From Regulars to ANTLR4: Three Architectural Decisions on Business Formula Parsing

Three years ago, I developed my first commercial parser using ANTLR4 for an analytical system focused on business formulas. In hindsight, I recognize that there are three significant aspects I would approach differently today. The first is the implementation of left recursion to manage operator precedence more effectively. This adjustment would enhance the parser's ability to handle complex expressions. The second area involves the choice between using Listener and Visitor patterns. Each has its advantages, but understanding the context of the parsing task is crucial for making the right decision. Lastly, I want to address the use of getText() for re-parsing, which I now consider an anti-pattern. This practice can lead to inefficiencies and complications in the parsing process. By revisiting these areas, I believe a more robust and efficient parser could be developed, ultimately improving the analytical capabilities of the system.