How to map new REV cycling paths in Montreal (and elsewhere in Canada)

Jarek · 2024-11-22T18:16:29.895Z

One more consideration: when a cycle track transitions into a lane.

When tagging on the highway, it’s a pretty simple change from cycleway=track to cycleway=lane, and it’s obvious to data users that we stay on the same street:

image1226×569 185 KB

On a separate cycleway you’d have to either merge back into the main highway, or do something like highway=cycleway + cycleway=lane.

The New Zealand mapper who changed Toronto did the former and it made routing pretty useless as the router listed every merge back and forth on this straight street as “turn slight right”: https://bin.piorkowski.ca/2024/2024-06-14%20bike%20track%20routing.png (perhaps made worse by giving the cycleway a different name than the street?)

On the other hand having a highway=cycleway + cycleway=lane doesn’t seem optimal either – it’s not really a separate cycleway if it’s just paint separation, right? It’s one thing to have a cycleway=crossing through intersections, but another in places where the bike track is downgraded to a lane for several blocks, but then resumes as a track later.

So perhaps on routes that somewhat frequently transition from track to lane we should encourage not mapping as separate way.

Jarek · 2024-11-22T18:25:12.370Z

ChaireMobiliteKaligrafy:

May be worth watching:

It appears there’s no transcript of this 30 minute video? That’s a shame. Do you know if there are any proposed solutions in the talk, is there maybe a blog post announcing and discussing them somewhere?

ChaireMobiliteKaligrafy · 2024-11-22T18:36:53.208Z

under the video there is a text summary

Jarek · 2024-11-22T18:41:59.474Z

Yes, I saw - it seems to be the talk pitch, which ends with:

“This talk will discuss these cases, and provide a starting point for discussion on what should be done to improve the situation. As people are ever keener to add more detail to the map, and as more and more mainstream users look to OSM, we have to ask whether the current model is arguably creaking too heavily. Is there a way that we can represent spaces as a set of interconnected flows in some way?”

So what was the discussion in the talk? Was there a new “way that we can represent spaces as a set of interconnected flows in some way” presented? Any potential solutions discussed, with pros and cons? Or was the talk only asking questions?

ChaireMobiliteKaligrafy · 2024-11-22T19:03:39.001Z

I’m not done yet listening to it, I will come back to you

hoserab · 2024-11-22T19:18:45.739Z

Jarek:

One more consideration: when a cycle track transitions into a lane.

When tagging on the highway, it’s a pretty simple change from cycleway=track to cycleway=lane, and it’s obvious to data users that we stay on the same street:

* [image clipped for brevity] *

On a separate cycleway you’d have to either merge back into the main highway, or do something like highway=cycleway + cycleway=lane.

The New Zealand mapper who changed Toronto did the former and it made routing pretty useless as the router listed every merge back and forth on this straight street as “turn slight right”: https://bin.piorkowski.ca/2024/2024-06-14%20bike%20track%20routing.png (perhaps made worse by giving the cycleway a different name than the street?)

Uh-oh… Changeset: 157775696 | OpenStreetMap

I for one like mapping them separately—likewise I map sidewalks separately—because of the aforementioned comparative ease in tagging them instead of adding a myriad tags to the highway. To do them “properly” also often means splitting the highway way into preposterously small segments, especially when factoring for other reasons why highways are split into small segments: for lanes, lanes:forward/backward, turn:lanes, turn restrictions, speed limits, surface, etc., etc.

And don’t even get me started about parking…

ChaireMobiliteKaligrafy · 2024-11-22T19:28:11.797Z

In the video: it is a discussion of the problems, but there is no solution provided: there may be no solution in fact except mapping all streets as areas and add relations but this would be too intensive for the available resources… So we need to decide between us

ChaireMobiliteKaligrafy · 2024-11-22T19:38:39.460Z

Here are some examples of how it is mapped right now (my work, but there are similar intersections mapped by others in Montreal):

image1920×1150 354 KB

image1920×1011 256 KB

skyper · 2024-11-22T21:55:41.215Z

In my eyes, we are well beyond the point that sidewalks and cycleways are tagged in details on the main road. I am fine with that as long as there is a physical separation like some kind of barrier or a grass strip. Though we should still connect the separate ways and the main road to each other with tags and we need to take care of each changing opportunity between the separate ways. E.g. if the curb is lowered at one place or there is an opening in the barrier we need to add a link. This gets tricky if the barrier has lots of openings or is a line of bollards.

