JSON Codec Performance Guide¶

Overview¶

PyProc supports multiple JSON codec implementations for optimal performance:

• Standard Library (json-stdlib): Default, most compatible
• goccy/go-json (json-goccy): High performance, drop-in replacement
• segmentio/encoding (json-segmentio): Fastest performance

Build Tags¶

Select the JSON codec at compile time using build tags:

# Default (standard library)
go build ./...

# Use goccy/go-json
go build -tags=json_goccy ./...

# Use segmentio/encoding (recommended for production)
go build -tags=json_segmentio ./...

Benchmark Results¶

On Apple M2 Max (arm64):

Runtime Detection¶

Check which codec is being used:

import "github.com/YuminosukeSato/pyproc/pkg/pyproc"

func main() {
    codecType := pyproc.GetJSONCodecType()
    fmt.Printf("Using JSON codec: %s\n", codecType)
}

Recommendations¶

• Development: Use default (standard library) for maximum compatibility
• Production: Use segmentio/encoding for best performance
• Compatibility Issues: Fall back to goccy/go-json or standard library

Building for Production¶

# Build with fastest JSON codec
go build -tags=json_segmentio -o myapp ./cmd/myapp

# Verify codec in use
./myapp --version

Performance Impact¶

With segmentio/encoding, the JSON codec overhead is reduced by ~60%, contributing to achieving our target latencies: - p50 < 100µs (JSON codec ~0.5-1µs) - p99 < 500µs (JSON codec ~1-2µs)