feat(cpp <> pairs) auto add <> for templates,includes and templated types

[Sam-programs]

this issue #405 inspired me to make this pr
this pr adds cpp_pairs to the completion pairs
which does

press < 
template|    -> template<|>
#include|    -> #include <|> // extra space for include
std::vector| -> std::vector<|>

cpp_sort_cmp is there to make sure templated classes are picked over constructors for templated classes
without this we wouldn't detect templated classes

[Sam-programs]

windpw this uses set_current_line i could migrate it to only use keys but that would make the code much less readable

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[kuator]

Awesome feature, I have a question on whether it would be possible to have both autopair's default confirm_done callback which completes function braces callback with your cpp_pairs callback which completees generics. And would it be possible to have cpp_pairs work with other languages that have generics as well?

[Sam-programs]

I have a question on whether it would be possible to have both autopair's default confirm_done callback which completes function braces callback with your cpp_pairs callback which completees generic

I think that would make the code a bit messy, Because it would be language-specific code in the general confirm_done callback.

And would it be possible to have cpp_pairs work with other languages that have generics as well?

I think it would have to be manually added for each language. templates and #include have a hardcoded check for cpp.

[kuator]

I have a question on whether it would be possible to have both autopair's default confirm_done callback which completes function braces callback with your cpp_pairs callback which completees generic

I think that would make the code a bit messy, Because it would be language-specific code in the general confirm_done callback.

And would it be possible to have cpp_pairs work with other languages that have generics as well?

I think it would have to be manually added for each language. templates and #include have a hardcoded check for cpp.

Thank you for the quick answer!

[kuator]

I have a question on whether it would be possible to have both autopair's default confirm_done callback which completes function braces callback with your cpp_pairs callback which completees generic

I think that would make the code a bit messy, Because it would be language-specific code in the general confirm_done callback.

And would it be possible to have cpp_pairs work with other languages that have generics as well?

I think it would have to be manually added for each language. templates and #include have a hardcoded check for cpp.

So I decided to keep it stupid and wrote my simple callback:

  local function on_confirm_done(evt)
    local entry = evt.entry
    local item = entry:get_completion_item()
    local types = require("cmp.types")

    if evt.commit_character then
      return
    end

    if item.label:find('<>') and vim.bo.filetype == "cs" and entry:get_kind() == types.lsp.CompletionItemKind.Class then
      local keys = vim.api.nvim_replace_termcodes('<><left>', false, false, true)
      vim.api.nvim_feedkeys(keys, "i", true)
      return
    end

    if
        vim.bo.filetype == "cs"
        and (
          entry:get_kind() == types.lsp.CompletionItemKind.Function
          or entry:get_kind() == types.lsp.CompletionItemKind.Method
          or entry:get_kind() == types.lsp.CompletionItemKind.Class
        )
    then
      local keys = vim.api.nvim_replace_termcodes('()<left>', false, false, true)
      vim.api.nvim_feedkeys(keys, "i", true)
      return
    end

  end

Now I can ditch nvim-autopairs as this was the only reason for me to use in the first place.

[Sam-programs]

@kuator
feedkey's 'i' mode still uses mappings because there is no 'n' flag, That might cause issues.

   {escape_ks}  If true, escape K_SPECIAL bytes in `keys`. This should be
                false if you already used nvim_replace_termcodes(), and
                true otherwise.

Also the last parameter should be false, That shouldn't matter thought.

Now I can ditch nvim-autopairs as this was the only reason for me to use in the first place.

You don't manually write pairs anywhere even for []?
I am just curious.

[kuator]

@kuator feedkey's 'i' mode still uses mappings because there is no 'n' flag, That might cause issues.

   {escape_ks}  If true, escape K_SPECIAL bytes in `keys`. This should be
                false if you already used nvim_replace_termcodes(), and
                true otherwise.

Also the last parameter should be false, That shouldn't matter thought.

Now I can ditch nvim-autopairs as this was the only reason for me to use in the first place.

You don't manually write pairs anywhere even for []? I am just curious.

I tried to use auto-pair plugins multiple times throughout my vim journey starting from https://github.com/Raimondi/delimitMate and now nvim-autopairs and the behaviour ends up being more annoying than useful. It's simpler for me type the pairs myself. What I also did is I set up the luasnip snippets. One example is qq which converts into "" or bb which translates into (). Have already deleted nvim-autopairs, now there's no unexpected behaviour going when I type.

[kuator]

@kuator feedkey's 'i' mode still uses mappings because there is no 'n' flag, That might cause issues.

   {escape_ks}  If true, escape K_SPECIAL bytes in `keys`. This should be
                false if you already used nvim_replace_termcodes(), and
                true otherwise.

Also the last parameter should be false, That shouldn't matter thought.

Now I can ditch nvim-autopairs as this was the only reason for me to use in the first place.

You don't manually write pairs anywhere even for []? I am just curious.

@kuator feedkey's 'i' mode still uses mappings because there is no 'n' flag, That might cause issues.

   {escape_ks}  If true, escape K_SPECIAL bytes in `keys`. This should be
                false if you already used nvim_replace_termcodes(), and
                true otherwise.

Also the last parameter should be false, That shouldn't matter thought.

Now I can ditch nvim-autopairs as this was the only reason for me to use in the first place.

You don't manually write pairs anywhere even for []? I am just curious.

Honestly, I have no idea how feedkeys works. I did a bit more research and now instead call luasnip:

  local function on_confirm_done(evt)
    local entry = evt.entry
    local item = entry:get_completion_item()
    local types = require("cmp.types")

    if evt.commit_character then
      return
    end

    if item.label:find('<>') and vim.bo.filetype == "cs" and entry:get_kind() == types.lsp.CompletionItemKind.Class then
      local ls = require('luasnip')
      ls.snip_expand(ls.s("trig", { ls.t"<", ls.i(1, ""), ls.t">" }) )
      return
    end

    if
        vim.bo.filetype == "cs"
        and (
          entry:get_kind() == types.lsp.CompletionItemKind.Function
          or entry:get_kind() == types.lsp.CompletionItemKind.Method
          or entry:get_kind() == types.lsp.CompletionItemKind.Class
        )
    then
      local ls = require('luasnip')
      ls.snip_expand(ls.s("trig", { ls.t"(", ls.i(1, ""), ls.t")" }) )

      return
    end

  end