| Home | Trees | Indices | Help |
|
|---|
|
|
1 """Base classes for match providers.
2
3 They are used by business objects to give
4 phrasewheels the ability to guess phrases.
5
6 Copyright (C) GNUMed developers
7 license: GPL v2 or later
8 """
9 __version__ = "$Revision: 1.34 $"
10 __author__ = "K.Hilbert <Karsten.Hilbert@gmx.net>, I.Haywood <ihaywood@gnu.org>, S.J.Tan <sjtan@bigpond.com>"
11
12 # std lib
13 import re as regex, logging
14
15
16 # GNUmed
17 from Gnumed.pycommon import gmPG2
18
19
20 _log = logging.getLogger('gm.ui')
21 _log.info(__version__)
22
23
24 # these are stripped from the fragment passed to the
25 # match provider before looking for matches:
26 default_ignored_chars = "[?!.'\\(){}\[\]<>~#*$%^_]+" + '"'
27
28 # these are used to detect word boundaries which is,
29 # in turn, used to normalize word boundaries in the
30 # input fragment
31 default_word_separators = '[- \t=+&:@]+'
32 #============================================================
34 """Base class for match providing objects.
35
36 Match sources might be:
37 - database tables
38 - flat files
39 - previous input
40 - config files
41 - in-memory list created on the fly
42 """
43 print_queries = False
44 #--------------------------------------------------------
46 self.setThresholds()
47
48 self._context_vals = {}
49 self.__ignored_chars = regex.compile(default_ignored_chars)
50 # used to normalize word boundaries:
51 self.__word_separators = regex.compile(default_word_separators)
52 #--------------------------------------------------------
53 # actions
54 #--------------------------------------------------------
56 """Return matches according to aFragment and matching thresholds.
57
58 FIXME: design decision: we dont worry about data source changes
59 during the lifetime of a MatchProvider
60 FIXME: append _("*get all items*") on truncation
61 """
62 # sanity check
63 if aFragment is None:
64 raise ValueError, 'Cannot find matches without a fragment.'
65
66 # user explicitly wants all matches
67 if aFragment == u'*':
68 return self.getAllMatches()
69
70 # case insensitivity
71 tmpFragment = aFragment.lower()
72 # remove ignored chars
73 if self.__ignored_chars is not None:
74 tmpFragment = self.__ignored_chars.sub('', tmpFragment)
75 # normalize word separators
76 if self.__word_separators is not None:
77 tmpFragment = u' '.join(self.__word_separators.split(tmpFragment))
78 # length in number of significant characters only
79 lngFragment = len(tmpFragment)
80
81 # order is important !
82 if lngFragment >= self.__threshold_substring:
83 return self.getMatchesBySubstr(tmpFragment)
84 elif lngFragment >= self.__threshold_word:
85 return self.getMatchesByWord(tmpFragment)
86 elif lngFragment >= self.__threshold_phrase:
87 return self.getMatchesByPhrase(tmpFragment)
88 else:
89 return (False, [])
90 #--------------------------------------------------------
93 #--------------------------------------------------------
96 #--------------------------------------------------------
99 #--------------------------------------------------------
102 #--------------------------------------------------------
105 #--------------------------------------------------------
106 # configuration
107 #--------------------------------------------------------
109 """Set match location thresholds.
110
111 - the fragment passed to getMatches() must contain at least this many
112 characters before it triggers a match search at:
113 1) phrase_start - start of phrase (first word)
114 2) word_start - start of any word within phrase
115 3) in_word - _inside_ any word within phrase
116 """
117 # sanity checks
118 if aSubstring < aWord:
119 _log.error('Setting substring threshold (%s) lower than word-start threshold (%s) does not make sense. Retaining original thresholds (%s:%s, respectively).' % (aSubstring, aWord, self.__threshold_substring, self.__threshold_word))
120 return False
121 if aWord < aPhrase:
122 _log.error('Setting word-start threshold (%s) lower than phrase-start threshold (%s) does not make sense. Retaining original thresholds (%s:%s, respectively).' % (aSubstring, aWord, self.__threshold_word, self.__threshold_phrase))
123 return False
124
125 # now actually reassign thresholds
126 self.__threshold_phrase = aPhrase
127 self.__threshold_word = aWord
128 self.__threshold_substring = aSubstring
129
130 return True
131 #--------------------------------------------------------
133 if word_separators is None:
134 self.__word_separators = None
135 else:
136 self.__word_separators = regex.compile(word_separators)
137
142
143 word_separators = property(_get_word_separators, _set_word_separators)
144 #--------------------------------------------------------
146 if ignored_chars is None:
147 self.__ignored_chars = None
148 else:
149 self.__ignored_chars = regex.compile(ignored_chars)
150
155
156 ignored_chars = property(_get_ignored_chars, _set_ignored_chars)
157 #--------------------------------------------------------
159 """Set value to provide context information for matches.
