Question

I have a somewhat interesting query..

I may have over simplified the example, but ill do my best to describe my problem.

I am building a very simple implementation of a wiki from scratch, everything going well until I realized I need Cycle Detection to prevent endless loops of data populating the page and well Overflowing the stack heap.

Database structure is basic, well its more intricate that what is shown but for the purposes of this post the two columns is all we need.

The Content field is straight forward, it stores the Content of the Page or WikiPart links i.e [[n]] to link to another part and includes, links are shopwn as [[n]] and includes are {{n}}.

+---------------------------+
| id    |  Content          |
+---------------------------+
|  1    | see {{2}} here    | 
+---------------------------+
|  2    | {{1}} here [[4]]  | 
+---------------------------+
|  4    | {{1}}             | 
+---------------------------+



$html_for_screen = readData($this->Content);

function readData($wikipage) {

    $str = "";

    //Convert any wiki links to HTML Links
    $wikipage = Converter::convertWikink($wikipage);

    //Get ALL Include Link matches into array
    $wiki_inc = RegEx::getMatches(wikipage); 

    //Iterate through the Matches
    foreach($wiki_inc as $wiki) {
         //traverse through each match. 
         //but I assume here is where I would eventually have the trouble
         //With infinant loops
         $str .= readData($wiki);
    }

    return $str;

}

The Question:
How would I prevent Wiki parts endlessly including eachother.
i.e WikiPart 1 includes WikiPart2.. but WikiPart 2 includes WikiPart1

The parse or readData() function would just continue looping.

regards

Answer 1

Actually if you encounter a cycle, you can’t resolve any longer. Example:

1: {{2}}
2: {{1}}

This will create an endless loop:

1 -> 2 -> 1 -> 2 -> ...

As any computers resources are limited, endless loops will result in a crash.

So what can you do? You could detect that and then error out by using a stack:

function readData($wikipage)
{
    static $stack = array();
    if (in_array($wikipage, $stack))
    {
        throw new Exception(sprintf('Circular reference detected: %s -> %s', implode(' -> ', $stack), $wikipage));
    }
    $stack[] = $wikipage;

    ... (your existing code)

    array_pop($stack);
}

Additionally you can control recursion limit by using count($stack) to determine the nesting level.

Actually throwing an exception might not be the proper reaction to the cyclic reference, but it shows how the detection work. You can decide on your own how you’d like to deal with the case, e.g. returning FALSE or not resolving the field any longer etc..

Edit: Getting creative here:

If the output is HTML you could make the user resolve the problem as well. If such a cyclic reference is detected, some AJAX marker could be inserted that would request in some form of overlay within the browser that snippet which was unable to obtain on the server side. Such an overlay will then contain the cyclic reference again (being able to overlay again) so that the user would be able to see the cyclic reference interactively.