Data Contracts: Consensus as Code
summary
Argues that data contracts as code turn fuzzy agreements into enforceable expectations by generating models, tests and infrastructure from a single source of truth. Implements a pipeline that reads spreadsheets with pandas, validates and types with pydantic, refactors with rope and runs tests with pytest and black, producing SQL data-quality checks derived from contracts. After about a month and a half the team auto-generates tests and finds mismatches earlier, proving the approach scales; tooling list: pandas pydantic rope pytest black.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 3:34Data contracts defined
- 7:24Schema vs semantics
- 10:32Practical applications
- 13:03Data contracts in a real-world ERP system
- 16:12Implement as code
- 20:03Implementation principles
- 23:14Spreadsheets vs code
- 27:05Tooling and codegen
- 29:15Key takeaways and closing remarks
related talks
20mData Contracts: Keep Your Weekend Work-Free!
47mData Contracts
26mData contracts: driving data culture change
50mData Contracts VS Data Chaos: A Survival Story
44mBring data teams together with data contracts
39mTaming the Chaos: Open Standards for Data Contracts and Products Are the Back...
48mContract Driven Development: The Death of Integration Hell
17mOpen Source for Data Quality
