[Code scan] JSON Schema uses array items for dict-style repeated arguments

[njzjz]

This issue was found by a Codex global code scan of the repository.

Affected code:

dargs/dargs/json_schema.py

Lines 116 to 128 in b4db564

	if not argument.repeat:
	data["properties"] = properties
	data["required"] = required
	if allof:
	data["allOf"] = allof
	else:
	data["items"] = {
	"type": "object",
	"properties": properties,
	"required": required,
	}
	if allof:
	data["items"]["allOf"] = allof

Problem:
For repeat=True, schema generation always puts subfield validation under items. That works for list-style repeated arguments, but for dict-style repeated arguments the top-level type is "object"; JSON Schema ignores items on objects.

Reproducer:

from dargs import Argument
from dargs.json_schema import generate_json_schema
from jsonschema import validate

arg = Argument("base", dict, [Argument("sub1", int), Argument("sub2", str)], repeat=True)
schema = generate_json_schema(arg)
validate({"item1": {"sub1": 1, "sub2": None}}, schema)
validate({"item1": {"sub1": 1}}, schema)
print(schema)

Observed behavior:
Both invalid objects validate because the generated schema has type: ["object"] plus items, and the item schema is ignored.

Expected behavior:
Dict-style repeated arguments should use object-key validation, for example additionalProperties or patternProperties, so every repeated entry is checked against the subfield schema.