gocept.testdb 1.2.1

General

The same procedure also works for PostgreSQL:

Note: Instead of using POSTGRES_PASS, use the ~/.pgpass mechanism provided by postgres itself.

>>> db = gocept.testdb.PostgreSQL(schema_path=schema)
>>> db.create()
>>> engine = sqlalchemy.create_engine(db.dsn)
>>> conn = engine.connect()
>>> ignore = conn.execute('SELECT * from tmp_functest')
>>> ignore = conn.execute('SELECT * from foo')
>>> conn.invalidate()
>>> db.drop()
>>> engine.connect().execute('SELECT * from tmp_functest')
Traceback (most recent call last):
  ...
OperationalError:...

Encoding

For Postgres an optional encoding parameter can be specified in the constructor. It is used when creating the database.

>>> db = gocept.testdb.PostgreSQL(schema_path=schema, encoding='UTF8')
>>> db.create()
>>> engine = sqlalchemy.create_engine(db.dsn)
>>> conn = engine.connect()
>>> encoding = conn.execute(
...     '''SELECT pg_catalog.pg_encoding_to_char(encoding) as encoding
...        FROM pg_catalog.pg_database
...        WHERE datname = %s''', db.dsn.split('/')[-1]).fetchall()
>>> encoding in ([(u'UTF8',)], [('UTF8',)],)
True
>>> conn.invalidate()
>>> db.drop()

DB name

You can specify the name of the database:

>>> db = gocept.testdb.PostgreSQL(schema_path=schema, db_name='mytestdb')
>>> db.dsn
'postgresql://localhost/mytestdb'

Templates

For Postgres, an optional template parameter can be passed to the constructor. It specifies the name of a template db which is used for the creation of the database. If the template db does not exist, it is created with the specified schema.

The first time you create the database with the db_template argument, the template db is created (if it does not exist already) along with the requested db:

>>> db = gocept.testdb.PostgreSQL(
...     schema_path=schema, db_template=db_template)
>>> db.create()
>>> table_names(db.get_dsn(db_template))
[u'foo', u'tmp_functest']
>>> table_names(db.dsn)
[u'foo', u'tmp_functest']

Now with the template available, the schema is not used anymore to create the database (it’s re-created from the template). Let’s modify the template db before the next db creation run to demonstrate this:

>>> _ = execute(db.get_dsn(db_template), 'DROP TABLE foo;')
>>> db2 = gocept.testdb.PostgreSQL(
...     schema_path=schema, db_template=db_template)
>>> db2.create()
>>> table_names(db2.get_dsn(db_template))
[u'tmp_functest']
>>> table_names(db2.dsn)
[u'tmp_functest']

When creating the database, we can, however, force the template db to be created afresh from the schema. Doing so now will leave us with both a test db and a template db according to the modified schema:

>>> db3 = gocept.testdb.PostgreSQL(
...     schema_path=schema, db_template=db_template, force_template=True)
>>> db3.create()
>>> table_names(db3.get_dsn(db_template))
[u'foo', u'tmp_functest']
>>> table_names(db3.dsn)
[u'foo', u'tmp_functest']

The template db (and with it, the test db) ist also created anew if the schema file is newer than the existing template db:

>>> import time; time.sleep(1)  # XXX hack; either fake db mtime or see #12431
>>> write(schema, 'CREATE TABLE bar (dummy int);')
>>> db4 = gocept.testdb.PostgreSQL(
...     schema_path=schema, db_template=db_template, force_template=False)
>>> db4.create()
>>> table_names(db4.get_dsn(db_template))
[u'bar', u'tmp_functest']
>>> table_names(db4.dsn)
[u'bar', u'tmp_functest']

If, however, the template db cannot be set up properly, it is removed altogether to avoid a broken template db interfering with subsequent tests:

>>> write(schema+'-broken', 'foobar')
>>> db_broken = gocept.testdb.PostgreSQL(
...     schema_path=schema+'-broken', db_template=db_template,
...     force_template=True)
>>> db_broken.create()
Traceback (most recent call last):
SystemExit: Could not initialize schema in database
            'gocept.testdb.tests-...-templatetest'.
>>> db_template in db_broken.list_db_names()
False

Clean up:

>>> db.drop()
>>> db2.drop()
>>> db3.drop()
>>> db4.drop()

1.2.1 (2013-09-03)

• Improve retry logic for dropping databases: ensure that db actually exists before trying to drop it (#12706).

1.2 (2013-06-21)

• Provisional compatibility to Python 3 (no compatible mysql driver exists yet, though).

• Fixed test code that made implicit assumptions about existing databases on the PostgreSQL server used or depended on timing conditions.

1.1.1 (2013-03-19)

• Use template0 when changing LC_COLLATE which was introduced in v1.1.

1.1 (2013-03-19)

• Add possibility to set the LC_COLLATE when creating a database.

1.0 (2013-03-12)

• Allow a postfix for mysql-commands.

1.0b5 (2011-12-22)

• Use timestamp for randomizing database names to avoid collisions.

1.0b4 (2011-09-23)

• drop-all entry point now also works, if PostgreSQL or MySQL is not installed/running.

1.0b3 (2011-05-13)

• Added connect convenience method.

1.0b2 (2011-05-04)

• Added is_testing property as a convenience API.

• Added exists property as a convenience API.

1.0b1 (2011-04-12)

• Changed the protocol for using test databases: separated instantiating a Database instance from creating the database on the server. Creating Database instances is cheap now so they can be interacted with and passed around, deferring the expensive database creation on the server to the moment db.create() is called explicitly. This is also more symmetric with respect to db.drop().

• Added functionality for dropping all databases matching the testdb naming scheme which may have been left on the server by earlier test runs. Optionally also drops the template db.

• Added option to specify the name of the database for MySQL and PostgreSQL.

• Added an option to specify a template database for PostgreSQL. If it does not exist, it is created from the specified schema. It is also possible to force the creation of the template even if it exists (dropping the current template database). Whenever the schema has changed since the template db was last created, the template db is updated automatically.

0.4 (2010-12-15)

• Added option to specify the encoding for the PostgreSQL database.

• Updated PostgreSQL protocol from postgres: to postgresql: as the former one is deprecated in SQLAlchemy 0.6, thus requiring atleast version 0.5.6 of SQLAlchemy.

• Added documentation how to develop this package further.

• Added doumentation about usage of Database prefix

0.3 (2010-10-15)

• PostgreSQL: Don’t call createdb/dropdb with --quite but only psql.

0.2 (2009-02-26)

• implemented authentication with password for mysql. Passwords for postgres are still not supported, though.

0.1 (2008-09-26)

• first release