What is pypq

pypq is a DBAPI 2.0 compliant Python driver for PostgreSQL, made using ctypes and libpq.so library.

This means, that it can be used with PyPy, or any other circumstances, where ctypes is available, but psycopg2, the most popular PostgreSQL driver, is not.

Features

• full DBAPI 2.0 compliance

• fast enough to be compared to psycopg2

• a psycopg2-like extension API

• has a Django backend

• casting of all standard python types implemented, including datetime, date and timedelta, except converting lists into postgres arrays and vice versa

Installation

Just do:

easy_install pypq

Or to download the most recent version from BitBucket:

hg clone ssh://hg@bitbucket.org/descent/pypq
cd pypyq
python setup.py install

Example usage

Basic operations:

import pypq
connection = pypq.connect('dbname=dbname user=user host=host port=port password=password')
# or
connection = pypq.connect(dbname='dbname', user='user', host='host', \
    port='port', password='password')
# for complete reference look at PQconnectdbParams section at
# http://www.postgresql.org/docs/9.1/static/libpq-connect.htm
cursor = connection.cursor()
cursor.execute('select * from mytable where x = %s', [x])
for row in cursor:
    print row

To use pypq as a django backend, simply change the ‘ENGINE’ line to “pypq.django_backend” in settings.py file. Something like:

DATABASES = {
    'default': {
        'ENGINE': 'pypq.django_backend',
        'NAME': 'dbname',
        'USER': 'user',
        'PASSWORD': 'password',
        'HOST': 'host',
        'PORT': '',
    }
}

# If using south, change the database adapter

SOUTH_DATABASE_ADAPTERS = {'default': 'pypq.django_backend.south_adapter'}

Extending with custom types is done in simimlar way to psycopg2. Here are some examples, but you can look at pypq/datatypes.py for additional ways of usage:

def adapter(myclass_instance):
    """A function to adapt MyClass to postgres

    It should return a string and a postgres OID of
    resulting datatype or 0 if you do not know it
     (it's ok to do that)
    """
    return adapted, oid

# This will allow doing this:
# cursor.execute('select %s', [myclass_instance])
pypq.datatypes.register_adapter(MyClass, adapter)

# And this will allow doing this, if you have a special type
# in postgres which you want to handle manually, and for
# which you know postgres OIDs:

# cursor.execute('create table test(a some_special_type)')
# cursor.execute('select * from test')
# and it will fetch your launch to_python(value)
# for every special instance

def to_python(value):
    """A function to convert a postgres value into Python object""""

mytype = pypq.datatypes.new_type((1,2,3), 'MyAwesomeType', to_python))
pypq.datatypes.register_type(mytype)

# or

class MyType(pypq.datatypes.PyPQDataType):

    oids = (1,2,3)

    @classmethod
    def to_python(cls, value):
        return value

Bugs

The project currently has some limitations - no support of postgres arrays yet - this does not work somewhy, though it works in psycopg2:

cur.execute('select %s', [None])
DatabaseError: ERROR: could not determine data type of parameter $1

• Internally, OID’s are currently not passed to the database, when executing PQexecParams. I am not sure, if it is ok, but django works flawleslly this way, and if we pass the OID’s to PQexecParams, it does not, cause some postgres type casting fails.

• not tested on other python versions than 2.7 and recent pypy

• not tested on windows (it probably won’t work)

• not thread-safe, you cannot use same cursors or connections in different threads