cassiopee 0.1.0
Sign up to get free protection for your applications and to get access to all the features.
- data/Changelog +1 -0
- data/LICENSE +1 -0
- data/README +2 -0
- data/demo.rb +31 -0
- data/lib/cassiopee.rb +411 -0
- data/tests/test-suite.rb +38 -0
- metadata +72 -0
data/Changelog
ADDED
@@ -0,0 +1 @@
|
|
1
|
+
v0.1.0 : First version
|
data/LICENSE
ADDED
@@ -0,0 +1 @@
|
|
1
|
+
Software is distributed under LGPL v3
|
data/README
ADDED
data/demo.rb
ADDED
@@ -0,0 +1,31 @@
|
|
1
|
+
require File.join(File.dirname(__FILE__), 'lib/cassiopee')
|
2
|
+
require 'rubygems'
|
3
|
+
require 'logger'
|
4
|
+
|
5
|
+
# Instanciate a new crawler
|
6
|
+
crawler = Cassiopee::Crawler.new
|
7
|
+
#crawler.use_store = true
|
8
|
+
|
9
|
+
# String to index
|
10
|
+
crawler.indexString('sallou sallu')
|
11
|
+
# Search pattern in indexed string
|
12
|
+
crawler.searchExact('llo')
|
13
|
+
# Search it again, using already loaded indexed data
|
14
|
+
crawler.searchExact('llo')
|
15
|
+
|
16
|
+
|
17
|
+
test= "my string"
|
18
|
+
# Extend to use match algorithms
|
19
|
+
test.extend(Cassiopee)
|
20
|
+
test.computeDistance('test',0,0)
|
21
|
+
puts "Hamming: " << test.computeHamming("my strigg",1).to_s
|
22
|
+
|
23
|
+
puts "Levenshtein: " << test.computeLevenshtein("mystriigg",3).to_s
|
24
|
+
|
25
|
+
# Approcimate search, edit distance = 1
|
26
|
+
crawler.searchApproximate("llu",-1)
|
27
|
+
|
28
|
+
# Go through matches
|
29
|
+
while((match = crawler.next())!=nil)
|
30
|
+
puts "got a match " << match.inspect
|
31
|
+
end
|
data/lib/cassiopee.rb
ADDED
@@ -0,0 +1,411 @@
|
|
1
|
+
require 'digest/md5'
|
2
|
+
require 'logger'
|
3
|
+
require 'zlib'
|
4
|
+
|
5
|
+
module Cassiopee
|
6
|
+
|
7
|
+
# Calculate the edit or hamming distance between String and pattern
|
8
|
+
# Extend a String
|
9
|
+
# Return -1 if max is reached
|
10
|
+
|
11
|
+
def computeDistance(pattern,hamming,edit)
|
12
|
+
if(edit==0)
|
13
|
+
return computeHamming(pattern,hamming)
|
14
|
+
else
|
15
|
+
return computeLevenshtein(pattern,edit)
|
16
|
+
end
|
17
|
+
end
|
18
|
+
|
19
|
+
# Calculate number of substitution between string and pattern
|
20
|
+
# Extend a String
|
21
|
+
# Return -1 if max is reached
|
22
|
+
|
23
|
+
def computeHamming(pattern,hamming)
|
24
|
+
pattern = pattern.downcase
|
25
|
+
nberr = 0
|
26
|
+
(0..(self.length-1)).each do |c|
|
27
|
+
if(pattern[c] != self[c])
|
28
|
+
nberr = nberr+1
|
29
|
+
if(nberr>hamming.to_i)
|
30
|
+
return -1
|
31
|
+
end
|
32
|
+
end
|
33
|
+
end
|
34
|
+
return nberr
|
35
|
+
end
|
36
|
+
|
37
|
+
# Calculate the edit distance between string and pattern
|
38
|
+
# Extend a String
|
39
|
+
# Return -1 if max is reached
|
40
|
+
|
41
|
+
def computeLevenshtein(pattern,edit)
|
42
|
+
pattern = pattern.downcase
|
43
|
+
matrix= Array.new(2)
|
44
|
+
matrix[0] = Array.new(pattern.length+1)
|
45
|
+
matrix[1] = Array.new(pattern.length+1)
|
46
|
+
(0..(pattern.length)).each do |i|
|
47
|
