http://stackoverflow.com/questions/13640538/vim-syntax-files-add-to-cterm
Example:
syntax region Italic start='_' end='_' modifier contains=Bold
syntax region Bold start='\*' end='\*' modifier contains=Italic
highlight Italic cterm=italic
highlight Bold cterm=bold
_this is both italic *and bold*_
The code is surprisingly simple, as vim already has a stack of syntax matches which it walks up to determine the attributes. It's trivial to add a flag which merges attributes instead of clobbering them.
I'm not sure how to add tests that test text color and format.
Hmm, I think "combine" would express better what this is doing:
combining the existing attributes with those in the current
syntax/highlight item. One can also combine a fg and bg color, but a
new fg color would overrule an existing fg color (we don't mix colors).
We don't need the patch for += and -= then, right?
I'm not clear on how synIDattr() will work with combined syntax items, or synIDtrans() for that matter.
In the example you give, synIDtrans() needs to somehow return an ID for both Bold AND Italic.
Do we need to add a new function or modify synIDattr to get whether a syntax ID stands on its own or needs to be combined with the previous item on the syntax stack? Then scripts can call synstack() and check the combine attribute of each item in the stack.
Or maybe modify synIDtrans() to return a List of items if a combined item is used. Then scripts can just check each item actually used to highlight the text.
I think I like the idea of synIDtrans() returning a list in this case a little better.
I'm bringing this up because I maintain TOhtml, which needs to build CSS information for each highlight attribute. It is easy enough to combine CSS classes to obtain a combined syntax highlight, but the information to know which classes to combine needs to be available.
I do think this sounds like a useful feature.
The problem is that synID() returns the id of the *highlight group*, not the syntax match. In the following example:
syntax match Foo 'foo' combine
syntax match Bar 'bar'
highlight link Foo Comment
highlight link Bar Comment
if you have the highlight id for 'Comment', I can't reliably tell you whether or not it's going to combine with attributes further up the stack.
The best I can think to do is one of:
1. Add synmatchID() and synmatchIDattr()
2. Add syncombines(line, col) which tells you if you also need to consult the rest of the synstack() for attributes.
Honestly I think that these functions are a bit misnamed and would prefer them to be 'hlID', 'hlIDattr', hlIDtrans', 'synID', 'synIDattr', and 'synIDtrans' (the *trans functions always returning lists), but that's backwards compatibility breaking.
Does anyone have preference where we go? I don't have the experience with synID* to choose the 'right' decision.
I don't understand. synIDattr(synID(...),"name") on foo returns "Foo". synIDattr(synIDtrans(synID(...)),"name") on foo returns "Comment". Can't we just add something like synIDattr(synID(...),"combine") which returns true/false? Determining the highlight would then be a matter of going up the syntax stack until a non-combining item is found, then walking back up the stack getting the attributes for the translated syntax ID of each item.
That said, doing all that in vimscript would probably be quite inefficient. The TOHtml implementation even has a comment on the loop that finds all characters in a particular highlight group, that the loop needs to be kept small. A traversal up and down the syntax stack for every character in the document would be a very bad thing. Doing the same thing for every entire highlighted item found would also probably not be very good. So I think we want a new set of functions to help out.
> The best I can think to do is one of:
>
> 1. Add synmatchID() and synmatchIDattr()
What would these do?
> 2. Add syncombines(line, col) which tells you if you also need to consult the rest of the synstack() for attributes.
>
This could work but would likewise affect the performance of a tight inner loop requiring "if" logic and a vimscript stack traversal on each character.
Syntax highlighting internally uses a unique number for each individually highlighted item. What if we add a function to return this number (synconcealed() already returns it as one item in a list for different reasons), and another function to get a list of syntax IDs which apply to an item's highlighting (in order of application), given this item number?
I've updated the patch to add the "combine" {what} to synIDattr, which is the bare minimum we need here.
synID should definitely return only one ID in this case, both because we don't want to duplicate synstack and because it's the only way to get interoperability with the extended links patch.
I'm still not sure how best to deal with synIDattr. How would you feel about adding a [, {flatten}] argument to it that flattens out the stack before checking the attr (analagous to [, {trans}] in synID)?
I like that idea, but in order for it to work it would need a row+col because the attributes for combined items depend on what they are combined with which depends on where in the document you are.
So I guess I could get the attrs flattened always at a particular row and column, then if the "combine" item shows that at least one item in the flattened view is a combining syntax item, look at the syntax stack for the names used.
The problem I see is in detecting two regions side-by-side with the same top-level (combining) syntax item but different underlying syntax stacks which therefore need different highlighting. I'd like to do this without checking the entire stack for every character in the document which I believe will be much slower than otherwise. I suppose I could keep a list of synIDs which combine and only check the stack for those but I'd still like a faster method.
They both still handle the single-id case, but the code's a bit messy and the name is stale. I think we should add synActive and synActiveAttr, which take/return lists instead of single numbers, and then deprecate synID*.