autottlcache 1.0.8

Time-to-live (TTL) cache with auto-expiry

autottlcache

A time-to-live (TTL) cache dictionary with auto-expiry.

Installation

To install from PyPI:

$ pip install autottlcache

Usage

The TTL cache behaves very much like a dictionary. In fact, it is a MutableMapping so supports the dictionary interface. It has one important difference though, which is that the items within it will expire and automatically be removed.

Let's create an AutoTTLCache with a a maximum size of twelve entires, and a TTL of 30 seconds:

>>> from autottlcache import AutoTTLCache
>>> cache = AutoTTLCache(maxsize=12, ttl=30.0)

We can add entries to the cache:

>>> cache[1] = 56
>>> cache[2] = 89
>>> cache[3] = 99
>>> cache[4] = 10
>>> cache[5] = 20

And watch them expire, by repeatedly asking the cache to list its keys:

>>> list(cache.keys())
[2, 3, 4, 5]

We took long enough adding the keys that the first one had already expired:

>>> list(cache.keys())
[3, 4, 5]

Now the second one has gone too:

>>> list(cache.keys())
[3, 4, 5]

This time we're quick enough to get in before the third one expires:

>>> list(cache.keys())
[3, 4, 5]
>>> list(cache.keys())
[3, 4, 5]
>>> list(cache.keys())
[3, 4, 5]
>>> list(cache.keys())
[4, 5]
>>> list(cache.keys())
[4, 5]
>>> list(cache.keys())
[4, 5]
>>> list(cache.keys())
[5]
>>> list(cache.keys())
[5]
>>> list(cache.keys())
[5]
>>> list(cache.keys())
[5]
>>> list(cache.keys())
[5]
>>> list(cache.keys())
[]

All gone.

Stale items are removed from the cache either when an operation on the cache object is invoked, such as when an item is retrieved, or by a continuously running background thread that periodically removes expired items. The cache will be checked with a frequency of ten times the TTL period, so an item in a cache with a TTL of 30 seconds should live no longer than 33 seconds.

Programming style

Owing to the fact that items will auto-expire, to avoid race conditions you should strongly prefer an Easier to Ask for Forgiveness than Permission (EAFP) programming style, rather than a Look Before You Leap (LBYL) style.

Risky race-condition:

from autottlcache import AutoTTLCache
cache = AutoTTLCache(maxsize=100, ttl=30)
cache["daffodils"] = Image("daffodils-3280x2040.png")

# ...

if "daffodils" in cache:
    # What happens if "daffodils" expires here?
    image = cache["daffodils"]  # This could raise a key error which is unhandled
else:
    print("No flowers for you!")

Better to use a single operation to retrieve an object, and handle the failure:

from autottlcache import AutoTTLCache
cache = AutoTTLCache(maxsize=100, ttl=30)
cache["daffodils"] = Image("daffodils-3280x2040.png")

# ...

try:
    image = cache["daffodils"]
except KeyError:
    print("No flowers for you!")
else:
    display(image)

Resource use

A single additional thread will be running during any periods for which one or more AutoTTLCache objects are extant. The thread will start when the first AutoTTLCache is created, and will terminate shortly after the last AutoTTLCache is finalized.

Development

To release, there is a short manual process:

$ bumpversion patch
$ python setup.py sdist bdist_wheel
$ twine upload dist/* --config-file=path/to/sixty-north.pypirc