Total Complexity | 44 |
Total Lines | 431 |
Duplicated Lines | 17.63 % |
Coverage | 100% |
Changes | 0 |
Duplicate code is one of the most pungent code smells. A rule that is often used is to re-structure code once it is duplicated in three or more places.
Common duplication problems, and corresponding solutions are:
Complex classes like abydos.distance._discounted_levenshtein often do a lot of different things. To break such a class down, we need to identify a cohesive component within that class. A common approach to find such a component is to look for fields/methods that share the same prefixes, or suffixes.
Once you have determined the fields that belong together, you can apply the Extract Class refactoring. If the component makes sense as a sub-class, Extract Subclass is also a candidate, and is often faster.
1 | # -*- coding: utf-8 -*- |
||
2 | |||
3 | # Copyright 2019 by Christopher C. Little. |
||
4 | # This file is part of Abydos. |
||
5 | # |
||
6 | # Abydos is free software: you can redistribute it and/or modify |
||
7 | # it under the terms of the GNU General Public License as published by |
||
8 | # the Free Software Foundation, either version 3 of the License, or |
||
9 | # (at your option) any later version. |
||
10 | # |
||
11 | # Abydos is distributed in the hope that it will be useful, |
||
12 | # but WITHOUT ANY WARRANTY; without even the implied warranty of |
||
13 | # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
||
14 | # GNU General Public License for more details. |
||
15 | # |
||
16 | # You should have received a copy of the GNU General Public License |
||
17 | # along with Abydos. If not, see <http://www.gnu.org/licenses/>. |
||
18 | |||
19 | 1 | """abydos.distance._discounted_levenshtein. |
|
20 | |||
21 | Discounted Levenshtein edit distance |
||
22 | """ |
||
23 | |||
24 | 1 | from __future__ import ( |
|
25 | absolute_import, |
||
26 | division, |
||
27 | print_function, |
||
28 | unicode_literals, |
||
29 | ) |
||
30 | |||
31 | 1 | from math import log |
|
32 | |||
33 | 1 | from numpy import float as np_float |
|
34 | 1 | from numpy import zeros as np_zeros |
|
35 | |||
36 | 1 | from six.moves import range |
|
37 | |||
38 | 1 | from ._distance import _Distance |
|
39 | |||
40 | 1 | __all__ = ['DiscountedLevenshtein'] |
|
41 | |||
42 | |||
43 | 1 | class DiscountedLevenshtein(_Distance): |
|
44 | """Discounted Levenshtein distance. |
||
45 | |||
46 | This is a variant of Levenshtein distance for which edits later in a string |
||
47 | have discounted cost, on the theory that earlier edits are less likely |
||
48 | than later ones. |
||
49 | |||
50 | .. versionadded:: 0.4.1 |
||
51 | """ |
||
52 | |||
53 | 1 | def __init__( |
|
54 | self, |
||
55 | mode='lev', |
||
56 | normalizer=max, |
||
57 | discount_from=1, |
||
58 | discount_func='log', |
||
59 | vowels='aeiou', |
||
60 | **kwargs |
||
61 | ): |
||
62 | """Initialize DiscountedLevenshtein instance. |
||
63 | |||
64 | Parameters |
||
65 | ---------- |
||
66 | mode : str |
||
67 | Specifies a mode for computing the discounted Levenshtein distance: |
||
68 | |||
