Re: [xsl] comparing nodesets to each other

["Aron Bock" <aronbock@xxxxxxxxxxx>]

Kai,

IMO the general problem of finding the differences between any 2 XML 
documents is, shall we say, challenging.  Something that helps such an 
operation is being extremely precise about what constitutes a difference, 
and being able to formulate precedence rules in comparision operations.  An 
earlier respondent illustrated the need for this with an example that 
"added" a node in the second document.  It's very likely *you* have a good 
idea of what you're after, but in these types of problems you'll get the 
most help if you can express your "rules for comparision" in [formal] 
written form.

Consider the following documents:

doc1.xml
=======
<doc>
<chapter n="1"/>
<chapter n="2"/>
</doc>

doc2.xml
=======
<doc>
<chapter n="1"/>
<chapter n="2">
 <para n="1"/>
</chapter>
</doc>

What *exactly* would you like in your final output?  Do you want to see only 
the node <para n="1"/>?  Do you want to see <para n="1"/> and all its parent 
nodes?  You see where this is going?  It helps to be precise.

Also, while writing a "general" differencing algorithm would be worthwhile, 
it's probably not simple.  To start you'll have better luck if you constrain 
your problem, as it relates to your domain.  One way to do this is by 
identifying a least granular level for your purposes--perhaps a node or 
"level" below which identifying differences is superfluous.  In the example 
above, you could say:

--chapter nodes are compared by their "n" attribute
--if there are any differences betweein 2 <chapter> nodes or any of their 
descendents, the entire <chapter> node is considered "changed", and that of 
doc2.xml is output

I've done this type of "constrained" comparision with success.

Here's another approach to consider: preprocess each xml document to a 
"standard" format, then use a textual diff tool.  The idea here is that you 
apply an XSL transform to doc1.xml so that <chapter> nodes are sequential, 
their descendents are ordered is a specific way, etc.  Do the same with 
doc2.xml.  Then use a diff tool ( eg: beyondcompare, from 
http://www.scootersoftware.com/ ) to check differences.  Note, this method 
is susceptible to line-breaks, so it's not trivial to implement.

Regards

--A



> From: "Kai Hackemesser" <kaha@xxxxxx>
> Reply-To: xsl-list@xxxxxxxxxxxxxxxxxxxxxx
> To: xsl-list@xxxxxxxxxxxxxxxxxxxxxx
> Subject: Re: [xsl] comparing nodesets to each other
> Date: Mon, 11 Apr 2005 18:18:47 +0200 (MEST)
>
> Hello, David,
>
> Thanks for the response. The errors you mentioned already have happened,
> that's why I'm currently clueless how to solve it.
>
> I try to show the structure of the recipe (eased):
>
> <object>
>   <relation>
>     <Attribute Type="string" Name="FindNumber">
>       <Value><![CDATA[0005]]></Value>
>     <Attribute>
>     <Attribute Type="float" Name="...
>     <object>
>       <Attribute Type="string" Name="PartNumber">
>         <Value><![CDATA[Part1]]></Value>
>       </Attribute>
>     </object>
>   </relation>
>   <relation>
>     <Attribute Type="string" Name="FindNumber">
>       <Value><![CDATA[0010]]></Value>
>     <Attribute>
>     <Attribute Type="float" Name="...
>     <object>
>       <Attribute Type="string" Name="PartNumber">
>         <Value><![CDATA[Part2]]></Value>
>       </Attribute>
>     </object>
>   </relation>
>   <relation>
>     <Attribute Type="string" Name="FindNumber">
>       <Value><![CDATA[0015]]></Value>
>     <Attribute>
>     <Attribute Type="float" Name="...
>     <object>
>       <Attribute Type="string" Name="PartNumber">
>         <Value><![CDATA[Part3]]></Value>
>       </Attribute>
>     </object>
>   </relation>
> </object>
>
> needs to be compared against a similar structure:
> <object>
>   <relation>
>     <Attribute Type="string" Name="FindNumber">
>       <Value><![CDATA[0005]]></Value>
>     <Attribute>
>     <Attribute Type="float" Name="...
>     <object>
>       <Attribute Type="string" Name="PartNumber">
>         <Value><![CDATA[Part1]]></Value>
>       </Attribute>
>     </object>
>   </relation>
>   <relation>
>     <Attribute Type="string" Name="FindNumber">
>       <Value><![CDATA[0015]]></Value>
>     <Attribute>
>     <Attribute Type="float" Name="...
>     <object>
>       <Attribute Type="string" Name="PartNumber">
>         <Value><![CDATA[Part3b]]></Value>
>       </Attribute>
>     </object>
>   </relation>
> </object>
>
> (Attribute nodes are more than one per object or relation node)
>
> So I need to extract all differences like attribute change, missing nodes,
> altered nodes, added nodes. To identify a node I use the findnumber
> Attribute node of each relation node. A missing node is one, where the
> corresponding Findnumber Attribute value is missing in nodelist 'b'. An
> added node is one where the corresponding Findnumber Attribute value is
> missing in nodelist 'a'. An altered node means the Findnumber Attribute
> value is there in bothe nodelists, but the Attribute nodes or the
> object/Attribute nodes are different. I think a simple text compare would 
> be
> enough for the test of alternation.
>
> Regards,
> Kai

_________________________________________________________________
Dont just search. Find. Check out the new MSN Search! 
http://search.msn.click-url.com/go/onm00200636ave/direct/01/