Jarek:

Some of my thoughts:

Pros of tagging on the main highway:

It is clear what street/road the track is next to or on. This can help determine some properties of the bike paths, for example, even if there is physical separation, riding on a bike track next to a 60 km/h highway=primary is quite different from riding on a bike track next to a 30 km/h highway=residential. [1]

Sounds like is_sidepath=* and its subkeys.

Jarek:

Also, the street/road itself would also know that it’s next to a bike track, so car routers can give appropriate warnings or prefer to avoid it.

cycleway[:side]=separate and bicycle[:direction]=use_sidepath

Jarek:

In particular, name is always set, and routers can give directions like “take the cycle track on Bloor Street”. We can and should add name to highway=cycleways mapped separately, but in practice this hasn’t happened a lot, perhaps because it creates clutter on the carto map styles

Not a pretty tag but is_sidepath:of:name=* can be used instead.

Jarek:

Pros of mapping separately:

Geometry could be shown more accurately (though it sets up potential problems at transition to and from the separate mapping)

How about placement=transition? Routers can ignore the short links and calculate the angle between the way segments before and after the way with this tag.

Jarek:

A lot of mappers seem to like this style (perhaps so they can see the bicycle infrastructure shown on the default carto style?) so you won’t be potentially swimming against the stream and repeatedly having discussions about this. (In Toronto we had a mapper from New Zealand (?!) come in and map some tracks separate ways, with several issues , but if you set up the guidelines for mapping separately well, maybe that won’t happen as much.)

We have a guide for tagging sidewalks separately but still I often have to remind (new) users to tag it properly with footway=sidewalk, is_sidepath=yes plus change the value of sidewalk[:side]=* to separate on the main road (and add foot=use_sidepath in Germany). Not talking about lots of missing links between the sidewalk and the main road.

Jarek · 2024-11-23T01:48:45.075Z

ChaireMobiliteKaligrafy:

Here are some examples of how it is mapped right now (my work, but there are similar intersections mapped by others in Montreal):

Looks good!

Do you have any recommendations on tagging lower-quality infrastructure mixed separation, shifting from paint- to parking- to curb-delimited and back again? Or would you prefer that to be a separate discussion?

skyper:

This gets tricky if the barrier has lots of openings or is a line of bollards.

Which is the case in most of Toronto bike infra.

image1920×743 140 KB

image1920×546 94.6 KB

image1967×303 133 KB

Maybe elsewhere in Canada it’s better?

skyper:

Sounds like is_sidepath=* and its subkeys.

There are currently zero uses of is_sidepath:of:maxspeed but you could be the first one

skyper:

Not a pretty tag but is_sidepath:of:name=* can be used instead.

Which incidentally isn’t meaningfully supported by any data users, see Haben straßenbegleitende Rad- und Fußwege Namen? - #30 by Jarek

I would love to use it. Please get some data users to add support (25k current uses) then I will join in.

skyper:

How about placement=transition? Routers can ignore the short links and calculate the angle between the way segments before and after the way with this tag.

Do routers or renderers actually do this? The proposal has been around since 2012 - has anyone implemented using this?

Jarek · 2024-11-23T01:57:11.304Z

hoserab:

Uh-oh… Changeset: 157775696 | OpenStreetMap

At least you were mapping in your province, not across the globe!

Though to get from the Red Arrow stop to Celebration Square by bike you now continue on 9 Avenue SE, then turn right, then turn left, then turn left, then turn right, then turn left. Easy!

image348×538 25.1 KB

Compare with car directions:

hoserab:

To do them “properly” also often means splitting the highway way into preposterously small segments, especially when factoring for other reasons why highways are split into small segments: for lanes, lanes:forward/backward, turn:lanes, turn restrictions, speed limits, surface, etc., etc.

Honestly, I kind of agree, but I think that battle was lost at turn:lanes.

Minh_Nguyen · 2024-11-23T02:50:40.220Z

Jarek:

Honestly, I kind of agree, but I think that battle was lost at turn:lanes.

And then parking:lane=* happened…

KevinOs · 2024-11-23T05:13:29.514Z

I’m a big fan of mapping the cycling infrastructure out individually where possible. We have excellent mapping imagery resolutions now in most cities why not make the most of it? Additionally, cycling infrastructure is not entirely bound to roadways with paths and routes take departures through greeways, parks etc… There are specifics of crossings that the roadway tagging does not capture for pedestrians and cyclists that i feel should be there to help users. Finally its completely lost on new mappers that these road tags even exist, it’s easy to see and map a new way with what its function is.

