pjpersist 1.7.1

PostgreSQL/JSONB Persistence Backend

PostGreSQL/JSONB Data Persistence

This document outlines the general capabilities of the pjpersist package. pjpersist is a PostGreSQL/JSONB storage implementation for persistent Python objects. It is not a storage for the ZODB.

The goal of pjpersist is to provide a data manager that serializes objects to JSONB blobs at transaction boundaries. The PJ data manager is a persistent data manager, which handles events at transaction boundaries (see transaction.interfaces.IDataManager) as well as events from the persistency framework (see persistent.interfaces.IPersistentDataManager).

An instance of a data manager is supposed to have the same life time as the transaction, meaning that it is assumed that you create a new data manager when creating a new transaction:

>>> import transaction

Note: The conn object is a psycopg.Connection instance. In this case our tests use the pjpersist_test database.

Let’s now define a simple persistent object:

>>> import datetime
>>> import persistent

>>> class Person(persistent.Persistent):
...
...     def __init__(self, name, phone=None, address=None, friends=None,
...                  visited=(), birthday=None):
...         self.name = name
...         self.address = address
...         self.friends = friends or {}
...         self.visited = visited
...         self.phone = phone
...         self.birthday = birthday
...         self.today = datetime.datetime(2014, 5, 14, 12, 30)
...
...     def __str__(self):
...         return self.name
...
...     def __repr__(self):
...         return '<%s %s>' %(self.__class__.__name__, self)

We will fill out the other objects later. But for now, let’s create a new person and store it in PJ:

>>> stephan = Person(u'Stephan')
>>> stephan
<Person Stephan>

The datamanager provides a root attribute in which the object tree roots can be stored. It is special in the sense that it immediately writes the data to the DB:

>>> dm.root['stephan'] = stephan
>>> dm.root['stephan']
<Person Stephan>

Custom Persistence Tables

By default, persistent objects are stored in a table having the escaped Python path of the class:

>>> from pjpersist import serialize
>>> person_cn = serialize.get_dotted_name(Person, True)
>>> person_cn
'u__main___dot_Person'

>>> transaction.commit()
>>> dumpTable(person_cn)
[{'data': {u'_py_persistent_type': u'__main__.Person',
           u'address': None,
           u'birthday': None,
           u'friends': {},
           u'name': u'Stephan',
           u'phone': None,
           u'today': {u'_py_type': u'datetime.datetime',
                      u'value': u'2014-05-14T12:30:00.000000'},
           u'visited': []},
  'id': u'0001020304050607080a0b0c0'}]

As you can see, the stored document for the person looks very much like a natural JSON document. But oh no, I forgot to specify the full name for Stephan. Let’s do that:

>>> dm.root['stephan'].name = u'Stephan Richter'
>>> dm.root['stephan']._p_changed
True

This time, the data is not automatically saved:

>>> fetchone(person_cn)['data']['name']
u'Stephan'

So we have to commit the transaction first:

>>> dm.root['stephan']._p_changed
True
>>> transaction.commit()
>>> dm.root['stephan']._p_changed
>>> fetchone(person_cn)['data']['name']
u'Stephan Richter'

Let’s now add an address for Stephan. Addresses are also persistent objects:

>>> class Address(persistent.Persistent):
...     _p_pj_table = 'address'
...
...     def __init__(self, city, zip):
...         self.city = city
...         self.zip = zip
...
...     def __str__(self):
...         return '%s (%s)' %(self.city, self.zip)
...
...     def __repr__(self):
...         return '<%s %s>' %(self.__class__.__name__, self)

pjpersist supports a special attribute called _p_pj_table, which allows you to specify a custom table to use.

>>> stephan = dm.root['stephan']
>>> stephan.address = Address('Maynard', '01754')
>>> stephan.address
<Address Maynard (01754)>

Note that the address is not immediately saved in the database:

>>> dumpTable('address', isolate=True)
relation "address" does not exist
...

