X-Git-Url: https://git.mdrn.pl/wolnelektury.git/blobdiff_plain/57e98a5807465f9494e5006e16860f16abbf4136..480da6ff4ed8c04b205710c318b2fcfcb961bbc7:/apps/search/index.py diff --git a/apps/search/index.py b/apps/search/index.py index 0f69d0725..e7f28c9da 100644 --- a/apps/search/index.py +++ b/apps/search/index.py @@ -1,25 +1,7 @@ # -*- 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, 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 @@ -27,79 +9,18 @@ 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 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 = 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 = highlight.HLSolrInterface(settings.SOLR, mode=mode) class Snippets(object): @@ -110,7 +31,7 @@ class Snippets(object): """ 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: @@ -118,15 +39,32 @@ 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 @@ -155,85 +93,137 @@ class Snippets(object): """Close snippet file""" self.file.close() - -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, analyzer=None): - if self.index: - raise Exception("Index is already opened") - self.index = IndexWriter(self.store, self.analyzer,\ - IndexWriter.MaxFieldLength.LIMITED) - return self.index - - def optimize(self): - self.index.optimize() - - def close(self): + def remove(self): + self.revision = None try: - self.index.optimize() - except JavaError, je: - print "Error during optimize phase, check index: %s" % je - - self.index.close() - self.index = None - - def __enter__(self): - self.open() - return self - - def __exit__(self, type, value, tb): - self.close() + os.unlink(self.path) + self.revision = 0 + while True: + self.revision += 1 + os.unlink(self.path) + except OSError: + pass -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): + 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 """ - 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(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) + 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 + } + 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 + } + else: + 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): + 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.remove() def index_book(self, book, book_info=None, overwrite=True): """ @@ -242,21 +232,27 @@ class Index(BaseIndex): 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, 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) + 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'] + 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', @@ -280,7 +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 """ @@ -289,12 +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 @@ -304,40 +302,38 @@ 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 - source = book_info.source_name - match = self.published_date_re.search(source) - print("published date is %s %s" % (match, match is not None and match.groups())) - if match is not None: - fields["published_date"] = Field("published_date", str(match.groups()[0]), Field.Store.YES, Field.Index.NOT_ANALYZED) + 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 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): """ @@ -347,7 +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. @@ -360,11 +356,18 @@ class Index(BaseIndex): return [] def walker(node, ignore_tags=[]): - yield node, None - for child in filter(lambda n: n.tag not in ignore_tags, list(node)): - for b, e in walker(child): - yield b, e - yield None, node + + 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): @@ -383,39 +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'] - 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): @@ -436,35 +431,51 @@ class Index(BaseIndex): # section content content = [] - footnote = None + footnote = [] + + def all_content(text): + for frag in fragments.values(): + frag['text'].append(text) + content.append(text) + handle_text = [all_content] - for start, end in walker(header, ignore_tags=self.ignore_content_tags): + 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 = ' '.join(start.itertext()) - # elif end is not None and footnote is not None and end.tag in self.footnote_tags: - # doc = add_part(snippets, header_index=position, header_type=header.tag, - # content=footnote) + if start is not None and start.tag in self.footnote_tags: + footnote = [] - # self.index.addDocument(doc) + def collect_footnote(t): + footnote.append(t) - # footnote = None + 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) + + 