Bruno Le Floch
Bruno Le Floch
This issue also affects me.
A similar comment holds in `xunicode-addon.sty`, which if I understand correctly is also supported by this organization on Github. Let me know if I should open a separate issue.
I'm worried that this would add 5 commands (`:nnF`, `:nnN`, `:nnNT`, `:nnNF`, `:nnNTF`) for each of N/c/n-type tl/clist/str and N/c type seq/prop, leading us to add 5×(3×3+2×2)=65 control sequences, plus...
I think this is a confusing choice and I agree that `"Abc"` would be more expected, namely using ``\int_compare:nTF { 11 = \char_value_catcode:n {`#1} }`` instead of `\token_if_eq_catcode:NNTF X #1`
Check if `[a-zA-Z\c{alpha|beta|gamma}]` can be optimized (I'm not listing all greek letters here for brevity).
I strongly suspect there are still important bugs. Sorry no time atm.
> 3. How much slower does lualatex start up if it doesn't do the byte > compiling? Plausible to discern? IIRC we saved around 0.6 seconds on a reasonably good...
The current date-checking code is pretty much a hack, using dumb integer comparison to compare dates, so it's not very easy to allow a one-day difference.
Related to `l3tl_analysis` too: in platex and uplatex the kanji characters should be described with their kcatcode since the meaning does not include this informatioin.
Hopefully such a reimplementation of `l3tl-analysis` can also avoid issues with the special LuaTeX "bytes" such as `^^^^^^1100d3` which are tokens to which one cannot safely apply `\string` or `\meaning`...