Add support for serializing KeyValue lists

[juliuskoval]

According to the spec, attribute values should be AnyValue. One of the possible values of AnyValue is KeyValueList, which we currently aren't able to serialize properly.

Changes

This PR adds support for serializing KeyValue lists to the OTLP exporter.

Merge requirement checklist

• CONTRIBUTING guidelines followed (license requirements, nullable enabled, static analysis, etc.)
• Unit tests added/updated
• Appropriate CHANGELOG.md files updated for non-trivial changes
• Changes in public API reviewed (if applicable)

[juliuskoval]

Question: As far as I know, the spec doesn't say anything about limits to have many nested layers there can be, but shouldn't we try to guard against self-referencing IEnumerables?

[codecov]

Codecov Report

❌ Patch coverage is 30.76923% with 54 lines in your changes missing coverage. Please review.
✅ Project coverage is 89.77%. Comparing base (7ef0ebc) to head (d3141e4).
⚠️ Report is 19 commits behind head on main.
✅ All tests successful. No failed tests found.

Files with missing lines	Patch %	Lines
...xporter.Console/Implementation/ConsoleTagWriter.cs	0.00%	49 Missing ⚠️
....Exporter.Zipkin/Implementation/ZipkinTagWriter.cs	0.00%	5 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #7015      +/-   ##
==========================================
- Coverage   89.98%   89.77%   -0.21%     
==========================================
  Files         277      278       +1     
  Lines       14622    14778     +156     
==========================================
+ Hits        13157    13267     +110     
- Misses       1465     1511      +46     

Flag	Coverage Δ
unittests-Project-Experimental	89.73% <30.76%> (-0.20%)	⬇️
unittests-Project-Stable	89.74% <30.76%> (-0.12%)	⬇️
unittests-Solution	89.70% <30.76%> (-0.28%)	⬇️
unittests-UnstableCoreLibraries-Experimental	48.85% <ø> (+0.04%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
...Implementation/Serializer/ProtobufOtlpTagWriter.cs	97.22% <100.00%> (+0.79%)	⬆️
....Exporter.Zipkin/Implementation/ZipkinTagWriter.cs	50.00% <0.00%> (-10.87%)	⬇️
...xporter.Console/Implementation/ConsoleTagWriter.cs	31.25% <0.00%> (-49.40%)	⬇️

... and 28 files with indirect coverage changes

[martincostello]

shouldn't we try to guard against self-referencing IEnumerables

I think so, otherwise it would cause a stack overflow? Could just add a recursion limit to track such occurences.

[github-actions]

This PR was marked stale due to lack of activity and will be closed in 7 days. Commenting or pushing will instruct the bot to automatically remove the label. This bot runs once per day.

[juliuskoval]

I added a limit to recursion depth and a unit test for the behavior. For now I arbitrarily set the limit to 3, because I'm guessing it would be subject to discussion anyway.

[github-actions]

This PR was marked stale due to lack of activity and will be closed in 7 days. Commenting or pushing will instruct the bot to automatically remove the label. This bot runs once per day.

[martincostello]

Could you also extend the fuzz tests to cover this functionality please?

[martincostello]

Could this be simplified using Enum.GetValues()?

[juliuskoval]

Enum.GetValues() doesn't exist on .NET Framework, so we'd have to put it behind a preprocessor directive which wouldn't be very elegant.

[martincostello]

It does, you just have to use the GetValues(Type) version not GetValues<T>(). But we can leave it if it's still too ugly.

[SalvoDeveloper]

We are having trouble sending complex nested objects to an OpenTelemetryCollector. I have tested the fork of @juliuskoval by connecting NLog.Target.OpenTelemetryProtocol with the OpenTelemetry-dotnet. I only need to traslate the object into IEnumerable<KeyValuePair<string, object?>>, and the everything works. I would kindly ask you to merge this and release new version. This would be really important for us.

[martincostello]

@SalvoDeveloper I'm waiting for at least one other approving review from one of the other maintainers before merging this. Once merged it will be part of the next release (which has no specific date at this time).

After it's merged you can use consume prerelease builds from MyGet if you can't wait until then.

[rajkumar-rangaraj]

Before merging, I'd like us to align on:

1. Is implicit type-sniffing the path we want, or should we wait for typed AnyValue API so we don't ship two redundant contracts?
2. How do we address Blanch's reference-aliasing concern from [feature request] support extensible tag serialization #6052 - should this require SDK-level deep-copy on span end?

Strong support for the underlying use case, but want to avoid locking in behavior we can't revisit.

[juliuskoval]

@rajkumar-rangaraj @martincostello I added the benchmarks that we talked about 2 weeks ago. It measures serializing a log with a varying number of attributes and optionally a KvList attribute with no nesting and also with a varying number of attributes.

BenchmarkDotNet v0.15.8, Windows 11 (10.0.26200.8457/25H2/2025Update/HudsonValley2)
AMD Ryzen 7 9800X3D 4.70GHz, 1 CPU, 16 logical and 8 physical cores
.NET SDK 10.0.301
[Host] : .NET 10.0.9 (10.0.9, 10.0.926.27113), X64 RyuJIT x86-64-v4
DefaultJob : .NET 10.0.9 (10.0.9, 10.0.926.27113), X64 RyuJIT x86-64-v4

Method	AttributeCount	KvListAttributeCount	Mean	Error	StdDev	Median	Allocated
WriteLogRecord	4	0	113.8 ns	1.87 ns	1.57 ns	113.9 ns	32 B
WriteLogRecord	4	2	168.9 ns	3.15 ns	8.31 ns	166.4 ns	32 B
WriteLogRecord	4	4	224.2 ns	4.55 ns	12.84 ns	219.1 ns	32 B
WriteLogRecord	4	8	254.4 ns	5.11 ns	10.99 ns	252.6 ns	32 B
WriteLogRecord	8	0	193.0 ns	3.88 ns	9.58 ns	188.7 ns	32 B
WriteLogRecord	8	2	239.4 ns	2.89 ns	2.26 ns	239.3 ns	32 B
WriteLogRecord	8	4	297.5 ns	5.87 ns	10.58 ns	294.2 ns	32 B
WriteLogRecord	8	8	340.9 ns	8.21 ns	23.03 ns	332.2 ns	32 B
WriteLogRecord	16	0	365.1 ns	7.95 ns	22.16 ns	357.4 ns	32 B
WriteLogRecord	16	2	427.9 ns	11.96 ns	34.70 ns	417.5 ns	32 B
WriteLogRecord	16	4	476.7 ns	15.13 ns	44.62 ns	480.0 ns	32 B
WriteLogRecord	16	8	471.3 ns	9.11 ns	12.16 ns	466.0 ns	32 B

[martincostello]

Suggested change

	=> maxLength.HasValue && value.Length > maxLength
	? value.Slice(0, maxLength.Value)
	: value;
	=> maxLength is { } maxLengthValue && value.Length > maxLengthValue
	? value.Slice(0, maxLengthValue)
	: value;

[martincostello]

These changes seem reasonable to me. Raj is on PTO until August so let's wait on another +1 from another maintainer.

Can you also update the CHANGELOGs as-appropriate please?