Should c:insert/2, c:insert!/2, c:update/2, and others validate changes thru changesets?

[joeljuca]

I just noticed that most repo mutating callback functions validate structs through the schema module's respective changeset/2, which is different than what happens when Repo.insert/2 is used. I'm not sure what to do in this case.

Should it validate all changes through changeset/2 or should they dangerously allow inserts directly, without caring with changeset validations (effectively delegating it to the respective functions)?

[joeljuca]

I think it's wise to check how these functions differ in Ecto APIs (aka.: which ones validate changes w/ changesets and which do not) and replicate it in SwissSchema APIs.

Since it'll probably cause breaking changes, it's important to advertise, both in docs and social media, that a version bringing this change (probably v1.0.0) will be an API-breaking one.

[joeljuca]

To be continued on #20.