Build scalable Laravel apps without compromising code quality and the least amount of technical debt. Effortless code navigation, onboarding and reviews. Tailored for Microservices and mindful of Monoliths.
Lucid is a software architecture to build scalable Laravel projects. It incorporates Command Bus and Domain Driven Design
at the core, upon which it builds a stack of directories and classes to organize business logic.
It also derives from SOA (Service Oriented Architecture) the notion of encapsulating functionality
within a service and enriches the concept with more than the service being a class.
Use Lucid to:
This architecture is in an amalgamation of best practices, design patterns and proven methods.
Feature
, Job
or Operation
.If you prefer a video, watch the announcement at Laracon EU 2016:
In a typical MVC application, Lucid will be the bond between the application’s entrypoints and the units that do the work,
securing code form meandring in drastic directions:
At a glance…
Provides the “kernel” to do the heavy lifting of the tedious stuff such as request/response lifecycle, dependency
injection, and other core functionalities.
Extends the framework to provide higher level abstractions that are custom to the application and can be shared
across the entire stack rather than being case-specific.
Examples of what could go into foundation are:
DateTime
a support class for common date and time functionsJsonSerializableInterface
that is used to identify an object to be serializable from and to JSON formatProvide separation to categorize jobs and corresponding classes that belong to the same topic. A domain operates in isolation
from other domains and exposes its functionalities to features and operations through Lucid jobs only.
Consider the structure below for an example on what a domain may look like:
app/Domains/GitHub
├── GitHubClient
├── Jobs
│ ├── FetchGitHubRepoInfoJob
│ └── LoginWithGitHubJob
├── Exceptions
│ ├── InvalidTokenException
│ └── RepositoryNotFoundException
└── Tests
└── GitHubClientTest
└── Jobs
├── FetchGitHubReposJobTest
└── LoginWithGitHubJobTest
documentation contains more details on working with domains.
Are directories rich in functionality, used to separate a [Monolith]({{<ref “/micro-vs-monolith/#monolith”>}}) into
areas of focus in a multi-purpose application.
Consider the example of an application where we enter food recipes and would want our members to have discussions in a forum,
we would have two services: 1) Kitchen, 2) Forum where the kitchen would manage all that’s related to recipes, and forum is obvious:
app/Services
├── Forum
└── Kitchen
and following is a single service’s structure, highlighted are the Lucid specific directories:
app/Services/Forum ├── Console │ └── Commands ├── Features ├── Operations ├── Http │ ├── Controllers │ └── Middleware ├── Providers │ ├── KitchenServiceProvider │ ├── BroadcastServiceProvider │ └── RouteServiceProvider ├── Tests │ └── Features │ └── Operations ├── database │ ├── factories │ ├── migrations │ └── seeds ├── resources │ ├── lang │ └── views └── routes ├── api ├── channels ├── console └── web
documentation has more examples of services and their contents.
Represent a human-readable application feature in a class. It contains the logic that implements the feature but with the least
amount of detail, by running jobs from domains and operations at the application or service level.
Serving the Feature class will be the only line in a controller’s method (in MVC), consequently achieving the thinnest form of controllers.
class AddRecipeFeature extends Feature
{
public function handle(AddRecipe $request)
{
$price = $this->run(CalculateRecipePriceOperation::class, [
'ingredients' => $request->input('ingredients'),
]);
$this->run(SaveRecipeJob::class, [
'price' => $price,
'user' => Auth::user(),
'title' => $request->input('title'),
'ingredients' => $request->input('ingredients'),
'instructions' => $request->input('instructions'),
]);
return $this->run(RedirectBackJob::class);
}
}
documentation about features expands on how to serve them as classes from anywhere.
Their purpose is to increase the degree of code reusability by piecing jobs together to provide composite functionalities from across domains.
class NotifySubscribersOperation extends Operation
{
private int $authorId;
public function __construct(int $authorId)
{
$this->authorId = $authorId;
}
/**
* Sends notifications to subscribers.
*
* @return int Number of notification jobs enqueued.
