Crash with invalid String.slice length if file has emoji

[nshafer]

Dorgan asked me in Discord to report bugs with Lexical, since Expert may have the same problem. This is a weird one I experience in a Phoenix app I'm working on, where when typing just d when I start to type def will crash lexical with:

2025-08-04 12:04:19.717 [info] [Error - 12:04:19 PM] Request textDocument/completion failed.
2025-08-04 12:04:19.717 [info]   Message: ** (FunctionClauseError) no function clause matching in String.slice/3
    (elixir 1.18.3) lib/string.ex:2333: String.slice("  d", 0, -1)
    (lx_common 0.7.2) lib/lexical/code_unit.ex:23: LXical.CodeUnit.utf8_position_to_utf16_offset/2
    (lx_protocol 0.7.2) lib/lexical/protocol/conversions.ex:146: LXical.Protocol.Conversions.extract_lsp_character/1
    (lx_protocol 0.7.2) lib/lexical/protocol/conversions.ex:120: LXical.Protocol.Conversions.to_lsp/1
    (lx_protocol 0.7.2) lib/lexical/protocol/conversions.ex:98: LXical.Protocol.Conversions.to_lsp/1
    (lx_protocol 0.7.2) lib/lexical/protocol/convertibles/lexical.document.edit.ex:11: LXical.Convertible.Lexical.Document.Edit.to_lsp/1
    (lx_lexical_shared 0.5.0) lib/lexical/convertible.ex:9: anonymous fn/3 in LXical.Convertible.Helpers.apply/2
    (elixir 1.18.3) lib/enum.ex:4968: Enumerable.List.reduce/3

  Code: -32603 

I was able to reproduce this in a new project created with mix phx.new lxtest (1.8.0-rc.4). After install and dep install/compile so everything is happy, in VS Code, create a new file, "lib/testlx_web/components/test_component.ex" and paste:

defmodule TestlxWeb.Components.TestComponent do
  use TestlxWeb, :live_component

  def render(assigns) do
    ~H"""
    <div>
      <h1>Hello, {@name}! ⚠️</h1>
    </div>
    """
  end
end

Now try to create a new function, as soon as you type d it will crash. If you remove the emoji, then it works as expected. I'm guessing this is due to counting codepoints instead of graphemes, as that emoji is 2 codepoints.

iex(6)> s = "Hello⚠️"
"Hello⚠️"
iex(7)> s
"Hello⚠️"
iex(8)> String.codepoints(s)
["H", "e", "l", "l", "o", "⚠", ""]
iex(9)> String.graphemes(s)
["H", "e", "l", "l", "o", "⚠"]

Thanks!