should we use == or === in tests?

[lucaferranti]

In the latest commit I switched to use === instead of ==, mainly because the comparison of decorated intervals ignores the decoration part. This helped revealing a couple of bugs that have been silent in IntervalArithmetic so far, e.g.

julia> DecoratedInterval(-Inf..Inf, com)
[-∞, ∞]_com

As a downside, since 0.0 === -0.0 evaluates to false, now a few more tests fail, e.g.

julia> @test +(interval(-0.0,2.0)) === Interval(0.0,2.0)
Test Failed at REPL[36]:1
  Expression: +(interval(-0.0, 2.0)) === Interval(0.0, 2.0)
   Evaluated: [-0, 2] === [0, 2]
ERROR: There was an error during testing

julia> @test +(interval(-0.0, 2.0)) == Interval(0.0, 2.0)
Test Passed

I am not sure what the standard says about +0.0 vs -0.0, so I am not sure what to think about this.

One alternative would be to modify the parser so that whenever it encounters an expression in the form something == DecoratedInterval(something) then it creates the test
something == DecoratedInterval(something) && decoration(something) == decoration(DecoratedInterval(something))

[Kolaru]

Among other things in JuliaIntervals/IntervalArithmetic.jl#271, I plan to introduce a new symbol for interval equality (X ≛ Y) to represent isEqual as defined in the standard.

This should solve the problem you have here as well.

[dpsanders]

I think -0.0 should be normalized to 0.0 when it's a lower bound, and maybe vice versa when it's an upper bound.

[lbenet]

Now that #500 is merged, can we check again if things are now working with ===?