"""Base class for entity objects manipulated in clients
:organization: Logilab
:copyright: 2001-2009 LOGILAB S.A. (Paris, FRANCE), all rights reserved.
:contact: http://www.logilab.fr/ -- mailto:contact@logilab.fr
"""
__docformat__ = "restructuredtext en"
from logilab.common import interface
from logilab.common.compat import all
from logilab.common.decorators import cached
from logilab.mtconverter import TransformData, TransformError, html_escape
from rql.utils import rqlvar_maker
from cubicweb import Unauthorized
from cubicweb.rset import ResultSet
from cubicweb.selectors import yes
from cubicweb.appobject import AppRsetObject
from cubicweb.schema import RQLVocabularyConstraint, RQLConstraint, bw_normalize_etype
try:
from cubicweb.common.uilib import printable_value, soup2xhtml
from cubicweb.common.mixins import MI_REL_TRIGGERS
from cubicweb.common.mttransforms import ENGINE
except ImportError:
# missing -common
MI_REL_TRIGGERS = {}
_marker = object()
def greater_card(rschema, subjtypes, objtypes, index):
for subjtype in subjtypes:
for objtype in objtypes:
card = rschema.rproperty(subjtype, objtype, 'cardinality')[index]
if card in '+*':
return card
return '1'
class RelationTags(object):
MODE_TAGS = frozenset(('link', 'create'))
CATEGORY_TAGS = frozenset(('primary', 'secondary', 'generic', 'generated',
'inlineview'))
def __init__(self, eclass, tagdefs):
self.eclass = eclass
self._tagdefs = {}
for relation, tags in tagdefs.iteritems():
# tags must become a set
if isinstance(tags, basestring):
tags = set((tags,))
elif not isinstance(tags, set):
tags = set(tags)
# relation must become a 3-uple (rtype, targettype, role)
if isinstance(relation, basestring):
self._tagdefs[(relation, '*', 'subject')] = tags
self._tagdefs[(relation, '*', 'object')] = tags
elif len(relation) == 1: # useful ?
self._tagdefs[(relation[0], '*', 'subject')] = tags
self._tagdefs[(relation[0], '*', 'object')] = tags
elif len(relation) == 2:
rtype, ttype = relation
ttype = bw_normalize_etype(ttype) # XXX bw compat
self._tagdefs[rtype, ttype, 'subject'] = tags
self._tagdefs[rtype, ttype, 'object'] = tags
elif len(relation) == 3:
relation = list(relation) # XXX bw compat
relation[1] = bw_normalize_etype(relation[1])
self._tagdefs[tuple(relation)] = tags
else:
raise ValueError('bad rtag definition (%r)' % (relation,))
def __initialize__(self):
# eclass.[*]schema are only set when registering
self.schema = self.eclass.schema
eschema = self.eschema = self.eclass.e_schema
rtags = self._tagdefs
# expand wildcards in rtags and add automatic tags
for rschema, tschemas, role in sorted(eschema.relation_definitions(True)):
rtype = rschema.type
star_tags = rtags.pop((rtype, '*', role), set())
for tschema in tschemas:
tags = rtags.setdefault((rtype, tschema.type, role), set(star_tags))
if role == 'subject':
X, Y = eschema, tschema
card = rschema.rproperty(X, Y, 'cardinality')[0]
composed = rschema.rproperty(X, Y, 'composite') == 'object'
else:
X, Y = tschema, eschema
card = rschema.rproperty(X, Y, 'cardinality')[1]
composed = rschema.rproperty(X, Y, 'composite') == 'subject'
# set default category tags if needed
if not tags & self.CATEGORY_TAGS:
if card in '1+':
if not rschema.is_final() and composed:
category = 'generated'
elif rschema.is_final() and (
rschema.type.endswith('_format')
or rschema.type.endswith('_encoding')):
category = 'generated'
else:
category = 'primary'
elif rschema.is_final():
if (rschema.type.endswith('_format')
or rschema.type.endswith('_encoding')):
category = 'generated'
else:
category = 'secondary'
else:
category = 'generic'
tags.add(category)
if not tags & self.MODE_TAGS:
if card in '?1':
# by default, suppose link mode if cardinality doesn't allow
# more than one relation
mode = 'link'
elif rschema.rproperty(X, Y, 'composite') == role:
# if self is composed of the target type, create mode
mode = 'create'
else:
# link mode by default
mode = 'link'
tags.add(mode)
def _default_target(self, rschema, role='subject'):
eschema = self.eschema
if role == 'subject':
return eschema.subject_relation(rschema).objects(eschema)[0]
else:
return eschema.object_relation(rschema).subjects(eschema)[0]
# dict compat
def __getitem__(self, key):
if isinstance(key, basestring):
key = (key,)
return self.get_tags(*key)
__contains__ = __getitem__
def get_tags(self, rtype, targettype=None, role='subject'):
rschema = self.schema.rschema(rtype)
if targettype is None:
tschema = self._default_target(rschema, role)
else:
tschema = self.schema.eschema(targettype)
return self._tagdefs[(rtype, tschema.type, role)]
__call__ = get_tags
def get_mode(self, rtype, targettype=None, role='subject'):