But once we commit the transaction, everything is available:

>>> transaction.commit()
>>> dumpTable('address')
[{'data': {u'_py_persistent_type': u'__main__.Address',
           u'city': u'Maynard',
           u'zip': u'01754'},
  'id': u'0001020304050607080a0b0c0'}]

>>> dumpTable(person_cn)
[{'data': {u'_py_persistent_type': u'__main__.Person',
           u'address': {u'_py_type': u'DBREF',
                        u'database': u'pjpersist_test',
                        u'id': u'0001020304050607080a0b0c0',
                        u'table': u'address'},
           u'birthday': None,
           u'friends': {},
           u'name': u'Stephan Richter',
           u'phone': None,
           u'today': {u'_py_type': u'datetime.datetime',
                      u'value': u'2014-05-14T12:30:00.000000'},
           u'visited': []},
  'id': u'0001020304050607080a0b0c0'}]

>>> dm.root['stephan'].address
<Address Maynard (01754)>

Non-Persistent Objects

As you can see, even the reference looks nice and all components are easily visible. But what about arbitrary non-persistent, but picklable, objects? Well, let’s create a phone number object for that:

>>> class Phone(object):
...
...     def __init__(self, country, area, number):
...         self.country = country
...         self.area = area
...         self.number = number
...
...     def __str__(self):
...         return '%s-%s-%s' %(self.country, self.area, self.number)
...
...     def __repr__(self):
...         return '<%s %s>' %(self.__class__.__name__, self)

>>> dm.root['stephan'].phone = Phone('+1', '978', '394-5124')
>>> dm.root['stephan'].phone
<Phone [phone]>

Let’s now commit the transaction and look at the JSONB document again:

>>> transaction.commit()
>>> dm.root['stephan'].phone
<Phone [phone]>

>>> dumpTable(person_cn)
[{'data': {u'_py_persistent_type': u'__main__.Person',
           u'address': {u'_py_type': u'DBREF',
                        u'database': u'pjpersist_test',
                        u'id': u'0001020304050607080a0b0c0',
                        u'table': u'address'},
           u'birthday': None,
           u'friends': {},
           u'name': u'Stephan Richter',
           u'phone': {u'_py_type': u'__main__.Phone',
                      u'area': u'978',
                      u'country': u'+1',
                      u'number': u'394-5124'},
           u'today': {u'_py_type': u'datetime.datetime',
                      u'value': u'2014-05-14T12:30:00.000000'},
           u'visited': []},
  'id': u'0001020304050607080a0b0c0'}]

As you can see, for arbitrary non-persistent objects we need a small hint in the sub-document, but it is very minimal. If the __reduce__ method returns a more complex construct, more meta-data is written. We will see that next when storing a date and other arbitrary data:

>>> dm.root['stephan'].friends = {'roy': Person(u'Roy Mathew')}
>>> dm.root['stephan'].visited = (u'Germany', u'USA')
>>> dm.root['stephan'].birthday = datetime.date(1980, 1, 25)

>>> transaction.commit()
>>> dm.root['stephan'].friends
{u'roy': <Person Roy Mathew>}
>>> dm.root['stephan'].visited
[u'Germany', u'USA']
>>> dm.root['stephan'].birthday
datetime.date(1980, 1, 25)

As you can see, a dictionary key is always converted to unicode and tuples are always maintained as lists, since JSON does not have two sequence types.

