What ryl won't do¶
ryl lints general YAML and can validate user-supplied JSON Schemas. Its fixes never change what a document means. Check the boundaries below before filing a request.
Rules for one YAML flavour¶
ryl has no built-in rules for GitHub Actions, Kubernetes, Ansible or any other YAML dialect. Use the dialect's own linter alongside ryl: actionlint, kube-linter or ansible-lint.
Rules without evidence, or for contested style¶
A new rule needs more than one reputable source, such as the YAML 1.2.2 spec, yaml.org, PyYAML's documentation or the yamllint maintainers, calling it best practice. Blog posts alone don't count. Style with no broad consensus stays out, for example:
nullversus~- key-naming conventions
- flow versus block sequences
- exact blank-line counts
Rules about application content¶
Rules check how a document is written and what YAML itself makes of it, such as
a truthy word, an implicit octal or a merge key overriding a value. They don't
include built-in application policies. The optional schema rule
checks allowed keys and value ranges against user-supplied JSON Schemas.
Mapping key order is form,
since YAML gives it no meaning, so key-ordering fits.
Fixes that change meaning or reprint the document¶
--fix edits the document in place rather than reprinting it, and keeps its
resolved value, comments and anchors. That rules out:
- Fixes that guess intent, such as rewriting
notofalse: the author may have meant the string. One would only ever sit behind an explicit opt-in. - Whole-document reprinting, the way Prettier formats YAML. Use a
reprinting formatter if that is what you want; ryl won't become one.
ryl formatedits in place under the same guarantees.
JSON Schema validation¶
User-supplied JSON Schemas are in scope. Remote fetching requires independent user origin grants. In-file comment directives require explicit TOML opt-in; catalogs and built-in dialect policies are outside the current implementation.