# XXX: should we make an assertion on rtype not being final ?
# assert not rschema.is_final()
tags = self.get_tags(rtype, targettype, role)
# do not change the intersection order !
modes = tags & self.MODE_TAGS
assert len(modes) == 1
return modes.pop()
def get_category(self, rtype, targettype=None, role='subject'):
tags = self.get_tags(rtype, targettype, role)
categories = tags & self.CATEGORY_TAGS
assert len(categories) == 1
return categories.pop()
def is_inlined(self, rtype, targettype=None, role='subject'):
# return set(('primary', 'secondary')) & self.get_tags(rtype, targettype)
return 'inlineview' in self.get_tags(rtype, targettype, role)
class metaentity(type):
"""this metaclass sets the relation tags on the entity class
and deals with the `widgets` attribute
"""
def __new__(mcs, name, bases, classdict):
# collect baseclass' rtags
tagdefs = {}
widgets = {}
for base in bases:
tagdefs.update(getattr(base, '__rtags__', {}))
widgets.update(getattr(base, 'widgets', {}))
# update with the class' own rtgas
tagdefs.update(classdict.get('__rtags__', {}))
widgets.update(classdict.get('widgets', {}))
# XXX decide whether or not it's a good idea to replace __rtags__
# good point: transparent support for inheritance levels >= 2
# bad point: we loose the information of which tags are specific
# to this entity class
classdict['__rtags__'] = tagdefs
classdict['widgets'] = widgets
eclass = super(metaentity, mcs).__new__(mcs, name, bases, classdict)
# adds the "rtags" attribute
eclass.rtags = RelationTags(eclass, tagdefs)
return eclass
class Entity(AppRsetObject, dict):
"""an entity instance has e_schema automagically set on
the class and instances has access to their issuing cursor.
A property is set for each attribute and relation on each entity's type
class. Becare that among attributes, 'eid' is *NEITHER* stored in the
dict containment (which acts as a cache for other attributes dynamically
fetched)
:type e_schema: `cubicweb.schema.EntitySchema`
:ivar e_schema: the entity's schema
:type rest_var: str
:cvar rest_var: indicates which attribute should be used to build REST urls
If None is specified, the first non-meta attribute will
be used
:type skip_copy_for: list
:cvar skip_copy_for: a list of relations that should be skipped when copying
this kind of entity. Note that some relations such
as composite relations or relations that have '?1' as object
cardinality
"""
__metaclass__ = metaentity
__registry__ = 'etypes'
__select__ = yes()
widgets = {}
id = None
e_schema = None
eid = None
rest_attr = None
skip_copy_for = ()
@classmethod
def registered(cls, registry):
"""build class using descriptor at registration time"""
assert cls.id is not None
super(Entity, cls).registered(registry)
if cls.id != 'Any':
cls.__initialize__()
return cls
MODE_TAGS = set(('link', 'create'))
CATEGORY_TAGS = set(('primary', 'secondary', 'generic', 'generated')) # , 'metadata'))
@classmethod
def __initialize__(cls):
"""initialize a specific entity class by adding descriptors to access
entity type's attributes and relations
"""
etype = cls.id
assert etype != 'Any', etype
cls.e_schema = eschema = cls.schema.eschema(etype)
for rschema, _ in eschema.attribute_definitions():
if rschema.type == 'eid':
continue
setattr(cls, rschema.type, Attribute(rschema.type))
mixins = []
for rschema, _, x in eschema.relation_definitions():
if (rschema, x) in MI_REL_TRIGGERS:
mixin = MI_REL_TRIGGERS[(rschema, x)]
if not (issubclass(cls, mixin) or mixin in mixins): # already mixed ?
