An empirical evaluation of Lex/Yacc and ANTLR parser generation tools

Area of Science:

• Computer Science
• Software Engineering
• Programming Language Theory

Background:

• Parsers are fundamental in software development for tasks like compiler construction and natural language processing.
• Lex/Yacc and ANTLR are prominent parser generator tools, with Lex/Yacc often favored in academic settings despite ANTLR's advanced features.
• Existing comparisons are qualitative; empirical data on implementor productivity and tool usability is lacking.

Purpose of the Study:

• To empirically evaluate and compare the performance of Lex/Yacc and ANTLR.
• To assess empirical features including language implementor productivity, tool simplicity, intuitiveness, and maintainability.
• To provide data-driven insights for selecting parser generator tools in academic and practical contexts.

Main Methods:

• An experiment was conducted with undergraduate Software Engineering students.
• Students were randomly assigned to implement the same language using either Lex/Yacc or ANTLR.
• Statistical analysis was performed on various performance measures to compare the two tools.

Main Results:

• ANTLR showed statistically significant improvements across most measured empirical features compared to Lex/Yacc.
• The study identified key differences in implementor productivity and perceived tool quality.
• Findings suggest ANTLR offers a more advantageous development experience in the evaluated aspects.

Conclusions:

• ANTLR presents significant empirical advantages over Lex/Yacc in terms of productivity and usability for parser generation.
• The findings challenge the traditional preference for Lex/Yacc in some academic curricula.
• This study provides valuable empirical evidence for tool selection in software engineering education and practice.