Re: Another fix candidate for issue 1613. (issue4426072)

[mtsolo]

On 2011/05/01 02:44:31, hanwenn wrote:
> Can you document what this is doing?

> My idea was to have a datatype

> class RealSubset {
>    // Starts out as full interval
>    Real_subset();

>    // Remove an interval
>    remove(Interval v);

>    // Return sorted, non-overlapping list of allowed intervals
>    vector<Interval> allowed() const;

>    // implement with list or tree of non-overlapping intervals
> }


> You create such an object, remove the extents of all collisions (duly
> transformed to be corresponding to left beam edge).  Finally, pick the
> center of the most sensible interval from allowed(), one which is not
> too small and as close as possible to the original left Y.



Hey Han-Wen,

I put some more comments in a new patch in hopes that it is enough to
help you make an implementation of the type of thing you're talking
about.  Conversely, if you think that the patch in its current state is
admissible, let me know.

Cheers,
MS

http://codereview.appspot.com/4426072/