python3 - 2to3 + miscell obvious tweaks
[sfa.git] / sfa / storage / model.py
index da66716..a2e3bd4 100644 (file)
@@ -1,6 +1,6 @@
-from types import StringTypes
 from datetime import datetime
 
+from sqlalchemy import or_, and_
 from sqlalchemy import Column, Integer, String, DateTime
 from sqlalchemy import Table, Column, MetaData, join, ForeignKey
 from sqlalchemy.orm import relationship, backref
@@ -9,13 +9,16 @@ from sqlalchemy.orm import object_mapper
 from sqlalchemy.orm import validates
 from sqlalchemy.ext.declarative import declarative_base
 
+from sfa.storage.record import Record
 from sfa.util.sfalogging import logger
-from sfa.util.xml import XML 
+from sfa.util.sfatime import utcparse, datetime_to_string
+from sfa.util.xml import XML
+from sfa.util.py23 import StringType
 
 from sfa.trust.gid import GID
 
 ##############################
-Base=declarative_base()
+Base = declarative_base()
 
 ####################
 # dicts vs objects
@@ -24,13 +27,13 @@ Base=declarative_base()
 # sqlalchemy however offers an object interface, meaning that you write obj.id instead of obj['id']
 # which is admittedly much nicer
 # however we still need to deal with dictionaries if only for the xmlrpc layer
-# 
-# here are a few utilities for this 
-# 
+#
+# here are a few utilities for this
+#
 # (*) first off, when an old pieve of code needs to be used as-is, if only temporarily, the simplest trick
 # is to use obj.__dict__
 # this behaves exactly like required, i.e. obj.__dict__['field']='new value' does change obj.field
-# however this depends on sqlalchemy's implementation so it should be avoided 
+# however this depends on sqlalchemy's implementation so it should be avoided
 #
 # (*) second, when an object needs to be exposed to the xmlrpc layer, we need to convert it into a dict
 # remember though that writing the resulting dictionary won't change the object
@@ -45,49 +48,18 @@ Base=declarative_base()
 # (*) finally for converting a dictionary into an sqlalchemy object, we provide
 # obj.load_from_dict(dict)
 
-class AlchemyObj:
-    def __iter__(self): 
+
+class AlchemyObj(Record):
+
+    def __iter__(self):
         self._i = iter(object_mapper(self).columns)
-        return self 
-    def next(self): 
+        return self
+
+    def __next__(self):
         n = self._i.next().name
         return n, getattr(self, n)
-    def todict (self):
-        d=self.__dict__
-        keys=[k for k in d.keys() if not k.startswith('_')]
-        return dict ( [ (k,d[k]) for k in keys ] )
-    def load_from_dict (self, d):
-        for (k,v) in d.iteritems():
-            # experimental
-            if isinstance(v, StringTypes) and v.lower() in ['true']: v=True
-            if isinstance(v, StringTypes) and v.lower() in ['false']: v=False
-            setattr(self,k,v)
-    
-    # in addition we provide convenience for converting to and from xml records
-    # for this purpose only, we need the subclasses to define 'fields' as either 
-    # a list or a dictionary
-    def xml_fields (self):
-        fields=self.fields
-        if isinstance(fields,dict): fields=fields.keys()
-        return fields
-
-    def save_as_xml (self):
-        # xxx not sure about the scope here
-        input_dict = dict( [ (key, getattr(self.key), ) for key in self.xml_fields() if getattr(self,key,None) ] )
-        xml_record=XML("<record />")
-        xml_record.parse_dict (input_dict)
-        return xml_record.toxml()
-
-    def dump(self, dump_parents=False):
-        for key in self.fields:
-            if key == 'gid' and self.gid:
-                gid = GID(string=self.gid)
-                print "    %s:" % key
-                gid.dump(8, dump_parents)
-            elif getattr(self,key,None):    
-                print "    %s: %s" % (key, getattr(self,key))
-    
-#    # only intended for debugging 
+
+#    # only intended for debugging
 #    def inspect (self, logger, message=""):
 #        logger.info("%s -- Inspecting AlchemyObj -- attrs"%message)
 #        for k in dir(self):
@@ -103,121 +75,284 @@ class AlchemyObj:
 # various kinds of records are implemented as an inheritance hierarchy
 # RegRecord is the base class for all actual variants
 # a first draft was using 'type' as the discriminator for the inheritance
-# but we had to define another more internal column (classtype) so we 
+# but we had to define another more internal column (classtype) so we
 # accomodate variants in types like authority+am and the like
 
