PHP - 为什么按引用传递使用相同数量的内存？

Question

I am getting 'memory exhausted' in my PHP script. I cannot figure out why. The limit is 128Mb. I am sending, from a javascript xmlrequest(), a json string that is maximum 10MB (usually under 1MB). The PHP script itself is 43K. Even copying everything a dozen times over I shouldn't run out of memory. I do a few calls to the database at getParentFolders() but that produces only a few rows. Certainly not the 62 megabytes it claims I am using. I have used Xdebug (see picture) but this tells me nothing useful, only that yup, I'm using a lot of memory.

So at this point I am trying to do 'best practices' to minimize memory usage. A simple fix, or so I thought, was to pass in values by reference. So I put a '&' before every variable in every function. To my surprise, there was no change in memory consumption. It was slightly worse by a few bytes, in fact. I have also tried using every variable as a global, but again, to my surprise, there was little difference.

So what's going on? Why is passing by reference and using globals not producing the performance benefits I expected? (see images)

Xdebug 'pass by value'

Xdebug 'pass by reference'

Note they are pretty much identical.

For those who want code, here is the getParentFolders() function, which returns just a short string but somehow uses 70Mb!

function getParentFolders(&$node) {  //returns a string of folders
    debugprint("GetParentFOlders()");  //prints if $DEBUG flag is on
    $parent = getParent($node);
    $path = "";
    while ($parent) {  //goes till we hit root folder
        $path = $parent->title . '/' . $path;  //prepend it
        $parent = getParent($parent);
    }
    return $path;
}

function getParent(&$node) {  //return node that is the parent 
    global $email;
    $parentId = $node->parentId;
    $clientId = $node->parentClient;
    $idCol = $clientId . "_id";
    $tablename = $email . "_bookmarks";
    $query = "SELECT * FROM `$tablename`
                        WHERE $idCol = '$parentId'";  //only return one row since id is unique
    $result = sendquery($query);
    return (object) $result[0];
}

Edit:

Just to clarify, I am looking for a technical explanation of PHP memory usage and best practices - specifically why I am not seeing memory differences - not a workaround for the issue.

Answer 1

Make sure the query result returns one row by using LIMIT 1 ; If no result found then make the getParent() return false or null so that it can be set to $parent to exit from while loop. And I don't think so that you need to pass the $node argument by reference in your case.

Answer 2

The value of $node is a reference, ie a pointer to an object. In PHP 5+, "objects" are not values directly. You always work with objects through pointers to them: when you do new ClassName , it evaluates to a pointer to an object; when you use the -> operator, it takes a pointer to an object on the left side.

So if both of these functions were pass-by-value, the size of the thing passed is just the size of a pointer. When you passed by reference, that basically wraps the pointer in another level of pointer, which is the same size (though the total memory is greater since there are two pointers). The only point of passing by reference is that it allows you to assign to $node inside the function to point to a different object or hold a different type of value, and have it be reflected in the passed variable in the calling function, but you do not assign to $node anywhere here (modifying a field of the object pointed to $node is not the same as assigning to $node ), so pass-by-reference is pointless here.

(Plus, even if you passed a value type, like arrays in PHP, they have copy-on-write semantics, and is not copied until you write to it. And even if you wrote to it and caused it to make a copy, that copy would only last for the lifetime of this local variable, which ends at the end of the function, so it would not create persistent memory usage.)

Answer 3

The answer is that objects cannot be passed by value. They are always passed by reference. The 'pass by reference' operator, '&', will do nothing other than wrap the object reference (ie a pointer) in a wrapper, then pass that new pointer.

As for the memory usage, there is no simple answer other than database query results seem to come with a lot of overhead, and you want to make your 'footprint' (ie any results returned from a database, or any variables) as small as possible.