improve support for infinite domains

[daanhb]

trying out #106

[codecov[bot]]

Codecov Report

Merging #107 (dbcac16) into master (60bb050) will decrease coverage by 0.14%. The diff coverage is 86.48%.

@@            Coverage Diff             @@
##           master     #107      +/-   ##
==========================================
- Coverage   85.80%   85.65%   -0.15%     
==========================================
  Files          30       30              
  Lines        2479     2586     +107     
==========================================
+ Hits         2127     2215      +88     
- Misses        352      371      +19     

Impacted Files	Coverage Δ
src/domains/interval.jl	90.10% <86.39%> (-4.56%)	:arrow_down:
src/generic/lazy.jl	84.31% <100.00%> (+0.31%)	:arrow_up:

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

[daanhb]

I think these would be okay to have. I'm pretty sure I fixed some bugs along the way (set operations on fixed intervals). Two questions remain:

• how to name this. Right now HalfLine() and HalfLine{T}() is left-closed at 0 as before. There is no special syntax for HalfLine{T,:open}() or NegativeHalfLine{T,:closed}()
• should the half line be the canonical domain for all half-infinite intervals? That would make canonicaldomain type-unsafe for intervals, because the result would depends on the value of the endpoint. But right now the canonical domain for an infinite interval is [-1,1], which just doesn't work.

Perhaps this would make a 0.6 version.

[daanhb]

• should the half line be the canonical domain for all half-infinite intervals? That would make canonicaldomain type-unsafe for intervals

Turns out not to be true, mostly. Currently, a canonical domain [-1,1] was only defined for a closed interval. Infinite domains are at least open at the infinite end, hence they have a different type, hence adding a canonical domain for them did not introduce type unsafety. (It does compared to other open intervals which are bounded, for which the canonical domain is also (-1,1).)

[daanhb]

This will go into v0.6 instead, into dev-0.6 branch for now