-class RegRecord (Base,AlchemyObj):
-    __tablename__       = 'records'
-    record_id           = Column (Integer, primary_key=True)
+class RegRecord(Base, AlchemyObj):
+    __tablename__ = 'records'
+    record_id = Column(Integer, primary_key=True)
     # this is the discriminator that tells which class to use
-    classtype           = Column (String)
-    type                = Column (String)
-    hrn                 = Column (String)
-    gid                 = Column (String)
-    authority           = Column (String)
-    peer_authority      = Column (String)
-    pointer             = Column (Integer, default=-1)
-    date_created        = Column (DateTime)
-    last_updated        = Column (DateTime)
+    classtype = Column(String)
+    # in a first version type was the discriminator
+    # but that could not accomodate for 'authority+sa' and the like
+    type = Column(String)
+    hrn = Column(String)
+    gid = Column(String)
+    authority = Column(String)
+    peer_authority = Column(String)
+    pointer = Column(Integer, default=-1)
+    date_created = Column(DateTime)
+    last_updated = Column(DateTime)
     # use the 'type' column to decide which subclass the object is of
-    __mapper_args__     = { 'polymorphic_on' : classtype }
-
-    fields = [ 'type', 'hrn', 'gid', 'authority', 'peer_authority' ]
-    def __init__ (self, type=None, hrn=None, gid=None, authority=None, peer_authority=None, 
-                  pointer=None, dict=None):
-        if type:                                self.type=type
-        if hrn:                                 self.hrn=hrn
-        if gid: 
-            if isinstance(gid, StringTypes):    self.gid=gid
-            else:                               self.gid=gid.save_to_string(save_parents=True)
-        if authority:                           self.authority=authority
-        if peer_authority:                      self.peer_authority=peer_authority
-        if pointer:                             self.pointer=pointer
-        if dict:                                self.load_from_dict (dict)
+    __mapper_args__ = {'polymorphic_on': classtype}
+
+    fields = ['type', 'hrn', 'gid', 'authority', 'peer_authority']
+
+    def __init__(self, type=None, hrn=None, gid=None, authority=None, peer_authority=None,
+                 pointer=None, dict=None):
+        if type:
+            self.type = type
+        if hrn:
+            self.hrn = hrn
+        if gid:
+            if isinstance(gid, StringType):
+                self.gid = gid
+            else:
+                self.gid = gid.save_to_string(save_parents=True)
+        if authority:
+            self.authority = authority
+        if peer_authority:
+            self.peer_authority = peer_authority
+        if pointer:
+            self.pointer = pointer
+        if dict:
+            self.load_from_dict(dict)
 
     def __repr__(self):
-        result="[Record id=%s, type=%s, hrn=%s, authority=%s, pointer=%s" % \
-                (self.record_id, self.type, self.hrn, self.authority, self.pointer)
+        result = "<Record id=%s, type=%s, hrn=%s, authority=%s" % \
+            (self.record_id, self.type, self.hrn, self.authority)
+#        for extra in ('pointer', 'email', 'name'):
+#        for extra in ('email', 'name'):
+# displaying names at this point it too dangerous, because of unicode
+        for extra in ('email'):
+            if hasattr(self, extra):
+                result += " {}={},".format(extra, getattr(self, extra))
         # skip the uniform '--- BEGIN CERTIFICATE --' stuff
-        if self.gid: result+=" gid=%s..."%self.gid[28:36]
-        else: result+=" nogid"
-        result += "]"
+        if self.gid:
+            result += " gid=%s..." % self.gid[28:36]
+        else:
+            result += " nogid"
+        result += ">"
         return result
 
