unmashing.py

from __future__ import division

# layout: group keys that are typed by the same finger.
dvorak_layout = '`1!\'a; 2@,oq 3#ej $4puk%5yix ^6fdb&7ghm *8ctw (9rnv )0lsz[{]}/?+=|\\-_'.split()
qwerty_layout = '''`~!1qaz @2wsx #3edc $4rfv%5tgb ^6yhn&7ujm *8ik,< (9ol.> )0p;:?/-_[{'"=+]}|\\'''.split()

layout = qwerty_layout  #dvorak_layout
finger = {ch: i
          for i, chars in enumerate(layout)
          for ch in chars}

def mash_naive(word):
    return ''.join(set(word))

max_fingers = 4   # 5

def mash_first_fingers(word):
    # Use up to max_fingers of the first letters you'd type:
    fingers = set()
    keys = []
    for ch in word:
        f = finger[ch]
        if f not in fingers:
            keys.append(ch)
            fingers.add(f)
            if max_fingers == len(fingers): break
    keys.sort()
    return ''.join(keys)

mash = mash_first_fingers  # mash_naive

def gen_common_words():
    with open('count_1w_100k.txt') as wordfile:
        for line in wordfile:
            word, count = line.split('\t')
            yield word, int(count)

freqs = dict(gen_common_words())

unmashed = {}
for word, count in freqs.iteritems():
    key = mash(word)
    unmashed.setdefault(key, []).append(word)
for candidates in unmashed.itervalues():
    candidates.sort(key=freqs.get, reverse=True)

freqs_right = [0] * 3
total_freq = 0
for word, count in freqs.iteritems():
    total_freq += count
    candidates = unmashed[mash(word)]
    for n_off in range(len(freqs_right)):
        if word in candidates[:n_off+1]:
            freqs_right[n_off] += count

for n_off in range(len(freqs_right)):
    print "Correct within %d guesses: %.2f%%" % (n_off+1, 100*freqs_right[n_off]/total_freq)

 $ python mashing.py
Correct within 1 guesses: 58.34%
Correct within 2 guesses: 71.48%
Correct within 3 guesses: 77.74%