69 | - ``lev`` (default) computes the ordinary Levenshtein distance, |
||
70 | in which edits may include inserts, deletes, and |
||
71 | substitutions |
||
72 | - ``osa`` computes the Optimal String Alignment distance, in |
||
73 | which edits may include inserts, deletes, substitutions, and |
||
74 | transpositions but substrings may only be edited once |
||
75 | |||
76 | normalizer : function |
||
77 | A function that takes an list and computes a normalization term |
||
78 | by which the edit distance is divided (max by default). Another |
||
79 | good option is the sum function. |
||
80 | discount_from : int or str |
||
81 | If an int is supplied, this is the first character whose edit cost |
||
82 | will be discounted. If the str ``coda`` is supplied, discounting |
||
83 | will start with the first non-vowel after the first vowel (the |
||
84 | first syllable coda). |
||
85 | discount_func : str or function |
||
86 | The two supported str arguments are ``log``, for a logarithmic |
||
87 | discount function, and ``exp`` for a exponential discount function. |
||
88 | See notes below for information on how to supply your own |
||
89 | discount function. |
||
90 | vowels : str |
||
91 | These are the letters to consider as vowels when discount_from is |
||
92 | set to ``coda``. It defaults to the English vowels 'aeiou', but |
||
93 | it would be reasonable to localize this to other languages or to |
||
94 | add orthographic semi-vowels like 'y', 'w', and even 'h'. |
||
95 | **kwargs |
||
96 | Arbitrary keyword arguments |
||
97 | |||
98 | Notes |
||
99 | ----- |
||
100 | This class is highly experimental and will need additional tuning. |
||
101 | |||
102 | The discount function can be passed as a callable function. It should |
||
103 | expect an integer as its only argument and return a float, ideally |
||
104 | less than or equal to 1.0. The argument represents the degree of |
||
105 | discounting to apply. |
||
106 | |||
107 | |||
108 | .. versionadded:: 0.4.1 |
||
109 | |||
110 | """ |
||
111 | 1 | super(DiscountedLevenshtein, self).__init__(**kwargs) |
|
112 | 1 | self._mode = mode |
|
113 | 1 | self._normalizer = normalizer |
|
114 | 1 | self._discount_from = discount_from |
|
115 | 1 | self._vowels = set(vowels.lower()) |
|
116 | 1 | if callable(discount_func): |
|
117 | 1 | self._cost = discount_func |
|
118 | 1 | elif discount_func == 'exp': |
|
119 | 1 | self._cost = self._exp_discount |
|
120 | else: |
||
121 | 1 | self._cost = self._log_discount |
|
122 | |||
123 | 1 | @staticmethod |
|
124 | def _log_discount(discounts): |
||
125 | 1 | return 1 / (log(1 + discounts / 5) + 1) |
|
126 | |||
127 | 1 | @staticmethod |
|
128 | def _exp_discount(discounts): |
||
129 | 1 | return 1 / (discounts + 1) ** 0.2 |
|
130 | |||
131 | 1 | def _alignment_matrix(self, src, tar): |
|
132 | """Return the Levenshtein alignment matrix. |
||
133 | |||
134 | Parameters |
||
135 | ---------- |
||
136 | src : str |
||
137 | Source string for comparison |
||
138 | tar : str |
||
139 | Target string for comparison |
||
140 | |||
141 | Returns |
||
142 | ------- |
