Skip to content

Khulan is a cache driver and content cache with NoSQL interface for Kirby using MongoDB

License

Notifications You must be signed in to change notification settings

bnomei/kirby-mongodb

Repository files navigation

🐎 Khulan - Kirby MongoDB

Release Downloads Coverage Maintainability Discord

Khulan is a cache driver and content cache with NoSQL interface for Kirby using MongoDB.

Installation

  • unzip master.zip as folder site/plugins/kirby-mongodb or
  • git submodule add https://github.com/bnomei/kirby-mongodb.git site/plugins/kirby-mongodb or
  • composer require bnomei/kirby-mongodb

MongoDB

There are various ways to install MongoDB. This is one way to do it on localhost for Mac OSX using Homebrew and MongoDB Atlas.

brew install mongodb-atlas
atlas setup # create account and sign-in
atlas deployments setup # -> select localhost
atlas deployments start # start the local mongodb

Usecase

The plugin caches all content files and keeps the cache up to date when you add/remove or update content. This cache will be used when constructing page/file/user objects making everything that involves model faster (even the Panel).

It will also allow you to query the content cache directly as a NoSQL database which might be very useful for some use-cases like filtering or searching content.

Setup

For each template you want to be cached you need to use a model to add the content cache logic using a trait.

site/models/default.php

class DefaultPage extends \Kirby\Cms\Page
{
    use \Bnomei\ModelWithKhulan;
}

Note

You can also use the trait for user models. File models are patched automatically.

Kirby's Content goes NoSQL

The plugin writes the content cache to a collection named khulan in the database. You can query this collection directly. It is not wrapped in an Cache object. This allows you to treat all your Kirby content as a NoSQL database.

// using the collection
$document = khulan()->find(['uuid' => 'XXX']);

// find a single page by uuid string
$page = khulan((string) $myIdOrUUID);

// find a single page by uuid by field-key
$page = khulan(['uuid' => 'page://betterharder']);

// find all pages with a template
$pages = khulan(['template' => 'post']);

// find a user by email or Id
$user = khulan($emailOrId);

// find a file by filename and template
$file = khulan(['filename' => 'my-image.jpg', 'template' => 'image']);

Special Fields with [], {} and [,]

The plugin creates modified copies of a few field types to make the YAML based content from Kirby ready for queries. You can use

  • fieldname[] to query for a value in an array from pages/files/user fields and
  • fieldname{} to query for an objectId in an array from pages/files/user fields and
  • fieldname[,] to query for fields in comma separated strings like tags/select/radio/checkbox/multiselect fields.
// find all pages that have another page linked
$pages = khulan([
    'related[]' => ['$in' => ['page://fasterstronger']],
]);
$pages = khulan([
    'related{}' => ['$in' => ['dc9f7835c2400cc4']], // objectId
]);

// find all products in the category 'books' or 'movies'
// that had been modified within the last 7 days
$pages = khulan([
    'template' => 'product',
    'category[,]' => ['$in' => ['books', 'movies']],
    'modified' => ['$gt' => time() - (7 * 24 * 60 * 60)]
]);

Example: Filtering and Resolving Relations

With Kirby's content cache as a NoSQL database you can do some advanced filtering. If you would do the same with Kirby's filtering on collection you would end up loading a lot pages and that is not as efficient. We can load the information we need directly from the cache without the need to load the full page object.

Why would we bother to read the content from a NoSQL based cache and not just put a cache around the native Kirby logic? Because the hard thing with caches is to know when to invalidate them. With the content cache we can query the NoSQL based cache directly and updating any model will be reflected instantly. Whereas with a cache around the native Kirby logic we would need rebuild the cache on every change (like with the pages cache).

Let's assume we have two models: film and actor. The film model has a field actors which is a pages field with linked actor pages. We want to list all films with their actors.

