Caching

Using the cache is very easy in Nette, while it also covers very advanced caching needs. It is designed for performance and 100% durability. Basically, you will find adapters for the most common backend storage. Allows tags-based invalidation, cache stampede protection, time expiration, etc.

Installation

Download and install the package using Composer:

composer require nette/caching

Basic Usage

The center of work with the cache is the object Nette\Caching\Cache. We create its instance and pass the so-called storage to the constructor as a parameter. Which is an object representing the place where the data will be physically stored (database, Memcached, files on disk, …). You get the storage object by passing it using dependency injection with type Nette\Caching\Storage. You will find out all the essentials in section Storage.

In version 3.0, the interface still had the I prefix, so the name was Nette\Caching\IStorage. Also, the constants of the Cache class were capitalized, so for example Cache::EXPIRE instead of Cache::Expire.

For the following examples, suppose we have an alias Cache and a storage in the variable $storage.

use Nette\Caching\Cache;

$storage = /* ... */; // instance of Nette\Caching\Storage

The cache is actually a key–value store, so we read and write data under keys just like associative arrays. Applications consist of a number of independent parts, and if they all used one storage (for idea: one directory on a disk), sooner or later there would be a key collision. The Nette Framework solves the problem by dividing the entire space into namespaces (subdirectories). Each part of the program then uses its own space with a unique name and no collisions can occur.

The name of the space is specified as the second parameter of the constructor of the Cache class:

$cache = new Cache($storage, 'Full Html Pages');

We can now use object $cache to read and write from the cache. The method load() is used for both. The first argument is the key and the second is the PHP callback, which is called when the key is not found in the cache. The callback generates a value, returns it and caches it:

$value = $cache->load($key, function () use ($key) {
	$computedValue = /* ... */; // heavy computations
	return $computedValue;
});

If the second parameter is not specified $value = $cache->load($key), the null is returned if the item is not in the cache.

The great thing is that any serializable structures can be cached, not only strings. And the same applies for keys.

The item is cleared from the cache using method remove():

$cache->remove($key);

You can also cache an item using method $cache->save($key, $value, array $dependencies = []). However, the above method using load() is preferred.

Memoization

Memoization means caching the result of a function or method so you can use it next time instead of calculating the same thing again and again.

Methods and functions can be called memoized using call(callable $callback, ...$args):

$result = $cache->call('gethostbyaddr', $ip);

The function gethostbyaddr() is called only once for each parameter $ip and the next time the value from the cache will be returned.

It is also possible to create a memoized wrapper for a method or function that can be called later:

function factorial($num)
{
	return /* ... */;
}

$memoizedFactorial = $cache->wrap('factorial');

$result = $memoizedFactorial(5); // counts it
$result = $memoizedFactorial(5); // returns it from cache

Expiration & Invalidation

With caching, it is necessary to address the question that some of the previously saved data will become invalid over time. Nette Framework provides a mechanism, how to limit the validity of data and how to delete them in a controlled way (“to invalidate them”, using the framework's terminology).

The validity of the data is set at the time of saving using the third parameter of the method save(), eg:

$cache->save($key, $value, [
	$cache::Expire => '20 minutes',
]);

Or using the $dependencies parameter passed by reference to the callback in the load() method, eg:

$value = $cache->load($key, function (&$dependencies) {
	$dependencies[Cache::Expire] = '20 minutes';
	return /* ... */;
});

Or using the 3rd parameter in the load() method, eg:

$value = $cache->load($key, function () {
	return ...;
}, [Cache::Expire => '20 minutes']);

In the following examples, we will assume the second variant and thus the existence of a variable $dependencies.

Expiration

The simplest expiration is the time limit. Here's how to cache data valid for 20 minutes:

// it also accepts the number of seconds or the UNIX timestamp
$dependencies[Cache::Expire] = '20 minutes';

If we want to extend the validity period with each reading, it can be achieved this way, but beware, this will increase the cache overhead:

$dependencies[Cache::Sliding] = true;

The handy option is the ability to let the data expire when a particular file is changed or one of several files. This can be used, for example, for caching data resulting from procession these files. Use absolute paths.

