feat: specify_any_of(), specify_one_of(), and specify_all_of()

[jonthegeek]

Summary

As a package developer, in order to build reusable, pre-configured composition validators, I would like specify_*() factories for the any_of, one_of, and all_of families.

Details

specify_any_of(), specify_one_of(), and specify_all_of() each capture a set of specs and return a pre-configured function (of class stbl_specified_fn) that calls the corresponding stabilize_*_of(). Their implementations are nearly identical — they differ only in which stabilize_*_of() they wrap — so they are grouped into one issue.

These factories are what make composition composable: e.g. stabilize_each(x, specify_any_of(specify_int(), specify_chr())), and they map directly onto nested JSON Schema composition (an object property whose schema is itself an anyOf / oneOf / allOf).

Proposed signature

specify_any_of(...)
specify_one_of(...)
specify_all_of(...)

Arguments

• ... — Unnamed stabilizer / to_* / specify_*() specs, captured and forwarded to the underlying stabilize_*_of().

Returns a function of class c("stbl_specified_fn", "function") that calls the corresponding stabilize_*_of() with the captured specs.

Behavior

• Follow the existing specify_lst() pattern: capture element_specs <- list(...) and return a structure(function(x, ..., x_arg, call, x_class) rlang::inject(stabilize_*_of(x, !!!element_specs, ...)), class = c("stbl_specified_fn", "function")).
• Depends on the three underlying functions: stabilize_any_of() (the *_one_of() → *_any_of() rename), the new exactly-one stabilize_one_of(), and stabilize_all_of() (feat: stabilize_all_of() family #278).
• Keep behavior/argument-echoing consistent with the other specify_*() functions (e.g. Show actual args in specify_*() #161).

References

• Existing pattern: R/specify_lst.R, R/specify_cls.R.
• JSON Schema draft 2020-12, anyOf / oneOf / allOf: https://www.ietf.org/archive/id/draft-bhutton-json-schema-validation-01.html