160
161 The matching code may ignore it depending on its exact
162 implementation. Names and values of the context depend
163 on what is being matched.
164
165 <context> -- the *placeholder* key *inside* the context
166 definition, not the context *definition* key
167 """
168 if context is None:
169 return False
170 self._context_vals[context] = val
171 return True
172 #--------------------------------------------------------
178 #------------------------------------------------------------
179 # usable instances
180 #------------------------------------------------------------
182 """Match provider where all possible options can be held
183 in a reasonably sized, pre-allocated list.
184 """
186 """aSeq must be a list of dicts. Each dict must have the keys (data, label, weight)
187 """
188 if not type(aSeq) in [type(None), type([]), type(())]:
189 _log.error('fixed list match provider argument must be a list/tuple of dicts/None')
190 raise TypeError('fixed list match provider argument must be a list/tuple of dicts/None')
191
192 self.__items = aSeq
193 cMatchProvider.__init__(self)
194 #--------------------------------------------------------
195 # internal matching algorithms
196 #
197 # if we end up here:
198 # - aFragment will not be "None"
199 # - aFragment will be lower case
200 # - we _do_ deliver matches (whether we find any is a different story)
201 #--------------------------------------------------------
203 """Return matches for aFragment at start of phrases."""
204 matches = []
205 # look for matches
206 for item in self.__items:
207 # at start of phrase, that is
208 if item['list_label'].lower().startswith(aFragment.lower()):
209 matches.append(item)
210 # no matches found
211 if len(matches) == 0:
212 return (False, [])
213
214 matches.sort(self.__cmp_items)
215 return (True, matches)
216 #--------------------------------------------------------
218 """Return matches for aFragment at start of words inside phrases."""
219 matches = []
220 # look for matches
221 for item in self.__items:
222 item_label = item['list_label'].lower()
223 fragment_pos = item_label.find(aFragment.lower())
224 # found at start of phrase
225 if fragment_pos == 0:
226 matches.append(item)
227 # found as a true substring
228 elif fragment_pos > 0:
229 # but use only if substring is at start of a word
230 if item_label[fragment_pos-1] == u' ':
231 matches.append(item)
232 # no matches found
233 if len(matches) == 0:
234 return (False, [])
235
236 matches.sort(self.__cmp_items)
237 return (True, matches)
238 #--------------------------------------------------------
240 """Return matches for aFragment as a true substring."""
241 matches = []
242 # look for matches
243 for item in self.__items:
244 if item['list_label'].lower().find(aFragment.lower()) != -1:
245 matches.append(item)
246 # no matches found
247 if len(matches) == 0:
248 return (False, [])
249
250 matches.sort(self.__cmp_items)
251 return (True, matches)
252 #--------------------------------------------------------
254 """Return all items."""
255 matches = self.__items
256 # no matches found
257 if len(matches) == 0:
258 return (False, [])
259
260 matches.sort(self.__cmp_items)
261 return (True, matches)
262 #--------------------------------------------------------
264 """items must be a list of dicts. Each dict must have the keys (data, list_label, weight)"""
265 self.__items = items
266 #--------------------------------------------------------
277 # ===========================================================
279 """Match provider which searches matches
280 in the results of a function call.
281 """
283 """get_candidates() must return a list of strings."""
284 if get_candidates is None:
285 _log.error('must define function to retrieve match candidates list')
286 raise ValueError('must define function to retrieve match candidates list')
287
288 self._get_candidates = get_candidates
289 cMatchProvider.__init__(self)
290 #--------------------------------------------------------
291 # internal matching algorithms
292 #
293 # if we end up here:
294 # - aFragment will not be "None"
295 # - aFragment will be lower case
296 # - we _do_ deliver matches (whether we find any is a different story)
297 #--------------------------------------------------------
299 """Return matches for aFragment at start of phrases."""
