mis-handling oneOf selection in blaze/valid_default & blaze/valid_examples

[robbat2]

Possibly related to issue #411:

{
  "$schema": "https://json-schema.org/draft/2020-12/schema",
  "default": "unknown",
  "examples": [
    "unknown",
    "infinite",
    "PT1H",
    "P1D"
  ],
  "oneOf": [
    { "const": "unknown" },
    { "const": "infinite" },
    { "type": "string", "format": "duration" }
  ]
}

$ jsonschema lint test.schema.json
test.schema.json:3:3:
  Only set a `default` value that validates against the schema (blaze/valid_default)
    at location "/default"
    The string value was expected to validate against one and only one of the 3 given subschemas
      at instance location ""
      at evaluate path "/oneOf"
test.schema.json:5:5:
  Only include instances in the `examples` array that validate against the schema (blaze/valid_examples)
    at location "/examples/0"
    Invalid example instance at index 0
      The string value was expected to validate against one and only one of the 3 given subschemas
        at instance location ""
        at evaluate path "/oneOf"

[jviotti]

Oh very interesting. Let me check

[jviotti]

Actually the behaviour is correct. The example and default value unknown WILL match both the first and the third branch on the oneOf. Remember that format does not validate, only annotate, so in the third branch you are only checking that the instance is a string, which is true.

[robbat2]

Thanks; can I request that the error message has improved clarity:
"Default was expected to match against exactly one subschemas: matched N" (where N says how many)
"Example was expected to match against exactly one subschemas: matched N" (where N says how many)

Even better would be saying which sub-schemas it matched againsts, but I don't think that would be easy.

Solved in my underlying use case with a regex for iso8601 durations in the third branch.

[jviotti]

Yeah, that makes sense. It should be possible to get that level of information. Let me play with it

Solved in my underlying use case with a regex for iso8601 durations in the third branch.

Talking about ISO durations, you might find https://www.sourcemeta.com/products/std/ interesting. We are trying to create a growing collection of standard schemas, and we already have all the possible ISO duration variants :) https://schemas.sourcemeta.com/sourcemeta/std/v0/iso/datetime/2019/duration

[robbat2]

Thanks! I decided to only accept a subset of it to constrain the users as well, but I will keep it in mind for future use cases.