The validity of the data present in the cache depends on the implementation logic and hugely depends on the nature of the application itself. For example, In an online store, the cache data containing the list of products available in the store can be valid for even 2 hours as there is very less chance of a product being added every 2 hours. But the cache data holding the availability (stock) of products can be valid for probably 2 minutes ( assuming , it’s a real online store where business happens) as the availability keeps changing.

Today we’d talk about one of the techniques in output caching – Record set caching. It’s one of the straight forward approach, that can be implemented. In layman terms, The record-set after a complicated query would be stored separately so that, the future requests can be served from the temporary storage instead of running the time consuming complicated query again every time.

To start with, Let us consider an example. A page delivering search results in JSON format from a product catalogue accepting a search query. Let us put down the algorithm to implement this

Step 1 : Check if a valid cache data exists in cache
Step 2 : If Yes read, the cache data and serve it to the user / client
Step 3 : If No, Read it from the DB and serve the user.
Step 4 : Update / Create the valid cache data with the one got in step 3

Let’s get into coding rightaway..

Source code

$mcdir = "cache/";
$mcache_validity=10;
$mfnam = $mcdir . "ch_" . $search_data ."_file.txt";
$mread_from_cache = false;
if (file_exists($mfnam)) {
	$mctime = (filectime($mfnam));
	$mtime_offset = time() - $mctime;
	if ($mtime_offset < $mcache_validity * 60) {
		$mread_from_cache = true;
	}
}

In the above code snippet , we just check if there is a valid cache data available. The logic behind is to have a separate file to store the result-set for individual search strings. I’ve set the validity of the cache data to 10 minutes. The next step would be based on the decision made in step-1

Source code

if (!$mread_from_cache) {
    $msql = "select product_id,product_name,product_img,product_desc,product_rate
            from product_master where product_name like '%$search_data%'
            where flag_dele=0";
    $mres = mysql_query($msql);
    $mresult = array();
    while ($mdata = mysql_fetch_object($mres)) {
        $mresult[] = $mdata;
    }
    $cacheData = readCache($mfnam);
    (md5(serialize($cacheData)) == md5(serialize($data))) ? $data = $cacheData : writeCache($data, $mfnam);
}else{
    $mresult = readCache($mfnam);
}
echo json_encode($mresult);
flush();

In the above code, we based on the value of ‘$mread_from_cache‘ , either read from the database or from the cache data. I’ve got a simple query for illustration. It may not be the case in an application. Here we have used two user defined function, one to write and other to read from the cache. The below is the code for them.

Source code

function writeCache($data, $cacheFile='default_cache.txt') {
    if (!$fp = fopen($cacheFile, 'w')) {
        trigger_error('Error opening cache file');
        exit();
    }
    if (!flock($fp, LOCK_EX)) {
        trigger_error('Unable to lock file');
        exit();
    }
    if (!fwrite($fp, serialize($data))) {
        trigger_error('Error writing to cache file');
        exit();
    }
    flock($fp, LOCK_UN);
    fclose($fp);
}
 
function readCache($cacheFile) {
    if (!file_exists($cacheFile)) {
        return '';
    }
    return unserialize(file_get_contents($cacheFile));
}

This is a very basic implementation of record set caching. This can be improved based on the application demand and other load factors. We’ve taken a more procedural approach in this code. In the next article I’d talk about handling the same in a more object oriented approach.

Do share your comments and suggestions.

The complete script can be configured by setting some variables in it. I’ll explain the configuration parameters below.

    Once you have the configuration done. Open the Windows Scheduler (from control panel) add this batch file and schedule it with the interval you want. You are done. You have your backup process running. The same script can be modified slightly to handle multiple databases if you need. This script can be downloaded from here

Do share your comments.