>>> import pprint
>>> pprint.pprint(dict(
...     fetchone(person_cn, """data @> '{"name": "Stephan Richter"}'""")))
{'data': {u'_py_persistent_type': u'__main__.Person',
          u'address': {u'_py_type': u'DBREF',
                       u'database': u'pjpersist_test',
                       u'id': u'0001020304050607080a0b0c0',
                       u'table': u'address'},
          u'birthday': {u'_py_type': u'datetime.date',
                        u'value': u'1980-01-25'},
          u'friends': {u'roy': {u'_py_type': u'DBREF',
                                u'database': u'pjpersist_test',
                                u'id': u'0001020304050607080a0b0c0',
                                u'table': u'u__main___dot_Person'}},
          u'name': u'Stephan Richter',
          u'phone': {u'_py_type': u'__main__.Phone',
                     u'area': u'978',
                     u'country': u'+1',
                     u'number': u'394-5124'},
          u'today': {u'_py_type': u'datetime.datetime',
                     u'value': u'2014-05-14T12:30:00.000000'},
          u'visited': [u'Germany', u'USA']},
 'id': u'0001020304050607080a0b0c0'}

Custom Serializers

(A patch to demonstrate)

>>> dm.root['stephan'].birthday = datetime.date(1981, 1, 25)
>>> transaction.commit()

>>> pprint.pprint(
...     fetchone(person_cn,
...         """data @> '{"name": "Stephan Richter"}'""")['data']['birthday'])
{u'_py_type': u'datetime.date', u'value': u'1981-01-25'}

As you can see, the serialization of the birthay is an ISO string. We can, however, provide a custom serializer that uses the ordinal to store the data.

>>> class DateSerializer(serialize.ObjectSerializer):
...
...     def can_read(self, state):
...         return isinstance(state, dict) and \
...                state.get('_py_type') == 'custom_date'
...
...     def read(self, state):
...         return datetime.date.fromordinal(state['ordinal'])
...
...     def can_write(self, obj):
...         return isinstance(obj, datetime.date)
...
...     def write(self, obj):
...         return {'_py_type': 'custom_date',
...                 'ordinal': obj.toordinal()}

>>> serialize.SERIALIZERS.append(DateSerializer())
>>> dm.root['stephan']._p_changed = True
>>> transaction.commit()

Let’s have a look again:

>>> dm.root['stephan'].birthday
datetime.date(1981, 1, 25)

>>> pprint.pprint(dict(
...     fetchone(person_cn, """data @> '{"name": "Stephan Richter"}'""")))
{'data': {u'_py_persistent_type': u'__main__.Person',
          u'address': {u'_py_type': u'DBREF',
                       u'database': u'pjpersist_test',
                       u'id': u'0001020304050607080a0b0c0',
                       u'table': u'address'},
          u'birthday': {u'_py_type': u'custom_date', u'ordinal': 723205},
          u'friends': {u'roy': {u'_py_type': u'DBREF',
                                u'database': u'pjpersist_test',
                                u'id': u'0001020304050607080a0b0c0',
                                u'table': u'u__main___dot_Person'}},
          u'name': u'Stephan Richter',
          u'phone': {u'_py_type': u'__main__.Phone',
                     u'area': u'978',
                     u'country': u'+1',
                     u'number': u'394-5124'},
          u'today': {u'_py_type': u'custom_date', u'ordinal': 735367},
          u'visited': [u'Germany', u'USA']},
 'id': u'0001020304050607080a0b0c0'}

Much better!

>>> del serialize.SERIALIZERS[:]

Persistent Objects as Sub-Documents

In order to give more control over which objects receive their own tables and which do not, the developer can provide a special flag marking a persistent class so that it becomes part of its parent object’s document:

>>> class Car(persistent.Persistent):
...     _p_pj_sub_object = True
...
...     def __init__(self, year, make, model):
...         self.year = year
...         self.make = make
...         self.model = model
...
...     def __str__(self):
...         return '%s %s %s' %(self.year, self.make, self.model)
...
...     def __repr__(self):
...         return '<%s %s>' %(self.__class__.__name__, self)

The _p_pj_sub_object is used to mark a type of object to be just part of another document:

>>> dm.root['stephan'].car = car = Car('2005', 'Ford', 'Explorer')
>>> transaction.commit()

