chore: Renamed changed/created_on to changed/created_at

[john-bodley]

SUMMARY

As part of [SIP-27] Proposal for Paranoid Deletes with the addition of paranoid deletes it seems that, per the SIP, the sqla-paranoid package would be useful which provides the deleted_at column.

Currently for various audit tables have changed_on and created_on yet these refer to timestamps as opposed to dates and the tendency is to use _at for timestamps and _on for dates (example). To ensure consistency with sqla-paranoid I thought there would be merit in renaming the various changed_on and created_on columns to changed_at and created_at respectively.

I probably should have dug a little further before actioning on this PR as it seems that the auditability comes from FAB which explicitly defines the changed_on and created_on columns which are used in the ab_user table. I'm uncertain how people feel about this change. Note as a work around I needed to make the AuditMixinNullable not inherit from flask_appbuilder.models.mixins.AuditMixin.

Thoughts @bkyryliuk, @dpgaspar, @villebro, et al?

Note as a side note the mypy checks are failing when deprecating the flask_appbuilder.models.mixins.AuditMixin. Irrespective of whether we go ahead with said change it's highlighted a number of previously silenced (due to inheritance) typing issues which should be resolved.

BEFORE/AFTER SCREENSHOTS OR ANIMATED GIF

TEST PLAN

CI and manual testing.

ADDITIONAL INFORMATION

• Has associated issue:
• Changes UI
• Requires DB Migration.
• Confirm DB Migration upgrade and downgrade tested.
• Introduces new feature or API
• Removes existing feature or API

[bkyryliuk]

@john-bodley I just briefly looked into the https://github.com/jeanphix/sqla-paranoid and it feels to me that reimplementing it could be more straightforward vs migrating superset dbs to follow it's convention. There are only couple hundred lines.

[john-bodley]

@bkyryliuk I agree it’s not much work to reimplement, though I do sense _at is more accurate than _on and thus adheres to other systems.

[john-bodley]

@dpgaspar what are your thoughts from a FAB perspective about introducing a breaking change which renames the audit columns from _on to _at?

[bkyryliuk]

@bkyryliuk I agree it’s not much work to reimplement, though I do sense _at is more accurate than _on and thus adheres to other systems.

yep, agreed. curious to learn how hard would it be to do this rename in the FAB. It could be a good topic to discuss in slack.

[villebro]

I'm not familiar with _at vs _on, but if it is indeed established convention, I think it's a good change, but I'd prefer to see it happen via FAB. Wrt reimplementing vs using sqla-paranoid, I find reimplementing a better alternative in this case, as it doesn't seem to be a very actively maintained project.

[john-bodley]

Closing per @bkyryliuk and @villebro's feedback.