Duplicated Node IDs in annotation

[drogers0]

I am running into some strange behavior regarding nodes that I can't find a great explanation for.
When I query with annotations on, a lot of times I will get duplicated node ids with geometries that don't match each other.

For example when querying ...driving/...?annotations=true&geometries=geojson&overview=full

I get 9/19 unique nodes: [A, A, B, A, A, C, A, A, A, D, A, A, E, F, F, G, F, H, I] and 19/19 unique Geometries

{
  "code": "Ok",
  "routes": [
    {
      "legs": [
        {
          "distance": 956.4,
          "annotation": {
            "metadata": {
              "datasource_names": [
                "lua profile"
              ]
            },
            "datasources": [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
            "weight": [5, 3.2, 4.6, 2.9, 2.2, 3.9, 4.4, 3, 2.4, 4, 11.7, 3.2, 1.8, 2.5, 1.5, 2.4, 2.9, 0.7],
            "nodes": [11074276640, 11074276640, 202132925, 11074276640, 11074276640, 202169303, 11074276640, 11074276640, 11074276640, 202169299, 11074276640, 11074276640, 202169298, 11074276630, 11074276630, 202169294, 11074276630, 202134268, 202169290],
            "distance": [77.46360473, 49.52062204, 70.81260108, 44.99539377, 34.10624539, 60.29819477, 66.7624365, 45.62389966, 36.66196298, 61.52715243, 178.6535481, 49.12974121, 27.07614971, 37.80968054, 22.8711949, 37.24471037, 43.80780856, 12.02025699],
            "duration": [5, 3.2, 4.6, 2.9, 2.2, 3.9, 4.4, 3, 2.4, 4, 11.7, 3.2, 1.8, 2.5, 1.5, 2.4, 2.9, 0.7],
            "speed": [15.5, 15.5, 15.4, 15.5, 15.5, 15.5, 15.2, 15.2, 15.3, 15.4, 15.3, 15.4, 15, 15.1, 15.2, 15.5, 15.1, 17.2]
          },
          "duration": 62.3,
          "summary": "",
          "weight": 62.3,
          "steps": []
        }
      ],
      "weight_name": "routability",
      "geometry": {
        "coordinates": [
          [-89.22045, 35.809735],
          [-89.219593, 35.809718],
          [-89.219045, 35.809722],
          [-89.218262, 35.809748],
          [-89.217769, 35.809805],
          [-89.217396, 35.809852],
          [-89.216749, 35.809985],
          [-89.216041, 35.810157],
          [-89.215569, 35.810303],
          [-89.215196, 35.810433],
          [-89.214577, 35.810664],
          [-89.212793, 35.811358],
          [-89.212292, 35.81153],
          [-89.212013, 35.811619],
          [-89.211613, 35.811719],
          [-89.211366, 35.811764],
          [-89.210959, 35.811817],
          [-89.210476, 35.811851],
          [-89.210343, 35.811849]
        ],
        "type": "LineString"
      },
      "weight": 62.3,
      "duration": 62.3,
      "distance": 956.4
    }
  ],
  "waypoints": [
    {
      "hint": "QqCsgp7PrIIyAAAAXQAAACAAAADMAAAAXu2aQta7DEMeFUZC4smbQzIAAABdAAAAIAAAAMwAAADXcAAAnpqu-sdpIgKbmq76UmkiAgIA_xAAAAAA",
      "location": [-89.22045, 35.809735],
      "name": "State Route 88",
      "distance": 12.98465404
    },
    {
      "hint": "VaKsgpjPrIIHAAAABgAAAAAAAAAAAAAA-VJAQVIiAkEAAAAAAAAAAAcAAAAGAAAAAAAAAAAAAADXcAAAGcKu-glyIgIZwq7693EiAgAArwUAAAAA",
      "location": [-89.210343, 35.811849],
      "name": "State Route 88",
      "distance": 1.997203679
    }
  ]
}

Plotting and color coding for node id I get a strange looking result that shows the same node appearing at different locations sometimes disjoint from one another. One of these locations does match the location associated with the node at openstreetmap.org but the rest are a mystery to me.

Someone who knows how the annotations work, please help me out!

[shinomontaz]

Hi!

The short answer as I understand is:
Node sequence is about the routing edges traversed, not the explicit sequence of visible points on the drawn route.

Why?

What I've learned on osrm prepare procedure is:

• Node annotations are segmen boundaries:
  Each entry in annotation.nodes marks the start (or end) of a segment. If a route passes through the same intersection (OSM node) multiple times, that node will appear each time it is crossed, resulting in duplicates.

• Segment geometry does not equals node:
  Each geometry segment (the line between two coordinates in the main geometry.coordinates array) may correspond to a unique path between two nodes, but the node ID alone doesn't define the geometry—it defines the endpoint of that segment for routing purposes.

• In case of same node it can be different context:
  If the route leaves and returns to the same node (e.g., a loop, or a "U-turn"), the same node ID will appear with different preceding and following geometries/distances.

• Internal edge spliting:
  Sometimes, OSRM splits ways/edges internally (e.g., due to restrictions, turn costs, or data interpolation), causing the same node to appear as a break-point for multiple short edges even if there's no visible "real" difference in the map.

Plotting and color coding for node id I get a strange looking result that shows the same node appearing at different locations sometimes disjoint from one another

I found that the order of nodes can be weird to see, but as we said before, geometry is not a equals to node list. It instead lists the OSM node IDs at the boundaries of each route segment OSRM calculates, often based on the OSM graph, not on the polyline geometry.
These segments are based on the routing graph, which can have different splits than the visible geometry due to restrictions, turn costs, etc.

It's possible (and common) for the same node to appear multiple times, especially if the algorithm splits the way at that node for different logical graph edges (e.g., U-turns, intersections, or way membership).

In your case, i found only one position, that really can looks weird - a presence of 202132925 node at the beginnig of node list after 11074276640.
The explanation for 202132925 I see is that this node relates to 2 "relations" (lines) and lays on intersection, so OSRM may "jump" between ways at nodes that are part of multiple ways (e.g., 202132925 is in both State Route 88 and Hillcrest Street), and OSRM may use this as a segment boundary due to an internal decision about route graph traversal (e.g., even if you don’t visibly turn onto Hillcrest Street, the graph edge may split there).

For 11074276640, that appears so many times it seems, that it is due to virtual nodes and segments that OSRM created on preprocessing stage and the 11074276640 used as some mark or segment boundary on this virtual nodes/segments ( i've mention this process early in internal edge splitting ).

Seems like that.