Build a Blazing-Fast JSON-to-CSV API with Bun + Express (2025)

NOTE

Bun 1.2 now supports Node HTTP APIs and Express 5 out-of-the-box, giving you 2-3× faster throughput versus Node 20.

In this guide you’ll:

Scaffold a Bun project with Express 5 middleware.
Implement a /convert POST endpoint that streams CSV.
Add validation with zod.
Benchmark vs Node with wrk.
Deploy to Fly.io with a tiny 7 MB container.

Table of Contents
1. Why Bun + Express?
2. Prerequisites
3. Initialize Project
4. Express Server (index.ts)
5. Test Locally
6. Benchmark
7. Dockerfile (7 MB)
8. Input Validation Tips
9. Error Handling Middleware
10. Security
11. Conclusion

1. Why Bun + Express?

Bun’s JIT + Zig HTTP = <1 ms latency.
You reuse thousands of Express tutorials/middleware.
Built-in native CSV streaming via fast-csv works unchanged.

2. Prerequisites

Bun 1.2+ (brew install bun).
wrk or autocannon for local bench.

3. Initialize Project

bun create express-json-csv
cd express-json-csv
bun add express fast-csv zod

tsconfig.json is auto-generated; Bun compiles TS instantly.

4. Express Server (index.ts)

import express from 'express'
import { json } from 'express'
import { z } from 'zod'
import { format } from '@fast-csv/format'

const app = express()
app.use(json({ limit: '1mb' }))

const rowSchema = z.record(z.any())
const bodySchema = z.object({ data: z.array(rowSchema) })

app.post('/convert', (req, res) => {
  const parse = bodySchema.safeParse(req.body)
  if (!parse.success) return res.status(400).json({ error: parse.error.flatten() })

  res.setHeader('Content-Type', 'text/csv')
  res.setHeader('Content-Disposition', 'attachment; filename="data.csv"')

  const csvStream = format({ headers: true })
  csvStream.pipe(res)
  parse.data.data.forEach((row) => csvStream.write(row))
  csvStream.end()
})

const PORT = process.env.PORT || 3000
app.listen(PORT, () => console.log(`🚀 Server ready on :${PORT}`))

Bun handles TS without transpile step; bun run index.ts boots instantly.

5. Test Locally

bun run index.ts &
curl -X POST http://localhost:3000/convert \
  -H 'Content-Type: application/json' \
  -d '{"data":[{"name":"Alice","age":30},{"name":"Bob","age":25}]}' -o out.csv
cat out.csv

Output:

name,age
Alice,30
Bob,25

6. Benchmark

wrk -t4 -c200 -d15s --latency -s ./scripts/json.lua http://localhost:3000/convert

Expect ~140k req/s on M1 vs ~55k with Node 20.

7. Dockerfile (7 MB)

FROM oven/bun:slim-1.2
WORKDIR /app
COPY package.json bun.lockb tsconfig.json ./
RUN bun install --production
COPY . ./
EXPOSE 3000
CMD ["bun","run","index.ts"]

Build & push:

docker build -t json-csv-api .

Fly.io fly launch → detects port 3000.

8. Input Validation Tips

Reject rows > 10 000 to avoid memory spikes.
Use streaming JSON parser (clarinet) for 50 MB+ bodies.

9. Error Handling Middleware

app.use((err, _req, res, _next) => {
  console.error(err)
  res.status(500).json({ error: 'Internal' })
})

10. Security

Add helmet() for headers (works in Bun).
Limit body size.
Run behind Cloudflare or fly-proxy for rate-limiting.

11. Conclusion

With Bun + Express you get Node’s ecosystem plus Go-level speed. Ship your JSON-to-CSV micro-service in minutes and slash compute costs. ⚡️

Table of Contents