DEV Community

Sulthon Zainul Habib
Sulthon Zainul Habib

Posted on

I Don't Know Zig — But I Built a Production Log Tool With GitHub Copilot

#devchallenge #githubchallenge #zig #beginners

GitHub “Finish-Up-A-Thon” Challenge Submission

This is a submission for the GitHub Finish-Up-A-Thon Challenge

This is a submission for the GitHub Finish-Up-A-Thon Challenge

⭐ GitHub Repository

👉 github.com/sulthonzh/logchef-zig — Star it if you find it useful!

What I Built

logchef — a 592KB binary that parses, filters, and pretty-prints log files in JSON, logfmt, or plain text. Think jq meets tail meets grep, purpose-built for logs.

I'm a TypeScript/Node.js developer. I'd never written a line of Zig before this project. I built the entire thing using GitHub Copilot.

Demo 

# Filter errors from any log format
logchef app.log -l error

# Search + follow in real-time
logchef app.log -c "timeout" -F

# Field filtering
logchef app.log -f status=500

# JSON output for piping
logchef app.log -l error -o json | jq '.message'

# Pipe from kubectl
kubectl logs my-pod | logchef -c panic
Enter fullscreen mode Exit fullscreen mode

🔗 GitHub: github.com/sulthonzh/logchef-zig
⭐ Give it a star! It helps other developers discover the project.

The Comeback Story

BEFORE (May 6, 2026)

I started logchef during a weekend sprint. I had an idea: replace bloated Node.js log viewers with a tiny Zig binary. I didn't know Zig, so I opened VS Code, enabled Copilot, and started learning.

The first version was rough:

  • 1,218 lines of code
  • Basic JSON parsing and level filtering worked
  • Text search worked
  • Color output worked
  • 21 tests passing

But the README listed features that didn't exist:

  • ❌ Follow mode (tail -f)
  • ❌ Logfmt format support
  • ❌ No benchmarks
  • ❌ No cross-platform builds
  • ❌ No release automation

Life happened. The project sat for 2 weeks.

AFTER (May 25, 2026)

I picked it back up for the GitHub Finish-Up-A-Thon. Using GitHub Copilot CLI (copilot -p), I created 6 issues and implemented them one by one:

  1. ✅ Follow mode — polls file every 500ms, handles rotation
  2. ✅ Logfmt support — auto-detects and parses key=value pairs
  3. ✅ JSON output — NDJSON format for pipeline integration
  4. ✅ Benchmarks — compared vs jq, grep, awk
  5. ✅ Cross-platform CI — Linux + macOS, x86 + ARM
  6. ✅ Polished README — badges, comparison table, benchmarks

Final stats:

  • 1,991 lines of Zig code (+64%)
  • 23 tests
  • 592KB binary
  • 4-platform CI
  • Zero dependencies

My Experience with GitHub Copilot

Learning Zig Through Copilot

Zig has concepts I'd never seen in TypeScript:

  • Explicit allocators — every allocation needs an allocator passed in
  • Error unions — !Type syntax for functions that can fail
  • Comptime — code that runs at compile time
  • No hidden control flow — no exceptions, no hidden allocations

I didn't learn these from docs. I learned them by writing prompts like:

copilot -p "How do I read a file line by line in Zig? Show me with explicit allocators."
Enter fullscreen mode Exit fullscreen mode

And Copilot would generate code that taught me the pattern:

var gpa = std.heap.GeneralPurposeAllocator(.{}){};
const allocator = gpa.allocator();
const file = try std.fs.cwd().openFile(path, .{});
const content = try file.readToEndAlloc(allocator, max_size);
defer allocator.free(content);
Enter fullscreen mode Exit fullscreen mode

Copilot CLI as My Zig Tutor

The most valuable tool was copilot -p — non-interactive mode that reads your codebase and generates matching code:

# Implement follow mode
copilot -p "Add follow mode that polls a file for new lines every 500ms" \
  --allow-all -C . -s

# Add logfmt parsing
copilot -p "Add logfmt key=value parsing to the existing parser" \
  --allow-all -C . -s

# Fix build errors
copilot -p "The build fails with: [error]. Fix it." \
  --allow-all -C . -s
Enter fullscreen mode Exit fullscreen mode

Every prompt taught me something new about Zig. By the end, I understood allocators, error handling, and even cross-compilation — all through Copilot-guided implementation.

What Copilot Got Right

  • Pattern matching — it read my existing code and matched the style perfectly
  • Zig idioms — it knew to use defer for cleanup, try for error propagation
  • Test writing — it generated tests that followed my existing test patterns
  • Cross-platform — it knew the right Zig target triples for all 4 platforms

What I Still Needed to Do

  • Architecture decisions — Copilot writes code, but I decided what to build
  • Code review — I read every line before committing
  • Integration — I connected the pieces Copilot built
  • Problem framing — the quality of Copilot's output depends on how you describe the problem

Results

Metric Before After
Lines of code 1,218 1,991
Tests 21 23
Binary size 592KB 592KB (unchanged!)
Formats supported JSON, plain text JSON, logfmt, plain text
Follow mode ❌ ✅
JSON output ❌ ✅
Cross-platform CI ❌ ✅ (4 platforms)
Benchmarks ❌ ✅

Key Takeaway

You don't need to know a language to build something useful in it. GitHub Copilot won't replace understanding — but it dramatically lowers the barrier to entry. I went from "what is Zig?" to "I shipped a cross-platform CLI tool in Zig" in 3 weeks, part-time.

The trick isn't blindly accepting Copilot's output. It's:

  1. Having a clear idea of what you want to build
  2. Reading and understanding every line Copilot generates
  3. Using Copilot to learn, not just to code

Thanks to GitHub and DEV for the challenge! 🙏

Built with GitHub Copilot (free tier). No paid subscriptions used.

Top comments (0)