proper licensing information (LGPL-2.1). Hope I get it right this time.
# copyright 2003-2010 LOGILAB S.A. (Paris, FRANCE), all rights reserved.
# contact http://www.logilab.fr/ -- mailto:contact@logilab.fr
#
# This file is part of CubicWeb.
#
# CubicWeb is free software: you can redistribute it and/or modify it under the
# terms of the GNU Lesser General Public License as published by the Free
# Software Foundation, either version 2.1 of the License, or (at your option)
# any later version.
#
# logilab-common is distributed in the hope that it will be useful, but WITHOUT
# ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
# FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more
# details.
#
# You should have received a copy of the GNU Lesser General Public License along
# with CubicWeb. If not, see <http://www.gnu.org/licenses/>.
"""provide a minimal RQL support for google appengine dbmodel
"""
__docformat__ = "restructuredtext en"
from datetime import datetime
from rql import RQLHelper, nodes
from cubicweb import Binary
from cubicweb.rset import ResultSet
from cubicweb.server import SQL_CONNECT_HOOKS
from google.appengine.api.datastore import Key, Get, Query, Entity
from google.appengine.api.datastore_types import Text, Blob
from google.appengine.api.datastore_errors import EntityNotFoundError, BadKeyError
def etype_from_key(key):
return Key(key).kind()
def poss_var_types(myvar, ovar, kind, solutions):
return frozenset(etypes[myvar] for etypes in solutions
if etypes[ovar] == kind)
def expand_result(results, result, myvar, values, dsget=None):
values = map(dsget, values)
if values:
result[myvar] = values.pop(0)
for value in values:
newresult = result.copy()
newresult[myvar] = value
results.append(newresult)
else:
results.remove(result)
def _resolve(restrictions, solutions, fixed):
varname = restrictions[0].searched_var
objs = []
for etype in frozenset(etypes[varname] for etypes in solutions):
gqlargs = {}
query = Query(etype)
for restriction in restrictions:
restriction.fill_query(fixed, query)
pobjs = query.Run()
if varname in fixed:
value = fixed[varname]
objs += (x for x in pobjs if x == value)
else:
objs += pobjs
if varname in fixed and not objs:
raise EidMismatch(varname, value)
return objs
def _resolve_not(restrictions, solutions, fixed):
restr = restrictions[0]
constrvarname = restr.constraint_var
if len(restrictions) > 1 or not constrvarname in fixed:
raise NotImplementedError()
varname = restr.searched_var
objs = []
for etype in frozenset(etypes[varname] for etypes in solutions):
gqlargs = {}
for operator in ('<', '>'):
query = Query(etype)
restr.fill_query(fixed, query, operator)
objs += query.Run()
return objs
def _print_results(rlist):
return '[%s]' % ', '.join(_print_result(r) for r in rlist)
def _print_result(rdict):
string = []
for k, v in rdict.iteritems():
if isinstance(v, Entity):
string.append('%s: %s' % (k, v.key()))#_print_result(v)))
elif isinstance(v, list):
string.append('%s: [%s]' % (k, ', '.join(str(i) for i in v)))
else:
string.append('%s: %s' % (k, v))
return '{%s}' % ', '.join(string)
class EidMismatch(Exception):
def __init__(self, varname, value):
self.varname = varname
self.value = value
class Restriction(object):
supported_operators = ('=',)
def __init__(self, rel):
operator = rel.children[1].operator
if not operator in self.supported_operators:
raise NotImplementedError('unsupported operator')
self.rel = rel
self.operator = operator
self.rtype = rel.r_type
self.var = rel.children[0]
def __repr__(self):
return '<%s for %s>' % (self.__class__.__name__, self.rel)
@property
def rhs(self):
return self.rel.children[1].children[0]