+
matrix[0][i]=i
|
48
|
+
matrix[1][i]=i
|
49
|
+
end
|
50
|
+
c=0
|
51
|
+
p=1
|
52
|
+
(1..(self.length)).each do |i|
|
53
|
+
c = i.modulo(2)
|
54
|
+
p = (i+1).modulo(2)
|
55
|
+
matrix[c][0] = i
|
56
|
+
(1..(pattern.length)).each do |j|
|
57
|
+
# Bellman's principle of optimality
|
58
|
+
weight = 0
|
59
|
+
if(pattern[i-1] != self[j-1])
|
60
|
+
weight = 1
|
61
|
+
end
|
62
|
+
weight = matrix[p][j-1] + weight
|
63
|
+
if(weight > matrix[p][j] +1)
|
64
|
+
weight = matrix[p][j] +1
|
65
|
+
end
|
66
|
+
if(weight > matrix[c][j-1] +1)
|
67
|
+
weight = matrix[c][j-1] +1
|
68
|
+
end
|
69
|
+
matrix[c][j] = weight
|
70
|
+
end
|
71
|
+
end
|
72
|
+
p = c
|
73
|
+
c = (c + 1).modulo(2)
|
74
|
+
if(matrix[p][pattern.length]>edit)
|
75
|
+
return -1
|
76
|
+
end
|
77
|
+
return matrix[p][pattern.length]
|
78
|
+
|
79
|
+
end
|
80
|
+
|
81
|
+
# Base class to index and search through a string
|
82
|
+
|
83
|
+
class Crawler
|
84
|
+
|
85
|
+
# Use alphabet ambiguity (dna/rna) in search
|
86
|
+
attr_accessor :useAmbiguity
|
87
|
+
# Suffix files name/path
|
88
|
+
attr_accessor :file_suffix
|
89
|
+
# Max number fo threads to use (not yet used)
|
90
|
+
attr_accessor :maxthread
|
91
|
+
# Use persistent suffix file ?
|
92
|
+
attr_accessor :use_store
|
93
|
+
|
94
|
+
FILE_SUFFIX_EXT = ".sfx"
|
95
|
+
FILE_SUFFIX_POS = ".sfp"
|
96
|
+
|
97
|
+
SUFFIXLEN = 'suffix_length'
|
98
|
+
|
99
|
+
$maxthread = 1
|
100
|
+
|
101
|
+
$log = Logger.new(STDOUT)
|
102
|
+
$log.level = Logger::DEBUG
|
103
|
+
|
104
|
+
def initialize
|
105
|
+
@useAmbiguity = false
|
106
|
+
@file_suffix = "crawler"
|
107
|
+
|
108
|
+
@suffix = nil
|
109
|
+
@suffixmd5 = nil
|
110
|
+
@position = 0
|
111
|
+
|
112
|
+
@suffixes = Hash.new
|
113
|
+
|
114
|
+
@matches = nil
|
115
|
+
@curmatch = 0
|
116
|
+
@use_store = false
|
117
|
+
|
118
|
+
@sequence = nil
|
119
|
+
end
|
120
|
+
|
121
|
+
# Clear suffixes in memory
|
122
|
+
|
123
|
+
def clear
|
124
|
+
@suffixes = Hash.new
|
125
|
+
end
|
126
|
+
|
127
|
+
# Set Logger level
|
128
|
+
|
129
|
+
def setLogLevel(level)
|
130
|
+
$log.level = level
|
131
|
+
end
|
132
|
+
|
133
|
+
# Index an input file
|
134
|
+
|
135
|
+
def indexFile(f)
|
136
|
+
# Parse file, map letters to reduced alphabet
|
137
|
+
# Later on, use binary map instead of ascii map
|
138
|
+
# Take all suffix, order by length, link to position map on other file
|
139
|
+
# Store md5 for easier compare? + 20 bytes per suffix
|
140
|
+
@sequence = readSequence(f)
|
141
|
+
|
142
|
+
end
|
143
|
+
|
144
|
+
# Index an input string
|
145
|
+
|
146
|
+
def indexString(s)
|
147
|
+
@sequence = s
|
148
|
+
File.open(@file_suffix+FILE_SUFFIX_EXT, 'w') do |data|
|
149
|
+
data.puts(@sequence)
|
150
|
+
end
|
151
|
+
|
152
|
+
|
153
|
+
end
|
154
|
+
|
155
|
+
# Search exact match
|
156
|
+
|
157
|
+
def searchExact(pattern)
|
158
|
+
pattern = pattern.downcase
|
159
|
+
parseSuffixes(@sequence,pattern.length,pattern.length)
