after 3 weeks of perserverance, i am victorious!
the biggest problem i was having was with the v.clean portion to tie together the connections with the streets. I used GRASS 6.0.2.
- i took buildings (points) and used v.distance to the streets (lines) to create connectors (lines)
- added category values to connectors, (otherwise v.patch would put them on layer ZERO.
- used v.patch to patch buildings, connectors, and streets together.
- ran v.clean - this was the biggest problem. I was expecting the connectors to break the line where it “touches” it. it would not break the line. it seems that the connectors would only snap to nearby nodes or vertexes (cant recall if it did snap to verticies). so i had to go through 208 points and verify that the line had a node where the connector was meeting the road. i used the split tool in QGIS editing the GRASS vectors for each one.
- ran v.clean again to snap to the manually created nodes.
- ran v.net.steiner.
depending on how long it takes on one’s machine to compute the steiner tree, i highly recommend verifying all points manually that they are snapped to streets. using QGIS i was able to do this relatively quickly, by looking for red X’s that indicated no connection.
i had to strip down the streets because 28K of them and 208 buildings took 8 days till i bailed. so i cleaned out roads that i knew it would not route along (mostly dead ends). after doing this, it ran with about 9K street lines, and took about 2 days. the biggest hangup was v.clean not breaking lines where connectors met. i assumed it was doing this but learned the hard way it did not.
one solution i forsee is to use v.split in 6.1, which can add (unnecessary) verticies every X distance. I believe this would help the connectors find vertex points to snap too. if it doesnt snap to vertexes, then it may not help.
if i can help anyone with this process feel free to ask!
Great work from the GRASS developers on this one. A network engineer/planner is reviewing the results and is impressed