Where i typically see the breakpoint from moving from road tags to dedicated mapping is going from “safer streets”, sharrows and painted lanes to a separated lane, cycle track or MUP.

This can confuse routers. What i would encourage you to do when mapping is think of the segment as a bus route to ensure there are no gaps in the lines. Ideally we are turning these into bike routes using the meta tagging. (there is a whole question i need to bring up to you all regarding naming schemas and route creating in the absence of formalized ones for cities without)

I also see the Montreal REV mappers don’t make use of the forward tag on their routes as well you might want to consider that to indicate directionality on your one-way cycle lanes routes. Here is an example from Edmonton i did a while back to help

OpenStreetMap - Edmonton 76/106 routes

anyway excited to see how others are settling in on cycling infrastructure mapping its a bit challenging to find gold standards to copy from as every city seems to have its own takes and quirks.

ChaireMobiliteKaligrafy · 2024-11-23T16:34:27.649Z

I added the placement=transition to all motorways in the Montreal area. I plan to use them in my research, but I am not aware of any tool that parse them correctly right now.

KevinOs · 2024-11-23T19:24:22.052Z

I watched the linked video… Am still equally confused on a path forward for the OSM people but I see now the inherent problems of dedicated mapping I kind of overlooked. Sidewalks and bike routes probably need some extra work from what I or other locals have done to help. I still think referring to entirely road based tagging is just accommodating a flawed OSM architecture and routing issues is a dead end.

Really the core issue is in testing routing and connectivity. Is there any way to test this efficiently before uploading? The wiki suggests using an export and OSMand but that is pretty convoluted too and reliant on manual checks.

Equally are there tools to flag routing issues on the existing map to investigate further?

It’s really hard to sort out why a certain router doesn’t do what a normal human would expect or gravitate towards safer options and I would love to know why.

tallcoleman · 2024-11-23T23:11:15.442Z

KevinOs:

Really the core issue is in testing routing and connectivity. Is there any way to test this efficiently before uploading?

I would also love to know if there are good solutions to this. In suburban contexts, there’s a lot of stuff that is perfectly navigable by bike in real life (e.g. going through one parking lot to another in a way that is not possible for a car) but the routers will put you on a 4-lane 60kph street instead.

Jarek · 2024-11-24T00:23:47.320Z

KevinOs:

Really the core issue is in testing routing and connectivity. Is there any way to test this efficiently before uploading?

Generally, my impression is that “testing locally” isn’t really feasible or usually done with OpenStreetMap these days. Exceptions could include big data users, but for individual editors, the usual way I’ve assumed is to try to get it as decent as possible, upload, then see how it works in the tools once they index in your change, and make adjustments if necessary. It’s a very iterative-improvement process, but I feel that works with the ethos of OSM in general.

ChaireMobiliteKaligrafy · 2024-11-24T15:29:38.499Z

I think we can wrap this by stating which method we prefer/suggest for each scenario. I’ll start with my suggestion, which is the method I use since 2020 I would say:

draw as a separate cycleway when: there is a clear separation (like a permanent parking lane, bollards, fence, or concrete curbs at intersections or along the whole way), or a clear change in surface (concrete surface instead or paving stones alongside an asphalt road)
add cycleway tags to the road itself when: there is only paint all the way (barred area and/or solid/dashed/dotted lane, green/blue paint with no physical separation, or flex posts which can be removed easily (usually in winter, and sometimes they put them back only late in the spring or in the summer), and no concrete/fence at intersection.

So as a summary:

draw as a separate cycleway when there is partial (only intersections or short segments) or complete PERMANENT/NOT SEASONAL physical separation
use cycleway tag on road otherwise (including flex posts)

Jarek · 2024-11-24T17:08:39.874Z

ChaireMobiliteKaligrafy:

draw as a separate cycleway when there is partial (only intersections or short segments) or complete PERMANENT/NOT SEASONAL physical separation

I would word this as:

draw as a separate cycleway when there is substantial (most of the route) or complete PERMANENT/NOT SEASONAL physical separation

Otherwise we could get into nonsense 5 metre long cycleways like this stuff in London (England):

image592×577 51 KB

I would also like to have an official recommendation/requirement to add the street name as the cycleway’s name as well. You do this in Montreal, but it’s not done in Toronto and I’d like to start seeing that.

ChaireMobiliteKaligrafy · 2024-11-24T18:07:57.894Z

Yes, merging/unmerging from separate cycleway to road cycleway tags should not be done unless there is at least 100m or so. Otherwise, keep the previous method. THat is why the REV should use the separate method all the way, to make sure there is no merging repetition between each intersection.