def metadata_from_xmp_packet(raw_bytes): root = parse_xmp_packet(raw_bytes) mi = Metadata(_('Unknown')) title = first_alt('//dc:title', root) if title: if title.startswith(r'\376\377'): # corrupted XMP packet generated by Nitro PDF. See # https://bugs.launchpad.net/calibre/+bug/1541981 raise ValueError( 'Corrupted XMP metadata packet detected, probably generated by Nitro PDF' ) mi.title = title authors = multiple_sequences('//dc:creator', root) if authors: mi.authors = [au for aus in authors for au in string_to_authors(aus)] tags = multiple_sequences('//dc:subject', root) or multiple_sequences( '//pdf:Keywords', root) if tags: mi.tags = tags comments = first_alt('//dc:description', root) if comments: mi.comments = comments publishers = multiple_sequences('//dc:publisher', root) if publishers: mi.publisher = publishers[0] try: pubdate = parse_date(first_sequence('//dc:date', root) or first_simple('//xmp:CreateDate', root), assume_utc=False) except: pass else: mi.pubdate = pubdate bkp = first_simple('//xmp:CreatorTool', root) if bkp: mi.book_producer = bkp md = safe_parse_date(first_simple('//xmp:MetadataDate', root)) mod = safe_parse_date(first_simple('//xmp:ModifyDate', root)) fd = more_recent(md, mod) if fd is not None: mi.metadata_date = fd rating = first_simple('//calibre:rating', root) if rating is not None: try: rating = float(rating) if 0 <= rating <= 10: mi.rating = rating except (ValueError, TypeError): pass series, series_index = read_series(root) if series: mi.series, mi.series_index = series, series_index for x in ('title_sort', 'author_sort'): for elem in XPath('//calibre:' + x)(root): val = read_simple_property(elem) if val: setattr(mi, x, val) break for x in ('author_link_map', 'user_categories'): val = first_simple('//calibre:' + x, root) if val: try: setattr(mi, x, json.loads(val)) except: pass languages = multiple_sequences('//dc:language', root) if languages: languages = list(filter(None, map(canonicalize_lang, languages))) if languages: mi.languages = languages identifiers = {} for xmpid in XPath('//xmp:Identifier')(root): for scheme, value in read_xmp_identifers(xmpid): if scheme and value: identifiers[scheme.lower()] = value for namespace in ('prism', 'pdfx'): for scheme in KNOWN_ID_SCHEMES: if scheme not in identifiers: val = first_simple('//%s:%s' % (namespace, scheme), root) scheme = scheme.lower() if scheme == 'isbn': val = check_isbn(val) elif scheme == 'doi': val = check_doi(val) if val: identifiers[scheme] = val # Check Dublin Core for recognizable identifier types for scheme, check_func in iteritems({ 'doi': check_doi, 'isbn': check_isbn }): if scheme not in identifiers: val = check_func(first_simple('//dc:identifier', root)) if val: identifiers['doi'] = val if identifiers: mi.set_identifiers(identifiers) read_user_metadata(mi, root) return mi
def metadata_from_xmp_packet(raw_bytes): root = parse_xmp_packet(raw_bytes) mi = Metadata(_("Unknown")) title = first_alt("//dc:title", root) if title: mi.title = title authors = multiple_sequences("//dc:creator", root) if authors: mi.authors = authors tags = multiple_sequences("//dc:subject", root) or multiple_sequences("//pdf:Keywords", root) if tags: mi.tags = tags comments = first_alt("//dc:description", root) if comments: mi.comments = comments publishers = multiple_sequences("//dc:publisher", root) if publishers: mi.publisher = publishers[0] try: pubdate = parse_date( first_sequence("//dc:date", root) or first_simple("//xmp:CreateDate", root), assume_utc=False ) except: pass else: mi.pubdate = pubdate bkp = first_simple("//xmp:CreatorTool", root) if bkp: mi.book_producer = bkp md = first_simple("//xmp:MetadataDate", root) if md: try: mi.metadata_date = parse_date(md) except: pass rating = first_simple("//calibre:rating", root) if rating is not None: try: rating = float(rating) if 0 <= rating <= 10: mi.rating = rating except (ValueError, TypeError): pass series, series_index = read_series(root) if series: mi.series, mi.series_index = series, series_index for x in ("title_sort", "author_sort"): for elem in XPath("//calibre:" + x)(root): val = read_simple_property(elem) if val: setattr(mi, x, val) break for x in ("author_link_map", "user_categories"): val = first_simple("//calibre:" + x, root) if val: try: setattr(mi, x, json.loads(val)) except: pass languages = multiple_sequences("//dc:language", root) if languages: languages = filter(None, map(canonicalize_lang, languages)) if languages: mi.languages = languages identifiers = {} for xmpid in XPath("//xmp:Identifier")(root): for scheme, value in read_xmp_identifers(xmpid): if scheme and value: identifiers[scheme.lower()] = value for namespace in ("prism", "pdfx"): for scheme in KNOWN_ID_SCHEMES: if scheme not in identifiers: val = first_simple("//%s:%s" % (namespace, scheme), root) scheme = scheme.lower() if scheme == "isbn": val = check_isbn(val) elif scheme == "doi": val = check_doi(val) if val: identifiers[scheme] = val # Check Dublin Core for recognizable identifier types for scheme, check_func in {"doi": check_doi, "isbn": check_isbn}.iteritems(): if scheme not in identifiers: val = check_func(first_simple("//dc:identifier", root)) if val: identifiers["doi"] = val if identifiers: mi.set_identifiers(identifiers) read_user_metadata(mi, root) return mi
