Release 0.22 ?!

[martin-g]

Shall we make a new release ?
We have several PRs with API breaks and I'm afraid this may frustrate some users while upgrading.

[Kriskras99]

Do we expect more breaking changes? I think it's best to have as many breaking changes as possible in one release. This way a user only has the change the code once and can then continue with upgrading without having to change the code.

[Kriskras99]

If we do start the release process we should take special care with the release notes and maybe update migration_guide.md (or delete it entirely, having it but incomplete will confuse users)

[martin-g]

Do we expect more breaking changes?

I didn't expect even the ones we have so far 😄

I think it's best to have as many breaking changes as possible in one release

👍🏻

f we do start the release process we should take special care with the release notes and maybe update migration_guide.md (or delete it entirely, having it but incomplete will confuse users)

Right! I think it would be better to add upgrade notes for 0.21 -> 0.22

[Kriskras99]

Decisions to make before 0.22

• fix!: Name and Alias incorrectly implement From<&str> #423
• feat: RecordSchema/RecordField::builder().name() accepts Into<String> #429
• Why does AvroSchemaComponent::get_schema_in_ctxt take a Names? #400
  • It would be good to change the trait to take a HashSet<Name> instead of a HashMap<Name, Schema>. It would be more efficient by reducing unnecessary clones
  • I expect that this wouldn't break a lot of code, as most usage will be via the AvroSchema trait and most implementations via the AvroSchema derive
• Schema resolution lacks named type checking and allows invalid coercions #383
  • Value::resolve(reader_schema: &Schema) should become Value::resolve(writer_schema: &Schema, reader_schema: &Schema)
• Missing support for custom attributes on primitive types #368
  • To fix this issue the Schema type would need to be changed
• Enums don't serialize correctly #365
  • To properly support enums in to_value it should take a writer_schema: &Schema and then use SchemaAwareWriteSerializer
  • To properly support enums in from_value it should also take a writer_schema: &Schema and we would need a SchemaAwareReadDeserializer (feat: Serde driven schema aware deserialization #237) (We don't actually need to implement this yet, just change from_value to take the schema and ignore it)
• Move the bytes module to the serde module #430
• Do we have any deprecated items that we want to remove now?

I would also like to update the Serde documentation before this release. I think enum support will go to the next release.

[Kriskras99]

#431 would also be a breaking change (but also a bugfix) so might also be nice to include in 0.22 (if it is actually a bug and not as intended)