-A
When a syntax error is detected, there are, in general, several ways to repair it. Most compilers will select a particular repair by using heuristic criteria of repair cost and/or validation. Some authors have argued that merely putting a sequence of marks into an incorrect program such that each text between marks is individually guaranteed to be a substring of SOME syntactically correct program, provides a more useful output of the parser - mainly because incorrect repair choices CANNOT occur in this technique. The approach has been called 'non-correcting recovery'. Until recently, the time demand of doing this kind of parsing was an open problem. Rather surprisingly, some new results have led to the observation that linear time suffices. The speaker will try to present this talk with as few technical details as possible.