def metadata_from_xmp_packet(raw_bytes): root = parse_xmp_packet(raw_bytes) mi = Metadata(_('Unknown')) title = first_alt('//dc:title', root) if title.startswith(r'\376\377'): # corrupted XMP packet generated by Nitro PDF. See # https://bugs.launchpad.net/calibre/+bug/1541981 raise ValueError('Corrupted XMP metadata packet detected, probably generated by Nitro PDF') if title: mi.title = title authors = multiple_sequences('//dc:creator', root) if authors: mi.authors = authors tags = multiple_sequences('//dc:subject', root) or multiple_sequences('//pdf:Keywords', root) if tags: mi.tags = tags comments = first_alt('//dc:description', root) if comments: mi.comments = comments publishers = multiple_sequences('//dc:publisher', root) if publishers: mi.publisher = publishers[0] try: pubdate = parse_date(first_sequence('//dc:date', root) or first_simple('//xmp:CreateDate', root), assume_utc=False) except: pass else: mi.pubdate = pubdate bkp = first_simple('//xmp:CreatorTool', root) if bkp: mi.book_producer = bkp md = safe_parse_date(first_simple('//xmp:MetadataDate', root)) mod = safe_parse_date(first_simple('//xmp:ModifyDate', root)) fd = more_recent(md, mod) if fd is not None: mi.metadata_date = fd rating = first_simple('//calibre:rating', root) if rating is not None: try: rating = float(rating) if 0 <= rating <= 10: mi.rating = rating except (ValueError, TypeError): pass series, series_index = read_series(root) if series: mi.series, mi.series_index = series, series_index for x in ('title_sort', 'author_sort'): for elem in XPath('//calibre:' + x)(root): val = read_simple_property(elem) if val: setattr(mi, x, val) break for x in ('author_link_map', 'user_categories'): val = first_simple('//calibre:'+x, root) if val: try: setattr(mi, x, json.loads(val)) except: pass languages = multiple_sequences('//dc:language', root) if languages: languages = filter(None, map(canonicalize_lang, languages)) if languages: mi.languages = languages identifiers = {} for xmpid in XPath('//xmp:Identifier')(root): for scheme, value in read_xmp_identifers(xmpid): if scheme and value: identifiers[scheme.lower()] = value for namespace in ('prism', 'pdfx'): for scheme in KNOWN_ID_SCHEMES: if scheme not in identifiers: val = first_simple('//%s:%s' % (namespace, scheme), root) scheme = scheme.lower() if scheme == 'isbn': val = check_isbn(val) elif scheme == 'doi': val = check_doi(val) if val: identifiers[scheme] = val # Check Dublin Core for recognizable identifier types for scheme, check_func in {'doi':check_doi, 'isbn':check_isbn}.iteritems(): if scheme not in identifiers: val = check_func(first_simple('//dc:identifier', root)) if val: identifiers['doi'] = val if identifiers: mi.set_identifiers(identifiers) read_user_metadata(mi, root) return mi
def metadata_from_xmp_packet(raw_bytes): root = parse_xmp_packet(raw_bytes) mi = Metadata(_('Unknown')) title = first_alt('//dc:title', root) if title: mi.title = title authors = multiple_sequences('//dc:creator', root) if authors: mi.authors = authors tags = multiple_sequences('//dc:subject', root) or multiple_sequences( '//pdf:Keywords', root) if tags: mi.tags = tags comments = first_alt('//dc:description', root) if comments: mi.comments = comments publishers = multiple_sequences('//dc:publisher', root) if publishers: mi.publisher = publishers[0] try: pubdate = parse_date(first_sequence('//dc:date', root) or first_simple('//xmp:CreateDate', root), assume_utc=False) except: pass else: mi.pubdate = pubdate bkp = first_simple('//xmp:CreatorTool', root) if bkp: mi.book_producer = bkp md = first_simple('//xmp:MetadataDate', root) if md: try: mi.metadata_date = parse_date(md) except: pass rating = first_simple('//calibre:rating', root) if rating is not None: try: rating = float(rating) if 0 <= rating <= 10: mi.rating = rating except (ValueError, TypeError): pass series, series_index = read_series(root) if series: mi.series, mi.series_index = series, series_index for x in ('title_sort', 'author_sort'): for elem in XPath('//calibre:' + x)(root): val = read_simple_property(elem) if val: setattr(mi, x, val) break for x in ('author_link_map', 'user_categories'): val = first_simple('//calibre:' + x, root) if val: try: setattr(mi, x, json.loads(val)) except: pass languages = multiple_sequences('//dc:language', root) if languages: languages = filter(None, map(canonicalize_lang, languages)) if languages: mi.languages = languages identifiers = {} for xmpid in XPath('//xmp:Identifier')(root): for scheme, value in read_xmp_identifers(xmpid): if scheme and value: identifiers[scheme.lower()] = value for namespace in ('prism', 'pdfx'): for scheme in KNOWN_ID_SCHEMES: if scheme not in identifiers: val = first_simple('//%s:%s' % (namespace, scheme), root) scheme = scheme.lower() if scheme == 'isbn': val = check_isbn(val) elif scheme == 'doi': val = check_doi(val) if val: identifiers[scheme] = val # Check Dublin Core for recognizable identifier types for scheme, check_func in { 'doi': check_doi, 'isbn': check_isbn }.iteritems(): if scheme not in identifiers: val = check_func(first_simple('//dc:identifier', root)) if val: identifiers['doi'] = val if identifiers: mi.set_identifiers(identifiers) read_user_metadata(mi, root) return mi