feat: Add shebang support in lexer

lib/lua/lexer.ex

@@ -37,10 +37,23 @@ defmodule Lua.Lexer do
   """
   @spec tokenize(String.t()) :: {:ok, [token()]} | {:error, term()}
   def tokenize(code) when is_binary(code) do
+    # Handle shebang on first line (Unix convention: #! means interpreter directive)
+    code = strip_shebang(code)
     pos = %{line: 1, column: 1, byte_offset: 0}
     do_tokenize(code, [], pos)
   end
 
+  # Strip shebang (#!) if it's the first line
+  defp strip_shebang(<<"#!", rest::binary>>) do
+    # Skip entire first line (everything up to and including the newline)
+    case String.split(rest, ~r/\r\n|\r|\n/, parts: 2) do
+      [_shebang_line, remaining] -> remaining
+      [_only_shebang] -> ""
+    end
+  end
+
+  defp strip_shebang(code), do: code
+
   # End of input
   defp do_tokenize(<<>>, acc, pos) do
     {:ok, Enum.reverse([{:eof, pos} | acc])}

test/lua/lexer_test.exs

@@ -340,6 +340,28 @@ defmodule Lua.LexerTest do
       assert {:ok, [{:comment, :multi, " comment ", _}, {:identifier, "x", _}, {:eof, _}]} =
                Lexer.tokenize("--[[ comment ]]x")
     end
+
+    test "ignores shebang on first line" do
+      # Lua ignores the first line if it starts with #
+      code = """
+      #!/usr/bin/env lua
+      local x = 1
+      return x
+      """
+
+      assert {:ok, tokens} = Lexer.tokenize(code)
+      # Shebang should be treated as a comment and stripped
+      assert [{:keyword, :local, _} | _] = tokens
+
+      # Shebang only on first line
+      code2 = """
+      local x = 1
+      #!/usr/bin/env lua
+      """
+
+      # Second line with # should cause error (not a shebang)
+      assert {:error, _} = Lexer.tokenize(code2)
+    end
   end
 
   describe "whitespace" do