mixins.append(mixin)
for iface in getattr(mixin, '__implements__', ()):
if not interface.implements(cls, iface):
interface.extend(cls, iface)
if x == 'subject':
setattr(cls, rschema.type, SubjectRelation(rschema))
else:
attr = 'reverse_%s' % rschema.type
setattr(cls, attr, ObjectRelation(rschema))
if mixins:
cls.__bases__ = tuple(mixins + [p for p in cls.__bases__ if not p is object])
cls.debug('plugged %s mixins on %s', mixins, etype)
cls.rtags.__initialize__()
@classmethod
def fetch_rql(cls, user, restriction=None, fetchattrs=None, mainvar='X',
settype=True, ordermethod='fetch_order'):
"""return a rql to fetch all entities of the class type"""
restrictions = restriction or []
if settype:
restrictions.append('%s is %s' % (mainvar, cls.id))
if fetchattrs is None:
fetchattrs = cls.fetch_attrs
selection = [mainvar]
orderby = []
# start from 26 to avoid possible conflicts with X
varmaker = rqlvar_maker(index=26)
cls._fetch_restrictions(mainvar, varmaker, fetchattrs, selection,
orderby, restrictions, user, ordermethod)
rql = 'Any %s' % ','.join(selection)
if orderby:
rql += ' ORDERBY %s' % ','.join(orderby)
rql += ' WHERE %s' % ', '.join(restrictions)
return rql
@classmethod
def _fetch_restrictions(cls, mainvar, varmaker, fetchattrs,
selection, orderby, restrictions, user,
ordermethod='fetch_order', visited=None):
eschema = cls.e_schema
if visited is None:
visited = set((eschema.type,))
elif eschema.type in visited:
# avoid infinite recursion
return
else:
visited.add(eschema.type)
_fetchattrs = []
for attr in fetchattrs:
try:
rschema = eschema.subject_relation(attr)
except KeyError:
cls.warning('skipping fetch_attr %s defined in %s (not found in schema)',
attr, cls.id)
continue
if not user.matching_groups(rschema.get_groups('read')):
continue
var = varmaker.next()
selection.append(var)
restriction = '%s %s %s' % (mainvar, attr, var)
restrictions.append(restriction)
if not rschema.is_final():
# XXX this does not handle several destination types
desttype = rschema.objects(eschema.type)[0]
card = rschema.rproperty(eschema, desttype, 'cardinality')[0]
if card not in '?1':
selection.pop()
restrictions.pop()
continue
if card == '?':
restrictions[-1] += '?' # left outer join if not mandatory
destcls = cls.vreg.etype_class(desttype)
destcls._fetch_restrictions(var, varmaker, destcls.fetch_attrs,
selection, orderby, restrictions,
user, ordermethod, visited=visited)
orderterm = getattr(cls, ordermethod)(attr, var)
if orderterm:
orderby.append(orderterm)
return selection, orderby, restrictions
def __init__(self, req, rset, row=None, col=0):
AppRsetObject.__init__(self, req, rset)
dict.__init__(self)
self.row, self.col = row, col
self._related_cache = {}
if rset is not None:
self.eid = rset[row][col]
else:
self.eid = None
self._is_saved = True
def __repr__(self):
return '<Entity %s %s %s at %s>' % (
self.e_schema, self.eid, self.keys(), id(self))
def __nonzero__(self):
return True
def __hash__(self):
return id(self)
def pre_add_hook(self):
"""hook called by the repository before doing anything to add the entity
(before_add entity hooks have not been called yet). This give the
occasion to do weird stuff such as autocast (File -> Image for instance).
This method must return the actual entity to be added.
"""
return self
def set_eid(self, eid):
self.eid = self['eid'] = eid
def has_eid(self):
"""return True if the entity has an attributed eid (False
meaning that the entity has to be created
"""
try:
int(self.eid)
return True
except (ValueError, TypeError):
return False
def is_saved(self):
"""during entity creation, there is some time during which the entity
has an eid attributed though it's not saved (eg during before_add_entity
hooks). You can use this method to ensure the entity has an eid *and* is
saved in its source.
