[FrameworkBundle] Semantic config for app/system/pool caches #18667

nicolas-grekas · 2016-04-28T19:18:41Z

Q	A
Branch?	3.1
Bug fix?	no
New feature?	no
BC breaks?	no
Deprecations?	no
Tests pass?	yes
Fixed tickets	#18625
License	MIT
Doc PR	-

nicolas-grekas · 2016-04-28T19:19:12Z

I'd happily accept some help to fix the XSD please, that should be the only remaining thing left.

GuilhemN · 2016-04-29T15:24:56Z

src/Symfony/Bundle/FrameworkBundle/DependencyInjection/Configuration.php

                                    ->scalarNode('adapter')
-                                        ->info('The cache pool adapter service to use as template definition.')
-                                        ->defaultValue('cache.adapter.shared')
+                                        ->info('The create a new pool based on this adapter service.')


should be To create

nicolas-grekas · 2016-04-29T15:26:06Z

Hold on, I'm going to change a few things after some discussions with @javiereguiluz

nicolas-grekas · 2016-04-29T19:28:46Z

Following suggestions from @javiereguiluz and @fabpot, configuration is now the following:

Pre-configured cache pool services are now: cache.app, cache.system, cache.serializer and cache.validator. Cache adapters (abstract pools) are prefixed by cache.adapter.*. ~~cache.adapter.local~~ is replaced by cache.system. ~~cache.adapter.shared~~ is replaced by cache.app. They both can be reused as a parent to define new "app" or "system" pools reusing the same base adapter. They both default to cache.adapter.filesystem.

Configuration is made easy for those two, using full service names to define which adapter should be used for each of them.

E.g. to use APCu for all system caches:

framework:
  cache:
    system: cache.adapter.apcu

To use Redis for all app related caches:

framework:
  cache:
    app: cache.adapter.redis
    default_redis_provider: redis://1.2.3.4:5678

As you can see, this PR also adds a factory for creating Redis connection based on a DSN.

Then of course you can go to more advanced configurations to redefine either custom adapters or regular pools (pre-defined or not).

Adapter and pool configurations are merged in a single framework.cache.pools option. Note that pool names should be given as full service names.

E.g. to define a new system cache pool accessible with $container->get('my_local_cache_pool'):

framework:
  cache:
    pools:
      my_local_cache_pool:
        adapter: cache.system # overwrite the default which is "cache.app"
        public: true # new cache pools are private by default

E.g. to use Redis for all system pools and use https://github.com/snc/SncRedisBundle to configure the connection:

framework:
  cache:
    system: cache.adapter.redis
    default_redis_provider: snc_redis_connection_service

a verbose variant of this could be:

framework:
  cache:
    system: app.redis_cache
    pools:
      app.redis_cache:
        adapter: cache.adapter.redis
        provider: snc_redis_connection_service

Or to use a Doctrine Cache provider for all app related caches:

framework:
  cache:
    app: cache.adapter.doctrine
    default_doctrine_provider: doctrine_cache_provider_service

or verbose variant:

framework:
  cache:
    app: app.doctrine_cache
    pools:
      app.doctrine_cache:
        adapter: cache.adapter.doctrine
        provider: doctrine_cache_provider_service

The same would go for configuring a pool on top of a PSR-6 cache adapter.

nicolas-grekas · 2016-04-29T22:03:29Z

Last tweaks pushed, green, ready for review+merge:

cache.system is now a public pool by default (alongside with cache.app)
those two special pools are reserved names in the pools config section: they must be configured through their dedicated options under cache (see examples above)
new pools are private by default to allow easily changing the adapter of predefined pools (cache.serializer & cache.validation today). IMHO creating custom pools is an even more advanced use case that won't suffer from having this requirement of adding public: true when needed.

xabbuh · 2016-04-30T09:36:41Z

src/Symfony/Component/Cache/Adapter/RedisAdapter.php

    private $redis;

    public function __construct(\Redis $redisConnection, $namespace = '', $defaultLifetime = 0)
    {
        if (preg_match('#[^-+_.A-Za-z0-9]#', $namespace, $match)) {
            throw new InvalidArgumentException(sprintf('RedisAdapter namespace contains "%s" but only characters in [-+_.A-Za-z0-9] are allowed.', $match[0]));
        }
+        if (!$redisConnection->isConnected()) {


Are you sure that it is a good idea to require an established connection in the constructor?

good catch, check removed

xabbuh · 2016-04-30T09:37:49Z

@nicolas-grekas I'd be happy to split the PR in two (one for the Redis related changes and one for the semantic config). That would make reviewing a bit easier.

xabbuh · 2016-04-30T09:38:46Z

UPGRADE-3.1.md

@@ -107,7 +107,7 @@ FrameworkBundle
   framework:
       cache:
           pools:
-               serializer:
+               cache.serializer:


Why do we repeat the cache. prefix here. Imo this looks weird as it repeats what we already know from the config tree (I mean we are configuring the cache, why do we want to reflect that in the name here too?)?

because I've been told that Symfony core prefers explicit over implicit & "magic" conventions. This in fact is much more flexible and doesn't force any service prefix. See examples above where an app.doctrine_cache pool is created. We don't have to teach any special convention here. Just give a service name.

