### What needs to get done #### Acceptance criteria * A redesign of the project models (see section below) * Migrate to a schema based validation * Per-base reference documentation of the schema * Automatically generated base-aware and type-aware JSON schema * This should be generated declaratively from the project models. The schema should not be post-processed beyond [what Charmcraft does](https://github.com/canonical/charmcraft/blob/main/tools/schema.py) #### ## Rough design The hierarchy would look something along the lines of this pseudocode: ```py class Project: ... class BaseProject(Project): type: Literal["base"] build-base: BaseEnum class AppProject(Project): base: BaseEnum class BareProject(AppProject): base: BaseEnum.BARE class Core22Project(AppProject): base: BaseEnum.CORE22 class Core22KernelProject(Core22Project): type: kernel UnionAppProject = ... UnionKernelProject = ... SnapcraftProject = Annotated[ BaseProject | SnapdProject | GadgetProject | UnionAppProject | UnionKernelProject, Field(discriminator="type") ] ``` ### Why it needs to get done Snapcraft has a monolithic project model that represents core22 and core24 snaps in addition to different types of snaps (e.g. base snaps, kernel snaps, gadget snaps). There are 3 major problems with a monolithic project model: Documentation * It’s not possible to generate per-base documentation of the project model JSON schema * It’s not possible to generate a base-aware JSON schema, so editing a snapcraft.yaml in an IDE isn’t a good experience. IDEs inform users that valid keys are invalid and vise-versa. Maintainability * Instead of using modern programming constructs like inheritance, model validation is driven by sets of if/else statements. This is bug prone and burdensome to maintain.