"""
return self.has_eid() and self._is_saved
@cached
def metainformation(self):
res = dict(zip(('type', 'source', 'extid'), self.req.describe(self.eid)))
res['source'] = self.req.source_defs()[res['source']]
return res
def clear_local_perm_cache(self, action):
for rqlexpr in self.e_schema.get_rqlexprs(action):
self.req.local_perm_cache.pop((rqlexpr.eid, (('x', self.eid),)), None)
def check_perm(self, action):
self.e_schema.check_perm(self.req, action, self.eid)
def has_perm(self, action):
return self.e_schema.has_perm(self.req, action, self.eid)
def view(self, vid, __registry='views', **kwargs):
"""shortcut to apply a view on this entity"""
return self.vreg.render(__registry, vid, self.req, rset=self.rset,
row=self.row, col=self.col, **kwargs)
def absolute_url(self, method=None, **kwargs):
"""return an absolute url to view this entity"""
# in linksearch mode, we don't want external urls else selecting
# the object for use in the relation is tricky
# XXX search_state is web specific
if getattr(self.req, 'search_state', ('normal',))[0] == 'normal':
kwargs['base_url'] = self.metainformation()['source'].get('base-url')
if method is None or method == 'view':
kwargs['_restpath'] = self.rest_path()
else:
kwargs['rql'] = 'Any X WHERE X eid %s' % self.eid
return self.build_url(method, **kwargs)
def rest_path(self):
"""returns a REST-like (relative) path for this entity"""
mainattr, needcheck = self._rest_attr_info()
etype = str(self.e_schema)
if mainattr == 'eid':
value = self.eid
else:
value = getattr(self, mainattr)
if value is None:
return '%s/eid/%s' % (etype.lower(), self.eid)
if needcheck:
# make sure url is not ambiguous
rql = 'Any COUNT(X) WHERE X is %s, X %s %%(value)s' % (etype, mainattr)
if value is not None:
nbresults = self.req.execute(rql, {'value' : value})[0][0]
# may an assertion that nbresults is not 0 would be a good idea
if nbresults != 1: # no ambiguity
return '%s/eid/%s' % (etype.lower(), self.eid)
return '%s/%s' % (etype.lower(), self.req.url_quote(value))
@classmethod
def _rest_attr_info(cls):
mainattr, needcheck = 'eid', True
if cls.rest_attr:
mainattr = cls.rest_attr
needcheck = not cls.e_schema.has_unique_values(mainattr)
else:
for rschema in cls.e_schema.subject_relations():
if rschema.is_final() and rschema != 'eid' and cls.e_schema.has_unique_values(rschema):
mainattr = str(rschema)
needcheck = False
break
if mainattr == 'eid':
needcheck = False
return mainattr, needcheck
@cached
def formatted_attrs(self):
"""returns the list of attributes which have some format information
(i.e. rich text strings)
"""
attrs = []
for rschema, attrschema in self.e_schema.attribute_definitions():
if attrschema.type == 'String' and self.has_format(rschema):
attrs.append(rschema.type)
return attrs
def format(self, attr):
"""return the mime type format for an attribute (if specified)"""
return getattr(self, '%s_format' % attr, None)
def text_encoding(self, attr):
"""return the text encoding for an attribute, default to site encoding
"""
encoding = getattr(self, '%s_encoding' % attr, None)
return encoding or self.vreg.property_value('ui.encoding')
def has_format(self, attr):
"""return true if this entity's schema has a format field for the given
attribute
"""
return self.e_schema.has_subject_relation('%s_format' % attr)
def has_text_encoding(self, attr):
"""return true if this entity's schema has ab encoding field for the
given attribute
"""
return self.e_schema.has_subject_relation('%s_encoding' % attr)
def printable_value(self, attr, value=_marker, attrtype=None,
format='text/html', displaytime=True):
"""return a displayable value (i.e. unicode string) which may contains
html tags
"""
attr = str(attr)
if value is _marker:
value = getattr(self, attr)
if isinstance(value, basestring):
value = value.strip()
if value is None or value == '': # don't use "not", 0 is an acceptable value
return u''
if attrtype is None:
attrtype = self.e_schema.destination(attr)
props = self.e_schema.rproperties(attr)
if attrtype == 'String':
