Robert Muir

I'm not worried, according to the JEP: https://openjdk.org/jeps/411 ``` In feature releases after Java 18, we will degrade other Security Manager APIs so that they remain in place but with...

This change looks good to me! I just want to clone the branch and try to figure how you removed 25,000 lines of code from the generated jflex :)

I don't think you did anything wrong. The changes look fine, it is just a real puzzler. I even cloned your repo and inspected the enormous diff to the generated...

looks like an antlr problem, if they broke backwards compat, they prolly should have named it `5.x`? let's be careful about upgrading to new versions. newer antlr versions have historically...

i'd prefer not changing anything without addressing the testing. I need to reiterate just how insanely trappy antlr v4 is. for painless to work with v4 and prevent insanely slow...

An alternative solution could be to only tableize latin1 or the bmp (int[128] or int[64k]) and use binary search for supplementary characters. For example: ``` if (cp < table.length) {...

We hit this issue too with apache lucene :) One practical problem with the current `@SuppressWarnings("FallThrough")` is that this isn't recognized by `javac` in this mixed-case form. So we had...

> Just had a look at `UAX29URLEmailTokenizerImpl.jflex`: it uses `!` to work around the no-macros-in-charclasses restriction. This is now implemented in JFlex master (#654) and will be in the next...

one issue with "fancy" copies (e.g. for bulk merging) is that we can't make them super-fancy anyway (e.g. zero-copy from fd to fd) as we still have to calculate CRC32...

the performance with e.g. 768 is incredibly painful. hours and hours to index just 1M documents. Already doesn't scale with the current limit!