Kjuhwa Skills Hub Type Safe Enum Generation Per Language

Type Safe Enum Generation Per Language

Trigger: You support 300+ string-keyed types and want compile-time validation in every language without hand-maintaining four enum lists.

Steps

• Parse content_types_kb.min.json once; produce a deterministically sorted list of labels.
• Generate per language: Rust pub enum (with non_exhaustive), Python StrEnum, TS string literal union, Go const iota with String().
• Apply consistent name normalization (3gp → _3gp in Rust due to no leading-digit identifiers).
• Generate from_string / to_string + a SIZE/COUNT constant.
• At codegen time, validate every label is present in the target model — fail the build on drift.
• Add a 'DO NOT EDIT' header to every generated file and check generated files into git.

Counter / Caveats

• Label normalization rules differ per language; keep them in one place.
• Rust enums with 300+ variants slow compile time noticeably; non_exhaustive helps downstream evolution.
• Index stability matters if downstream uses int indices for array lookups; renumbering breaks model output mappings.
• Tooling (rust-analyzer, pyright) needs the generated files to exist before it can resolve symbols — generate before checkout-time tasks.

Source

Extracted from magika (https://github.com/google/magika.git @ main).

Files of interest:

• rust/gen/src/main.rs:40-100
• python/src/magika/types/content_type_label.py:19-50
• js/src/content-type-label.ts

Source: kjuhwa/skills-hub — distributed by TomeVault.