Faster backend algorithms

[pczarn]

At every parsing step, we can check whether the parse will be temporarily deterministic. We could also find all symbols that encode regular languages, and put them in one big NFA VM.

• regular grammars
  • DFA
  • NFA (perhaps not worth adding) (nondeterministic!)
• LL
  • classification
  • backend algorithm
• LALR
  • classification
  • backend algorithm
• LR
  • lane table method, or IELR

Of all grammar classes these algorithms work for, some will be subsets of others. For example, LL is contained in LR, and regular grammars (DFA) are contained in LL. Better only choose those algorithms that give substantial speedup, then.

Ensure that all of the following features do not collide with alternative algorithms: semantic actions, tracing, events, error messages, parse exhaustion, and LATM.