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() 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 - # import pdb; pdb.set_trace() frag = fragments[fid] if frag['themes'] == []: continue # empty themes list. @@ -475,157 +486,66 @@ 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']) - - self.index.addDocument(doc) + #print '@ FRAG %s' % frag['content'] + self.index.add(doc) # Collect content. - elif start is not None: - for frag in fragments.values(): - frag['content'].append(start.text) - content.append(start.text) - elif end is not None: - for frag in fragments.values(): - frag['content'].append(end.tail) - content.append(end.tail) + + if text is not None and handle_text is not []: + hdl = handle_text[-1] + if hdl is not None: + 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, threads=4): - if ReusableIndex.index is not None: - self.index = ReusableIndex.index - else: - print("opening index") - 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: - ReusableIndex.index.optimize() - ReusableIndex.index.close() - ReusableIndex.index = None - - def close(self): - pass - - -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"]) - header_type = stored.get("header_type") - if not header_type: - return - - 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 + pd = doc["published_date"] + try: + self.published_date = int(pd) + except ValueError: + self.published_date = 0 - fragment = stored.get("fragment_anchor") + 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 - pd = stored.get("published_date") - if pd is None: - print "published_date is none for book %d" % self.book_id - pd = 0 - self.published_date = int(pd) + 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 []}) + if snippets: + snippets = snippets.replace("/\n", "\n") + hit = (sec + (header_span,), fragment, self._score, {'how_found': how_found, 'snippets': snippets and [snippets] or []}) - self._hits.append(hit) + self._hits.append(hit) self.search = search self.searched = searched @@ -644,6 +564,8 @@ class SearchResult(object): return self def get_book(self): + if hasattr(self, '_book'): + return self._book return catalogue.models.Book.objects.get(id=self.book_id) book = property(get_book) @@ -662,7 +584,10 @@ class SearchResult(object): # to sections and fragments frags = filter(lambda r: r[FRAGMENT] is not None, self._hits) + sect = filter(lambda r: r[FRAGMENT] is None, self._hits) + + # sections not covered by fragments sect = filter(lambda s: 0 == len(filter( lambda f: s[POSITION][POSITION_INDEX] >= f[POSITION][POSITION_INDEX] and s[POSITION][POSITION_INDEX] < f[POSITION][POSITION_INDEX] + f[POSITION][POSITION_SPAN], @@ -670,15 +595,20 @@ class SearchResult(object): hits = [] - # remove duplicate fragments - fragments = {} - for f in frags: - fid = f[FRAGMENT] - if fid in fragments: - if fragments[fid][SCORE] >= f[SCORE]: - continue - fragments[fid] = f - frags = fragments.values() + 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[FRAGMENT], lambda a, b: cmp(a[SCORE], b[SCORE])) + 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 sections = {} @@ -700,7 +630,7 @@ class SearchResult(object): for f in frags: try: - frag = catalogue.models.Fragment.objects.get(anchor=f[FRAGMENT]) + frag = catalogue.models.Fragment.objects.get(anchor=f[FRAGMENT], book__id=self.book_id) except catalogue.models.Fragment.DoesNotExist: # stale index continue @@ -743,7 +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() @@ -852,20 +781,39 @@ 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() - self.searcher = IndexSearcher(self.store, True) + reader = IndexReader.open(self.store, True) + self.searcher = IndexSearcher(reader) self.parser = QueryParser(Version.LUCENE_34, default_field, self.analyzer) self.parent_filter = TermsFilter() self.parent_filter.addTerm(Term("is_book", "true")) + index_changed.connect(self.reopen) + + def close(self): + reader = self.searcher.getIndexReader() + self.searcher.close() + reader.close() + super(Search, self).close() + index_changed.disconnect(self.reopen) + + def reopen(self, **unused): + reader = self.searcher.getIndexReader() + rdr = reader.reopen() + if not rdr.equals(reader): + log.debug('Reopening index') + oldsearch = self.searcher + self.searcher = IndexSearcher(rdr) + oldsearch.close() + reader.close() def query(self, query): """Parse query in default Lucene Syntax. (for humans) @@ -884,7 +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) @@ -909,7 +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 @@ -918,7 +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. """ @@ -930,7 +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) @@ -947,7 +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 @@ -1054,7 +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: @@ -1087,7 +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 @@ -1095,20 +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']: @@ -1120,7 +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 @@ -1166,7 +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. """ @@ -1179,18 +1126,36 @@ class Search(IndexStore): length = stored.get('snippets_length') if position is None or length is None: return None + revision = stored.get('snippets_revision') + if revision: revision = int(revision) + # locate content. - snippets = Snippets(stored.get('book_id')).open() + book_id = int(stored.get('book_id')) + snippets = Snippets(book_id, revision=revision) + try: - text = snippets.get((int(position), - int(length))) - finally: - snippets.close() + 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), + int(length))) + finally: + snippets.close() - tokenStream = TokenSources.getAnyTokenStream(self.searcher.getIndexReader(), scoreDoc.doc, field, self.analyzer) - # highlighter.getBestTextFragments(tokenStream, text, False, 10) - snip = highlighter.getBestFragments(tokenStream, text, 3, "...") + tokenStream = TokenSources.getAnyTokenStream(self.searcher.getIndexReader(), scoreDoc.doc, field, self.analyzer) + # highlighter.getBestTextFragments(tokenStream, text, False, 10) + snip = highlighter.getBestFragments(tokenStream, text, 3, "...") + except Exception, e: + e2 = e + if hasattr(e, 'getJavaException'): + e2 = unicode(e.getJavaException()) + raise Exception("Problem fetching snippets for book %d, @%d len=%d" % (book_id, int(position), int(length)), + e2) return snip @staticmethod @@ -1210,33 +1175,56 @@ class Search(IndexStore): if terms: return JArray('object')(terms, Term) - def search_tags(self, query, filter=None, max_results=40): + def search_tags(self, query, filt=None, max_results=40, pdcounter=False): """ Search for Tag objects using query. """ - tops = self.searcher.search(query, filter, max_results) + if not pdcounter: + 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: 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) + is_pdcounter = doc.get('is_pdcounter') + 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: + tag = catalogue.models.Tag.objects.get(id=doc.get("tag_id")) + # don't add the pdcounter tag if same tag already exists + + tags.append(tag) + + except catalogue.models.Tag.DoesNotExist: pass + except PDCounterAuthor.DoesNotExist: pass + except PDCounterBook.DoesNotExist: pass + + 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) - bks.append(catalogue.models.Book.objects.get(id=doc.get("book_id"))) + try: + bks.append(catalogue.models.Book.objects.get(id=doc.get("book_id"))) + except catalogue.models.Book.DoesNotExist: pass return bks - def create_prefix_phrase(self, toks, field): + def make_prefix_phrase(self, toks, field): q = MultiPhraseQuery() for i in range(len(toks)): t = Term(field, toks[i]) @@ -1262,7 +1250,7 @@ class Search(IndexStore): return only_term - def hint_tags(self, string, max_results=50): + def hint_tags(self, string, max_results=50, pdcounter=True, prefix=True, fuzzy=False): """ Return auto-complete hints for tags using prefix search. @@ -1271,14 +1259,17 @@ class Search(IndexStore): top = BooleanQuery() for field in ['tag_name', 'tag_name_pl']: - q = self.create_prefix_phrase(toks, field) + if prefix: + q = self.make_prefix_phrase(toks, field) + else: + q = self.make_term_query(toks, field, fuzzy=fuzzy) top.add(BooleanClause(q, BooleanClause.Occur.SHOULD)) no_book_cat = self.term_filter(Term("tag_category", "book"), inverse=True) - return self.search_tags(top, no_book_cat, max_results=max_results) + return self.search_tags(top, no_book_cat, max_results=max_results, pdcounter=pdcounter) - def hint_books(self, string, max_results=50): + def hint_books(self, string, max_results=50, prefix=True, fuzzy=False): """ Returns auto-complete hints for book titles Because we do not index 'pseudo' title-tags. @@ -1286,12 +1277,15 @@ class Search(IndexStore): """ toks = self.get_tokens(string, field='SIMPLE') - q = self.create_prefix_phrase(toks, 'title') + if prefix: + q = self.make_prefix_phrase(toks, 'title') + else: + q = self.make_term_query(toks, 'title', fuzzy=fuzzy) 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 """