- For PHP ^7.2 and Symfony <6, use version 3.x
- For PHP ^8.0 and Symfony >6, use version 4.x
Symfony's HttpCache
store implementation is rather old and was developed
when there were no separate components for locking and caching yet. Moreover,
expired cache entries are never pruned and thus causes your cache directory
to continue to grow forever until you delete it manually.
Along the way, I needed support for cache invalidation based on tags which was pretty easy to implement thanks to the Symfony Cache component.
This bundle thus provides an alternative StoreInterface
implementation
that…
- …instead of re-implementing locking and caching mechanisms again, uses the well tested Symfony Cache and Lock components, both with the local filesystem adapters by default.
- …thanks to the
TagAwareAdapterInterface
of the Cache component, supports tag based cache invalidation. - …thanks to the
PrunableInterface
of the Cache component, supports auto-pruning of expired entries on the filesystem trying to prevent flooding the filesystem. - …allows you to use a different PSR-6 cache adapters as well as a different lock adapter than the local filesystem ones. However, be careful about choosing the right adapters, see warning below.
- …supports
BinaryFileResponse
instances.
$ composer require toflar/psr6-symfony-http-cache-store
For the Symfony 4/Flex structure, you need to adjust your index.php
like this:
<?php
// public/index.php
$kernel = new Kernel($env, $debug);
$kernel = new HttpCache(
$kernel,
new Psr6Store(['cache_directory' => $kernel->getCacheDir()]),
null,
['debug' => $debug]
);
That's it, that's all there is to do. The Psr6Store
will automatically
create the best caching and locking adapters available for your local filesystem.
If you want to go beyond this point, the Psr6Store
can be configured by
passing an array of $options
in the constructor:
-
cache_directory: Path to the cache directory for the default cache adapter and lock factory.
Either this or both
cache
andlock_factory
are required.Type:
string
-
cache: Explicitly specify the cache adapter you want to use.
Note that if you want to make use of cache tagging, this cache must implement the
Symfony\Component\Cache\Adapter\TagAwareAdapterInterface
Make sure thatlock
andcache
have the same scope. See warning below!Type:
Symfony\Component\Cache\Adapter\AdapterInterface
Default:FilesystemAdapter
instance withcache_directory
-
lock_factory: Explicitly specify the lock factory you want to use. Make sure that lock and cache have the same scope. See warning below!
Type:
Symfony\Component\Lock\Factory
Default:Factory
withSemaphoreStore
if supported,FlockStore
otherwise -
prune_threshold: Configure the number of write actions until the store will prune the expired cache entries. Pass
0
to disable automated pruning.Type:
int
Default:500
-
cache_tags_header: The HTTP header name that's used to check for tags.
Type:
string
Default:Cache-Tags
-
generate_content_digests: Whether or not content digests should be generated. See "Generating Content Digests" for more information.
Type:
boolean
Default:true
By default, this cache implementation generates content digests. This means that the response meta data is stored separately from the response content. If multiple responses share the same content, it is stored in the cache only once. Compare the following illustrations to see the difference:
With generating content digests:
Without generating content digests:
Generating content digests optimizes the cache so it uses up less storage. Using them, however, also comes at the costs of requiring a second round trip to fetch the content digest from the cache during the lookup process.
Whether or not you want to use content digests depends on your PSR-6 cache back end. If lookups are fast and storage is rather limited (e.g. Redis), you might want to use content digests. If lookups are rather slow and storage is less of an issue (e.g. Filesystem), you might want to disable them.
You can control the behaviour using the generate_content_digests
configuration
option.
This cache implementation allows to cache BinaryFileResponse
instances but
the files are not actually copied to the cache directory. It will just try to
fetch the original file and if that does not exist anymore, the store returns
null
, causing HttpCache to deal with it as a cache miss and continue normally.
It is ideal for use cases such as caching /favicon.ico
requests where you would
like to prevent the application from being started and thus deliver the response
from HttpCache.
Tag cache entries by adding a response header with the tags as a comma
separated value. By default, that header is called Cache-Tags
, this can be
overwritten in cache_tags_header
.
To invalidate tags, call the method Psr6Store::invalidateTags
or use the
PurgeTagsListener
from the FOSHttpCache library to handle tag
invalidation requests.
By default, this store removes expired entries from the cache after every 500
cache write operations. Fetching data does not affect performance.
You can change the automated pruning frequency with the prune_threshold
configuration setting.
You can also manually trigger pruning by calling the prune()
method on the
cache. With this, you could for example implement a cron job that loads the store
and prunes it at a configured interval, to prevent slowing down random requests
that were cache misses because they have to wait for the pruning to happen. If you
have set up a cron job, you should disable auto pruning by setting the threshold
to 0
.
It is possible to configure other cache adapters or lock stores than the
filesystem ones. Only do this if you are sure of what you are doing. In
this pull request Fabien refused to add PSR-6 store support to
the Symfony AppCache
with the following arguments:
- Using a filesystem allows for
opcache
to make the cache very effective; - The cache contains some PHP (when using ESI for instance) and storing
PHP in anything else than a filesystem would mean
eval()
-ing strings coming from Redis / Memcache /...; - HttpCache is triggered very early and does not have access to the container or anything else really. And it should stay that way to be efficient.
While the first and third point depend on what you do and need, be sure to respect the second point. If you use network enabled caches like Redis or Memcache, make sure that they are not shared with other systems to avoid code injection!
I would like to thank David for his invaluable feedback on this library while we were working on an integration for the awesome FOSHttpCache library.