# internalinalized *and* formatted string such as schema
# description...
if props.get('internationalizable'):
value = self.req._(value)
attrformat = self.format(attr)
if attrformat:
return self.mtc_transform(value, attrformat, format,
self.req.encoding)
elif attrtype == 'Bytes':
attrformat = self.format(attr)
if attrformat:
try:
encoding = getattr(self, '%s_encoding' % attr)
except AttributeError:
encoding = self.req.encoding
return self.mtc_transform(value.getvalue(), attrformat, format,
encoding)
return u''
value = printable_value(self.req, attrtype, value, props, displaytime)
if format == 'text/html':
value = html_escape(value)
return value
def mtc_transform(self, data, format, target_format, encoding,
_engine=ENGINE):
trdata = TransformData(data, format, encoding, appobject=self)
data = _engine.convert(trdata, target_format).decode()
if format == 'text/html':
data = soup2xhtml(data, self.req.encoding)
return data
# entity cloning ##########################################################
def copy_relations(self, ceid):
"""copy relations of the object with the given eid on this object
By default meta and composite relations are skipped.
Overrides this if you want another behaviour
"""
assert self.has_eid()
execute = self.req.execute
for rschema in self.e_schema.subject_relations():
if rschema.meta or rschema.is_final():
continue
# skip already defined relations
if getattr(self, rschema.type):
continue
if rschema.type in self.skip_copy_for:
continue
if rschema.type == 'in_state':
# if the workflow is defining an initial state (XXX AND we are
# not in the managers group? not done to be more consistent)
# don't try to copy in_state
if execute('Any S WHERE S state_of ET, ET initial_state S,'
'ET name %(etype)s', {'etype': str(self.e_schema)}):
continue
# skip composite relation
if self.e_schema.subjrproperty(rschema, 'composite'):
continue
# skip relation with card in ?1 else we either change the copied
# object (inlined relation) or inserting some inconsistency
if self.e_schema.subjrproperty(rschema, 'cardinality')[1] in '?1':
continue
rql = 'SET X %s V WHERE X eid %%(x)s, Y eid %%(y)s, Y %s V' % (
rschema.type, rschema.type)
execute(rql, {'x': self.eid, 'y': ceid}, ('x', 'y'))
self.clear_related_cache(rschema.type, 'subject')
for rschema in self.e_schema.object_relations():
if rschema.meta:
continue
# skip already defined relations
if getattr(self, 'reverse_%s' % rschema.type):
continue
# skip composite relation
if self.e_schema.objrproperty(rschema, 'composite'):
continue
# skip relation with card in ?1 else we either change the copied
# object (inlined relation) or inserting some inconsistency
if self.e_schema.objrproperty(rschema, 'cardinality')[0] in '?1':
continue
rql = 'SET V %s X WHERE X eid %%(x)s, Y eid %%(y)s, V %s Y' % (
rschema.type, rschema.type)
execute(rql, {'x': self.eid, 'y': ceid}, ('x', 'y'))
self.clear_related_cache(rschema.type, 'object')
# data fetching methods ###################################################
@cached
def as_rset(self):
"""returns a resultset containing `self` information"""
rset = ResultSet([(self.eid,)], 'Any X WHERE X eid %(x)s',
{'x': self.eid}, [(self.id,)])
return self.req.decorate_rset(rset)
def to_complete_relations(self):
"""by default complete final relations to when calling .complete()"""
for rschema in self.e_schema.subject_relations():
if rschema.is_final():
continue
if len(rschema.objects(self.e_schema)) > 1:
# ambigous relations, the querier doesn't handle
# outer join correctly in this case
continue
if rschema.inlined:
matching_groups = self.req.user.matching_groups
if matching_groups(rschema.get_groups('read')) and \
all(matching_groups(es.get_groups('read'))
for es in rschema.objects(self.e_schema)):
yield rschema, 'subject'
def to_complete_attributes(self, skip_bytes=True):
for rschema, attrschema in self.e_schema.attribute_definitions():
# skip binary data by default
if skip_bytes and attrschema.type == 'Bytes':
continue
attr = rschema.type
if attr == 'eid':
continue
# password retreival is blocked at the repository server level
if not self.req.user.matching_groups(rschema.get_groups('read')) \
or attrschema.type == 'Password':
self[attr] = None
continue
yield attr
def complete(self, attributes=None, skip_bytes=True):
"""complete this entity by adding missing attributes (i.e. query the
repository to fill the entity)
:type skip_bytes: bool
:param skip_bytes:
if true, attribute of type Bytes won't be considered
"""
assert self.has_eid()
varmaker = rqlvar_maker()
V = varmaker.next()
rql = ['WHERE %s eid %%(x)s' % V]
selected = []
for attr in (attributes or self.to_complete_attributes(skip_bytes)):
# if attribute already in entity, nothing to do
if self.has_key(attr):
continue
# case where attribute must be completed, but is not yet in entity
var = varmaker.next()
rql.append('%s %s %s' % (V, attr, var))
selected.append((attr, var))
# +1 since this doen't include the main variable
lastattr = len(selected) + 1
if attributes is None:
# fetch additional relations (restricted to 0..1 relations)
for rschema, role in self.to_complete_relations():
rtype = rschema.type
if self.relation_cached(rtype, role):
continue
var = varmaker.next()
if role == 'subject':
targettype = rschema.objects(self.e_schema)[0]
card = rschema.rproperty(self.e_schema, targettype,
'cardinality')[0]
if card == '1':
rql.append('%s %s %s' % (V, rtype, var))
else: # '?"