class MultipleRestriction(object):
def __init__(self, restrictions):
self.restrictions = restrictions
def resolve(self, solutions, fixed):
return _resolve(self.restrictions, solutions, fixed)
class VariableSelection(Restriction):
def __init__(self, rel, dsget, prefix='s'):
Restriction.__init__(self, rel)
self._dsget = dsget
self._not = self.rel.neged(strict=True)
self._prefix = prefix + '_'
def __repr__(self):
return '<%s%s for %s>' % (self._prefix[0], self.__class__.__name__, self.rel)
@property
def searched_var(self):
if self._prefix == 's_':
return self.var.name
return self.rhs.name
@property
def constraint_var(self):
if self._prefix == 's_':
return self.rhs.name
return self.var.name
def _possible_values(self, myvar, ovar, entity, solutions, dsprefix):
if self.rtype == 'identity':
return (entity.key(),)
value = entity.get(dsprefix + self.rtype)
if value is None:
return []
if not isinstance(value, list):
value = [value]
vartypes = poss_var_types(myvar, ovar, entity.kind(), solutions)
return (v for v in value if v.kind() in vartypes)
def complete_and_filter(self, solutions, results):
myvar = self.rhs.name
ovar = self.var.name
rtype = self.rtype
if self.schema.rschema(rtype).final:
# should be detected by rql.stcheck: "Any C WHERE NOT X attr C" doesn't make sense
#if self._not:
# raise NotImplementedError()
for result in results:
result[myvar] = result[ovar].get('s_'+rtype)
elif self.var.name in results[0]:
if self.rhs.name in results[0]:
self.filter(solutions, results)
else:
if self._not:
raise NotImplementedError()
for result in results[:]:
values = self._possible_values(myvar, ovar, result[ovar],
solutions, 's_')
expand_result(results, result, myvar, values, self._dsget)
else:
assert self.rhs.name in results[0]
self.object_complete_and_filter(solutions, results)
def filter(self, solutions, results):
myvar = self.rhs.name
ovar = self.var.name
newsols = {}
for result in results[:]:
entity = result[ovar]
key = entity.key()
if not key in newsols:
values = self._possible_values(myvar, ovar, entity, solutions, 's_')
newsols[key] = frozenset(v for v in values)
if self._not:
if result[myvar].key() in newsols[key]:
results.remove(result)
elif not result[myvar].key() in newsols[key]:
results.remove(result)
def object_complete_and_filter(self, solutions, results):
if self._not:
raise NotImplementedError()
myvar = self.var.name
ovar = self.rhs.name
for result in results[:]:
values = self._possible_values(myvar, ovar, result[ovar],
solutions, 'o_')
expand_result(results, result, myvar, values, self._dsget)
class EidRestriction(Restriction):
def __init__(self, rel, dsget):
Restriction.__init__(self, rel)
self._dsget = dsget
def resolve(self, kwargs):
value = self.rel.children[1].children[0].eval(kwargs)
return self._dsget(value)
class RelationRestriction(VariableSelection):
def _get_value(self, fixed):
return fixed[self.constraint_var].key()
def fill_query(self, fixed, query, operator=None):
restr = '%s%s %s' % (self._prefix, self.rtype, operator or self.operator)
query[restr] = self._get_value(fixed)
def resolve(self, solutions, fixed):
if self.rtype == 'identity':
if self._not:
raise NotImplementedError()
return [fixed[self.constraint_var]]
if self._not:
return _resolve_not([self], solutions, fixed)
return _resolve([self], solutions, fixed)
class NotRelationRestriction(RelationRestriction):
def _get_value(self, fixed):
return None
def resolve(self, solutions, fixed):
if self.rtype == 'identity':
raise NotImplementedError()
return _resolve([self], solutions, fixed)
class AttributeRestriction(RelationRestriction):
supported_operators = ('=', '>', '>=', '<', '<=', 'ILIKE')
def __init__(self, rel, kwargs):
