History-Sensitive Error Recovery

We present a novel approach to incremental recovery from lexical and syntactic errors in an interactive software development environment. Unlike existing techniques, we utilize the history of changes to the program to discover the natural correlation between user modifications and errors detected during incremental lexical and syntactic analysis. Our technique is non-correcting—the analysis refuses to incorporate invalid modifications, while still permitting correct changes to be applied. Errors are presented to the user simply by highlighting the invalid changes. The approach is automated—no user action is required to detect or recover from errors. Multiple textual and structural edits, arbitrary timing of incremental analysis, multiple errors per analysis, and nested errors are supported. Historybased error recovery is language independent and is compatible with the best known methods for incremental lexing and parsing, adding neither time nor space overhead to those algorithms. Effective integration with the environment’s history services ensures that other tools can efficiently discover regions of the program (un)affected by errors, and that any transformations of the program required to isolate or present errors are themselves efficiently reversible operations. Keywords— Error recovery, software development environments, incremental parsing, incremental lexing, development log, program presentation