*/
public function handle(): int
{
$author = $this->run(GetAuthorByIDJob::class, [
'id' => $this->authorId,
]);
do {
$result = $this->run(PaginateSubscribersJob::class, [
'authorId' => $this->authorId,
]);
if ($result->subscribers->isNotEmpty()) {
// it's a queueable job so it will be enqueued, no waiting time
$this->run(SendNotificationJob::class, [
'from' => $author,
'to' => $result->subscribers,
'notification' => 'article.published',
]);
}
} while ($result->hasMorePages());
return $result->total;
}
}
documentation goes over this simple yet powerful concept.
For a scalable set of interconnected data elements, we’ve created a place for them in app/Data
,
because most likely over time writing the application there could develop a need for more than Models in data,
such as Repositories, Value Objects, Collections and more.
app/Data
├── Models
├── Values
├── Collections
└── Repositories
There are valuable advantages to what may seem as overengineering.
By dissecting our application into small building blocks of code - a.k.a units - we’ve instantly opened the door for a high
degree of code sharing across the application with Data and Domains, as well as replaceability with the least amount of friction
and technical debt.
By setting boundaries you would’ve taken a step towards proetcting application code from growing unbearably large
and made it easier for new devs to onboard. Most importantly, that you’ve reduced technical debt to the minimum so that you don’t
have to pay with bugs and sleepless nights; code doesn’t run on good intentions nor wishes.
When our application scales we’d typically have a bunch of instances of it running in different locations,
at some point we would want to activate certain parts of our codebase in some areas and shut off others.
Here’s a humble example of running Api, Back Office and Web App instances of the same application, which in Lucid terminology
are services that share functionality through data and domains:
To encourage active collaboration, Lucid strongly encourages contribution through pull requests.
“Bug reports” may be searched or created in issues or sent in the form of a pull request containing a failing test or steps to reproduce the bug.
If you file a bug report, your issue should contain a title and a clear description of the issue. You should also include as much relevant information as possible and a code sample that demonstrates the issue. The goal of a bug report is to make it easy for yourself - and others - to replicate the bug and develop a fix.
⏱ PRs and issues are usually checked about three times a week so there is a high chance yours will be picked up soon.
The Lucid Architecture source code is on GitHub as lucidarch/lucid.
Lucid Architecture’s GitHub issue trackers are not intended to provide help or support. Instead, use one of the following channels:
#support
channelYou may propose new features or improvements of existing Lucid Architecture behaviour in the Lucid Discussins.
If you propose a new feature, please be willing to implement at least some of the code that would be needed to complete the feature, or collaborate on active ideation in the meantime.
Informal discussion regarding bugs, new features, and implementation of existing features takes place in the #internals
channel of the Lucid Slack workspace.
Abed Halawi, the maintainer of Lucid, is typically present in the channel on weekdays from 8am-5pm EEST (Eastern European Summer Time), and sporadically present in the channel at other times.
The main
branch is what contains the latest live version and is the one that gets released.
feature/queueable-units
, fix/issue-31
)git push origin feature/queueable-units
main
branch, which will run tests for your edits⏱ PRs and issues are usually checked about three times a week.
Following are the steps to setup for development on Lucid:
Assuming we’re in
~/dev
directory…
[your username]/lucid
which will create a lucid
folder at ~/dev/lucid
composer create-project laravel/laravel myproject
composer.json
"require": {
"...": "",
"lucidarch/lucid": "@dev"
},
"repositories": [
{
"type": "path",
"url": "~/dev/lucid",
"options": {
"symlink": true
}
}
],
"minimum-stability": "dev",
Make sure you change the
url
to the absolute path of your directory
composer update
to create the symlinkNow all your changes in the lucid directory will take effect automatically in the project.
If you discover a security vulnerability within Lucid, please send an email to Abed Halawi at [email protected].
All security vulnerabilities will be promptly addressed.
Lucid Architecture follows the PSR-2 coding standard and the PSR-4 autoloading standard.
Below is an example of a valid Lucid Architecture documentation block. Note that the @param
attribute is followed by two spaces, the argument type, two more spaces, and finally the variable name:
/**
* Register a binding with the container.
*
* @param string|array $abstract
* @param \Closure|string|null $concrete
* @param bool $shared
* @return void
*
* @throws \Exception
*/
public function bind($abstract, $concrete = null, $shared = false)
{
//
}
The Lucid Architecture code of conduct is derived from the Laravel code of conduct. Any violations of the code of conduct may be reported to Abed Halawi ([email protected]):