Improve ordinal and clock notation hot paths

[clairernovotny]

Summary

• avoid eager allocation/formatting in ordinal, clock-notation, transformer, metric, phrase-join, and single-part TimeSpan hot paths
• preserve existing public API behavior with focused tests around optimized branches, culture-sensitive formatting, and boundary cases
• address Copilot review feedback for culture-specific ordinal formatting, adjacent-hyphen normalization, and context-sensitive title casing
• cache default ordinal number formatting by culture name to remove the large int.Ordinalize(culture) positive-number regression found in the benchmark artifacts
• repair benchmark comparison reporting so ResultsComparer restores with an isolated NuGet config and no longer uploads a hidden build-failure fragment as a successful report
• expand BenchmarkDotNet coverage for optimized paths and control paths:
  • metric default int small/kilo/mega plus boundary, giga, formatted fallback, and double milli cases
  • formatter first-touch/date/data-unit cases plus single-part, multi-part, zero, and words TimeSpan humanization

Benchmark Evidence

Local net10 short BenchmarkDotNet runs on Apple M2 Max after b959f088:

• MetricNumeralBenchmarks: small 59.05 ns / 56 B, kilo 117.68 ns / 112 B, mega 113.29 ns / 112 B, boundary 149.74 ns / 128 B, giga 158.14 ns / 152 B, formatted fallback 205.21 ns / 128 B, milli 125.57 ns / 136 B
• FormatterBenchmarks: Russian single-part TimeSpan 95.84 ns / 64 B, multi-part control 423.60 ns / 600 B, zero 161.50 ns / 80 B, words 131.71 ns / 224 B, Romanian Date 140.24 ns / 136 B, Arabic DataUnit 39.56 ns / 80 B

Short local review-fix spot checks after 58084558:

• OrdinalBenchmarks: English ordinalize 29.93 ns / 64 B; custom-culture paths unchanged in shape for Dutch/Turkish/Greek/Finnish control cases
• TransformersBenchmarks: title case 10 chars 201.24 ns / 696 B, 100 chars 2.424 us / 7264 B, 1000 chars 24.483 us / 73200 B

Short local ordinal regression check after 3221a0ad:

• OrdinalBenchmarks: English 27.63 ns / 64 B, Dutch 44.51 ns / 176 B, Turkish 45.18 ns / 176 B, Greek 46.37 ns / 176 B, Finnish 44.53 ns / 176 B
• The prior GitHub artifact for 58084558 showed Dutch/Turkish/Greek around 110-111 us; the cache change removes that regression locally.

GitHub benchmark workflow:

• b959f088 completed successfully: https://github.com/Humanizr/Humanizer/actions/runs/24616985676
• 58084558 completed successfully, but its comparison artifact exposed the ResultsComparer restore/reporting bug: https://github.com/Humanizr/Humanizer/actions/runs/24617740031
• latest head 3221a0ad is running from the push and PR events:
  • https://github.com/Humanizr/Humanizer/actions/runs/24618892125
  • https://github.com/Humanizr/Humanizer/actions/runs/24618892631

