]> git.decadent.org.uk Git - dak.git/blobdiff - daklib/dbconn.py
Merge branch 'master' into dbtests
[dak.git] / daklib / dbconn.py
index 4abfcc596f934fdefa52b4381d26a9e147bfc4f7..7221d243f226970ad91fcfce89992098d04dbe7a 100755 (executable)
@@ -38,6 +38,14 @@ import re
 import psycopg2
 import traceback
 import commands
+
+try:
+    # python >= 2.6
+    import json
+except:
+    # python <= 2.5
+    import simplejson as json
+
 from datetime import datetime, timedelta
 from errno import ENOENT
 from tempfile import mkstemp, mkdtemp
@@ -45,8 +53,10 @@ from tempfile import mkstemp, mkdtemp
 from inspect import getargspec
 
 import sqlalchemy
-from sqlalchemy import create_engine, Table, MetaData, Column, Integer
-from sqlalchemy.orm import sessionmaker, mapper, relation
+from sqlalchemy import create_engine, Table, MetaData, Column, Integer, desc, \
+    Text, ForeignKey
+from sqlalchemy.orm import sessionmaker, mapper, relation, object_session, \
+    backref, MapperExtension, EXT_CONTINUE, object_mapper
 from sqlalchemy import types as sqltypes
 
 # Don't remove this, we re-export the exceptions to scripts which import us
@@ -57,7 +67,17 @@ from sqlalchemy.orm.exc import NoResultFound
 # in the database
 from config import Config
 from textutils import fix_maintainer
-from dak_exceptions import NoSourceFieldError
+from dak_exceptions import DBUpdateError, NoSourceFieldError
+
+# suppress some deprecation warnings in squeeze related to sqlalchemy
+import warnings
+warnings.filterwarnings('ignore', \
+    "The SQLAlchemy PostgreSQL dialect has been renamed from 'postgres' to 'postgresql'.*", \
+    SADeprecationWarning)
+# TODO: sqlalchemy needs some extra configuration to correctly reflect
+# the ind_deb_contents_* indexes - we ignore the warnings at the moment
+warnings.filterwarnings("ignore", 'Predicate of partial index', SAWarning)
+
 
 ################################################################################
 
@@ -146,7 +166,196 @@ __all__.append('session_wrapper')
 
 ################################################################################
 
-class Architecture(object):
+class ORMObject(object):
+    """
+    ORMObject is a base class for all ORM classes mapped by SQLalchemy. All
+    derived classes must implement the properties() method.
+    """
+
+    def properties(self):
+        '''
+        This method should be implemented by all derived classes and returns a
+        list of the important properties. The properties 'created' and
+        'modified' will be added automatically. A suffix '_count' should be
+        added to properties that are lists or query objects. The most important
+        property name should be returned as the first element in the list
+        because it is used by repr().
+        '''
+        return []
+
+    def json(self):
+        '''
+        Returns a JSON representation of the object based on the properties
+        returned from the properties() method.
+        '''
+        data = {}
+        # add created and modified
+        all_properties = self.properties() + ['created', 'modified']
+        for property in all_properties:
+            # check for list or query
+            if property[-6:] == '_count':
+                real_property = property[:-6]
+                if not hasattr(self, real_property):
+                    continue
+                value = getattr(self, real_property)
+                if hasattr(value, '__len__'):
+                    # list
+                    value = len(value)
+                elif hasattr(value, 'count'):
+                    # query
+                    value = value.count()
+                else:
+                    raise KeyError('Do not understand property %s.' % property)
+            else:
+                if not hasattr(self, property):
+                    continue
+                # plain object
+                value = getattr(self, property)
+                if value is None:
+                    # skip None
+                    continue
+                elif isinstance(value, ORMObject):
+                    # use repr() for ORMObject types
+                    value = repr(value)
+                else:
+                    # we want a string for all other types because json cannot
+                    # encode everything
+                    value = str(value)
+            data[property] = value
+        return json.dumps(data)
+
+    def classname(self):
+        '''
+        Returns the name of the class.
+        '''
+        return type(self).__name__
+
+    def __repr__(self):
+        '''
+        Returns a short string representation of the object using the first
+        element from the properties() method.
+        '''
+        primary_property = self.properties()[0]
+        value = getattr(self, primary_property)
+        return '<%s %s>' % (self.classname(), str(value))
+
+    def __str__(self):
+        '''
+        Returns a human readable form of the object using the properties()
+        method.
+        '''
+        return '<%s %s>' % (self.classname(), self.json())
+
+    def not_null_constraints(self):
+        '''
+        Returns a list of properties that must be not NULL. Derived classes
+        should override this method if needed.
+        '''
+        return []
+
+    validation_message = \
+        "Validation failed because property '%s' must not be empty in object\n%s"
+
+    def validate(self):
+        '''
+        This function validates the not NULL constraints as returned by
+        not_null_constraints(). It raises the DBUpdateError exception if
+        validation fails.
+        '''
+        for property in self.not_null_constraints():
+            # TODO: It is a bit awkward that the mapper configuration allow
+            # directly setting the numeric _id columns. We should get rid of it
+            # in the long run.
+            if hasattr(self, property + '_id') and \
+                getattr(self, property + '_id') is not None:
+                continue
+            if not hasattr(self, property) or getattr(self, property) is None:
+                raise DBUpdateError(self.validation_message % \
+                    (property, str(self)))
+
+    @classmethod
+    @session_wrapper
+    def get(cls, primary_key,  session = None):
+        '''
+        This is a support function that allows getting an object by its primary
+        key.
+
+        Architecture.get(3[, session])
+
+        instead of the more verbose
+
+        session.query(Architecture).get(3)
+        '''
+        return session.query(cls).get(primary_key)
+
+    def session(self, replace = False):
+        '''
+        Returns the current session that is associated with the object. May
+        return None is object is in detached state.
+        '''
+
+        return object_session(self)
+
+    def clone(self, session = None):
+        '''
+        Clones the current object in a new session and returns the new clone. A
+        fresh session is created if the optional session parameter is not
+        provided. The function will fail if a session is provided and has
+        unflushed changes.
+
+        RATIONALE: SQLAlchemy's session is not thread safe. This method clones
+        an existing object to allow several threads to work with their own
+        instances of an ORMObject.
+
+        WARNING: Only persistent (committed) objects can be cloned. Changes
+        made to the original object that are not committed yet will get lost.
+        The session of the new object will always be rolled back to avoid
+        ressource leaks.
+        '''
+
+        if self.session() is None:
+            raise RuntimeError( \
+                'Method clone() failed for detached object:\n%s' % self)
+        self.session().flush()
+        mapper = object_mapper(self)
+        primary_key = mapper.primary_key_from_instance(self)
+        object_class = self.__class__
+        if session is None:
+            session = DBConn().session()
+        elif len(session.new) + len(session.dirty) + len(session.deleted) > 0:
+            raise RuntimeError( \
+                'Method clone() failed due to unflushed changes in session.')
+        new_object = session.query(object_class).get(primary_key)
+        session.rollback()
+        if new_object is None:
+            raise RuntimeError( \
+                'Method clone() failed for non-persistent object:\n%s' % self)
+        return new_object
+
+__all__.append('ORMObject')
+
+################################################################################
+
+class Validator(MapperExtension):
+    '''
+    This class calls the validate() method for each instance for the
+    'before_update' and 'before_insert' events. A global object validator is
+    used for configuring the individual mappers.
+    '''
+
+    def before_update(self, mapper, connection, instance):
+        instance.validate()
+        return EXT_CONTINUE
+
+    def before_insert(self, mapper, connection, instance):
+        instance.validate()
+        return EXT_CONTINUE
+
+validator = Validator()
+
+################################################################################
+
+class Architecture(ORMObject):
     def __init__(self, arch_string = None, description = None):
         self.arch_string = arch_string
         self.description = description
@@ -163,8 +372,11 @@ class Architecture(object):
         # This signals to use the normal comparison operator
         return NotImplemented
 
