Slug generation performance improvement

[RomainMazB]

The EXISTS sql statement will return true at the first existing value where the COUNT will uselessly continue to loop through the whole table which is inaccurate for this use case.

The syntax is more readable too IMO.

[bennothommo]

I've given this a test by running the following in the Sluggable tests:

public function testSlugGenerationLargeDataset()
{
    // Create 2500 records
    for ($i = 0; $i < 2500; ++$i) {
        TestModelSluggable::create([
            'name' => 'test row ' . $i,
        ]);
    }

    $testModel1 = TestModelSluggable::create(['name' => 'test']);

    // Create another 2500 records
    for ($i = 2500; $i < 5000; ++$i) {
        TestModelSluggable::create([
            'name' => 'test row ' . $i,
        ]);
    }

    DB::listen(function ($query) {
        print_r([
            'query' => $query->sql,
            'bindings' => $query->bindings,
            'time' => $query->time,
        ]);
    });

    $testModel2 = TestModelSluggable::create(['name' => 'test']);
}

I found a very, very, very small performance benefit for using exists() over count(). I mean incredibly minuscule - even with 5000 records. exists() ran in 0.05 seconds, count() ran in 0.06 seconds, but it fluctuated, so I assume there's rounding going on. This is with SQLite.

The logic is sound, so I'm not against merging it - it makes sense that we only care if a record exists with a given slug, we don't need to know how many records have the slug, but I just wanted to make it clear this is a micro-optimisation.

[LukeTowers]

While the gains are small, the change does take it from an issue that gets slower with scale and turns it into a consistent performance operation instead which is always better IMO.