Implement grammar for Avro - Githubissues

mozilla / jsonschema-transpiler

Compile JSON Schema into Avro and BigQuery schemas

Mozilla Public License 2.0

43 stars 9 forks source link

Implement grammar for Avro #29

Closed acmiyaguchi closed 5 years ago

acmiyaguchi commented 5 years ago

This fixes #14. I learned from the other grammars, making the avro schema cleaner than the other implementations. The spec was useful during this stage.

https://avro.apache.org/docs/current/spec.html

Unions are implemented using standard serde attributes. In the jsonschema module, I left the type as a Value and implemented a helper function to disambiguate Type from a Vec<Type>.

https://github.com/acmiyaguchi/jsonschema-transpiler/blob/2de4ca494d8488fac1d4041ddcf88d189bbf3671/src/jsonschema.rs#L79-L85

I also avoided the issue with flattening a nested enum by treating the enum as an untagged union. I ran into issues treating the main Type as a tagged-union and adding the serde(tag = "type") attribute on all involved types. This will flatten to the correct json format, but it requires custom deserialization. In the bigquery module, I had to implement a custom deserializer.

https://github.com/acmiyaguchi/jsonschema-transpiler/blob/2de4ca494d8488fac1d4041ddcf88d189bbf3671/src/bigquery.rs#L52-L55

I avoided using HashMaps for things that are actually Vecs. In bigquery, I did this because it makes traversal and insertions into the schema simpler. However, most of the manipulation is done within ast now, so having an easily traversable schema by path is no longer of concern.

https://github.com/acmiyaguchi/jsonschema-transpiler/blob/2de4ca494d8488fac1d4041ddcf88d189bbf3671/src/bigquery.rs#L69-L72

badboy commented 5 years ago

(I have no idea why my review was submitted twice)