||
143 | numpy.ndarray |
||
144 | The alignment matrix |
||
145 | |||
146 | |||
147 | .. versionadded:: 0.4.1 |
||
148 | |||
149 | """ |
||
150 | 1 | src_len = len(src) |
|
151 | 1 | tar_len = len(tar) |
|
152 | |||
153 | 1 | if self._discount_from == 'coda': |
|
154 | 1 | discount_from = [0, 0] |
|
155 | |||
156 | 1 | src_voc = src.lower() |
|
157 | 1 | for i in range(len(src_voc)): |
|
158 | 1 | if src_voc[i] in self._vowels: |
|
159 | 1 | discount_from[0] = i |
|
160 | 1 | break |
|
161 | 1 | for i in range(discount_from[0], len(src_voc)): |
|
162 | 1 | if src_voc[i] not in self._vowels: |
|
163 | 1 | discount_from[0] = i |
|
164 | 1 | break |
|
165 | else: |
||
166 | 1 | discount_from[0] += 1 |
|
167 | |||
168 | 1 | tar_voc = tar.lower() |
|
169 | 1 | for i in range(len(tar_voc)): |
|
170 | 1 | if tar_voc[i] in self._vowels: |
|
171 | 1 | discount_from[1] = i |
|
172 | 1 | break |
|
173 | 1 | for i in range(discount_from[1], len(tar_voc)): |
|
174 | 1 | if tar_voc[i] not in self._vowels: |
|
175 | 1 | discount_from[1] = i |
|
176 | 1 | break |
|
177 | else: |
||
178 | 1 | discount_from[1] += 1 |
|
179 | |||
180 | 1 | elif isinstance(self._discount_from, int): |
|
181 | 1 | discount_from = [self._discount_from, self._discount_from] |
|
182 | else: |
||
183 | 1 | discount_from = [1, 1] |
|
184 | |||
185 | 1 | d_mat = np_zeros((src_len + 1, tar_len + 1), dtype=np_float) |
|
186 | 1 | for i in range(1, src_len + 1): |
|
187 | 1 | d_mat[i, 0] = d_mat[i - 1, 0] + self._cost( |
|
188 | max(0, i - discount_from[0]) |
||
189 | ) |
||
190 | 1 | for j in range(1, tar_len + 1): |
|
191 | 1 | d_mat[0, j] = d_mat[0, j - 1] + self._cost( |
|
192 | max(0, j - discount_from[1]) |
||
193 | ) |
||
194 | |||
195 | 1 | for i in range(src_len): |
|
196 | 1 | i_extend = self._cost(max(0, i - discount_from[0])) |
|
197 | 1 | for j in range(tar_len): |
|
198 | 1 | cost = min(i_extend, self._cost(max(0, j - discount_from[1]))) |
|
199 | 1 | d_mat[i + 1, j + 1] = min( |
|
200 | d_mat[i + 1, j] + cost, # ins |
||
201 | d_mat[i, j + 1] + cost, # del |
||
202 | d_mat[i, j] + (cost if src[i] != tar[j] else 0), # sub/== |
||
203 | ) |
||
204 | |||
205 | 1 | if self._mode == 'osa': |
|
206 | 1 | if ( |
|
207 | i + 1 > 1 |
||
208 | and j + 1 > 1 |
||
209 | and src[i] == tar[j - 1] |
||
210 | and src[i - 1] == tar[j] |
||
211 | ): |
||
212 | # transposition |
||
213 | 1 | d_mat[i + 1, j + 1] = min( |
|
214 | d_mat[i + 1, j + 1], d_mat[i - 1, j - 1] + cost |
||
215 | ) |
||
216 | |||
217 | 1 | return d_mat |
|
218 | |||
219 | 1 | View Code Duplication | def alignment(self, src, tar): |
|
|||
220 | """Return the Levenshtein alignment of two strings. |
||
221 | |||
222 | Parameters |
||
223 | ---------- |
||
224 | src : str |
||
225 | Source string for comparison |
||
226 | tar : str |
||
227 | Target string for comparison |
||
228 | |||
229 | Returns |
||
230 | ------- |
||
231 | tuple |
||
232 | A tuple containing the Levenshtein distance and the two strings, |
||
233 | aligned. |
||
234 | |||
235 | Examples |
||
236 | -------- |
||
237 | >>> cmp = DiscountedLevenshtein() |
||
