ponysay/src/spellocorrecter.py
Mattias Andrée 5df821d7cd splitting into multiple files
Signed-off-by: Mattias Andrée <maandree@operamail.com>
2013-04-02 11:21:33 +02:00

216 lines
8 KiB
Python

#!/usr/bin/env python3
# -*- coding: utf-8 -*-
'''
ponysay - Ponysay, cowsay reimplementation for ponies
Copyright (C) 2012, 2013 Erkin Batu Altunbaş et al.
This program is free software. It comes without any warranty, to
the extent permitted by applicable law. You can redistribute it
and/or modify it under the terms of the Do What The Fuck You Want
To Public License, Version 2, as published by Sam Hocevar. See
http://sam.zoy.org/wtfpl/COPYING for more details.
'''
from common import *
'''
Class used for correcting spellos and typos,
Note that this implementation will not find that correctly spelled word are correct faster than it corrects words.
It is also limited to words of size 0 to 127 (inclusive)
'''
class SpelloCorrecter(): # Naïvely and quickly ported and adapted from optimised Java, may not be the nicest, or even fast, Python code
'''
Constructor
@param directories:list<str> List of directories that contains the file names with the correct spelling
@param ending:str The file name ending of the correctly spelled file names, this is removed for the name
'''
def __init__(self, directories, ending):
self.weights = {'k' : {'c' : 0.25, 'g' : 0.75, 'q' : 0.125},
'c' : {'k' : 0.25, 'g' : 0.75, 's' : 0.5, 'z' : 0.5, 'q' : 0.125},
's' : {'z' : 0.25, 'c' : 0.5},
'z' : {'s' : 0.25, 'c' : 0.5},
'g' : {'k' : 0.75, 'c' : 0.75, 'q' : 0.9},
'o' : {'u' : 0.5},
'u' : {'o' : 0.5, 'v' : 0.75, 'w' : 0.5},
'b' : {'v' : 0.75},
'v' : {'b' : 0.75, 'w' : 0.5, 'u' : 0.7},
'w' : {'v' : 0.5, 'u' : 0.5},
'q' : {'c' : 0.125, 'k' : 0.125, 'g' : 0.9}}
self.corrections = None
self.dictionary = [None] * 513
self.reusable = [0] * 512
self.dictionaryEnd = 512
self.closestDistance = 0
self.M = [None] * 128
for y in range(0, 128):
self.M[y] = [0] * 128
self.M[y][0] = y
m0 = self.M[0]
x = 127
while x > -1:
m0[x] = x
x -= 1
previous = ''
self.dictionary[-1] = previous;
for directory in directories:
for filename in os.listdir(directory):
if (not endswith(filename, ending)) or (len(filename) - len(ending) > 127):
continue
proper = filename[:-len(ending)]
if self.dictionaryEnd == 0:
self.dictionaryEnd = len(self.dictionary)
self.reusable = [0] * self.dictionaryEnd + self.reusable
self.dictionary = [None] * self.dictionaryEnd + self.dictionary
self.dictionaryEnd -= 1
self.dictionary[self.dictionaryEnd] = proper
prevCommon = min(len(previous), len(proper))
for i in range(0, prevCommon):
if previous[i] != proper[i]:
prevCommon = i
break
previous = proper
self.reusable[self.dictionaryEnd] = prevCommon
#part = self.dictionary[self.dictionaryEnd : len(self.dictionary) - 1]
#part.sort()
#self.dictionary[self.dictionaryEnd : len(self.dictionary) - 1] = part
#
#index = len(self.dictionary) - 1
#while index >= self.dictionaryEnd:
# proper = self.dictionary[index]
# prevCommon = min(len(previous), len(proper))
# for i in range(0, prevCommon):
# if previous[i] != proper[i]:
# prevCommon = i
# break
# previous = proper
# self.reusable[self.dictionaryEnd] = prevCommon
# index -= 1;
'''
Finds the closests correct spelled word
@param used:str The word to correct
@return (words, distance):(list<string>, int) A list the closest spellings and the weighted distance
'''
def correct(self, used):
if len(used) > 127:
return ([used], 0)
self.__correct(used)
return (self.corrections, self.closestDistance)
'''
Finds the closests correct spelled word
@param used:str The word to correct, it must satisfy all restrictions
'''
def __correct(self, used):
self.closestDistance = 0x7FFFFFFF
previous = self.dictionary[-1]
prevLen = 0
usedLen = len(used)
proper = None
prevCommon = 0
d = len(self.dictionary) - 1
while d > self.dictionaryEnd:
d -= 1
proper = self.dictionary[d]
if abs(len(proper) - usedLen) <= self.closestDistance:
if previous == self.dictionary[d + 1]:
prevCommon = self.reusable[d];
else:
prevCommon = min(prevLen, len(proper))
for i in range(0, prevCommon):
if previous[i] != proper[i]:
prevCommon = i
break
skip = min(prevLen, len(proper))
i = prevCommon
while i < skip:
for u in range(0, usedLen):
if (used[u] == previous[i]) or (used[u] == proper[i]):
skip = i
break
i += 1
common = min(skip, min(usedLen, len(proper)))
for i in range(0, common):
if used[i] != proper[i]:
common = i
break
distance = self.__distance(proper, skip, len(proper), used, common, usedLen)
if self.closestDistance > distance:
self.closestDistance = distance
self.corrections = [proper]
elif self.closestDistance == distance:
self.corrections.append(proper)
previous = proper;
if distance >= 0x7FFFFF00:
prevLen = distance & 255
else:
prevLen = len(proper)
'''
Calculate the distance between a correct word and a incorrect word
@param proper:str The correct word
@param y0:int The offset for `proper`
@param yn:int The length, before applying `y0`, of `proper`
@param used:str The incorrect word
@param x0:int The offset for `used`
@param xn:int The length, before applying `x0`, of `used`
@return :float The distance between the words
'''
def __distance(self, proper, y0, yn, used, x0, xn):
my = self.M[y0]
for y in range(y0, yn):
best = 0x7FFFFFFF
p = proper[y]
myy = self.M[y + 1] # only one array bound check, and at most one + ☺
x = x0
while x < xn:
change = my[x]
u = used[x]
if p == u:
# commence black magick … twilight would be so disappointed
x += 1
myy[x] = change
best = min(best, change)
remove = myy[x]
add = my[x + 1]
cw = 1
if my[x] in self.weights:
if p in self.weights[u]:
cw = self.weights[u][p]
x += 1
myy[x] = min(cw + change, 1 + min(remove, add))
if best > myy[x]:
best = myy[x]
if best > self.closestDistance:
return 0x7FFFFF00 | y
my = myy
return my[xn]