cpp lexer adjustments

[snoopy]

I'm trying to make some adjustments to the cpp lexer and ran into some problems.

I'd like to remove the highlight for std:: from STL types. Highlighting that part only increases clutter and takes away from the info you are looking for IMO.
Removing the std:: part from the lexer works but will also highlight unrelated tokens of the same name. Is there a way to match the full token but only highlight the part without std::? E.g. match std::vector but only highlight vector.

Related to this question, I also want to highlight custom types at the end of a scope chain. E.g. in foo::bar::type, type should be highlighted. This can be achieved with a rule like the following:

lex:add_rule('custom_type', lex:tag(lexer.TYPE .. '.custom', '::' * lexer.word * -S('(:')))

The problem is that this also matches the leading :: part which it should not. Any way to avoid this? I've tried #('::') and #P('::'), both of which produce wrong results.

---

Some other questions that might make sense for the default lexer as well.

• How to highlight angle braces?
  • In C++ it makes sense to highlight < and > but it doesn't work out of the box. I've tried adding lexer.property['scintillua.angle.braces'] = '1' which will do the trick but also adds autopairing which is not wanted here.
• Should template functions be highlighted as such?
  • E.g. things like std::make_shared<int>() are currently not highlighted as functions

Lastly, are there still plans to use treesitter over or maybe alongside lua lexers? @orbitalquark

[orbitalquark]

To match 'std::' but not highlight it, you could do something like this:

lex:tag(lexer.DEFAULT, 'std::') * lex:tag(lexer.TYPE .. '.stl', lex:word_match(lexer.TYPE .. '.stl'))

For the related question, maybe

lex:add_rule('custom_type', lex:tag(lexer.DEFAULT, '::') * lex:tag(lexer.TYPE .. '.custom', lexer.word * -S('(:')))

? I'm not sure I understand exactly what you're trying to do.

You can highlight angle braces using lexer.property['scintillua.angle.braces'] = '1' as you've found. You can disable auto-pairing by adding to your ~/.textadept/init.lua:

events.connect(events.LEXER_LOADED, function(lang)
  if lang == 'cpp' then textadept.editing.auto_pairs['<'] = nil end
end

(Basically, overriding https://github.com/orbitalquark/textadept/blob/81180915559448f60101f1d3da61819b40591e83/modules/textadept/editing.lua#L104-L110)

It would be nice to highlight std::make_shared<int>() as a function, but the fact that anything can be between <...> for templated things, lexing it is not straightforward.

As for treesitter, I experimented with Scintilla and treesitter (https://github.com/orbitalquark/tslexia), but nothing has come of it. A majority of compiled grammar shared objects are larger than 1MB in size each. Compare that to Scintillua lexers, which are between 1-5KB on average. All 100+ of them cumulatively total 1MB in size. Once I realized this, I kind of lost interest in trying to further explore this. Also, last I checked, developing and compiling treesitter grammars requires node.JS, which feels so dirty.

[snoopy]

Thanks, that works nicely.
I was using textadept.editing.auto_pairs = {} which did not prevent auto pairing as well as I expected.

developing and compiling treesitter grammars requires node.JS

I had no idea. That's really lame.