|
160
|
+
|
161
|
+
@matches = Array.new
|
162
|
+
# Search required length, compare (compare md5?)
|
163
|
+
# MD5 = 128 bits, easier to compare for large strings
|
164
|
+
matchsize = pattern.length
|
165
|
+
matchmd5 = Digest::MD5.hexdigest(pattern)
|
166
|
+
@suffixes.each do |md5val,posArray|
|
167
|
+
if (md5val == matchmd5)
|
168
|
+
match = Array[md5val, 0, posArray]
|
169
|
+
$log.debug "Match: " << match.inspect
|
170
|
+
@matches << match
|
171
|
+
end
|
172
|
+
end
|
173
|
+
return @matches
|
174
|
+
|
175
|
+
end
|
176
|
+
|
177
|
+
# Search an approximate string
|
178
|
+
#
|
179
|
+
# * support insertion, deletion, substitution
|
180
|
+
# * If edit > 0, use Hamming
|
181
|
+
# * Else use Levenshtein
|
182
|
+
|
183
|
+
|
184
|
+
def searchApproximate(s,edit)
|
185
|
+
if(edit==0)
|
186
|
+
return searchExact(s)
|
187
|
+
end
|
188
|
+
|
189
|
+
if(edit>0)
|
190
|
+
useHamming = true
|
191
|
+
minmatchsize = s.length
|
192
|
+
maxmatchsize = s.length
|
193
|
+
else
|
194
|
+
useHamming = false
|
195
|
+
edit = edit * (-1)
|
196
|
+
minmatchsize = s.length - edit
|
197
|
+
maxmatchsize = s.length + edit
|
198
|
+
end
|
199
|
+
|
200
|
+
parseSuffixes(@sequence,minmatchsize,maxmatchsize)
|
201
|
+
|
202
|
+
matchmd5 = Digest::MD5.hexdigest(s)
|
203
|
+
|
204
|
+
@matches = Array.new
|
205
|
+
|
206
|
+
@suffixes.each do |md5val,posArray|
|
207
|
+
if(md5val == SUFFIXLEN)
|
208
|
+
next
|
209
|
+
end
|
210
|
+
if (md5val == matchmd5)
|
211
|
+
match = Array[md5val, 0, posArray]
|
212
|
+
$log.debug "Match: " << match.inspect
|
213
|
+
@matches << match
|
214
|
+
else
|
215
|
+
if(posArray[0]>= minmatchsize && posArray[0] <= maxmatchsize)
|
216
|
+
# Get string
|
217
|
+
seq = extractSuffix(posArray[1],posArray[0])
|
218
|
+
seq.extend(Cassiopee)
|
219
|
+
if(useHamming)
|
220
|
+
errors = seq.computeHamming(s,edit)
|
221
|
+
else
|
222
|
+
errors = seq.computeLevenshtein(s,edit)
|
223
|
+
end
|
224
|
+
if(errors>=0)
|
225
|
+
match = Array[md5val, errors, posArray]
|
226
|
+
$log.debug "Match: " << match.inspect
|
227
|
+
@matches << match
|
228
|
+
end
|
229
|
+
end
|
230
|
+
end
|
231
|
+
|
232
|
+
end
|
233
|
+
|
234
|
+
return @matches
|
235
|
+
end
|
236
|
+
|
237
|
+
# Extract un suffix from suffix file based on md5 match
|
238
|
+
|
239
|
+
def extractSuffix(start,len)
|
240
|
+
sequence = ''
|
241
|
+
begin
|
242
|
+
file = File.new(@file_suffix+FILE_SUFFIX_EXT, "r")
|
243
|
+
file.pos = start
|
244
|
+
sequence = file.read(len)
|
245
|
+
file.close
|
246
|
+
rescue => err
|
247
|
+
puts "Exception: #{err}"
|
248
|
+
return nil
|
249
|
+
end
|
250
|
+
