bug#32599: 25.2; Feature request: input PUA characters by name

[Eli Zaretskii]

> From: jsbien@mimuw.edu.pl (Janusz S. Bień)
> Cc: 32599@debbugs.gnu.org
> Date: Sun, 26 May 2019 17:48:43 +0200
> 
> >> The UnicodeData.txt file is compiled into Emacs,
> >
> > I know and I'm curious whether it is really needed.
> 
> It was a bad wording as I can guess it is done for efficiency.

Indeed.  Both processing efficiency and (perhaps more importantly)
memory efficiency.  Some of the resulting data is accessed by core
features that must be very efficient: the display engine, the regexp
search functions, etc.

> How the character data are stored and accessed?

The data is stored in char-tables of special kind.  See the files
uni-*.el which are produced as part of the Emacs build process.

> > Why it cannot be loaded at the startup?
> 
> Or, better, partially overriden or supplemented? Is is prohibited by the
> way the character data are stored?

You can override the char-tables at run time, of course, but first you
need to generate the new ones.

And there's the other problem I mentioned: with general treatment of
PUA codepoints.