X-Git-Url: https://git.mdrn.pl/librarian.git/blobdiff_plain/a34b95aa7ba5fd4838541d1cdcd28358fb808062..5b1dcc7d247996752fa566c7150a45037b068565:/librarian/dcparser.py diff --git a/librarian/dcparser.py b/librarian/dcparser.py index 21244ef..7418f70 100644 --- a/librarian/dcparser.py +++ b/librarian/dcparser.py @@ -6,6 +6,8 @@ from xml.parsers.expat import ExpatError from datetime import date import time +import re +from librarian.util import roman_to_int from librarian import (ValidationError, NoDublinCore, ParseError, DCNS, RDFNS, WLURI) @@ -62,15 +64,30 @@ class Person(object): def as_date(text): try: - try: - t = time.strptime(text, '%Y-%m-%d') - except ValueError: - t = time.strptime(text, '%Y') + # check out the "N. poł X w." syntax + if isinstance(text, str): text = text.decode("utf-8") + m = re.match(u"(?:([12]) *poł[.]? )?([MCDXVI]+) *w[.]?", text) + if m: + + half = m.groups()[0] + if half is not None: + half = int(half) + else: + half = 1 + century = roman_to_int(str(m.groups()[1])) + t = ((century*100 + (half-1)*50), 1, 1) + else: + text = re.sub(r"(po|ok[.]?) *", "", text) + try: + t = time.strptime(text, '%Y-%m-%d') + except ValueError: + t = time.strptime(re.split(r'[-/]', text)[0], '%Y') return date(t[0], t[1], t[2]) except ValueError, e: raise ValueError("Unrecognized date format. Try YYYY-MM-DD or YYYY.") -as_person = Person.from_text +def as_person(text): + return Person.from_text(text) def as_unicode(text): if isinstance(text, unicode): @@ -78,12 +95,15 @@ def as_unicode(text): else: return text.decode('utf-8') +def as_wluri_strict(text): + return WLURI.strict(text) + class Field(object): def __init__(self, uri, attr_name, validator=as_unicode, strict=None, multiple=False, salias=None, **kwargs): self.uri = uri self.name = attr_name - self.validator = lambda x: validator(x) - self.strict = lambda x: strict(x) + self.validator = validator + self.strict = strict self.multiple = multiple self.salias = salias @@ -111,10 +131,21 @@ class Field(object): except ValueError, e: raise ValidationError("Field '%s' - invald value: %s" % (self.uri, e.message)) - def validate(self, fdict, strict=False): + def validate(self, fdict, fallbacks=None, strict=False): + if fallbacks is None: + fallbacks = {} if not fdict.has_key(self.uri): if not self.required: - f = self.default + # Accept single value for single fields and saliases. + if self.name in fallbacks: + if self.multiple: + f = fallbacks[self.name] + else: + f = [fallbacks[self.name]] + elif self.salias and self.salias in fallbacks: + f = [fallbacks[self.salias]] + else: + f = self.default else: raise ValidationError("Required field %s not found" % self.uri) else: @@ -156,6 +187,9 @@ class WorkInfo(object): as_person, salias='editor', multiple=True, default=[]), Field( DCNS('contributor.technical_editor'), 'technical_editors', as_person, salias='technical_editor', multiple=True, default=[]), + Field( DCNS('contributor.funding'), 'funders', + salias='funder', multiple=True, default=[]), + Field( DCNS('contributor.thanks'), 'thanks', required=False), Field( DCNS('date'), 'created_at', as_date), Field( DCNS('date.pd'), 'released_to_public_domain_at', as_date, required=False), @@ -166,7 +200,7 @@ class WorkInfo(object): Field( DCNS('source'), 'source_name', required=False), Field( DCNS('source.URL'), 'source_url', required=False), - Field( DCNS('identifier.url'), 'url', WLURI, strict=WLURI.strict), + Field( DCNS('identifier.url'), 'url', WLURI, strict=as_wluri_strict), Field( DCNS('rights.license'), 'license', required=False), Field( DCNS('rights'), 'license_description'), ) @@ -220,7 +254,7 @@ class WorkInfo(object): return cls(desc.attrib, field_dict, *args, **kwargs) - def __init__(self, rdf_attrs, dc_fields, strict=False): + def __init__(self, rdf_attrs, dc_fields, fallbacks=None, strict=False): """rdf_attrs should be a dictionary-like object with any attributes of the RDF:Description. dc_fields - dictionary mapping DC fields (with namespace) to list of text values for the given field. """ @@ -229,7 +263,8 @@ class WorkInfo(object): self.fmap = {} for field in self.FIELDS: - value = field.validate(dc_fields, strict=strict) + value = field.validate(dc_fields, fallbacks=fallbacks, + strict=strict) setattr(self, 'prop_' + field.name, value) self.fmap[field.name] = field if field.salias: self.fmap[field.salias] = field @@ -244,7 +279,7 @@ class WorkInfo(object): if not field.multiple: raise "OUCH!! for field %s" % name - return value[0] + return value[0] if value else None except (KeyError, AttributeError): return object.__getattribute__(self, name) @@ -343,20 +378,23 @@ class BookInfo(WorkInfo): Field( DCNS('audience'), 'audiences', salias='audience', multiple=True, required=False), - Field( DCNS('subject.period'), 'epochs', salias='epoch', multiple=True), - Field( DCNS('subject.type'), 'kinds', salias='kind', multiple=True), - Field( DCNS('subject.genre'), 'genres', salias='genre', multiple=True), + Field( DCNS('subject.period'), 'epochs', salias='epoch', multiple=True, + required=False), + Field( DCNS('subject.type'), 'kinds', salias='kind', multiple=True, + required=False), + Field( DCNS('subject.genre'), 'genres', salias='genre', multiple=True, + required=False), Field( DCNS('contributor.translator'), 'translators', \ as_person, salias='translator', multiple=True, default=[]), Field( DCNS('relation.hasPart'), 'parts', - WLURI, strict=WLURI.strict, multiple=True, required=False), + WLURI, strict=as_wluri_strict, multiple=True, required=False), Field( DCNS('relation.isVariantOf'), 'variant_of', - WLURI, strict=WLURI.strict, required=False), + WLURI, strict=as_wluri_strict, required=False), - Field( DCNS('relation.cover_image.url'), 'cover_url', required=False), - Field( DCNS('relation.cover_image.attribution'), 'cover_by', required=False), - Field( DCNS('relation.cover_image.source'), 'cover_source', required=False), + Field( DCNS('relation.coverImage.url'), 'cover_url', required=False), + Field( DCNS('relation.coverImage.attribution'), 'cover_by', required=False), + Field( DCNS('relation.coverImage.source'), 'cover_source', required=False), )