Added text-repeat-distance for highway names

[Penegal]

This PR adds text-repeat-distance for highways, which currently unnecessarily clutter the map on dense areas. It uses a smaller value of 50 for highway=unclassified and lower-importance highways, and 200 for higher-importance highways. This also achieves the split of #3295 in smaller PR as requested by @kocio-pl, and consequently closes #3111.

Before/after examples:


https://www.openstreetmap.org/#map=17/48.66397/6.17061

https://www.openstreetmap.org/#map=17/48.68110/6.14494

https://www.openstreetmap.org/#map=17/48.69825/6.18554

https://www.openstreetmap.org/#map=16/48.1898/6.4266

Edit: added links to test locations.

[Penegal]

@kocio-pl: this is the last version I did for #3295; it comprises the value of 50 you asked for residential/unclassified roads, and the decrease of text-repeat-distance to 200 for more important highways.

These examples make me think that the current value of 200 is still too many repetitions of the name for important highways, as they change name or direction much less often than residential/unclassified ones; IMHO, 200 still displays them much more than is necessary for following such highways on the map. As you recommended 200, what do you think about this suggested increase?

[kocio-pl]

You need to fix the roads.mss conflict now.

[kocio-pl]

I did not ask for any change for residential roads - they look worse (incomplete) with this code:

Example place

Before

[Penegal]

Ok, I'll fix the residential roads with the conflict; what about the 200 value for more important roads?

[Penegal]

Reverted the highway=residential code to the previous version.

[kocio-pl]

I will review this. What value would you propose for major roads?

I think every road type change needs separate consideration, at least one test rendering, because you never know and roads are important element of the map.

[Penegal]

I'll do some tests to find a correct value. At first glance, I would say 600 as I remember I tested this value, but, since then, you pointed some elements to care about that I did not care about back then, so I'll see what the tests suggest.

BTW, good remark about testing the rendering of each road type; I would probably not have tested that.

[kocio-pl]

Great!

Good testing pattern is trying not only the value you think should be right, but also something "too low" and "too high". This way I was able to detect "switch" behavior in the previous code (every global value had exactly the same effect), it also hints if bigger or smaller value might be better than first guess.

[Penegal]

OK, this version could do it: I played with examples to find good values. It does not significantly alter highway=residential rendering, while still dropping the most obvious duplicates. Here comes before/after comparisons, with URLs and, if relevant, notes on the displayed items:

https://www.openstreetmap.org/#map=18/48.05304/6.94675

https://www.openstreetmap.org/#map=18/48.05304/6.94675 (highway=pedestrian)

https://www.openstreetmap.org/#map=17/48.18681/6.39958

https://www.openstreetmap.org/#map=16/48.1936/6.4205

https://www.openstreetmap.org/#map=17/48.30746/6.02281

https://www.openstreetmap.org/#map=19/49.35006/6.12876

https://www.openstreetmap.org/#map=17/48.51451/6.64924

https://www.openstreetmap.org/#map=18/48.65392/6.13826 (these are highway=unclassified, not highway=residential)

https://www.openstreetmap.org/#map=18/48.66386/6.17092 (along the Boulevard de l’Europe, highway=living_street)

https://www.openstreetmap.org/#map=18/48.68220/6.17323

https://www.openstreetmap.org/#map=17/48.68156/6.21596

https://www.openstreetmap.org/#map=18/48.69006/6.17514

https://www.openstreetmap.org/#map=19/48.69787/6.15558

https://www.openstreetmap.org/#map=15/48.8924/6.1029

[kocio-pl]

Thanks, it will take me some time to review and test and I'm not sure if this PR should be maybe cut into separate PRs to make discussion more focused and easier to follow. What do you think about it?

[Penegal]

I currently use one PR because the current code uses only 2 different values for text-repeat-distance; if the number of different values increases, or if there are still numerous issues, I'd fragment it, but I don't think it would be relevant if there are only one or two issues left. Are you OK with that?

[kocio-pl]

OK, we might try testing case by case in this PR.

[Penegal]

@kocio-pl: were you able to test this?

[kocio-pl]

I plan to test it soon, however piece by piece, not all at once.

[matthijsmelissen]

@kocio-pl What is the status of this PR? If I read it correctly, it is still in your hands?

[kocio-pl]

Yes, it's just tedious to check and there was explosion of other PRs.

[kocio-pl]

I think these repeat distance values (500 for major and 100 for minor roads) are way too big, I would go with something 10x smaller. It's enough to avoid 2 carriage roads having the same labels in the same place, but nothing more. For smaller ones I don't like to make any special changes, they are basically OK for me.

Example of major roads - 2-carriage and multiple carriage way (extreme case)

current code

distance 500

distance 50

Example of minor roads - multiple carriage way (extreme case)

current code

distance 100

distance 10

[kocio-pl]

Example of minor roads - 2 carriage way

current code

distance 100

distance 10

[kocio-pl]

Example of a more typical major (2 carriage) and minor (1 carriage, small segments) roads

current code

distances: major 500 + minor 100

distances: major 50 + minor 10

[kocio-pl]

@Penegal Could you update this PR? I would make some more tests to check if nothing is broken and if it will pass, I would merge it soon.

[kocio-pl]

I think that text-repeat-distance for living street should be in separate line, so the brackets should create a small inner block.

[kocio-pl]

Could you fix this remaining issue? Than it would ready to merge.

[Penegal]

@kocio-pl : done.

[Penegal]

@kocio-pl : the code should now be as you asked; warn me if there are some problems left.

[kocio-pl]

It looks good now, thanks a lot! I hope this will make street experience more pleasant - at least I'm happy with the output I tested.

[kocio-pl]

Related to #497.

[geostonemarten]

Is it possible to ignore the label creation for footway if it contain footway=sidewalk because there is also the same name on major highway.

https://www.openstreetmap.org/#map=17/50.52778/1.58588

[SomeoneElseOSM]

Is it possible to ignore the label creation for footway if it contain footway=sidewalk because there is also the same name on major highway.

That sounds like a question unrelated to this pull request, but something that achieves a similar effect is technically possible. Example from a related style that uses similar software components: https://github.com/SomeoneElseOSM/SomeoneElse-style/blob/master/style.lua#L974 .