And it's well worth the effort - I've already found a few stupid typos and such.
The "stacked models" branch is of course still there too. I'm most likely going to cherry-pick things from it, once this release has settled a bit.
I'm currently working on merging the `xstream-new-buffer-implementation` branch into the master
didn't know what branch to start from.
- mostly working on this locally, but already made a few commits to the master, I think.
This is the milestone I'm working towards, btw.
Re the sandbox: yeah, how it was _planned_ to work, and how it _actually_ ended up being used are two slightly different things.
I think by now you've been diving deeper into lua meta-methods than I have, so I'm keen to hear your input/opinions.
AFAIK the "table as a switch" (using associative, string based keys for access) is very efficient in lua, comparable with environment tables.
But perhaps you have evidence that shows otherwise? I must admit, profiling code is often the last thing I do ...