Linting controls

[dkolbly]

Related Issue:

#1061 Support linting of enum and sibling conventions

Description of changes:

• Adds a suppress_checks option to the metaschema to configure turning off certain linting rules
• Turns off those linting checks for places where we have violated the conventions (there are about 3)
• Fixes data_lifecycle_state_id to follow the enum convention by adding a 99 (Other) enumerand and articulating that it should be used for "other"

[dkolbly]

I will note that the validator in ocsf-validator does not yet understand lint and hence is failing against this PR. If discussion proceeds, I can probably address that issue as well at least at a surface level.

[rmouritzen-splunk]

The ocsf/ocsf-validator needs a change to support the new lint attribute. I believe the change will be in ocsf_validator/types.py, adding a lint key to OcsfAttr:

OcsfAttr = TypedDict(
    "OcsfAttr",
    {
        "$include": NotRequired[str],
        # "caption": NotRequired[str],
        "caption": str,
        "default": NotRequired[Any],
        "description": NotRequired[str],
        "enum": NotRequired[Dict[str, OcsfEnumMember]],
        "group": NotRequired[str],
        "is_array": NotRequired[bool],
        "lint": NotRequired[Sequence[str]],
        "max_len": NotRequired[int],
        "name": NotRequired[str],
        "notes": NotRequired[str],
        "observable": NotRequired[int],
        "range": NotRequired[Sequence[int]],
        "regex": NotRequired[str],
        "requirement": NotRequired[str],
        "sibling": NotRequired[str],
        "type": NotRequired[str],
        "type_name": NotRequired[str],
        "profile": NotRequired[Optional[Sequence[str]]],
        "values": NotRequired[Sequence[Any]],
        "@deprecated": NotRequired[OcsfDeprecationInfo],
    },
)

(I say "I believe..." because I haven't tried this change against the modified schema myself.)

[dkolbly]

I will try it out!

[alanisaac]

Do we have rules for naming conventions for fields like this? Thinking about:

• We use a $ prefix for references to other fields (which is similar to JSON schema's conventions, as well as comments)
• We also have a @ prefix in @deprecated as an annotation for the schema. I don't know the rationale behind that prefix, but it reminds me of a Python decorator (such as @deprecated here).

Should lint follow one of these conventions (@lint or $lint) since it's more of a validation directive or annotation than something that influences the final schemas themselves?

[dkolbly]

@alanisaac great question... I don't have an opinion on that, really. There are lots of other properties that directed at the compiler (i.e., that do not appear in the produced concrete schema, e.g., JSONSchema) like extends and the top level name property in files, so I didn't think anything of using an undecorated property name for lint. The notion of linting the schema itself is new, so I'm not sure where else to look for a comparative example...

[rmouritzen-splunk]

I see a couple things:

1. The suppressions use skewer-case instead of snake_case. Perhaps they should be snake_case for consistency.
2. It's not clear to me what check each is actually suppressing. Can these be described somewhere?

(I'm quite interested because I'm adding a new event validation API, and I'll want to add support for these suppressions to that API.)