We are happy to announce that we have created the Pony ORM mailing list
firstname.lastname@example.org. This mailing list is the place where you can ask questions, share experiences, or discuss the Pony ORM with other users and the Pony ORM developers. Please subscribe to the list using this link: ponyorm-list.ponyorm.com
Starting with this release Pony starts transactions with the database in the standard mode by default. Previously the default mode was the optimistic mode.
Standard and optimistic transaction modes
Both modes require that all interaction with the database should happen within the db_session. In order to achieve that you should either decorate the function which interacts with the database with the @db_session decorator or use the db_session context manager. However what happens inside the db_session differs in the standard and optimistic mode.
When an application makes changes in the database, those changes are accumulated in the db_session cache. In the standard mode the changes which are made in the current transaction become visible to the subsequent SELECTs in the same transaction. This happens because Pony automatically flushes accumulated changes before each SELECT.
The optimistic transactions mode is designed to provide better scalability. The main goal of the optimistic mode it to minimize the time when the database or its part is locked due to the data update. In this mode Pony splits one transaction into two. The first transaction is used only for reading data from the database. All the changes are accumulated in the db_session cache. When it is time to do commit() Pony sends rollback() to the database in order to finish the read-only transaction and then starts short write-only transaction and flushes the cache. This mode is called optimistic because Pony hopes that no other transaction changes the same objects concurrently. During the write-only transaction Pony checks that the object’s attributes which are seen by the user have the same values as during the first transaction. If any change was detected, Pony raises the exception “UnrepeatableReadError: ‘Object Person was updated outside of current transaction’“
– Select for update. Now you can add “.for_update()“ at the end of the query in order to generate “SELECT .. FOR UPDATE“ SQL statement:
x = 'john' u = select(u for u in User if u.login == x).for_update()[:]
Since SQLite doesn’t support the SELECT … FOR UPDATE syntax, Pony emulates such behavior by ensuring that such SELECT is executed within the transaction.
– Inline foreign keys definitions for MySQL bug workaround.
MySQL ignores inline foreign keys definitions. Pony provides a workaround for this bug replacing the inline definitions with foreign key constraints declared separately.
– Reconnect on connection failure. If the connection to the database was lost, now Pony will try to reconnect.
– Automatic foreign key indexes generation.
– Ability to control the column indexes creation. Now you can specify the keyword parameter “index“ during the attribute declaration.
“index=False“ – skip index creation (MySQL InnoDB will still create foreign key indexes automatically)
“index=True“ – index will be created with the default name
“index=’index_name’“ – create index with the specified name
If no “index“ option is specified then Pony still creates index for foreign keys with the default name.
– UUID datatype support has been added:
>>> from pony.orm import * >>> from uuid import UUID, uuid4 >>> db = Database('sqlite', ':memory:') >>> class MyEntity(db.Entity): ... uid = PrimaryKey(UUID, default=uuid4) ... name = Required(unicode) ... >>> db.generate_mapping(create_tables=True) >>> with db_session: ... obj1 = MyEntity(name='First') ... obj2 = MyEntity(name='Second') ... >>> for obj in MyEntity.select(): ... print obj.uid, obj.name ... c719327f-128e-45b5-8c64-5af451e70925 First 6ede633c-63e8-4ad8-9ff8-89d56f822588 Second >>>
– The parameter ‘check_tables’ in the “Database.generate_mapping()“ method is now deprecated. Starting with this release Pony always checks the existence of tables and columns for each entity.
– The support of PyGreSQL module is discontinued. Please use psycopg2 for connecting to the PostgreSQL database.
– Bug fixes.