X-Git-Url: https://git.mdrn.pl/wolnelektury.git/blobdiff_plain/dbb6eb3883a5f5e371f4bf7c89e74326feca0fd1..480da6ff4ed8c04b205710c318b2fcfcb961bbc7:/apps/search/index.py diff --git a/apps/search/index.py b/apps/search/index.py index 9d6d59861..e7f28c9da 100644 --- a/apps/search/index.py +++ b/apps/search/index.py @@ -1,26 +1,7 @@ # -*- coding: utf-8 -*- from django.conf import settings -from django.dispatch import Signal -from lucene import SimpleFSDirectory, NIOFSDirectory, IndexWriter, IndexReader, IndexWriterConfig, CheckIndex, \ - File, Field, Integer, \ - NumericField, Version, Document, JavaError, IndexSearcher, \ - QueryParser, PerFieldAnalyzerWrapper, \ - SimpleAnalyzer, PolishAnalyzer, ArrayList, \ - KeywordAnalyzer, NumericRangeQuery, NumericRangeFilter, 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 @@ -29,82 +10,17 @@ from librarian.parser import WLDocument from lxml import etree import catalogue.models from pdcounter.models import Author as PDCounterAuthor, BookStub as PDCounterBook -from multiprocessing.pool import ThreadPool -from threading import current_thread -import atexit +from itertools import chain import traceback +import logging +log = logging.getLogger('search') +import sunburnt +import highlight -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("authors", simple) - self.addAnalyzer("title", simple) - - self.addAnalyzer("is_book", keyword) - # shouldn't the title have two forms? _pl and simple? - - self.addAnalyzer("themes", simple) - self.addAnalyzer("themes_pl", polish) - - self.addAnalyzer("tag_name", simple) - self.addAnalyzer("tag_name_pl", polish) - - self.addAnalyzer("translators", simple) - - self.addAnalyzer("KEYWORD", keyword) - self.addAnalyzer("SIMPLE", simple) - self.addAnalyzer("POLISH", polish) - - -class IndexStore(object): - """ - Provides access to search index. - - self.store - lucene index directory - """ - def __init__(self): - self.make_index_dir() - self.store = NIOFSDirectory(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 - - def close(self): - self.store.close() - - -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 = highlight.HLSolrInterface(settings.SOLR, mode=mode) class Snippets(object): @@ -147,7 +63,6 @@ class Snippets(object): if not os.path.exists(self.path): break self.revision += 1 - print "using %s" % self.path self.file = open(self.path, mode) self.position = 0 @@ -190,60 +105,38 @@ class Snippets(object): pass -class BaseIndex(IndexStore): - """ - Base index class. - Provides basic operations on index: opening, closing, optimizing. - """ - def __init__(self, analyzer=None): - super(BaseIndex, self).__init__() - self.index = None - if not analyzer: - analyzer = WLAnalyzer() - self.analyzer = analyzer - - def open(self, timeout=None): - if self.index: - raise Exception("Index is already opened") - conf = IndexWriterConfig(Version.LUCENE_34, self.analyzer) - if timeout: - conf.setWriteLockTimeout(long(timeout)) - self.index = IndexWriter(self.store, conf) - return self.index - - def optimize(self): - self.index.optimize() - - 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 - - index_changed.send_robust(self) - - super(BaseIndex, self).close() - - def __enter__(self): - self.open() - return self - - def __exit__(self, type, value, tb): - self.close() - - -index_changed = Signal() - - -class Index(BaseIndex): +class Index(SolrIndex): """ Class indexing books. """ - def __init__(self, analyzer=None): - super(Index, self).__init__(analyzer) + def __init__(self): + super(Index, self).