Convert complex query from Laravel to SQL - php

Question

I'm trying to move some statistics logic entirely to SQL to vastly improve performance, however it's quite complex and I'm not sure how to achieve this using procedures/functions in SQL - or whether it's even possible.
Stats table looks like this:
Row in the table looks like this:
This is the code I'm trying to convert. $this->stats is an Eloquent Collection of all the rows in the table:
return $this->stats
->groupBy(function ($stat) {
return $stat->created_at->format('Y-m-d');
})
->map(function ($stats) {
$times = [];
for ($hour = 0; $hour < 24; $hour++) {
$thisHour = $stats->filter(function ($stat) use ($hour) {
return (int) $stat->created_at->format('H') === $hour;
});
$times[$hour] = $thisHour->isNotEmpty()
? $thisHour->sum(function ($stat) {
return $stat->data->count;
}) : 0;
}
return $times;
});
This outputs something like this:
{
"2018-12-20": {
0: 54,
1: 87,
2: 18,
3: 44,
4: 35,
...
}
}
So it's grouped by date, and each date contains 0-23 (the hours of the day) with the corresponding value (in this case the summation of the row's data->count property).
In a different query I've been able to get the summation of the data->count property by using this:
SELECT SUM(data->>\"$.count\") AS value
So is this even possible to do in SQL? I'm imaging there to be a date columns, plus the hours column, so hour_0, hour_1 etc. with the values underneath.
Any help would be greatly appreciated!

Answer 1

The first step I would take to understand how Laravel is doing it would be enabling query logging and dumping the query;
DB::connection()->enableQueryLog();
// ... do your query
$queries = DB::getQueryLog();
That will allow you to recreated the Eloquent query in plain SQL. From there, you can work on optimizing and condensing.
https://laravel.com/docs/5.0/database#query-logging

Answer 2

UDPATED
As you have now included the table structure, the following query should give you required results in tabular format. You need to parse the results to convert to json.
Updated SQL
SELECT DATE_FORMAT(stats.updated_at, "%Y-%m-%d") AS `date`,
HOUR(stats.updated_at) AS `hour`,
COUNT(stats.id) AS record_count
FROM stats
GROUP BY `date`, `hour`
ORDER BY 1 ASC, 2 ASC
The resulting tabular data would look similar to this.