>>> dm.root['stephan'].car
<Car 2005 Ford Explorer>

>>> pprint.pprint(dict(
...     fetchone(person_cn, """data @> '{"name": "Stephan Richter"}'""")))
{'data': {u'_py_persistent_type': u'__main__.Person',
          u'address': {u'_py_type': u'DBREF',
                       u'database': u'pjpersist_test',
                       u'id': u'0001020304050607080a0b0c0',
                       u'table': u'address'},
          u'birthday': {u'_py_type': u'datetime.date',
                        u'value': u'1981-01-25'},
          u'car': {u'_py_persistent_type': u'__main__.Car',
                   u'make': u'Ford',
                   u'model': u'Explorer',
                   u'year': u'2005'},
          u'friends': {u'roy': {u'_py_type': u'DBREF',
                                u'database': u'pjpersist_test',
                                u'id': u'0001020304050607080a0b0c0',
                                u'table': u'u__main___dot_Person'}},
          u'name': u'Stephan Richter',
          u'phone': {u'_py_type': u'__main__.Phone',
                     u'area': u'978',
                     u'country': u'+1',
                     u'number': u'394-5124'},
          u'today': {u'_py_type': u'datetime.date', u'value': u'2014-05-14'},
          u'visited': [u'Germany', u'USA']},
 'id': u'0001020304050607080a0b0c0'}

The reason we want objects to be persistent is so that they pick up changes automatically:

>>> dm.root['stephan'].car.year = '2004'
>>> transaction.commit()
>>> dm.root['stephan'].car
<Car 2004 Ford Explorer>

Table Sharing

Since PostGreSQL/JSONB is so flexible, it sometimes makes sense to store multiple types of (similar) objects in the same table. In those cases you instruct the object type to store its Python path as part of the document.

Warning: Please note though that this method is less efficient, since the document must be loaded in order to create a ghost causing more database access.

>>> class ExtendedAddress(Address):
...
...     def __init__(self, city, zip, country):
...         super(ExtendedAddress, self).__init__(city, zip)
...         self.country = country
...
...     def __str__(self):
...         return '%s (%s) in %s' %(self.city, self.zip, self.country)

In order to accomplish table sharing, you simply create another class that has the same _p_pj_table string as another (sub-classing will ensure that).

So let’s give Stephan two extended addresses now.

>>> dm.root['stephan'].address2 = ExtendedAddress(
...     'Tettau', '01945', 'Germany')
>>> dm.root['stephan'].address2
<ExtendedAddress Tettau (01945) in Germany>

>>> dm.root['stephan'].address3 = ExtendedAddress(
...     'Arnsdorf', '01945', 'Germany')
>>> dm.root['stephan'].address3
<ExtendedAddress Arnsdorf (01945) in Germany>

>>> transaction.commit()

When loading the addresses, they should be of the right type:

>>> dm.root['stephan'].address
<Address Maynard (01754)>
>>> dm.root['stephan'].address2
<ExtendedAddress Tettau (01945) in Germany>
>>> dm.root['stephan'].address3
<ExtendedAddress Arnsdorf (01945) in Germany>

Persistent Serialization Hooks

When persistent components implement the IPersistentSerializationHooks, it is possible for the object to conduct some custom storage function.

>>> from pjpersist.persistent import PersistentSerializationHooks
>>> class Usernames(PersistentSerializationHooks):
...     _p_pj_table = 'usernames'
...     format = 'email'
...
...     def _pj_after_store_hook(self, conn):
...         print('After Store Hook')
...
...     def _pj_after_load_hook(self, conn):
...         print('After Load Hook')

When we store the object, the hook is called: (actually twice, because this is a new object)

>>> dm.root['stephan'].usernames = Usernames()
>>> transaction.commit()
After Store Hook
After Store Hook

When loading, the same happens:

>>> dm.root['stephan'].usernames.format
After Load Hook
'email'

The store hook fires just once if the object is not new:

>>> dm.root['stephan'].usernames.format = 'snailmail'
>>> transaction.commit()
After Store Hook

Column Serialization

pjpersist also allows for the object to specify values, usually attributes or properties, to be stored as columns on the object’s storage table.