rql.append('%s %s %s?' % (V, rtype, var))
else:
targettype = rschema.subjects(self.e_schema)[1]
card = rschema.rproperty(self.e_schema, targettype,
'cardinality')[1]
if card == '1':
rql.append('%s %s %s' % (var, rtype, V))
else: # '?"
rql.append('%s? %s %s' % (var, rtype, V))
assert card in '1?', '%s %s %s %s' % (self.e_schema, rtype,
role, card)
selected.append(((rtype, role), var))
if selected:
# select V, we need it as the left most selected variable
# if some outer join are included to fetch inlined relations
rql = 'Any %s,%s %s' % (V, ','.join(var for attr, var in selected),
','.join(rql))
execute = getattr(self.req, 'unsafe_execute', self.req.execute)
rset = execute(rql, {'x': self.eid}, 'x', build_descr=False)[0]
# handle attributes
for i in xrange(1, lastattr):
self[str(selected[i-1][0])] = rset[i]
# handle relations
for i in xrange(lastattr, len(rset)):
rtype, x = selected[i-1][0]
value = rset[i]
if value is None:
rrset = ResultSet([], rql, {'x': self.eid})
self.req.decorate_rset(rrset)
else:
rrset = self.req.eid_rset(value)
self.set_related_cache(rtype, x, rrset)
def get_value(self, name):
"""get value for the attribute relation <name>, query the repository
to get the value if necessary.
:type name: str
:param name: name of the attribute to get
"""
try:
value = self[name]
except KeyError:
if not self.is_saved():
return None
rql = "Any A WHERE X eid %%(x)s, X %s A" % name
# XXX should we really use unsafe_execute here??
execute = getattr(self.req, 'unsafe_execute', self.req.execute)
try:
rset = execute(rql, {'x': self.eid}, 'x')
except Unauthorized:
self[name] = value = None
else:
assert rset.rowcount <= 1, (self, rql, rset.rowcount)
try:
self[name] = value = rset.rows[0][0]
except IndexError:
# probably a multisource error
self.critical("can't get value for attribute %s of entity with eid %s",
name, self.eid)
if self.e_schema.destination(name) == 'String':
self[name] = value = self.req._('unaccessible')
else:
self[name] = value = None
return value
def related(self, rtype, role='subject', limit=None, entities=False):
"""returns a resultset of related entities
:param role: is the role played by 'self' in the relation ('subject' or 'object')
:param limit: resultset's maximum size
:param entities: if True, the entites are returned; if False, a result set is returned
"""
try:
return self.related_cache(rtype, role, entities, limit)
except KeyError:
pass
assert self.has_eid()
rql = self.related_rql(rtype, role)
rset = self.req.execute(rql, {'x': self.eid}, 'x')
self.set_related_cache(rtype, role, rset)
return self.related(rtype, role, limit, entities)
def related_rql(self, rtype, role='subject'):
rschema = self.schema[rtype]
if role == 'subject':
targettypes = rschema.objects(self.e_schema)
restriction = 'E eid %%(x)s, E %s X' % rtype
card = greater_card(rschema, (self.e_schema,), targettypes, 0)
else:
targettypes = rschema.subjects(self.e_schema)
restriction = 'E eid %%(x)s, X %s E' % rtype
card = greater_card(rschema, targettypes, (self.e_schema,), 1)
if len(targettypes) > 1:
fetchattrs_list = []
for ttype in targettypes:
etypecls = self.vreg.etype_class(ttype)
fetchattrs_list.append(set(etypecls.fetch_attrs))
fetchattrs = reduce(set.intersection, fetchattrs_list)
rql = etypecls.fetch_rql(self.req.user, [restriction], fetchattrs,
settype=False)
else:
etypecls = self.vreg.etype_class(targettypes[0])
rql = etypecls.fetch_rql(self.req.user, [restriction], settype=False)
# optimisation: remove ORDERBY if cardinality is 1 or ? (though
# greater_card return 1 for those both cases)
if card == '1':
if ' ORDERBY ' in rql:
rql = '%s WHERE %s' % (rql.split(' ORDERBY ', 1)[0],
rql.split(' WHERE ', 1)[1])
elif not ' ORDERBY ' in rql:
args = tuple(rql.split(' WHERE ', 1))
rql = '%s ORDERBY Z DESC WHERE X modification_date Z, %s' % args
return rql
# generic vocabulary methods ##############################################
def vocabulary(self, rtype, role='subject', limit=None):
"""vocabulary functions must return a list of couples
(label, eid) that will typically be used to fill the
edition view's combobox.
