liam-wiltshire/laravel-jit-loader is an extension to the default Laravel Eloquent model to 'very lazy eager load' relationships with performance comparable with eager loading.
liam-wiltshire/laravel-jit-loader is available as a composer package:
composer require liam-wiltshire/laravel-jit-loader
Once installed, use the \LiamWiltshire\LaravelJitLoader\Concerns\AutoloadsRelationships
trait in your model, or have your models extend the \LiamWiltshire\LaravelJitLoader\Model
class instead of the default eloquent model, and JIT loading will be automatically enabled.
In order to avoid N+1 issues, you'd normally load your required relationships while building your collection:
$books = App\Book::with(['author', 'publisher'])->get();
Or otherwise after the fact, but before use:
$books = App\Book::all();
if ($someCondition) {
$books->load('author', 'publisher');
}
In some situations however, this may not be possible - perhaps front-end developers are able to make changes to templates without touching the code, or perhaps during development you know don't which relationships you'll need anyway. This change will track if your models belong to a collection, and if they do and a relationship is called that hasn't already been loaded, the relationship will be loaded across the whole collection just in time for use.
This is used in a number of production applications with no issues. It's also been tested against a (rather constructed) test, pulling out staff, companies and addresses - while this isn't a 'real life' representation, it should give an idea of what it can do:
public function handle()
{
//Count the number of queries
$querycount = 0;
DB::listen(function ($query) use (&$querycount) {
$querycount++;
});
$startTime = microtime(true);
$staff = Staff::where('name', 'LIKE', 'E%')->orWhere('name', 'LIKE', 'P%')->get();
/**
* @var Staff $st
*/
foreach ($staff as $st) {
/**
* @var Company $company
*/
$company = $st->company;
echo "\n\nName: {$st->name}\n";
echo "Company Name: {$company->name}\n";
foreach ($company->address as $address) {
echo "Addresses: {$address->address_1}, ";
}
}
$endTime = microtime(true);
echo "\n\n=========================\n\n";
echo "Queries Run: {$querycount}\n";
echo "Execution Time: " . ($endTime - $startTime) . "\n";
echo "Memory:" . memory_get_peak_usage(true)/1024/1024 . "MiB";
echo "\n\n";
}
Running this locally against a database with 200 companies, 1157 addresses and 39685 staff:
Queries Run: 10739
Execution Time: 17.090859889984
Memory: 70MiB
Queries Run: 3
Execution Time: 1.7261669635773
Memory: 26MiB
Queries Run: 3
Execution Time: 1.659285068512
Memory: 26MiB
As you can see the different between JIT loading and traditional eager loading is small (c. 0.067 seconds in our above test), so you can likely rely on JIT loader to protect you.
However, if you want to log when the JIT loader is used so that you can do back and correct them later, you can add a $logChannel
property to your models to ask the trait to log into that channel as configured in Laravel
class Address extends Model
{
use AutoloadsRelationships;
/**
* @var string
*/
protected $logChannel = 'jit-logger';
public function company()
{
return $this->belongsTo(Company::class);
}
}
This is an early release based on specific use cases. At the moment the autoloading will only be used when the relationship is loaded like a property e.g. $user->company->name
instead of $user->company()->first()->name
. I am working on supporting relations loaded in alternate ways, however there is more complexity in that so there isn't a fixed timescale as of yet!
With any eager loading, a sufficiently large collection can cause memory issues. The JIT model specifies a threshold for autoloading. This is set to 6000 by default, but can be changed by overriding the $autoloadThreshold
property on a model-by-model basis.