How do I make regexes mutually contain each other?

Question

I've written regexes for markdown for emphasis and bold as follows:

syntax match PandocEmphasisMarker "\v(^|\s+)\zs_[^\n\t_ ]+([^_\n]|(\S_\S)|(\s_\S)|(\s+_\s+))+[^\n\t_ ]+_\ze($|\s+)" contains=PandocEmphasis
syntax match PandocEmphasis "\v\zs[^\n\t_ ]+([^_\n]|(\S_\S)|(\s_\S)|(\s+_\s+))+[^\n\t_ ]+\ze" contained

syntax match PandocStrongMarker "\v\zs\*[^\n*]+\*\ze" contains=PandocStrong
syntax match PandocStrong "\v\zs[^\n*]+\ze" contained

This yields the following (in my opinion correct) highlighting:

However, in Markdown there could be emphasis inside strong and strong inside emphasis as well. Since I already have the regexes so far, I'd like to allow them to mutually include each other.

I've tried with clusters as follows:

syntax cluster PandocEmphasisCluster contains=PandocEmphasis,PandocStrongMarker
syntax cluster PandocStrongCluster contains=PandocStrong,PandocEmphasisMarker

And then referenced those clusters from the "Marker" expressions using contains, but that did not work, it messed up the strong highlighting in my test file and the strong part inside an emphasized part was not made bold.

How can / should I make them contain each other?

This is also important for other things like ` inside emphasized or strong. Maybe I should use region somehow?

Clarification:

• emphasis inside strong should still be strong, but also become emphasized.
• strong inside emphasis should still be emphasized, but also become strong
• effects should add up, that's what I want to achieve

Answer 1

If you have known start and end delimiters, you should use regions (:h :syn-region):

syntax region Emphasis start="\\\@<!\z(_\)" skip="\%(\s\|\\\)\@<=\z1\|\z1\k" end="\\\@<!\z1" keepend contains=Strong,Strike
syntax region Strong start="\\\@<!\z(\*\)" skip="\%(\s\|\\\)\@<=\z1\|\z1\k" end="\\\@<!\z1" keepend contains=Emphasis,Strike
syntax region Strike start="\\\@<!\z(\~\~\)" skip="\%(\s\|\\\)\@<=\z1\|\z1\k" end="\\\@<!\z1" keepend contains=Emphasis,Strong

highlight default link Emphasis WarningMsg
highlight default link Strong ErrorMsg
highlight default link Strike Conceal

The regions above use the syntax-specific capture group (:h /\z() which is great for broad patterns (e.g. if..endif, while..endwhile, etc.), but in this case, it's mainly for readability and avoiding some repetition.

Each region starts with the \z(\) atom to define your delimiter and they must end with what was captured (:h /\z1). Additionally, the delimiters can't be preceded by a backslash.

To skip the delimiters under certain circumstances, the \%(\s\|\\\)\@<=\z1\|\z1\k pattern is used: If it's preceded by whitespace or a backslash, or if it's followed by a keyword character, it should not be considered the end of the region.

Each region contains the others which allows nesting, giving us highlighting like this:

---

Existing plugins

I would like to point out that there are already a number of markdown syntax plugins available for Vim. There's nothing wrong with making your own, but there's plenty of reference material you could draw from, including one specifically made for pandoc.

Answer 2

You can simply add the contains=<...> option to both the PandocEmphasis and PandocStrong to allow for nested matches:

syntax match PandocEmphasisMarker "\v(^|\s+)\zs_[^\n\t_ ]+([^_\n]|(\S_\S)|(\s_\S)|(\s+_\s+))+[^\n\t_ ]+_\ze($|\s+)" contains=PandocEmphasis
syntax match PandocEmphasis "\v\zs[^\n\t_ ]+([^_\n]|(\S_\S)|(\s_\S)|(\s+_\s+))+[^\n\t_ ]+\ze" contained contains=PandocStrongMarker

syntax match PandocStrongMarker "\v\zs\*[^\n*]+\*\ze" contains=PandocStrong
syntax match PandocStrong "\v\zs[^\n*]+\ze" contained  contains=PandocEmphasisMarker