-    @validates ('gid')
-    def validate_gid (self, key, gid):
-        if gid is None:                     return
-        elif isinstance(gid, StringTypes):  return gid
-        else:                               return gid.save_to_string(save_parents=True)
+    # shortcut - former implem. was record-based
+    def get(self, field, default):
+        return getattr(self, field, default)
+
+    @validates('gid')
+    def validate_gid(self, key, gid):
+        if gid is None:
+            return
+        elif isinstance(gid, StringType):
+            return gid
+        else:
+            return gid.save_to_string(save_parents=True)
+
+    def validate_datetime(self, key, incoming):
+        if isinstance(incoming, datetime):
+            return incoming
+        elif isinstance(incoming, (int, float)):
+            return datetime.fromtimestamp(incoming)
+        else:
+            logger.info("Cannot validate datetime for key %s with input %s" %
+                        (key, incoming))
+
+    @validates('date_created')
+    def validate_date_created(self, key, incoming):
+        return self.validate_datetime(key, incoming)
 
-    # xxx - there might be smarter ways to handle get/set'ing gid using validation hooks 
-    def get_gid_object (self):
-        if not self.gid: return None
-        else: return GID(string=self.gid)
+    @validates('last_updated')
+    def validate_last_updated(self, key, incoming):
+        return self.validate_datetime(key, incoming)
 
-    def just_created (self):
-        now=datetime.now()
-        self.date_created=now
-        self.last_updated=now
+    # xxx - there might be smarter ways to handle get/set'ing gid using
+    # validation hooks
+    def get_gid_object(self):
+        if not self.gid:
+            return None
+        else:
+            return GID(string=self.gid)
 
-    def just_updated (self):
-        now=datetime.now()
-        self.last_updated=now
+    def just_created(self):
+        now = datetime.utcnow()
+        self.date_created = now
+        self.last_updated = now
+
+    def just_updated(self):
+        now = datetime.utcnow()
+        self.last_updated = now
+
+# cross-relations tables
+# authority x user (pis) association
+authority_pi_table = \
+    Table('authority_pi', Base.metadata,
+          Column('authority_id', Integer, ForeignKey(
+              'records.record_id'), primary_key=True),
+          Column('pi_id', Integer, ForeignKey(
+              'records.record_id'), primary_key=True),
+          )
+# slice x user (researchers) association
+slice_researcher_table = \
+    Table('slice_researcher', Base.metadata,
+          Column('slice_id', Integer, ForeignKey(
+              'records.record_id'), primary_key=True),
+          Column('researcher_id', Integer, ForeignKey(
+              'records.record_id'), primary_key=True),
+          )
 
 ##############################
-class RegAuthority (RegRecord):
-    __tablename__       = 'authorities'
-    __mapper_args__     = { 'polymorphic_identity' : 'authority' }
-    record_id           = Column (Integer, ForeignKey ("records.record_id"), primary_key=True)
-    
+# all subclasses define a convenience constructor with a default value for type,
+# and when applicable a way to define local fields in a kwd=value argument
+####################
+
+
+class RegAuthority(RegRecord):
+    __tablename__ = 'authorities'
+    __mapper_args__ = {'polymorphic_identity': 'authority'}
+    record_id = Column(Integer, ForeignKey(
+        "records.record_id"), primary_key=True)
+    # extensions come here
+    name = Column('name', String)
+    # extensions come here
+    reg_pis             = relationship \
+        ('RegUser',
+         secondary=authority_pi_table,
+         primaryjoin=RegRecord.record_id == authority_pi_table.c.authority_id,
+         secondaryjoin=RegRecord.record_id == authority_pi_table.c.pi_id,
+         backref='reg_authorities_as_pi',
+         )
+
+    def __init__(self, **kwds):
+        # handle local settings
+        if 'name' in kwds:
+            self.name = kwds.pop('name')
+        # fill in type if not previously set
+        if 'type' not in kwds:
+            kwds['type'] = 'authority'
+        # base class constructor
+        RegRecord.__init__(self, **kwds)
+
     # no proper data yet, just hack the typename
-    def __repr__ (self):
-        return RegRecord.__repr__(self).replace("Record","Authority")
+    def __repr__(self):
+        result = RegRecord.__repr__(self).replace("Record", "Authority")
+# here again trying to display names that can be utf8 is too dangerous
+#        result.replace(">", " name={}>".format(self.name))
+        return result
 
