Add boundary cases for integers #774
Conversation
There was a problem hiding this comment.
The "small and large float are integers" tests are not valid. You cannot make any assumptions about the architecture the tests are running on, and the expectation is that all integers, no matter the size, no matter whether they fit into the native integer size for your architecture, will still be treated as integers by JSON Schema.
tests/draft-next/type.json
Outdated
| "description": "a small float whose non-zero fractional part is truncated in a binary64 float is an integer", | ||
| "data": 1.00000000000001, | ||
| "valid": true | ||
| }, | ||
| { | ||
| "description": "a large float whose non-zero fractional part is truncated in a binary64 float is an integer", | ||
| "data": 9007199254740991.5, | ||
| "valid": true | ||
| }, |
There was a problem hiding this comment.
@karenetheridge is correct. These are not integers. JSON Schema uses an object model based on the one defined by the JSON specification, RFC 8259, which does not put limits on numeric values. Range does not matter. See Core 6.2 and Validation 4.2.
We do already have some tests for extremely large numbers. They are in the optional folders as we recognize that they typically are not fully supported and are generally edge-case. An argument could be made that they should not be optional.
hudlow
force-pushed
the
integer-boundary-cases
branch
from
6affa5c to
d319467
Compare
hudlow
force-pushed
the
integer-boundary-cases
branch
from
d319467 to
fe9a287
Compare
hudlow
force-pushed
the
integer-boundary-cases
branch
2 times, most recently
from
5f6a91d to
cca4442
Compare
hudlow
force-pushed
the
integer-boundary-cases
branch
from
cca4442 to
1f81212
Compare
I certainly don't like all of these behaviors,
but if we want truly consistent behavior across languages, we'll have to match the least-common-denominator behavior for JSON parsers (particularly ECMAScript's.JSON.parse())