I understand that we want the service to be cache.serializer. But imo we should then add the cache. prefix here: https://github.com/nicolas-grekas/symfony/blob/cache-config/src/Symfony/Bundle/FrameworkBundle/DependencyInjection/FrameworkExtension.php#L1059

Otherwise, the full config path we use here is framework.cache.pools.cache.serializer which imo is not really clear why I should use the cache part twice here.

And by the way, adding the cache. prefix in the extension when registering the definition will also prevent us from potentially having conflicting service names because when people configure their own cache pools, they will not always think of adding the prefix here explicitly.

I think I prefer to be explicit. Whenever possible we try to use the full service name in configuration and that's what is done here. If the end user does not want to prefix their service name with cache., it's not a problem.

But we then still have the point that the configured names are directly used as service ids which can easily lead to conflicts (we don't do that anywhere else, do we?).

After re-reading #18667 (comment) think we can stop the discussion here as it was based on a misunderstand on my end.

👍 for how the PR is right now.

nicolas-grekas · 2016-04-30T12:09:04Z

To be rebased on top of #18681 (which is included for now)

…s-grekas) This PR was merged into the 3.1-dev branch. Discussion ---------- [Cache] Add DSN based Redis connection factory | Q | A | ------------- | --- | Branch? | 3.1 | Bug fix? | no | New feature? | yes | BC breaks? | no | Deprecations? | no | Tests pass? | yes | Fixed tickets | - | License | MIT | Doc PR | - Required by #18667 Commits ------- 3073efb [Cache] Add DSN based Redis connection factory

fabpot · 2016-05-01T11:01:04Z

#18681 merged now

nicolas-grekas · 2016-05-01T12:04:20Z

rebased

nicolas-grekas · 2016-05-02T06:37:01Z

src/Symfony/Bundle/FrameworkBundle/Resources/config/cache.xml

            <argument /> <!-- PSR-6 provider service -->
            <argument /> <!-- namespace -->
            <argument /> <!-- default lifetime -->
        </service>

-        <service id="cache.adapter.redis" class="Symfony\Component\Cache\Adapter\RedisAdapter" abstract="true">
+        <service id="cache.adapter.redis" class="Symfony\Component\Cache\Adapter\RedisAdapter" abstract="true" lazy="true">


I just made cache.adapter.redis lazy so that no connection is opened unless needed.

but I'm not sure this is required, wdyt?

we don't have any lazy service in Symfony core as it depends on a deps that is not a hard dep. Lazy means that the service is wrapped with a proxy, which adds its own layer of complexity, so I would avoid that if possible.

xabbuh · 2016-05-02T07:51:10Z

Apart from how cache pool names are now treated I am 👍 for the changes.

dunglas · 2016-05-02T20:53:10Z

👍

fabpot · 2016-05-04T15:43:51Z

Thank you @nicolas-grekas.

…caches (tgalopin, nicolas-grekas) This PR was merged into the 3.1-dev branch. Discussion ---------- [FrameworkBundle] Semantic config for app/system/pool caches | Q | A | ------------- | --- | Branch? | 3.1 | Bug fix? | no | New feature? | no | BC breaks? | no | Deprecations? | no | Tests pass? | yes | Fixed tickets | #18625 | License | MIT | Doc PR | - Commits ------- a2a567d [FrameworkBundle] Simplify config for app/system/pool caches 80a5508 [FrameworkBundle] Add cache adapters in semantic configuration

carsonbot added the Status: Needs Review label Apr 28, 2016

nicolas-grekas mentioned this pull request Apr 28, 2016

[FrameworkBundle] Add cache adapters in semantic configuration #18625

Closed

nicolas-grekas force-pushed the cache-config branch from 09320b9 to bb8e5d7 Compare April 29, 2016 06:37

GuilhemN reviewed Apr 29, 2016
View reviewed changes

nicolas-grekas force-pushed the cache-config branch from bb8e5d7 to b2df9d6 Compare April 29, 2016 19:10

nicolas-grekas force-pushed the cache-config branch 4 times, most recently from 22bd298 to b80a8cc Compare April 29, 2016 21:58

nicolas-grekas added the Feature label Apr 29, 2016

nicolas-grekas force-pushed the cache-config branch 4 times, most recently from eeb561e to f84e588 Compare April 30, 2016 08:57

nicolas-grekas changed the title ~~[FrameworkBundle] Semantic configuration for redis/apcu/pools/adapters caches~~ [FrameworkBundle] Semantic config for app/system/pool caches Apr 30, 2016

nicolas-grekas force-pushed the cache-config branch 2 times, most recently from 2318540 to a34a005 Compare April 30, 2016 09:20

xabbuh reviewed Apr 30, 2016
View reviewed changes

javiereguiluz added the FrameworkBundle label Apr 30, 2016

nicolas-grekas force-pushed the cache-config branch from a34a005 to 40265e5 Compare April 30, 2016 12:06

nicolas-grekas mentioned this pull request Apr 30, 2016

[Cache] Add DSN based Redis connection factory #18681

Merged

nicolas-grekas force-pushed the cache-config branch 2 times, most recently from 51bfb86 to 493c337 Compare May 1, 2016 08:49