-##############################
-class RegSlice (RegRecord):
-    __tablename__       = 'slices'
-    __mapper_args__     = { 'polymorphic_identity' : 'slice' }
-    record_id           = Column (Integer, ForeignKey ("records.record_id"), primary_key=True)
-    
-    def __repr__ (self):
-        return RegRecord.__repr__(self).replace("Record","Slice")
+    def update_pis(self, pi_hrns, dbsession):
+        # strip that in case we have <researcher> words </researcher>
+        pi_hrns = [x.strip() for x in pi_hrns]
+        request = dbsession.query(RegUser).filter(RegUser.hrn.in_(pi_hrns))
+        logger.info("RegAuthority.update_pis: %d incoming pis, %d matches found"
+                    % (len(pi_hrns), request.count()))
+        pis = dbsession.query(RegUser).filter(RegUser.hrn.in_(pi_hrns)).all()
+        self.reg_pis = pis
 
-##############################
-class RegNode (RegRecord):
-    __tablename__       = 'nodes'
-    __mapper_args__     = { 'polymorphic_identity' : 'node' }
-    record_id           = Column (Integer, ForeignKey ("records.record_id"), primary_key=True)
-    
-    def __repr__ (self):
-        return RegRecord.__repr__(self).replace("Record","Node")
+####################
 
-##############################
-class RegUser (RegRecord):
-    __tablename__       = 'users'
+
+class RegSlice(RegRecord):
+    __tablename__ = 'slices'
+    __mapper_args__ = {'polymorphic_identity': 'slice'}
+    record_id = Column(Integer, ForeignKey(
+        "records.record_id"), primary_key=True)
+    # extensions come here
+    reg_researchers     = relationship \
+        ('RegUser',
+         secondary=slice_researcher_table,
+         primaryjoin=RegRecord.record_id == slice_researcher_table.c.slice_id,
+         secondaryjoin=RegRecord.record_id == slice_researcher_table.c.researcher_id,
+         backref='reg_slices_as_researcher',
+         )
+
+    def __init__(self, **kwds):
+        if 'type' not in kwds:
+            kwds['type'] = 'slice'
+        RegRecord.__init__(self, **kwds)
+
+    def __repr__(self):
+        return RegRecord.__repr__(self).replace("Record", "Slice")
+
+    def update_researchers(self, researcher_hrns, dbsession):
+        # strip that in case we have <researcher> words </researcher>
+        researcher_hrns = [x.strip() for x in researcher_hrns]
+        request = dbsession.query(RegUser).filter(
+            RegUser.hrn.in_(researcher_hrns))
+        logger.info("RegSlice.update_researchers: %d incoming researchers, %d matches found"
+                    % (len(researcher_hrns), request.count()))
+        researchers = dbsession.query(RegUser).filter(
+            RegUser.hrn.in_(researcher_hrns)).all()
+        self.reg_researchers = researchers
+
+    # when dealing with credentials, we need to retrieve the PIs attached to a slice
+    # WARNING: with the move to passing dbsessions around, we face a glitch here because this
+    # helper function is called from the trust/ area that
+    def get_pis(self):
+        from sqlalchemy.orm import sessionmaker
+        Session = sessionmaker()
+        dbsession = Session.object_session(self)
+        from sfa.util.xrn import get_authority
+        authority_hrn = get_authority(self.hrn)
+        auth_record = dbsession.query(
+            RegAuthority).filter_by(hrn=authority_hrn).first()
+        return auth_record.reg_pis
+
+    @validates('expires')
+    def validate_expires(self, key, incoming):
+        return self.validate_datetime(key, incoming)
+
+####################
+
+
+class RegNode(RegRecord):
+    __tablename__ = 'nodes'
+    __mapper_args__ = {'polymorphic_identity': 'node'}
+    record_id = Column(Integer, ForeignKey(
+        "records.record_id"), primary_key=True)
+
+    def __init__(self, **kwds):
+        if 'type' not in kwds:
+            kwds['type'] = 'node'
+        RegRecord.__init__(self, **kwds)
+
+    def __repr__(self):
+        return RegRecord.__repr__(self).replace("Record", "Node")
+
+####################
+
+
+class RegUser(RegRecord):
+    __tablename__ = 'users'
     # these objects will have type='user' in the records table
-    __mapper_args__     = { 'polymorphic_identity' : 'user' }
-    record_id           = Column (Integer, ForeignKey ("records.record_id"), primary_key=True)
-    email               = Column ('email', String)
+    __mapper_args__ = {'polymorphic_identity': 'user'}
+    record_id = Column(Integer, ForeignKey(
+        "records.record_id"), primary_key=True)
+    # extensions come here
+    email = Column('email', String)
     # can't use name 'keys' here because when loading from xml we're getting
-    # a 'keys' tag, and assigning a list of strings in a reference column like this crashes
-    reg_keys                = relationship ('RegKey', backref='reg_user')
-    
-    def __init__ (self, **kwds):
+    # a 'keys' tag, and assigning a list of strings in a reference column like
+    # this crashes
+    reg_keys            = relationship \
+        ('RegKey', backref='reg_user',
+         cascade="all, delete, delete-orphan",
+         )
+
+    # so we can use RegUser (email=.., hrn=..) and the like
+    def __init__(self, **kwds):
         # handle local settings
-        if 'email' in kwds: self.email=kwds.pop('email')
-        # fill in type if not previously set
-        if 'type' not in kwds: kwds['type']='user'
+        if 'email' in kwds:
+            self.email = kwds.pop('email')
+        if 'type' not in kwds:
+            kwds['type'] = 'user'
         RegRecord.__init__(self, **kwds)
 
