Profiling codepoints

README

Instructions:

 - Apply the patch
 - Compile elixir
 - Run the tests

    $ git apply /path/to/patch.diff
    $ make compile
    $ ./bin/elixir native.ex > /tmp/native.ex
    $ ./bin/elixir unicode.ex > /tmp/unicode.ex

In my machine, native implementation was ~1.3x faster. Feel free to add more tests.

codepoints_profiler.ex

defmodule CodepointsProfiler do

  def string, do: "áéíàà ¨&$#"

  def unicode do
    Enum.map 1..100, fn(x) -> String.codepoints_unicode("#{x}#{string}") end
  end

  def native do
    Enum.map 1..100, fn(x) -> String.codepoints("#{x}#{string}") end
  end

end

native.ex

Code.require_file "codepoints_profiler.ex"

Erlang.fprof.apply fn -> CodepointsProfiler.native end, []
Erlang.fprof.profile
Erlang.fprof.analyse

patch.diff

diff --git a/lib/elixir/lib/string.ex b/lib/elixir/lib/string.ex
index 08868fc..5c3d2b7 100644
--- a/lib/elixir/lib/string.ex
+++ b/lib/elixir/lib/string.ex
@@ -287,6 +287,33 @@ defmodule String do
     Erlang.binary.copy(subject, n)
   end
 
+  def codepoints_unicode(string) do
+    to_char = fn(char) -> :unicode.characters_to_binary([char], :utf8, :utf8) end
+    Enum.map(:unicode.characters_to_list(string, :utf8), to_char)
+  end
+
+  @doc """
+  Returns a list with codepoints
+  """
+  def codepoints(string) do
+    codepoints(string, [])
+  end
+
+  def codepoints(<<194, char, rest :: binary>>, buffer) when char in 161..191 do
+    codepoints(rest, buffer ++ [<<194, char>>])
+  end
+
+  def codepoints(<<195, char, rest :: binary>>, buffer) when char in 128..191 do
+    codepoints(rest, buffer ++ [<<195, char>>])
+  end
+
+  def codepoints(<<other, rest :: binary>>, buffer) do
+    codepoints(rest, buffer ++ [<<other>>])
+  end
+
+  def codepoints(<<>>, buffer) do
+    buffer
+  end
 
   defp translate_replace_options([]), do: []
   defp translate_replace_options(raw_options) do
diff --git a/lib/elixir/test/elixir/string_test.exs b/lib/elixir/test/elixir/string_test.exs
index c9afa9f..dc4d3b5 100644
--- a/lib/elixir/test/elixir/string_test.exs
+++ b/lib/elixir/test/elixir/string_test.exs
@@ -77,4 +77,10 @@ defmodule StringTest do
        assert String.duplicate("&ã$", 2) == "&ã$&ã$"
   end
 
+  test :codepoins do
+    assert String.codepoints("&é%Á#") == ["&","é","%","Á","#"]
+    assert String.codepoints("Élysés") == ["É","l","y","s","é","s"]
+    assert String.codepoints_unicode("áá") == ["á", "á"]
+  end
+
 end

unicode.ex

Code.require_file "codepoints_profiler.ex"

Erlang.fprof.apply fn -> CodepointsProfiler.unicode end, []
Erlang.fprof.profile
Erlang.fprof.analyse