CODEWITHSUNDEEP
phpmysql
LeoSam
LeoSam

Reputation: 4941

PHP/MYSQL ranking code cause slowness loading page PHP/MYSQL

I have one test page running #-tag trending which is stored in my database. The page takes a long time to load (up to 2 minutes) before the content is displayed, only one function used a that page

URL: http://www.sudanesetweeps.com/trendingtopics.php

How can I adjust the statement so the page load in less time?

This is my code:

<?php 

require_once("dbconnect.php");
require_once("lib_isarabic.php");

$query = "SELECT COUNT( * ) cnt, hashtags
FROM  tweets
WHERE tweeted_at >= DATE_SUB( NOW( ) , INTERVAL 2 DAY ) 
AND hashtags !=  ''
GROUP BY hashtags
ORDER BY cnt DESC LIMIT 100";

$res = mysql_query($query);

while($row = mysql_fetch_assoc($res) ) {

    $count = $row['cnt'];
    $hashtags = explode( " ", $row['hashtags'] );

    foreach($hashtags as $hashtag ) {
        if( strtolower($hashtag) != 'sudan' && strtolower($hashtag) != 'new' && strtolower($hashtag) != 'new' )     
            if( is_arabic($hashtag) ) 
                $topics_ara[strtolower( trim($hashtag) )] += $count;            
            else    
                $topics_eng[strtolower( trim($hashtag) )] += $count;        

    }
}

array_multisort($topics_ara, SORT_DESC);
array_multisort($topics_eng, SORT_DESC);

$index = 0;
foreach($topics_eng as $key=>$value) {

    $query = "SELECT count(*) cnt FROM ( 
            SELECT count(*), tweeted_by FROM tweets
            WHERE hashtags like '%$key%'
                      AND tweeted_at >= DATE_SUB( NOW( ) , INTERVAL 2 DAY ) 
            GROUP BY tweeted_by 
             ) AS T";


/*              $query = "
            SELECT count(*) FROM tweets
            WHERE hashtags like '%$key%'
                      AND tweeted_at > DATE_SUB( NOW( ) , INTERVAL 1 DAY ) ";
*/

        $res = mysql_query($query);               
        $row = mysql_fetch_assoc($res);

        if($row['cnt'] > 1) {

        $index++;
        if($key != "" ) {
            $trending_eng[$key] = $value;
        }
    }

    if($index > 30) 
        break;
}


$index = 0;
foreach($topics_ara as $key=>$value) {

    $query = "SELECT count(*) cnt FROM ( 
            SELECT count(*), tweeted_by FROM tweets
            WHERE hashtags like '%$key%'
                      AND tweeted_at >= DATE_SUB( NOW( ) , INTERVAL 2 DAY ) 
            GROUP BY tweeted_by 
             ) AS T";
        $res = mysql_query($query);
        $row = mysql_fetch_assoc($res);

        if($row['cnt'] > 1) {

        $index++;
        if($key != "" ) {
            $trending_ara[$key] = $value;
        }
    }

    if($index > 30) 
        break;
}


 //var_dump($trending_eng) ;
 //var_dump($trending_ara) ;

?>

Upvotes: 0

Views: 316

Answers (1)

Eugen Rieck
Eugen Rieck

Reputation: 65304

Sorry, but your data model is defective.

You do not normalize the tweets, but do a fulltext search for the hashtags (via hashtags like '%$key%'), which means that the complete text of all tweets in the time interval has to run through a CPU-intensive process - not only one time, but through two foreach() loops of 30 iterations each.

So you do 60 full text scans - good luck with that.

The correct way would be to norlize the tweets on receiving them, splitting the hashtags and creating a table similar to hashtag | user | count

Upvotes: 3

Related Questions