Is this the right way to handle LS_SELECT_RAW/TM_SELECTED_TEXT not being populated on snippet init?

[xav-ie]

When a snippet with a function node is expanded for the first time in a session, the function node appears to undergo a pre-evaluation phase using some placeholder variables. After this initial phase, it is reinvoked with the "correct"/expected variables. I am attempting to define a snippet as follows:

s("co", {
  d(1, function(_, snip) 
    local empty_node = sn(nil, {})
    local ls_select_raw = snip.env.LS_SELECT_RAW or {}
    if type(ls_select_raw) == "string" then
      return empty_node
    end
    local selected_text = table.concat(ls_select_raw, "\n")

    local register_data = vim.fn.getreg() .. ""
    local selected_and_register_data = selected_text .. register_data
    local first_match = string.match(selected_and_register_data, "-?%d+,%s*-?%d+")

    if first_match then
      return sn(nil, {
        t("position([" .. first_match .. "])")
      })
    else
      vim.notify("Register does not contain the pattern", vim.log.levels.INFO, { title = "LuaSnippet" })
      return empty_node
    end
  end),
}),

It seems that my snippet function is being called once during its initial setup. On this first invocation, snip.env.LS_SELECTED_TEXT is not of the expected type and instead holds the string "$LS_SELECTED_TEXT". I'm not entirely sure why this happens during initialization, so I handle it by returning an empty node. Is there a more appropriate way to handle this situation?

I saw #647
where the author uses:

local ok, val = pcall(vim.api.nvim_buf_get_var, 0, "LUASNIP_SELECT_RAW");

But that has the behavior of only containing the expected text the very first time the snippet is invoked, so I don't think that is what I want.

I have read :h luasnip-selection, :h luasnip-functionnode, :h luasnip-dynamicnode, I went through the examples, I watched some tutorials, and I even now know there is this fake_env function that is causing this behavior, but I do not know why it is this way and not mentioned.

[L3MON4D3]

Hi!
First of all, sorry for the late response 😅

What you're observing is the mechanism luasnip uses to generate a string-representation of a snippet-object, it's explained here.

So, the correct handling depends on what you'd like to see in the docstring, you could for example do

s("co", {
  d(1, function(_, snip)
    -- check for fake_expand
    if snip.snippet.env.____INVALIDVARNAME then
      return sn(nil, {t"position[$x,$y]"})
    end

    local ls_select_raw = snip.env.LS_SELECT_RAW or {}

    local selected_text = table.concat(ls_select_raw, "\n")

    local register_data = vim.fn.getreg() .. ""
    local selected_and_register_data = selected_text .. register_data
    local first_match = string.match(selected_and_register_data, "-?%d+,%s*-?%d+")

    if first_match then
      return sn(nil, {
        t("position([" .. first_match .. "])")
      })
    else
      vim.notify("Register does not contain the pattern", vim.log.levels.INFO, { title = "LuaSnippet" })
      return sn(nil, {})
    end
  end),
})

which would give the docstring "${1:position[$x,$y]}$0"