Skip to main content

"A comprehensive, fast, pure Python memcached client"

Project description

pymemcache Master Documentation Status

A comprehensive, fast, pure-Python memcached client.

pymemcache supports the following features:

  • Complete implementation of the memcached text protocol.
  • Configurable timeouts for socket connect and send/recv calls.
  • Access to the “noreply” flag, which can significantly increase the speed of writes.
  • Flexible, modular and simple approach to serialization and deserialization.
  • The (optional) ability to treat network and memcached errors as cache misses.

Installing pymemcache

Install from pip:

pip install pymemcache

For development, clone from github and run the tests:

git clone
cd pymemcache

Run the tests (make sure you have a local memcached server running):



See the documentation here:


If you’re planning on using pymemcache with Django, you might be interested in django-pymemcache. It provides a Django cache backend that is built on pymemcache.

Comparison with Other Libraries


The pylibmc library is a wrapper around libmemcached, implemented in C. It is fast, implements consistent hashing, the full memcached protocol and timeouts. It does not provide access to the “noreply” flag. It also isn’t pure Python, so using it with libraries like gevent is out of the question, and its dependency on libmemcached poses challenges (e.g., it must be built against the same version of libmemcached that it will use at runtime).


The python-memcache library implements the entire memcached text protocol, has a single timeout for all socket calls and has a flexible approach to serialization and deserialization. It is also written entirely in Python, so it works well with libraries like gevent. However, it is tied to using thread locals, doesn’t implement “noreply”, can’t treat errors as cache misses and is slower than both pylibmc and pymemcache. It is also tied to a specific method for handling clusters of memcached servers.


The team at mixpanel put together a pure Python memcached client as well. It has more fine grained support for socket timeouts, only connects to a single host. However, it doesn’t support most of the memcached API (just get, set, delete and stats), doesn’t support “noreply”, has no serialization or deserialization support and can’t treat errors as cache misses.

We’re Hiring!

Are you really excited about open-source? Or great software engineering? Pinterest is hiring!


New in version 2.2.2

  • Fix long_description string in Python packaging.

New in version 2.2.1

  • Fix flags when setting multiple differently-typed values at once.

New in version 2.2.0

  • Drop official support for Python 3.4.
  • Use setup.cfg metadata instead config to generate package.
  • Add default_noreply parameter to HashClient.
  • Add encoding parameter to Client constructors (defaults to ascii).
  • Add flags parameter to write operation methods.
  • Handle unicode key values in MockMemcacheClient correctly.
  • Improve ASCII encoding failure exception.

New in version 2.1.1

  • Fix dependency on six already being installed.

New in version 2.1.0

  • Public classes and exceptions can now be imported from the top-level pymemcache package (e.g. pymemcache.Client). #197
  • Add UNIX domain socket support and document server connection options. #206
  • Add support for the cache_memlimit command. #211
  • Commands key are now always sent in their original order. #209

New in version 2.0.0

  • Change set_many and set_multi api return value. #179
  • Fix support for newbytes from python-future. #187
  • Add support for Python 3.7, and drop support for Python 3.3
  • Properly batch Client.set_many() call. #182
  • Improve _check_key() and _store_cmd() performance. #183
  • Properly batch Client.delete_many() call. #184
  • Add option to explicitly set pickle version used by serde. #190

New in version 1.4.4

  • pypy3 to travis test matrix

  • full benchmarks in test

  • fix flake8 issues

  • Have mockmemcacheclient support non-ascii strings

  • Switch from using pickle format 0 to the highest available version. See #156

    Warning: different versions of python have different highest pickle versions:

New in version 1.4.3

  • Documentation improvements
  • Fixed cachedump stats command, see #103
  • Honor default_value in HashClient

New in version 1.4.2

  • Drop support for python 2.6, see #109

New in version 1.4.1

  • Python 3 serializations fixes #131
  • Drop support for pypy3
  • Comment cleanup
  • Add gets_many to hash_client
  • Better checking for illegal chars in key

New in version 1.4.0

  • Unicode keys support. It is now possible to pass the flag allow_unicode_keys when creating the clients, thanks @jogo!
  • Fixed a bug where PooledClient wasn’t following default_noreply arg set on init, thanks @kols!
  • Improved documentation

New in version 1.3.8

  • use cpickle instead of pickle when possible (python2)

New in version 1.3.7

  • default parameter on get(key, default=0)
  • fixed docs to autogenerate themselves with sphinx
  • fix linter to work with python3
  • improve error message on illegal Input for the key
  • refactor stat parsing
  • fix MockMemcacheClient
  • fix unicode char in middle of key bug

New in version 1.3.6

  • Fix flake8 and cleanup tox building
  • Fix security vulnerability by sanitizing key input

New in version 1.3.5

  • Bug fix for HashClient when retries is set to zero.
  • Adding the VERSION command to the clients.

New in version 1.3.4

  • Bug fix for the HashClient that corrects behavior when there are no working servers.

New in version 1.3.3

  • Adding caching to the Travis build.
  • A bug fix for pluggable hashing in HashClient.
  • Adding a default_noreply argument to the Client ctor.

New in version 1.3.2

  • Making the location of Memcache Exceptions backwards compatible.

New in version 1.3.0

  • Python 3 Support
  • Introduced HashClient that uses consistent hasing for allocating keys across many memcached nodes. It also can detect servers going down and rebalance keys across the available nodes.
  • Retry sock.recv() when it raises EINTR

New in version 1.2.9

  • Introduced PooledClient a thread-safe pool of clients

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Files for pymemcache, version 2.2.2
Filename, size & hash File type Python version Upload date
pymemcache-2.2.2-py2.py3-none-any.whl (44.2 kB) View hashes Wheel py2.py3
pymemcache-2.2.2.tar.gz (41.3 kB) View hashes Source None

Supported by

Elastic Elastic Search Pingdom Pingdom Monitoring Google Google BigQuery Sentry Sentry Error logging AWS AWS Cloud computing DataDog DataDog Monitoring Fastly Fastly CDN SignalFx SignalFx Supporter DigiCert DigiCert EV certificate StatusPage StatusPage Status page