[Propagators] Improve JaegerPropagator performance

[martincostello]

Changes

Improve the performance of JaegerPropagator for .NET 8+ by using SearchValues<T>.

Benchmark Results

Copilot Summary

JaegerPropagator-perf improves all common Extract benchmarks, while Inject is mixed. Allocations are unchanged across every shared suite.

Suite	Method	main	JaegerPropagator-perf	Duration delta	Duration ratio	Allocation delta	Allocation ratio
Sampled=False, UseEncodedDelimiter=False	Extract	212.57 ns	98.98 ns	-53.4%	0.47x	312 B -> 312 B (0.0%)	1.00x
Sampled=False, UseEncodedDelimiter=False	Inject	54.12 ns	28.58 ns	-47.2%	0.53x	128 B -> 128 B (0.0%)	1.00x
Sampled=False, UseEncodedDelimiter=True	Extract	200.61 ns	92.99 ns	-53.6%	0.46x	312 B -> 312 B (0.0%)	1.00x
Sampled=False, UseEncodedDelimiter=True	Inject	27.55 ns	28.85 ns	+4.7%	1.05x	128 B -> 128 B (0.0%)	1.00x
Sampled=True, UseEncodedDelimiter=False	Extract	119.99 ns	89.37 ns	-25.5%	0.74x	312 B -> 312 B (0.0%)	1.00x
Sampled=True, UseEncodedDelimiter=False	Inject	26.59 ns	28.38 ns	+6.7%	1.07x	128 B -> 128 B (0.0%)	1.00x
Sampled=True, UseEncodedDelimiter=True	Extract	120.75 ns	89.50 ns	-25.9%	0.74x	312 B -> 312 B (0.0%)	1.00x
Sampled=True, UseEncodedDelimiter=True	Inject	27.28 ns	27.20 ns	-0.3%	1.00x	128 B -> 128 B (0.0%)	1.00x

Overall

• Extract: faster in every shared suite on JaegerPropagator-perf, with the biggest wins in non-sampled cases (~54% lower duration).
• Inject: one large win (-47.2%), two small regressions (+4.7% and +6.7%), and one effectively unchanged result.
• Allocations: no change in any shared benchmark.

Detailed Results

Expand to view

main

BenchmarkDotNet v0.15.8, Windows 11 (10.0.26200.8246/25H2/2025Update/HudsonValley2)
13th Gen Intel Core i7-13700H 2.90GHz, 1 CPU, 20 logical and 14 physical cores
.NET SDK 10.0.203
  [Host]    : .NET 10.0.7 (10.0.7, 10.0.726.21808), X64 RyuJIT x86-64-v3
  .NET 10.0 : .NET 10.0.7 (10.0.7, 10.0.726.21808), X64 RyuJIT x86-64-v3

Job=.NET 10.0  Runtime=.NET 10.0  Toolchain=net10.0  

Method	Sampled	UseEncodedDelimiter	Mean	Error	StdDev	Ratio	RatioSD	Gen0	Allocated	Alloc Ratio
Extract	False	False	212.57 ns	4.066 ns	3.804 ns	1.00	0.02	0.0248	312 B	1.00
Inject	False	False	54.12 ns	2.795 ns	7.975 ns	0.25	0.04	0.0102	128 B	0.41
Extract	False	True	200.61 ns	3.607 ns	3.542 ns	1.00	0.02	0.0248	312 B	1.00
Inject	False	True	27.55 ns	0.585 ns	1.220 ns	0.14	0.01	0.0102	128 B	0.41
Extract	True	False	119.99 ns	2.045 ns	1.708 ns	1.00	0.02	0.0248	312 B	1.00
Inject	True	False	26.59 ns	0.560 ns	1.144 ns	0.22	0.01	0.0102	128 B	0.41
Extract	True	True	120.75 ns	2.199 ns	2.057 ns	1.00	0.02	0.0248	312 B	1.00
Inject	True	True	27.28 ns	0.571 ns	0.984 ns	0.23	0.01	0.0102	128 B	0.41

