I wonder about the name. Often something that bypasses some rules is called "dirty". Would it make sense to call the function validate_dirty_uniqueness or dirty_validate_uniqueness?

@michalmuskala "dirty" implies dirty tracking (i.e. check if an object was touched)

I prefer the name "validate_optimistic_uniqueness". This is after all an "optimistic" check.

@michalmuskala I tried the process dictionary approach - see latest commit.

Personally I like this approach less than the previous one. I had to guess to discover the values {1, [true]} and {0, []} that the adapter returns, and relying on side effects elsewhere doesn't seem as clean as passing something like FakeRepoWithDup to the validation. But I'm happy to use this approach if it's more in line with Ecto test conventions. What do you think?

Regarding the name: all the existing validation names are of the pattern validate_{some property of the changeset}, so I like the idea of having this one start with validate_unique, then adding an adverb. tentatively seems to fit well; the definition is "subject to further confirmation", which is exactly what we want to convey: this needs to be followed up with a constraint check.

optimistically also seems OK given how the word is used in programming contexts, but to me it's less obvious what it means here; the point is not that we have a sunny outlook on life 😆 , but that we already know we'll need to do some further checking.

I should say - I'm not adamant about the name. Happy to call it whatever y'all think best.

My primary concern with "tentatively" is that it's not a very common word and might be confusing for non-native speakers.

Perhaps validate_unique_naively to fit in with NaiveDateTime

Perhaps validate_unique_naively to fit in with NaiveDateTime

That could work. But it might convey "this approach isn't very sophisticated" as opposed to "this is intended to be the first check, to be followed up somehow with a final check." "Tentative" has the built-in idea of "needs confirmation".

My primary concern with "tentatively" is that it's not a very common word and might be confusing for non-native speakers.

Ah. Maybe so. It's also not super easy to type. 😆

"Preliminary" means something similar to "tentative", but "preliminarily" is awful. 🤣

I still believe "optimistic" is the correct word in the computer engineering context (validate_optimistic_uniqueness)

https://en.wikipedia.org/wiki/Optimistic_concurrency_control
"While running, transactions use data resources without acquiring locks on those resources."

Conversely, a proper "pessimistic" uniqueness validation at insert time is one that locks the database so that no other transaction can modify the data (e.g. create a duplicate value) while the insert is occurring.

@johnnyshields I see what you're saying, but I think a unique constraint in the db is the optimistic check here - we don't lock the table before inserting, but if, at the last moment, it turns out that there's a duplicate, we roll back the transaction. The check is optimistic (no up-front locks), but it's reliable.

Validating uniqueness doesn't offer any guarantees, just a better user experience in the 99.99% case where the duplicate already exists at the time of validation.

❤️ 💚 💙 💛 💜