If `eid` is None in one of these couples, it should be
interpreted as a separator in case vocabulary results are grouped
"""
try:
vocabfunc = getattr(self, '%s_%s_vocabulary' % (role, rtype))
except AttributeError:
vocabfunc = getattr(self, '%s_relation_vocabulary' % role)
# NOTE: it is the responsibility of `vocabfunc` to sort the result
# (direclty through RQL or via a python sort). This is also
# important because `vocabfunc` might return a list with
# couples (label, None) which act as separators. In these
# cases, it doesn't make sense to sort results afterwards.
return vocabfunc(rtype, limit)
def unrelated_rql(self, rtype, targettype, role, ordermethod=None,
vocabconstraints=True):
"""build a rql to fetch `targettype` entities unrelated to this entity
using (rtype, role) relation
"""
ordermethod = ordermethod or 'fetch_unrelated_order'
if isinstance(rtype, basestring):
rtype = self.schema.rschema(rtype)
if role == 'subject':
evar, searchedvar = 'S', 'O'
subjtype, objtype = self.e_schema, targettype
else:
searchedvar, evar = 'S', 'O'
objtype, subjtype = self.e_schema, targettype
if self.has_eid():
restriction = ['NOT S %s O' % rtype, '%s eid %%(x)s' % evar]
else:
restriction = []
constraints = rtype.rproperty(subjtype, objtype, 'constraints')
if vocabconstraints:
# RQLConstraint is a subclass for RQLVocabularyConstraint, so they
# will be included as well
restriction += [cstr.restriction for cstr in constraints
if isinstance(cstr, RQLVocabularyConstraint)]
else:
restriction += [cstr.restriction for cstr in constraints
if isinstance(cstr, RQLConstraint)]
etypecls = self.vreg.etype_class(targettype)
rql = etypecls.fetch_rql(self.req.user, restriction,
mainvar=searchedvar, ordermethod=ordermethod)
# ensure we have an order defined
if not ' ORDERBY ' in rql:
before, after = rql.split(' WHERE ', 1)
rql = '%s ORDERBY %s WHERE %s' % (before, searchedvar, after)
return rql
def unrelated(self, rtype, targettype, role='subject', limit=None,
ordermethod=None):
"""return a result set of target type objects that may be related
by a given relation, with self as subject or object
"""
rql = self.unrelated_rql(rtype, targettype, role, ordermethod)
if limit is not None:
before, after = rql.split(' WHERE ', 1)
rql = '%s LIMIT %s WHERE %s' % (before, limit, after)
if self.has_eid():
return self.req.execute(rql, {'x': self.eid})
return self.req.execute(rql)
# relations cache handling ################################################
def relation_cached(self, rtype, role):
"""return true if the given relation is already cached on the instance
"""
return '%s_%s' % (rtype, role) in self._related_cache
def related_cache(self, rtype, role, entities=True, limit=None):
"""return values for the given relation if it's cached on the instance,
else raise `KeyError`
"""
res = self._related_cache['%s_%s' % (rtype, role)][entities]
if limit:
if entities:
res = res[:limit]
else:
res = res.limit(limit)
return res
def set_related_cache(self, rtype, role, rset, col=0):
"""set cached values for the given relation"""
if rset:
related = list(rset.entities(col))
rschema = self.schema.rschema(rtype)
if role == 'subject':
rcard = rschema.rproperty(self.e_schema, related[0].e_schema,
'cardinality')[1]
target = 'object'
else:
rcard = rschema.rproperty(related[0].e_schema, self.e_schema,
'cardinality')[0]
target = 'subject'
if rcard in '?1':
for rentity in related:
rentity._related_cache['%s_%s' % (rtype, target)] = (self.as_rset(), [self])
else:
related = []
self._related_cache['%s_%s' % (rtype, role)] = (rset, related)
def clear_related_cache(self, rtype=None, role=None):
"""clear cached values for the given relation or the entire cache if
no relation is given
"""
if rtype is None:
self._related_cache = {}
else:
assert role
self._related_cache.pop('%s_%s' % (rtype, role), None)
