LEARNING LAB / CONTRACTS

Data Contract Lab

Will this upstream change break a reader?

Compare producer promises with consumer expectations, replay the same events against both versions, and inspect the evidence. All checks run in your browser.

Browser-local · No uploads · Synthetic examples

Four experiments. One explicit direction.

New producer → existing consumer. A changed contract is a promise; replayed samples are observations. Neither executes your production implementation.

DCL/1 is a restricted contract language, not a full JSON Schema validator. Unsupported keywords are input errors.

Follow the six-unit reading and repair exercise →

Paste an input bundle

Input bundle JSON

Limits: 2 MB per input; 500 JSONL rows; 12 consumers; schema depth 8 / 160 nodes. A trailing newline is allowed. Interior blank lines are errors.

Old producer contractImport local file

New producer contractImport local file

Consumer expectationsImport local file

Sample events · JSONLImport local file

Load a synthetic scenario to begin.

Validation report

Samples can expose a failure, but cannot establish universal behavioral compatibility. “Proven inclusion” refers only to structural value sets in DCL/1. Semantic checks rely on explicit declarations.

Producer → consumers

Both producer-contract directions

Same rows, both contracts

Line	Result	Old: structure / semantics	New: structure / semantics	Introduced / resolved / unchanged

All validation errors

Line	Stage	JSON Pointer	Code	Message

Save evidence

JSON and readable exports include the supplied contracts and names, generated witnesses and diagnostics. Raw event values are omitted. CSV prefixes formula-like cells and quotes every field; do not remove those guards when importing into a spreadsheet.