__init__() + + 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 index_tags(self, *tags, **kw): """ @@ -254,76 +147,82 @@ class Index(BaseIndex): remove_only = kw.get('remove_only', False) # first, remove tags from index. if tags: - q = BooleanQuery() + tag_qs = [] for tag in tags: - b_id_cat = BooleanQuery() - - q_id = NumericRangeQuery.newIntRange("tag_id", tag.id, tag.id, True, True) - b_id_cat.add(q_id, BooleanClause.Occur.MUST) + q_id = self.index.Q(tag_id=tag.id) if isinstance(tag, PDCounterAuthor): - q_cat = TermQuery(Term('tag_category', 'pd_author')) + q_cat = self.index.Q(tag_category='pd_author') elif isinstance(tag, PDCounterBook): - q_cat = TermQuery(Term('tag_category', 'pd_book')) + q_cat = self.index.Q(tag_category='pd_book') else: - q_cat = TermQuery(Term('tag_category', tag.category)) - b_id_cat.add(q_cat, BooleanClause.Occur.MUST) + q_cat = self.index.Q(tag_category=tag.category) - q.add(b_id_cat, BooleanClause.Occur.SHOULD) + q_id_cat = self.index.Q(q_id & q_cat) + tag_qs.append(q_id_cat) + self.delete_query(tag_qs) else: # all - q = NumericRangeQuery.newIntRange("tag_id", 0, Integer.MAX_VALUE, True, True) - self.index.deleteDocuments(q) + 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 = catalogue.models.Tag.objects.exclude(category='set') + \ - PDCounterAuthor.objects.all() + \ - PDCounterBook.objects.all() + tags = chain(catalogue.models.Tag.objects.exclude(category='set'), \ + PDCounterAuthor.objects.all(), \ + PDCounterBook.objects.all()) for tag in tags: if isinstance(tag, PDCounterAuthor): - doc = Document() - doc.add(NumericField("tag_id", Field.Store.YES, True).setIntValue(int(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", 'pd_author', Field.Store.YES, Field.Index.NOT_ANALYZED)) - doc.add(Field("is_pdcounter", 'true', Field.Store.YES, Field.Index.NOT_ANALYZED)) - self.index.addDocument(doc) + doc = { + "tag_id": int(tag.id), + "tag_name": tag.name, + "tag_name_pl": tag.name, + "tag_category": 'pd_author', + "is_pdcounter": True + } elif isinstance(tag, PDCounterBook): - doc = Document() - doc.add(NumericField("tag_id", Field.Store.YES, True).setIntValue(int(tag.id))) - doc.add(Field("tag_name", tag.title, Field.Store.NO, Field.Index.ANALYZED)) - doc.add(Field("tag_name_pl", tag.title, Field.Store.NO, Field.Index.ANALYZED)) - doc.add(Field("tag_category", 'pd_book', Field.Store.YES, Field.Index.NOT_ANALYZED)) - doc.add(Field("is_pdcounter", 'true', Field.Store.YES, Field.Index.NOT_ANALYZED)) - self.index.addDocument(doc) + doc = { + "tag_id": int(tag.id), + "tag_name": tag.title, + "tag_name_pl": tag.title, + "tag_category": 'pd_book', + "is_pdcounter": True + } else: - doc = Document() - doc.add(NumericField("tag_id", Field.Store.YES, True).setIntValue(int(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) + doc = { + "tag_id": int(tag.id), + "tag_name": tag.name, + "tag_name_pl": tag.name, + "tag_category": tag.category, + "is_pdcounter": False + } + doc['uid'] = "tag%d" % tag.id + self.index.add(doc) def create_book_doc(self, book): """ Create a lucene document referring book id. """ - doc = Document() - doc.add(NumericField("book_id", Field.Store.YES, True).setIntValue(int(book.id))) + doc = { + 'book_id': int(book.id), + } if book.parent is not None: - doc.add(NumericField("parent_id", Field.Store.YES, True).setIntValue(int(book.parent.id))) + doc["parent_id"] = int(book.parent.id) return doc - def remove_book(self, book, remove_snippets=True): + def remove_book(self, book_or_id, remove_snippets=True): """Removes a book from search index. book - Book instance.""" - q = NumericRangeQuery.newIntRange("book_id", book.id, book.id, True, True) - self.index.deleteDocuments(q) + 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 = Snippets(book_id) snippets.remove() def index_book(self, book, book_info=None, overwrite=True): @@ -338,17 +237,22 @@ class Index(BaseIndex): self.remove_book(book, remove_snippets=False) book_doc = self.create_book_doc(book) - meta_fields = self.extract_metadata(book, book_info) - 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) - self.index.addDocument(book_doc) + 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'] + + for n, f in meta_fields.items(): + book_doc[n] = f + + book_doc['uid'] = "book%s" % book_doc['book_id'] + self.index.add(book_doc) del book_doc - self.index_content(book, book_fields=[meta_fields['title'], meta_fields['authors'], meta_fields['published_date']]) + self.index_content(book, book_fields={ + 'title': meta_fields['title'], + 'authors': meta_fields['authors'], + 'published_date': meta_fields['published_date']}) master_tags = [ 'opowiadanie', @@ -372,7 +276,7 @@ class Index(BaseIndex): published_date_re = re.compile("([0-9]+)[\]. ]*$") - def