Error recovery and LR(1)/LALR(1)

Hi,


I am in the process of writing an LR(1)/LALR(1) parser generator called
YooParse, which will be used with YooLex, a C++ lexer generator. While I
successfully generated the LR(1)/LALR(1) DFA states, I encountered
problems dealing with error states. This is a long post.


Situation 1:
For example, a DFA state contains the following LR(1) items
A -> alpha X . , lookahead = 'a'
B -> beta X . gamma , lookahead = 'b'
My question is really what to do if the lookahead is neither 'a'
or 'b'? Should A be reduced?


Situation 2:
For example, a DFA state contains the following LR(1) items
A -> alpha X . , lookahead = 'a'
B -> beta X . , lookahead = 'b'
What if the lookahead is neither 'a' nor 'b'? Do I assume default
transitions on A since the rule for A preceeds B?


My concern is the loss of chance of early recovery.


Thanks


Heng Yuan
heng@ag.arizona.edu

Return to the comp.compilers page.
Search the comp.compilers archives again.