Analysis
...
v4 uses adaptive LL(*) or ALL(*). It is pronounced "all star". 
Parsing
- Left recursion elimination and expression grammar simplification [ANTLR4:DONE] (Actually tested in v3 playground, not in v4 yet)
- Adaptive LL(*), ALL(*) "all star", parsing DONE
- Scannerless parsing
- Error alternatives
- Maybe ambiguous alt parsing, yielding parse forest
- Improve tree pattern matching, rewrites; e.g., <"0":primary>
- syntactic predicates ala PEGs; needed in token rules. ~ and ! are taken at the moment. perhaps the negate operator '-LetterOrDigit'; might as well at the positive version also: &LetterOrDigit.
Lexing
- DFA-based / NFA-based Lexers
- Lexer modes / contexts. "context STRING"
- Long int based tokens instead of objects; use token adaptor
- Unbuffered I/O
...
- Generate code that uses ST v4 [ANTLR4:DONE]
- Make default output=template action for single rule ref to be that rule's return value then perhaps $text for more complicated alts w/o -> actions.
Actions
- Allow $FIRST, $FOLLOW, $LOOK
Runtime
- Improve error messages when backtracking or using arbitrary lookahead
- Add Jim Idle's sync rule concept
- Make a real interpreter using NFA; allow access as library from Java code
- Tokens and Trees should both know their start/stop line, start/stop char position to make IDEs easier.
- Tree parser error handling should skip subtrees not nodes; these are programming errors not input errors. The flat stream makes it hard to resync.
...
- Use ST v4 to generate code [ANTLR4:DONE]
- Create new code generation mechanism
...