#!/usr/bin/env python
# -*- coding: utf-8 -*-
#  Make a guess at a transliteration mapping, given a lexicon.
from unicodedata import name
import sys, codecs
from textual import freq
sys.stdout = codecs.getwriter('utf-8')(sys.stdout)
sys.stdin = codecs.getreader('utf-8')(sys.stdin)

# '/media/SEAGATE/pat/l/lex/en2hi.txt'

lexfile = sys.argv[1]

# Lexicons are extracted from Wikipedia using extractterms.py
# examples are English/Hindi
source2target = open(lexfile).read()
source2target = source2target.decode('utf-8')

s2t = {} 
for line in source2target.splitlines():
    # lines look like:
    # en	बिहार	Bihar
    try: 
        code, target, source = line.split('\t')
        source = source.lower()
        s2t[source] = target
    except ValueError:
        continue


# now s2t contains a lexicon:
# { u'bihar' : u'बिहार' , ... } 

correspondences = [] # this will hold all letter-letter correspondences: [('b', 'ब'), ('h', 'ह') ... ]
for e,h in s2t.items():
    all = []
    all.extend(zip(list(e),list(h)))
    # because letter-letter correspondences may be better at the end of the word than the 
    # beginning, we zip reversed versions of the words too:
    all.extend(zip(reversed(list(e)), reversed(list(h))))
    correspondences.extend(all)

fq = freq(correspondences)
qf = [(fq,pair) for pair,fq in freq(correspondences).items()]
sqf = sorted(qf)

for f, pair in sqf[-100:]:
    print name(pair[0]), pair[0], pair[1]