How can I check whether a JSON number is an integer or a decimal value?

After json.loads() or json.load(), check the Python value’s type. Python’s standard JSON decoder maps integer-form number tokens such as 7 to int, and real or fractional-form tokens such as 7.0 to float by default. [S1]

For an exact distinction, use type(value) is int and type(value) is float. This checks what type the decoder produced; it does not decide which type your application ought to accept. That rule belongs to the schema you choose.

Suppose, as an example schema, each record must have an integer count and a fractional-number ratio. The example below requires ratio to decode as a float; that is a schema choice, not a rule imposed by JSON. Python’s bool is an integer subtype, so isinstance(True, int) is true; the exact-type check avoids treating true as an ordinary integer. [S1]

import json

def valid_record(record):
    return (
        isinstance(record, dict)
        and type(record.get("count")) is int
        and type(record.get("ratio")) is float
    )

samples = [
    '{"count": 3, "ratio": 0.5}',
    '{"count": 3.0, "ratio": 0.5}',
    '{"count": true, "ratio": 0.5}',
    '{"count": 3, "ratio": 1}',
]

for text in samples:
    record = json.loads(text)
    print(record, valid_record(record))

The first sample is accepted. The second has a floating-point count, the third has a boolean count, and the fourth has an integer-form ratio; all three are rejected by this particular schema.

A missing field makes record.get(...) return None, which fails these exact-type checks. A string such as "3" also fails: this code validates the decoded type and deliberately does not convert text into a number. If your schema permits conversion, define that separately rather than silently accepting different input forms.

For clearer feedback, check fields one at a time and report the expected type. For example, you could distinguish “missing count” from “count must be an integer” by testing whether the key exists before checking its value. Decide whether additional fields are allowed too; the example schema above only checks the two named fields and does not reject extras.

No. The type check answers whether Python decoded a token as an integer or float; it does not establish your field rules, required fields, allowed ranges, or whether extra fields are acceptable. The Python documentation describes the decoder mappings, but it does not define your application’s schema. [S1]

Before relying on the check, write down each field’s expected type and whether it is required. Then try accepted and rejected examples that match those rules. In particular, test an integer token, a fractional token, true, a missing field, and a numeric string. Keep the sample schema explicit so another reader can see why each input passes or fails.