Package translate :: Package search :: Package indexing :: Module PyLuceneIndexer
[hide private]
[frames] | no frames]

Source Code for Module translate.search.indexing.PyLuceneIndexer

  1  # -*- coding: utf-8 -*- 
  2  # 
  3  # Copyright 2008 Zuza Software Foundation 
  4  # 
  5  # This file is part of translate. 
  6  # 
  7  # translate is free software; you can redistribute it and/or modify 
  8  # it under the terms of the GNU General Public License as published by 
  9  # the Free Software Foundation; either version 2 of the License, or 
 10  # (at your option) any later version. 
 11  # 
 12  # translate is distributed in the hope that it will be useful, 
 13  # but WITHOUT ANY WARRANTY; without even the implied warranty of 
 14  # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the 
 15  # GNU General Public License for more details. 
 16  # 
 17  # You should have received a copy of the GNU General Public License 
 18  # along with translate; if not, write to the Free Software 
 19  # Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA 
 20  # 
 21   
 22   
 23  """ 
 24  interface for the PyLucene (v2.x) indexing engine 
 25   
 26  take a look at PyLuceneIndexer1.py for the PyLucene v1.x interface 
 27  """ 
 28   
 29  __revision__ = "$Id: PyLuceneIndexer.py 14174 2010-04-08 23:14:10Z alaaosh $" 
 30   
 31  import CommonIndexer 
 32  import re 
 33  import os 
 34  import time 
 35  import logging 
 36   
 37  # try to import the PyLucene package (with the two possible names) 
 38  # remember the type of the detected package (compiled with jcc (>=v2.3) or 
 39  # with gcj (<=v2.2) 
 40  try: 
 41      import PyLucene 
 42      _COMPILER = 'gcj' 
 43  except ImportError: 
 44      # if this fails, then there is no pylucene installed 
 45      import lucene 
 46      PyLucene = lucene 
 47      PyLucene.initVM(PyLucene.CLASSPATH) 
 48      _COMPILER = 'jcc' 
 49   
 50   
 51  UNNAMED_FIELD_NAME = "FieldWithoutAName" 
 52  MAX_FIELD_SIZE = 1048576 
 53   
 54   
55 -def is_available():
56 return _get_pylucene_version() == 2
57 58
59 -class PyLuceneDatabase(CommonIndexer.CommonDatabase):
60 """manage and use a pylucene indexing database""" 61 62 QUERY_TYPE = PyLucene.Query 63 INDEX_DIRECTORY_NAME = "lucene" 64
65 - def __init__(self, basedir, analyzer=None, create_allowed=True):
66 """initialize or open an indexing database 67 68 Any derived class must override __init__. 69 70 @raise ValueError: the given location exists, but the database type 71 is incompatible (e.g. created by a different indexing engine) 72 @raise OSError: the database failed to initialize 73 74 @param basedir: the parent directory of the database 75 @type basedir: str 76 @param analyzer: bitwise combination of possible analyzer flags 77 to be used as the default analyzer for this database. Leave it empty 78 to use the system default analyzer (self.ANALYZER_DEFAULT). 79 see self.ANALYZER_TOKENIZE, self.ANALYZER_PARTIAL, ... 80 @type analyzer: int 81 @param create_allowed: create the database, if necessary; default: True 82 @type create_allowed: bool 83 """ 84 jvm = PyLucene.getVMEnv() 85 jvm.attachCurrentThread() 86 super(PyLuceneDatabase, self).__init__(basedir, analyzer=analyzer, 87 create_allowed=create_allowed) 88 self.pyl_analyzer = PyLucene.StandardAnalyzer() 89 self.writer = None 90 self.reader = None 91 self.index_version = None 92 try: 93 # try to open an existing database 94 tempreader = PyLucene.IndexReader.open(self.location) 95 tempreader.close() 96 except PyLucene.JavaError, err_msg: 97 # Write an error out, in case this is a real problem instead of an absence of an index 98 # TODO: turn the following two lines into debug output 99 #errorstr = str(e).strip() + "\n" + self.errorhandler.traceback_str() 100 #DEBUG_FOO("could not open index, so going to create: " + errorstr) 101 # Create the index, so we can open cached readers on it 102 if not create_allowed: 103 raise OSError("Indexer: skipping database creation") 104 try: 105 # create the parent directory if it does not exist 106 parent_path = os.path.dirname(self.location) 107 if not os.path.isdir(parent_path): 108 # recursively create all directories up to parent_path 109 os.makedirs(parent_path) 110 except IOError, err_msg: 111 raise OSError("Indexer: failed to create the parent " \ 112 + "directory (%s) of the indexing database: %s" \ 113 % (parent_path, err_msg)) 114 try: 115 tempwriter = PyLucene.IndexWriter(self.location, 116 self.pyl_analyzer, True) 117 tempwriter.close() 118 except PyLucene.JavaError, err_msg: 119 raise OSError("Indexer: failed to open or create a Lucene" \ 120 + " database (%s): %s" % (self.location, err_msg)) 121 # the indexer is initialized - now we prepare the searcher 122 # windows file locking seems inconsistent, so we try 10 times 123 numtries = 0 124 #self.dir_lock.acquire(blocking=True) 125 # read "self.reader", "self.indexVersion" and "self.searcher" 126 try: 127 while numtries < 10: 128 try: 129 self.reader = PyLucene.IndexReader.open(self.location) 130 self.indexVersion = self.reader.getCurrentVersion( 131 self.location) 132 self.searcher = PyLucene.IndexSearcher(self.reader) 133 break 134 except PyLucene.JavaError, e: 135 # store error message for possible later re-raise (below) 136 lock_error_msg = e 137 time.sleep(0.01) 138 numtries += 1 139 else: 140 # locking failed for 10 times 141 raise OSError("Indexer: failed to lock index database" \ 142 + " (%s)" % lock_error_msg) 143 finally: 144 pass 145 # self.dir_lock.release() 146 # initialize the searcher and the reader 147 self._index_refresh()
148
149 - def __del__(self):
150 """remove lock and close writer after loosing the last reference""" 151 self._writer_close()
152
153 - def flush(self, optimize=False):
154 """flush the content of the database - to force changes to be written 155 to disk 156 157 some databases also support index optimization 158 159 @param optimize: should the index be optimized if possible? 160 @type optimize: bool 161 """ 162 if self._writer_is_open(): 163 try: 164 if optimize: 165 self.writer.optimize() 166 finally: 167 # close the database even if optimizing failed 168 self._writer_close() 169 # the reader/searcher needs an update, too 170 self._index_refresh()
171
172 - def _create_query_for_query(self, query):
173 """generate a query based on an existing query object 174 175 basically this function should just create a copy of the original 176 177 @param query: the original query object 178 @type query: PyLucene.Query 179 @return: resulting query object 180 @rtype: PyLucene.Query 181 """ 182 # TODO: a deep copy or a clone would be safer 183 # somehow not working (returns "null"): copy.deepcopy(query) 184 return query
185
186 - def _create_query_for_string(self, text, require_all=True, 187 analyzer=None):
188 """generate a query for a plain term of a string query 189 190 basically this function parses the string and returns the resulting 191 query 192 193 @param text: the query string 194 @type text: str 195 @param require_all: boolean operator 196 (True -> AND (default) / False -> OR) 197 @type require_all: bool 198 @param analyzer: the analyzer to be used 199 possible analyzers are: 200 - L{CommonDatabase.ANALYZER_TOKENIZE} 201 the field value is splitted to be matched word-wise 202 - L{CommonDatabase.ANALYZER_PARTIAL} 203 the field value must start with the query string 204 - L{CommonDatabase.ANALYZER_EXACT} 205 keep special characters and the like 206 @type analyzer: bool 207 @return: resulting query object 208 @rtype: PyLucene.Query 209 """ 210 if analyzer is None: 211 analyzer = self.analyzer 212 if analyzer == self.ANALYZER_EXACT: 213 analyzer_obj = PyLucene.KeywordAnalyzer() 214 else: 215 text = _escape_term_value(text) 216 analyzer_obj = PyLucene.StandardAnalyzer() 217 qp = PyLucene.QueryParser(UNNAMED_FIELD_NAME, analyzer_obj) 218 if (analyzer & self.ANALYZER_PARTIAL > 0): 219 # PyLucene uses explicit wildcards for partial matching 220 text += "*" 221 if require_all: 222 qp.setDefaultOperator(qp.Operator.AND) 223 else: 224 qp.setDefaultOperator(qp.Operator.OR) 225 return qp.parse(text)
226
227 - def _create_query_for_field(self, field, value, analyzer=None):
228 """generate a field query 229 230 this functions creates a field->value query 231 232 @param field: the fieldname to be used 233 @type field: str 234 @param value: the wanted value of the field 235 @type value: str 236 @param analyzer: the analyzer to be used 237 possible analyzers are: 238 - L{CommonDatabase.ANALYZER_TOKENIZE} 239 the field value is splitted to be matched word-wise 240 - L{CommonDatabase.ANALYZER_PARTIAL} 241 the field value must start with the query string 242 - L{CommonDatabase.ANALYZER_EXACT} 243 keep special characters and the like 244 @type analyzer: bool 245 @return: resulting query object 246 @rtype: PyLucene.Query 247 """ 248 if analyzer is None: 249 analyzer = self.analyzer 250 if analyzer == self.ANALYZER_EXACT: 251 analyzer_obj = PyLucene.KeywordAnalyzer() 252 else: 253 value = _escape_term_value(value) 254 analyzer_obj = PyLucene.StandardAnalyzer() 255 qp = PyLucene.QueryParser(field, analyzer_obj) 256 if (analyzer & self.ANALYZER_PARTIAL > 0): 257 # PyLucene uses explicit wildcards for partial matching 258 value += "*" 259 return qp.parse(value)
260
