[60secs]

Because helm charts lack a schema. Programming by yaml configuration means you spend time searching github for an example kind of like what you're trying to do instead of having actual tools which allow auto-completion.

Editing helm charts it the ultimate form of throwing spaghetti at the wall and praying it works, without being able to actually test anything locally.

[belmont_sup]

At this point, if it’s painful enough, why isn’t compiling-to-yml tools more popular?

Example: https://github.com/dhall-lang/dhall-kubernetes

Haven’t used dhall myself but I’d definitely prefer a DSL on top of yaml.

[vbezhenar]

I use plain yamls and kustomize. It works for me. Not ideal, but definitely not on the level to replace it with the whole another language.

Yamls do have a scheme, editors can autocomplete it.

I've heard that helm is painful. Never used it.

[zerbinxx]

I know this may be controversial, but ChatGPT is an absolute godsend for this exact problem. K8s YAML an overly flexible markdown language (itself with infinite options that you may or may not choose to RTFM about), complex custom resources that may or may not be Done Right (wrt abstraction, readability, general best practices, unsurprising-ness), and multiple biomes worth of competing tools and possible applications, it’s fairly safe to say that, like natural language, the entire world (of k8s) is overdetermined and suffers badly from Yet Another Framework/Tool overload. Generative AI, thankfully, is really good at at least giving your workable examples of some of the features, and can make up for my own lack of interest in RTFM at 9 on Sunday night when stage goes down and we have barbarous developers eager to push their Rube Goldberg machines to production bright and early tomorrow morning.