Dump SeaORM Dependency from SeaSchema's Migrator

[billy1624]

PR Info

• Closes Dump SeaORM Dependency from SeaSchema's Migrator #57
• Dependents:
  • Dump SeaORM Dependency from SeaSchema's Migrator sea-orm#666

Adds

Multiple traits, structs and enums are added to attach with SeaORM

• E.g. ConnectionTrait, QueryResultTrait, Statement, DatabaseBackend, MigrationErr

Breaking Changes

• Migrator and CLI are moved to SeaORM leaving only the SchemaManager behind
• Migration prelude module are moved to SeaORM as well

[tyt2y3]

I think it might be easier to just remove the migrator from sea-schema altogether, keeping only the SchemaManager.
And it is actually possible to just share the SQLx connections directly between the two crates, that way we don't have to split out SeaORM's connections.

[billy1624]

That would be better. This way we can avoid introducing additional connection related traits on sea-schema side

[billy1624]

I think it might be easier to just remove the migrator from sea-schema altogether, keeping only the SchemaManager.

But then we still need a way to get the connection out of SchemaManager for user to seed the db with SeaORM for instance. And, this is why I have...

impl<'c, C> SchemaManager<'c, C>
where
    C: MigrationConnection,
{
    pub fn get_connection(&self) -> &'c C::Connection {
        self.conn.get_connection()
    }
}

And it is actually possible to just share the SQLx connections directly between the two crates, that way we don't have to split out SeaORM's connections.

If we only keep SchemaManager we don't need a SQLx connection in sea-schema. Btw... if we only keep SchemaManager, why don't we move all the migration things into SeaORM? SchemaManager is just a wrapper of SeaQuery statements and some db utilities (column / table exists).

[tyt2y3]

But then we still need a way to get the connection out of SchemaManager for user to seed the db with SeaORM for instance

It should work the other way around, downstream (sea-orm) lend the connection to SeaSchema instead of SeaSchema owning it

[tyt2y3]

I think it's worthwhile to have clear-cut crate boundary between things

[billy1624]

It should work the other way around, downstream (sea-orm) lend the connection to SeaSchema instead of SeaSchema owning it

SeaORM lend the inner SQLx connection to SeaSchema?

[tyt2y3]

Yes

[tyt2y3]

Hopefully we can untangle this in 0.8.0

[billy1624]

Hopefully we can untangle this in 0.8.0

That's the target! Please review this PR @tyt2y3

[tyt2y3]

we should drop this file and clap dependency entirely

[billy1624]

The clap subcommand was used on sea-orm-cli and sea-orm-migration, both of them shared the save subcommand that's why I put it in sea-schema.

I have removed clap dependency and rewrite it as macros, 644741a.

[billy1624]

See if that looks good @tyt2y3

[tyt2y3]

A library crate cannot have clap as dependency

[tyt2y3]

Otherwise it looks good on sea-schema side