annotate roundup/backends/indexer_dbm.py @ 3068:c4e76c84f43d

another fix to indexargs
author Richard Jones <richard@users.sourceforge.net>
date Wed, 05 Jan 2005 21:57:46 +0000
parents 1c063814d567
children 2817a4db901d
Ignore whitespace changes - Everywhere: Within whitespace: At end of lines:
rev   line source
2089
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
1 #
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
2 # This module is derived from the module described at:
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
3 # http://gnosis.cx/publish/programming/charming_python_15.txt
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
4 #
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
5 # Author: David Mertz (mertz@gnosis.cx)
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
6 # Thanks to: Pat Knight (p.knight@ktgroup.co.uk)
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
7 # Gregory Popovitch (greg@gpy.com)
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
8 #
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
9 # The original module was released under this license, and remains under
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
10 # it:
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
11 #
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
12 # This file is released to the public domain. I (dqm) would
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
13 # appreciate it if you choose to keep derived works under terms
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
14 # that promote freedom, but obviously am giving up any rights
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
15 # to compel such.
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
16 #
3058
1c063814d567 Move search method duplicated in indexer_dbm and indexer_tsearch2...
Johannes Gijsbers <jlgijsbers@users.sourceforge.net>
parents: 2872
diff changeset
17 #$Id: indexer_dbm.py,v 1.3 2005-01-04 16:48:46 jlgijsbers Exp $
2089
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
18 '''This module provides an indexer class, RoundupIndexer, that stores text
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
19 indices in a roundup instance. This class makes searching the content of
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
20 messages, string properties and text files possible.
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
21 '''
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
22 __docformat__ = 'restructuredtext'
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
23
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
24 import os, shutil, re, mimetypes, marshal, zlib, errno
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
25 from roundup.hyperdb import Link, Multilink
3058
1c063814d567 Move search method duplicated in indexer_dbm and indexer_tsearch2...
Johannes Gijsbers <jlgijsbers@users.sourceforge.net>
parents: 2872
diff changeset
26 from roundup.backends.indexer_common import Indexer
2089
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
27
2872
d530b68e4b42 don't index common words [SF#1046612]
Richard Jones <richard@users.sourceforge.net>
parents: 2089
diff changeset
28 stopwords = [
d530b68e4b42 don't index common words [SF#1046612]
Richard Jones <richard@users.sourceforge.net>
parents: 2089
diff changeset
29 "A", "AND", "ARE", "AS", "AT", "BE", "BUT", "BY",
d530b68e4b42 don't index common words [SF#1046612]
Richard Jones <richard@users.sourceforge.net>
parents: 2089
diff changeset
30 "FOR", "IF", "IN", "INTO", "IS", "IT",
d530b68e4b42 don't index common words [SF#1046612]
Richard Jones <richard@users.sourceforge.net>
parents: 2089
diff changeset
31 "NO", "NOT", "OF", "ON", "OR", "SUCH",
d530b68e4b42 don't index common words [SF#1046612]
Richard Jones <richard@users.sourceforge.net>
parents: 2089
diff changeset
32 "THAT", "THE", "THEIR", "THEN", "THERE", "THESE",
d530b68e4b42 don't index common words [SF#1046612]
Richard Jones <richard@users.sourceforge.net>
parents: 2089
diff changeset
33 "THEY", "THIS", "TO", "WAS", "WILL", "WITH"
d530b68e4b42 don't index common words [SF#1046612]
Richard Jones <richard@users.sourceforge.net>
parents: 2089
diff changeset
34 ]
d530b68e4b42 don't index common words [SF#1046612]
Richard Jones <richard@users.sourceforge.net>
parents: 2089
diff changeset
35 is_stopword = {}
d530b68e4b42 don't index common words [SF#1046612]
Richard Jones <richard@users.sourceforge.net>
parents: 2089
diff changeset
36 for word in stopwords:
d530b68e4b42 don't index common words [SF#1046612]
Richard Jones <richard@users.sourceforge.net>
parents: 2089
diff changeset
37 is_stopword[word] = None
d530b68e4b42 don't index common words [SF#1046612]
Richard Jones <richard@users.sourceforge.net>
parents: 2089
diff changeset
38 is_stopword = is_stopword.has_key
d530b68e4b42 don't index common words [SF#1046612]
Richard Jones <richard@users.sourceforge.net>
parents: 2089
diff changeset
39
3058
1c063814d567 Move search method duplicated in indexer_dbm and indexer_tsearch2...
Johannes Gijsbers <jlgijsbers@users.sourceforge.net>
parents: 2872
diff changeset
40 class Indexer(Indexer):
2089
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
41 '''Indexes information from roundup's hyperdb to allow efficient
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
42 searching.