238 | >>> cmp.alignment('cat', 'hat') |
||
239 | (1.0, 'cat', 'hat') |
||
240 | >>> cmp.alignment('Niall', 'Neil') |
||
241 | (2.526064024369237, 'N-iall', 'Neil--') |
||
242 | >>> cmp.alignment('aluminum', 'Catalan') |
||
243 | (5.053867269967515, '-aluminum', 'Catalan--') |
||
244 | >>> cmp.alignment('ATCG', 'TAGC') |
||
245 | (2.594032108779918, 'ATCG-', '-TAGC') |
||
246 | |||
247 | >>> cmp = DiscountedLevenshtein(mode='osa') |
||
248 | >>> cmp.alignment('ATCG', 'TAGC') |
||
249 | (1.7482385137517997, 'ATCG', 'TAGC') |
||
250 | >>> cmp.alignment('ACTG', 'TAGC') |
||
251 | (3.342270622531718, '-ACTG', 'TAGC-') |
||
252 | |||
253 | |||
254 | .. versionadded:: 0.4.1 |
||
255 | |||
256 | """ |
||
257 | 1 | d_mat = self._alignment_matrix(src, tar) |
|
258 | |||
259 | 1 | src_aligned = [] |
|
260 | 1 | tar_aligned = [] |
|
261 | |||
262 | 1 | src_pos = len(src) |
|
263 | 1 | tar_pos = len(tar) |
|
264 | |||
265 | 1 | distance = d_mat[src_pos, tar_pos] |
|
266 | |||
267 | 1 | while src_pos and tar_pos: |
|
268 | 1 | up = d_mat[src_pos, tar_pos - 1] |
|
269 | 1 | left = d_mat[src_pos - 1, tar_pos] |
|
270 | 1 | diag = d_mat[src_pos - 1, tar_pos - 1] |
|
271 | |||
272 | 1 | if diag <= min(up, left): |
|
273 | 1 | src_pos -= 1 |
|
274 | 1 | tar_pos -= 1 |
|
275 | 1 | src_aligned.append(src[src_pos]) |
|
276 | 1 | tar_aligned.append(tar[tar_pos]) |
|
277 | 1 | elif up <= left: |
|
278 | 1 | tar_pos -= 1 |
|
279 | 1 | src_aligned.append('-') |
|
280 | 1 | tar_aligned.append(tar[tar_pos]) |
|
281 | else: |
||
282 | 1 | src_pos -= 1 |
|
283 | 1 | src_aligned.append(src[src_pos]) |
|
284 | 1 | tar_aligned.append('-') |
|
285 | 1 | while tar_pos: |
|
286 | 1 | tar_pos -= 1 |
|
287 | 1 | tar_aligned.append(tar[tar_pos]) |
|
288 | 1 | src_aligned.append('-') |
|
289 | 1 | while src_pos: |
|
290 | 1 | src_pos -= 1 |
|
291 | 1 | src_aligned.append(src[src_pos]) |
|
292 | 1 | tar_aligned.append('-') |
|
293 | |||
294 | 1 | return distance, ''.join(src_aligned[::-1]), ''.join(tar_aligned[::-1]) |
|
295 | |||
296 | 1 | def dist_abs(self, src, tar): |
|
297 | """Return the Levenshtein distance between two strings. |
||
298 | |||
299 | Parameters |
||
300 | ---------- |
||
301 | src : str |
||
302 | Source string for comparison |
||
303 | tar : str |
||
304 | Target string for comparison |
||
305 | |||
306 | Returns |
||
307 | ------- |
||
308 | int (may return a float if cost has float values) |
||
309 | The Levenshtein distance between src & tar |
||
310 | |||
311 | Examples |
||
312 | -------- |
||
313 | >>> cmp = DiscountedLevenshtein() |
||
314 | >>> cmp.dist_abs('cat', 'hat') |
||
315 | 1 |
||
316 | >>> cmp.dist_abs('Niall', 'Neil') |
||
317 | 2.526064024369237 |
||
318 | >>> cmp.dist_abs('aluminum', 'Catalan') |
||
319 | 5.053867269967515 |
||
320 | >>> cmp.dist_abs('ATCG', 'TAGC') |
||
321 | 2.594032108779918 |
||
322 | |||
323 | >>> cmp = DiscountedLevenshtein(mode='osa') |
||
324 | >>> cmp.dist_abs('ATCG', 'TAGC') |
||
325 | 1.7482385137517997 |
||
326 | >>> cmp.dist_abs('ACTG', 'TAGC') |
||
327 | 3.342270622531718 |
||
328 | |||
329 | |||
330 | .. versionadded:: 0.4.1 |
