Re: [glob2-devel] idea: make existing features work well before adding new features (was: Nicowar Update!)

[Kai Antweiler]

> >> 1. Nicowar could really benefit from better handling of resource
> >>    overgrowth that cuts off individual buildings (and sometimes traps
> >>    individual globules) or which divides the main base into two
> >>    portions.  I would suggest computing some kind of minimal spanning
> >>    tree that connects all of Nicowar's buildings and placing clearing
> >>    areas (and maybe also clearing flags of radius 0 and 1) on this
> >>    spanning tree.
> >
> > Instead of clearing areas I suggest to use leos pavements, but I think
> > clearing areas are fine too.
> > We could try to implement a greedy algorithm that - given two points
> > on the map - tells
> > the globs which field to transform to a pavement next.  This could be
> > used for real
> > players too.  In the heuristics we would have to decide whether we only care 
> for
> > noswim globs, or whether shorter distances are more important than
> > fewer pavement.
>
> My personal preference is to make things that are already in the game
> work well, rather than adding new features (like paving areas).  Glob2
> already has plenty of features; it just needs for them to work well.

I think in most cases it is best to solve a task in a abstract way.  And then
assign some glob2 structure that fits the job best.  Instead of
planting clearing
areas, we plant path-pieces.  The algorithm does not need to know whether they
are clearing areas or pavements or clearing flags.  For the success of
the algorithm,
it only needs to know whether this structure can be used for a path.

I think in this case it is not necessary, because clearing areas would
become obsolete
by pavements.  But generally I think it is best to be abstract.
Otherwise you end up not
being able to adjust something, because it already has a meaning that
it shouldn't have.

--
Kai Antweiler