     # append stuff at the end of the record __repr__
-    def __repr__ (self): 
-        result = RegRecord.__repr__(self).replace("Record","User")
-        result.replace ("]"," email=%s"%self.email)
-        result += "]"
+    def __repr__(self):
+        result = RegRecord.__repr__(self).replace("Record", "User")
+        result.replace(">", " email={}>".format(self.email))
         return result
-    
-    @validates('email') 
+
+    @validates('email')
     def validate_email(self, key, address):
         assert '@' in address
         return address
@@ -225,27 +360,119 @@ class RegUser (RegRecord):
 ####################
 # xxx tocheck : not sure about eager loading of this one
 # meaning, when querying the whole records, we expect there should
-# be a single query to fetch all the keys 
-class RegKey (Base):
-    __tablename__       = 'keys'
-    key_id              = Column (Integer, primary_key=True)
-    record_id             = Column (Integer, ForeignKey ("records.record_id"))
-    key                 = Column (String)
-    pointer             = Column (Integer, default = -1)
-    
-    def __init__ (self, key, pointer=None):
-        self.key=key
-        if pointer: self.pointer=pointer
-
-    def __repr__ (self):
-        result="[key key=%s..."%self.key[8:16]
-        try:    result += " user=%s"%self.user.record_id
-        except: result += " <orphan>"
-        result += "]"
+# be a single query to fetch all the keys
+# or, is it enough that we issue a single query to retrieve all the keys
+
+
+class RegKey(Base):
+    __tablename__ = 'keys'
+    key_id = Column(Integer, primary_key=True)
+    record_id = Column(Integer, ForeignKey("records.record_id"))
+    key = Column(String)
+    pointer = Column(Integer, default=-1)
+
+    def __init__(self, key, pointer=None):
+        self.key = key
+        if pointer:
+            self.pointer = pointer
+
+    def __repr__(self):
+        result = "<key id=%s key=%s..." % (self.key_id, self.key[8:16],)
+        try:
+            result += " user=%s" % self.reg_user.record_id
+        except:
+            result += " no-user"
+        result += ">"
+        return result
+
+
+class SliverAllocation(Base, AlchemyObj):
+    __tablename__ = 'sliver_allocation'
+    sliver_id = Column(String, primary_key=True)
+    client_id = Column(String)
+    component_id = Column(String)
+    slice_urn = Column(String)
+    allocation_state = Column(String)
+
+    def __init__(self, **kwds):
+        if 'sliver_id' in kwds:
+            self.sliver_id = kwds['sliver_id']
+        if 'client_id' in kwds:
+            self.client_id = kwds['client_id']
+        if 'component_id' in kwds:
+            self.component_id = kwds['component_id']
+        if 'slice_urn' in kwds:
+            self.slice_urn = kwds['slice_urn']
+        if 'allocation_state' in kwds:
+            self.allocation_state = kwds['allocation_state']
+
+    def __repr__(self):
+        result = "<sliver_allocation sliver_id=%s allocation_state=%s"\
+                 % (self.sliver_id, self.allocation_state)
         return result
 
