Manhattan - Easy, High Performance Analytics

Scott Torborg - Cart Logic

Manhattan is a Python infrastructure block to provide basic server-side analytics and multivariate testing. It is:

• Easy to deploy and develop on
• Scalable
• Not slow, in-band request latency < 1ms and can process > 2k events/sec
• Customizable and flexible to varying rollup needs
• Robust to server failures, migrating between cluster topologies, and backend reconfiguration

It is also:

• 2 oz rye whiskey
• 1/2 oz sweet vermouth
• 2 dashes bitters
• 1 cherry

Example Usage

from manhattan.middleware import ManhattanMiddleware
from manhattan.log.timerotating import TimeRotatingLog

log = TimeRotatingLog('/tmp/manhattan.log')
app = ManhattanMiddleware(app, log, secret='s3krit')

def checkout_action(request):
    request.visitor.goal('Began Checkout')
    if request.visitor.split('Blue Button'):
        ...

Contents

• API Reference
• Quick Start
  • Installation
  • Architecture
  • Goal Conversions and Split Tests
  • Backend Setup
  • Next Steps
• Advanced Usage
  • Site Specific Analysis
  • Configurable Rollups
  • Complex Goals
• Contributing

License

Manhattan is licensed under an MIT license. Please see the LICENSE file for more information.