-    def __repr__(self):
-        return '<Architecture %s>' % self.arch_string
+    def properties(self):
+        return ['arch_string', 'arch_id', 'suites_count']
+
+    def not_null_constraints(self):
+        return ['arch_string']
 
 __all__.append('Architecture')
 
@@ -193,6 +405,7 @@ def get_architecture(architecture, session=None):
 
 __all__.append('get_architecture')
 
+# TODO: should be removed because the implementation is too trivial
 @session_wrapper
 def get_architecture_suites(architecture, session=None):
     """
@@ -209,13 +422,7 @@ def get_architecture_suites(architecture, session=None):
     @return: list of Suite objects for the given name (may be empty)
     """
 
-    q = session.query(Suite)
-    q = q.join(SuiteArchitecture)
-    q = q.join(Architecture).filter_by(arch_string=architecture).order_by('suite_name')
-
-    ret = q.all()
-
-    return ret
+    return get_architecture(architecture, session).suites
 
 __all__.append('get_architecture_suites')
 
@@ -259,34 +466,41 @@ __all__.append('get_archive')
 
 ################################################################################
 
-class BinAssociation(object):
-    def __init__(self, *args, **kwargs):
-        pass
-
-    def __repr__(self):
-        return '<BinAssociation %s (%s, %s)>' % (self.ba_id, self.binary, self.suite)
-
-__all__.append('BinAssociation')
+class BinContents(ORMObject):
+    def __init__(self, file = None, binary = None):
+        self.file = file
+        self.binary = binary
 
-################################################################################
-
-class BinContents(object):
-    def __init__(self, *args, **kwargs):
-        pass
-
-    def __repr__(self):
-        return '<BinContents (%s, %s)>' % (self.binary, self.filename)
+    def properties(self):
+        return ['file', 'binary']
 
 __all__.append('BinContents')
 
 ################################################################################
 
-class DBBinary(object):
-    def __init__(self, *args, **kwargs):
-        pass
-
-    def __repr__(self):
-        return '<DBBinary %s (%s, %s)>' % (self.package, self.version, self.architecture)
+class DBBinary(ORMObject):
+    def __init__(self, package = None, source = None, version = None, \
+        maintainer = None, architecture = None, poolfile = None, \
+        binarytype = 'deb'):
+        self.package = package
+        self.source = source
+        self.version = version
+        self.maintainer = maintainer
+        self.architecture = architecture
+        self.poolfile = poolfile
+        self.binarytype = binarytype
+
+    def properties(self):
+        return ['package', 'version', 'maintainer', 'source', 'architecture', \
+            'poolfile', 'binarytype', 'fingerprint', 'install_date', \
+            'suites_count', 'binary_id', 'contents_count']
+
+    def not_null_constraints(self):
+        return ['package', 'version', 'maintainer', 'source',  'poolfile', \
+            'binarytype']
+
+    def get_component_name(self):
+        return self.poolfile.location.component.component_name
 
 __all__.append('DBBinary')
 
@@ -302,129 +516,42 @@ def get_suites_binary_in(package, session=None):
     @return: list of Suite objects for the given package
     """
 
-    return session.query(Suite).join(BinAssociation).join(DBBinary).filter_by(package=package).all()
+    return session.query(Suite).filter(Suite.binaries.any(DBBinary.package == package)).all()
 
 __all__.append('get_suites_binary_in')
 
 @session_wrapper
-def get_binary_from_id(binary_id, session=None):
-    """
-    Returns DBBinary object for given C{id}
-
-    @type binary_id: int
-    @param binary_id: Id of the required binary
-
-    @type session: Session
-    @param session: Optional SQLA session object (a temporary one will be
-    generated if not supplied)
-
-    @rtype: DBBinary
-    @return: DBBinary object for the given binary (None if not present)
-    """
-
-    q = session.query(DBBinary).filter_by(binary_id=binary_id)
-
-    try:
-        return q.one()
-    except NoResultFound:
-        return None
-
-__all__.append('get_binary_from_id')
-
-@session_wrapper
-def get_binaries_from_name(package, version=None, architecture=None, session=None):
-    """
-    Returns list of DBBinary objects for given C{package} name
+def get_component_by_package_suite(package, suite_list, arch_list=[], session=None):
+    '''
+    Returns the component name of the newest binary package in suite_list or
+    None if no package is found. The result can be optionally filtered by a list
+    of architecture names.
 
