[jira] [Commented] (AVRO-3759) [Rust] Schema types inconsistency

["Martin Tzvetanov Grigorov (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/AVRO-3759?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17724144#comment-17724144 ] 

Martin Tzvetanov Grigorov commented on AVRO-3759:
-------------------------------------------------

Please send a Pull request with your suggested changes! It will make it more clear what would be the benefit.

> [Rust] Schema types inconsistency
> ---------------------------------
>
>                 Key: AVRO-3759
>                 URL: https://issues.apache.org/jira/browse/AVRO-3759
>             Project: Apache Avro
>          Issue Type: Improvement
>          Components: rust
>            Reporter: Fedor Telnov
>            Priority: Major
>
> That is how Unions and Records are defined in apache-avro Rust crate:
>  
> ```
>     /// A `union` Avro schema.
>     Union(UnionSchema),
>     /// A `record` Avro schema.
>     ///
>     /// The `lookup` table maps field names to their position in the `Vec`
>     /// of `fields`.
>     Record {
>         name: Name,
>         aliases: Aliases,
>         doc: Documentation,
>         fields: Vec<RecordField>,
>         lookup: BTreeMap<String, usize>,
>     },
> ```
>  
> That is inconsistent - one variant is defined with pattern Type(Type) which is VERY convenient(for instance, it allows one to use UnionSchema as type), and the other one(Record) is simply a struct variant, which is not very convenient. My proposition is to hide Record's body in RecordSchema type and use it here - just as you do with union. That would significantly help users, as we can't use enum variants as types in Rust. It can also be done with other schema types as well.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)