Dec 11, 2009, 7:45:33 PM12/11/09
to Matthew Heberger, sf-bike-planner
Been looking at your code-- a few initial questions
1. How did you generate graph.pkl? Am I just not seeing the code or is it not checked in?
2. What is douglas_peucker_distance ?
I like that calculate the type of turn to take (right left) for each path segment. I think this is easier w/ sql representation of geometry, which is another advantage to doing the backend store that way.
I definitely think Python is the way to go as far as implementation language... it allows really fast iteration relative to a compiled language. If we don't want to port Matt's stuff, we could create Python bindings that wrap the C code. But I think long term it will aid development if we port some of it at least.
For a unified planner, one thing we'll need to do (I think) is write some code to determine if the input start/address both fall within the boundaries of one of our disjoint networks (e.g. all SF or all east bay). I honestly don't even know what my SF planner does if you put in start address in arkansas and end address in livermore! Do you guys handle this?