We're about to submit a new version of ellmer to CRAN but we have test failures for constructive.
Failure (test-s7-elmer_TypeBasic.R:3:3): ellmer::TypeBasic
Expected `recreated` to equal `expected`.
Differences:
actual vs expected
`ellmer::type_array(items = ellmer::type_object(x = ellmer::type_boolean(), `
- ` y = ellmer::type_string(), z = ellmer::type_number(), json = ellmer::type_from_schema("[1,2]"), `
- ` .additional_properties = FALSE))`
+ ` y = ellmer::type_string(), z = ellmer::type_number(), json = ellmer::type_from_schema("[1,2]")))`
They stem from the fact that we've deprecated the .additional_properties argument of type_object(), changing the formal default from FALSE to lifecycle::deprecated() (tidyverse/ellmer#976).
I asked Claude to analyse the constructive source and it suggested adding a version gate in test-s7-elmer_TypeBasic.R, following the existing with_version() pattern for the same argument.
As a heads up, I'm planning to release this ellmer update within the next two weeks.
We're about to submit a new version of ellmer to CRAN but we have test failures for constructive.
They stem from the fact that we've deprecated the
.additional_propertiesargument oftype_object(), changing the formal default fromFALSEtolifecycle::deprecated()(tidyverse/ellmer#976).I asked Claude to analyse the constructive source and it suggested adding a version gate in
test-s7-elmer_TypeBasic.R, following the existingwith_version()pattern for the same argument.As a heads up, I'm planning to release this ellmer update within the next two weeks.