     @type package: str
     @param package: DBBinary package name to search for
 
-    @type version: str or None
-    @param version: Version to search for (or None)
-
-    @type architecture: str, list or None
-    @param architecture: Architectures to limit to (or None if no limit)
-
-    @type session: Session
-    @param session: Optional SQL session object (a temporary one will be
-    generated if not supplied)
-
-    @rtype: list
-    @return: list of DBBinary objects for the given name (may be empty)
-    """
-
-    q = session.query(DBBinary).filter_by(package=package)
-
-    if version is not None:
-        q = q.filter_by(version=version)
+    @type suite_list: list of str
+    @param suite_list: list of suite_name items
 
-    if architecture is not None:
-        if not isinstance(architecture, list):
-            architecture = [architecture]
-        q = q.join(Architecture).filter(Architecture.arch_string.in_(architecture))
-
-    ret = q.all()
-
-    return ret
-
-__all__.append('get_binaries_from_name')
-
-@session_wrapper
-def get_binaries_from_source_id(source_id, session=None):
-    """
-    Returns list of DBBinary objects for given C{source_id}
-
-    @type source_id: int
-    @param source_id: source_id to search for
-
-    @type session: Session
-    @param session: Optional SQL session object (a temporary one will be
-    generated if not supplied)
-
-    @rtype: list
-    @return: list of DBBinary objects for the given name (may be empty)
-    """
+    @type arch_list: list of str
+    @param arch_list: optional list of arch_string items that defaults to []
 
-    return session.query(DBBinary).filter_by(source_id=source_id).all()
+    @rtype: str or NoneType
+    @return: name of component or None
+    '''
 
-__all__.append('get_binaries_from_source_id')
-
-@session_wrapper
-def get_binary_from_name_suite(package, suitename, session=None):
-    ### For dak examine-package
-    ### XXX: Doesn't use object API yet
-
-    sql = """SELECT DISTINCT(b.package), b.version, c.name, su.suite_name
-             FROM binaries b, files fi, location l, component c, bin_associations ba, suite su
-             WHERE b.package='%(package)s'
-               AND b.file = fi.id
-               AND fi.location = l.id
-               AND l.component = c.id
-               AND ba.bin=b.id
-               AND ba.suite = su.id
-               AND su.suite_name %(suitename)s
-          ORDER BY b.version DESC"""
-
-    return session.execute(sql % {'package': package, 'suitename': suitename})
-
-__all__.append('get_binary_from_name_suite')
-
-@session_wrapper
-def get_binary_components(package, suitename, arch, session=None):
-    # Check for packages that have moved from one component to another
-    query = """SELECT c.name FROM binaries b, bin_associations ba, suite s, location l, component c, architecture a, files f
-    WHERE b.package=:package AND s.suite_name=:suitename
-      AND (a.arch_string = :arch OR a.arch_string = 'all')
-      AND ba.bin = b.id AND ba.suite = s.id AND b.architecture = a.id
-      AND f.location = l.id
-      AND l.component = c.id
-      AND b.file = f.id"""
-
-    vals = {'package': package, 'suitename': suitename, 'arch': arch}
-
-    return session.execute(query, vals)
+    q = session.query(DBBinary).filter_by(package = package). \
+        join(DBBinary.suites).filter(Suite.suite_name.in_(suite_list))
+    if len(arch_list) > 0:
+        q = q.join(DBBinary.architecture). \
+            filter(Architecture.arch_string.in_(arch_list))
+    binary = q.order_by(desc(DBBinary.version)).first()
+    if binary is None:
+        return None
+    else:
+        return binary.get_component_name()
 
-__all__.append('get_binary_components')
+__all__.append('get_component_by_package_suite')
 
 ################################################################################
 
@@ -767,9 +894,9 @@ __all__.append('ChangePendingSource')
 
 ################################################################################
 
-class Component(object):
-    def __init__(self, *args, **kwargs):
-        pass
+class Component(ORMObject):
+    def __init__(self, component_name = None):
+        self.component_name = component_name
 
     def __eq__(self, val):
         if isinstance(val, str):
@@ -783,8 +910,12 @@ class Component(object):
         # This signals to use the normal comparison operator
         return NotImplemented
 
-    def __repr__(self):
-        return '<Component %s>' % self.component_name
+    def properties(self):
+        return ['component_name', 'component_id', 'description', \
+            'location_count', 'meets_dfsg', 'overrides_count']
+
+    def not_null_constraints(self):
+        return ['component_name']
 
 
 __all__.append('Component')
@@ -1064,24 +1195,35 @@ __all__.append('get_dscfiles')
 
 ################################################################################
 
-class PoolFile(object):
-    def __init__(self, *args, **kwargs):
-        pass
-
-    def __repr__(self):
-        return '<PoolFile %s>' % self.filename
+class PoolFile(ORMObject):
+    def __init__(self, filename = None, location = None, filesize = -1, \
+        md5sum = None):
+        self.filename = filename
+        self.location = location
+        self.filesize = filesize
+        self.md5sum = md5sum
 
     @property
     def fullpath(self):
         return os.path.join(self.location.path, self.filename)
 
+    def is_valid(self, filesize = -1, md5sum = None):
+        return self.filesize == long(filesize) and self.md5sum == md5sum
+
+    def properties(self):
+        return ['filename', 'file_id', 'filesize', 'md5sum', 'sha1sum', \
+            'sha256sum', 'location', 'source', 'binary', 'last_used']
+
+    def not_null_constraints(self):
+        return ['filename', 'md5sum', 'location']
+
 __all__.append('PoolFile')
 
 @session_wrapper
 def check_poolfile(filename, filesize, md5sum, location_id, session=None):
     """
     Returns a tuple:
-    (ValidFileFound [boolean or None], PoolFile object or None)
+    (ValidFileFound [boolean], PoolFile object or None)
 
     @type filename: string
     @param filename: the filename of the file to check against the DB
@@ -1097,34 +1239,24 @@ def check_poolfile(filename, filesize, md5sum, location_id, session=None):
 