+    @validates('allocation_state')
+    def validate_allocation_state(self, key, state):
+        allocation_states = ['geni_unallocated',
+                             'geni_allocated', 'geni_provisioned']
+        assert state in allocation_states
+        return state
+
+    @staticmethod
+    def set_allocations(sliver_ids, state, dbsession):
+        if not isinstance(sliver_ids, list):
+            sliver_ids = [sliver_ids]
+        sliver_state_updated = {}
+        constraint = SliverAllocation.sliver_id.in_(sliver_ids)
+        sliver_allocations = dbsession.query(
+            SliverAllocation).filter(constraint)
+        sliver_ids_found = []
+        for sliver_allocation in sliver_allocations:
+            sliver_allocation.allocation_state = state
+            sliver_ids_found.append(sliver_allocation.sliver_id)
+
+        # Some states may not have been updated becuase no sliver allocation state record
+        # exists for the sliver. Insert new allocation records for these slivers and set
+        # it to geni_allocated.
+        sliver_ids_not_found = set(sliver_ids).difference(sliver_ids_found)
+        for sliver_id in sliver_ids_not_found:
+            record = SliverAllocation(
+                sliver_id=sliver_id, allocation_state=state)
+            dbsession.add(record)
+        dbsession.commit()
+
+    @staticmethod
+    def delete_allocations(sliver_ids, dbsession):
+        if not isinstance(sliver_ids, list):
+            sliver_ids = [sliver_ids]
+        constraint = SliverAllocation.sliver_id.in_(sliver_ids)
+        sliver_allocations = dbsession.query(
+            SliverAllocation).filter(constraint)
+        for sliver_allocation in sliver_allocations:
+            dbsession.delete(sliver_allocation)
+        dbsession.commit()
+
+    def sync(self, dbsession):
+        constraints = [SliverAllocation.sliver_id == self.sliver_id]
+        results = dbsession.query(SliverAllocation).filter(and_(*constraints))
+        records = []
+        for result in results:
+            records.append(result)
+
+        if not records:
+            dbsession.add(self)
+        else:
+            record = records[0]
+            record.sliver_id = self.sliver_id
+            record.client_id = self.client_id
+            record.component_id = self.component_id
+            record.slice_urn = self.slice_urn
+            record.allocation_state = self.allocation_state
+        dbsession.commit()
+
+
 ##############################
-# although the db needs of course to be reachable,
+# although the db needs of course to be reachable for the following functions
 # the schema management functions are here and not in alchemy
 # because the actual details of the classes need to be known
 # migrations: this code has no notion of the previous versions
@@ -256,40 +483,106 @@ def init_tables(engine):
     logger.info("Initializing db schema from current/latest model")
     Base.metadata.create_all(engine)
 
+
 def drop_tables(engine):
     logger.info("Dropping tables from current/latest model")
     Base.metadata.drop_all(engine)
 
 ##############################
 # create a record of the right type from either a dict or an xml string
-def make_record (dict={}, xml=""):
-    if dict:    return make_record_dict (dict)
-    elif xml:   return make_record_xml (xml)
-    else:       raise Exception("make_record has no input")
+
+
+def make_record(dict=None, xml=""):
+    if dict is None:
+        dict = {}
+    if dict:
+        return make_record_dict(dict)
+    elif xml:
+        return make_record_xml(xml)
+    else:
+        raise Exception("make_record has no input")
 
