Breaking changes for v2.0

[cljoly]

A wishlist of breaking changes for the next major version:

• Change return types so that we can fix the cargo clippy pedantic warnings, like casting `i64` to `usize` may lose the sign of the value or casting `usize` to `u32` may truncate the value on targets with 64-bit wide pointers
• Introduce intermediary types so that we could write M::up(…).hook(…).down(…).hook(…) giving up on this due to the complexity it would introduce.
• Remove deprecated items

Other things to decide on:

• Remove Eq implementation on M Comparison on migrations with hooks causes a stack overflow #45
• Remove PartialEq from our Error type?
• Remove the async support (Re-evaluate Async Support #226)

I don’t know if a breaking version will ever be released though, there are costs to forcing users to migrate.

[czocher]

We can probably also remove PartialEq from our Error implementation.

Maybe also migrate to thiserror?

[cljoly]

We can probably also remove PartialEq from our Error implementation.

That’s a good point, I’ve added it to the checklist at the top.

Maybe also migrate to thiserror?

What prevents us from doing it? We could do it in a way that doesn’t break the API compatibility, couldn’t we?

[czocher]

What prevents us from doing it? We could do it in a way that doesn’t break the API compatibility, couldn’t we?

I believe so, we can create a task for that if we deem it worthwhile.

[cljoly]

I’ve created #95 for this.

[cljoly]

Going over the list of proposed breaking change, we have implemented a few of them and we have enough breaking changes as it is.

The other breaking changes are more disruption than they are worth.