Search, XPATH and XPATH namespaces

[DWRoelands]

I have an index populated with data from a Log4Net trace log.  Each Splunk event in the index is a block of XML with an XML namespace:

<E2ETraceEvent xmlns="http://schemas.microsoft.com/2004/06/E2ETraceEvent">
	<System xmlns="http://schemas.microsoft.com/2004/06/windows/eventlog/system">
		<EventID>0</EventID>
		<Type>3</Type>
		<SubType Name="Information">0</SubType>
		<Level>8</Level>
		<TimeCreated SystemTime="04/22/2023 12:30:45.0456293Z"/>
		<Source Name="Bar"/>
		<Correlation ActivityID="{459d276d-8255-47be-be1d-9acd903fd3f0}"/>
		<Execution ProcessName="NA" ProcessID="1124" ThreadID="9"/>
		<Channel/>
		<Computer>NA</Computer>
	</System>
	<ApplicationData>
		<TraceData>
			<DataItem>
				<TraceRecord Severity="Information">
					<TraceIdentifier/>
					<Description><![CDATA[Start Operation: foo]]></Description>
					<Activity><![CDATA[Start Operation: foo]]></Activity>
					<Duration>0</Duration>
				</TraceRecord>
			</DataItem>
		</TraceData>
	</ApplicationData>
</E2ETraceEvent>

I am trying to search these events, and I want the contents of the "Description" XML element.  Does Splunk's implementation of XPATH support namespaces?   This search is returning no records:

index="lmstracelogs" xpath "//E2ETraceEvent/ApplicationData/TraceData/DataItem/TraceRecord/Description"

Almost every example I've found for working with XML suggests regular expressions, which seems inelegant.

[PickleRick]

You can't match any events this way since you're trying to find "xpath" as a literal term included in your event.

[danspav]

Hi @DWRoelands,

This may not be the most elegant, but to avoid regex you can use:

| xpath outfield=description "/*[name()='E2ETraceEvent' and  namespace-uri()='http://schemas.microsoft.com/2004/06/E2ETraceEvent' ]/*[name()='ApplicationData']/*[name()='TraceData']/*[name()='DataItem']/*[name()='TraceRecord']/*[name()='Description']"

That creates a description field with the correct text:

It does obey the namespace, but it's not the easiest to read.

Falling back to regex, if you run a sedcmd you can strip the namespaces and use your original xpath:

| rex mode=sed "s/xmlns=\"[^\"]+\"//g"
| xpath outfield=description "//E2ETraceEvent/ApplicationData/TraceData/DataItem/TraceRecord/Description"

Cheers,
Daniel

[danspav]

Alternatively, just run:

| spath path="E2ETraceEvent.ApplicationData.TraceData.DataItem.TraceRecord.Description" output=description