 # convert an incoming record - typically from xmlrpc - into an object
-def make_record_dict (record_dict):
+
+
+def make_record_dict(record_dict):
     assert ('type' in record_dict)
-    type=record_dict['type'].split('+')[0]
-    if type=='authority':
-        result=RegAuthority (dict=record_dict)
-    elif type=='user':
-        result=RegUser (dict=record_dict)
-    elif type=='slice':
-        result=RegSlice (dict=record_dict)
-    elif type=='node':
-        result=RegNode (dict=record_dict)
+    type = record_dict['type'].split('+')[0]
+    if type == 'authority':
+        result = RegAuthority(dict=record_dict)
+    elif type == 'user':
+        result = RegUser(dict=record_dict)
+    elif type == 'slice':
+        result = RegSlice(dict=record_dict)
+    elif type == 'node':
+        result = RegNode(dict=record_dict)
     else:
         logger.debug("Untyped RegRecord instance")
-        result=RegRecord (dict=record_dict)
-    logger.info ("converting dict into Reg* with type=%s"%type)
-    logger.info ("returning=%s"%result)
+        result = RegRecord(dict=record_dict)
+    logger.info("converting dict into Reg* with type=%s" % type)
+    logger.info("returning=%s" % result)
     # xxx todo
     # register non-db attributes in an extensions field
     return result
-        
-def make_record_xml (xml):
-    xml_record = XML(xml)
-    xml_dict = xml_record.todict()
-    logger.info("load from xml, keys=%s"%xml_dict.keys())
-    return make_record_dict (xml_dict)
+
+
+def make_record_xml(xml_str):
+    xml = XML(xml_str)
+    xml_dict = xml.todict()
+    logger.info("load from xml, keys=%s" % list(xml_dict.keys()))
+    return make_record_dict(xml_dict)
+
+####################
+# augment local records with data from builtin relationships
+# expose related objects as a list of hrns
+# we pick names that clearly won't conflict with the ones used in the old approach,
+# were the relationships data came from the testbed side
+# for each type, a dict of the form {<field-name-exposed-in-record>:<alchemy_accessor_name>}
+# so after that, an 'authority' record will e.g. have a 'reg-pis' field
+# with the hrns of its pi-users
+augment_map = {'authority': {'reg-pis': 'reg_pis', },
+               'slice': {'reg-researchers': 'reg_researchers', },
+               'user': {'reg-pi-authorities': 'reg_authorities_as_pi',
+                        'reg-slices': 'reg_slices_as_researcher', },
+               }
+
+
+# xxx mystery
+# the way we use sqlalchemy might be a little wrong
+# in any case what has been observed is that (Reg)Records as returned by an sqlalchemy
+# query not always have their __dict__ properly adjusted
+# typically a RegAuthority object would have its object.name set properly, but
+# object.__dict__ has no 'name' key
+# which is an issue because we rely on __dict__ for many things, in particular this
+# is what gets exposed to the drivers (this is historical and dates back before sqlalchemy)
+# so it is recommended to always run this function that will make sure
+# that such built-in fields are properly set in __dict__ too
+#
+def augment_with_sfa_builtins(local_record):
+    # don't ruin the import of that file in a client world
+    from sfa.util.xrn import Xrn
+    # add a 'urn' field
+    setattr(local_record, 'reg-urn',
+            Xrn(xrn=local_record.hrn, type=local_record.type).urn)
+    # users have keys and this is needed to synthesize 'users' sent over to
+    # CreateSliver
+    fields_to_check = []
+    if local_record.type == 'user':
+        user_keys = [key.key for key in local_record.reg_keys]
+        setattr(local_record, 'reg-keys', user_keys)
+        fields_to_check = ['email']
+    elif local_record.type == 'authority':
+        fields_to_check = ['name']
+    for field in fields_to_check:
+        if not field in local_record.__dict__:
+            logger.debug("augment_with_sfa_builtins: hotfixing missing '{}' in {}"
+                         .format(field, local_record.hrn))
+            local_record.__dict__[field] = getattr(local_record, field)
+    # search in map according to record type
+    type_map = augment_map.get(local_record.type, {})
+    # use type-dep. map to do the job
+    for (field_name, attribute) in list(type_map.items()):
+        # get related objects
+        related_records = getattr(local_record, attribute, [])
+        hrns = [r.hrn for r in related_records]
+        setattr(local_record, field_name, hrns)