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
43
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
44 Three structures are created by the indexer::
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
45
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
46 files {identifier: (fileid, wordcount)}
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
47 words {word: {fileid: count}}
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
48 fileids {fileid: identifier}
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
49
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
50 where identifier is (classname, nodeid, propertyname)
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
51 '''
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
52 def __init__(self, db_path):
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
53 self.indexdb_path = os.path.join(db_path, 'indexes')
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
54 self.indexdb = os.path.join(self.indexdb_path, 'index.db')
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
55 self.reindex = 0
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
56 self.quiet = 9
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
57 self.changed = 0
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
58
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
59 # see if we need to reindex because of a change in code
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
60 version = os.path.join(self.indexdb_path, 'version')
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
61 if (not os.path.exists(self.indexdb_path) or
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
62 not os.path.exists(version)):
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
63 # for now the file itself is a flag
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
64 self.force_reindex()
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
65 elif os.path.exists(version):
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
66 version = open(version).read()
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
67 # check the value and reindex if it's not the latest
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
68 if version.strip() != '1':
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
69 self.force_reindex()
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
70
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
71 def force_reindex(self):
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
72 '''Force a reindex condition
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
73 '''
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
74 if os.path.exists(self.indexdb_path):
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
75 shutil.rmtree(self.indexdb_path)
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
76 os.makedirs(self.indexdb_path)
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
77 os.chmod(self.indexdb_path, 0775)
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
78 open(os.path.join(self.indexdb_path, 'version'), 'w').write('1\n')
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
79 self.reindex = 1
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
80 self.changed = 1
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
81
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
82 def should_reindex(self):
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
83 '''Should we reindex?
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
84 '''
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
85 return self.reindex
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
86
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
87 def add_text(self, identifier, text, mime_type='text/plain'):
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
88 '''Add some text associated with the (classname, nodeid, property)
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
89 identifier.
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
90 '''
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
91 # make sure the index is loaded
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
92 self.load_index()
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
93
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
94 # remove old entries for this identifier
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
95 if self.files.has_key(identifier):
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
96 self.purge_entry(identifier)
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
97
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
98 # split into words
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
99 words = self.splitter(text, mime_type)
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
100
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
101 # Find new file index, and assign it to identifier
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
102 # (_TOP uses trick of negative to avoid conflict with file index)
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
103 self.files['_TOP'] = (self.files['_TOP'][0]-1, None)
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
104 file_index = abs(self.files['_TOP'][0])
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
105 self.files[identifier] = (file_index, len(words))
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
106 self.fileids[file_index] = identifier
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
107
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
108 # find the unique words
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
109 filedict = {}
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
110 for word in words:
2872
d530b68e4b42 don't index common words [SF#1046612]
Richard Jones <richard@users.sourceforge.net>
parents: 2089
diff changeset
111 if is_stopword(word):
d530b68e4b42 don't index common words [SF#1046612]
Richard Jones <richard@users.sourceforge.net>
parents: 2089
diff changeset
112 continue
2089
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
113 if filedict.has_key(word):
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
114 filedict[word] = filedict[word]+1
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
115 else:
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
116 filedict[word] = 1
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
117
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
118 # now add to the totals
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
119 for word in filedict.keys():
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
120 # each word has a dict of {identifier: count}
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
121 if self.words.has_key(word):
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
122 entry = self.words[word]
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
123 else:
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
124 # new word
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
125 entry = {}
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
126 self.words[word] = entry
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
127
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
128 # make a reference to the file for this word
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
129 entry[file_index] = filedict[word]
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
130
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
131 # save needed
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
132 self.changed = 1
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
133
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
134 def splitter(self, text, ftype):
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
135 '''Split the contents of a text string into a list of 'words'
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
136 '''
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
137 if ftype == 'text/plain':
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
138 words = self.text_splitter(text)
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
139 else:
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
140 return []
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
141 return words
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
142
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
143 def text_splitter(self, text):
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
144 """Split text/plain string into a list of words
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
145 """
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
146 # case insensitive
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
147 text = str(text).upper()
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
148
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
149 # Split the raw text, losing anything longer than 25 characters
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
150 # since that'll be gibberish (encoded text or somesuch) or shorter
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
151 # than 3 characters since those short words appear all over the
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
152 # place
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
153 return re.findall(r'\b\w{2,25}\b', text)
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
154
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
155 # we override this to ignore not 2 < word < 25 and also to fix a bug -
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
156 # the (fail) case.
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
157 def find(self, wordlist):
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
158 '''Locate files that match ALL the words in wordlist
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
159 '''
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
160 if not hasattr(self, 'words'):
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
161 self.load_index()
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
162 self.load_index(wordlist=wordlist)
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
163 entries = {}
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
164 hits = None
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
165 for word in wordlist:
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
166 if not 2 < len(word) < 25:
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
167 # word outside the bounds of what we index - ignore
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
168 continue
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
169 word = word.upper()
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
170 entry = self.words.get(word) # For each word, get index
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
171 entries[word] = entry # of matching files
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
172 if not entry: # Nothing for this one word (fail)
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
173 return {}
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
174 if hits is None:
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
175 hits = {}
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
176 for k in entry.keys():
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
177 if not self.fileids.has_key(k):
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
178 raise ValueError, 'Index is corrupted: re-generate it'
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
179 hits[k] = self.fileids[k]
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
180 else:
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
181 # Eliminate hits for every non-match
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
182 for fileid in hits.keys():
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
183 if not entry.has_key(fileid):
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
184 del hits[fileid]
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
185 if hits is None:
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
186 return {}
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
187 return hits
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
188
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
189 segments = "0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZ#_-!"
