davsclaus opened a new pull request, #27231:
URL: https://github.com/apache/camel/pull/27231
**Stacked on #27227 (CAMEL-25194)** — both touch `YamlValidator`; merge that
one first.
The branches of a pick-one `oneOf` such as `unmarshal` each require their
own key, so a file that chose one collects "required property X not found" for
every key it did **not** write. The key it *did* write is missing from that
list, because its branch got past the required check and failed, if at all,
further in. `filterOneOfNoise` then keeps one branch for all the places
validated against the same construct, so a second place inherits the first
one's winner:
```yaml
- choice:
when:
- simple: "${header.encoding} == 'JSON'"
steps: [ { unmarshal: { fhirJson: { fhirVersion: "{{fhirVersion}}" }
} } ]
- simple: "${header.encoding} == 'XML'"
steps: [ { unmarshal: { fhirXml: { fhirVersion: "{{fhirVersion}}" }
} } ]
```
```
Line 16: /0/.../when/1/steps/0/unmarshal: required property 'fhirJson' not
found
```
The file is correct — `fhirXml` is a data format of `unmarshal` and
`fhirVersion` is a placeholder the runtime resolves. `fhir-sink.kamelet.yaml`
of camel-kamelets has exactly this shape, which is where it was found.
## The change
The `required` errors of a pick-one construct are dropped where the object
at that place has **something in it**: a branch was chosen, so the others only
say which keys were not used. An empty object, or a key no branch knows, still
gets the construct's own error and the unknown-property error — which are the
ones that tell the author what to do.
## Why it is this narrow
The general fix — a winner per place rather than one per construct — also
works, and I measured it: it takes the fhir shape and all 250 kamelets to zero.
But it costs `YamlValidatorTest.testTypeMismatchFiltersOneOfNoise`, which goes
from 3 errors to 6. That assertion *is* the filter's purpose, and it fails
because picking a winner per place stops a place borrowing evidence from
another place — which is sometimes exactly what produces the clean report.
Making that work wants the ranking improved first, so it is recorded on the
issue as its own piece of work rather than attempted here.
## Verified
- **All 250 out-of-the-box kamelets of camel-kamelets validate clean**,
where before `fhir-sink` did not.
- The documentation corpus, the EIP examples and the XML round-trip corpus
of this module are unchanged — 180 tests green, including
`testTypeMismatchFiltersOneOfNoise`.
- `OneOfAlternativesTest` adds five cases: the two-branch fhir shape; one
branch chosen in several places; choosing nothing is still reported; choosing
something that is not a data format is still reported; an unrelated typo is
untouched.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
https://claude.ai/code/session_01Bp3538HRBPMQkb5ta9xRaj
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
