Oh, yes, that's a good point - with manhattan distance, if a point's
region reaches the border, nothing can stop it from reaching infinity.

So my code could be simplified slightly, since I don't need that extra
row of padding.

(I am also wondering if I am guaranteed to have a conflicted region on
the border. I am pretty sure that I am guaranteed that, but I haven't
proven it to myself, and it was only a couple of characters to let me
ignore that issue.)

Thanks,

-- 
Raul
On Wed, Dec 12, 2018 at 3:51 PM Jimmy Gauvin <[email protected]> wrote:
>
> Hi,
>
> I went with the assumption that inner regions are characterized by an
> absence of points on the border of the spanning box.
> A region that has a point on the border can "escape" out of the box.
> This might not be true for Euclidian distance but seems to hold for
> Manhattan distance.
>
> On Wed, Dec 12, 2018 at 1:40 PM Raul Miller <[email protected]> wrote:
>
> > On Wed, Dec 12, 2018 at 1:25 PM 'Mike Day' via Programming
> > <[email protected]> wrote:
> > > It occurred to me yesterday, though I haven’t coded it, that you can
> > work through twice, first for the minimum suitable spanning box, and second
> > for that box extended one row up and down and one column left and right.
> > The inner regions are those whose size doesn’t change. Probably pretty
> > inefficient, but easier to understand?
> >
> > What I did was ensure my grid had at least one row of empty grid cells
> > on the border.
> >
> > I am not sure that that's sufficient. Intuitively, I would think the
> > grid should be twice times the size necessary to contain the named
> > coordinates for the worst cases (with the named coordinates in the
> > central part), but it worked for every example I've encountered, so
> > far.  (On the other hand, if that single extra grid coordinate is
> > sufficient, for the worst case, it might be that I don't even need
> > that single extra row.)
> >
> > "Fortunately", ... Advent of Code doesn't require you solve for the
> > general case.
> >
> > FYI,
> >
> > --
> > Raul
> > ----------------------------------------------------------------------
> > For information about J forums see http://www.jsoftware.com/forums.htm
> ----------------------------------------------------------------------
> For information about J forums see http://www.jsoftware.com/forums.htm
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

Reply via email to