Load 1000 films and their actors, total of 6429 pages accessed in 250ms

$films = page('films')->children();
foreach ($films as $film) {
    echo $film->title();
    $actors = [];
    foreach ($film->actors()->toPages() as $actor) {
        $actors[] = $actor->title();
    }
    echo implode(', ', $actors);
}

Query the cache instead to get the same information in under 100ms

/** @var \MongoDB\Driver\Cursor $films */
$films = khulan()->aggregate([
    [
        // only get pages with template 'film'
        '$match' => ['template' => 'film'],
    ],
    [
        // lookup actors by their mongodb objectId
        // this will create an array of objects
        '$lookup' => [
            'from' => 'kirby',
            'localField' => 'actors{}',
            'foreignField' => '_id',
            'as' => 'actor_details',
        ],
    ],
    [
        // only get the fields we need
        // to make the query even faster
        '$project' => [
            'title' => 1,
            'id' => 1,
            'actor_details.title' => 1,
        ],
    ],
]);

/** @var \MongoDB\BSON\Document $film */
foreach ($films as $film) { ?>
<div>
    <a href="<?= url($film->id) ?>"><?= html($film->title) ?></a>
    <ul>
        <?php foreach ($film->actor_details as $actor) { ?>
            <li><?= html($actor->title) ?></li>
        <?php } ?>
    </ul>
</div>
<?php }

Note

This example is from my Sakila DB kit. You can use similar queries to filter and resolve relations of files and user objects.

MongoDB Client

You can access the underlying MongoDB client directly.

/** @var \MongoDB\Client $client */
$client = \Bnomei\Mongodb::singleton()->client();
$client = mongo()->client();

$collection = $client->selectCollection('my-database', 'my-collection');

Cache

You can either use the cache directly or use it as a cache driver for Kirby.

The MongoDB based cache will, compared to the default file-based cache, perform worse! This is to be expected as web-servers are optimized for handling requests to a couple of hundred files and keep them in memory.

$cache = mongo()->cache();

$cache->set('mykey', 'myvalue', 5); // ttl in minutes
$value = $cache->get('mykey');

$cache->set('tagging', [
    'tags' => ['tag1', 'tag2'],
    'other_value' => 'myothervalue',
]);

As with regular Kirby cache you can also use the getOrSet method to wrap your time expensive code into a closure and only execute it when the cache is empty.

$cache = mongo()->cache();

$value = $cache->getOrSet('mykey', function() {
    sleep(5); // like a API call, database query or filtering content
    return 'myvalue';
}, 5); 

Using the MongoDB-based cache will allow you to perform NoSQL queries on the cache and do advanced stuff like filtering my tags or invalidating many cache entries at once.

// NOTE: we are using the cacheCollection() method here
$collection = mongo()->cacheCollection();

// find all that have the tag 'tag1' and are NOT expired
$documents = $collection->find([
    'value.tags' => ['$in' => ['tag1']],
    'expires_at' => ['$gt' => time()],
]);
foreach ($documents as $document) {
    // $document->value->tags
    // $document->value->other_value
}

// delete any cache entry older than 5 minutes
$deleteResult = $collection->deleteMany([
    'expires_at' => ['$lt' => time() - 5*60]
]);
$deletedCount = $deleteResult->getDeletedCount();

// you can also manually clean expired cache entries
mongo()->clean();

Note

Querying the cache directly is not recommended for regular use-cases. It will not check for the expiration of the cache entries and delete them automatically.

Using the Cache Driver in Kirby

You can also use the MongoDB-based cache as a cache driver for Kirby. This will allow you to use it for caching of other extensions in Kirby.

site/config/config.php

return [
    // ... other options
    
    // example of using mongodb as cache driver for storing uuids
    // instead of the default file-based cache
    'cache' => [
        'uuid' => [
            'type' => 'mongodb',
        ],
    ],
];

Settings

bnomei.mongodb. Default Description
host 127.0.0.1
port 27017
username null
password null
database kirby you can give it any name you want and MongoDB will create it for you
uriOptions []
driverOptions []
auto-clean-cache true will clean the cache once before the first get()
khulan.read false read from cache is disabled by default as loading from file might be faster
khulan.write true write to cache for all models that use the ModelWithKhulan trait
khulan.patch-files-class true monkey-patch the \Kirby\CMS\Files class to use Khulan for caching its content

Disclaimer

This plugin is provided "as is" with no guarantee. Use it at your own risk and always test it yourself before using it in a production environment. If you find any issues, please create a new issue.

License

MIT

It is discouraged to use this plugin in any project that promotes racism, sexism, homophobia, animal abuse, violence or any other form of hate speech.