v0.3.1

Polyglot v0.3.1 (draft)

Highlights

• Benchmark charts land in the main README (SVG) with regenerable PNG/SVG from summary.json
• Python vs stdlib logging and .NET vs Serilog competitor harnesses + bindings chart
• Bench suite reliability: scale CSV actually produced, Node bindings built before Node benches, gofmt import fix

Benchmarks & docs

• Embed throughput / P99 latency / scale / FFI / Python·.NET charts in README.md
• Expand bench/README.md: rich vs cross vs FFI payload shapes, fairness notes, Node FFI tax (stringify → C string → Go unmarshal → rematchal)
• Chart pipeline writes SVG + PNG; empty scale data shows an explicit placeholder instead of a flat zero line
• Scale chart uses even x-spacing for 1→64 writers and y-axis labels
• make bench / bench-smoke build bindings/node before running Node benches (fixes CI MODULE_NOT_FOUND on dist/index.js)
• BENCH_SCALE_CSV=1 runs TestScaleCSV so scale.csv / scale chart are real
• New: bench/python/bench.py, bench/dotnet/ (Serilog competitor)
• Summarizer/charts ingest python.txt / dotnet.txt → bindings_ops → bindings.svg

Fixes

• internal/logger/logger.go: gofmt import order (maps alphabetized) — unblocks CI format check
• Release workflow: build Node bindings before release benches; emit scale CSV on release runners

Notes / caveats

• Node still trails Pino on sync file: expected while every call double-serializes across FFI. Buffer/Uint8Array ABI path is documented as next work, not shipped
• Binding benches include FFI tax; do not compare Go Zap numbers to Node/Python/.NET as one runtime
• Absolute ops/s vary by machine; compare relative deltas on the same runner (bench/HISTORY.md)