Revert dotdot

[dpsanders]

• Change definition of .. back to give tight intervals.
• Add convert(Interval, s::String)

[codecov-io]

Codecov Report

Merging #97 into master will decrease coverage by 0.01%.
The diff coverage is 80%.

@@            Coverage Diff             @@
##           master      #97      +/-   ##
==========================================
- Coverage   91.95%   91.93%   -0.02%     
==========================================
  Files          21       21              
  Lines         969      967       -2     
==========================================
- Hits          891      889       -2     
  Misses         78       78

Impacted Files	Coverage Δ
src/intervals/conversion.jl	76.19% <0%> (+1.19%)	⬆️
src/intervals/intervals.jl	85.29% <100%> (-1.2%)	⬇️
src/parsing.jl	96.96% <100%> (ø)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 1b59cc6...e5da6a1. Read the comment docs.

[lbenet]

LGTM. Let's wait for travis....

[lbenet]

Perhaps, we need interval here.

[coveralls]

Coverage decreased (-0.02%) to 91.934% when pulling e5da6a1 on revert_dotdot into 1b59cc6 on master.

[lbenet]

Some tests need revision...

[lbenet]

Three tests are failing, related to the mid-point radius notation (using ±). The first two are easily solved by:

    @test parse(Interval{Float64}, "3 ±  4") == Interval(-1, 7)
    @test parse(Interval{Float64}, "0.2 ± 0.1") == @interval(0.1,0.3)

which is intuitively what they should do.

The last one is trickier and I think some changes in the code itself are needed, perhaps here.

In my opinion, ± should behave (for BigFloats) just as @biginterval does, so that test would correspond to

@test parse(Interval{BigFloat}, "0.2 ± 0.1") == @biginterval(0.1,0.3)

Yet, this test also fails, because a (and b) is parsed as a BigFloat. In this case, (a-b)..(a+b) will not contain the true lower bound due to rounding issues.

[dpsanders]

I think the latest changes fix this.

[lbenet]

Now, other tests are failing: this one and this one, at least...

[lbenet]

Some tests still fail. The failure, as far as I have checked locally, is only related to the display of intervals (i.e., display_tests/display.jl). Hopefully, some lefts-over only...

[lbenet]

Locally, everything seems to work fine!

I propose to wait for travis, merge and tag a new version! The question is if it should be a patch or a minor version?

[dpsanders]

Since we changed the meaning of .. I think it should be a minor version.

[lbenet]

Excellent!

[dpsanders]

The string parsing code in general needs an overhaul.
This should be done by checking what the basic standard and the full standard require and implementing exactly that.

E.g. they require being able to parse 2/3 inside an interval string. Currently this is done by using Julia's parse, but we should avoid that in the future and do it ourselves.

I have a candidate in mind to do all of this work...

[lbenet]

I agree with you. But, let's leave it for another PR! The first require to make JuliaIntervals/IntervalRootFinding.jl#24 work is tag a new release here!

[dpsanders]

Tests are passing on Travis! Though the doc generation seems to be broken... :/

[lbenet]

Thanks a lot!