Note that we support only a one-way transformation, because object state will be always deserialized from the data jsonb field.

>>> import zope.schema
>>> class IPerson(zope.interface.Interface):
...
...     name = zope.schema.TextLine(title=u'Name')
...     address = zope.schema.TextLine(title=u'Address')
...     visited = zope.schema.Datetime(title=u'Visited')
...     phone = zope.schema.TextLine(title=u'Phone')

Initially, we are storing only the name in a column:

>>> from pjpersist.persistent import SimpleColumnSerialization, select_fields
>>> @zope.interface.implementer(IPerson)
... class ColumnPerson(SimpleColumnSerialization, Person):
...     _p_pj_table = 'cperson'
...     _pj_column_fields = select_fields(IPerson, 'name')

So once I create such a person and commit the transaction, the person table is extended to store the attribute and the person is added to the table:

>>> dm.root['anton'] = anton = ColumnPerson(u'Anton')
>>> transaction.commit()

>>> dumpTable('cperson')
[{'data': {u'_py_persistent_type': u'__main__.ColumnPerson',
           u'address': None,
           u'birthday': None,
           u'friends': {},
           u'name': u'Anton',
           u'phone': None,
           u'today': {u'_py_type': u'datetime.datetime',
                      u'value': u'2014-05-14T12:30:00.000000'},
           u'visited': []},
  'id': u'0001020304050607080a0b0c0',
  'name': u'Anton'}]

Tricky Cases

Changes in Basic Mutable Type

Tricky, tricky. How do we make the framework detect changes in mutable objects, such as lists and dictionaries? Answer: We keep track of which persistent object they belong to and provide persistent implementations.

>>> type(dm.root['stephan'].friends)
 <class 'pjpersist.serialize.PersistentDict'>

>>> dm.root['stephan'].friends[u'roger'] = Person(u'Roger')
>>> transaction.commit()
>>> sorted(dm.root['stephan'].friends.keys())
[u'roger', u'roy']

The same is true for lists:

>>> type(dm.root['stephan'].visited)
 <class 'pjpersist.serialize.PersistentList'>

>>> dm.root['stephan'].visited.append('France')
>>> transaction.commit()
>>> dm.root['stephan'].visited
[u'Germany', u'USA', u'France']

Circular Non-Persistent References

Any mutable object that is stored in a sub-document, cannot have multiple references in the object tree, since there is no global referencing. These circular references are detected and reported:

>>> class Top(persistent.Persistent):
...     foo = None

>>> class Foo(object):
...     bar = None

>>> class Bar(object):
...     foo = None

>>> top = Top()
>>> foo = Foo()
>>> bar = Bar()
>>> top.foo = foo
>>> foo.bar = bar
>>> bar.foo = foo

>>> dm.root['top'] = top
Traceback (most recent call last):
...
CircularReferenceError: <...>

Circular Persistent References

In general, circular references among persistent objects are not a problem, since we always only store a link to the object. However, there is a case when the circular dependencies become a problem.

If you set up an object tree with circular references and then add the tree to the storage at once, it must insert objects during serialization, so that references can be created. However, care needs to be taken to only create a minimal reference object, so that the system does not try to recursively reduce the state.

>>> class PFoo(persistent.Persistent):
...     bar = None

>>> class PBar(persistent.Persistent):
...     foo = None

>>> top = Top()
>>> foo = PFoo()
>>> bar = PBar()
>>> top.foo = foo
>>> foo.bar = bar
>>> bar.foo = foo

>>> dm.root['ptop'] = top

Containers and Tables

Now that we have talked so much about the gory details on storing one object, what about mappings that reflect an entire table, for example a table of people.