     @rtype: tuple
     @return: Tuple of length 2.
-                 - If more than one file found with that name: (C{None},  C{None})
                  - If valid pool file found: (C{True}, C{PoolFile object})
                  - If valid pool file not found:
                      - (C{False}, C{None}) if no file found
                      - (C{False}, C{PoolFile object}) if file found with size/md5sum mismatch
     """
 
-    q = session.query(PoolFile).filter_by(filename=filename)
-    q = q.join(Location).filter_by(location_id=location_id)
+    poolfile = session.query(Location).get(location_id). \
+        files.filter_by(filename=filename).first()
+    valid = False
+    if poolfile and poolfile.is_valid(filesize = filesize, md5sum = md5sum):
+        valid = True
 
-    ret = None
-
-    if q.count() > 1:
-        ret = (None, None)
-    elif q.count() < 1:
-        ret = (False, None)
-    else:
-        obj = q.one()
-        if obj.md5sum != md5sum or obj.filesize != int(filesize):
-            ret = (False, obj)
-
-    if ret is None:
-        ret = (True, obj)
-
-    return ret
+    return (valid, poolfile)
 
 __all__.append('check_poolfile')
 
+# TODO: the implementation can trivially be inlined at the place where the
+# function is called
 @session_wrapper
 def get_poolfile_by_id(file_id, session=None):
     """
@@ -1137,41 +1269,10 @@ def get_poolfile_by_id(file_id, session=None):
     @return: either the PoolFile object or None
     """
 
-    q = session.query(PoolFile).filter_by(file_id=file_id)
-
-    try:
-        return q.one()
-    except NoResultFound:
-        return None
+    return session.query(PoolFile).get(file_id)
 
 __all__.append('get_poolfile_by_id')
 
-
-@session_wrapper
-def get_poolfile_by_name(filename, location_id=None, session=None):
-    """
-    Returns an array of PoolFile objects for the given filename and
-    (optionally) location_id
-
-    @type filename: string
-    @param filename: the filename of the file to check against the DB
-
-    @type location_id: int
-    @param location_id: the id of the location to look in (optional)
-
-    @rtype: array
-    @return: array of PoolFile objects
-    """
-
-    q = session.query(PoolFile).filter_by(filename=filename)
-
-    if location_id is not None:
-        q = q.join(Location).filter_by(location_id=location_id)
-
-    return q.all()
-
-__all__.append('get_poolfile_by_name')
-
 @session_wrapper
 def get_poolfile_like_name(filename, session=None):
     """
@@ -1226,12 +1327,16 @@ __all__.append('add_poolfile')
 
 ################################################################################
 
-class Fingerprint(object):
+class Fingerprint(ORMObject):
     def __init__(self, fingerprint = None):
         self.fingerprint = fingerprint
 
-    def __repr__(self):
-        return '<Fingerprint %s>' % self.fingerprint
+    def properties(self):
+        return ['fingerprint', 'fingerprint_id', 'keyring', 'uid', \
+            'binary_reject']
+
+    def not_null_constraints(self):
+        return ['fingerprint']
 
 __all__.append('Fingerprint')
 
@@ -1516,12 +1621,19 @@ __all__.append('get_dbchange')
 
 ################################################################################
 
-class Location(object):
-    def __init__(self, *args, **kwargs):
-        pass
+class Location(ORMObject):
+    def __init__(self, path = None, component = None):
+        self.path = path
+        self.component = component
+        # the column 'type' should go away, see comment at mapper
+        self.archive_type = 'pool'
 
-    def __repr__(self):
-        return '<Location %s (%s)>' % (self.path, self.location_id)
+    def properties(self):
+        return ['path', 'location_id', 'archive_type', 'component', \
+            'files_count']
+
+    def not_null_constraints(self):
+        return ['path', 'archive_type']
 
 __all__.append('Location')
 
@@ -1561,12 +1673,15 @@ __all__.append('get_location')
 
 ################################################################################
 
-class Maintainer(object):
-    def __init__(self, *args, **kwargs):
-        pass
+class Maintainer(ORMObject):
+    def __init__(self, name = None):
+        self.name = name
 
-    def __repr__(self):
-        return '''<Maintainer '%s' (%s)>''' % (self.name, self.maintainer_id)
+    def properties(self):
+        return ['name', 'maintainer_id']
+
+    def not_null_constraints(self):
+        return ['name']
 
     def get_split_maintainer(self):
         if not hasattr(self, 'name') or self.name is None:
@@ -1758,12 +1873,15 @@ __all__.append('get_override')
 
 ################################################################################
 
-class OverrideType(object):
-    def __init__(self, *args, **kwargs):
-        pass
+class OverrideType(ORMObject):
+    def __init__(self, overridetype = None):
+        self.overridetype = overridetype
 
-    def __repr__(self):
-        return '<OverrideType %s>' % self.overridetype
+    def properties(self):
+        return ['overridetype', 'overridetype_id', 'overrides_count']
+
+    def not_null_constraints(self):
+        return ['overridetype']
 
 __all__.append('OverrideType')
 
@@ -1960,9 +2078,16 @@ __all__.append('get_policy_queue_from_path')
 
 ################################################################################
 
-class Priority(object):
-    def __init__(self, *args, **kwargs):
-        pass
+class Priority(ORMObject):
+    def __init__(self, priority = None, level = None):
+        self.priority = priority
+        self.level = level
+
+    def properties(self):
+        return ['priority', 'priority_id', 'level', 'overrides_count']
+
+    def not_null_constraints(self):
+        return ['priority', 'level']
 
     def __eq__(self, val):
         if isinstance(val, str):
@@ -1976,9 +2101,6 @@ class Priority(object):
         # This signals to use the normal comparison operator
         return NotImplemented
 
-    def __repr__(self):
-        return '<Priority %s (%s)>' % (self.priority, self.priority_id)
-
 __all__.append('Priority')
 
 @session_wrapper
@@ -2030,9 +2152,15 @@ __all__.append('get_priorities')
 
