-
Improve definition parse efficiency by using slices more than vectors.
-
Add useful builtins (especially whitespace, lowercase, uppercase, numbers).
-
Actually do parser validation.
-
Improve actual parse efficiency by using slices more than vectors
-
Merge parser processes that reach the same state, or maybe just declare an ambiguous parse?
-
Handle ambiguous parse are failed parse. Ideally a failed parse should identify where the parse goes wrong, i.e. how far along parsing stopped.
-
Calls to
matches()could be memoized, though it is unclear if this would be worth it for most users. -
Nota Bene: With current algorithm, rules can be skipped in final parse tree if surrounding Optional or Many operators consume no tokens. I guess this is ultimately fine, afterall the alternative seems to be letting certain recursive formulae have infinite parse trees: " Expr : Term Expr? "
- We could kinda solve this though if we examined the backtrace in a more principled way, examining every node insteaed of just ancestors of terminals. Maybe we should actually do that. Also, I don't the example I cited was actually problematic.
PS: Here is a C grammar for when we really want to stress test: