming rewrite.
The existing system required significant manual effort for code transformation, testing, and documentation. The client sought to streamline the process, reducing time and costs while ensuring high code quality and functionality.
Automating the Transformation
Ascendion addressed these challenges by deploying our GenAI platform, Ascendion AVA+ Digital Ascender. This platform utilized GenAI-driven automated code transformation by seamlessly integrating all essential components.
The approach encompassed these essential actions:
- Direct transformation of T-SQL code to Java, using both single and multi-instruction prompting (conversational and non-conversational) addressed various coding patterns.
- Key steps included prompt management, model selection, contextualization, and a structured conversion workflow.
- The Digital Ascender ecosystem and IDE plugins enabled iterative code refinement, unit testing, and tuning, and helped implement coding best practices post-transformation.
- GenAI capabilities enabled the generation of automated tests and documentation, drastically minimizing manual intervention.
The results:
- There were effort savings of approximately 50% overall and 90% in code creation compared to traditional methods.
- Initial analysis time was reduced from 2-3 days to about 1 hour.
- Code transformation time was drastically shortened from 5 days to just 4-7 hours.
- Fine-tuning and optimization were streamlined to under 30 minutes.
- Unit test case generation and documentation time were reduced to approximately 1 hour and 5-10 minutes, respectively.
- The modernization process yielded high reusability and minimal effort, setting a new benchmark for legacy code transformation.