This PR

BenchmarkDotNet v0.15.8, Windows 11 (10.0.26200.8246/25H2/2025Update/HudsonValley2)
13th Gen Intel Core i7-13700H 2.90GHz, 1 CPU, 20 logical and 14 physical cores
.NET SDK 10.0.203
  [Host]    : .NET 10.0.7 (10.0.7, 10.0.726.21808), X64 RyuJIT x86-64-v3
  .NET 10.0 : .NET 10.0.7 (10.0.7, 10.0.726.21808), X64 RyuJIT x86-64-v3

Job=.NET 10.0  Runtime=.NET 10.0  Toolchain=net10.0  

Method	Sampled	UseEncodedDelimiter	Mean	Error	StdDev	Ratio	RatioSD	Gen0	Allocated	Alloc Ratio
Extract	False	False	98.98 ns	2.006 ns	2.746 ns	1.00	0.04	0.0248	312 B	1.00
Inject	False	False	28.58 ns	0.590 ns	1.245 ns	0.29	0.01	0.0102	128 B	0.41
Extract	False	True	92.99 ns	1.827 ns	2.438 ns	1.00	0.04	0.0248	312 B	1.00
Inject	False	True	28.85 ns	0.609 ns	1.143 ns	0.31	0.01	0.0102	128 B	0.41
Extract	True	False	89.37 ns	1.693 ns	1.738 ns	1.00	0.03	0.0248	312 B	1.00
Inject	True	False	28.38 ns	0.564 ns	1.238 ns	0.32	0.01	0.0102	128 B	0.41
Extract	True	True	89.50 ns	1.760 ns	1.807 ns	1.00	0.03	0.0248	312 B	1.00
Inject	True	True	27.20 ns	0.569 ns	1.353 ns	0.30	0.02	0.0102	128 B	0.41

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)

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 89.08%. Comparing base (8bee1f4) to head (e7514a2).
✅ All tests successful. No failed tests found.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #7171      +/-   ##
==========================================
- Coverage   89.10%   89.08%   -0.03%     
==========================================
  Files         271      271              
  Lines       13042    13063      +21     
==========================================
+ Hits        11621    11637      +16     
- Misses       1421     1426       +5     

Flag	Coverage Δ
unittests-Project-Experimental	88.67% <100.00%> (-0.29%)	⬇️
unittests-Project-Stable	89.08% <100.00%> (-0.01%)	⬇️

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

Files with missing lines	Coverage Δ
...lemetry.Extensions.Propagators/JaegerPropagator.cs	96.03% <100.00%> (+0.79%)	⬆️

... and 2 files with indirect coverage changes

[Copilot]

Pull request overview

This PR aims to improve JaegerPropagator extraction performance on modern .NET by optimizing how the propagator splits the uber-trace-id header into components.

Changes:

• Adds a .NET-specific fast-path for parsing header components using SearchValues<char> + IndexOfAny.
• Introduces a cached delimiter hint set (: and %) for faster delimiter discovery.
• Keeps the existing parsing implementation for non-.NET TFMs via conditional compilation.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[cijothomas]

Given the component is to be deprecated, should we do perf improvements for it?

[martincostello]

This was "free" as part of some performance investigation work I asked Copilot to do. I figured they were worth keeping rather than throwing away as we still ship this anyway.

[cijothomas]

This was "free" as part of some performance investigation work I asked Copilot to do. I figured they were worth keeping rather than throwing away as we still ship this anyway.

I don't think this is "free". While copilot did a lot of heavy lifting, the review cycle isn't free for us — approver bandwidth is one of our biggest bottlenecks across OTel repos. I'd rather we save that energy for higher-impact work, like accelerating the deprecation/removal of this component.

Let's not make it harder for users to say goodbye to JaegerPropagator by making it faster! 🤣

[martincostello]

If it had been a large change, I would have likely agreed, but for ~50 lines I considered it worth the improvement considering we still ship this package and it's mostly just adopting some newer APIs from .NET itself.

[rajkumar-rangaraj]

I want to acknowledge @cijothomas's point. Approver bandwidth across the OTel repos is tight right now, and I agree we should be prioritizing the higher-impact items.

[martincostello]

Next time I get Copilot to do some speculative search for improvements, I'll make sure I tell it to exclude any areas marked as deprecated.