Robert Smallshire

Preparation

• This is a workshop, not a lecture, so come with a laptop or pair-up with somebody who has one.

• This workshop requires at least Python 3.4 so you should have a suitable interpreter installed.

• Bring your favourite Python code editor.

• Have the workshop student materials installed from Bitbucket. Further instructions are to be found there.

• You can also look at the introductory slides which were used in the workshop.

Synopsis

In first part of this workshop we explore how to implement rich domain models using plain-old Python objects which are completely independent of any particular persistence technology such as an object-relational mapper. Domain models often embody the core value of software systems, so implementing models independently of - often ephemeral - technology choices is an important strategy for long-lived, high-value systems.

In the second part of this workshop we implement an event-sourced architecture for persistence of our domain model, whereby all changes applied to the model are recorded as events in a simple, append-only event-store. From this event store we can reconstitute the model state as it was at any historical time, something that is difficult – if not impossible – to do with most object-relational solutions. Furthermore, we can project the event stream into other representations to support queries which are not conveniently supported by the entities in our model, or which roll-up historical data. Among many other benefits, such projected read-models gives very high scalability on the read-side.

This workshop is aimed at Python developers who want to step beyond the limits of canned framework-based solutions such as Django models, or toolkits such as SQLAlchemy, (which are ultimately about managing the horrors of shared-mutable state). Event-sourced domain models facilitate allow us to combine the best aspects of object-oriented design with robust and simple persistence inspired by functional-programming practice.