The non-compiled XslTransform supports XmlReader pull model, so you could do 
this:

XslTransform transform = new XslTransform();
XmlDocument doc = new XmlDocument();
doc.Load( transform.Transform(...));

But for high throughput transformations you really want to use 
XslCompiledTransform in which case I would suggest using Oleg's suggestion 
about using XmlWriter to build an XmlDocument as follows:
XslCompiledTransform tran = new XslCompiledTransform();
tran.Load("xsl1.xslt");
XmlDocument doc = new XmlDocument();
doc.PreserveWhitespace = true;
XmlWriter writer = doc.CreateNavigator().AppendChild();
tran.Transform("xml1.xml", null, writer );
writer.Close();

XmlDocument will be faster than reparsing the XML from a StringBuilder .



"Slava Sedov" <SlavaSedov@d...> wrote in message 
news:26AEE002-80C1-4240-84AE-78C98BD3EF4C@m......
>
>
> "Oleg Tkachenko [MVP]" wrote:
>
>> Slava Sedov wrote:
>> > need to effectively perform 2 sequential xslt transforms on some data
>> > ((xm1+xsl1)+xsl2) and can`t find a way to do it without intermediate
>> > transform intermediate xml to text - i use XslCompiledTransform with
>> > XPathDocument, is anyone know good way to pass result data after first
>> > xslt-transformation as source data for second xslt-transformation?
>>
>> As XslCompiledTransform doesn't support XmlReader output, there is no
>> such easy way. The only way that doesn't incur reparsing is transforming
>> to XmlNodeWriter, which writes to XmlDocument. But then your next
>> transformation will deal with XmlDocument, which is slower than
>> XPathDocument.
>> We (XML MVPs) are working on this problem.
>
> it is very actual problem, temporary i will use StringBuilder and cached
> XPathDocument but must mark this part of code as "weak" and set "review 
> date"
> for this problem, when MS going to solve it? .NET 2.0 or later?
>
>>
>> Another solution is to avoid double stylesheets and instead combine them
>> into one (set mode for each template to distinguish them), then store
>> first transformation in a variable, convert it to a nodeset and run
>> second one.
>
> i used it very long in SQLXML (without ASP.NET) - but it is not acceptable
> for current work (i must divide content, presentation and interaction into
> totally independent parts of code) maybe i need not only 2 sequential of
> transformations, what if i include PI stylesheet into resulting document? 
> can
> Microsoft in this case provide support for "precompilied instant chained
> transformations"? :)
>
>>
>> > p.s. also i am interesting about best practice to this scenario - i get
>> > xml1, then apply xslt1 to it and in result i get another xslt2 and then
>> > perform tranformation of xml2 using xslt2. I don`t understand why i 
>> > need
>> > store intermediate results as strings instead of storing it as 
>> > XPathDocument
>> > or at least XMLDocument.
>> You don't have to. Transform to XmlWriter instead.
>
> can you present any code example?
>
>>
>> -- 
>> Oleg Tkachenko [XML MVP, MCAD]
>> http://www.xmllab.net
>> http://blog.tkachenko.com
>>