feat(benchmark): compare supplied Expressif versions

[Seddryck]

Summary

• add a Windows-only .NET 10 BenchmarkDotNet solution isolated under benchmark/
• discover manually supplied v1/v2 binary folders and load each version in its own assembly load context
• compare parse-and-bind plus simple, text, and coercion evaluation workloads with v1 as the ratio baseline
• validate supplied binaries with isolated parsing, binding, evaluation, and coercion adapter tests

Validation

• dotnet build benchmark/Expressif.Benchmark.sln -c Release
• dotnet build Expressif.sln -c Release
• dotnet test Expressif.sln -c Release --no-build
• dotnet test benchmark/Expressif.Benchmark.sln -c Release — 3 passed with staged v1.51.0 and next-major binaries
• all eight benchmark combinations completed in a short smoke run using replace-chars(" ", "-") | first-chars(10)

Close #600

[coderabbitai]

Important

• 🔍 Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: 7fd8cd08-8c87-4643-aed0-3012b96e4b13

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[Seddryck]

Added an isolated benchmark test project in commit 5fb4f52. Validation against the manually staged binaries now confirms: (1) the prescribed complex expression parses successfully on NuGet v1.51.0 and fails specifically during construction/binding on the unavailable
eplace\ function; and (2) the implicit-coercion pipeline returns \€1493.82\ as a string on both v1.51.0 and latest next-major. \dotnet test benchmark/Expressif.Benchmark.sln -c Release: 3 passed.

[Seddryck]

Added parse-only benchmarks for both workload sources. With 3 warmups and 5 measurement iterations: complex text parsing is 32.21 µs / 191.3 KB on v1.51.0 versus 62.46 µs / 8.05 KB on next-major (1.94x time, 0.04x allocation); coercion-source parsing is 30.30 µs / 177.7 KB versus 58.74 µs / 7.6 KB (1.94x time, 0.04x allocation). Benchmark adapter tests now validate that both sources parse on both API generations (5 tests passed).

[Seddryck]

Added native parse+bind benchmarks for both sources (commit 0a84190). Each measurement constructs the expression from source through that version's public API, so it includes parsing and binding together. Results (5 iterations, 3 warmups): complex pipeline — v1 53.14 ms / 2.37 MB, v2 60.78 ms / 2.65 MB (1.14x time, 1.12x allocation); implicit-coercion pipeline — v1 53.30 ms / 2.36 MB, v2 60.99 ms / 2.65 MB (1.14x time, 1.13x allocation). Adapter tests now verify both expressions parse and parse+bind successfully on v1 and v2; 5 tests pass.

[sonarqubecloud]

Quality Gate failed

Failed conditions
C Reliability Rating on New Code (required ≥ A)

See analysis details on SonarQube Cloud

Catch issues before they fail your Quality Gate with our IDE extension SonarQube for IDE