261 - def _create_query_combined(self, queries, require_all=True):
262 """generate a combined query 263 264 @param queries: list of the original queries 265 @type queries: list of PyLucene.Query 266 @param require_all: boolean operator 267 (True -> AND (default) / False -> OR) 268 @type require_all: bool 269 @return: the resulting combined query object 270 @rtype: PyLucene.Query 271 """ 272 combined_query = PyLucene.BooleanQuery() 273 for query in queries: 274 combined_query.add( 275 PyLucene.BooleanClause(query, _occur(require_all, False))) 276 return combined_query
277
278 - def _create_empty_document(self):
279 """create an empty document to be filled and added to the index later 280 281 @return: the new document object 282 @rtype: PyLucene.Document 283 """ 284 return PyLucene.Document()
285
286 - def _add_plain_term(self, document, term, tokenize=True):
287 """add a term to a document 288 289 @param document: the document to be changed 290 @type document: PyLucene.Document 291 @param term: a single term to be added 292 @type term: str 293 @param tokenize: should the term be tokenized automatically 294 @type tokenize: bool 295 """ 296 if tokenize: 297 token_flag = PyLucene.Field.Index.TOKENIZED 298 else: 299 token_flag = PyLucene.Field.Index.UN_TOKENIZED 300 document.add(PyLucene.Field(str(UNNAMED_FIELD_NAME), term, 301 PyLucene.Field.Store.YES, token_flag))
302
303 - def _add_field_term(self, document, field, term, tokenize=True):
304 """add a field term to a document 305 306 @param document: the document to be changed 307 @type document: PyLucene.Document 308 @param field: name of the field 309 @type field: str 310 @param term: term to be associated to the field 311 @type term: str 312 @param tokenize: should the term be tokenized automatically 313 @type tokenize: bool 314 """ 315 if tokenize: 316 token_flag = PyLucene.Field.Index.TOKENIZED 317 else: 318 token_flag = PyLucene.Field.Index.UN_TOKENIZED 319 document.add(PyLucene.Field(str(field), term, 320 PyLucene.Field.Store.YES, token_flag))
321
322 - def _add_document_to_index(self, document):
323 """add a prepared document to the index database 324 325 @param document: the document to be added 326 @type document: PyLucene.Document 327 """ 328 self._writer_open() 329 self.writer.addDocument(document)
330
331 - def begin_transaction(self):
332 """PyLucene does not support transactions 333 334 Thus this function just opens the database for write access. 335 Call "cancel_transaction" or "commit_transaction" to close write 336 access in order to remove the exclusive lock from the database 337 directory. 338 """ 339 self._writer_open()
340
341 - def cancel_transaction(self):
342 """PyLucene does not support transactions 343 344 Thus this function just closes the database write access and removes 345 the exclusive lock. 346 347 See 'start_transaction' for details. 348 """ 349 self._writer_close()
350
351 - def commit_transaction(self):
352 """PyLucene does not support transactions 353 354 Thus this function just closes the database write access and removes 355 the exclusive lock. 356 357 See 'start_transaction' for details. 358 """ 359 self._writer_close() 360 self._index_refresh()
361
362 - def get_query_result(self, query):
363 """return an object containing the results of a query 364 365 @param query: a pre-compiled query 366 @type query: a query object of the real implementation 367 @return: an object that allows access to the results 368 @rtype: subclass of CommonEnquire 369 """ 370 return PyLuceneHits(self.searcher.search(query))
371
372 - def delete_document_by_id(self, docid):
373 """delete a specified document 374 375 @param docid: the document ID to be deleted 376 @type docid: int 377 """ 378 self._delete_stale_lock() 379 self.reader.deleteDocument(docid) 380 self.reader.flush() 381 # TODO: check the performance impact of calling "refresh" for each id 382 self._index_refresh()
383
384 - def search(self, query, fieldnames):
385 """return a list of the contents of specified fields for all matches of 386 a query 387 388 @param query: the query to be issued 389 @type query: a query object of the real implementation 390 @param fieldnames: the name(s) of a field of the document content 391 @type fieldnames: string | list of strings 392 @return: a list of dicts containing the specified field(s) 393 @rtype: list of dicts 394 """ 395 if isinstance(fieldnames, basestring): 396 fieldnames = [fieldnames] 397 hits = self.searcher.search(query) 398 if _COMPILER == 'jcc': 399 # add the ranking number and the retrieved document to the array 400 hits = [(hit, hits.doc(hit)) for hit in range(hits.length())] 401 result = [] 402 for hit, doc in hits: 403 fields = {} 404 for fieldname in fieldnames: 405 # take care for the special field "None" 406 if fieldname is None: 407 pyl_fieldname = UNNAMED_FIELD_NAME 408 else: 409 pyl_fieldname = fieldname 410 fields[fieldname] = doc.getValues(pyl_fieldname) 411 result.append(fields) 412 return result