There are many approaches that can be taken. The following implementation defines an attribute in the document as the mapping key and names a table:

>>> from pjpersist import mapping
>>> class People(mapping.PJTableMapping):
...     __pj_table__ = person_cn
...     __pj_mapping_key__ = 'short_name'

The mapping takes the data manager as an argument. One can easily create a sub-class that assigns the data manager automatically. Let’s have a look:

>>> People(dm).keys()
[]

The reason no person is in the list yet, is because no document has the key yet or the key is null. Let’s change that:

>>> People(dm)['stephan'] = dm.root['stephan']
>>> transaction.commit()

>>> People(dm).keys()
[u'stephan']
>>> People(dm)['stephan']
<Person Stephan Richter>

Also note that setting the “short-name” attribute on any other person will add it to the mapping:

>>> dm.root['stephan'].friends['roy'].short_name = 'roy'
>>> transaction.commit()
>>> sorted(People(dm).keys())
[u'roy', u'stephan']

CHANGES

1.7.1 (2019-06-19)

• Fixed an edge case when the serializer gets a mapping with a key dict_data. Reading such object failed.

• Fixed an edge case with the serializer, when an object’s state living in a persistent object became ‘empty’. Basically the state was just {‘_py_persistent_type’: ‘SomeClass’} SomeClass.__setstate__ was not called, thus the object could miss attributes. Like a subclass of UserDict would miss the data attribute.

• Removed checking for 0x00 chars in dict keys. Turns out PostGreSQL just can not store 0x00.

1.7.0 (2019-05-29)

• Support for sub-second datetime and time resolution during serialization.

• Add use_cache argument to PJContainer._load_one() to support ignoring the cache. (This became handy if a container keeps track of multiple versions of an item and you try to load all old revisions.)

1.6.0 (2019-05-29)

• Make id and data column name configurable via _pj_id_column and _pj_data_column attributes in PJContainer, respectively.

• Auto-assign a name to objects when using PJContainer, not just IdNamesPJContainer.

1.5.0 (2018-10-10)

• Support for Python 3.7. Removed Python 3.5 testing from tox.

1.4.1 (2018-09-13)

• No need to log in tpc_finish.

1.4.0 (2018-09-13)

• Implemented skipping tpc_prepare when DM has no writes. We found out that AWS Aurora is dog slow at the moment on tpc_prepare. When the DataManager has no writes, there’s no need to call tpc_prepare. See CALL_TPC_PREPARE_ON_NO_WRITE_TRANSACTION, by default True for backwards compatibility.

• Added ability to log whether the transaction had writes. See LOG_READ_WRITE_TRANSACTION, by default False

1.3.2 (2018-04-19)

• More precise flushing of datamanager to avoid unnecessary database writes.

1.3.1 (2018-04-11)

• Enabled concurrent adds to IdNamesPJContainer by eliminating a query that was causing transaction conflicts.

1.3.0 (2018-03-22)

• Python 3 compatibility fixes

• More efficient PJContainer.values() implementation

1.2.2 (2017-12-12)

• Need to protect all DB calls against DatabaseDisconnected

1.2.1 (2017-12-12)

• psycopg2.OperationalError and psycopg2.InterfaceError will be caught on SQL command execution and reraised as DatabaseDisconnected

1.2.0 (2017-10-24)

• Added a new helper function to link subobject to main doc object. This is needed when a custom __getstate__() and __setstate__() is implemented. A detailed example is provided.

• Implemented flush_hint argument for IDataManager.execute() to allow flushing only some objects during query. flush_hints is a list table names that need to be flushed for the query to return a correct result.

• The Zope-specific containers use the flush_hint to only flush objects they manage when a query is run on the container.

• While flushing objects, every main document object is now only flushed once. Before that fix, any subobject would cause its doc object to be dumped again.

Note: These optimizations provide a 15% performance improvements in real-world applications.

1.1.2 (2017-09-14)

• Make sure changed objects aren’t _p_changed anymore after commit.

1.1.1 (2017-07-03)

• Nothing changed yet.

1.0.0 (2017-03-18)

• Initial Public Release

• Project forked from mongopersist to work with PostGreSQL and JSONB data type. The main motiviation is the ability to utilize PostGreSQL’s great transactional support.