Rewrite road layer

[1ec5]

This is a full rewrite of the road layer code based on aaroads-wiki#2.

Tasks:

• Merge and rebase changes to roads layers from AARoads fork
• Restore road widths and colors based on functional classification
• Restore space-filling streets at high zoom levels
• Restore casing on freeways
• Check whether Unnecessary components to road line-sort-key #779 is fixed
• Eliminate bridge casing and always sort filled road lines over stroked road lines at lower zoom levels (possible tail work)
• Experiment with single offset lines for freeway and expressway casing instead of line gap (possible tail work)

Fixes #746, fixes #992, fixes #993.

[ZeLonewolf]

How are you addressing #746 (exits to nowhere)?

[1ec5]

How are you addressing #746 (exits to nowhere)?

This design reduces freeway ramps to thin black strokes, even when they connect freeways to freeways, but tries to keep them visible at any zoom level where freeways are visible (OpenMapTiles is a limiting factor here). See the before-and-after screenshots in aaroads-wiki#2 (comment) aaroads-wiki#2 (comment) where I first implemented this approach, and also this discussion in Slack. I think there will be more continuity between freeways and their ramps once we restore the black casing to freeways.

The treatment of freeway-to-freeway ramps may come as a surprise to anyone who’s been focused on highway connectivity as part of the U.S. mapping community’s reclassification effort, with its emphasis on identifying a coherent, well-connected network of freeways and trunk roads. However, the effect of “pinching” the lines at major freeway interchanges has plenty of precedent among American maps at scales where interchanges are more than mere diamonds but roads are still stroked. At least this treatment is more prevalent on these maps than the “exits to nowhere” that are typical of OSM-based maps.

Some maps do keep freeway–freeway ramps as filled lines that look identical to the freeway, but they’re in the minority. (If anything, these maps would be arguments for tagging these connections as highway=motorway rather than highway=motorway_link.)

Anecdotally, I’ve found that deemphasizing ramps highlights many misclassified ramps that would potentially affect turn-by-turn navigation. I think this benefit would outweigh the downside of tempting mappers to reclassify freeway–freeway ramps as freeways.

[claysmalley]

I like the offset effect, but it looks really clunky in left-hand-drive localities, including the US Virgin Islands. Perhaps an asymmetric oneway icon could make it look a little better.

[1ec5]

I was thinking this might fall under the notion that we’re primarily optimizing for North America, but a regression in left-hand driving countries would be far from ideal, and you have a point about USVI. We could bundle a GeoJSON of left-hand driving countries and reference it within a within expression, similar to #749. Performance might be an issue, however.

[1ec5]

Really the effect I’m going for is how printed maps traditionally would append an arrow to the (offset) label of each one-way street, rather than treating it as a separately placed repeating element. That would have the benefit of less repetition and probably handling divided highways better in left-hand driving countries. But I don’t think that’s feasible here, because the oneway field is only present on the transportation layer, not the transportation_name layer.

[1ec5]

I guess 25.7 kilobytes or 57.2 kB isn’t too bad, though it sure would be nice if every road just had its country code on it.