extract_metadata(self, book, book_info=None): + def extract_metadata(self, book, book_info=None, dc_only=None): """ Extract metadata from book and returns a map of fields keyed by fieldname """ @@ -381,12 +285,14 @@ class Index(BaseIndex): 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 @@ -396,21 +302,17 @@ class Index(BaseIndex): 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 @@ -419,19 +321,19 @@ class Index(BaseIndex): if match is not None: pd = str(match.groups()[0]) if not pd: pd = "" - fields["published_date"] = Field("published_date", pd, Field.Store.YES, Field.Index.NOT_ANALYZED) + 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 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): """ @@ -441,7 +343,7 @@ class Index(BaseIndex): if master.tag in self.master_tags: return master - def index_content(self, book, book_fields=[]): + 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. @@ -484,43 +386,31 @@ class Index(BaseIndex): 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'] - print ">>[%s]>%s<<<" % (fields.get('fragment_anchor', ''), fields['content']) + doc['text'] = fields['text'] - doc.add(Field('content', fields["content"], Field.Store.NO, Field.Index.ANALYZED, \ - Field.TermVector.WITH_POSITIONS_OFFSETS)) + # snippets + snip_pos = snippets.add(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])) + doc['snippets_position'] = snip_pos[0] + doc['snippets_length'] = snip_pos[1] if snippets.revision: - doc.add(NumericField("snippets_revision", Field.Store.YES, True).setIntValue(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): @@ -545,39 +435,40 @@ class Index(BaseIndex): def all_content(text): for frag in fragments.values(): - frag['content'].append(text) + 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, - content=u''.join(footnote), - is_footnote=Field("is_footnote", 'true', Field.Store.NO, Field.Index.NOT_ANALYZED)) - - self.index.addDocument(doc) + text=u''.join(footnote), + is_footnote=True) + + 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] = {'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:] handle_text.append(None) if start.text is not None: - fragments[fid]['themes'] += map(str.strip, map(give_me_utf8, start.text.split(','))) + fragments[fid]['themes'] += map(unicode.strip, map(unicode, (start.text.split(',')))) elif end is not None and end.tag == 'motyw': handle_text.pop() @@ -595,10 +486,10 @@ class Index(BaseIndex): header_index=frag['start_section'], header_span=position - frag['start_section'] + 1, fragment_anchor=fid, - content=fix_format(frag['content']), + text=fix_format(frag['text']), themes=frag['themes']) #print '@ FRAG %s' % frag['content'] - self.index.addDocument(doc) + self.index.add(doc) # Collect content. @@ -608,143 +499,51 @@ class Index(BaseIndex): hdl(text) # in the end, add a section text. - doc = add_part(snippets, header_index=position, header_type=header.tag, - content=fix_format(content)) + doc = add_part(snippets, header_index=position, + header_type=header.tag, text=fix_format(content)) #print '@ CONTENT: %s' % fix_format(content) - self.index.addDocument(doc) + self.index.add(doc) finally: snippets.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 - - def open(self, analyzer=None, **kw): - if ReusableIndex.index: - self.index = ReusableIndex.index - else: - print("opening index") - Index.open(self, analyzer, **kw) - 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: - print("closing index") - ReusableIndex.index.optimize() - ReusableIndex.index.close() - ReusableIndex.index = None - - index_changed.send_robust(None) - - def close(self): - if ReusableIndex.index: - ReusableIndex.index.commit() - - -class JoinSearch(object): - """ - This mixin could be used to handle block join queries. - (currently unused) - """ - def __init__(self, *args, **kw): - super(JoinSearch, self).