Tests

• dotnet build src/Benchmarks/Benchmarks.csproj -c Release -f net10.0
• dotnet run --project src/Benchmarks/Benchmarks.csproj -c Release -f net10.0 -- --filter '*MetricNumeralBenchmarks*' --job short --warmupCount 1 --iterationCount 3
• dotnet run --project src/Benchmarks/Benchmarks.csproj -c Release -f net10.0 -- --filter '*FormatterBenchmarks*' --job short --warmupCount 1 --iterationCount 3
• dotnet run --project src/Benchmarks/Benchmarks.csproj -c Release -f net10.0 -- --filter '*OrdinalBenchmarks*' --job short --warmupCount 1 --iterationCount 3
• dotnet run --project src/Benchmarks/Benchmarks.csproj -c Release -f net10.0 -- --filter '*TransformersBenchmarks*' --job short --warmupCount 1 --iterationCount 3
• dotnet test --project tests/Humanizer.Tests/Humanizer.Tests.csproj --framework net10.0 --no-restore -- --filter-class '*OrdinalizeTests' --filter-class '*TransformersTests' --filter-class '*CoverageGapTests'
• dotnet test --project tests/Humanizer.Tests/Humanizer.Tests.csproj --framework net10.0 --no-restore -- --filter-class '*OrdinalizeTests' --filter-class '*DutchGenderedOrdinalTests'
• dotnet test --project tests/Humanizer.Tests/Humanizer.Tests.csproj --framework net11.0 --no-restore -- --filter-class '*OrdinalizeTests' --filter-class '*DutchGenderedOrdinalTests'
• dotnet run --project src/Benchmarks/Benchmarks.csproj -c Release -f net10.0 -- --filter '*OrdinalBenchmarks*' --job short --warmupCount 1 --iterationCount 1
• dotnet format Humanizer.slnx --verify-no-changes --verbosity diagnostic --no-restore
• ruby -e 'require "yaml"; YAML.load_file(".github/workflows/benchmarks-baseline-vs-current.yml"); puts "yaml ok"'
• dotnet pack src/Humanizer/Humanizer.csproj -c Release -o artifacts/packages/ordinal-benchmark-fix
• dotnet test --project tests/Humanizer.Tests/Humanizer.Tests.csproj --framework net10.0 --no-restore
• dotnet test --project tests/Humanizer.Tests/Humanizer.Tests.csproj --framework net11.0 --no-restore
• git diff --check

Note: dotnet test --project tests/Humanizer.Tests/Humanizer.Tests.csproj --framework net8.0 --no-restore could not run locally because this machine does not have the .NET 8 runtime installed; CI should cover it.

[Copilot]

Pull request overview

Optimizes Humanizer’s ordinalization and phrase-based clock-notation conversion hot paths to reduce allocations and avoid unnecessary culture/word-resolution work.

Changes:

• Refactors int.Ordinalize(..., CultureInfo) to avoid loading CultureInfo.NumberFormat on the positive-number path and to better preserve custom negative-sign behavior for negative values.
• Updates phrase clock-notation template expansion to lazily resolve placeholder values only when used by the selected template.
• Adds a regression test covering custom NegativeSign handling for negative ordinalization.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.

File	Description
tests/Humanizer.Tests/OrdinalizeTests.cs	Adds a test ensuring custom culture negative-sign is preserved for negative ordinals.
src/Humanizer/OrdinalizeExtensions.cs	Introduces a new ordinal number-string formatting path intended to reduce culture formatting overhead.
src/Humanizer/Localisation/TimeToClockNotation/PhraseClockNotationConverter.cs	Lazily resolves template placeholders to avoid computing unused hour/minute/article/day-period strings.

---

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

[Copilot]

Pull request overview

Copilot reviewed 6 out of 6 changed files in this pull request and generated 1 comment.

---

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

[Copilot]

Pull request overview

Copilot reviewed 14 out of 14 changed files in this pull request and generated 2 comments.

---

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

[Copilot]

Pull request overview

Copilot reviewed 17 out of 17 changed files in this pull request and generated 2 comments.

---

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

[Copilot]

Pull request overview

Copilot reviewed 19 out of 19 changed files in this pull request and generated 3 comments.

---

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

[Copilot]

Pull request overview

Copilot reviewed 14 out of 14 changed files in this pull request and generated 3 comments.

---

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

[Copilot]

Pull request overview

Copilot reviewed 41 out of 41 changed files in this pull request and generated 2 comments.

---

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

[Copilot]

Pull request overview

Copilot reviewed 51 out of 51 changed files in this pull request and generated 1 comment.

---

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

[Copilot]

Pull request overview

Copilot reviewed 59 out of 59 changed files in this pull request and generated no new comments.

---

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