divergent migration handling is missleading

[ignatz]

Was just playing around with divergent migrations, i.e. same version but different name.

The Runner has a configuration flag "set_abort_divergent" that does what the name suggests. However, it may come as a surprise that divergent migration are then simply not applied (they also cannot, since the version column in the migration schema has a unique constraint).

There's a lot of error logging going on, it might be worth letting the user know that while the migration proceeds, it is only partially applied 🤷

[jxs]

Hi, and thanks for your interest! Would you be open to submit a PR addressing it?
Thanks!

[ignatz]

Would you be open to submit a PR addressing it?

I normally would by default and since you asked explicitly I did: #341 .

That said, let me share some thoughts and please don't take this personal, maintaining open source is hard and refinery is much appreciated. Ok, here it comes...

I held myself from sending a PR because I wasn't sure if this project/respository is still alive:

• A unit test is failing at head
• Formatting isn't applied
• Some of the code could benefit from some streamlinging (The PR is so long because I felt the urge to opportunistically refactor)
• (and least importantly and understandably so, you seem to be rejecting new drivers. Which makes sense but means that I'm maintaining my own fork anyways)

All that said, this is all very normal. We're all busy and have other things going on. I'm mostly just wondering where you see the future of refinery? Do you need help?

[ignatz]

Coming back to this, I guess I'm full of crap. I did break a bunch of tests. Should be fixed now