__init__(*args, **kw) - - 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 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) - class SearchResult(object): - def __init__(self, search, scoreDocs, score=None, how_found=None, snippets=None, searched=None, tokens_cache=None): + def __init__(self, search, doc, how_found=None, snippets=None, searched=None, tokens_cache=None): if tokens_cache is None: tokens_cache = {} - if score: - self._score = score + if 'score' in doc: + self._score = doc['score'] else: - self._score = scoreDocs.score + self._score = 0 self.boost = 1.0 self._hits = [] self._processed_hits = None # processed hits - stored = search.searcher.doc(scoreDocs.doc) - self.book_id = int(stored.get("book_id")) + self.book_id = int(doc["book_id"]) - pd = stored.get("published_date") + pd = doc["published_date"] try: self.published_date = int(pd) except ValueError: self.published_date = 0 - header_type = stored.get("header_type") + 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(stored.get("header_index"))) - header_span = stored.get('header_span') + 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 = stored.get("fragment_anchor") + fragment = doc.get("fragment_anchor", None) if snippets: snippets = snippets.replace("/\n", "\n") - hit = (sec + (header_span,), fragment, scoreDocs.score, {'how_found': how_found, 'snippets': snippets and [snippets] or []}) + hit = (sec + (header_span,), fragment, self._score, {'how_found': how_found, 'snippets': snippets and [snippets] or []}) self._hits.append(hit) @@ -808,7 +607,7 @@ class SearchResult(object): # remove fragments with duplicated fid's and duplicated snippets frags = remove_duplicates(frags, lambda f: f[FRAGMENT], lambda a, b: cmp(a[SCORE], b[SCORE])) - frags = remove_duplicates(frags, lambda f: f[OTHER]['snippets'] and f[OTHER]['snippets'][0] or hash(f), + frags = remove_duplicates(frags, lambda f: f[OTHER]['snippets'] and f[OTHER]['snippets'][0] or f[FRAGMENT], lambda a, b: cmp(a[SCORE], b[SCORE])) # remove duplicate sections @@ -874,7 +673,6 @@ 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() @@ -983,11 +781,11 @@ class Hint(object): return some -class Search(IndexStore): +class Search(SolrIndex): """ Search facilities. """ - def __init__(self, default_field="content"): + def __init__(self, default_field="text"): IndexStore.__init__(self) self.analyzer = WLAnalyzer() # PolishAnalyzer(Version.LUCENE_34) # self.analyzer = WLAnalyzer() @@ -1010,9 +808,8 @@ class Search(IndexStore): def reopen(self, **unused): reader = self.searcher.getIndexReader() rdr = reader.reopen() - print "got signal to reopen index" if not rdr.equals(reader): - print "will reopen index" + log.debug('Reopening index') oldsearch = self.searcher self.searcher = IndexSearcher(rdr) oldsearch.close() @@ -1035,7 +832,7 @@ class Search(IndexStore): bks.append(catalogue.models.Book.objects.get(id=doc.get("book_id"))) return (bks, tops.totalHits) - def get_tokens(self, searched, field='content', cached=None): + 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) @@ -1060,7 +857,8 @@ class Search(IndexStore): return toks - def fuzziness(self, fuzzy): + @staticmethod + def fuzziness(fuzzy): """Helper method to sanitize fuzziness""" if not fuzzy: return None @@ -1069,7 +867,7 @@ class Search(IndexStore): else: return 0.5 - def make_phrase(self, tokens, field='content', slop=2, fuzzy=False): + def make_phrase(self, tokens, field='text', slop=2, fuzzy=False): """ Return a PhraseQuery with a series of tokens. """ @@ -1081,7 +879,6 @@ class Search(IndexStore): fuzzterms = [] while True: - # print("fuzz %s" % unicode(fuzzterm.term()).encode('utf-8')) ft = fuzzterm.term() if ft: fuzzterms.append(ft) @@ -1098,7 +895,8 @@ class Search(IndexStore): phrase.add(term) return phrase - def make_term_query(self, tokens, field='content', modal=BooleanClause.Occur.SHOULD, fuzzy=False): + @staticmethod + def make_term_query(tokens, field='text', modal='BooleanClause.Occur.SHOULD XXX', fuzzy=False): """ Returns term queries joined by boolean query. modal - applies to boolean query @@ -1205,7 +1003,7 @@ class Search(IndexStore): 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 ['content']] + qrys = [self.make_phrase(self.get_tokens(searched), field=fld, fuzzy=fuzzy) for fld in ['text']] flt = None if hint: @@ -1238,7 +1036,7 @@ class Search(IndexStore): # content only query : themes x content q = BooleanQuery() - tokens_pl = self.get_tokens(searched, field='content', cached=tokens_cache) + tokens_pl = self.get_tokens(searched, field='text', cached=tokens_cache) tokens = self.get_tokens(searched, field='SIMPLE', cached=tokens_cache) # only search in themes when we do not already filter by themes @@ -1246,20 +1044,19 @@ class Search(IndexStore): q.add(BooleanClause(self.make_term_query(tokens_pl, field='themes_pl', fuzzy=fuzzy), BooleanClause.Occur.MUST)) - q.add(BooleanClause(self.make_term_query(tokens_pl, field='content', + q.add(BooleanClause(self.make_term_query(tokens_pl, field='text', fuzzy=fuzzy), BooleanClause.Occur.SHOULD)) topDocs = self.searcher.search(q, only_in, max_results) for found in topDocs.scoreDocs: books.append(SearchResult(self, found, how_found='search_everywhere_themesXcontent', searched=searched)) - print "* %s theme x content: %s" % (searched, books[-1]._hits) # query themes/content x author/title/tags q = BooleanQuery() in_content = BooleanQuery() in_meta = BooleanQuery() - for fld in ['themes_pl', 'content']: + for fld in ['themes_pl', 'text']: in_content.add(BooleanClause(self.make_term_query(tokens_pl, field=fld, fuzzy=False), BooleanClause.Occur.SHOULD)) for fld in ['tags', 'authors', 'title']: @@ -1271,7 +1068,6 @@ class Search(IndexStore): topDocs = self.searcher.search(q, only_in, max_results) for found in topDocs.scoreDocs: books.append(SearchResult(self, found, how_found='search_everywhere', searched=searched)) - print "* %s scatter search: %s" % (searched, books[-1]._hits) return books @@ -1317,7 +1113,7 @@ class Search(IndexStore): # return None - def get_snippets(self, scoreDoc, query, field='content'): + def get_snippets(self, scoreDoc, query, field='text'): """ Returns a snippet for found scoreDoc. """ @@ -1332,9 +1128,17 @@ class Search(IndexStore): return None revision = stored.get('snippets_revision') if revision: revision = int(revision) + # locate content. book_id = int(stored.get('book_id')) - snippets = Snippets(book_id, revision=revision).open() + snippets = Snippets(book_id, revision=revision) + + try: + snippets.open() + except IOError, e: + log.error("Cannot open snippet file for book id = %d [rev=%d], %s" % (book_id, revision, e)) + return [] + try: try: text = snippets.get((int(position), @@ -1371,13 +1175,13 @@ class Search(IndexStore): if terms: return JArray('object')(terms, Term) - def search_tags(self, query, filters=None, max_results=40, pdcounter=False): + def search_tags(self, query, filt=None, max_results=40, pdcounter=False): """ Search for Tag objects using query. """ if not pdcounter: - filters = self.chain_filters([filter, self.term_filter(Term('is_pdcounter', 'true'), inverse=True)]) - tops = self.searcher.search(query, filters, max_results) + filters = self.chain_filters([filt, self.term_filter(Term('is_pdcounter', 'true'), inverse=True)]) + tops = self.searcher.search(query, filt, max_results) tags = [] for found in tops.scoreDocs: @@ -1396,22 +1200,23 @@ class Search(IndexStore): else: tag = catalogue.models.Tag.objects.get(id=doc.get("tag_id")) # don't add the pdcounter tag if same tag already exists - if not (is_pdcounter and filter(lambda t: tag.slug == t.slug, tags)): - tags.append(tag) + + tags.append(tag) + except catalogue.models.Tag.DoesNotExist: pass except PDCounterAuthor.DoesNotExist: pass except PDCounterBook.DoesNotExist: pass - # print "%s (%d) -> %f" % (tag, tag.id, found.score) - print 'returning %s' % tags + log.debug('search_tags: %s' % tags) + return tags - def search_books(self, query, filter=None, max_results=10): + def search_books(self, query, filt=None, max_results=10): """ Searches for Book objects using query """ bks = [] - tops = self.searcher.search(query, filter, max_results) + tops = self.searcher.search(query, filt, max_results) for found in tops.scoreDocs: doc = self.searcher.doc(found.doc) try: @@ -1480,7 +1285,7 @@ class Search(IndexStore): return self.search_books(q, self.term_filter(Term("is_book", "true")), max_results=max_results) @staticmethod - def chain_filters(filters, op=ChainedFilter.AND): + def chain_filters(filters, op='XXXChainedFilter.AND'): """ Chains a filter list together """