 ################################################################################
 
-class Section(object):
-    def __init__(self, *args, **kwargs):
-        pass
+class Section(ORMObject):
+    def __init__(self, section = None):
+        self.section = section
+
+    def properties(self):
+        return ['section', 'section_id', 'overrides_count']
+
+    def not_null_constraints(self):
+        return ['section']
 
     def __eq__(self, val):
         if isinstance(val, str):
@@ -2046,9 +2174,6 @@ class Section(object):
         # This signals to use the normal comparison operator
         return NotImplemented
 
-    def __repr__(self):
-        return '<Section %s>' % self.section
-
 __all__.append('Section')
 
 @session_wrapper
@@ -2100,12 +2225,24 @@ __all__.append('get_sections')
 
 ################################################################################
 
-class DBSource(object):
-    def __init__(self, *args, **kwargs):
-        pass
-
-    def __repr__(self):
-        return '<DBSource %s (%s)>' % (self.source, self.version)
+class DBSource(ORMObject):
+    def __init__(self, source = None, version = None, maintainer = None, \
+        changedby = None, poolfile = None, install_date = None):
+        self.source = source
+        self.version = version
+        self.maintainer = maintainer
+        self.changedby = changedby
+        self.poolfile = poolfile
+        self.install_date = install_date
+
+    def properties(self):
+        return ['source', 'source_id', 'maintainer', 'changedby', \
+            'fingerprint', 'poolfile', 'version', 'suites_count', \
+            'install_date', 'binaries_count']
+
+    def not_null_constraints(self):
+        return ['source', 'version', 'install_date', 'maintainer', \
+            'changedby', 'poolfile', 'install_date']
 
 __all__.append('DBSource')
 
@@ -2136,10 +2273,14 @@ def source_exists(source, source_version, suites = ["any"], session=None):
     """
 
     cnf = Config()
-    ret = 1
+    ret = True
+
+    from daklib.regexes import re_bin_only_nmu
+    orig_source_version = re_bin_only_nmu.sub('', source_version)
 
     for suite in suites:
-        q = session.query(DBSource).filter_by(source=source)
+        q = session.query(DBSource).filter_by(source=source). \
+            filter(DBSource.version.in_([source_version, orig_source_version]))
         if suite != "any":
             # source must exist in suite X, or in some other suite that's
             # mapped to X, recursively... silent-maps are counted too,
@@ -2150,28 +2291,17 @@ def source_exists(source, source_version, suites = ["any"], session=None):
             maps = [ (x[1], x[2]) for x in maps
                             if x[0] == "map" or x[0] == "silent-map" ]
             s = [suite]
-            for x in maps:
-                if x[1] in s and x[0] not in s:
-                    s.append(x[0])
+            for (from_, to) in maps:
+                if from_ in s and to not in s:
+                    s.append(to)
 
-            q = q.join(SrcAssociation).join(Suite)
-            q = q.filter(Suite.suite_name.in_(s))
+            q = q.filter(DBSource.suites.any(Suite.suite_name.in_(s)))
 
-        # Reduce the query results to a list of version numbers
-        ql = [ j.version for j in q.all() ]
-
-        # Try (1)
-        if source_version in ql:
-            continue
-
-        # Try (2)
-        from daklib.regexes import re_bin_only_nmu
-        orig_source_version = re_bin_only_nmu.sub('', source_version)
-        if orig_source_version in ql:
+        if q.count() > 0:
             continue
 
         # No source found so return not ok
-        ret = 0
+        ret = False
 
     return ret
 
@@ -2189,7 +2319,7 @@ def get_suites_source_in(source, session=None):
     @return: list of Suite objects for the given source
     """
 
-    return session.query(Suite).join(SrcAssociation).join(DBSource).filter_by(source=source).all()
+    return session.query(Suite).filter(Suite.sources.any(source=source)).all()
 
 __all__.append('get_suites_source_in')
 
@@ -2228,10 +2358,12 @@ def get_sources_from_name(source, version=None, dm_upload_allowed=None, session=
 
 __all__.append('get_sources_from_name')
 
+# FIXME: This function fails badly if it finds more than 1 source package and
+# its implementation is trivial enough to be inlined.
 @session_wrapper
 def get_source_in_suite(source, suite, session=None):
     """
-    Returns list of DBSource objects for a combination of C{source} and C{suite}.
+    Returns a DBSource object for a combination of C{source} and C{suite}.
 
       - B{source} - source package name, eg. I{mailfilter}, I{bbdb}, I{glibc}
       - B{suite} - a suite name, eg. I{unstable}
@@ -2247,12 +2379,9 @@ def get_source_in_suite(source, suite, session=None):
 
     """
 
-    q = session.query(SrcAssociation)
-    q = q.join('source').filter_by(source=source)
-    q = q.join('suite').filter_by(suite_name=suite)
-
+    q = get_suite(suite, session).get_sources(source)
     try:
-        return q.one().source
+        return q.one()
     except NoResultFound:
         return None
 
@@ -2288,15 +2417,10 @@ def add_dsc_to_db(u, filename, session=None):
 
     source.poolfile_id = entry["files id"]
     session.add(source)
-    session.flush()
 
-    for suite_name in u.pkg.changes["distribution"].keys():
-        sa = SrcAssociation()
-        sa.source_id = source.source_id
-        sa.suite_id = get_suite(suite_name).suite_id
-        session.add(sa)
-
-    session.flush()
+    suite_names = u.pkg.changes["distribution"].keys()
+    source.suites = session.query(Suite). \
+        filter(Suite.suite_name.in_(suite_names)).all()
 
     # Add the source files to the DB (files and dsc_files)
     dscfile = DSCFile()
@@ -2346,8 +2470,6 @@ def add_dsc_to_db(u, filename, session=None):
         df.poolfile_id = files_id
         session.add(df)
 
-    session.flush()
-
     # Add the src_uploaders to the DB
     uploader_ids = [source.maintainer_id]
     if u.pkg.dsc.has_key("uploaders"):
@@ -2417,14 +2539,10 @@ def add_deb_to_db(u, filename, session=None):
 
