Laravel scoreboard of more than 1 million users - php

Question

I'm working in a biggest application ( more than 1 million users ) and I try to get the ranking of each user in the scoreboard section but had this problem: the result is very very slow
This is the architecture of my database:
Schema::create('users', function (Blueprint $table) {
$table->increments('id');
...
});
Schema::create('topics', function (Blueprint $table) {
$table->increments('id');
...
});
The topics table have than 20 row
Schema::create('user_scores', function (Blueprint $table) {
$table->increments('id');
$table->integer('user_id')->unsigned();
$table->integer('topic_id')->unsigned();
$table->unique(['user_id', 'topic_id']);
$table->float('timer');
$table->integer('score');
});
The query to make rank for users
User::where('type',0)->get()->each(function ($user) {
$user->topics= $user->scores->sum('score');
$user->timing= $user->scores->sum('timer');
})->sort(function ($a, $b){
return ($b->topics - $a->topics) == 0
? ($a->timing - $b->timing)
: ($b->topics - $a->topics);
})->values()->each(function($user, $key){
$user->rank = $key +1;
});
Any optimization should I make the get the result quicker? Thanks.

Answer 1

As soon as you call get(), all(), find() or first() on a query builder, you'll ask the Eloquent engine to perform the query and return you the result. So in your case, all the sorting and grouping is performed in memory, which comes with incredibly bad performance.
What you can do is to improve your query:
User::query()
->where('type', 0)
->withCount('scores as topics')
->withCount(['scores as timing' => function ($query) {
$query->selectRaw('SUM(timer)'); // might look weird, but works...
}])
->orderBy('topics', 'desc')
->orderBy('timing', 'desc')
->get()
For the row number (or rank, or however you wanna call it), you might want to search through existing questions and answers. Answering that as well would be too much for this answer, to be honest. Clearly you should not use your approach though, as it will also calculate the row number in memory.
But obviously it is also important what you are doing with the query results. Are you displaying one million rows to the user? If so, the bottleneck will be the browser in the end for sure. You might want to consider using pagination with paginate() instead of get().