Wrong Expected Output for Duration w/ Different Precisions

[cmoesel]

The test CqlDateTimeOperatorsTest.Uncertainty tests.TimeDurationBetweenHourDiffPrecision2 specifies the following CQL:

hours between @T06 and @T07:00:00

and expected result 1.

Since duration counts "whole calendar periods", and @T06 represents some time from 6:00:00.0 to 6:59:59.999, this should expect the result Uncertainty[0, 1] (represented as Interval[0, 1] in this test framework).

[raleigh-g-thompson]

While it is not clear the statement hours between implies duration or difference the spec does clarify how to treat precision differences.

In addition, if either date or time value involved is not specified to the level of precision for the duration or difference being calculated, the result will be an uncertainty covering the range of possible values for the duration. Subsequent comparisons using this uncertainty may result in null rather than true or false. For a detailed discussion of the behavior of uncertainties, refer to the [Uncertainty](https://cql.hl7.org/05-languagesemantics.html#uncertainty) section.

Since the precision is hours and both values include at least hours precision there shouldn't be an uncertainty.

[cmoesel]

The phrase hours between indicates duration. For difference, you would have to say difference in hours between.

Note that the bit of spec that you quoted indicates one condition under which an uncertainty may occur, but it shouldn't be interpreted to mean that is the only condition under which an uncertainty may occur. That said, I can see how it is confusing, especially when you consider the next matter:

The CQL Reference (Appendix B) indicates that duration should count "whole calendar periods" and Appendix H defines duration in hours this way:

In CQL, an hour is defined as 60 minutes. The duration in hours between two dates is the number of minutes between the two dates, divided by 60. The result is truncated to the unit.

The examples in Appendix B also show that when calculating duration in hours, minutes matter:

Example 3:
Date 1: 2012-03-01 03:10
Date 2: 2012-03-01 04:00
Duration = 0 hours

The spec also indicates that when Dates, DateTimes, or Times are not fully specified, they represent an uncertainty over the range at the first unspecified precision. Appendix B provides this example:

For example, the datetime @2014-01-01T03 represents some instant during the hour of 3:00 on January 1st, 2014.

Taking all of that into account... In order to calculate the duration calculation hours between @T06 and @T07:00:00, we need to count the minutes, as described in Appendix H. According to how CQL defines times, @T06 represents an uncertainty over the range of @T06:00:00.0 to @T06:59:59.999. This means the number of minutes elapsed is the uncertain range of 0 to 60 minutes. Dividing by 60, as instructed in Appendix H, gives us an uncertain range of 0 to 1 hours.

[raleigh-g-thompson]

Agreed this is messy and needs clarification.

@brynrhodes What are your thoughts on this?

[brynrhodes]

Agree it should be Uncertainty[0, 1]

And yes, we are using Intervals in the tests to represent the uncertainty, but in each of those tests there is also a todo about how we actually represent uncertainty. The spec says it's only an internal engine thing, which engines would be expected to resolve at the point of actually returning a value to the calling environment.

I'll note the Java engine currently returns 1 for this test. @cmoesel what does the JS engine do?

[cmoesel]

The JS engine returns Uncertainty[0, 1]. I noticed the incorrect expectation here because we were failing the test.