300 matches = []
301 candidates = self._get_candidates()
302 # look for matches
303 for candidate in candidates:
304 # at start of phrase, that is
305 if aFragment.startswith(candidate['list_label'].lower()):
306 matches.append(candidate)
307 # no matches found
308 if len(matches) == 0:
309 return (False, [])
310
311 matches.sort(self.__cmp_candidates)
312 return (True, matches)
313 #--------------------------------------------------------
315 """Return matches for aFragment at start of words inside phrases."""
316 matches = []
317 candidates = self._get_candidates()
318 # look for matches
319 for candidate in candidates:
320 pos = candidate['list_label'].lower().find(aFragment)
321 # pos = string.find(string.lower(candidate['list_label']), aFragment)
322 # found as a true substring
323 # but use only if substring is at start of a word
324 # FIXME: use word seps
325 if (pos == 0) or (candidate['list_label'][pos-1] == u' '):
326 matches.append(candidate)
327 # no matches found
328 if len(matches) == 0:
329 return (False, [])
330
331 matches.sort(self.__cmp_candidates)
332 return (True, matches)
333 #--------------------------------------------------------
335 """Return matches for aFragment as a true substring."""
336 matches = []
337 candidates = self._get_candidates()
338 # look for matches
339 for candidate in candidates:
340 if candidate['list_label'].lower().find(aFragment) != -1:
341 # if string.find(string.lower(candidate['list_label']), aFragment) != -1:
342 matches.append(candidate)
343 # no matches found
344 if len(matches) == 0:
345 return (False, [])
346
347 matches.sort(self.__cmp_candidates)
348 return (True, matches)
349 #--------------------------------------------------------
353 #--------------------------------------------------------
357 # FIXME: do ordering
358 # if candidate1 < candidate2:
359 # return -1
360 # if candidate1 == candidate2:
361 # return 0
362 # return 1
363
364 # ===========================================================
366 """Match provider which searches matches
367 in possibly several database tables.
368
369 queries:
370 - a list of unicode strings
371 - each string is a query
372 - each string must contain: "... WHERE <column> %(fragment_condition)s ..."
373 - each string can contain in the where clause: "... %(<ctxt_key1>)s ..."
374 - each query must return (data, list_label, field_label)
375
376 context definitions to be used in the queries, example:
377 {'ctxt_key1': {'where_part': 'AND country = %(country)s', 'placeholder': 'country'}}
378
379 client code using .set_context() must use the 'placeholder':
380 <phrasewheel>/<match provider>.set_context('country', 'Germany')
381
382 full example query:
383
384 query = u" " "
385 SELECT DISTINCT ON (list_label)
386 pk_encounter
387 AS data,
388 to_char(started, 'YYYY Mon DD (HH24:MI)') || ': ' || l10n_type || ' [#' || pk_encounter || ']'
389 AS list_label,
390 to_char(started, 'YYYY Mon DD') || ': ' || l10n_type
391 AS field_label
392 FROM
393 clin.v_pat_encounters
394 WHERE
395 (
396 l10n_type %(fragment_condition)s
397 OR
398 type %(fragment_condition)s
399 ) %(ctxt_patient)s
400 ORDER BY
401 list_label
402 LIMIT
403 30
404 " " "
405 context = {'ctxt_patient': {
406 'where_part': u'AND pk_patient = %(PLACEHOLDER)s',
407 'placeholder': u'PLACEHOLDER'
408 }}
409 self.mp = gmMatchProvider.cMatchProvider_SQL2(queries = query, context = context)
410 self.set_context(context = 'PLACEHOLDER', val = '<THE VALUE>')
411
412 _SQL_data2match:
413 SQL to retrieve a match by, say, primary key
414 wherein the only keyword argument is 'pk'
415 """
417
418 cMatchProvider.__init__(self)
419
420 if type(queries) == type([]):
421 self._queries = queries
422 else:
423 self._queries = [queries]
424
425 if context is None:
426 self._context = {}
427 else:
428 self._context = context
429
430 self._args = {}
431
432 self._SQL_data2match = None
433 #--------------------------------------------------------
434 # internal matching algorithms
435 #
436 # if we end up here:
437 # - aFragment will not be "None"
438 # - aFragment will be lower case
439 # - we _do_ deliver matches (whether we find any is a different story)
440 #--------------------------------------------------------
442 """Return matches for aFragment at start of phrases."""
