Avro Schema Evolution
Check whether the new reader can read data written with the old schema, and whether the old reader can read new data. The inspector applies a bounded Apache Avro 1.12 writer–reader resolution profile to every possible writer union branch and enum symbol. It reports the actual matching rule and a path for each promotion, default, alias or incompatibility.
Key features
- Analyze old-writer to new-reader and new-writer to old-reader separately
- Resolve records, enums, fixed, arrays, maps, unions and earlier named references
- Show reader defaults, writer-only fields, numeric/string-byte promotions and reader aliases
- Require every possible writer union branch and enum symbol to resolve
- Reject invalid defaults and unsupported logical types rather than guessing
- Download a versioned JSON report after local analysis
How to use
- Load the Account example or paste the old and new Avro schema JSON documents.
- Select Analyze compatibility to parse both schemas under the displayed limits.
- Read the two directional verdicts and each rule/path before planning a rollout.
- Download the report if you need to review the decision with your team.
Use cases
- Check whether a new consumer reads older records
- Review a producer rollout against older consumers
- Investigate why a renamed field needs a reader alias
- Compare union branches and enum fallback before changing a schema
Frequently asked questions
Does adding a field always break older data?
A new reader can fill a missing writer field only when the reader field has a valid default. The reverse direction is checked separately.
Are aliases directional?
Yes. This profile applies type and field aliases declared by the reader to names used by the writer.
Does one compatible union branch make the schema compatible?
No. Every possible writer branch must find a matching reader branch for an all-data compatibility verdict.
Does this decode Avro files or cover logical types?
No. It compares bounded schema contracts; binary/JSON records and logical types are outside this profile.
Is a compatible result a production certification?
No. Check real producer and consumer implementations, stored writer schemas and representative data before deployment.
Privacy
Schema text is analyzed in the browser and is not sent to a tool API or stored automatically. A downloaded report includes both input schemas, so handle it as sensitive project data.
Comments & questions