dogpile.core provides a locking interface around a "value creation" and "value retrieval" pair of functions.
The primary interface is the :class:`.Lock` object, which provides for the invocation of the creation function by only one thread and/or process at a time, deferring all other threads/processes to the "value retrieval" function until the single creation thread is completed.
Do I Need to Learn the dogpile.core API Directly?
It's anticipated that most users of dogpile.core will be using it indirectly via the dogpile.cache caching front-end. If you fall into this category, then the short answer is no.
dogpile.core provides core internals to the dogpile.cache package, which provides a simple-to-use caching API, rudimental backends for Memcached and others, and easy hooks to add new backends. Users of dogpile.cache don't need to know or access dogpile.core's APIs directly, though a rough understanding the general idea is always helpful.
Using the core dogpile.core APIs described here directly implies you're building your own resource-usage system outside, or in addition to, the one dogpile.cache provides.
The primary API dogpile provides is the :class:`.Lock` object. This object allows for functions that provide mutexing, value creation, as well as value retrieval.
An example usage is as follows:
from dogpile.core import Lock, NeedRegenerationException import threading import time # store a reference to a "resource", some # object that is expensive to create. the_resource = [None] def some_creation_function(): # call a value creation function value = create_some_resource() # get creationtime using time.time() creationtime = time.time() # keep track of the value and creation time in the "cache" the_resource = tup = (value, creationtime) # return the tuple of (value, creationtime) return tup def retrieve_resource(): # function that retrieves the resource and # creation time. # if no resource, then raise NeedRegenerationException if the_resource is None: raise NeedRegenerationException() # else return the tuple of (value, creationtime) return the_resource # a mutex, which needs here to be shared across all invocations # of this particular creation function mutex = threading.Lock() with Lock(mutex, some_creation_function, retrieve_resource, 3600) as value: # some function that uses # the resource. Won't reach # here until some_creation_function() # has completed at least once. value.do_something()
Above, some_creation_function() will be called when :class:`.Lock` is first invoked as a context manager. The value returned by this function is then passed into the with block, where it can be used by application code. Concurrent threads which call :class:`.Lock` during this initial period will be blocked until some_creation_function() completes.
Once the creation function has completed successfully the first time, new calls to :class:`.Lock` will call retrieve_resource() in order to get the current cached value as well as its creation time; if the creation time is older than the current time minus an expiration time of 3600, then some_creation_function() will be called again, but only by one thread/process, using the given mutex object as a source of synchronization. Concurrent threads/processes which call :class:`.Lock` during this period will fall through, and not be blocked; instead, the "stale" value just returned by retrieve_resource() will continue to be returned until the creation function has finished.
The :class:`.Lock` API is designed to work with simple cache backends like Memcached. It addresses such issues as:
- Values can disappear from the cache at any time, before our expiration time is reached. The :class:`.NeedRegenerationException` class is used to alert the :class:`.Lock` object that a value needs regeneration ahead of the usual expiration time.
- There's no function in a Memcached-like system to "check" for a key without actually retrieving it. The usage of the retrieve_resource() function allows that we check for an existing key and also return the existing value, if any, at the same time, without the need for two separate round trips.
- The "creation" function used by :class:`.Lock` is expected to store the newly created value in the cache, as well as to return it. This is also more efficient than using two separate round trips to separately store, and re-retrieve, the object.
Using dogpile.core for Caching
dogpile.core is part of an effort to "break up" the Beaker package into smaller, simpler components (which also work better). Here, we illustrate how to approximate Beaker's "cache decoration" function, to decorate any function and store the value in Memcached. We create a Python decorator function called cached() which will provide caching for the output of a single function. It's given the "key" which we'd like to use in Memcached, and internally it makes usage of :class:`.Lock`, along with a thread based mutex (we'll see a distributed mutex in the next section):
import pylibmc import threading import time from dogpile.core import Lock, NeedRegenerationException mc_pool = pylibmc.ThreadMappedPool(pylibmc.Client("localhost")) def cached(key, expiration_time): """A decorator that will cache the return value of a function in memcached given a key.""" mutex = threading.Lock() def get_value(): with mc_pool.reserve() as mc: value_plus_time = mc.get(key) if value_plus_time is None: raise NeedRegenerationException() # return a tuple (value, createdtime) return value_plus_time def decorate(fn): def gen_cached(): value = fn() with mc_pool.reserve() as mc: # create a tuple (value, createdtime) value_plus_time = (value, time.time()) mc.put(key, value_plus_time) return value_plus_time def invoke(): with Lock(mutex, gen_cached, get_value, expiration_time) as value: return value return invoke return decorate
Using the above, we can decorate any function as:
@cached("some key", 3600) def generate_my_expensive_value(): return slow_database.lookup("stuff")
The :class:`.Lock` object will ensure that only one thread at a time performs slow_database.lookup(), and only every 3600 seconds, unless Memcached has removed the value, in which case it will be called again as needed.
In particular, dogpile.core's system allows us to call the memcached get() function at most once per access, instead of Beaker's system which calls it twice, and doesn't make us call get() when we just created the value.
For the mutex object, we keep a threading.Lock object that's local to the decorated function, rather than using a global lock. This localizes the in-process locking to be local to this one decorated function. In the next section, we'll see the usage of a cross-process lock that accomplishes this differently.
Using a File or Distributed Lock with Dogpile
The examples thus far use a threading.Lock() object for synchronization. If our application uses multiple processes, we will want to coordinate creation operations not just on threads, but on some mutex that other processes can access.
In this example we'll use a file-based lock as provided by the lockfile package, which uses a unix-symlink concept to provide a filesystem-level lock (which also has been made threadsafe). Another strategy may base itself directly off the Unix os.flock() call, or use an NFS-safe file lock like flufl.lock, and still another approach is to lock against a cache server, using a recipe such as that described at Using Memcached as a Distributed Locking Service.
What all of these locking schemes have in common is that unlike the Python threading.Lock object, they all need access to an actual key which acts as the symbol that all processes will coordinate upon. So here, we will also need to create the "mutex" which we pass to :class:`.Lock` using the key argument:
import lockfile import os from hashlib import sha1 # ... other imports and setup from the previous example def cached(key, expiration_time): """A decorator that will cache the return value of a function in memcached given a key.""" lock_path = os.path.join("/tmp", "%s.lock" % sha1(key).hexdigest()) # ... get_value() from the previous example goes here def decorate(fn): # ... gen_cached() from the previous example goes here def invoke(): # create an ad-hoc FileLock mutex = lockfile.FileLock(lock_path) with Lock(mutex, gen_cached, get_value, expiration_time) as value: return value return invoke return decorate
For a given key "some_key", we generate a hex digest of the key, then use lockfile.FileLock() to create a lock against the file /tmp/53def077a4264bd3183d4eb21b1f56f883e1b572.lock. Any number of :class:`.Lock` objects in various processes will now coordinate with each other, using this common filename as the "baton" against which creation of a new value proceeds.
Unlike when we used threading.Lock, the file lock is ultimately locking on a file, so multiple instances of FileLock() will all coordinate on that same file - it's often the case that file locks that rely upon flock() require non-threaded usage, so a unique filesystem lock per thread is often a good idea in any case.