     # Add and flush object so it has an ID
     session.add(bin)
-    session.flush()
 
-    # Add BinAssociations
-    for suite_name in u.pkg.changes["distribution"].keys():
-        ba = BinAssociation()
-        ba.binary_id = bin.binary_id
-        ba.suite_id = get_suite(suite_name).suite_id
-        session.add(ba)
+    suite_names = u.pkg.changes["distribution"].keys()
+    bin.suites = session.query(Suite). \
+        filter(Suite.suite_name.in_(suite_names)).all()
 
     session.flush()
 
@@ -2452,17 +2570,6 @@ __all__.append('SourceACL')
 
 ################################################################################
 
-class SrcAssociation(object):
-    def __init__(self, *args, **kwargs):
-        pass
-
-    def __repr__(self):
-        return '<SrcAssociation %s (%s, %s)>' % (self.sa_id, self.source, self.suite)
-
-__all__.append('SrcAssociation')
-
-################################################################################
-
 class SrcFormat(object):
     def __init__(self, *args, **kwargs):
         pass
@@ -2501,12 +2608,19 @@ SUITE_FIELDS = [ ('SuiteName', 'suite_name'),
                  ('CopyChanges', 'copychanges'),
                  ('OverrideSuite', 'overridesuite')]
 
-class Suite(object):
-    def __init__(self, *args, **kwargs):
-        pass
+# Why the heck don't we have any UNIQUE constraints in table suite?
+# TODO: Add UNIQUE constraints for appropriate columns.
+class Suite(ORMObject):
+    def __init__(self, suite_name = None, version = None):
+        self.suite_name = suite_name
+        self.version = version
 
-    def __repr__(self):
-        return '<Suite %s>' % self.suite_name
+    def properties(self):
+        return ['suite_name', 'version', 'sources_count', 'binaries_count', \
+            'overrides_count']
+
+    def not_null_constraints(self):
+        return ['suite_name', 'version']
 
     def __eq__(self, val):
         if isinstance(val, str):
@@ -2529,38 +2643,48 @@ class Suite(object):
 
         return "\n".join(ret)
 
-__all__.append('Suite')
+    def get_architectures(self, skipsrc=False, skipall=False):
+        """
+        Returns list of Architecture objects
 
-@session_wrapper
-def get_suite_architecture(suite, architecture, session=None):
-    """
-    Returns a SuiteArchitecture object given C{suite} and ${arch} or None if it
-    doesn't exist
+        @type skipsrc: boolean
+        @param skipsrc: Whether to skip returning the 'source' architecture entry
+        (Default False)
 
-    @type suite: str
-    @param suite: Suite name to search for
+        @type skipall: boolean
+        @param skipall: Whether to skip returning the 'all' architecture entry
+        (Default False)
 
-    @type architecture: str
-    @param architecture: Architecture name to search for
+        @rtype: list
+        @return: list of Architecture objects for the given name (may be empty)
+        """
 
-    @type session: Session
-    @param session: Optional SQL session object (a temporary one will be
-    generated if not supplied)
+        q = object_session(self).query(Architecture).with_parent(self)
+        if skipsrc:
+            q = q.filter(Architecture.arch_string != 'source')
+        if skipall:
+            q = q.filter(Architecture.arch_string != 'all')
+        return q.order_by(Architecture.arch_string).all()
 
-    @rtype: SuiteArchitecture
-    @return: the SuiteArchitecture object or None
-    """
+    def get_sources(self, source):
+        """
+        Returns a query object representing DBSource that is part of C{suite}.
 
-    q = session.query(SuiteArchitecture)
-    q = q.join(Architecture).filter_by(arch_string=architecture)
-    q = q.join(Suite).filter_by(suite_name=suite)
+          - B{source} - source package name, eg. I{mailfilter}, I{bbdb}, I{glibc}
 
-    try:
-        return q.one()
-    except NoResultFound:
-        return None
+        @type source: string
+        @param source: source package name
+
+        @rtype: sqlalchemy.orm.query.Query
+        @return: a query of DBSource
 
-__all__.append('get_suite_architecture')
+        """
+
+        session = object_session(self)
+        return session.query(DBSource).filter_by(source = source). \
+            with_parent(self)
+
+__all__.append('Suite')
 
 @session_wrapper
 def get_suite(suite, session=None):
@@ -2589,15 +2713,7 @@ __all__.append('get_suite')
 
 ################################################################################
 
-class SuiteArchitecture(object):
-    def __init__(self, *args, **kwargs):
-        pass
-
-    def __repr__(self):
-        return '<SuiteArchitecture (%s, %s)>' % (self.suite_id, self.arch_id)
-
-__all__.append('SuiteArchitecture')
-
+# TODO: should be removed because the implementation is too trivial
 @session_wrapper
 def get_suite_architectures(suite, skipsrc=False, skipall=False, session=None):
     """
@@ -2622,19 +2738,7 @@ def get_suite_architectures(suite, skipsrc=False, skipall=False, session=None):
     @return: list of Architecture objects for the given name (may be empty)
     """
 
-    q = session.query(Architecture)
-    q = q.join(SuiteArchitecture)
-    q = q.join(Suite).filter_by(suite_name=suite)
-
-    if skipsrc:
-        q = q.filter(Architecture.arch_string != 'source')
-
-    if skipall:
-        q = q.filter(Architecture.arch_string != 'all')
-
-    q = q.order_by('arch_string')
-
-    return q.all()
+    return get_suite(suite, session).get_architectures(skipsrc, skipall)
 
 __all__.append('get_suite_architectures')
 
@@ -2676,7 +2780,7 @@ __all__.append('get_suite_src_formats')
 
 ################################################################################
 
-class Uid(object):
+class Uid(ORMObject):
     def __init__(self, uid = None, name = None):
         self.uid = uid
         self.name = name
@@ -2693,8 +2797,11 @@ class Uid(object):
         # This signals to use the normal comparison operator
         return NotImplemented
 
-    def __repr__(self):
-        return '<Uid %s (%s)>' % (self.uid, self.name)
+    def properties(self):
+        return ['uid', 'name', 'fingerprint']
+
+    def not_null_constraints(self):
+        return ['uid']
 