[FrameworkBundle] Add cache adapters in semantic configuration

80a5508

nicolas-grekas force-pushed the cache-config branch from 493c337 to c24745b Compare May 1, 2016 12:04

nicolas-grekas force-pushed the cache-config branch from c24745b to 8c5e9ae Compare May 2, 2016 06:33

nicolas-grekas reviewed May 2, 2016
View reviewed changes

nicolas-grekas force-pushed the cache-config branch 2 times, most recently from e4abdc8 to 8c5e9ae Compare May 2, 2016 06:42

nicolas-grekas mentioned this pull request May 3, 2016

[FrameworkBundle] Wire PhpArrayAdapter with a new cache warmer for annotations #18533

Closed

nicolas-grekas force-pushed the cache-config branch from 8c5e9ae to 893afda Compare May 3, 2016 11:42

nicolas-grekas mentioned this pull request May 3, 2016

[PropertyAccess] Add PSR-6 cache #16838

Closed

nicolas-grekas added the Ready label May 3, 2016

[FrameworkBundle] Simplify config for app/system/pool caches

a2a567d

nicolas-grekas force-pushed the cache-config branch from 893afda to a2a567d Compare May 4, 2016 15:36

fabpot merged commit a2a567d into symfony:master May 4, 2016

GuilhemN mentioned this pull request May 4, 2016

Do not use validator.cache and serializer.cache anymore symfony/symfony-standard#965

Merged

nicolas-grekas deleted the cache-config branch May 4, 2016 19:16

fabpot mentioned this pull request May 13, 2016

Release v3.1.0-BETA1 #18776

Merged

This was referenced Jun 14, 2016

Switch to the new Symfony cache infrastructure api-platform/core#577

Merged

Documentation for the new PSR-6 Cache component symfony/symfony-docs#6171

Closed

HypeMC mentioned this pull request Nov 5, 2023

[FrameworkBundle][Cache] The cache.[type] services are never created #52461

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[FrameworkBundle] Semantic config for app/system/pool caches #18667

[FrameworkBundle] Semantic config for app/system/pool caches #18667

nicolas-grekas commented Apr 28, 2016 •

edited

Loading

nicolas-grekas commented Apr 28, 2016 •

edited

Loading

GuilhemN Apr 29, 2016 •

edited

Loading

nicolas-grekas commented Apr 29, 2016

nicolas-grekas commented Apr 29, 2016 •

edited

Loading

nicolas-grekas commented Apr 29, 2016 •

edited

Loading

xabbuh Apr 30, 2016

nicolas-grekas Apr 30, 2016

xabbuh commented Apr 30, 2016

xabbuh Apr 30, 2016

nicolas-grekas Apr 30, 2016

xabbuh May 2, 2016

xabbuh May 2, 2016

fabpot May 2, 2016

xabbuh May 2, 2016

xabbuh May 2, 2016

nicolas-grekas commented Apr 30, 2016

fabpot commented May 1, 2016

nicolas-grekas commented May 1, 2016

nicolas-grekas May 2, 2016

nicolas-grekas May 2, 2016

fabpot May 4, 2016

xabbuh commented May 2, 2016

dunglas commented May 2, 2016

fabpot commented May 4, 2016

[FrameworkBundle] Semantic config for app/system/pool caches #18667

[FrameworkBundle] Semantic config for app/system/pool caches #18667

Conversation

nicolas-grekas commented Apr 28, 2016 • edited Loading

nicolas-grekas commented Apr 28, 2016 • edited Loading

GuilhemN Apr 29, 2016 • edited Loading

Choose a reason for hiding this comment

nicolas-grekas commented Apr 29, 2016

nicolas-grekas commented Apr 29, 2016 • edited Loading

nicolas-grekas commented Apr 29, 2016 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

xabbuh commented Apr 30, 2016

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

nicolas-grekas commented Apr 30, 2016

fabpot commented May 1, 2016

nicolas-grekas commented May 1, 2016

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

xabbuh commented May 2, 2016

dunglas commented May 2, 2016

fabpot commented May 4, 2016

nicolas-grekas commented Apr 28, 2016 •

edited

Loading

nicolas-grekas commented Apr 28, 2016 •

edited

Loading

GuilhemN Apr 29, 2016 •

edited

Loading

nicolas-grekas commented Apr 29, 2016 •

edited

Loading

nicolas-grekas commented Apr 29, 2016 •

edited

Loading