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
190 def load_index(self, reload=0, wordlist=None):
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
191 # Unless reload is indicated, do not load twice
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
192 if self.index_loaded() and not reload:
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
193 return 0
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
194
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
195 # Ok, now let's actually load it
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
196 db = {'WORDS': {}, 'FILES': {'_TOP':(0,None)}, 'FILEIDS': {}}
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
197
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
198 # Identify the relevant word-dictionary segments
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
199 if not wordlist:
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
200 segments = self.segments
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
201 else:
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
202 segments = ['-','#']
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
203 for word in wordlist:
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
204 segments.append(word[0].upper())
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
205
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
206 # Load the segments
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
207 for segment in segments:
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
208 try:
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
209 f = open(self.indexdb + segment, 'rb')
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
210 except IOError, error:
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
211 # probably just nonexistent segment index file
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
212 if error.errno != errno.ENOENT: raise
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
213 else:
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
214 pickle_str = zlib.decompress(f.read())
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
215 f.close()
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
216 dbslice = marshal.loads(pickle_str)
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
217 if dbslice.get('WORDS'):
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
218 # if it has some words, add them
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
219 for word, entry in dbslice['WORDS'].items():
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
220 db['WORDS'][word] = entry
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
221 if dbslice.get('FILES'):
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
222 # if it has some files, add them
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
223 db['FILES'] = dbslice['FILES']
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
224 if dbslice.get('FILEIDS'):
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
225 # if it has fileids, add them
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
226 db['FILEIDS'] = dbslice['FILEIDS']
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
227
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
228 self.words = db['WORDS']
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
229 self.files = db['FILES']
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
230 self.fileids = db['FILEIDS']
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
231 self.changed = 0
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
232
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
233 def save_index(self):
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
234 # only save if the index is loaded and changed
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
235 if not self.index_loaded() or not self.changed:
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
236 return
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
237
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
238 # brutal space saver... delete all the small segments
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
239 for segment in self.segments:
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
240 try:
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
241 os.remove(self.indexdb + segment)
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
242 except OSError, error:
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
243 # probably just nonexistent segment index file
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
244 if error.errno != errno.ENOENT: raise
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
245
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
246 # First write the much simpler filename/fileid dictionaries
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
247 dbfil = {'WORDS':None, 'FILES':self.files, 'FILEIDS':self.fileids}
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
248 open(self.indexdb+'-','wb').write(zlib.compress(marshal.dumps(dbfil)))
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
249
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
250 # The hard part is splitting the word dictionary up, of course
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
251 letters = "0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZ#_"
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
252 segdicts = {} # Need batch of empty dicts
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
253 for segment in letters:
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
254 segdicts[segment] = {}
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
255 for word, entry in self.words.items(): # Split into segment dicts
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
256 initchar = word[0].upper()
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
257 segdicts[initchar][word] = entry
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
258
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
259 # save
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
260 for initchar in letters:
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
261 db = {'WORDS':segdicts[initchar], 'FILES':None, 'FILEIDS':None}
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
262 pickle_str = marshal.dumps(db)
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
263 filename = self.indexdb + initchar
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
264 pickle_fh = open(filename, 'wb')
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
265 pickle_fh.write(zlib.compress(pickle_str))
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
266 os.chmod(filename, 0664)
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
267
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
268 # save done
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
269 self.changed = 0
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
270
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
271 def purge_entry(self, identifier):
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
272 '''Remove a file from file index and word index
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
273 '''
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
274 self.load_index()
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
275
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
276 if not self.files.has_key(identifier):
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
277 return
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
278
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
279 file_index = self.files[identifier][0]
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
280 del self.files[identifier]
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
281 del self.fileids[file_index]
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
282
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
283 # The much harder part, cleanup the word index
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
284 for key, occurs in self.words.items():
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
285 if occurs.has_key(file_index):
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
286 del occurs[file_index]
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
287
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
288 # save needed
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
289 self.changed = 1
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
290
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
291 def index_loaded(self):
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
292 return (hasattr(self,'fileids') and hasattr(self,'files') and
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
293 hasattr(self,'words'))
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
294
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
295
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
296 def rollback(self):
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
297 ''' load last saved index info. '''
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
298 self.load_index(reload=1)
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
299
93f03c6714d8 A few big changes in this commit:
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
300 # vim: set filetype=python ts=4 sw=4 et si

Roundup Issue Tracker: http://roundup-tracker.org/