 __all__.append('Uid')
 
@@ -2780,7 +2887,9 @@ class DBConn(object):
             'binary_acl',
             'binary_acl_map',
             'build_queue',
+            'build_queue_files',
             'changelogs_text',
+            'changes',
             'component',
             'config',
             'changes_pending_binaries',
@@ -2807,15 +2916,9 @@ class DBConn(object):
             'suite',
             'uid',
             'upload_blocks',
-            # The following tables have primary keys but sqlalchemy
-            # version 0.5 fails to reflect them correctly with database
-            # versions before upgrade #41.
-            #'changes',
-            #'build_queue_files',
         )
 
         tables_no_primary = (
-            'bin_contents',
             'changes_pending_files_map',
             'changes_pending_source_files',
             'changes_pool_files',
@@ -2825,9 +2928,6 @@ class DBConn(object):
             'suite_src_formats',
             'suite_build_queue_copy',
             'udeb_contents',
-            # see the comment above
-            'changes',
-            'build_queue_files',
         )
 
         views = (
@@ -2867,25 +2967,30 @@ class DBConn(object):
             table = Table(table_name, self.db_meta, autoload=True)
             setattr(self, 'tbl_%s' % table_name, table)
 
+        # bin_contents needs special attention until update #41 has been
+        # applied
+        self.tbl_bin_contents = Table('bin_contents', self.db_meta, \
+            Column('file', Text, primary_key = True),
+            Column('binary_id', Integer, ForeignKey('binaries.id'), \
+                primary_key = True),
+            autoload=True, useexisting=True)
+
         for view_name in views:
             view = Table(view_name, self.db_meta, autoload=True)
             setattr(self, 'view_%s' % view_name, view)
 
     def __setupmappers(self):
         mapper(Architecture, self.tbl_architecture,
-               properties = dict(arch_id = self.tbl_architecture.c.id))
+            properties = dict(arch_id = self.tbl_architecture.c.id,
+               suites = relation(Suite, secondary=self.tbl_suite_architectures,
+                   order_by='suite_name',
+                   backref=backref('architectures', order_by='arch_string'))),
+            extension = validator)
 
         mapper(Archive, self.tbl_archive,
                properties = dict(archive_id = self.tbl_archive.c.id,
                                  archive_name = self.tbl_archive.c.name))
 
-        mapper(BinAssociation, self.tbl_bin_associations,
-               properties = dict(ba_id = self.tbl_bin_associations.c.id,
-                                 suite_id = self.tbl_bin_associations.c.suite,
-                                 suite = relation(Suite),
-                                 binary_id = self.tbl_bin_associations.c.bin,
-                                 binary = relation(DBBinary)))
-
         mapper(PendingBinContents, self.tbl_pending_bin_contents,
                properties = dict(contents_id =self.tbl_pending_bin_contents.c.id,
                                  filename = self.tbl_pending_bin_contents.c.filename,
@@ -2924,17 +3029,18 @@ class DBConn(object):
                                  maintainer_id = self.tbl_binaries.c.maintainer,
                                  maintainer = relation(Maintainer),
                                  source_id = self.tbl_binaries.c.source,
-                                 source = relation(DBSource),
+                                 source = relation(DBSource, backref='binaries'),
                                  arch_id = self.tbl_binaries.c.architecture,
                                  architecture = relation(Architecture),
                                  poolfile_id = self.tbl_binaries.c.file,
-                                 poolfile = relation(PoolFile),
+                                 poolfile = relation(PoolFile, backref=backref('binary', uselist = False)),
                                  binarytype = self.tbl_binaries.c.type,
                                  fingerprint_id = self.tbl_binaries.c.sig_fpr,
                                  fingerprint = relation(Fingerprint),
                                  install_date = self.tbl_binaries.c.install_date,
-                                 binassociations = relation(BinAssociation,
-                                                            primaryjoin=(self.tbl_binaries.c.id==self.tbl_bin_associations.c.bin))))
+                                 suites = relation(Suite, secondary=self.tbl_bin_associations,
+                                     backref=backref('binaries', lazy='dynamic'))),
+                extension = validator)
 
         mapper(BinaryACL, self.tbl_binary_acl,
                properties = dict(binary_acl_id = self.tbl_binary_acl.c.id))
@@ -2946,7 +3052,8 @@ class DBConn(object):
 
         mapper(Component, self.tbl_component,
                properties = dict(component_id = self.tbl_component.c.id,
-                                 component_name = self.tbl_component.c.name))
+                                 component_name = self.tbl_component.c.name),
+               extension = validator)
 
         mapper(DBConfig, self.tbl_config,
                properties = dict(config_id = self.tbl_config.c.id))
@@ -2962,7 +3069,12 @@ class DBConn(object):
                properties = dict(file_id = self.tbl_files.c.id,
                                  filesize = self.tbl_files.c.size,
                                  location_id = self.tbl_files.c.location,
-                                 location = relation(Location)))
+                                 location = relation(Location,
+                                     # using lazy='dynamic' in the back
+                                     # reference because we have A LOT of
+                                     # files in one location
+                                     backref=backref('files', lazy='dynamic'))),
+                extension = validator)
 
         mapper(Fingerprint, self.tbl_fingerprint,
                properties = dict(fingerprint_id = self.tbl_fingerprint.c.id,
@@ -2971,7 +3083,8 @@ class DBConn(object):
                                  keyring_id = self.tbl_fingerprint.c.keyring,
                                  keyring = relation(Keyring),
                                  source_acl = relation(SourceACL),
-                                 binary_acl = relation(BinaryACL)))
+                                 binary_acl = relation(BinaryACL)),
+               extension = validator)
 
         mapper(Keyring, self.tbl_keyrings,
                properties = dict(keyring_name = self.tbl_keyrings.c.name,
@@ -3032,29 +3145,42 @@ class DBConn(object):
         mapper(Location, self.tbl_location,
                properties = dict(location_id = self.tbl_location.c.id,
                                  component_id = self.tbl_location.c.component,
-                                 component = relation(Component),
+                                 component = relation(Component, backref='location'),
                                  archive_id = self.tbl_location.c.archive,
                                  archive = relation(Archive),
-                                 archive_type = self.tbl_location.c.type))
+                                 # FIXME: the 'type' column is old cruft and
+                                 # should be removed in the future.
+                                 archive_type = self.tbl_location.c.type),
+               extension = validator)
 
