> Does anyone know if there's a performance difference if this ID is an
> attribute or an element?

Have you personally do a test to find out the answer? Time stamp before and
after SelectSingleNode for searching your ID as attribute and element. If
you are using VC++, you can do something like this:

DWORD dwElapsedTime = GetTickCount();
....
SelectSingleNode(XPath)
....
dwElapsedTime = GetTickCount() - dwElapsedTime;
print out the value of dwElapsedTime variable to compare the performance.

"James Hahn" <JamesHahn@h...> wrote in message
news:e1ylfOO2EHA.3336@T......
> I am processing an XML file where performance is important. Most of the
> searching is for an element within the current node, but there is one
point
> where I have to do a SelectSingleNode from the document root, searching
for
> a unique ID, and I suspect this is a slow procedure.
>
> Does anyone know if there's a performance difference if this ID is an
> attribute or an element? Is there any optimising I can do to ensure the
> search is as efficient as possible?
>
> I could pre-process the file to de-normalize it, but for a typical data
set
> this about quadruples the size, and I fear that I will not get any overall
> speed improvement.
>
>