优化查询PHP中几个键的值总和

Question

I have a table in Postgres with SIP-log.

|   in    |       out      |   sec   |
|   112   | sip/113-random |    12   | 
|   113   | sip/112-random |    45   | 
|   112   | sip/114-random |    40   | 
|   113   | sip/114-random |    35   | 
|   117   | sip/113-random |    11   | 
|   117   | sip/113-random |    25   | 
|   115   | sip/112-random |    98   | 
|   115   | sip/117-random |    78   | 
|   112   | sip/113-random |    18   | 

I need to sum all sec in "in" and "out" with single SIP and group by SIP number.

Example:

....
112 - 54
113 - 152
115 - 25
....

Now I use this code. But it runs in a loop and takes very long.

$array_sip=array(110,111 ....  199);

foreach ($array_sip as $sip) {
    $sum_out=0;
    $sum_in=0;
    $sql = "SELECT  SUM(sec) AS sec 
            FROM public.cdr 
            WHERE calldate::text like '".$date."%'  
            AND disposition='ANSWERED'
            AND out like 'SIP/".$sip."%'
            AND sec > 15
            ";  
            foreach ($db->query($sql) as $row) {
                $sum_out=$row['sec'];
            }
    $sql = "SELECT  SUM(sec) AS sec 
            FROM public.cdr 
            WHERE calldate::text like '".$date."%'  
            AND disposition='ANSWERED'
            AND in = '".$sip."'
            AND sec > 3";   

            foreach ($db->query($sql) as $row) {
                $sum_in=$row['sec'];
            }
    $sum=round((($sum_out+$sum_in)/60),1);
    $loading_level.= $sip.' - '.$sum;
}

How can I optimise the query?

Answer 1

You should compose and run a query like this:

SELECT substring(out FROM '\d+'),
       SUM(sec) AS sec
FROM public.cdr
WHERE calldate::text LIKE '<your date>%'
AND disposition = 'ANSWERED'
AND out LIKE ANY ('SIP/<sip1>-%', 'SIP/<sip2>-%', ...)
AND sec > 15
GROUP BY substring(out FROM '\d+');

You'll have to replace the <...> in the above with your actual values.

The LIKE ANY condition will pass everything that matches one of the patterns.

This way you can get it all in a single query, which will be more efficient.

Indexes on calldate::text and out may speed up processing.