Ivan Kochurkin

> No. I don't understand, the first is safe or the second is also should be safe for removing? > Out of curiosity, how did you generate the diagrams? The...

> Hiya! What problem are we trying to solve here? Improving quality and performance of ANTLR. Especially considering current PR also fixes some potential bugs that I've described above. For...

> Why don't we just throw the optimization implemented in this pull request behind a command line flag? Something like -remove-redundant-epsilon-transitions? * It compicates testing (with or without option) *...

> Sam, how can an ATN state which has exactly one incoming and one outgoing transition, with no labels attached to them impact anything surrounding it? Actually, optimizer also relinks...

> I would be much more interested in knowing how it affects lexing performance across a variety of problems and across targets. Ok, it makes sense to write some benchmark...

> @KvanTTT I was wondering about the new 15-state ATN (in https://github.com/antlr/antlr4/pull/3676#issue-1213298256) why states 11 and 12 are not removed and replaced with a single transition for the non-terminals C...

Let's move discussion about inlining to the separated issue: #3685

I think it's up for runtime.

Yes, separated class looks better.

Thanks, but it should be (if it will) implemented via ANTLR tool like https://github.com/antlr/antlr4/pull/3701