[Xastir-dev] first cut at geocoder

[Jeff Barlow]

"Curt, WE7U" <archer at eskimo.com>  wrote:

>I think we'd need to know which points are connected to which, and
>the distance for each line segment.  Using that we could calculate
>the shortest path between points A and B.  Trying to do this
>on-the-fly from shapefiles would be rather slow.
>
>We'd also want to try to be smart about it and not head out too far
>in the wrong direction checking each and every path.  Perhaps that
>optimization would get us into trouble?
>
>First thoughts are that we'd have to have a separate database
>containing the graph points/distances.  From that we could more
>quickly compute the shortest path between any two points.
>
>Perhaps that database and the geocoding database could be combined
>into one.  Maybe the geocoding database has the info we need buried
>in it already?  I doubt it though.

Garmin's MetroGuide does not do a simple shortest path thing. It seems
to try to put you on main roads for most of the trip. It will take you a
little ways the 'wrong' way to get you on a freeway. Their database must
contain that info. 

I suspect the best approach to this sort of thing involves very careful
design of the structure of the data files. One needs to be able to ask
questions like: "Where's the nearest main road?" and not have to do a 15
minute file search to get the answer.

Jeff Barlow
WB6CSV