443
444 fragment_condition = u"ILIKE %(fragment)s"
445 self._args['fragment'] = u"%s%%" % aFragment
446
447 return self._find_matches(fragment_condition)
448 #--------------------------------------------------------
450 """Return matches for aFragment at start of words inside phrases."""
451
452 fragment_condition = u"~* %(fragment)s"
453 aFragment = gmPG2.sanitize_pg_regex(expression = aFragment, escape_all = False)
454 self._args['fragment'] = u"( %s)|(^%s)" % (aFragment, aFragment)
455
456 return self._find_matches(fragment_condition)
457 #--------------------------------------------------------
459 """Return matches for aFragment as a true substring."""
460
461 fragment_condition = u"ILIKE %(fragment)s"
462 self._args['fragment'] = u"%%%s%%" % aFragment
463
464 return self._find_matches(fragment_condition)
465 #--------------------------------------------------------
469 #--------------------------------------------------------
471 if self._SQL_data2match is None:
472 return None
473
474 query = {'cmd': self._SQL_data2match, 'args': {'pk': data}}
475 try:
476 rows, idx = gmPG2.run_ro_queries(queries = [query], get_col_idx = False)
477 except:
478 _log.exception('[%s]: error running _SQL_data2match, dropping query', self.__class__.__name__)
479 self._SQL_data2match = None
480 return None
481
482 # hopefully the most frequent case:
483 if len(rows) == 1:
484 return rows[0]
485
486 _log.error('[%s]: 0 or >1 rows found by running _SQL_data2match, ambiguous, ignoring', self.__class__.__name__)
487 return None
488 #--------------------------------------------------------
490 if self.print_queries:
491 print "----------------------"
492 matches = []
493 for query in self._queries:
494 where_fragments = {'fragment_condition': fragment_condition}
495
496 for context_key, context_def in self._context.items():
497 try:
498 placeholder = context_def['placeholder']
499 where_part = context_def['where_part']
500 self._args[placeholder] = self._context_vals[placeholder]
501 # we do have a context value for this key, so add the where condition
502 where_fragments[context_key] = where_part
503 if self.print_queries:
504 print "ctxt ph:", placeholder
505 print "ctxt where:", where_part
506 print "ctxt val:", self._context_vals[placeholder]
507 except KeyError:
508 # we don't have a context value for this key, so skip the where condition
509 where_fragments[context_key] = u''
510 if self.print_queries:
511 print "invalid ctxt key:", context_key
512
513 cmd = query % where_fragments
514
515 if self.print_queries:
516 print "class:", self.__class__.__name__
517 print "ctxt:", self._context_vals
518 print "args:", self._args
519 print "query:", cmd
520
521 try:
522 rows, idx = gmPG2.run_ro_queries(queries = [{'cmd': cmd, 'args': self._args}], get_col_idx = False)
523 except:
524 _log.exception('[%s]: error running match provider SQL, dropping query', self.__class__.__name__)
525 idx = self._queries.index(query)
526 del self._queries[idx]
527 break
528
529 # no matches found: try next query
530 if len(rows) == 0:
531 continue
532
533 for row in rows:
534 match = {'weight': 0}
535
536 try:
537 match['data'] = row['data']
538 except KeyError:
539 match['data'] = row[0]
540
541 try:
542 match['list_label'] = row['list_label']
543 except KeyError:
544 match['list_label'] = row[1]
545
546 # explicit "field_label" in result ?
547 try:
548 match['field_label'] = row['field_label']
549 # no
550 except KeyError:
551 # but does row[2] exist ?
552 try:
553 match['field_label'] = row[2]
554 # no: reuse "list_label"
555 except IndexError:
556 match['field_label'] = match['list_label']
557
558 # try:
559 # match['label'] = row['label']
560 # except KeyError:
561 # match['label'] = match['list_label']
562
563 matches.append(match)
564
565 return (True, matches)
566
567 # none found whatsoever
568 return (False, [])
569 #================================================================
570 if __name__ == '__main__':
571 pass
572
573 #================================================================
574
| Home | Trees | Indices | Help |
|
|---|
| Generated by Epydoc 3.0.1 on Sat Oct 5 03:57:14 2013 | http://epydoc.sourceforge.net |