due to out-of-order (?!?) nodes by the Blackfriars bridge. Somehow the
.osm file included https://www.openstreetmap.org/node/2934336887, which
has been deleted for over a year. Not sure how this wound up in the
geofabrik extract in the first place, but grabbing fresh seems to fix.
This code stopped working around July 2020 when I attempted to tackle #190. It's sat dormant since then, with most bus and light rail routes not imported correctly at all. I'm going to (eventually) start another attempt at public transit in A/B Street by treating GTFS as the main source of truth, not trying to understand route relations mapped in OSM. It's simplest to just rip out all this old code first. Some of it may be useful later, but version control preserves it.
Regenerating everything; this is a binary format change.
* add turn_on_red config option which is false for nyc
* when making stages, don't add movements that don't share a stage with a protected movement
* Fix a renamed DirectedRoadID field from
cb3693bd7a and the new turn_on_red field
in the test crate
* Update all map config files with the new option
for x in importer/config/*/*/cfg.json; do
cat $x | jq '.map_config += {turn_on_red: true}' > tmp
mv -f tmp $x
done
(And then undoing NYC)
* turn on red banned on all non-us maps
* regenerate all maps
Co-authored-by: Marcel Dejean <marcel@dejean.nyc>
Co-authored-by: Dustin Carlino <dabreegster@gmail.com>
- Use the new filter_crosswalks option to downgrade zebra crossings into
unmarked
- Add a routing penalty
- Render differently than zebras
- They otherwise act the same as marked crosswalks
Simulation behavior (like pedestrians not having priority over vehicles)
not yet implemented. No change to any existing maps, since
filter_crosswalks isn't enabled anywhere.
Accomplishing this through RoutingParams, rather than actually editing
the map. I don't want to commit to a map model representation of modal
filters and actually edit the map until we think through diagonal
filters. This effectively revives pathfind_avoiding_roads from the old
congestion capping experiment, but organized better.
It's extremely unergonomic to do this. Issues:
1) Hard to keep track of which Results include good context and which
don't. Make a new type?
2) Lack of try blocks forcing use of closures
3) async try blocks aren't stabilized
