Profile Party

Your Code is Pretty. Your CPU Doesn't Care.

A side-by-side comparison of linter-perfect code vs CPU-perfect code, profiled with pprof, benchmarked with real numbers. One million transactions, two architectures, one uncomfortable truth: 63% of the runtime was garbage collection. The code review said "looks good to me."

Quick Start

go build -o profile-party .
./profile-party

Full Profiling Suite

chmod +x profile.sh
./profile.sh

This generates CPU + heap profiles for both versions, runs benchmarks, and tells you how to explore with pprof.

The Numbers

	Pretty	Fast	Factor
Time	107 ms	28 ms	3.8x
Memory	101.5 MB	3.6 MB	28x
Allocations	282	7	40x
GC CPU	63%	0%	---

Read the Article

The full tutorial lives at the GitHub Pages site:

https://copyleftdev.github.io/profile-party

Project Structure

pretty.go        # The linter's favorite — interfaces, pipelines, string types
fast.go          # The CPU's favorite — struct-of-arrays, integers, one pass
main.go          # Generates 1M records, runs both, prints the verdict
bench_test.go    # go test -bench with memory reporting
profile.sh       # One script to profile everything
article.md       # Article source (Markdown)
docs/            # GitHub Pages site

Setup GitHub Pages

1. Push this repo to GitHub
2. Go to Settings > Pages
3. Source: Deploy from a branch
4. Branch: main, folder: /docs
5. Save — your site goes live in ~60 seconds

CI/CD

Every push to main:

1. Build & Benchmark — compiles the Go code, runs the comparison, executes benchmarks
2. Deploy — publishes docs/ to GitHub Pages via actions/deploy-pages

PRs get the build + benchmark gate without deploying.

License

MIT