$dependencies[Cache::Files] = '/path/to/data.yaml';
// or
$dependencies[Cache::Files] = ['/path/to/data1.yaml', '/path/to/data2.yaml'];

We can let an item in the cache expired when another item (or one of several others) expires. This can be used when we cache the entire HTML page and fragments of it under other keys. Once the snippet changes, the entire page becomes invalid. If we have fragments stored under keys such as frag1 and frag2, we will use:

$dependencies[Cache::Items] = ['frag1', 'frag2'];

Expiration can also be controlled using custom functions or static methods, which always decide when reading whether the item is still valid. For example, we can let the item expire whenever the PHP version changes. We will create a function that compares the current version with the parameter, and when saving we will add an array in the form [function name, ...arguments] to the dependencies:

function checkPhpVersion($ver): bool
{
	return $ver === PHP_VERSION_ID;
}

$dependencies[Cache::Callbacks] = [
	['checkPhpVersion', PHP_VERSION_ID] // expire when checkPhpVersion(...) === false
];

Of course, all criteria can be combined. The cache then expires when at least one criterion is not met.

$dependencies[Cache::Expire] = '20 minutes';
$dependencies[Cache::Files] = '/path/to/data.yaml';

Invalidation Using Tags

Tags are a very useful invalidation tool. We can assign a list of tags, which are arbitrary strings, to each item stored in the cache. For example, suppose we have an HTML page with an article and comments, which we want to cache. So we specify tags when saving to cache:

$dependencies[Cache::Tags] = ["article/$articleId", "comments/$articleId"];

Now, let's move to the administration. Here we have a form for article editing. Together with saving the article to a database, we call the clean() command, which will delete cached items by tag:

$cache->clean([
	$cache::Tags => ["article/$articleId"],
]);

Likewise, in the place of adding a new comment (or editing a comment), we will not forget to invalidate the relevant tag:

$cache->clean([
	$cache::Tags => ["comments/$articleId"],
]);

What have we achieved? That our HTML cache will be invalidated (deleted) whenever the article or comments change. When editing an article with ID = 10, the tag article/10 is forced to be invalidated and the HTML page carrying the tag is deleted from the cache. The same happens when you insert a new comment under the relevant article.

Tags require Journal.

Invalidation by Priority

We can set the priority for individual items in the cache, and it will be possible to delete them in a controlled way when, for example, the cache exceeds a certain size:

$dependencies[Cache::Priority] = 50;

Delete all items with a priority equal to or less than 100:

$cache->clean([
	$cache::Priority => 100,
]);

Priorities require so-called Journal.

Clear Cache

The Cache::All parameter clears everything:

$cache->clean([
	$cache::All => true,
]);

Bulk Reading

For bulk reading and writing to cache, the bulkLoad() method is used, where we pass an array of keys and obtain an array of values:

$values = $cache->bulkLoad($keys);

Method bulkLoad() works similarly to load() with the second callback parameter, to which the key of the generated item is passed:

$values = $cache->bulkLoad($keys, function ($key, &$dependencies) {
	$computedValue = /* ... */; // heavy computations
	return $computedValue;
});

Using with PSR-16

To use Nette Cache with the PSR-16 interface, you can utilize the PsrCacheAdapter. It allows seamless integration between Nette Cache and any code or library that expects a PSR-16 compatible cache.

$psrCache = new Nette\Bridges\Psr\PsrCacheAdapter($storage);

Now you can use $psrCache as a PSR-16 cache:

$psrCache->set('key', 'value', 3600); // stores the value for 1 hour
$value = $psrCache->get('key', 'default');

The adapter supports all methods defined in PSR-16, including getMultiple(), setMultiple(), and deleteMultiple().

Output Caching

The output can be captured and cached very elegantly:

if ($capture = $cache->capture($key)) {

	echo ... // printing some data

	$capture->end(); // save the output to the cache
}

In case that the output is already present in the cache, the capture() method prints it and returns null, so the condition will not be executed. Otherwise, it starts to buffer the output and returns the $capture object using which we finally save the data to the cache.

In version 3.0 the method was called $cache->start().

Caching in Latte

Caching in Latte templates is very easy, just wrap part of the template with tags {cache}...{/cache}. The cache is automatically invalidated when the source template changes (including any included templates within the {cache} tags). Tags {cache} can be nested, and when a nested block is invalidated (for example, by a tag), the parent block is also invalidated.

