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
import sqlalchemy
from sqlalchemy import create_engine, Table, MetaData, Column, Integer
-from sqlalchemy.orm import sessionmaker, mapper, relation, object_session
+from sqlalchemy.orm import sessionmaker, mapper, relation, object_session, \
+ backref, MapperExtension, EXT_CONTINUE
from sqlalchemy import types as sqltypes
# Don't remove this, we re-export the exceptions to scripts which import us
# 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)
+
################################################################################
################################################################################
-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)
+
+__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
# 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')
################################################################################
-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 == filesize and self.md5sum == md5sum
+
+ def properties(self):
+ return ['filename', 'file_id', 'filesize', 'md5sum', 'sha1sum', \
+ 'sha256sum', 'location', 'source', '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
@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):
"""
@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):
"""
################################################################################
-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')
################################################################################
-class Location(object):
- def __init__(self, *args, **kwargs):
- pass
+class Location(ORMObject):
+ def __init__(self, path = None):
+ self.path = path
+ # 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', 'archive_type', 'component', 'files_count']
+
+ def not_null_constraints(self):
+ return ['path', 'archive_type']
__all__.append('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:
################################################################################
-class DBSource(object):
- def __init__(self, *args, **kwargs):
- pass
+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 __repr__(self):
- return '<DBSource %s (%s)>' % (self.source, self.version)
+ def properties(self):
+ return ['source', 'source_id', 'maintainer', 'changedby', \
+ 'fingerprint', 'poolfile', 'version', 'suites_count', \
+ 'install_date']
+
+ def not_null_constraints(self):
+ return ['source', 'version', 'install_date', 'maintainer', \
+ 'changedby', 'poolfile', 'install_date']
__all__.append('DBSource')
"""
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,
if x[1] in s and x[0] not in s:
s.append(x[0])
- q = q.join(SrcAssociation).join(Suite)
- q = q.filter(Suite.suite_name.in_(s))
-
- # Reduce the query results to a list of version numbers
- ql = [ j.version for j in q.all() ]
+ q = q.filter(DBSource.suites.any(Suite.suite_name.in_(s)))
- # 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
@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')
__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}
"""
- 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
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()
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"):
################################################################################
-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
('CopyChanges', 'copychanges'),
('OverrideSuite', 'overridesuite')]
-class Suite(object):
+# 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']
+
+ def not_null_constraints(self):
+ return ['suite_name', 'version']
def __eq__(self, val):
if isinstance(val, str):
@return: list of Architecture objects for the given name (may be empty)
"""
- q = object_session(self).query(Architecture). \
- filter(Architecture.suites.contains(self))
+ 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()
-__all__.append('Suite')
-
-@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
+ def get_sources(self, source):
+ """
+ Returns a query object representing DBSource that is part of C{suite}.
- @type suite: str
- @param suite: Suite name to search for
+ - B{source} - source package name, eg. I{mailfilter}, I{bbdb}, I{glibc}
- @type architecture: str
- @param architecture: Architecture name to search for
+ @type source: string
+ @param source: source package name
- @type session: Session
- @param session: Optional SQL session object (a temporary one will be
- generated if not supplied)
+ @rtype: sqlalchemy.orm.query.Query
+ @return: a query of DBSource
- @rtype: SuiteArchitecture
- @return: the SuiteArchitecture object or None
- """
+ """
- q = session.query(SuiteArchitecture)
- q = q.join(Architecture).filter_by(arch_string=architecture)
- q = q.join(Suite).filter_by(suite_name=suite)
+ session = object_session(self)
+ return session.query(DBSource).filter_by(source = source). \
+ with_parent(self)
- try:
- return q.one()
- except NoResultFound:
- return None
-
-__all__.append('get_suite_architecture')
+__all__.append('Suite')
@session_wrapper
def get_suite(suite, session=None):
################################################################################
-# TODO: remove SuiteArchitecture class
-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):
################################################################################
-class Uid(object):
+class Uid(ORMObject):
def __init__(self, uid = None, name = None):
self.uid = uid
self.name = name
# 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')
def __setupmappers(self):
mapper(Architecture, self.tbl_architecture,
- properties = dict(arch_id = self.tbl_architecture.c.id,
- suites = relation(Suite, secondary=self.tbl_suite_architectures, backref='architectures', order_by='suite_name')))
+ 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,
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,
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,
component = relation(Component),
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))
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='sources'),
+ 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))
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,
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,