Ignore RFC 1468 switch to ASCII code sequence (#225)

This ignores "\e(B" which sometimes shows up even when the other RFC
1468 Japanese character set switch codes aren't used. This also updates
defmodifier so that it can be used with non-CSI escape codes (the
ones that start with '[').

lib/livebook_web/ansi.ex

@@ -3,7 +3,7 @@ defmodule Livebook.ANSI.Modifier do
 
   defmacro defmodifier(modifier, code, terminator \\ "m") do
     quote bind_quoted: [modifier: modifier, code: code, terminator: terminator] do
-      defp ansi_prefix_to_modifier(unquote("#{code}#{terminator}") <> rest) do
+      defp ansi_prefix_to_modifier(unquote("[#{code}#{terminator}") <> rest) do
         {:ok, unquote(modifier), rest}
       end
     end
@@ -43,7 +43,7 @@ defmodule LivebookWeb.ANSI do
   def ansi_string_to_html(string, opts \\ []) do
     renderer = Keyword.get(opts, :renderer, &default_renderer/2)
 
-    [head | ansi_prefixed_strings] = String.split(string, "\e[")
+    [head | ansi_prefixed_strings] = String.split(string, "\e")
 
     {:safe, head_html} = Phoenix.HTML.html_escape(head)
     head_html = renderer.("", head_html)
@@ -58,7 +58,7 @@ defmodule LivebookWeb.ANSI do
               {modifiers, rest}
 
             {:error, _rest} ->
-              {modifiers, "\e[" <> string}
+              {modifiers, "\e" <> string}
           end
 
         {:safe, content} = Phoenix.HTML.html_escape(rest)
@@ -73,7 +73,7 @@ defmodule LivebookWeb.ANSI do
   end
 
   # Below goes a number of `ansi_prefix_to_modifier` function definitions,
-  # that take a string like "32msomething" (starting with ANSI code without the leading "\e[")
+  # that take a string like "[32msomething" (starting with ANSI code without the leading "\e")
   # and parse the prefix into the corresponding modifier.
   # The function returns either {:ok, modifier, rest} or {:error, rest}
 
@@ -105,18 +105,24 @@ defmodule LivebookWeb.ANSI do
   defmodifier({:text_decoration, :overline}, 53)
   defmodifier({:text_decoration, :reset}, 55)
 
-  defp ansi_prefix_to_modifier("38;5;" <> string) do
+  defp ansi_prefix_to_modifier("[38;5;" <> string) do
     with {:ok, color, rest} <- bit8_prefix_to_color(string) do
       {:ok, {:foreground_color, color}, rest}
     end
   end
 
-  defp ansi_prefix_to_modifier("48;5;" <> string) do
+  defp ansi_prefix_to_modifier("[48;5;" <> string) do
     with {:ok, color, rest} <- bit8_prefix_to_color(string) do
       {:ok, {:background_color, color}, rest}
     end
   end
 
+  # "\e(B" is RFC1468's switch to ASCII character set and can be ignored. This
+  # can appear even when JIS character sets aren't in use.
+  defp ansi_prefix_to_modifier("(B" <> rest) do
+    {:ok, :ignored, rest}
+  end
+
   defp bit8_prefix_to_color(string) do
     case Integer.parse(string) do
       {n, "m" <> rest} when n in 0..255 ->

test/livebook_web/ansi_test.exs

@@ -94,5 +94,9 @@ defmodule LivebookWeb.ANSITest do
                ANSI.ansi_string_to_html("cat", renderer: div_renderer)
                |> Phoenix.HTML.safe_to_string()
     end
+
+    test "ignores RFC 1468 switch to ASCII" do
+      assert ~s{cat} == ANSI.ansi_string_to_html("\e(Bcat") |> Phoenix.HTML.safe_to_string()
+    end
   end
 end