3 ms·
Thanks for the feedback. I'll definitely fix the pin accuracy and I'm thinking about what to do around street length & difficulty
by adamlynch 5y ago
Thanks for the feedback. I'll definitely fix the pin accuracy and I'm thinking about what to do around street length & difficulty
- dataflow 5y ago> I'm thinking about what to do around street length & difficulty Here's one idea if it might help you get started (though you can polish it): - Mark all "dead ends". - Mark all private streets. - Recursively (or perhaps I should say iteratively, depending on your POV) mark all roads as X that only lead to previously-marked roads. - Non-recursively, unmark each marked road that is both adjacent to an unmarked road AND that satisfies some reasonable heuristic (such as "being within 2 intersections of the center of your circle"). - Also unmark all freeways/highways/expressways/other reasonably-major roads (although few if any of them should be marked at this point) - Delete all marked roads from your list of candidates. This should ensure that every candidate road actually has a non-negligible chance of being at least seen by random drivers, and not just by residents who are forced to drive on it on the way home. You can obviously augment
- adamlynch 5y agoOK thanks but I do need to compute all of this. Plus some roads are broken into a lot of segments so I have to do more work first in finding & joining them up. Hopefully I could do all of it but just saying, it might be too much.