HHH-16311 - Migrate away from UserType for enum handling

[sebersole]

Simple proof for handling enum values with the UserType machinery

[gavinking]

Does this eliminate all usages of org.hibernate.type.EnumType? (I ask because you didn't delete it.)

[sebersole]

Well, that's the question isn't it ;) As I said, this is all just a quick proof. Still need to analyse all uses of EnumeratedValueResolution#getLegacyResolvedBasicType.

One use-case that comes to mind immediately is the SQM handling of enum value names as literals (i.e. ... gender=MALE).

[sebersole]

Actually this was really easy to test. There were just a few errors - looking through them

[gavinking]

One use-case that comes to mind immediately is the SQM handling of enum value names as literals (i.e. ... gender=MALE).

this is one of the things that was causing me a lot of pain when I tried to handle mapping to Postgres enum types.

[sebersole]

As a quick test I changed EnumeratedValueResolution to always return the ConvertedBasicType, even for the #getLegacyResolvedBasicType case. That resulted in a few errors, but all were related to tests casting to the CustomType and EnumType classes directly. I'll push here in a bit to see what affects it has across all the databases.

[sebersole]

This latest push should be pretty complete.

One change I'll probably revert is that this push limits "enum short name" handling to just implicit and explicit @Enumerated mappings. I was just playing around with hooks from EnumeratedValueResolution to think about native database type handling, aux database objects, etc.

[gavinking]

OK then, I'll try making #6232 work on top of this patch, and see if it helps.

[sebersole]

See especially InFlightMetadataCollector#registerEnumMapping...

I'm not yet sure what to think of the test failures from the last push. They stem from the fact that we now interpret mappings like @Enumerated(STRING) @Column(length=1) using Types.CHAR / Character. That change was never propagated into EnumType - meaning envers always saw Types.VARCHAR / String. I'm inclined to think the consistency is a good thing.

I have not yet looked into the Oracle test failure from the second push

[hibernate-github-bot]

Thanks for your pull request!

This pull request appears to follow the contribution rules.

› This message was automatically generated.

[sebersole]

The Oracle failure is ultimately https://hibernate.atlassian.net/browse/HHH-16333

Assuming all tests pass (other than the one failure on Oracle) in this latest push, I'm going to integrate this upstream

[sebersole]

Integrated upstream