(dev/core#174) CRM_Utils_Cache_Redis::flush() - Respect prefixes

[totten]

Overview

When flushing caches for an instance of CRM_Utils_Cache_Redis, you should
only delete the items with the matching prefix.

For reference, consider this example:

$js_strings = CRM_Utils_Cache::create(array(
  'name' => 'js_strings',
  'type' => array('*memory*', 'SqlGroup', 'ArrayCache'),
));

$community_messages = CRM_Utils_Cache::create(array(
  'name' => 'community_messages',
  'type' => array('*memory*', 'SqlGroup', 'ArrayCache'),
));

$js_strings->flush();

If the system is generally configured for Redis, this will give us two instances of CRM_Utils_Cache_Redis -- these instances are stored on the same server but use different prefixes.

Before

Flushing js_strings has the side-effect of flushing everything else in Redis, such as community_messages.

After

Flushing js_strings only flushes js_strings.

Comments

The example above focuses on two relatively obscure caches (js_strings and community_messages) because they actually exist today. In the future, as part of dev/core#174, it will become possible to store sessions in Redis. And then it will be more important -- e.g. if a user creates a custom-field, you do want to flush one set of data (e.g. the contact-fields cache), but you don't want to flush another (e.g. the user sessions).

The current commit fixes Redis. Although it's outside the scope of this
commit, here's an assessment of other drivers:

• APCcache looks fine -- it already consults the prefix.
• ArrayCache looks fine -- data is already scoped within the object.
• NoCache looks fine -- it's a null-op for everything.
• SqlGroup is loosely right. There's a similar-but-different quirk, but a separate PR will address that.
• Memcache and Memcached look buggy. I don't have these systems locally, so I haven't tested or patched.

[civibot]

(Standard links)

• If this is your first pull-request for CiviCRM, please browse CONTRIBUTING.md for information about the development and testing processes.
• If you are reviewing this pull-request, you may wish to consult the test sites and the Review Standards (long template, short template).

[eileenmcnaughton]

I stepped through this and it makes sense to me. I can't see a risk with it. Merging