Skip to content

nkonin/psycopg2cffi

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

https://travis-ci.org/chtd/psycopg2cffi.svg?branch=master

An implementation of the psycopg2 module using cffi. The module is currently compatible with Psycopg 2.5.

You can install psycopg2cffi from PyPI:

pip install psycopg2cffi

Or from source:

python setup.py install

Installation requirements are the same as for psycopg2 - you must have libpq headers, pg_config somewhere on your $PATH, Python development headers (python-dev or similar), and ffi-dev for cffi. Installation was tested on Ubuntu 12.04, Ubuntu 14.04, CentOS (RHEL 5.0), OS X 10.8 - 10.10. It should be possible to make it work on Windows, but I did not test it.

This module works under CPython 2.6+, CPython 3.2+, PyPy 2 and PyPy 3 (PyPy version should be at least 2.0, which is ancient history now).

To use this package with Django or SQLAlchemy invoke a compatibility hook (for example, from settings.py in case of Django, or from a psycopg2.py file in site-packages of your virtual env):

from psycopg2cffi import compat
compat.register()

This will map psycopg2cffi to psycopg2, so now any code that does import psycopg2 will use psycopg2cffi.

Submit issues to https://github.com/chtd/psycopg2cffi/issues

If you notice that psycopg2cffi under PyPy is noticeably slower than psycopg2 under CPython, submit this to the issues too - it should not be the case.

This is a port of (Michael van Tellingen port https://github.com/mvantellingen/psycopg2-ctypes of Alex Gaynor's RPython port (https://bitbucket.org/alex_gaynor/pypy-postgresql/overview) of psycopg2 to Python + ctypes) to cffi.

The main motivation for a port was speed - the writeup with benchmarks is here: http://chtd.ru/blog/bystraya-rabota-s-postgres-pod-pypy/?lang=en

To run tests, install pytest and run them with:

py.test psycopg2cffi

Note that for cffi>=1.0 you need to run python setup.py develop to rebuild _libpq.so (if you changed bindings). And for cffi<1.0 (only PyPy 2.5 and below) you need to run python setup.py install once to generate psycopg2cffi/_config.py, otherwise each import will run config and notify tests will fail.

You can also run Django tests. You need to checkout Django source, add psycopg2 compat as described above, and, from the root of the Django checkout:

PYTHONPATH=`pwd` ./tests/runtests.py \
    --settings=psycopg2cffi.tests.psycopg2_tests.testconfig

In case of problems with Django tests, see official Django docs https://docs.djangoproject.com/en/dev/internals/contributing/writing-code/unit-tests/#running-the-unit-tests

Bugfixes:

  • Support installation under Postgres 10 by jimbattin (#90)

Bugfixes:

  • Named cursors fix (affects newer Django) by danchr (#81)
  • Python 3 fixes in extras by avanov (#83)
  • Null check added in fast parser utils by malthe (#79)

Windows wheels support added by ryoyoko (#69).

Bugfixes:

  • Non-ascii notice messages fixed by asterite3 (#72)
  • AsIs with non-text/bytes fixed by jinty (#64)
  • Silent failures in copy_from and copy_to fixed by gobbledygook88 (#62)
  • Infinite recursion error fixed in errorcodes.lookup (#68)
  • Typos in README fixed by PavloKapyshin (#66)

Fix a regression with error handling when establishing the connection (#61)

Fix a bug with non-ascii error messages (#56)

Fixes for FreeBSD support by Andrew Coleman

JSONB support

Use cffi 1.0, which makes installation more robust, and import is faster. cffi<1.0 is used now only for PyPy 2.5 or below.

Fixing things that were broken in 2.6.0:

  • Fix issue #31 - bigint on 32-bit
  • Fix issue #32 - register_type and unicode
  • Python 3 support
  • A bit faster reading of int, long, float, double fields
  • Small bugfixes
  • Bugfixes and a lot of compatibility work by Daniele Varrazzo

Older releases lack release notes, first release of psycopg2cffi around Nov 2012.

About

Port to cffi with some speed improvements

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages

  • Python 100.0%