brekitomasson/laravel-tagged-cache

A wrapper around the cache implementation to reduce complexity regarding caching in your Models

v2.0.3 2023-02-22 03:10 UTC

This package is auto-updated.

Last update: 2024-04-22 05:31:12 UTC


README

A fairly straight-forward wrapper around the cache implementation to reduce complexity regarding caching in your Models. It has no external dependencies or configuration, and should "just work" out of the box.

Note, however, that this package requires you to be using a Cache implementation that supports tags. This means that it will not work if you are using file, dynamodb or database cache drivers.

Installation

composer require brekitomasson/laravel-tagged-cache

Usage

In any Laravel Models, just use BrekiTomasson\LaravelTaggedCache\HasTaggedCache, and you will be able to create reliable caches that do not require a bunch of special handling regarding your cache keys. Instead of needing to come up with complex naming rules for your cache keys, you can use the same cache key for everything, as the differentiation will be done using the cache tags instead. This allows you to do things like:

public function getDisplayNameAttribute(): string
{
  return $this->taggedCache()->remember(
      key: 'displayname', 
      ttl: now()->addHour(), 
      callback: fn () => $this->nickname ?? $this->name ?? $this->email,
  );
}

This method, if in your User model, will store the cache key displayname with the tags users and user:23 when used to get the display_name attribute for a user with the ID of 23.

You can also provide any amount of additional strings to the taggedCache() method, and those strings will be added to the list of tags. If, for example, you put something like this in your BlogEntry model:

public function getContentHtml(): string
{
    return $this->taggedCache('markdown')->remember(
        key: 'content:html',
        ttl: now()->addHour(),
        callback: fn () => Markdown::parse($this->content)->toHtml(),
    );
}

Then the content:html key will be tagged with blog_entries, blog_entry:25, and markdown. This allows you to do Cache::tags('markdown')->flush() if you've just made changes to your Markdown implementation and want all Markdown-related caches in all models to be cleared. Since a cache key will only get a hit if all tags match, this means that anything tagged with markdown in this way will automatically be forgotten, no matter which model it may be connected to.

Advanced Usage and Recommendations

1 - Cache Invalidation (optional, but recommended!)

They say that there are only two hard things in Computer Science: cache invalidation and naming things. Since this package makes it so much easier to cache things, you might be more willing to put things in your cache than you're used to, meaning you'll be more prone to get cache-related problems.

The normal way of working would be to build Observers that track changes to your models and flush caches based on which attributes have been modified, but this package offers a shortcut. By implementing a single method in your model, all caches for that specific instance of that model will be flushed. For example, if you were to put this in your Model:

public function flushTaggedCacheOnAttributeUpdate(): array
{
    return ['name', 'display_name', 'status'];
}

With this method in place, if you were to ModelName::find(132)->update(['display_name' => 'ACME Systems']); the package would automatically flush all caches related to that row, as the attribute display_name is listed inside the array returned by flushTaggedCacheOnAttributeUpdate(). Caches for other rows in that model will remain in place. This is the equivalent of building a Model Observer with a method like:

public function updated(ModelName $model): void
{
    if ($model->isDirty(['name', 'display_name', 'status'])) {
        $model->taggedCache()->flush();
    }
}

Note: Caches will always be flushed for rows that are deleted. This behavior cannot be disabled.

2 - Getting and Caching a Single Attribute

Instead of having to write $this->taggedCache()->remember('name', now()->addDay(), fn () => $this->name) to get a cached instance of a single attribute from a model, there's a simple helper function available. The code above can be rewritten: $this->getCachedAttribute('name'). The cache is stored with a key named after the attribute for 24 hours.

Under the hood, it uses Laravel's own getAttribute() method, which means you can use it to return relationships and attributes made available through Mutators and/or Accessor functions, and it takes your $casts into account.

3 - Personalizing the Name of the Cache Tags

By default, the name of the database table underlying the Model will be used as a basis for the Cache tags. If your model is Country and the underlying database table is countries, then Country::find(23)->taggedCache() will use the tags countries and country:23. The latter is generated using Laravel's Str::singular() method.

To override what base name to use in your cache tags, you can implement the getCacheTagIdentifier() method in your model. Any string returned by this method will automatically be converted into snake_case, but please try to keep any implementation of this to a plural string. For example:

public function getCacheTagIdentifier(): string
{
    return 'Comic Books';
}

If this method were in your model, then Model::find(42)->taggedCache() would use the tags comic_books and comic_book:42.

4 - Cache Duration Helper

Instead of relying on things like now()->addDays(3) when setting up how long a cache key should be remembered, this package contains an invokable and callable Enum class called TimeSpan that contains a whole host of pre-defined values, defined in seconds. Using this will save you the processing time required for Carbon to get the current system time, add your given duration to it, and for Laravel to then calculate the difference between the current time and the returned time. It may not seem like an expensive operation, but it all adds up quite quickly if you're using it often enough. To use the TimeSpan Enum in your cache statements, just do something like this:

public function getCountryNameAttribute(): string
{
    return $this->taggedCache()->remember(
        'country-name',
        TimeSpan::ONE_WEEK(),
        fn() => $this->country->name
    );
}

If there is not an appropriate case defined in the Enum, such as if you want exactly 32 minutes, or if you think it looks cleaner to write it that way, there are static methods for minutes, days, and weeks which you can invoke using syntax like TimeSpan::minutes(32) or TimeSpan::weeks(9). These always return the appropriate number of seconds as an integer.

public function getOpenTicketCount(): int
{
    return $this->taggedCache()->remember(
        'open-tickets',
        TimeSpan::minutes(3),
        fn() => $this->tickets->whereNotIn('status', [TicketStatus::CLOSED, TicketStatus::PENDING])->count();
    );
}

Note: The TimeSpan class does not have values or methods for months, as these are of variable length and cannot adequately be defined in seconds. The longest duration available as an Enum case is TimeSpan::FOUR_WEEKS(). To get a value approximating three months, you can use something like TimeSpan::weeks(12) or TimeSpan::days(90).

Future Development Ideas

These are things I've been thinking about when it comes to future development for this package. Some things listed below will be implemented, others won't. Pull Requests and suggestions are always welcome.

  • Is TimeSpan a good enough name for that class? I want to avoid potential naming conflicts, so I'm avoiding calling the class simply Seconds or something like that, but I also want the name to be descriptive enough for what the class is all about, which TimeSpan doesn't necessarily feel like.
  • Does TimeSpan really have to be an Enum? Feels like most of what I'm doing there can be done with just a normal class, and maybe even in a better way than currently.
  • Add ability to flush model caches on any change, not just to attributes listed in overridden method.

Copyright / License

This package is distributed under an MIT license. See the LICENSE file for more information.