[FEATURE]: Use a schema language abstraction.

[jspears]

Describe what you want

Coming to Drizzle from Prisma, I find I miss having a a succinct schema language to describe and annotate models.

In this end I wrote a PoC Drizz for using TypeSpec to generate drizzle/schema.ts from schema.tsp. Not sure if anybody else would be interested in such a thing, it's fairly complete for Postgres, and at least shows how it could be done. It would pretty easy to write an emitter for MySql, and PSSql.

Advantages of using TypeSpec is it already has IDE support, linter support and an existing extensible grammar. As such one could add their own annotations to a model and generate DB/Rest whatever from a model.

Some examples can be found in the tests

The schema looks like

       @table model User {
          @id id: integer;
          name:string;
        };
        @table model Book {
          @id id: integer;
          name: string;
        };
        @table("books_to_authors") 
        @id("booksToAuthor", #["authorId", "bookId"])
        model BooksToAuthors {
          authorId: integer;
          bookId: integer;
        };

      @table("users") model User {
          @uuid @id id: string; 
          @unique email: string;
          rank: int32;
          @map("cool_score") coolScore: int32;
        };
        @table("blogs") model Blog {
          @uuid @id id: string; 
          @unique name: string;
          posts: Post[];
        };
        @table("posts") model Post {
          @uuid @id id: string; 
          title: string;
          content: string;
          @relation(#{fields:#["blogId"]}) blog: Blog;
        };

Thanks for the great project.

[ehsan2003]

I disagree with this approach since it brings one more level of complexity and another build step to add