A typical operation in DHTML is to iterate through a collection of objects. Let’s say you’re writing an HTML application that indexes content. Your task is to collect all the H1 elements on a given page and to use them as index entries. Here’s an example of how you might go about this:

 function Iterate(aEntries){   for (var i=0; i < document.all.length; i++)   {      if (document.all(i).tagName == "H1")      {         aEntries[aEntries.length] = document.all(i).innerText;      }   }}

The code contains three calls through the all collection of the document. During each iteration through the loop, the scripting engine will get the length of the collection, get the tagName of the current object in the collection, and get the innerText of the current object in the collection. The overhead amounts to numerous unnecessary calls to the DHTML Object Model for information that you already know about. Here's an alternate version:

 function Iterate2(aEntries){   var oH1Coll = document.all.tags("H1");   var iLength = oH1Coll.length;   for (var i=0; i < iLength; i++)   {      aEntries[aEntries.length] = oH1Coll(i).innerText;   }}

This version removes the overhead of creating local variables to cache the all collection as well as the length of the collection. Using the tags method of the all collection puts the burden on the DHTML Object Model to do the filtering and allows you to eliminate the if condition on every iteration of the scripted loop. Also, it frees you from the ramifications of caching DHTML collections. DHTML collections are truly dynamic, and any code you author that creates new elements and inserts them into the document may cause the collection to grow or shrink.