Re: Recent releases of ConTeXt LMTX crash at \starttext when expl3 is loaded

On 05/08/2026 09:13, Max Chernoff via ntg-context wrote:
> Hi Vit, Hans,
> 
> On Tue, 2026-08-04 at 10:45 +0000, [email protected] wrote:
>> Cross-format modules like the Markdown package for TeX often benefit
>> from using expl3 instead of TeX,
> 
> The fundamental problems here are that:
> 
> 1. expl3 is relatively slow to load from scratch.
> 2. expl3 is not preloaded into the ConTeXt format.
> 
> I don't think that there's anything that can be done about (1), since
> expl3 isn't slow because it's bloated or poorly-written, but because it
> offers so much functionality. So unless expl3 loses functionality
> somehow, there's probably no way to fix this. But the expl3 load times
> in ConTeXt are _so_ much better since Joseph reimplemented the Unicode
> table stuff in Lua, so maybe there are some other similar places that
> can be sped up.

Marcel did that :-)

The change was how we parse the Unicode data files: in classical 
engines, we have no option but to do it in macros and store as 
\fontdimen data. With Lua available, parsing into some tables is doable 
from the Lua side. (This is for data that is not stored by the engine 
itself, so for example case changing data split out from 
\lccode/\uccode, as these are also needed for hyphenation control, and 
that can't be managed at the macro level.)

I don't think the same gains are available elsewhere. For me, loading 
expl3 into ConTeXt or plain LuaTeX takes <1 s nowadays.

> I don't think that there's anything to be done about (2) either: Hans
> has been focusing on reducing format build times, and adding expl3 would
> slow that down again; hardly any ConTeXt packages use expl3, so there's
> not much benefit to adding it; and expl3 conflicts with a few ConTeXt
> features and is occasionally broken, so adding it to the ConTeXt format
> might break things for non-expl3 users too.
It's not just ConTeXt, of course: any change in data loading would also 
need to be done for plain, OpTeX, etc., and the probability of agreement 
is very low. (We use the same data loading code in LaTeX and other 
formats, it's just that in LaTeX that's done in format-building and 
dumped. We really don't want multiple code paths.)

Joseph
___________________________________________________________________________________
If your question is of interest to others as well, please add an entry to the Wiki!

maillist : [email protected] / https://mailman.ntg.nl/mailman3/lists/ntg-context.ntg.nl
webpage  : https://www.pragma-ade.nl / https://context.aanhet.net (mirror)
archive  : https://github.com/contextgarden/context
wiki     : https://wiki.contextgarden.net
___________________________________________________________________________________