Roman Numerals kata implemented in elixir

roman_numerals.exs

defmodule RomanNumerals do

	@moduledoc """
	Task definition: Create a function taking a positive integer as its parameter and
	returning a string containing the Roman Numeral representation of that integer.
	"""
 
  	def convert(number) do
	    convert(number, [[10,'X'], [9,'IX'], [5,'V'], [4,'IV'], [1,'I']])
  	end  	

  	defp convert(0, _) do
	     ''
  	end

  	defp convert(number, [[arabic, roman]|_] = l) when number >= arabic do
	     roman ++ convert(number - arabic, l)
  	end

  	defp convert(number, [_|t]) do
	     convert(number, t)
  	end
end
 
ExUnit.start
 
defmodule RomanNumeralsTest do
  use ExUnit.Case
  
  test "converts arabic numbers to roman" do
  	Enum.each [ [1,'I'],
  		    [2,'II'],
		    [3,'III'],
  		    [4,'IV'],
  		    [5,'V'],
  		    [6,'VI'],
  		    [7,'VII'],
  		    [8,'VIII'],
  		    [9,'IX'],
  		    [10,'X'],
  		    [11,'XI'],
  		    [14,'XIV'],
  		    [19,'XIX'],
  		    [20,'XX'],
  		    [25,'XXV'],
  		    [27,'XXVII'] ], fn [arabic, roman] ->
  		assert roman == RomanNumerals.convert(arabic)
  	end
  end
end

You inspired me to work through with a different algorithm. See: https://github.com/gvaughn/elixir_kata/blob/master/roman/roman_numerals.exs

I am curious why you chose to use char lists instead of binaries.

Greg, the reason for that is quite simple: single quoted strings seem to me more elegant than double quoted ones. :)

I looked at your code... It's not clear how you came up with this solution as there is only one revision. Have you used TDD or any other approach?

In my case I wrote tests first. Then I created a straightforward solution to pass the tests (see revision # 1). And after that I refactored the code to get final solution (you're looking at it right now in revision # 2). So only TDD allowed me to do it!

Sorry for the delay. I had expect notification on a reply from github.

It's my understanding that in elixir single quoted strings (char lists) often perform worse than double quoted strings (binaries). Not that it really matters in this small of a problem.

I copied your tests and worked from them. I'm sorry I didn't do intermediate check ins. I feel I need to defend my TDD skills to you. I gave a talk at JavaOne 2002 entitled "Unit Testing Made Easy: Design Principles to Facilitate Testing". I'm having trouble finding much information online about it, but here's 2 links, the first is my comment on the refactoring mailing list about it, and the 2nd is someone's notes of talks they attended:
http://osdir.com/ml/programming.refactoring/2002-07/msg00052.html
http://ziggr.com/javaone2002/#z51

I could have incrementally done red/green/refactor for this exercise. However, given the limited complexity of it, and the fact that I've done the roman numeral kata in several languages, I wouldn't learn anything but new elixir syntax. So, I just focused on that. I built off your "when number >= arabic" idea, then learned elixir syntax for div, rem, and String.duplicate