We have a problem with the linking of stops that are inside areas with long edges.
For example:
http://otpdebug-test.entur.org/?module=planner&fromPlace=60.287862%2C11.169424&toPlace=60.287821%2C11.169582&mode=WALK&date=05-30-2017
http://otpdebug-test.entur.org/?module=planner&fromPlace=59.62100307780759%2C11.079298853874207&toPlace=59.62092982995759%2C11.079471856355667&time=2%3A19pm&date=05-30-2017
In both examples edges are added to the areas through WalkableAreaBuilder with areaVisibility set to true. In most cases this works fine. However, in some cases stops are linked to the middle of a long edge, creating a new vertex in that edge. This happens in the GTFS module, after the visibility graph has been built. The result is that there is no direct route from the stop to the platform entrance. You have to walk along that long edge before you are able to take a more direct route.
It seems to me that the problem here is the order in which the graph is built. When the OSM module runs, it has no knowledge of the stops. And as far as I know, the information about whether or not the edge is a part of an area is lost by the time the stops are linked.
We have a problem with the linking of stops that are inside areas with long edges.
For example:
http://otpdebug-test.entur.org/?module=planner&fromPlace=60.287862%2C11.169424&toPlace=60.287821%2C11.169582&mode=WALK&date=05-30-2017
http://otpdebug-test.entur.org/?module=planner&fromPlace=59.62100307780759%2C11.079298853874207&toPlace=59.62092982995759%2C11.079471856355667&time=2%3A19pm&date=05-30-2017
In both examples edges are added to the areas through WalkableAreaBuilder with areaVisibility set to true. In most cases this works fine. However, in some cases stops are linked to the middle of a long edge, creating a new vertex in that edge. This happens in the GTFS module, after the visibility graph has been built. The result is that there is no direct route from the stop to the platform entrance. You have to walk along that long edge before you are able to take a more direct route.
It seems to me that the problem here is the order in which the graph is built. When the OSM module runs, it has no knowledge of the stops. And as far as I know, the information about whether or not the edge is a part of an area is lost by the time the stops are linked.