# raw edition utilities ###################################################
def set_attributes(self, **kwargs):
assert kwargs
relations = []
for key in kwargs:
relations.append('X %s %%(%s)s' % (key, key))
# update current local object
self.update(kwargs)
# and now update the database
kwargs['x'] = self.eid
self.req.execute('SET %s WHERE X eid %%(x)s' % ','.join(relations),
kwargs, 'x')
def delete(self):
assert self.has_eid(), self.eid
self.req.execute('DELETE %s X WHERE X eid %%(x)s' % self.e_schema,
{'x': self.eid})
# server side utilities ###################################################
def set_defaults(self):
"""set default values according to the schema"""
self._default_set = set()
for attr, value in self.e_schema.defaults():
if not self.has_key(attr):
self[str(attr)] = value
self._default_set.add(attr)
def check(self, creation=False):
"""check this entity against its schema. Only final relation
are checked here, constraint on actual relations are checked in hooks
"""
# necessary since eid is handled specifically and yams require it to be
# in the dictionary
if self.req is None:
_ = unicode
else:
_ = self.req._
self.e_schema.check(self, creation=creation, _=_)
def fti_containers(self, _done=None):
if _done is None:
_done = set()
_done.add(self.eid)
containers = tuple(self.e_schema.fulltext_containers())
if containers:
yielded = False
for rschema, target in containers:
if target == 'object':
targets = getattr(self, rschema.type)
else:
targets = getattr(self, 'reverse_%s' % rschema)
for entity in targets:
if entity.eid in _done:
continue
for container in entity.fti_containers(_done):
yield container
yielded = True
if not yielded:
yield self
else:
yield self
def get_words(self):
"""used by the full text indexer to get words to index
this method should only be used on the repository side since it depends
on the indexer package
:rtype: list
:return: the list of indexable word of this entity
"""
from indexer.query_objects import tokenize
words = []
for rschema in self.e_schema.indexable_attributes():
try:
value = self.printable_value(rschema, format='text/plain')
except TransformError, ex:
continue
except:
self.exception("can't add value of %s to text index for entity %s",
rschema, self.eid)
continue
if value:
words += tokenize(value)
for rschema, role in self.e_schema.fulltext_relations():
if role == 'subject':
for entity in getattr(self, rschema.type):
words += entity.get_words()
else: # if role == 'object':
for entity in getattr(self, 'reverse_%s' % rschema.type):
words += entity.get_words()
return words
# attribute and relation descriptors ##########################################
class Attribute(object):
"""descriptor that controls schema attribute access"""
def __init__(self, attrname):
assert attrname != 'eid'
self._attrname = attrname
def __get__(self, eobj, eclass):
if eobj is None:
return self
return eobj.get_value(self._attrname)
def __set__(self, eobj, value):
# XXX bw compat
# would be better to generate UPDATE queries than the current behaviour
eobj.warning("deprecated usage, don't use 'entity.attr = val' notation)")
eobj[self._attrname] = value
class Relation(object):
"""descriptor that controls schema relation access"""
_role = None # for pylint
def __init__(self, rschema):
self._rschema = rschema
self._rtype = rschema.type
def __get__(self, eobj, eclass):
if eobj is None:
raise AttributeError('%s cannot be only be accessed from instances'
% self._rtype)
return eobj.related(self._rtype, self._role, entities=True)
def __set__(self, eobj, value):
raise NotImplementedError
class SubjectRelation(Relation):
"""descriptor that controls schema relation access"""
_role = 'subject'
class ObjectRelation(Relation):
"""descriptor that controls schema relation access"""
_role = 'object'
from logging import getLogger
from cubicweb import set_log_methods
set_log_methods(Entity, getLogger('cubicweb.entity'))