Getting all connectors/intersections for a distinct road

[Nolan Darilek (ndarilek)]

Hey folks, started playing with this data recently and had a few questions.

A bit of likely required context: I'm blind, so I literally can't see any visualizations and am working completely with numbers/theater of the mind. So apologies if these answers are obvious, or would be if I had working eyeballs. :) I'm currently building out my import/CLI tooling and it's very likely some or all of it has major issues I just don't know about.

I'm trying to identify intersection neighbors to the current intersection that are on a road. For my purposes, I define "road" as a segment likely named, likely to be traversed and oriented to. The goal is to describe someone's surroundings as them being on a specific known street rather than in some sidewalk/alley, at least for now. Ultimately I' trying to browse my map intersection-to-intersection.

Here's my Rust code for deciding that something is a "road," ultimately I want to pick out intersections only along these "roads," hopefully this makes it a bit clearer what the segment and ultimate intersection filter will look like.

    pub fn is_road(&self) -> bool { // `self` is guaranteed to contain segment data.
        if self.subtype == Some("road".to_string()) {
            if let Some(class) = &self.class {
                [
                    "motorway",
                    "primary",
                    "secondary",
                    "tertiary",
                    "residential",
                    "living_street",
                    "trunk",
                ]
                .contains(&class.as_str())
            } else {
                false
            }
        } else {
            false
        }
    }

The problem I'm running into is that, when I look at data for connectors/segments in my area, each segment appears to be a linestring from connector to connector. There are multiple points, it's just that I'd expect to see a longer segment with intersections along its length vs. just one short segment with intersections/connectors at each end. I don't know if this is right, or if I'm importing/processing data incorrectly. So I can browse between intersections like, say, streets intersecting driveways/walkways, but what I really need is some way to identify all segments as belonging to a specific "path", then I can iterate over all connectors along that super-segment and do the more detailed pattern-matching I need. So, specific questions:

1. Is my analysis correct--many streets are modelled as linestrings with intersections only at their endpoints? More specifically I'm looking at data around latitude 30.26488, longitude -97.746989, or connector ID 08f489e3460ed4f404755fde4614707e which, if I'm importing correctly, participates in 4 segments likely part of a larger street but looks like a point on much shorter lines.
2. Assuming so, is there some way other than name-matching to identify all segments as part of the same unique feature? I've checked the schema but nothing jumps out at me--at least not in the section on segments.

Thanks.

[Adam Lastowka (Rachmanin0xFF)]

Hi Nolan, maybe I can offer some insight; I'm an intern working on transportation at Overture.

This is an excellent question, and touches on something extensively discussed within Overture. The mereological problem you're facing is a consequence of Overture's segmentation: the way in which the real-world transportation network is partitioned into discrete elements. This is a slightly contentious topic, as the "best" segmentation depends on the the final use case.

At Overture, one of our segmentation rules is to split geometry on intersections between "major" roads, which loosely means "roads that cars drive on", and I think includes all of the road classes you listed. We're working on improving our docs for all this.

With respect to your specific points:

1. Your analysis is correct; the area you referred to consists of a lattice of primary and residential roads, both of which are considered "major" road types, and are consequently split at intersections.
2. Name matching is probably your best bet. Off the top of my head, I can think of a couple other options if that isn't good enough for your use case: first, you could try looking at the record_id property in sources, as the upstream OpenStreetMap data is sometimes segmented less-aggressively. The other would be some geometric heuristic (e.g. looking at the parallelness of segments attached to a connector). Also, this is probably way out-of-scope, but routing engines often generate these sorts of 'super-segments' (look up "contraction hierarchies").

Again, great question, and feel free to ask more!

[Nolan Darilek (ndarilek)]

Thanks, this is very helpful!

So it sounds like, for my purposes, it's likely sufficient to just traverse the entire segment and pick the relevant endpoint to find a major street. That works for me. Is there any documentation on which road subclasses serve as the division criteria? https://docs.overturemaps.org/schema/concepts/by-theme/transportation/roads/ suggests it's when a class transition occurs, so I guess what I'm looking for is which subclasses constitute a given segment class so I can align my definition of major roads with an Overture concept.

I'm probably munging my terms left and right here, so apologies if I'm getting it wrong, but hopefully it's clear what I'm asking for. :) thanks again.

[Adam Lastowka (Rachmanin0xFF)]

Hi, sorry for the delay! We're working on getting this in the docs. Currently, the subclasses we consider "minor" are:

• footway
• path
• cycleway
• steps
• pedestrian
• bridleway
• service
• track

If a major and a minor road intersect, we don't split apart the major road's geometry.