Compatibility for builtin syntax groups

[clason]

Vimtex 2.0 ships with a new syntax plugin with more consistent and structured syntax and highlight groups; this overrides the built-in syntax file (by default, unless let g:vimtex_syntax_enabled=0 disables the new syntax file and reverts to the old one).

However, it seems that other plugins rely on the specific groups and their names of the original syntax; see #1881 . To enable these, one can manually link the new groups to the old groups with, e.g., hi link texMathZoneZ texMathRegionEnsured.

It would be helpful to add a list of such commands to get back (as close as possible) to the old groups to the Wiki.

Even more convenient would be a function, e.g., vimtex#syntax#compat#set_groups() (or even an option g:vimtex_syntax_compat) that can be used to set these links from your vimrc/init.vim.

[lervag]

For completeness, @Konfekt opened a very related issue #1882. I decided to close that one and continue the discussion here.

... it seems that other plugins rely on the specific groups and their names of the original syntax; see #1881 . To enable these, one can manually link the new groups to the old groups with, e.g., hi link texMathZoneZ texMathRegionEnsured.

It would be helpful to add a list of such commands to get back (as close as possible) to the old groups to the Wiki.

I've started this, see the updated Wiki page here.

Even more convenient would be a function, e.g., vimtex#syntax#compat#set_groups() (or even an option g:vimtex_syntax_compat) that can be used to set these links from your vimrc/init.vim.

Yes, that might be a good solution. Let's first agree on which groups are relevant and necessary, then I'll add this function/option.

[Konfekt]

This linkage as given in the Wiki might restore compatibility with tex-conceal but the function that checks for a math zone by the built-in highlight groups still no longer works:

function! IsMath()
  return match(map(synstack(line('.'), max([col('.') - 1, 1])),
        \ 'synIDattr(v:val, ''name'')'), '^texMathZone\%([A-L]S\?\|[V-Z]\)$') >= 0
endfunction

[lervag]

Hmm. I don't think this trick will work, actually. highlight link only links highlight groups, it does not add syntax matches. So, a syntax item will not be matched by texMathZoneX after hi link ..., if the actual group is texMathRegionX.

With that, I think it seems like the hi link idea does not really give any compatbility benefit. Or am I wrong?

[Konfekt]

Fixed. It needs to follow linked groups by synIDtrans():

  return match(map(synstack(line('.'), max([col('.') - 1, 1])),
        \ 'synIDattr(synIDtrans(v:val), ''name'')'), '^texMathZone\%([A-L]S\?\|[V-Z]\)$') >= 0

[lervag]

Does that also work without the links?

[Konfekt]

Yes!

With that, I think it seems like the hi link idea does not really give any compatbility benefit. Or am I wrong?

At least it solves #1880

[Konfekt]

Does that also work without the links?

No, it does not.

[lervag]

My understanding is this:

• With synIDtrans, your function will work with Vimtex v2 as long as you also include the high link linkes suggested in the wiki.
• With synIDtrans, your function will not work with the built-in syntax plugin.

So, it still seems we are not really fixing anything right now. But I'll be glad to learn I'm wrong!

[clason]

And what about defining the new groups at the syntax level with something like syntax cluster <legacy-group> contains=<vimtex-group>?

[lervag]

I don't think that would work. Clusters don't work that way; they are used as arguments for contains=@cluster or containedin=@cluster.

No, for the desired compatibility, we actually do need to create duplicate syntax rules (if I understand things correctly, and I might not). If I'm right, then this is not something I think is really viable.

[Konfekt]

I agree. Thus, have mercy on the user and keep the syntax group names? (Even though the new names are obviously more pertinent, but backward compatibility is a valuable good.)

[lervag]

Thus, have mercy on the user and keep the syntax group names?

No, sorry, I wont keep the old syntax group names in general. I find the value of consistent group names much higher than the value of "backward" compatibility.

But ok, let's say I did agree to revert the change of Zone to Region, how much would this solve in terms of the actual issues people/you have? If I did this particular change, then I think it would solve your #1880, especially if you simplify your function in the direction I just suggested.