feat: ExceptionStatement raises custom SQL errors

[gronke]

PR Info

Allows defining SQL errors, for instance in an IfElseStatement #828 inside a TriggerStatement #824 that rejects an update on certain conditions.

New Features

• ExceptionStatement raises custom SQL errors

Breaking Changes

• Add SimpleExpr:: Exception variant to public enum

[Expurple]

Haven't reviewed the rest yet. Consider running cargo semver-checks to find and document the breaking changes.

Disclaimer: I'm not a maintainer

[Expurple]

You should mention that this is a breaking change (adding a new variant to a public enum not marked as #[non_exhaustive]).

Perhaps, we should add #[non_exhaustive] in the next major version. That's also a breaking change. But after that, changes like yours can me merged anytime with no problems

[tyt2y3]

Expurple is a big contributor :)

we should add #[non_exhaustive] in the next major version

I agree, and I am starting to merge breaking changes now

[Expurple]

As I look at the PostgreSQL docs, I'd rather have a powerful Postgres-specific PgRaiseStatement with an explicit

#[non_exhaustive]
pub enum PgRaiseLevel {
    Exception,
    ..
}

where later we could extend PgRaiseStatement in a backwards-compatible way and add methods for format args, SQLSTATE and USING (I'm not asking to contribute everything under the sun in this one PR).

We can keep your minimal "portable" API too, if you document that it's not a portable SQL feature, but an "artifitial" shortcut for the common subset of database-specific exception features. Which all seem very different from each other, judging by the SQL you generate in the implementation. And obviously, it'd switch the implementation to just delegate to the database-specific query builder: PgRaiseStatement, etc.

Personally, I develop against Postgres and don't care about portability. But I'm not sure how much the SeaQL community cares in general

[Expurple]

Oh, this doesn't look good to me. Why make it a provided method if you don't, in fact... provide it?

To me, such a provided method would look OK if it returned Result<(), ExceptionsNotImplemented> instead of panicking. But I see that the calling prepare_simple_expr_common doesn't return a Result anyway. Perhaps, we've hit some problem/limitation of the sea_query design that I don't fully understand yet. Perhaps, we shouldn't even try to pretend that exceptions are a portable and universally-supported feature (see #829 (comment))