Redis Tools (retools)
retools is a package of Redis tools. It’s aim is to provide a variety of Redis backed Python tools that are always 100% unit tested, fast, efficient, and utilize the capabilities of Redis.
Current tools in retools:
- Global Lock
On the horizon for future implementation:
- A worker/job processing system similar to Celery but based on how Ruby’s Resque system works.
A high performance caching system that can act as a drop-in replacement for Beaker’s caching. Unlike Beaker’s caching, this utilizes Redis for distributed write-locking dogpile prevention. It also collects hit/miss cache statistics along with recording what regions are used by which functions and arguments.
from retools.cache import CacheRegion, cache_region, invalidate_function CacheRegion.add_region('short_term', expires=3600) @cache_region('short_term') def slow_function(*search_terms): # Do a bunch of work return results my_results = slow_function('bunny') # Invalidate the cache for 'bunny' invalidate_function(slow_function, , 'bunny')
Differences from Beaker
Unlike Beaker’s caching system, this is built strictly for Redis. As such, it adds several features that Beaker doesn’t possess:
- A distributed write-lock so that only one writer updates the cache at a time across a cluster.
- Hit/Miss cache statistics to give you insight into what caches are less effectively utilized (and may need either higher expiration times, or just not very worthwhile to cache).
- Very small, compact code-base with 100% unit test coverage.
A Redis based lock implemented as a Python context manager, based on Chris Lamb’s example.
from retools.lock import Lock with Lock('a_key', expires=60, timeout=10): # do something that should only be done one at a time
retools is offered under the MIT license.
- Critical fix for caching that prevents old values from being displayed forever. Thanks to Daniel Holth for tracking down the problem-aware.
- Actually sets the Redis expiration for a value when setting the cached value in Redis. This defaults to 1 week.
- Statistics for the cache is now optional and can be disabled to slightly reduce the Redis queries used to store/retrieve cache data.
- Added first revision of worker/job Queue system, with event support.
- Heavily refactored Connection to not be a class singleton, instead a global_connection instance is created and used by default.
- Increased conditional coverage to 100% (via instrumental).
Changing the default global Redis connection has changed semantics, instead of using Connection.set_default, you should set the global_connection’s redis property directly:
import redis from retools import global_connection global_connection.redis = redis.Redis(host='myhost')
- Removed clear argument from invalidate_region, as removing keys from the set but not removing the hit statistics can lead to data accumulating in Redis that has no easy removal other than .keys() which should not be run in production environments.
- Removed deco_args from invalidate_callable (invalidate_function) as its not actually needed since the namespace is already on the callable to invalidate.
- Caching in a similar style to Beaker, with hit/miss statistics, backed by a Redis global write-lock with old values served to prevent the dogpile effect
- Redis global lock
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
|Filename, size||File type||Python version||Upload date||Hashes|
|Filename, size retools-0.2.tar.gz (27.5 kB)||File type Source||Python version None||Upload date||Hashes View|