In the tag it is possible to specify the keys to which the cache will be bound (here the variable $id) and set the expiration and invalidation tags

{cache $id, expire: '20 minutes', tags: [tag1, tag2]}
	...
{/cache}

All parameters are optional, so you don't have to specify expiration, tags, or keys.

The use of the cache can also be conditioned by if – the content will then be cached only if the condition is met:

{cache $id, if: !$form->isSubmitted()}
	{$form}
{/cache}

Storages

A storage is an object that represents where data is physically stored. We can use a database, a Memcached server, or the most available storage, which are files on disk.

You get the storage object by passing it using dependency injection with the Nette\Caching\Storage type. By default, Nette provides a FileStorage object that stores data in a subfolder cache in the directory for temporary files .

You can change the storage in the configuration:

services:
	cache.storage: Nette\Caching\Storages\DevNullStorage

FileStorage

Writes the cache to files on disk. The storage Nette\Caching\Storages\FileStorage is very well optimized for performance and above all ensures full atomicity of operations. What does it mean? That when using the cache, it cannot happen that we read a file that has not yet been completely written by another thread, or that someone would delete it “under your hands”. The use of the cache is therefore completely safe.

This storage also has an important built-in feature that prevents an extreme increase in CPU usage when the cache is cleared or cold (ie not created). This is cache stampede prevention. It happens that at one moment there are several concurrent requests that want the same thing from the cache (eg the result of an expensive SQL query) and because it is not cached, all processes start executing the same SQL query. The processor load is multiplied and it can even happen that no thread can respond within the time limit, the cache is not created and the application crashes. Fortunately, the cache in Nette works in such a way that when there are multiple concurrent requests for one item, it is generated only by the first thread, the others wait and then use the generated result.

Example of creating a FileStorage:

// the storage will be the directory '/path/to/temp' on the disk
$storage = new Nette\Caching\Storages\FileStorage('/path/to/temp');

MemcachedStorage

The server Memcached is a high-performance distributed storage system whose adapter is Nette\Caching\Storages\MemcachedStorage. In the configuration, specify the IP address and port if it differs from the standard 11211.

Requires PHP extension memcached.

services:
	cache.storage: Nette\Caching\Storages\MemcachedStorage('10.0.0.5')

MemoryStorage

Nette\Caching\Storages\MemoryStorage is a storage that stores data in a PHP array and is thus lost when the request is terminated.

SQLiteStorage

The SQLite database and adapter Nette\Caching\Storages\SQLiteStorage offer a way to cache in a single file on disk. The configuration will specify the path to this file.

Requires PHP extensions pdo and pdo_sqlite.

services:
	cache.storage: Nette\Caching\Storages\SQLiteStorage('%tempDir%/cache.db')

DevNullStorage

A special implementation of storage is Nette\Caching\Storages\DevNullStorage, which does not actually store data at all. It is therefore suitable for testing if we want to eliminate the effect of the cache.

Using Cache in Code

When using caching in code, you have two ways how to do it. The first is that you get the storage object by passing it using dependency injection and then create an object Cache:

use Nette;

class ClassOne
{
	private Nette\Caching\Cache $cache;

	public function __construct(Nette\Caching\Storage $storage)
	{
		$this->cache = new Nette\Caching\Cache($storage, 'my-namespace');
	}
}

The second way is that you get the storage object Cache:

class ClassTwo
{
	public function __construct(
		private Nette\Caching\Cache $cache,
	) {
	}
}

The Cache object is then created directly in the configuration as follows:

services:
	- ClassTwo( Nette\Caching\Cache(namespace: 'my-namespace') )

Journal

Nette stores tags and priorities in a so-called journal. By default, SQLite and file journal.s3db are used for this, and PHP extensions pdo and pdo_sqlite are required.

You can change the journal in the configuration:

services:
	cache.journal: MyJournal

DI Services

These services are added to the DI container:

Turning Off Cache

One of the ways to turn off caching in the application is to set the storage to DevNullStorage:

services:
	cache.storage: Nette\Caching\Storages\DevNullStorage

This setting does not affect the caching of templates in Latte or the DI container, as these libraries do not use the services of nette/caching and manage their cache independently. Moreover, their cache does not need to be turned off in development mode.