         mapper(Maintainer, self.tbl_maintainer,
-               properties = dict(maintainer_id = self.tbl_maintainer.c.id))
+               properties = dict(maintainer_id = self.tbl_maintainer.c.id,
+                   maintains_sources = relation(DBSource, backref='maintainer',
+                       primaryjoin=(self.tbl_maintainer.c.id==self.tbl_source.c.maintainer)),
+                   changed_sources = relation(DBSource, backref='changedby',
+                       primaryjoin=(self.tbl_maintainer.c.id==self.tbl_source.c.changedby))),
+                extension = validator)
 
         mapper(NewComment, self.tbl_new_comments,
                properties = dict(comment_id = self.tbl_new_comments.c.id))
 
         mapper(Override, self.tbl_override,
                properties = dict(suite_id = self.tbl_override.c.suite,
-                                 suite = relation(Suite),
+                                 suite = relation(Suite, \
+                                    backref=backref('overrides', lazy='dynamic')),
                                  package = self.tbl_override.c.package,
                                  component_id = self.tbl_override.c.component,
-                                 component = relation(Component),
+                                 component = relation(Component, \
+                                    backref=backref('overrides', lazy='dynamic')),
                                  priority_id = self.tbl_override.c.priority,
-                                 priority = relation(Priority),
+                                 priority = relation(Priority, \
+                                    backref=backref('overrides', lazy='dynamic')),
                                  section_id = self.tbl_override.c.section,
-                                 section = relation(Section),
+                                 section = relation(Section, \
+                                    backref=backref('overrides', lazy='dynamic')),
                                  overridetype_id = self.tbl_override.c.type,
-                                 overridetype = relation(OverrideType)))
+                                 overridetype = relation(OverrideType, \
+                                    backref=backref('overrides', lazy='dynamic'))))
 
         mapper(OverrideType, self.tbl_override_type,
                properties = dict(overridetype = self.tbl_override_type.c.type,
@@ -3074,31 +3200,21 @@ class DBConn(object):
                properties = dict(source_id = self.tbl_source.c.id,
                                  version = self.tbl_source.c.version,
                                  maintainer_id = self.tbl_source.c.maintainer,
-                                 maintainer = relation(Maintainer,
-                                                       primaryjoin=(self.tbl_source.c.maintainer==self.tbl_maintainer.c.id)),
                                  poolfile_id = self.tbl_source.c.file,
-                                 poolfile = relation(PoolFile),
+                                 poolfile = relation(PoolFile, backref=backref('source', uselist = False)),
                                  fingerprint_id = self.tbl_source.c.sig_fpr,
                                  fingerprint = relation(Fingerprint),
                                  changedby_id = self.tbl_source.c.changedby,
-                                 changedby = relation(Maintainer,
-                                                      primaryjoin=(self.tbl_source.c.changedby==self.tbl_maintainer.c.id)),
                                  srcfiles = relation(DSCFile,
                                                      primaryjoin=(self.tbl_source.c.id==self.tbl_dsc_files.c.source)),
-                                 srcassociations = relation(SrcAssociation,
-                                                            primaryjoin=(self.tbl_source.c.id==self.tbl_src_associations.c.source)),
-                                 srcuploaders = relation(SrcUploader)))
+                                 suites = relation(Suite, secondary=self.tbl_src_associations,
+                                     backref=backref('sources', lazy='dynamic')),
+                                 srcuploaders = relation(SrcUploader)),
+               extension = validator)
 
         mapper(SourceACL, self.tbl_source_acl,
                properties = dict(source_acl_id = self.tbl_source_acl.c.id))
 
-        mapper(SrcAssociation, self.tbl_src_associations,
-               properties = dict(sa_id = self.tbl_src_associations.c.id,
-                                 suite_id = self.tbl_src_associations.c.suite,
-                                 suite = relation(Suite),
-                                 source_id = self.tbl_src_associations.c.source,
-                                 source = relation(DBSource)))
-
         mapper(SrcFormat, self.tbl_src_format,
                properties = dict(src_format_id = self.tbl_src_format.c.id,
                                  format_name = self.tbl_src_format.c.format_name))
@@ -3115,13 +3231,9 @@ class DBConn(object):
         mapper(Suite, self.tbl_suite,
                properties = dict(suite_id = self.tbl_suite.c.id,
                                  policy_queue = relation(PolicyQueue),
-                                 copy_queues = relation(BuildQueue, secondary=self.tbl_suite_build_queue_copy)))
-
-        mapper(SuiteArchitecture, self.tbl_suite_architectures,
-               properties = dict(suite_id = self.tbl_suite_architectures.c.suite,
-                                 suite = relation(Suite, backref='suitearchitectures'),
-                                 arch_id = self.tbl_suite_architectures.c.architecture,
-                                 architecture = relation(Architecture)))
+                                 copy_queues = relation(BuildQueue,
+                                     secondary=self.tbl_suite_build_queue_copy)),
+                extension = validator)
 
         mapper(SuiteSrcFormat, self.tbl_suite_src_formats,
                properties = dict(suite_id = self.tbl_suite_src_formats.c.suite,
@@ -3131,13 +3243,20 @@ class DBConn(object):
 
         mapper(Uid, self.tbl_uid,
                properties = dict(uid_id = self.tbl_uid.c.id,
-                                 fingerprint = relation(Fingerprint)))
+                                 fingerprint = relation(Fingerprint)),
+               extension = validator)
 
         mapper(UploadBlock, self.tbl_upload_blocks,
                properties = dict(upload_block_id = self.tbl_upload_blocks.c.id,
                                  fingerprint = relation(Fingerprint, backref="uploadblocks"),
                                  uid = relation(Uid, backref="uploadblocks")))
 
+        mapper(BinContents, self.tbl_bin_contents,
+            properties = dict(
+                binary = relation(DBBinary,
+                    backref=backref('contents', lazy='dynamic')),
+                file = self.tbl_bin_contents.c.file))
+
     ## Connection functions
     def __createconn(self):
         from config import Config