GeezerVison

We need to rework the whole keyword management eventually. But in the meantime this can be a remedy.

Hi, @manticore-projects
I also agree that the current keyword handling could use a more systematic approach.
Out of curiosity, could you share your thoughts on what the ultimate direction for keyword management should look like? I'd like to align any future refactoring efforts accordingly.

The current implementation has grown more or less wild over 20 years and based on demand. It also suffers limitations from JavaCC 7.
At the same time it somehow just works good enough for most use-cases and rewriting it will take a lot of effort.

So my current idea was:

1. switch to JavaCC 8 (or even CongoCC)
2. then rework the DDL statements (they are a big mess)
3. then rewrite the KEYWORD handling

Hi, @manticore-projects
This sounds like quite a large-scale refactoring effort. I’m curious — do you have any rough timeline or plan in mind for when you'd like to tackle this?

Also, just wondering, have you ever considered adopting something like ANTLR instead, or is the plan to stick with JavaCC/CongoCC going forward?

ANTLR has been found producing slower code and so is out of scope for me. Although this claim could need a verification again.

Also, when rebasing to ANTLR then we would/should have to rewrite from scratch.

@manticore-projects
It's interesting that ANTLR was found to be slow. I wonder if that has something to do with the structural differences between LL and LR parsers. Of course, replacing ANTLR would come with the high cost of building everything from scratch, but I thought one of its strengths was the ability to support multiple languages.

I am not doubting the strength or superiority of ANTLR in general but only point out why I was more interested in sticking with the JavaCC origin as long as possible. JavaCC/CongoCC also supports multiple languages although I am really interested in Java only (and maybe C).

For my own purpose (real time parsing SQL while typing) I am interested in the fastest (automatically generated) parser possible -- and so far JavaCC holds that position.