||
331 | |||
332 | """ |
||
333 | 1 | src_len = len(src) |
|
334 | 1 | tar_len = len(tar) |
|
335 | |||
336 | 1 | if src == tar: |
|
337 | 1 | return 0 |
|
338 | |||
339 | 1 | if isinstance(self._discount_from, int): |
|
340 | 1 | discount_from = self._discount_from |
|
341 | else: |
||
342 | 1 | discount_from = 1 |
|
343 | |||
344 | 1 | if not src: |
|
345 | 1 | return sum( |
|
346 | self._cost(max(0, pos - discount_from)) |
||
347 | for pos in range(tar_len) |
||
348 | ) |
||
349 | 1 | if not tar: |
|
350 | 1 | return sum( |
|
351 | self._cost(max(0, pos - discount_from)) |
||
352 | for pos in range(src_len) |
||
353 | ) |
||
354 | |||
355 | 1 | d_mat = self._alignment_matrix(src, tar) |
|
356 | |||
357 | 1 | if int(d_mat[src_len, tar_len]) == d_mat[src_len, tar_len]: |
|
358 | 1 | return int(d_mat[src_len, tar_len]) |
|
359 | else: |
||
360 | 1 | return d_mat[src_len, tar_len] |
|
361 | |||
362 | 1 | def dist(self, src, tar): |
|
363 | """Return the normalized Levenshtein distance between two strings. |
||
364 | |||
365 | The Levenshtein distance is normalized by dividing the Levenshtein |
||
366 | distance (calculated by any of the three supported methods) by the |
||
367 | greater of the number of characters in src times the cost of a delete |
||
368 | and the number of characters in tar times the cost of an insert. |
||
369 | For the case in which all operations have :math:`cost = 1`, this is |
||
370 | equivalent to the greater of the length of the two strings src & tar. |
||
371 | |||
372 | Parameters |
||
373 | ---------- |
||
374 | src : str |
||
375 | Source string for comparison |
||
376 | tar : str |
||
377 | Target string for comparison |
||
378 | |||
379 | Returns |
||
380 | ------- |
||
381 | float |
||
382 | The normalized Levenshtein distance between src & tar |
||
383 | |||
384 | Examples |
||
385 | -------- |
||
386 | >>> cmp = DiscountedLevenshtein() |
||
387 | >>> cmp.dist('cat', 'hat') |
||
388 | 0.3513958291799864 |
||
389 | >>> cmp.dist('Niall', 'Neil') |
||
390 | 0.5909885886270658 |
||
391 | >>> cmp.dist('aluminum', 'Catalan') |
||
392 | 0.8348163322045603 |
||
393 | >>> cmp.dist('ATCG', 'TAGC') |
||
394 | 0.7217609721523955 |
||
395 | |||
396 | |||
397 | .. versionadded:: 0.4.1 |
||
398 | |||
399 | """ |
||
400 | 1 | if src == tar: |
|
401 | 1 | return 0 |
|
402 | |||
403 | 1 | if isinstance(self._discount_from, int): |
|
404 | 1 | discount_from = self._discount_from |
|
405 | else: |
||
406 | 1 | discount_from = 1 |
|
407 | |||
408 | 1 | src_len = len(src) |
|
409 | 1 | tar_len = len(tar) |
|
410 | |||
411 | 1 | normalize_term = self._normalizer( |
|
412 | [ |
||
413 | sum( |
||
414 | self._cost(max(0, pos - discount_from)) |
||
415 | for pos in range(src_len) |
||
416 | ), |
||
417 | sum( |
||
418 | self._cost(max(0, pos - discount_from)) |
||
419 | for pos in range(tar_len) |
||
420 | ), |
||
421 | ] |
||
422 | ) |
||
423 | |||
424 | 1 | return self.dist_abs(src, tar) / normalize_term |
|
425 | |||
426 | |||
427 | if __name__ == '__main__': |
||
428 | import doctest |
||
429 | |||
430 | doctest.testmod() |
||
431 |