Wrong Expected Output for TestInNullBoundaries

[cmoesel]

The test CqlIntervalOperatorsTest.In.TestInNullBoundaries specifies the following CQL:

5 in Interval[null, null]

and expected output false.

The CQL specification for In, however, says:

For closed interval boundaries, if the interval boundary is null, the result of the boundary comparison is considered true.

Therefore, the result of each boundary comparison in the test CQL is true and the overall result should also be true.

[raleigh-g-thompson]

There does seem to be a contradiction in the spec. A couple lines down from the For closed interval boundaries... statement, the spec also declares

If the first argument is null, the result is null. If the second argument is null, the result is false.

Are these statements exclusive?

• [null, ANY NON-NULL VALUE] == null
  • but the statement doesn't say anything about the 2nd value
  • does this only apply if the 2nd value is not null
• [ANY NON-NULL VALUE, null] == false
  • but the statement doesn't say anything about the 1st value
  • does this only apply if the 1st value is not null

What if both are null does the original statement about null interval boundary apply?
Does this only apply for certain interval type, i.e. closed vs open, or a mix of both???

@brynrhodes Can you provide clarification?

[cmoesel]

@raleigh-g-thompson - Just to be clear, I think the references to first argument and second argument are referring to the signature of In (for which first argument is a point and _second argument+ is an interval), not the values inside the interval.

That said, you're right to ask what the result should be when the point argument and interval argument are null (e.g., null in null).

[brynrhodes]

To me, that language is clear: "If the first argument is null, the result is null."

So null in null should return null.

And yes, I agree, 5 in Interval[null, null] should be true because the null boundaries are inclusive so the interval is the range from the beginning of the domain to the end of the domain.

[raleigh-g-thompson]

@brynrhodes

Spec

The in operator (can also be invoked using during) for intervals returns true if the given point is equal to the starting or ending point of the interval, or greater than the starting point and less than the ending point. For open interval boundaries, exclusive comparison operators are used. For closed interval boundaries, if the interval boundary is null, the result of the boundary comparison is considered true.*1

If precision is specified and the point type is a Date, DateTime, or Time type, comparisons used in the operation are performed at the specified precision.

If the first argument is null, the result is null. If the second argument is null, the result is false. *2

From the Spec, Note 1 covers the claim 5 in Interval[null, null] should be true.

When is Note 2 applied? That seems in direct conflict with Note 1.

Is there somewhere else in the spec that clarifies how to apply these rules?

[cmoesel]

@raleigh-g-thompson - Can you further clarify how/why you think Note 2 is in direct conflict with Note 1? Note 1 is talking about a non-null interval that has null boundaries; Note 2 is talking about cases where the interval itself is null (which is different than null boundaries).

@brynrhodes - "If the first argument is null, the result is null" would be clear if it was the only rule stated about null arguments. But there is another rule in the same paragraph: "If the second argument is null, the result is false." Perhaps the order in which those rules are stated implies precedence (i.e., the second rule has an implied "else if"), but I don't think the spec states that explicitly. So it probably should be clarified.

[raleigh-g-thompson]

@cmoesel Yes I see now that I mis-read Note 2 and there is no conflict.

Thanks for pointing that out.