php-fig / cache-util Goto Github PK

It was suggested to me to move this php-fig/cache#16 here, but not sure where to put in this codebase

It would be great to add boilerplate implementations of Psr\Cache\Exception and Psr\Cache\InvalidArgumentException to this library.

  41  - it may use psr6 compliant cache
      warning: assert(): Assertion "$expiration instanceof \DateTimeInterface" failed in
      /home/bartosz/PhpstormProjects/portal_customer/vendor/fig/cache-util/src/BasicCacheItemTrait.php line 75
       0 vendor/phpspec/phpspec/src/PhpSpec/Runner/Maintainer/ErrorMaintainer.php:121
         throw new PhpSpec\Exception\Example\ErrorException("warning: assert(): Assert...")
       1 vendor/fig/cache-util/src/BasicCacheItemTrait.php:75
         assert("$expiration instanceof \D...")
       2 src/CustomerPortalAdminClient/CustomerPortalAdminClient.php:59
         Fig\Cache\Memory\MemoryCacheItem->expiresAt([obj:DateTime])

assert checks if argument is an instance of \DateTimeInterface which has been added in php 5.5 which in turn causes warning on php5.4.
http://php.net/manual/en/class.datetimeinterface.php

Examples:

• https://github.com/symfony/cache/blob/master/Adapter/NullAdapter.php
• https://github.com/tedious/Stash/blob/master/src/Stash/Driver/BlackHole.php

Same as https://github.com/php-fig/log/blob/master/Psr/Log/NullLogger.php

Use case

Cache as an optional injectable dependency. When class user not provides it -- we take ready NullPool and use it.

As I understand, psr/cache dont' want to accept the idea of cache optionality. I will try to explain why optinality is a good idea.

Ok, let's talk about MySQL server.

As far as I'm concerned, MySQL server is good example of proper design.

According to this article you may see that cache is optional. It doesn't matter whereas user enabled it or not. The result of the query will be the same in any case.

It means that null-cache is a good idea.

From user's point of view the cache is not always needed. It's an optimization, infrastructure layer tool. When we design system -- we dont' start from optimization, we start from domain logic and then after some time, when we find bottlenecks, we optimize 'em. So from scratch we allow class's user to use it without any cache-setup issues out of the box. It's a friendly design. Same as we see in great PSR/Log.

Testability of objects with injected cache is an issue. What it should take? Some real pre-cleaned cache pool or home-grown null implementation?

The main advantage of null object implementation is that it's behaviour is do nothing. In case of cache pool it's just never hit/get/save behaviour. There is one way to do it.