Levenshtein algorithm implementation in CoffeeScript

levenshtein.coffee

levenshtein = (a, b) ->
  # Handle degenerate cases
  if a == b
    return 0

  unless a.length > 0 and b.length > 0
    return a.length or b.length

  # Normalize order for cache
  if b.length < a.length
    [a, b] = [b, a]
  
  unless levenshtein.cache[a]?
    levenshtein.cache[a] = {}
  
  if levenshtein.cache[a][b]?
    return levenshtein.cache[a][b]
  
  # Levenshtein only needs the current and previous rows of the matrix
  prevRow = [0..a.length]
  curRow = [0]

  for j in [1..b.length]
      curRow = [j]
      for i in [1..a.length]
          if a[i - 1] == b[j - 1]
            curRow[i] = prevRow[i - 1]
          else
            curRow[i] = Math.min(curRow[i - 1], prevRow[i], prevRow[i - 1]) + 1
      prevRow = curRow
  
  levenshtein.cache[a][b] = curRow[a.length]

levenshtein.cache = {}