Skip to content

Compound Keywords

Compound keywords combine or restrict multiple schemas.

allOf

All sub-schemas must be satisfied simultaneously. The generator merges them into a single schema before generating.

{
  "allOf": [
    {"type": "string", "minLength": 5},
    {"type": "string", "maxLength": 10}
  ]
}

"eDPVAeCOB"

Strategy: all sub-schemas are merged via the same constraint-merge logic used internally. Conflicting constraints (e.g. incompatible types, minimum > maximum) raise UnsatisfiableConstraintsError.

Limitations: merging is keyword-based. Semantic conflicts that cannot be detected statically (e.g. two mutually exclusive pattern values) will not be caught until generation fails.

anyOf

At least one sub-schema must be satisfied. The generator picks one sub-schema's type at random and generates from the subset of anyOf branches that share that type.

{
  "anyOf": [
    {"type": "string"},
    {"type": "integer"},
    {"type": "boolean"}
  ]
}

2233

Strategy: branches are grouped by type. One type is chosen at random, and a random subset of that type-group is merged and used for generation. If that merged schema turns out to be unsatisfiable, generation falls back to a single branch from the same group so the result still satisfies at least one anyOf branch.

Limitations: the generator does not search across every possible branch combination. It stays within one chosen type-group per call, so it may miss a satisfiable cross-group combination even when one exists.

oneOf

Exactly one sub-schema must be satisfied. The generator picks one branch at random and generates from it.

{
  "oneOf": [
    {"type": "integer", "multipleOf": 3},
    {"type": "integer", "multipleOf": 5}
  ]
}

16476

Strategy: picks a preferred branch, generates a candidate, and validates it against every oneOf branch. If the candidate matches more than one branch, generation retries with other branches until it finds a value that satisfies exactly one sub-schema.

Limitations: exact-one semantics use bounded retry (max_search). Highly overlapping oneOf branches can still exhaust the search budget and raise NoExampleFoundError.

not

The generated value must not conform to the given schema.

Excluding a type:

{"not": {"type": "string"}}

834844010457502.2

Excluding a pattern:

{"type": "string", "not": {"pattern": "^[0-9]+$"}}

"Forward even deal evidence by maintain."

Strategy: when the not schema declares a type, the generator preferentially picks a different type — since a value of a different type is guaranteed to fail validation against the not schema, this succeeds without any retry. With small probability it attempts same-type generation (for variety when the not schema has narrow constraints like {"type": "integer", "minimum": 0, "maximum": 10}), falling back to a different type after a few quick attempts.

Limitations: {"not": {}} and {"not": true} forbid all values and raise UnsatisfiableConstraintsError. Typeless not schemas (e.g. {"not": {"minimum": 5}}) use rejection sampling since any type might match.