Code review comment for lp:~zorba-coders/zorba/unordered-collection-tests

Hi Markos, I mostly agree with your analysis, just not quite on the conclusion. I think that the issue is precisely that it is implementation-dependent (and stable), but not implementation-defined, which means that there is no, and should not be any (documented) rule in the Sausalito store that says which one comes first. If we go in the direction of adding expected results, I am afraid we might need to add all of the possibly returned nodes.

When suggesting these changes, I had the feeling that specifying an id instead of a position did not decrease the relevancy of the test (which test, for instance, index probing), while keeping the test result deterministic in spite of the "unorderedness" of the collection.

I would be fine either way, there are just more changes involved with the first direction.

Does it make sense?