RelationRestriction.__init__(self, rel, None)
value = self.rhs.eval(kwargs)
self.value = value
if self.operator == 'ILIKE':
if value.startswith('%'):
raise NotImplementedError('LIKE is only supported for prefix search')
if not value.endswith('%'):
raise NotImplementedError('LIKE is only supported for prefix search')
self.operator = '>'
self.value = value[:-1]
def complete_and_filter(self, solutions, results):
# check lhs var first in case this is a restriction
assert self._not
myvar, rtype, value = self.var.name, self.rtype, self.value
for result in results[:]:
if result[myvar].get('s_'+rtype) == value:
results.remove(result)
def _get_value(self, fixed):
return self.value
class DateAttributeRestriction(AttributeRestriction):
"""just a thin layer on top af `AttributeRestriction` that
tries to convert date strings such as in :
Any X WHERE X creation_date >= '2008-03-04'
"""
def __init__(self, rel, kwargs):
super(DateAttributeRestriction, self).__init__(rel, kwargs)
if isinstance(self.value, basestring):
# try:
self.value = datetime.strptime(self.value, '%Y-%m-%d')
# except Exception, exc:
# from logging import error
# error('unable to parse date %s with format %%Y-%%m-%%d (exc=%s)', value, exc)
class AttributeInRestriction(AttributeRestriction):
def __init__(self, rel, kwargs):
RelationRestriction.__init__(self, rel, None)
values = []
for c in self.rel.children[1].iget_nodes(nodes.Constant):
values.append(c.eval(kwargs))
self.value = values
@property
def operator(self):
return 'in'
class TypeRestriction(AttributeRestriction):
def __init__(self, var):
self.var = var
def __repr__(self):
return '<%s for %s>' % (self.__class__.__name__, self.var)
def resolve(self, solutions, fixed):
objs = []
for etype in frozenset(etypes[self.var.name] for etypes in solutions):
objs += Query(etype).Run()
return objs
def append_result(res, descr, i, j, value, etype):
if value is not None:
if isinstance(value, Text):
value = unicode(value)
elif isinstance(value, Blob):
value = Binary(str(value))
if j == 0:
res.append([value])
descr.append([etype])
else:
res[i].append(value)
descr[i].append(etype)
class ValueResolver(object):
def __init__(self, functions, args, term):
self.functions = functions
self.args = args
self.term = term
self._solution = self.term.stmt.solutions[0]
def compute(self, result):
"""return (entity type, value) to which self.term is evaluated according
to the given result dictionnary and to query arguments (self.args)
"""
return self.term.accept(self, result)
def visit_function(self, node, result):
args = tuple(n.accept(self, result)[1] for n in node.children)
value = self.functions[node.name](*args)
return node.get_type(self._solution, self.args), value
def visit_variableref(self, node, result):
value = result[node.name]
try:
etype = value.kind()
value = str(value.key())
except AttributeError:
etype = self._solution[node.name]
return etype, value
def visit_constant(self, node, result):
return node.get_type(kwargs=self.args), node.eval(self.args)
class RQLInterpreter(object):
"""algorithm:
1. visit the restriction clauses and collect restriction for each subject
of a relation. Different restriction types are:
* EidRestriction
* AttributeRestriction
* RelationRestriction
* VariableSelection (not really a restriction)
-> dictionary {<variable>: [restriction...], ...}
2. resolve eid restrictions
3. for each select in union:
for each solution in select'solutions:
1. resolve variables which have attribute restriction
2. resolve relation restriction
3. resolve selection and add to global results
"""
def __init__(self, schema):
self.schema = schema
Restriction.schema = schema # yalta!