return sequence
|
251
|
+
end
|
252
|
+
|
253
|
+
# Iterates over matches
|
254
|
+
|
255
|
+
def next
|
256
|
+
if(@curmatch<@matches.length)
|
257
|
+
@curmatch = @curmatch + 1
|
258
|
+
return @matches[@curmatch-1]
|
259
|
+
else
|
260
|
+
@curmatch = 0
|
261
|
+
return nil
|
262
|
+
end
|
263
|
+
end
|
264
|
+
|
265
|
+
def to_s
|
266
|
+
puts '{ matches: "' << @matches.length << '" }'
|
267
|
+
end
|
268
|
+
|
269
|
+
private
|
270
|
+
|
271
|
+
# Parse input string
|
272
|
+
#
|
273
|
+
# * creates a suffix file
|
274
|
+
# * creates a suffix position file
|
275
|
+
|
276
|
+
def parseSuffixes(s,minlen,maxlen)
|
277
|
+
|
278
|
+
# Controls
|
279
|
+
if(minlen<=0)
|
280
|
+
minlen = 1
|
281
|
+
end
|
282
|
+
if(maxlen>@sequence.length)
|
283
|
+
maxlen = @sequence.length
|
284
|
+
end
|
285
|
+
|
286
|
+
|
287
|
+
suffixlen = nil
|
288
|
+
$log.info('Start indexing')
|
289
|
+
loaded = false
|
290
|
+
# Hash in memory already contain suffixes for searched lengths
|
291
|
+
if(@suffixes != nil && !@suffixes.empty?)
|
292
|
+
suffixlen = @suffixes[SUFFIXLEN]
|
293
|
+
if(suffixlen!=nil && !suffixlen.empty?)
|
294
|
+
loaded = true
|
295
|
+
(maxlen).downto(minlen) do |len|
|
296
|
+
if(suffixlen.index(len)==nil)
|
297
|
+
loaded = false
|
298
|
+
break
|
299
|
+
end
|
300
|
+
end
|
301
|
+
end
|
302
|
+
end
|
303
|
+
|
304
|
+
if(@use_store && loaded)
|
305
|
+
$log.debug('already in memory, skip file loading')
|
306
|
+
end
|
307
|
+
|
308
|
+
# If not already in memory
|
309
|
+
if(@use_store && !loaded)
|
310
|
+
@suffixes = loadSuffixes(@file_suffix+FILE_SUFFIX_POS)
|
311
|
+
suffixlen = @suffixes[SUFFIXLEN]
|
312
|
+
end
|
313
|
+
|
314
|
+
nbSuffix = 0
|
315
|
+
changed = false
|
316
|
+
|
317
|
+
# Load suffix between maxlen and minlen
|
318
|
+
(maxlen).downto(minlen) do |i|
|
319
|
+
$log.debug('parse for length ' << i.to_s)
|
320
|
+
if(suffixlen!=nil && suffixlen.index(i)!=nil)
|
321
|
+
$log.debug('length '<<i <<'already parsed')
|
322
|
+
next
|
323
|
+
end
|
324
|
+
changed = true
|
325
|
+
(0..(s.length-maxlen)).each do |j|
|
326
|
+
@suffix = s[j,i]
|
327
|
+
@suffixmd5 = Digest::MD5.hexdigest(@suffix)
|
328
|
+
@position = j
|
329
|
+
#$log.debug("add "+@suffix+" at pos "+@position.to_s)
|
330
|
+
nbSuffix += addSuffix(@suffixmd5, @position,i)
|
331
|
+
end
|
332
|
+
$log.debug("Nb suffix found: " << nbSuffix.to_s << ' for length ' << i.to_s)
|
333
|
+
end
|
334
|
+
|
335
|
+
|
336
|
+
if(@use_store && changed)
|
337
|
+
$log.info("Store suffixes")
|
338
|
+
marshal_dump = Marshal.dump(@suffixes)
|
339
|
+
sfxpos = File.new(@file_suffix+FILE_SUFFIX_POS,'w')
