feat: support dropping multiple types with schema.dropType(), cascade. - #1516

The latest updates on your projects. Learn more about Vercel for Git ↗︎

kysely_koa_example

npm i https://pkg.pr.new/kysely-org/kysely@1516

commit: 1ae6987

@igalklebanov Will this work for compatibility? I've never really looked at any Kysely plugins or compilers, so I'm not quite sure what I'm doing.

@igalklebanov Will this work for compatibility? I've never really looked at any Kysely plugins or compilers, so I'm not quite sure what I'm doing.

It's perfectly fine to ask and not get things right the first time. I'm more than happy to help!

A library operates within an ecosystem. Public APIs are important, and smoothening transitions is something that needs to be thought about.

When you rename a property in an operation node or change the signature of the function that creates such nodes - that's a breaking change that might affect (custom) dialects (their compilers) and plugins, potentially making them unuseable with the next version of this library.

Currently, the ecosystem knows of name, and that it always exists. Let's not break that promise.
It also knows that in order to create a DropTypeNode you provide a single name as the first argument. Let's not break that promise.

What we can do:

1. Keep node.name, untouched. It will represent the leading type name in the query. This ensures downstream code that accesses node.name doesn't break.
2. Add node.additionalNames?: SchemableIdentifierNode[]. It will represent any other type names (excluding the leading type name). The cost is slighly more complex compilation code, and it's a price we're willing to pay.
3. Add a second additionalNames?: SchemableIdentifierNode[] argument in DropTypeNode.create, leaving the first function argument (name) untouched. The cost is a slice operation before invoking DropTypeNode.create, and it's a price we're willing to pay. OR, turn the first argument into a SchemableIdentifierNode | SchemableIdentifierNode[] union, and handle both cases within the creation logic.

This should now be ready for re-review. Thanks for the help :)