Recreating UltiSnip's VISUAL

[Gargantuar314]

I thought about this a long time, but could not figure it out.
This is a very specific problem.

I recreated UltiSnip's ${1:${VISUAL}} using this post.
With the option store_selection_keys = '<Tab>', this is the snippet file:

local function insert_wrap(_, snip) -- Emulate Ultisnips' '${1:${VISUAL}}'
    local raw = snip.env.LS_SELECT_RAW
    if #raw == 0 then -- Insert text if no selection
        return sn(nil, i(1))
    else -- Else insert selection
        return sn(nil, t(raw))
    end
end

return {}, {

s({
    trig = 'mt',
    name = 'inline math mode'
}, {
    t('$'), i(1), t('$')
}),

s({
    trig = "(",
    dscr = "parenthesis",
    wordTrig = false
}, {
    t('('), d(1, insert_wrap), t(')')
}),

s({
    trig = '_',
    name = 'subscript',
    wordTrig = false,
}, {
    t('_'),
    m(1, '^.$', '', '{'), i(1), m(1, '^.$', '', '}')
}),

}

The idea is to make a subscript command in Latex, but only wrap the input in braces if the input is precisely one character.
This should give h_n, but h_{na} for instance.

Now, if I type mth_n<TAB>, I obtain $h_n$ as usual; the braces around the n disappear.
If I type mth_n<TAB>A_k<TAB>, I do obtain $h_nA_k$ as expected; after the first <TAB>, the braces of n dissapear, and after the second <TAB> the ones of k disappear.

But if I type mth_n<TAB>(A_k<TAB>, I obtain $h_{n(A_k)}$; after the first <TAB>, the braces of n disappear, but after the second <TAB>, braces suddenly appear including around the n.
What I would expect is $h_n(A_k)$, and all snippets finish after tabbing out a third and fourth time (because of the parenthesis and the $).
Notice that everything works as expected if I am not in a snippet, i.e. typing h_n<TAB>(A_k<TAB> really does give h_n(A_k).

Could someone explain to me why this is happening?
And how should I fix the function insert_wrap in order to remedy this?

[L3MON4D3]

Hey, nice catch, that's a bug!
The issue was that the extmark that marks the region of the i(1) was not adjusted correctly during expansion of the ()-snippet, and since it had end-rgrav true, it received all the text inserted into the ().
I've fixed this (straightforward with your example/MWE, thank you!), and it should all work correctly on master.
(this bug only occurs if the new snippet is expanded inside another snippet, immediately (so no text at all between them) behind another insertNode, I guess that's why it only popped up so late :D )

[Gargantuar314]

Thanks a lot, it works!