self.rqlhelper = RQLHelper(schema, {'eid': etype_from_key})
self._stored_proc = {'LOWER': lambda x: x.lower(),
'UPPER': lambda x: x.upper()}
for cb in SQL_CONNECT_HOOKS.get('sqlite', []):
cb(self)
# emulate sqlite connection interface so we can reuse stored procedures
def create_function(self, name, nbargs, func):
self._stored_proc[name] = func
def create_aggregate(self, name, nbargs, func):
self._stored_proc[name] = func
def execute(self, operation, parameters=None, eid_key=None, build_descr=True):
rqlst = self.rqlhelper.parse(operation, annotate=True)
try:
self.rqlhelper.compute_solutions(rqlst, kwargs=parameters)
except BadKeyError:
results, description = [], []
else:
results, description = self.interpret(rqlst, parameters)
return ResultSet(results, operation, parameters, description, rqlst=rqlst)
def interpret(self, node, kwargs, dsget=None):
if dsget is None:
self._dsget = Get
else:
self._dsget = dsget
try:
return node.accept(self, kwargs)
except NotImplementedError:
self.critical('support for query not implemented: %s', node)
raise
def visit_union(self, node, kwargs):
results, description = [], []
extra = {'kwargs': kwargs}
for child in node.children:
pres, pdescr = self.visit_select(child, extra)
results += pres
description += pdescr
return results, description
def visit_select(self, node, extra):
constraints = {}
if node.where is not None:
node.where.accept(self, constraints, extra)
fixed, toresolve, postresolve, postfilters = {}, {}, {}, []
# extract NOT filters
for vname, restrictions in constraints.items():
for restr in restrictions[:]:
if isinstance(restr, AttributeRestriction) and restr._not:
postfilters.append(restr)
restrictions.remove(restr)
if not restrictions:
del constraints[vname]
# add TypeRestriction for variable which have no restrictions at all
for varname, var in node.defined_vars.iteritems():
if not varname in constraints:
constraints[varname] = [TypeRestriction(var)]
#print node, constraints
# compute eid restrictions
kwargs = extra['kwargs']
for varname, restrictions in constraints.iteritems():
for restr in restrictions[:]:
if isinstance(restr, EidRestriction):
assert not varname in fixed
try:
value = restr.resolve(kwargs)
fixed[varname] = value
except EntityNotFoundError:
return [], []
restrictions.remove(restr)
#print 'fixed', fixed.keys()
# combine remaining restrictions
for varname, restrictions in constraints.iteritems():
for restr in restrictions:
if isinstance(restr, AttributeRestriction):
toresolve.setdefault(varname, []).append(restr)
elif isinstance(restr, NotRelationRestriction) or (
isinstance(restr, RelationRestriction) and
not restr.searched_var in fixed and restr.constraint_var in fixed):
toresolve.setdefault(varname, []).append(restr)
else:
postresolve.setdefault(varname, []).append(restr)
try:
if len(toresolve[varname]) > 1:
toresolve[varname] = MultipleRestriction(toresolve[varname])
else:
toresolve[varname] = toresolve[varname][0]
except KeyError:
pass
#print 'toresolve %s' % toresolve
#print 'postresolve %s' % postresolve
# resolve additional restrictions
if fixed:
partres = [fixed.copy()]
else:
partres = []
for varname, restr in toresolve.iteritems():
varpartres = partres[:]
try:
values = tuple(restr.resolve(node.solutions, fixed))
except EidMismatch, ex:
varname = ex.varname
value = ex.value
partres = [res for res in partres if res[varname] != value]
if partres:
continue
# some join failed, no possible results
return [], []
if not values:
# some join failed, no possible results
return [], []
if not varpartres:
# init results
for value in values:
partres.append({varname: value})
elif not varname in partres[0]:
# cartesian product
for res in partres:
res[varname] = values[0]
for res in partres[:]:
for value in values[1:]:
res = res.copy()
res[varname] = value
partres.append(res)
else:
# union
for res in varpartres:
for value in values:
res = res.copy()
res[varname] = value
partres.append(res)
#print 'partres', len(partres)
#print partres
# Note: don't check for empty partres since constant selection may still
# produce result at this point
# sort to get RelationRestriction before AttributeSelection
restrictions = sorted((restr for restrictions in postresolve.itervalues()
for restr in restrictions),
key=lambda x: not isinstance(x, RelationRestriction))
# compute stuff not doable in the previous step using datastore queries
for restr in restrictions + postfilters:
restr.complete_and_filter(node.solutions, partres)
if not partres:
# some join failed, no possible results
return [], []
if extra.pop('has_exists', False):
