X-Git-Url: https://git.mdrn.pl/wolnelektury.git/blobdiff_plain/919a5d574aac85d55b6cfd17f0a81dd5ff56265a..281a6f613c78ffd1fa81951db829a5551fd2e42d:/apps/search/index.py diff --git a/apps/search/index.py b/apps/search/index.py index a3a62de18..da54fe7f0 100644 --- a/apps/search/index.py +++ b/apps/search/index.py @@ -1,100 +1,38 @@ # -*- coding: utf-8 -*- from django.conf import settings -from lucene import SimpleFSDirectory, IndexWriter, CheckIndex, \ - File, Field, Integer, \ - NumericField, Version, Document, JavaError, IndexSearcher, \ - QueryParser, PerFieldAnalyzerWrapper, \ - SimpleAnalyzer, PolishAnalyzer, ArrayList, \ - KeywordAnalyzer, NumericRangeQuery, BooleanQuery, \ - BlockJoinQuery, BlockJoinCollector, Filter, TermsFilter, ChainedFilter, \ - HashSet, BooleanClause, Term, CharTermAttribute, \ - PhraseQuery, MultiPhraseQuery, StringReader, TermQuery, \ - FuzzyQuery, FuzzyTermEnum, PrefixTermEnum, Sort, Integer, \ - SimpleHTMLFormatter, Highlighter, QueryScorer, TokenSources, TextFragment, \ - BooleanFilter, TermsFilter, FilterClause, QueryWrapperFilter, \ - initVM, CLASSPATH, JArray, JavaError - # KeywordAnalyzer - -# Initialize jvm -JVM = initVM(CLASSPATH) - -import sys + import os import re import errno from librarian import dcparser from librarian.parser import WLDocument +from lxml import etree import catalogue.models -from multiprocessing.pool import ThreadPool -from threading import current_thread -import atexit +from pdcounter.models import Author as PDCounterAuthor, BookStub as PDCounterBook +from itertools import chain import traceback +import logging +log = logging.getLogger('search') +import sunburnt +import custom +import operator -class WLAnalyzer(PerFieldAnalyzerWrapper): - def __init__(self): - polish = PolishAnalyzer(Version.LUCENE_34) - # polish_gap.setPositionIncrementGap(999) - - simple = SimpleAnalyzer(Version.LUCENE_34) - # simple_gap.setPositionIncrementGap(999) - - keyword = KeywordAnalyzer(Version.LUCENE_34) - - # not sure if needed: there's NOT_ANALYZED meaning basically the same - - PerFieldAnalyzerWrapper.__init__(self, polish) - - self.addAnalyzer("tags", simple) - self.addAnalyzer("technical_editors", simple) - self.addAnalyzer("editors", simple) - self.addAnalyzer("url", keyword) - self.addAnalyzer("source_url", keyword) - self.addAnalyzer("source_name", simple) - self.addAnalyzer("publisher", simple) - self.addAnalyzer("author", simple) - self.addAnalyzer("is_book", keyword) - - self.addAnalyzer("themes", simple) - self.addAnalyzer("themes_pl", polish) - - self.addAnalyzer("tag_name", simple) - self.addAnalyzer("tag_name_pl", polish) - - self.addAnalyzer("KEYWORD", keyword) - self.addAnalyzer("SIMPLE", simple) - self.addAnalyzer("POLISH", polish) - - -class IndexStore(object): - def __init__(self): - self.make_index_dir() - self.store = SimpleFSDirectory(File(settings.SEARCH_INDEX)) - - def make_index_dir(self): - try: - os.makedirs(settings.SEARCH_INDEX) - except OSError as exc: - if exc.errno == errno.EEXIST: - pass - else: raise - - -class IndexChecker(IndexStore): - def __init__(self): - IndexStore.__init__(self) - - def check(self): - checker = CheckIndex(self.store) - status = checker.checkIndex() - return status +class SolrIndex(object): + def __init__(self, mode=None): + self.index = custom.CustomSolrInterface(settings.SOLR, mode=mode) class Snippets(object): + """ + This class manages snippet files for indexed object (book) + the snippets are concatenated together, and their positions and + lengths are kept in lucene index fields. + """ SNIPPET_DIR = "snippets" - def __init__(self, book_id): + def __init__(self, book_id, revision=None): try: os.makedirs(os.path.join(settings.SEARCH_INDEX, self.SNIPPET_DIR)) except OSError as exc: @@ -102,94 +40,228 @@ class Snippets(object): pass else: raise self.book_id = book_id + self.revision = revision self.file = None + @property + def path(self): + if self.revision: fn = "%d.%d" % (self.book_id, self.revision) + else: fn = "%d" % self.book_id + + return os.path.join(settings.SEARCH_INDEX, self.SNIPPET_DIR, fn) + def open(self, mode='r'): + """ + Open the snippet file. Call .close() afterwards. + """ if not 'b' in mode: mode += 'b' - self.file = open(os.path.join(settings.SEARCH_INDEX, self.SNIPPET_DIR, str(self.book_id)), mode) + + if 'w' in mode: + if os.path.exists(self.path): + self.revision = 1 + while True: + if not os.path.exists(self.path): + break + self.revision += 1 + + self.file = open(self.path, mode) self.position = 0 return self def add(self, snippet): + """ + Append a snippet (unicode) to the snippet file. + Return a (position, length) tuple + """ txt = snippet.encode('utf-8') l = len(txt) self.file.write(txt) pos = (self.position, l) self.position += l - print "Snip<%s>%s" %(pos, txt) return pos def get(self, pos): + """ + Given a tuple of (position, length) return an unicode + of the snippet stored there. + """ self.file.seek(pos[0], 0) txt = self.file.read(pos[1]).decode('utf-8') - print "got from snippets %d bytes from %s:" % (len(txt), pos) return txt def close(self): + """Close snippet file""" self.file.close() + def remove(self): + self.revision = None + try: + os.unlink(self.path) + self.revision = 0 + while True: + self.revision += 1 + os.unlink(self.path) + except OSError: + pass -class Index(IndexStore): - def __init__(self, analyzer=None): - IndexStore.__init__(self) - self.index = None - if not analyzer: - analyzer = WLAnalyzer() - self.analyzer = analyzer - def open(self, analyzer=None): - if self.index: - raise Exception("Index is already opened") - self.index = IndexWriter(self.store, self.analyzer,\ - IndexWriter.MaxFieldLength.LIMITED) - return self.index +class Index(SolrIndex): + """ + Class indexing books. + """ + def __init__(self): + super(Index, self).__init__(mode='rw') - def optimize(self): - self.index.optimize() + def delete_query(self, *queries): + """ + index.delete(queries=...) doesn't work, so let's reimplement it + using deletion of list of uids. + """ + uids = set() + for q in queries: + if isinstance(q, sunburnt.search.LuceneQuery): + q = self.index.query(q) + q.field_limiter.update(['uid']) + st = 0 + rows = 100 + while True: + ids = q.paginate(start=st, rows=rows).execute() + if not len(ids): + break + for res in ids: + uids.add(res['uid']) + st += rows + # print "Will delete %s" % ','.join([x for x in uids]) + if uids: + self.index.delete(uids) + return True + else: + return False - def close(self): - try: - self.index.optimize() - except JavaError, je: - print "Error during optimize phase, check index: %s" % je - - self.index.close() - self.index = None - - def index_tags(self): - q = NumericRangeQuery.newIntRange("tag_id", 0, Integer.MAX_VALUE, True, True) - self.index.deleteDocuments(q) - - for tag in catalogue.models.Tag.objects.all(): - doc = Document() - doc.add(NumericField("tag_id", Field.Store.YES, True).setIntValue(tag.id)) - doc.add(Field("tag_name", tag.name, Field.Store.NO, Field.Index.ANALYZED)) - doc.add(Field("tag_name_pl", tag.name, Field.Store.NO, Field.Index.ANALYZED)) - doc.add(Field("tag_category", tag.category, Field.Store.NO, Field.Index.NOT_ANALYZED)) - self.index.addDocument(doc) - - def remove_book(self, book): - q = NumericRangeQuery.newIntRange("book_id", book.id, book.id, True, True) - self.index.deleteDocuments(q) - - def index_book(self, book, overwrite=True): + def index_tags(self, *tags, **kw): + """ + Re-index global tag list. + Removes all tags from index, then index them again. + Indexed fields include: id, name (with and without polish stems), category + """ + remove_only = kw.get('remove_only', False) + # first, remove tags from index. + if tags: + tag_qs = [] + for tag in tags: + q_id = self.index.Q(tag_id=tag.id) + + if isinstance(tag, PDCounterAuthor): + q_cat = self.index.Q(tag_category='pd_author') + elif isinstance(tag, PDCounterBook): + q_cat = self.index.Q(tag_category='pd_book') + else: + q_cat = self.index.Q(tag_category=tag.category) + + q_id_cat = self.index.Q(q_id & q_cat) + tag_qs.append(q_id_cat) + self.delete_query(tag_qs) + else: # all + q = self.index.Q(tag_id__any=True) + self.delete_query(q) + + if not remove_only: + # then add them [all or just one passed] + if not tags: + tags = chain(catalogue.models.Tag.objects.exclude(category='set'), \ + PDCounterAuthor.objects.all(), \ + PDCounterBook.objects.all()) + + for tag in tags: + if isinstance(tag, PDCounterAuthor): + doc = { + "tag_id": int(tag.id), + "tag_name": tag.name, + "tag_name_pl": tag.name, + "tag_category": 'pd_author', + "is_pdcounter": True, + "uid": "tag%d_pd_a" % tag.id + } + elif isinstance(tag, PDCounterBook): + doc = { + "tag_id": int(tag.id), + "tag_name": tag.title, + "tag_name_pl": tag.title, + "tag_category": 'pd_book', + "is_pdcounter": True, + "uid": "tag%d_pd_b" % tag.id + } + else: + doc = { + "tag_id": int(tag.id), + "tag_name": tag.name, + "tag_name_pl": tag.name, + "tag_category": tag.category, + "is_pdcounter": False, + "uid": "tag%d" % tag.id + } + print "ADD 1 %s" % doc + self.index.add(doc) + + def create_book_doc(self, book): + """ + Create a lucene document referring book id. + """ + doc = { + 'book_id': int(book.id), + } + if book.parent is not None: + doc["parent_id"] = int(book.parent.id) + return doc + + def remove_book(self, book_or_id, remove_snippets=True): + """Removes a book from search index. + book - Book instance.""" + if isinstance(book_or_id, catalogue.models.Book): + book_id = book_or_id.id + else: + book_id = book_or_id + + self.delete_query(self.index.Q(book_id=book_id)) + + if remove_snippets: + snippets = Snippets(book_id) + snippets.remove() + + def index_book(self, book, book_info=None, overwrite=True): + """ + Indexes the book. + Creates a lucene document for extracted metadata + and calls self.index_content() to index the contents of the book. + """ if overwrite: - self.remove_book(book) + # we don't remove snippets, since they might be still needed by + # threads using not reopened index + self.remove_book(book, remove_snippets=False) book_doc = self.create_book_doc(book) - meta_fields = self.extract_metadata(book) - for f in meta_fields.values(): - if isinstance(f, list) or isinstance(f, tuple): - for elem in f: - book_doc.add(elem) - else: - book_doc.add(f) + meta_fields = self.extract_metadata(book, book_info, dc_only=['source_name', 'authors', 'title']) + # let's not index it - it's only used for extracting publish date + if 'source_name' in meta_fields: + del meta_fields['source_name'] - self.index.addDocument(book_doc) + for n, f in meta_fields.items(): + book_doc[n] = f + + book_doc['uid'] = "book%s" % book_doc['book_id'] + print "ADD 2 %s" % book_doc + self.index.add(book_doc) del book_doc + book_fields = { + 'title': meta_fields['title'], + 'authors': meta_fields['authors'], + 'published_date': meta_fields['published_date'] + } + if 'translators' in meta_fields: + book_fields['translators'] = meta_fields['translators'] - self.index_content(book, book_fields=[meta_fields['title'], meta_fields['author']]) + self.index_content(book, book_fields=book_fields) master_tags = [ 'opowiadanie', @@ -197,33 +269,39 @@ class Index(IndexStore): 'dramat_wierszowany_l', 'dramat_wierszowany_lp', 'dramat_wspolczesny', 'liryka_l', 'liryka_lp', - 'wywiad' + 'wywiad', + ] + + ignore_content_tags = [ + 'uwaga', 'extra', + 'zastepnik_tekstu', 'sekcja_asterysk', 'separator_linia', 'zastepnik_wersu', + 'didaskalia', + 'naglowek_aktu', 'naglowek_sceny', 'naglowek_czesc', ] - skip_header_tags = ['autor_utworu', 'nazwa_utworu', 'dzielo_nadrzedne'] + footnote_tags = ['pa', 'pt', 'pr', 'pe'] - def create_book_doc(self, book): + skip_header_tags = ['autor_utworu', 'nazwa_utworu', 'dzielo_nadrzedne', '{http://www.w3.org/1999/02/22-rdf-syntax-ns#}RDF'] + + published_date_re = re.compile("([0-9]+)[\]. ]*$") + + def extract_metadata(self, book, book_info=None, dc_only=None): """ - Create a lucene document connected to the book + Extract metadata from book and returns a map of fields keyed by fieldname """ - doc = Document() - doc.add(NumericField("book_id", Field.Store.YES, True).setIntValue(book.id)) - if book.parent is not None: - doc.add(NumericField("parent_id", Field.Store.YES, True).setIntValue(book.parent.id)) - return doc - - def extract_metadata(self, book): fields = {} - book_info = dcparser.parse(book.xml_file) - print("extract metadata for book %s id=%d, thread%d" % (book.slug, book.id, current_thread().ident)) + if book_info is None: + book_info = dcparser.parse(open(book.xml_file.path)) - fields['slug'] = Field("slug", book.slug, Field.Store.NO, Field.Index.ANALYZED_NO_NORMS) - fields['tags'] = self.add_gaps([Field("tags", t.name, Field.Store.NO, Field.Index.ANALYZED) for t in book.tags], 'tags') - fields['is_book'] = Field("is_book", 'true', Field.Store.NO, Field.Index.NOT_ANALYZED) + fields['slug'] = book.slug + fields['tags'] = [t.name for t in book.tags] + fields['is_book'] = True # validator, name for field in dcparser.BookInfo.FIELDS: + if dc_only and field.name not in dc_only: + continue if hasattr(book_info, field.name): if not getattr(book_info, field.name): continue @@ -233,90 +311,123 @@ class Index(IndexStore): s = getattr(book_info, field.name) if field.multiple: s = ', '.join(s) - try: - fields[field.name] = Field(field.name, s, Field.Store.NO, Field.Index.ANALYZED) - except JavaError as je: - raise Exception("failed to add field: %s = '%s', %s(%s)" % (field.name, s, je.message, je.args)) + fields[field.name] = s elif type_indicator == dcparser.as_person: p = getattr(book_info, field.name) if isinstance(p, dcparser.Person): persons = unicode(p) else: persons = ', '.join(map(unicode, p)) - fields[field.name] = Field(field.name, persons, Field.Store.NO, Field.Index.ANALYZED) + fields[field.name] = persons