4 from typing import Dict
10 cfg = config.add_commandline_args(
11 f'Unscramble! ({__file__})', 'A fast word unscrambler.'
14 "--unscramble_indexfile",
15 help="Path to a file of signature -> word index.",
17 default="/usr/share/dict/sparse_index",
20 logger = logging.getLogger(__name__)
23 letters_mask = 2 ** letters_bits - 1
26 fprint_mask = (2 ** fprint_bits - 1) << letters_bits
28 fprint_feature_bit = {
87 class Unscrambler(object):
88 """A class that unscrambles words quickly by computing a signature
89 (sig) for the word based on its position independent letter
90 population and then using a pregenerated index to look up known
91 words the same set of letters.
93 Note that each instance of Unscrambler caches its index to speed
94 up lookups number 2..N; careless reinstantiation will by slower.
96 Sigs are designed to cluster similar words near each other so both
97 lookup methods support a "fuzzy match" argument that can be set to
98 request similar words that do not match exactly in addition to any
104 # Cached index per instance.
108 if 'unscramble_indexfile' in config.config:
109 indexfile = config.config['unscramble_indexfile']
111 indexfile = "/usr/share/dict/sparse_index"
113 with open(indexfile, 'r') as rf:
114 lines = rf.readlines()
117 (fsig, word) = line.split('+')
119 self.sigs.append(fsig)
120 self.words.append(word)
124 def _compute_word_fingerprint(word: str, population) -> int:
126 for pair in sorted(population.items(), key=lambda x: x[1], reverse=True):
128 if letter in fprint_feature_bit:
132 shift = fprint_feature_bit[letter]
135 return fp << letters_bits
139 def _compute_word_letter_sig(letter_sigs, word: str, population) -> int:
141 for pair in sorted(population.items(), key=lambda x: x[1], reverse=True):
143 if letter not in letter_sigs:
145 s = letter_sigs[letter]
161 @decorator_utils.memoized
162 def compute_word_sig(word: str) -> int:
163 """Given a word, compute its signature for subsequent lookup
164 operations. Signatures are computed based on the letters in
165 the word and their frequencies. We try to cluster "similar"
166 words close to each other in the signature space.
168 >>> train = Unscrambler.compute_word_sig('train')
172 >>> retain = Unscrambler.compute_word_sig('retrain')
180 population = list_utils.population_counts(word)
181 fprint = Unscrambler._compute_word_fingerprint(word, population)
182 letter_sig = Unscrambler._compute_word_letter_sig(letter_sigs, word, population)
183 assert fprint & letter_sig == 0
184 sig = fprint | letter_sig
189 letter_sigs: Dict[str, int],
190 dictfile: str = '/usr/share/dict/words',
191 indexfile: str = '/usr/share/dict/sparse_index',
193 """Before calling this method, change letter_sigs from the default above
194 unless you want to populate the same exact files.
199 with open(dictfile, "r") as f:
201 word = word.replace('\n', '')
203 sig = Unscrambler.compute_word_sig(letter_sigs, word)
204 logger.debug("%s => 0x%x" % (word, sig))
208 if sig in words_by_sigs:
209 words_by_sigs[sig] += ",%s" % word
211 words_by_sigs[sig] = word
212 with open(indexfile, 'w') as f:
213 for sig in sorted(words_by_sigs.keys()):
214 word = words_by_sigs[sig]
215 print(f'0x{sig:x}+{word}', file=f)
217 def lookup(self, word: str, *, include_fuzzy_matches=False) -> Dict[str, bool]:
218 """Looks up a potentially scrambled word optionally including near
221 >>> u = Unscrambler()
222 >>> u.lookup('eanycleocipd', include_fuzzy_matches=False)
223 {'encyclopedia': True}
226 sig = Unscrambler.compute_word_sig(word)
227 return self.lookup_by_sig(sig, include_fuzzy_matches=include_fuzzy_matches)
229 def lookup_by_sig(self, sig, *, include_fuzzy_matches=False) -> Dict[str, bool]:
230 """Looks up a word that has already been translated into a signature by
231 a previous call to Unscrambler.compute_word_sig. Optionally returns
232 near "fuzzy" matches.
234 >>> sig = Unscrambler.compute_word_sig('sunepsapetuargiarin')
238 >>> u = Unscrambler()
239 >>> u.lookup_by_sig(sig, include_fuzzy_matches=True)
240 {'pupigerous': False, 'pupigenous': False, 'unpurposing': False, 'superpurgation': False, 'unsupporting': False, 'superseptuaginarian': True, 'purpurogallin': False, 'scuppaug': False, 'purpurigenous': False, 'purpurogenous': False, 'proppage': False}
244 (exact, location) = list_utils.binary_search(self.sigs, sig)
249 if end > len(self.words):
250 end = len(self.words)
252 for x in range(start, end):
255 if include_fuzzy_matches is True or (fsig == sig):
256 ret[word] = fsig == sig
261 # To repopulate, change letter_sigs and then call Unscrambler.repopulate.
262 # See notes above. See also ~/bin/unscramble.py --populate_destructively.
266 if __name__ == "__main__":