Creating invalid InclusiveInterval s

[stefnotch]

This crate requires InclusiveRanges to be valid when passing them to any of the functions, otherwise a panic will happen

	/// Is the range is valid, which according to this crate means `start()`
	/// <= `end()`
	fn is_valid(&self) -> bool
	where
		I: PointType,
	{
		self.start() <= self.end()
	}

Would it make sense to verify that in the constructor of InclusiveInterval? Then the implementation of InclusiveRange for InclusiveInterval could have is_valid be a no-op, and errors might be caught slightly sooner.

[ripytide]

	//only return valid range_bounds
	return CutResult {
		before_cut: result.before_cut.filter(|x| is_valid_range(*x)),
		inside_cut: result.inside_cut.filter(|x| is_valid_range(*x)),
		after_cut: result.after_cut.filter(|x| is_valid_range(*x)),
	};

In the implementation I sometimes create invalid range and then filter them out at the end as is makes some of the logic simpler but I would be fine with creating an unchecked variant of the constructor for my internal use. Catching invalid ranges earlier will be a great improvement, thanks.

[ripytide]

Hmm, I forgot but when going to implement this I remembered that if we ensure validity upon creation of a InclusiveInterval then we also have to make the start and end struct members private. This would mean you would no longer be able to write inclusive_interval.start and we would have to redirect it into a getter such as inclusive_interval.start(). This may still be an acceptable tradeoff, I'm not sure

[ripytide]

Hmm I think it would still be an acceptable tradeoff as I can still use the .start .end inside the implementation just fine and it would prevent all the panic sections on all the methods.

[ripytide]

Another caveat discovered upon implementation is that the panics on all the methods will still be necessary as on a custom range type the user can still create an invalid range. But I can continue the invariant constructors way on the InclusiveInterval struct.