elif type_indicator == dcparser.as_date: dt = getattr(book_info, field.name) - fields[field.name] = Field(field.name, "%04d%02d%02d" %\ - (dt.year, dt.month, dt.day), Field.Store.NO, Field.Index.NOT_ANALYZED) + fields[field.name] = dt + + # get published date + pd = None + if hasattr(book_info, 'source_name') and book_info.source_name: + match = self.published_date_re.search(book_info.source_name) + if match is not None: + pd = str(match.groups()[0]) + if not pd: pd = "" + fields["published_date"] = pd + return fields + # def add_gaps(self, fields, fieldname): + # """ + # Interposes a list of fields with gap-fields, which are indexed spaces and returns it. + # This allows for doing phrase queries which do not overlap the gaps (when slop is 0). + # """ + # def gap(): + # while True: + # yield Field(fieldname, ' ', Field.Store.NO, Field.Index.NOT_ANALYZED) + # return reduce(lambda a, b: a + b, zip(fields, gap()))[0:-1] + def get_master(self, root): + """ + Returns the first master tag from an etree. + """ for master in root.iter(): if master.tag in self.master_tags: return master - def add_gaps(self, fields, fieldname): - def gap(): - while True: - yield Field(fieldname, ' ', Field.Store.NO, Field.Index.NOT_ANALYZED) - return reduce(lambda a, b: a + b, zip(fields, gap()))[0:-1] - - def index_content(self, book, book_fields=[]): - wld = WLDocument.from_file(book.xml_file.path) + def index_content(self, book, book_fields={}): + """ + Walks the book XML and extract content from it. + Adds parts for each header tag and for each fragment. + """ + wld = WLDocument.from_file(book.xml_file.path, parse_dublincore=False) root = wld.edoc.getroot() master = self.get_master(root) if master is None: return [] - def walker(node): - yield node, None - for child in list(node): - for b, e in walker(child): - yield b, e - yield None, node + def walker(node, ignore_tags=[]): + + if node.tag not in ignore_tags: + yield node, None, None + if node.text is not None: + yield None, node.text, None + for child in list(node): + for b, t, e in walker(child): + yield b, t, e + yield None, None, node + + if node.tail is not None: + yield None, node.tail, None return def fix_format(text): - return re.sub("/$", "", text, flags=re.M) + # separator = [u" ", u"\t", u".", u";", u","] + if isinstance(text, list): + # need to join it first + text = filter(lambda s: s is not None, content) + text = u' '.join(text) + # for i in range(len(text)): + # if i > 0: + # if text[i][0] not in separator\ + # and text[i - 1][-1] not in separator: + # text.insert(i, u" ") + + return re.sub("(?m)/$", "", text) def add_part(snippets, **fields): doc = self.create_book_doc(book) - for f in book_fields: - doc.add(f) + for n, v in book_fields.items(): + doc[n] = v - doc.add(NumericField('header_index', Field.Store.YES, True).setIntValue(fields["header_index"])) - doc.add(NumericField("header_span", Field.Store.YES, True)\ - .setIntValue('header_span' in fields and fields['header_span'] or 1)) - doc.add(Field('header_type', fields["header_type"], Field.Store.YES, Field.Index.NOT_ANALYZED)) + doc['header_index'] = fields["header_index"] + doc['header_span'] = 'header_span' in fields and fields['header_span'] or 1 + doc['header_type'] = fields['header_type'] - doc.add(Field('content', fields["content"], Field.Store.NO, Field.Index.ANALYZED, \ - Field.TermVector.WITH_POSITIONS_OFFSETS)) + doc['text'] = fields['text'] - snip_pos = snippets.add(fields["content"]) - doc.add(NumericField("snippets_position", Field.Store.YES, True).setIntValue(snip_pos[0])) - doc.add(NumericField("snippets_length", Field.Store.YES, True).setIntValue(snip_pos[1])) + # snippets + snip_pos = snippets.add(fields["text"]) + + doc['snippets_position'] = snip_pos[0] + doc['snippets_length'] = snip_pos[1] + if snippets.revision: + doc["snippets_revision"] = snippets.revision if 'fragment_anchor' in fields: - doc.add(Field("fragment_anchor", fields['fragment_anchor'], - Field.Store.YES, Field.Index.NOT_ANALYZED)) + doc["fragment_anchor"] = fields['fragment_anchor'] if 'themes' in fields: - themes, themes_pl = zip(*[ - (Field("themes", theme, Field.Store.YES, Field.Index.ANALYZED, Field.TermVector.WITH_POSITIONS), - Field("themes_pl", theme, Field.Store.NO, Field.Index.ANALYZED, Field.TermVector.WITH_POSITIONS)) - for theme in fields['themes']]) - - themes = self.add_gaps(themes, 'themes') - themes_pl = self.add_gaps(themes_pl, 'themes_pl') - - for t in themes: - doc.add(t) - for t in themes_pl: - doc.add(t) - + doc['themes'] = fields['themes'] + doc['uid'] = "part%s%s%s" % (doc['header_index'], + doc['header_span'], + doc.get('fragment_anchor', '')) return doc + def give_me_utf8(s): + if isinstance(s, unicode): + return s.encode('utf-8') + else: + return s + fragments = {} snippets = Snippets(book.id).open('w') try: @@ -324,265 +435,259 @@ class Index(IndexStore): if header.tag in self.skip_header_tags: continue + if header.tag is etree.Comment: + continue - content = u' '.join([t for t in header.itertext()]) - content = fix_format(content) - - doc = add_part(snippets, header_index=position, header_type=header.tag, content=content) - - self.index.addDocument(doc) - - for start, end in walker(header): + # section content + content = [] + footnote = [] + + def all_content(text): + for frag in fragments.values(): + frag['text'].append(text) + content.append(text) + handle_text = [all_content] + + for start, text, end in walker(header, ignore_tags=self.ignore_content_tags): + # handle footnotes + if start is not None and start.tag in self.footnote_tags: + footnote = [] + + def collect_footnote(t): + footnote.append(t) + + handle_text.append(collect_footnote) + elif end is not None and footnote is not [] and end.tag in self.footnote_tags: + handle_text.pop() + doc = add_part(snippets, header_index=position, header_type=header.tag, + text=u''.join(footnote), + is_footnote=True) + print "ADD 3 %s" % doc + self.index.add(doc) + #print "@ footnote text: %s" % footnote + footnote = [] + + # handle fragments and themes. if start is not None and start.tag == 'begin': fid = start.attrib['id'][1:] - fragments[fid] = {'content': [], 'themes': [], 'start_section': position, 'start_header': header.tag} - fragments[fid]['content'].append(start.tail) + fragments[fid] = {'text': [], 'themes': [], 'start_section': position, 'start_header': header.tag} + + # themes for this fragment elif start is not None and start.tag == 'motyw': fid = start.attrib['id'][1:] - fragments[fid]['themes'].append(start.text) - fragments[fid]['content'].append(start.tail) + handle_text.append(None) + if start.text is not None: + fragments[fid]['themes'] += map(unicode.strip, map(unicode, (start.text.split(',')))) + elif end is not None and end.tag == 'motyw': + handle_text.pop() + elif start is not None and start.tag == 'end': fid = start.attrib['id'][1:] if fid not in fragments: continue # a broken node, skip it frag = fragments[fid] + if frag['themes'] == []: + continue # empty themes list. del fragments[fid] - def jstr(l): - return u' '.join(map( - lambda x: x == None and u'(none)' or unicode(x), - l)) - doc = add_part(snippets, header_type=frag['start_header'], header_index=frag['start_section'], header_span=position - frag['start_section'] + 1, fragment_anchor=fid, - content=u' '.join(filter(lambda s: s is not None, frag['content'])), + text=fix_format(frag['text']), themes=frag['themes']) + #print '@ FRAG %s' % frag['content'] + print "ADD 4 %s" % doc + self.index.add(doc) - self.index.addDocument(doc) - elif start is not None: - for frag in fragments.values(): - frag['content'].append(start.text) - elif end is not None: - for frag in fragments.values(): - frag['content'].append(end.tail) - finally: - snippets.close() - - - def __enter__(self): - self.open() - return self - - def __exit__(self, type, value, tb): - self.close() - - -def log_exception_wrapper(f): - def _wrap(*a): - try: - f(*a) - except Exception, e: - print("Error in indexing thread: %s" % e) - traceback.print_exc() - raise e - return _wrap - - -class ReusableIndex(Index): - """ - Works like index, but does not close/optimize Lucene index - until program exit (uses atexit hook). - This is usefull for importbooks command. - - if you cannot rely on atexit, use ReusableIndex.close_reusable() yourself. - """ - index = None - pool = None - pool_jobs = None - - def open(self, analyzer=None, threads=4): - if ReusableIndex.index is not None: - self.index = ReusableIndex.index - else: - print("opening index") - ReusableIndex.pool = ThreadPool(threads, initializer=lambda: JVM.attachCurrentThread() ) - ReusableIndex.pool_jobs = [] - Index.open(self, analyzer) - ReusableIndex.index = self.index - atexit.register(ReusableIndex.close_reusable) - - def index_book(self, *args, **kw): - job = ReusableIndex.pool.apply_async(log_exception_wrapper(Index.index_book), (self,) + args, kw) - ReusableIndex.pool_jobs.append(job) - - @staticmethod - def close_reusable(): - if ReusableIndex.index is not None: - print("wait for indexing to finish") - for job in ReusableIndex.pool_jobs: - job.get() - sys.stdout.write('.') - sys.stdout.flush() - print("done.") - ReusableIndex.pool.close() - - ReusableIndex.index.optimize() - ReusableIndex.index.close() - ReusableIndex.index = None - - def close(self): - pass - - -class Search(IndexStore): - def __init__(self, default_field="content"): - IndexStore.__init__(self) - self.analyzer = WLAnalyzer() #PolishAnalyzer(Version.LUCENE_34) - ## self.analyzer = WLAnalyzer() - self.searcher = IndexSearcher(self.store, True) - self.parser = QueryParser(Version.LUCENE_34, default_field, - self.analyzer) - - self.parent_filter = TermsFilter() - self.parent_filter.addTerm(Term("is_book", "true")) + # Collect content. - def query(self, query): - return self.parser.parse(query) + if text is not None and handle_text is not []: + hdl = handle_text[-1] + if hdl is not None: + hdl(text) - def wrapjoins(self, query, fields=[]): - """ - This functions modifies the query in a recursive way, - so Term and Phrase Queries contained, which match - provided fields are wrapped in a BlockJoinQuery, - and so delegated to children documents. - """ - if BooleanQuery.instance_(query): - qs = BooleanQuery.cast_(query) - for clause in qs: - clause = BooleanClause.cast_(clause) - clause.setQuery(self.wrapjoins(clause.getQuery(), fields)) - return qs - else: - termset = HashSet() - query.extractTerms(termset) - for t in termset: - t = Term.cast_(t) - if t.field() not in fields: - return query - return BlockJoinQuery(query, self.parent_filter, - BlockJoinQuery.ScoreMode.Total) - - def simple_search(self, query, max_results=50): - """Returns (books, total_hits) - """ - - tops = self.searcher.search(self.query(query), max_results) - bks = [] - for found in tops.scoreDocs: - doc = self.searcher.doc(found.doc) - bks.append(catalogue.models.Book.objects.get(id=doc.get("book_id"))) - return (bks, tops.totalHits) + # in the end, add a section text. + doc = add_part(snippets, header_index=position, + header_type=header.tag, text=fix_format(content)) + #print '@ CONTENT: %s' % fix_format(content) - def search(self, query, max_results=50): - query = self.query(query) - query = self.wrapjoins(query, ["content", "themes"]) + print "ADD 5 %s" % doc + self.index.add(doc) - tops = self.searcher.search(query, max_results) - bks = [] - for found in tops.scoreDocs: - doc = self.searcher.doc(found.doc) - bks.append(catalogue.models.Book.objects.get(id=doc.get("book_id"))) - return (bks, tops.totalHits) - - def bsearch(self, query, max_results=50): - q = self.query(query) - bjq = BlockJoinQuery(q, self.parent_filter, BlockJoinQuery.ScoreMode.Avg) - - tops = self.searcher.search(bjq, max_results) - bks = [] - for found in tops.scoreDocs: - doc = self.searcher.doc(found.doc) - bks.append(catalogue.models.Book.objects.get(id=doc.get("book_id"))) - return (bks, tops.totalHits) - -# TokenStream tokenStream = analyzer.tokenStream(fieldName, reader); -# OffsetAttribute offsetAttribute = tokenStream.getAttribute(OffsetAttribute.class); -# CharTermAttribute charTermAttribute = tokenStream.getAttribute(CharTermAttribute.class); - -# while (tokenStream.incrementToken()) { -# int startOffset = offsetAttribute.startOffset(); -# int endOffset = offsetAttribute.endOffset(); -# String term = charTermAttribute.toString(); -# } + finally: + snippets.close() class SearchResult(object): - def __init__(self, searcher, scoreDocs, score=None, how_found=None, snippets=None): + def __init__(self, doc, how_found=None, query=None, query_terms=None): + # self.search = search + self.boost = 1.0 + self._hits = [] + self._processed_hits = None # processed hits self.snippets = [] + self.query_terms = query_terms - if score: - self.score = score + if 'score' in doc: + self._score = doc['score'] else: - self.score = scoreDocs.score - - self.hits = [] + self._score = 0 - stored = searcher.doc(scoreDocs.doc) - self.book_id = int(stored.get("book_id")) - - header_type = stored.get("header_type") - if not header_type: - return + self.book_id = int(doc["book_id"]) - sec = (header_type, int(stored.get("header_index"))) - header_span = stored.get('header_span') - header_span = header_span is not None and int(header_span) or 1 + try: + self.published_date = int(doc.get("published_date")) + except ValueError: + self.published_date = 0 + + # content hits + header_type = doc.get("header_type", None) + # we have a content hit in some header of fragment + if header_type is not None: + sec = (header_type, int(doc["header_index"])) + header_span = doc['header_span'] + header_span = header_span is not None and int(header_span) or 1 + fragment = doc.get("fragment_anchor", None) + snippets_pos = (doc['snippets_position'], doc['snippets_length']) + snippets_rev = doc['snippets_revision'] + + hit = (sec + (header_span,), fragment, self._score, { + 'how_found': how_found, + 'snippets_pos': snippets_pos, + 'snippets_revision': snippets_rev, + 'themes': doc.get('themes', []), + 'themes_pl': doc.get('themes_pl', []) + }) + + self._hits.append(hit) - fragment = stored.get("fragment_anchor") + def __unicode__(self): + return u" self.score: - self.score = other.score + self._score = other._score return self def get_book(self): - return catalogue.models.Book.objects.get(id=self.book_id) + if hasattr(self, '_book'): + return self._book + self._book = catalogue.models.Book.objects.get(id=self.book_id) + return self._book book = property(get_book) - def get_parts(self): - book = self.book + POSITION = 0 + FRAGMENT = 1 + POSITION_INDEX = 1 + POSITION_SPAN = 2 + SCORE = 2 + OTHER = 3 + + @property + def hits(self): + if self._processed_hits is not None: + return self._processed_hits + + # to sections and fragments + frags = filter(lambda r: r[self.FRAGMENT] is not None, self._hits) + + sect = filter(lambda r: r[self.FRAGMENT] is None, self._hits) + + # sections not covered by fragments + sect = filter(lambda s: 0 == len(filter( + lambda f: s[self.POSITION][self.POSITION_INDEX] >= f[self.POSITION][self.POSITION_INDEX] + and s[self.POSITION][self.POSITION_INDEX] < f[self.POSITION][self.POSITION_INDEX] + f[self.POSITION][self.POSITION_SPAN], + frags)), sect) + + hits = [] + + def remove_duplicates(lst, keyfn, compare): + els = {} + for e in lst: + eif = keyfn(e) + if eif in els: + if compare(els[eif], e) >= 1: + continue + els[eif] = e + return els.values() + + # remove fragments with duplicated fid's and duplicated snippets + frags = remove_duplicates(frags, lambda f: f[self.FRAGMENT], lambda a, b: cmp(a[self.SCORE], b[self.SCORE])) + # frags = remove_duplicates(frags, lambda f: f[OTHER]['snippet_pos'] and f[OTHER]['snippet_pos'] or f[FRAGMENT], + # lambda a, b: cmp(a[SCORE], b[SCORE])) + + # remove duplicate sections + sections = {} + + for s in sect: + si = s[self.POSITION][self.POSITION_INDEX] + # skip existing + if si in sections: + if sections[si]['score'] >= s[self.SCORE]: + continue + + m = {'score': s[self.SCORE], + 'section_number': s[self.POSITION][self.POSITION_INDEX] + 1, + } + m.update(s[self.OTHER]) + sections[si] = m - def sections_covered(results): - frags = filter(lambda r: r[1] is not None, results) - sect = filter(lambda r: r[1] is None, results) - sect = filter(lambda s: 0 == len(filter( - lambda f: s[0][1] >= f[0][1] and s[0][1] < f[0][1] + f[0][2], - frags)), sect) - print "filtered, non overlapped sections: %s" % sect - return frags + sect + hits = sections.values() - parts = [{"header": s[0], "position": s[1], '_score_key': s} for s in self.sections] \ - + [{"fragment": book.fragments.get(anchor=f), '_score_key':f} for f in self.fragments] + for f in frags: + try: + frag = catalogue.models.Fragment.objects.get(anchor=f[self.FRAGMENT], book__id=self.book_id) + except catalogue.models.Fragment.DoesNotExist: + # stale index + continue + print f + # Figure out if we were searching for a token matching some word in theme name. + themes = frag.tags.filter(category='theme') + themes_hit = set() + if self.query_terms is not None: + for i in range(0, len(f[self.OTHER]['themes'])): + tms = f[self.OTHER]['themes'][i].split(r' +') + f[self.OTHER]['themes_pl'][i].split(' ') + tms = map(unicode.lower, tms) + for qt in self.query_terms: + if qt in tms: + themes_hit.add(f[self.OTHER]['themes'][i]) + break + + def theme_by_name(n): + th = filter(lambda t: t.name == n, themes) + if th: + return th[0] + else: + return None + themes_hit = filter(lambda a: a is not None, map(theme_by_name, themes_hit)) - parts.sort(lambda a, b: cmp(self.scores[a['_score_key']], self.scores[b['_score_key']])) - print("bookid: %d parts: %s" % (self.book_id, parts)) - return parts + m = {'score': f[self.SCORE], + 'fragment': frag, + 'section_number': f[self.POSITION][self.POSITION_INDEX] + 1, + 'themes': themes, + 'themes_hit': themes_hit + } + m.update(f[self.OTHER]) + hits.append(m) - parts = property(get_parts) + hits.sort(lambda a, b: cmp(a['score'], b['score']), reverse=True) - def __unicode__(self): - return u'SearchResult(book_id=%d, score=%d)' % (self.book_id, self.score) + self._processed_hits = hits + + return hits @staticmethod def aggregate(*result_lists): @@ -591,396 +696,414 @@ class SearchResult(object): for r in rl: if r.book_id in books: books[r.book_id].merge(r) - #print(u"already have one with score %f, and this one has score %f" % (books[book.id][0], found.score)) else: books[r.book_id] = r return books.values() def __cmp__(self, other): - return cmp(self.score, other.score) - - -class Hint(object): - def __init__(self, search): - self.search = search - self.book_tags = {} - self.part_tags = [] - self.book = None - - def book(self, book): - self.book = book - - def tags(self, tags): - for t in tags: - if t.category in ['author', 'title', 'epoch', 'genre', 'kind']: - lst = self.book_tags.get(t.category, []) - lst.append(t) - self.book_tags[t.category] = lst - if t.category in ['theme']: - self.part_tags.append(t) - - def tag_filter(self, tags, field='tags'): - q = BooleanQuery() - - for tag in tags: - toks = self.search.get_tokens(tag.name, field=field) - tag_phrase = PhraseQuery() - for tok in toks: - tag_phrase.add(Term(field, tok)) - q.add(BooleanClause(tag_phrase, BooleanClause.Occur.MUST)) - - return QueryWrapperFilter(q) - - def book_filter(self): - tags = reduce(lambda a, b: a + b, self.book_tags.values(), []) - if tags: - return self.tag_filter(tags) + c = cmp(self.score, other.score) + if c == 0: + # this is inverted, because earlier date is better + return cmp(other.published_date, self.published_date) else: - return None + return c - def part_filter(self): - if self.part_tags: - return self.tag_filter(self.part_tags, field='themes') - else: - return None + def __len__(self): + return len(self.hits) - def should_search_for_book(self): - return self.book is None + def snippet_pos(self, idx=0): + return self.hits[idx]['snippets_pos'] - def just_search_in(self, all): - """Holds logic to figure out which indexes should be search, when we have some hinst already""" - some = [] - for field in all: - if field == 'author' and 'author' in self.book_tags: - continue - if field == 'title' and self.book is not None: - continue - if (field == 'themes' or field == 'themes_pl') and self.part_tags: - continue - some.append(field) - return some - - -class MultiSearch(Search): - """Class capable of IMDb-like searching""" - def get_tokens(self, searched, field='content'): - """returns tokens analyzed by a proper (for a field) analyzer - argument can be: StringReader, string/unicode, or tokens. In the last case - they will just be returned (so we can reuse tokens, if we don't change the analyzer) - """ - if isinstance(searched, str) or isinstance(searched, unicode): - searched = StringReader(searched) - elif isinstance(searched, list): - return searched - - searched.reset() - tokens = self.analyzer.reusableTokenStream(field, searched) - toks = [] - while tokens.incrementToken(): - cta = tokens.getAttribute(CharTermAttribute.class_) - toks.append(cta.toString()) - return toks - - def fuzziness(self, fuzzy): - if not fuzzy: + def snippet_revision(self, idx=0): + try: + return self.hits[idx]['snippets_revision'] + except: return None - if isinstance(fuzzy, float) and fuzzy > 0.0 and fuzzy <= 1.0: - return fuzzy - else: - return 0.5 - def make_phrase(self, tokens, field='content', slop=2, fuzzy=False): - if fuzzy: - phrase = MultiPhraseQuery() - for t in tokens: - term = Term(field, t) - fuzzterm = FuzzyTermEnum(self.searcher.getIndexReader(), term, self.fuzziness(fuzzy)) - fuzzterms = [] - while True: - # print("fuzz %s" % unicode(fuzzterm.term()).encode('utf-8')) - ft = fuzzterm.term() - if ft: - fuzzterms.append(ft) - if not fuzzterm.next(): break - if fuzzterms: - phrase.add(JArray('object')(fuzzterms, Term)) - else: - phrase.add(term) - else: - phrase = PhraseQuery() - phrase.setSlop(slop) - for t in tokens: - term = Term(field, t) - phrase.add(term) - return phrase - - def make_term_query(self, tokens, field='content', modal=BooleanClause.Occur.SHOULD, fuzzy=False): - q = BooleanQuery() - for t in tokens: - term = Term(field, t) - if fuzzy: - term = FuzzyQuery(term, self.fuzziness(fuzzy)) - else: - term = TermQuery(term) - q.add(BooleanClause(term, modal)) - return q - - def content_query(self, query): - return BlockJoinQuery(query, self.parent_filter, - BlockJoinQuery.ScoreMode.Total) - - def search_perfect_book(self, searched, max_results=20, fuzzy=False, hint=None): - fields_to_search = ['author', 'title'] - only_in = None - if hint: - if not hint.should_search_for_book(): - return [] - fields_to_search = hint.just_search_in(fields_to_search) - only_in = hint.book_filter() - - qrys = [self.make_phrase(self.get_tokens(searched, field=fld), field=fld, fuzzy=fuzzy) for fld in fields_to_search] - - books = [] - for q in qrys: - top = self.searcher.search(q, - self.chain_filters([only_in, self.term_filter(Term('is_book', 'true'))]), - max_results) - for found in top.scoreDocs: - books.append(SearchResult(self.searcher, found)) - return books - - def search_perfect_parts(self, searched, max_results=20, fuzzy=False, hint=None): - qrys = [self.make_phrase(self.get_tokens(searched), field=fld, fuzzy=fuzzy) for fld in ['content']] - - flt = None - if hint: - flt = hint.part_filter() - - books = [] - for q in qrys: - top = self.searcher.search(q, - self.chain_filters([self.term_filter(Term('is_book', 'true'), inverse=True), - flt - ]), - max_results) - for found in top.scoreDocs: - books.append(SearchResult(self.searcher, found, snippets=self.get_snippets(found, q))) +class Search(SolrIndex): + """ + Search facilities. + """ + def __init__(self, default_field="text"): + super(Search, self).__init__(mode='r') + + # def get_tokens(self, searched, field='text', cached=None): + # """returns tokens analyzed by a proper (for a field) analyzer + # argument can be: StringReader, string/unicode, or tokens. In the last case + # they will just be returned (so we can reuse tokens, if we don't change the analyzer) + # """ + # if cached is not None and field in cached: + # return cached[field] + + # if isinstance(searched, str) or isinstance(searched, unicode): + # searched = StringReader(searched) + # elif isinstance(searched, list): + # return searched + + # searched.reset() + # tokens = self.analyzer.reusableTokenStream(field, searched) + # toks = [] + # while tokens.incrementToken(): + # cta = tokens.getAttribute(CharTermAttribute.class_) + # toks.append(cta.toString()) + + # if cached is not None: + # cached[field] = toks + + # return toks + + # @staticmethod + # def fuzziness(fuzzy): + # """Helper method to sanitize fuzziness""" + # if not fuzzy: + # return None + # if isinstance(fuzzy, float) and fuzzy > 0.0 and fuzzy <= 1.0: + # return fuzzy + # else: + # return 0.5 + + # def make_phrase(self, tokens, field='text', slop=2, fuzzy=False): + # """ + # Return a PhraseQuery with a series of tokens. + # """ + # if fuzzy: + # phrase = MultiPhraseQuery() + # for t in tokens: + # term = Term(field, t) + # fuzzterm = FuzzyTermEnum(self.searcher.getIndexReader(), term, self.fuzziness(fuzzy)) + # fuzzterms = [] + + # while True: + # ft = fuzzterm.term() + # if ft: + # fuzzterms.append(ft) + # if not fuzzterm.next(): break + # if fuzzterms: + # phrase.add(JArray('object')(fuzzterms, Term)) + # else: + # phrase.add(term) + # else: + # phrase = PhraseQuery() + # phrase.setSlop(slop) + # for t in tokens: + # term = Term(field, t) + # phrase.add(term) + # return phrase + + def make_term_query(self, query, field='text', modal=operator.or_): + """ + Returns term queries joined by boolean query. + modal - applies to boolean query + fuzzy - should the query by fuzzy. + """ + q = self.index.Q() + q = reduce(modal, map(lambda s: self.index.Q(**{field: s}), + query.split(r" ")), q) - return books + return q - def search_everywhere(self, searched, max_results=20, fuzzy=False, hint=None): + def search_phrase(self, searched, field='text', book=False, + filters=None, + snippets=False): + if filters is None: filters = [] + if book: filters.append(self.index.Q(is_book=True)) + + q = self.index.query(**{field: searched}) + q = self.apply_filters(q, filters).field_limit(score=True, all_fields=True) + res = q.execute() + return [SearchResult(found, how_found=u'search_phrase') for found in res] + + def search_some(self, searched, fields, book=True, + filters=None, snippets=True, query_terms=None): + assert isinstance(fields, list) + if filters is None: filters = [] + if book: filters.append(self.index.Q(is_book=True)) + + query = self.index.Q() + + for fld in fields: + query = self.index.Q(query | self.make_term_query(searched, fld)) + + query = self.index.query(query) + query = self.apply_filters(query, filters).field_limit(score=True, all_fields=True) + res = query.execute() + return [SearchResult(found, how_found='search_some', query_terms=query_terms) for found in res] + + # def search_perfect_book(self, searched, max_results=20, fuzzy=False, hint=None): + # """ + # Search for perfect book matches. Just see if the query matches with some author or title, + # taking hints into account. + # """ + # fields_to_search = ['authors', 'title'] + # only_in = None + # if hint: + # if not hint.should_search_for_book(): + # return [] + # fields_to_search = hint.just_search_in(fields_to_search) + # only_in = hint.book_filter() + + # qrys = [self.make_phrase(self.get_tokens(searched, field=fld), field=fld, fuzzy=fuzzy) for fld