#'!/usr/bin/env python
# -*- coding: utf-8 -*-
import sys
import codecs
sys.stdout = codecs.getwriter('utf-8')(sys.stdout)
from collections import defaultdict

def readlexicon(lexiconfile):
  ka = open(lexiconfile).read().decode('utf-8')
  pairs = [line.strip().split('\t') for line in ka.splitlines()]
  pairs = [pair for pair in pairs if len(pair) == 2]
  pairs = [(e,k) for e,k in pairs if e != k]
  pairs = [(e,k) for e,k in pairs if len(e) > 3 and len(k) > 3]
  pairs = [(e,k) for e,k in pairs if len(e) < 10 and len(k) < 10]
  return pairs

def patternize(word):
  pattern = [str(word.index(letter)) for letter in word]
  return ''.join(pattern)

pairs = readlexicon(sys.argv[1])

transliteratedpairs = [(e,k) for e,k in pairs if patternize(e) == patternize(k)]

#entry = """ \latintext \\textbf{%s} \hinditext %s"""
entry = """%s\n%s\n"""

"""
# interesting to look at the word pairs
for (e,k) in transliteratedpairs: 
  print '\n' + '\n'.join([e, patternize(e), k])
"""

matches = []
[matches.extend(zip(e,k)) for e,k in transliteratedpairs]

def freq(seq):
    fq = defaultdict(int)
    for elem in seq:
      fq[elem] += 1
    return fq

model = defaultdict(list)
for (e,k),fq in freq(matches).items():
  model[e].append((k,fq))

schema = {} # the ultimate transliteration scheme

for e in model:
  max = 0 
  for (letter, count) in model[e]:
    if count > max:
      max = count
      maxletter = letter
    model[e] = maxletter

for e in model:
  from unicodedata import name
  print  e, model[e]  
  #print  e, model[e], name(model[e]).replace('ARABIC LETTER','')