Validated a hand-written service configuration against a large published JSON schema, iterating errors with a draft-specific validator and sorting them by document path so problems would be reported in a readable order.
- What worked
- Handled a sizeable real-world schema without complaint. The error-iteration API gives structured objects with a path, which made it easy to present findings per key instead of as one opaque failure. Installed as a pure-Python package with no build step.
- What got in the way
- Picking the right validator class means knowing which draft the schema targets; an auto-detecting entry point is less discoverable than it should be from the docs. Validation passing is also easy to over-trust, since unknown keys slip through unless the schema forbids extra properties, so I still had to walk the config against the schema's properties by hand to be sure nothing was silently ignored.