in fields_to_search] + + # books = [] + # for q in qrys: + # top = self.searcher.search(q, + # self.chain_filters([only_in, self.term_filter(Term('is_book', 'true'))]), + # max_results) + # for found in top.scoreDocs: + # books.append(SearchResult(self, found, how_found="search_perfect_book")) + # return books + + # def search_book(self, searched, max_results=20, fuzzy=False, hint=None): + # fields_to_search = ['tags', 'authors', 'title'] + + # only_in = None + # if hint: + # if not hint.should_search_for_book(): + # return [] + # fields_to_search = hint.just_search_in(fields_to_search) + # only_in = hint.book_filter() + + # tokens = self.get_tokens(searched, field='SIMPLE') + + # q = BooleanQuery() + + # for fld in fields_to_search: + # q.add(BooleanClause(self.make_term_query(tokens, field=fld, + # fuzzy=fuzzy), BooleanClause.Occur.SHOULD)) + + # books = [] + # top = self.searcher.search(q, + # self.chain_filters([only_in, self.term_filter(Term('is_book', 'true'))]), + # max_results) + # for found in top.scoreDocs: + # books.append(SearchResult(self, found, how_found="search_book")) + + # return books + + # def search_perfect_parts(self, searched, max_results=20, fuzzy=False, hint=None): + # """ + # Search for book parts which contains a phrase perfectly matching (with a slop of 2, default for make_phrase()) + # some part/fragment of the book. + # """ + # qrys = [self.make_phrase(self.get_tokens(searched), field=fld, fuzzy=fuzzy) for fld in ['text']] + + # flt = None + # if hint: + # flt = hint.part_filter() + + # books = [] + # for q in qrys: + # top = self.searcher.search(q, + # self.chain_filters([self.term_filter(Term('is_book', 'true'), inverse=True), + # flt]), + # max_results) + # for found in top.scoreDocs: + # books.append(SearchResult(self, found, snippets=self.get_snippets(found, q), how_found='search_perfect_parts')) + + # return books + + def search_everywhere(self, searched, query_terms=None): + """ + Tries to use search terms to match different fields of book (or its parts). + E.g. one word can be an author survey, another be a part of the title, and the rest + are some words from third chapter. + """ books = [] - only_in = None - - if hint: - only_in = hint.part_filter() - # content only query : themes x content - q = BooleanQuery() - - tokens = self.get_tokens(searched) - if hint is None or hint.just_search_in(['themes_pl']) != []: - q.add(BooleanClause(self.make_term_query(tokens, field='themes_pl', fuzzy=fuzzy), BooleanClause.Occur.MUST)) - - q.add(BooleanClause(self.make_term_query(tokens, field='content', fuzzy=fuzzy), BooleanClause.Occur.SHOULD)) + q = self.make_term_query(searched, 'text') + q_themes = self.make_term_query(searched, 'themes_pl') - topDocs = self.searcher.search(q, only_in, max_results) - for found in topDocs.scoreDocs: - books.append(SearchResult(self.searcher, found)) + query = self.index.query(q).query(q_themes).field_limit(score=True, all_fields=True) + res = query.execute() - # joined query themes/content x author/title/epochs/genres/kinds - # q = BooleanQuery() - # in_meta = BooleanQuery() - # in_content = BooleanQuery() + for found in res: + books.append(SearchResult(found, how_found='search_everywhere_themesXcontent', query_terms=query_terms)) - # for fld in ['themes', 'content']: - # in_content.add(BooleanClause(self.make_term_query(tokens, field=fld, fuzzy=False), BooleanClause.Occur.SHOULD)) + # query themes/content x author/title/tags + in_content = self.index.Q() + in_meta = self.index.Q() - # in_meta.add(BooleanClause(self.make_term_query( - # self.get_tokens(searched, field='author'), field='author', fuzzy=fuzzy), BooleanClause.Occur.SHOULD)) + for fld in ['themes_pl', 'text']: + in_content |= self.make_term_query(searched, field=fld) - # for fld in ['title', 'epochs', 'genres', 'kinds']: - # in_meta.add(BooleanClause(self.make_term_query(tokens, field=fld, fuzzy=fuzzy), BooleanClause.Occur.SHOULD)) + for fld in ['tags', 'authors', 'title']: + in_meta |= self.make_term_query(searched, field=fld) - # q.add(BooleanClause(in_meta, BooleanClause.Occur.MUST)) - # in_content_join = self.content_query(in_content) - # q.add(BooleanClause(in_content_join, BooleanClause.Occur.MUST)) - # # import pdb; pdb.set_trace() - # collector = BlockJoinCollector(Sort.RELEVANCE, 100, True, True) + q = in_content & in_meta + res = self.index.query(q).field_limit(score=True, all_fields=True).execute() - # self.searcher.search(q, collector) + for found in res: + books.append(SearchResult(found, how_found='search_everywhere', query_terms=query_terms)) - # top_groups = collector.getTopGroups(in_content_join, Sort.RELEVANCE, 0, max_results, 0, True) - # if top_groups: - # for grp in top_groups.groups: - # for part in grp.scoreDocs: - # books.append(SearchResult(self.searcher, part, score=grp.maxScore)) return books - def multisearch(self, query, max_results=50): + def get_snippets(self, searchresult, query, field='text', num=1): """ - Search strategy: - - (phrase) OR -> content - -> title - -> author - - (keywords) -> author - -> motyw - -> tags - -> content + Returns a snippet for found scoreDoc. """ - # queryreader = StringReader(query) - # tokens = self.get_tokens(queryreader) - - # top_level = BooleanQuery() - # Should = BooleanClause.Occur.SHOULD - - # phrase_level = BooleanQuery() - # phrase_level.setBoost(1.3) - - # p_content = self.make_phrase(tokens, joined=True) - # p_title = self.make_phrase(tokens, 'title') - # p_author = self.make_phrase(tokens, 'author') - - # phrase_level.add(BooleanClause(p_content, Should)) - # phrase_level.add(BooleanClause(p_title, Should)) - # phrase_level.add(BooleanClause(p_author, Should)) - - # kw_level = BooleanQuery() - - # kw_level.add(self.make_term_query(tokens, 'author'), Should) - # j_themes = self.make_term_query(tokens, 'themes', joined=True) - # kw_level.add(j_themes, Should) - # kw_level.add(self.make_term_query(tokens, 'tags'), Should) - # j_con = self.make_term_query(tokens, joined=True) - # kw_level.add(j_con, Should) - - # top_level.add(BooleanClause(phrase_level, Should)) - # top_level.add(BooleanClause(kw_level, Should)) - - return None - - def book_search(self, query, filter=None, max_results=50, collector=None): - tops = self.searcher.search(query, filter, max_results) - #tops = self.searcher.search(p_content, max_results) - - bks = [] - for found in tops.scoreDocs: - doc = self.searcher.doc(found.doc) - b = catalogue.models.Book.objects.get(id=doc.get("book_id")) - bks.append(b) - print "%s (%d) -> %f" % (b, b.id, found.score) - return bks - - def get_snippets(self, scoreDoc, query, field='content'): - htmlFormatter = SimpleHTMLFormatter() - highlighter = Highlighter(htmlFormatter, QueryScorer(query)) - - stored = self.searcher.doc(scoreDoc.doc) - - # locate content. - snippets = Snippets(stored.get('book_id')).open() + maxnum = len(searchresult) + if