Fix cql engine test cases--arithmetic and cql types

[leima888]

No description provided.

[brynrhodes]

The unit calculations seem to be looking for and returning 1 as the unit and erroring otherwise, why would the calculation convert the unit to 1? Shouldn't the units be preserved?

[leima888]

The unit calculations seem to be looking for and returning 1 as the unit and erroring otherwise, why would the calculation convert the unit to 1? Shouldn't the units be preserved?

Which unit calculation are you referring to? For division if the dividend and divisor have the same unit the result should have the unit '1' (no unit) otherwise calculate the unit such as 'cm3' / 'cm2' = 'cm' while for multiplication the unit calculation goes like 'cm'*'cm' = 'cm2'. It should throw an error if we have 20 '1' / 5 'cm' but 20 'cm' / 5 '1' should result in 4 'cm'.
Please advice if there're issues in these calculations.

[JPercival]

This PR is largely the same as #570. Closing this one in favor of that one.

As a note to contributors, the source of the xml tests in the cql-engine repo is actually the CQL specification IG here:

https://github.com/HL7/cql/tree/master/tests/cql

If there are tests that are incorrectly specified they should be fixed in the CQL spec (i.e. edit the XML there) and we'll propagate that to the cql-engine repo. I looked through the tests and the only one I'm certain that's wrong is TruncatedDivide10d1ByNeg3D1Quantity. I added an issue to track that here: HL7/cql#77

It doesn't look to me like the other changes are correct. Those appear to me to be bugs in the cql-engine, and therefore there should be cql-engine changes to fix that as opposed to bugs in the tests.This PR duplicates effort in #560, not sure if that's intended or not.

The source of the xml tests in the cql-engine repo is actually the CQL specification IG here:

https://github.com/HL7/cql/tree/master/tests/cql

If there are tests that are incorrectly specified they should be fixed in the CQL spec (i.e. edit the XML there) and we'll propagate that to the cql-engine repo. I looked through the tests and the only one I'm certain that's wrong is TruncatedDivide10d1ByNeg3D1Quantity. I added an issue to track that here: HL7/cql#77

It doesn't look to me like the other changes are correct. Those appear to me to be bugs in the cql-engine, and therefore there should be cql-engine changes to fix that as opposed to bugs in the tests.