使用Linq和XElement检测XML中的结构差异

Question

I'm trying to audit some XML that is used in a bespoke piece of software. Im able to detect changes in identical structures using 'XNode.DeepEquals' and then adding an extra attribute to the elements that have changed so I can highlight them.

My problem is that, when the structure does change this methodology fails. ( I'm enumerating over both XElements at the same time performing a DeepEquals, if they are not equal - recursively calling the same method to filter out where the exact changes occurr )

Obviously this now falls apart when I'm enumerating and the nodes being compared are not the same. See Below Sample:

Before

<?xml version="1.0" encoding="utf-16"?>
<Prices xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns:xsd="http://www.w3.org/2001/XMLSchema">
<Price default="true">
    <Expression operator="Addition">
        <LeftOperand>
            <AttributeValue field="ccx_bandwidth" />
        </LeftOperand>
        <RightOperand>
            <Constant value="10" type="Integer" />
        </RightOperand>
    </Expression>
</Price>
<Price default="false">
    <Expression operator="Addition">
        <LeftOperand>
            <AttributeValue field="ccx_bandwidth" />
        </LeftOperand>
        <RightOperand>
            <Constant value="99" type="Integer" />
        </RightOperand>
    </Expression>
</Price>
<RollupChildren />

After

<?xml version="1.0" encoding="utf-16"?>
<Prices xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns:xsd="http://www.w3.org/2001/XMLSchema">
<Price default="true">
    <Expression operator="Addition">
        <LeftOperand>
            <AttributeValue field="ccx_bandwidth" />
        </LeftOperand>
        <RightOperand>
            <Constant value="10" type="Integer" />
        </RightOperand>
    </Expression>
</Price>
<RollupChildren />

So you can see that the latter Price Node has been removed and I need to show this change.

At the moment I have access to both pieces of xml and modify them on load of the audit application with an 'auditchanged' attribute which in my silverlight app i bind the background too with a converter.

I'd been playing around with Linq to Xml and looking at joining the two XElements in a query but wasn't sure how to proceed.

Ideally what I would like to do is merge the two XElements together but add a seperate attribute depending on if it's added or removed which i can then bind to with a converter to say highlight in red or green appropriately.

Does anyone have any bright ideas on this one? ( I'd been looking at XmlDiff however I can't use that in Silverlight, I don't think? )

Answer 1

The important part here is the descendants query. It turns every element in the first doc in a list of its ancestors, where every item contains the name of the element and it's index among its siblings of the same name. I think this can be somehow used for joining, though I have no idea how to do full outer join with linq. So instead i just use these lists to find elements in the second document, and then depending on the result, probably mark it as either deleted or changed.

var doc = XDocument.Load(in_A);
var doc2 = XDocument.Load(in_B);
var descendants = doc.Descendants().Select(d => 
    d.AncestorsAndSelf().Reverse().Select(el => 
        new {idx = el.ElementsBeforeSelf(el.Name).Count(), el, name = el.Name}).ToList());

foreach (var list in descendants) {
    XContainer el2 = doc2;
    var el = list.Last().el;
    foreach (var item in list) {
        if (el2 == null) break;
        el2 = el2.Elements(item.name).Skip(item.idx).FirstOrDefault();
    }
    string changed = "";
    if (el2 == null) changed += " deleted";
    else {
        var el2e = el2 as XElement;
        if (el2e.Attributes().Select(a => new { a.Name, a.Value })
            .Except(el.Attributes().Select(a => new { a.Name, a.Value })).Count() > 0) {
                changed += " attributes";
        }
        if (!el2e.HasElements && el2e.Value != el.Value) {
            changed += " value";
        }
        el2e.SetAttributeValue("found", "found");
    }
    if (changed != "") el.SetAttributeValue("changed", changed.Trim());
}
doc.Save(out_A);
doc2.Save(out_B);

Answer 2

I have a generic differ class in the codeblocks library http://codeblocks.codeplex.com

Loading your XML documents and treating each document as an IEnumerable (flattened XML tree) should allow you to use the differ as shown here: http://codeblocks.codeplex.com/wikipage?title=Differ%20Sample&referringTitle=Home

Here's the source code for differ.cs: http://codeblocks.codeplex.com/SourceControl/changeset/view/96119#1887406

Diff prototype is:

static IEnumerable<DiffEntry> Diff(IEnumerable<T> oldData, IEnumerable<T> newData, Comparison<T> identity, Comparison<T> different)