|
340
|
+
sfxpos = Zlib::GzipWriter.new(sfxpos)
|
341
|
+
sfxpos.write marshal_dump
|
342
|
+
sfxpos.close
|
343
|
+
end
|
344
|
+
$log.info('End of indexing')
|
345
|
+
end
|
346
|
+
|
347
|
+
|
348
|
+
# Add a suffix in Hashmap
|
349
|
+
|
350
|
+
def addSuffix(md5val,position,len)
|
351
|
+
if(@suffixes.has_key?(md5val))
|
352
|
+
# Add position
|
353
|
+
@suffixes[md5val] << position
|
354
|
+
else
|
355
|
+
# Add position, write new suffix
|
356
|
+
# First elt is size of elt
|
357
|
+
@suffixes[md5val] = Array[len, position]
|
358
|
+
if(@suffixes.has_key?(SUFFIXLEN))
|
359
|
+
@suffixes[SUFFIXLEN] << len
|
360
|
+
else
|
361
|
+
@suffixes[SUFFIXLEN] = Array[len]
|
362
|
+
end
|
363
|
+
end
|
364
|
+
return 1
|
365
|
+
end
|
366
|
+
|
367
|
+
# read input string, and concat content
|
368
|
+
|
369
|
+
def readSequence(s)
|
370
|
+
$log.debug('read input sequence')
|
371
|
+
counter = 1
|
372
|
+
sequence = ''
|
373
|
+
begin
|
374
|
+
file = File.new(s, "r")
|
375
|
+
File.delete(@file_suffix+FILE_SUFFIX_EXT) unless !File.exists?(@file_suffix+FILE_SUFFIX_EXT)
|
376
|
+
File.open(@file_suffix+FILE_SUFFIX_EXT, 'w') do |data|
|
377
|
+
while (line = file.gets)
|
378
|
+
counter = counter + 1
|
379
|
+
input = line.downcase.chomp
|
380
|
+
sequence << input
|
381
|
+
data.puts input
|
382
|
+
end
|
383
|
+
|
384
|
+
end
|
385
|
+
file.close
|
386
|
+
rescue => err
|
387
|
+
puts "Exception: #{err}"
|
388
|
+
err
|
389
|
+
end
|
390
|
+
$log.debug('data file created')
|
391
|
+
return sequence
|
392
|
+
end
|
393
|
+
|
394
|
+
# Load suffix position file in memory
|
395
|
+
|
396
|
+
def loadSuffixes(file_name)
|
397
|
+
return Hash.new unless File.exists?(@file_suffix+FILE_SUFFIX_POS)
|
398
|
+
begin
|
399
|
+
file = Zlib::GzipReader.open(file_name)
|
400
|
+
rescue Zlib::GzipFile::Error
|
401
|
+
file = File.open(file_name, 'r')
|
402
|
+
ensure
|
403
|
+
obj = Marshal.load file.read
|
404
|
+
file.close
|
405
|
+
return obj
|
406
|
+
end
|
407
|
+
end
|
408
|
+
|
409
|
+
end
|
410
|
+
|
411
|
+
end
|
data/tests/test-suite.rb
ADDED
@@ -0,0 +1,38 @@
|
|
1
|
+
require File.join(File.dirname(__FILE__), '../lib/cassiopee')
|
2
|
+
require 'rubygems'
|
3
|
+
require 'logger'
|
4
|
+
require 'test/unit'
|
5
|
+
|
6
|
+
class TestCrawler < Test::Unit::TestCase
|
7
|
+
|
8
|
+
|
9
|
+
def test_exactsearch
|
10
|
+
crawler = Cassiopee::Crawler.new
|
11
|
+
crawler.setLogLevel(Logger::ERROR)
|
12
|
+
crawler.indexString('my sample example')
|
13
|
+
matches = crawler.searchExact('ampl')
|
14
|
+
assert_equal(1,matches.length)
|
15
|
+
# Minus 1, because first element is len of match
|