num is None or num < 0 or num > maxnum: + num = maxnum + book_id = searchresult.book_id + revision = searchresult.snippet_revision() + snippets = Snippets(book_id, revision=revision) + snips = [None] * maxnum try: - text = snippets.get((int(stored.get('snippets_position')), - int(stored.get('snippets_length')))) + snippets.open() + idx = 0 + while idx < maxnum and num > 0: + position, length = searchresult.snippet_pos(idx) + if position is None or length is None: + continue + text = snippets.get((int(position), + int(length))) + print "== %s -- %s ==" % (query, text) + snip = self.index.highlight(text=text, field=field, q=query) + snips[idx] = snip + if snip: + num -= 1 + idx += 1 + + except IOError, e: + log.error("Cannot open snippet file for book id = %d [rev=%d], %s" % (book_id, revision, e)) + return [] finally: snippets.close() - tokenStream = TokenSources.getAnyTokenStream(self.searcher.getIndexReader(), scoreDoc.doc, field, self.analyzer) - # highlighter.getBestTextFragments(tokenStream, text, False, 10) - # import pdb; pdb.set_trace() - snip = highlighter.getBestFragments(tokenStream, text, 3, "...") - print('snips: %s' % snip) + # remove verse end markers.. + snips = map(lambda s: s and s.replace("/\n", "\n"), snips) - return [snip] + searchresult.snippets = snips + return snips - @staticmethod - def enum_to_array(enum): + def hint_tags(self, query, pdcounter=True, prefix=True): """ - Converts a lucene TermEnum to array of Terms, suitable for - addition to queries + Return auto-complete hints for tags + using prefix search. """ - terms = [] - - while True: - t = enum.term() - if t: - terms.append(t) - if not enum.next(): break + q = self.index.Q() + query = query.strip() + for field in ['tag_name', 'tag_name_pl']: + if prefix: + q |= self.index.Q(**{field: query + "*"}) + else: + q |= self.make_term_query(query, field=field) + qu = self.index.query(q).exclude(tag_category="book") - if terms: - return JArray('object')(terms, Term) + return self.search_tags(qu, pdcounter=pdcounter) - def search_tags(self, query, filter=None, max_results=40): - tops = self.searcher.search(query, filter, max_results) + def search_tags(self, query, filters=None, pdcounter=False): + """ + Search for Tag objects using query. + """ + if not filters: filters = [] + if not pdcounter: + filters.append(~self.index.Q(is_pdcounter=True)) + res = self.apply_filters(query, filters).execute() tags = [] - for found in tops.scoreDocs: - doc = self.searcher.doc(found.doc) - tag = catalogue.models.Tag.objects.get(id=doc.get("tag_id")) - tags.append(tag) - print "%s (%d) -> %f" % (tag, tag.id, found.score) - - return tags - - def create_prefix_phrase(self, toks, field): - q = MultiPhraseQuery() - for i in range(len(toks)): - t = Term(field, toks[i]) - if i == len(toks) - 1: - pterms = MultiSearch.enum_to_array(PrefixTermEnum(self.searcher.getIndexReader(), t)) - if pterms: - q.add(pterms) + for doc in res: + is_pdcounter = doc.get('is_pdcounter', False) + category = doc.get('tag_category') + try: + if is_pdcounter == True: + if category == 'pd_author': + tag = PDCounterAuthor.objects.get(id=doc.get('tag_id')) + elif category == 'pd_book': + tag = PDCounterBook.objects.get(id=doc.get('tag_id')) + tag.category = 'pd_book' # make it look more lik a tag. + else: + print "Warning. cannot get pdcounter tag_id=%d from db; cat=%s" % (int(doc.get('tag_id')), category) else: - q.add(t) - else: - q.add(t) - return q + tag = catalogue.models.Tag.objects.get(id=doc.get("tag_id")) + # don't add the pdcounter tag if same tag already exists - @staticmethod - def term_filter(term, inverse=False): - only_term = TermsFilter() - only_term.addTerm(term) - - if inverse: - neg = BooleanFilter() - neg.add(FilterClause(only_term, BooleanClause.Occur.MUST_NOT)) - only_term = neg + tags.append(tag) - return only_term + except catalogue.models.Tag.DoesNotExist: pass + except PDCounterAuthor.DoesNotExist: pass + except PDCounterBook.DoesNotExist: pass - def hint_tags(self, string, max_results=50): - toks = self.get_tokens(string, field='SIMPLE') - top = BooleanQuery() + log.debug('search_tags: %s' % tags) - for field in ['tag_name', 'tag_name_pl']: - q = self.create_prefix_phrase(toks, field) - top.add(BooleanClause(q, BooleanClause.Occur.SHOULD)) - - no_book_cat = self.term_filter(Term("tag_category", "book"), inverse=True) + return tags - return self.search_tags(top, no_book_cat, max_results=max_results) + def hint_books(self, query, prefix=True): + """ + Returns auto-complete hints for book titles + Because we do not index 'pseudo' title-tags. + Prefix search. + """ + q = self.index.Q() + query = query.strip() + if prefix: + q |= self.index.Q(title=query + "*") + else: + q |= self.make_term_query(query, field='title') + qu = self.index.query(q) + only_books = self.index.Q(is_book=True) + return self.search_books(qu, [only_books]) - def hint_books(self, string, max_results=50): - toks = self.get_tokens(string, field='SIMPLE') + def search_books(self, query, filters=None, max_results=10): + """ + Searches for Book objects using query + """ + bks = [] + res = self.apply_filters(query, filters).field_limit(['book_id']) + for r in res: + try: + bks.append(catalogue.models.Book.objects.get(id=r['book_id'])) + except catalogue.models.Book.DoesNotExist: pass + return bks + + # def make_prefix_phrase(self, toks, field): + # q = MultiPhraseQuery() + # for i in range(len(toks)): + # t = Term(field, toks[i]) + # if i == len(toks) - 1: + # pterms = Search.enum_to_array(PrefixTermEnum(self.searcher.getIndexReader(), t)) + # if pterms: + # q.add(pterms) + # else: + # q.add(t) + # else: + # q.add(t) + # return q + + # @staticmethod + # def term_filter(term, inverse=False): + # only_term = TermsFilter() + # only_term.addTerm(term) + + # if inverse: + # neg = BooleanFilter() + # neg.add(FilterClause(only_term, BooleanClause.Occur.MUST_NOT)) + # only_term = neg + + # return only_term - q = self.create_prefix_phrase(toks, 'title') - return self.book_search(q, self.term_filter(Term("is_book", "true")), max_results=max_results) @staticmethod - def chain_filters(filters, op=ChainedFilter.AND): + def apply_filters(query, filters): + """ + Apply filters to a query + """ + if filters is None: filters = [] filters = filter(lambda x: x is not None, filters) - if not filters: - return None - chf = ChainedFilter(JArray('object')(filters, Filter), op) - return chf - - def filtered_categories(self, tags): - cats = {} - for t in tags: - cats[t.category] = True - return cats.keys() - - def hint(self): - return Hint(self) + for f in filters: + query = query.query(f) + return query + + # def filtered_categories(self, tags): + # """ + # Return a list of tag categories, present in tags list. + # """ + # cats = {} + # for t in tags: + # cats[t.category] = True + # return cats.keys() + + # def hint(self): + # return Hint(self)