I ran both of the algorithms for 20 minutes in perfmon and found the new 
algorithm appeared to be much better in limiting handle counts (varying 
between 181-188) than the old algorithm (varying between 162-308).

However, the new algorithm had a higher byte count then the old algorithm.  
Would there be a reason for this other than the memory used by the load 
method call is NOT getting released?  Any comments on my algorithms above are 
welcomed.

Thanks.

"malhenry" wrote:

> I am using MSXML 4.0 with VC++ 6.0 and am trying to optimize my code which 
> processes xml files.  My real question is does the load method below free the 
> memory used by the prior xml doc that was loaded?
> 
> My original pseudo code was:
> 
> Loop
>  //start constructor
>  ::CoInitialize(NULL);
>  hr = ::CoCreateInstance(...);
>  //end constructor
>  // process the xml
>  ...
>  	CHECKHR(pDoc->put_async(VARIANT_FALSE));
> 	CHECKHR(pDoc->put_validateOnParse(VARIANT_FALSE));
> 
> 	// Load xml document from the given URL or file path
> 
> 	VariantInit(&vURL);
> 	vURL.vt		= VT_BSTR;
> 	V_BSTR(&vURL)	= pBURL;
> 	CHECKHR(pDoc->load(vURL, &vb));//****KEY call
> 
> 	CHECKHR(CheckLoad(pDoc));
> 
> 	CleanUp:
> 	SAFERELEASE(pXMLError);
> ...
> //start destructor
>  if (m_pDoc) 
>  {
>     unsigned long ul;
>     ul = (m_pDoc)->Release();
>     m_pDoc = NULL;
>  }
>  // free the string allocated by StringToBSTR that contains xml filename
>  if (m_pBURL != NULL)
>    ::SysFreeString(m_pBURL);
>  ::CoUninitialize(); //returns a void
> //end destructor
> 
> //ENDLOOP
> 
> What I did to try to improve the code above was to move the constructor and 
> destructor OUT of the loop so that they should only get called once.  I added 
> to the bottom of the loop a  if (m_pBURL != NULL)
>    ::SysFreeString(m_pBURL); so that the memory associated with the xml 
> filename would be freed.
> 
> I thought the load call would free the memory used by the prior xml doc 
> processed.  However, I am seeing a HIGHER usage of memory outstanding with 
> the second version of my code using a memory analysis tool.
> 
> Am I mistaken that load will also free memory?  Should I go back to the 
> original version of my code which seems inefficient to initialize and 
> de-initialize com for every xml file I must process?  Any advice?
> 
> Thanks!!