Zach Daniel
Zach Daniel
Yeah, that code is pretty convoluted as well, so make sure that any other usages of those functions are updated. Are they public? I forget.
Hmm...I hadn't anticipated this workflow. There are two options: 1. add an option to that section like `define_struct?` that you can set to `false` and then we skip that logic....
Yes, you're right actually. I'm open to alternative solutions but not being able to override `new!` which was an explicit part of the API has to take precedence over opening...
Nah, this one is different.
For now we can do this just for `upsert_skipped?`.
And this should happen as basically the first thing in `Ash.Changeset.for_update` and `Ash.Changeset.for_destroy`.
Yeah I think a reproduction will be necessary. I will revisit once a reproduction exists (if @ken-kost doesn't get to it first 😉 )
@ken-kost your repro was against ash core but this would only happen on ash_postgres tests.
Awesome thanks for that. In that case we'll need a reproduction from the original poster :)
Unlikely, I think it just sends you down a different code path. We should fix the original underlying bug.