# remove potential duplicates introduced by exists
toremovevars = [v.name for v in node.defined_vars.itervalues()
if not v.scope is node]
if toremovevars:
newpartres = []
for result in partres:
for var in toremovevars:
del result[var]
if not result in newpartres:
newpartres.append(result)
if not newpartres:
# some join failed, no possible results
return [], []
partres = newpartres
if node.orderby:
for sortterm in reversed(node.orderby):
resolver = ValueResolver(self._stored_proc, kwargs, sortterm.term)
partres.sort(reverse=not sortterm.asc,
key=lambda x: resolver.compute(x)[1])
if partres:
if node.offset:
partres = partres[node.offset:]
if node.limit:
partres = partres[:node.limit]
if not partres:
return [], []
#print 'completed partres', _print_results(partres)
# compute results
res, descr = [], []
for j, term in enumerate(node.selection):
resolver = ValueResolver(self._stored_proc, kwargs, term)
if not partres:
etype, value = resolver.compute({})
# only constant selected
if not res:
res.append([])
descr.append([])
res[0].append(value)
descr[0].append(etype)
else:
for i, sol in enumerate(partres):
etype, value = resolver.compute(sol)
append_result(res, descr, i, j, value, etype)
#print '--------->', res
return res, descr
def visit_and(self, node, constraints, extra):
for child in node.children:
child.accept(self, constraints, extra)
def visit_exists(self, node, constraints, extra):
extra['has_exists'] = True
self.visit_and(node, constraints, extra)
def visit_not(self, node, constraints, extra):
for child in node.children:
child.accept(self, constraints, extra)
try:
extra.pop(node)
except KeyError:
raise NotImplementedError()
def visit_relation(self, node, constraints, extra):
if node.is_types_restriction():
return
rschema = self.schema.rschema(node.r_type)
neged = node.neged(strict=True)
if neged:
# ok, we *may* process this Not node (not implemented error will be
# raised later if we can't)
extra[node.parent] = True
if rschema.final:
self._visit_final_relation(rschema, node, constraints, extra)
elif neged:
self._visit_non_final_neged_relation(rschema, node, constraints)
else:
self._visit_non_final_relation(rschema, node, constraints)
def _visit_non_final_relation(self, rschema, node, constraints, not_=False):
lhs, rhs = node.get_variable_parts()
for v1, v2, prefix in ((lhs, rhs, 's'), (rhs, lhs, 'o')):
#if not_:
nbrels = len(v2.variable.stinfo['relations'])
#else:
# nbrels = len(v2.variable.stinfo['relations']) - len(v2.variable.stinfo['uidrels'])
if nbrels > 1:
constraints.setdefault(v1.name, []).append(
RelationRestriction(node, self._dsget, prefix))
# just init an empty list for v2 variable to avoid a
# TypeRestriction being added for it
constraints.setdefault(v2.name, [])
break
else:
constraints.setdefault(rhs.name, []).append(
VariableSelection(node, self._dsget, 's'))
def _visit_non_final_neged_relation(self, rschema, node, constraints):
lhs, rhs = node.get_variable_parts()
for v1, v2, prefix in ((lhs, rhs, 's'), (rhs, lhs, 'o')):
stinfo = v2.variable.stinfo
if not stinfo['selected'] and len(stinfo['relations']) == 1:
constraints.setdefault(v1.name, []).append(
NotRelationRestriction(node, self._dsget, prefix))
constraints.setdefault(v2.name, [])
break
else:
self._visit_non_final_relation(rschema, node, constraints, True)
def _visit_final_relation(self, rschema, node, constraints, extra):
varname = node.children[0].name
if rschema.type == 'eid':
constraints.setdefault(varname, []).append(
EidRestriction(node, self._dsget))
else:
rhs = node.children[1].children[0]
if isinstance(rhs, nodes.VariableRef):
constraints.setdefault(rhs.name, []).append(
VariableSelection(node, self._dsget))
elif isinstance(rhs, nodes.Constant):
if rschema.objects()[0] in ('Datetime', 'Date'): # XXX
constraints.setdefault(varname, []).append(
DateAttributeRestriction(node, extra['kwargs']))
else:
constraints.setdefault(varname, []).append(
AttributeRestriction(node, extra['kwargs']))
elif isinstance(rhs, nodes.Function) and rhs.name == 'IN':
constraints.setdefault(varname, []).append(
AttributeInRestriction(node, extra['kwargs']))
else:
raise NotImplementedError()
def _not_implemented(self, *args, **kwargs):
raise NotImplementedError()
visit_or = _not_implemented
# shouldn't occurs
visit_set = _not_implemented
visit_insert = _not_implemented
visit_delete = _not_implemented
from logging import getLogger
from cubicweb import set_log_methods
set_log_methods(RQLInterpreter, getLogger('cubicweb.goa.rqlinterpreter'))
set_log_methods(Restriction, getLogger('cubicweb.goa.rqlinterpreter'))