Avoid coding in YAML, Use a DSL

Configuration in YAML is fine, especially for non-full-time programmers.

But avoid coding in YAML (if conditions, looping, etc.). Use a DSL in a programming language instead.

Why? Because a DSL is not restricted (since it builds on top of a general-purpose language), and accomodates customization / complexity.

Plus, the DSL can generate the YAML, if the goal is to use YAML as the format of the input to the system.

References

• YAML as data format is defensible. YAML as a programming language is not.
• Kelsey Hightower on Kubernetes & YAML
• Every Sufficiently Advanced Configuration Language is Wrong
• Why DSLs Fail
• A lesson in creating and using niche business DSLs at scale | Zerodha Tech
• https://matt-rickard.com/abstracting-the-infrastructure
• Why shouldn’t I just use python code for configuration?
• HN discussion : Don’t use a DSL where a full programming language is what you need
• HN discussion : infrastructure as code/yaml
• An alternative to explore is CUE lang (via).
• Hashicorp is providing general programming languages (TypeScript, Python) as an alternative to their custom config language (HCL) and JSON
• KubeFlow DSL compiler that generates YAML vs. Argo Workflows YAML examples
• How to build a DSL in Kotlin
• Gradle Kotlin DSL