16
|
+
match = crawler.next()
|
17
|
+
assert_equal(2,match[2].length-1)
|
18
|
+
end
|
19
|
+
|
20
|
+
|
21
|
+
def test_hammingsearch
|
22
|
+
crawler = Cassiopee::Crawler.new
|
23
|
+
crawler.indexString('my sample example')
|
24
|
+
matches = crawler.searchApproximate('ebampl',1)
|
25
|
+
assert_equal(1,matches.length)
|
26
|
+
end
|
27
|
+
|
28
|
+
def test_levenshteinsearch
|
29
|
+
crawler = Cassiopee::Crawler.new
|
30
|
+
crawler.indexString('my sample example')
|
31
|
+
matches = crawler.searchApproximate('ebampl',-1)
|
32
|
+
assert_equal(1,matches.length)
|
33
|
+
end
|
34
|
+
|
35
|
+
end
|
36
|
+
|
37
|
+
|
38
|
+
|
metadata
ADDED
@@ -0,0 +1,72 @@
|
|
1
|
+
--- !ruby/object:Gem::Specification
|
2
|
+
name: cassiopee
|
3
|
+
version: !ruby/object:Gem::Version
|
4
|
+
hash: 27
|
5
|
+
prerelease: false
|
6
|
+
segments:
|
7
|
+
- 0
|
8
|
+
- 1
|
9
|
+
- 0
|
10
|
+
version: 0.1.0
|
11
|
+
platform: ruby
|
12
|
+
authors:
|
13
|
+
- Olivier Sallou
|
14
|
+
autorequire:
|
15
|
+
bindir: bin
|
16
|
+
cert_chain: []
|
17
|
+
|
18
|
+
date: 2011-09-04 00:00:00 +02:00
|
19
|
+
default_executable:
|
20
|
+
dependencies: []
|
21
|
+
|
22
|
+
description: Cassiopee index one String and provide methods to search exact match or approximate matches with Hammming and/or edit distance.
|
23
|
+
email: olivier.sallou@gmail.com
|
24
|
+
executables: []
|
25
|
+
|
26
|
+
extensions: []
|
27
|
+
|
28
|
+
extra_rdoc_files: []
|
29
|
+
|
30
|
+
files:
|
31
|
+
- README
|
32
|
+
- Changelog
|
33
|
+
- LICENSE
|
34
|
+
- demo.rb
|
35
|
+
- lib/cassiopee.rb
|
36
|
+
- tests/test-suite.rb
|
37
|
+
has_rdoc: true
|
38
|
+
homepage: https://github.com/osallou/cassiopee
|
39
|
+
licenses:
|
40
|
+
- LGPL-3
|
41
|
+
post_install_message:
|
42
|
+
rdoc_options:
|
43
|
+
- --main lib/cassiopee.rb
|
44
|
+
require_paths:
|
45
|
+
- lib
|
46
|
+
required_ruby_version: !ruby/object:Gem::Requirement
|
47
|
+
none: false
|
48
|
+
requirements:
|
49
|
+
- - ">="
|
50
|
+
- !ruby/object:Gem::Version
|
51
|
+
hash: 3
|
52
|
+
segments:
|
53
|
+
- 0
|
54
|
+
version: "0"
|
55
|
+
required_rubygems_version: !ruby/object:Gem::Requirement
|
56
|
+
none: false
|
57
|
+
requirements:
|
58
|
+
- - ">="
|
59
|
+
- !ruby/object:Gem::Version
|
60
|
+
hash: 3
|
61
|
+
segments:
|
62
|
+
- 0
|
63
|
+
version: "0"
|
64
|
+
requirements: []
|
65
|
+
|
66
|
+
rubyforge_project:
|
67
|
+
rubygems_version: 1.3.7
|
68
|
+
signing_key:
|
69
|
+
specification_version: 3
|
70
|
+
summary: Cassiopee index strings and provide exact or approximate search.
|
71
|
+
test_files:
|
72
|
+
- tests/test-suite.rb
|