413
414 - def _delete_stale_lock(self):
415 if self.reader.isLocked(self.location): 416 #HACKISH: there is a lock but Lucene api can't tell us how old it 417 # is, will have to check the filesystem 418 try: 419 # in try block just in case lock disappears on us while testing it 420 stat = os.stat(os.path.join(self.location, 'write.lock')) 421 age = (time.time() - stat.st_mtime) / 60 422 if age > 15: 423 logging.warning("stale lock found in %s, removing.", self.location) 424 self.reader.unlock(self.reader.directory()) 425 except: 426 pass
427
428 - def _writer_open(self):
429 """open write access for the indexing database and acquire an 430 exclusive lock 431 """ 432 if not self._writer_is_open(): 433 self._delete_stale_lock() 434 self.writer = PyLucene.IndexWriter(self.location, self.pyl_analyzer, 435 False) 436 # "setMaxFieldLength" is available since PyLucene v2 437 # we must stay compatible to v1 for the derived class 438 # (PyLuceneIndexer1) - thus we make this step optional 439 if hasattr(self.writer, "setMaxFieldLength"): 440 self.writer.setMaxFieldLength(MAX_FIELD_SIZE)
441 # do nothing, if it is already open 442
443 - def _writer_close(self):
444 """close indexing write access and remove the database lock""" 445 if self._writer_is_open(): 446 self.writer.commit() 447 self.writer.close() 448 self.writer = None
449
450 - def _writer_is_open(self):
451 """check if the indexing write access is currently open""" 452 return not self.writer is None
453
454 - def _index_refresh(self):
455 """re-read the indexer database""" 456 try: 457 if self.reader is None or self.searcher is None: 458 self.reader = PyLucene.IndexReader.open(self.location) 459 self.searcher = PyLucene.IndexSearcher(self.reader) 460 elif self.index_version != self.reader.getCurrentVersion( \ 461 self.location): 462 self.searcher.close() 463 self.reader.close() 464 self.reader = PyLucene.IndexReader.open(self.location) 465 self.searcher = PyLucene.IndexSearcher(self.reader) 466 self.index_version = self.reader.getCurrentVersion(self.location) 467 except PyLucene.JavaError, e: 468 # TODO: add some debugging output? 469 #self.errorhandler.logerror("Error attempting to read index - try reindexing: "+str(e)) 470 pass
471 472
473 -class PyLuceneHits(CommonIndexer.CommonEnquire):
474 """an enquire object contains the information about the result of a request 475 """ 476
477 - def get_matches(self, start, number):
478 """return a specified number of qualified matches of a previous query 479 480 @param start: index of the first match to return (starting from zero) 481 @type start: int 482 @param number: the number of matching entries to return 483 @type number: int 484 @return: a set of matching entries and some statistics 485 @rtype: tuple of (returned number, available number, matches) 486 "matches" is a dictionary of:: 487 ["rank", "percent", "document", "docid"] 488 """ 489 # check if requested results do not exist 490 # stop is the lowest index number to be ommitted 491 stop = start + number 492 if stop > self.enquire.length(): 493 stop = self.enquire.length() 494 # invalid request range 495 if stop <= start: 496 return (0, self.enquire.length(), []) 497 result = [] 498 for index in range(start, stop): 499 item = {} 500 item["rank"] = index 501 item["docid"] = self.enquire.id(index) 502 item["percent"] = self.enquire.score(index) 503 item["document"] = self.enquire.doc(index) 504 result.append(item) 505 return (stop-start, self.enquire.length(), result)
506
507 -def _occur(required, prohibited):
508 if required == True and prohibited == False: 509 return PyLucene.BooleanClause.Occur.MUST 510 elif required == False and prohibited == False: 511 return PyLucene.BooleanClause.Occur.SHOULD 512 elif required == False and prohibited == True: 513 return PyLucene.BooleanClause.Occur.MUST_NOT 514 else: 515 # It is an error to specify a clause as both required 516 # and prohibited 517 return None
518
519 -def _get_pylucene_version():
520 """get the installed pylucene version 521 522 @return: 1 -> PyLucene v1.x / 2 -> PyLucene v2.x / 0 -> unknown 523 @rtype: int 524 """ 525 version = PyLucene.VERSION 526 if version.startswith("1."): 527 return 1 528 elif version.startswith("2."): 529 return 2 530 else: 531 return 0
532 533
534 -def _escape_term_value(text):
535 return re.sub("\*", "", text)
536