Split out geometry implementation about osm2streets HOT 3 CLOSED

If you can comfortably use georust::geo and get rid of geom, more power to you, but because georust is intended to be a sort of "lowest common denominator", I would be surprised if it meets your needs. I'd guess you'd end up with some kind of application specific geometry library to wrap it.

That said, I expect there'd be dividends in eventually making the types in geom a "New Type" wrapper for geo, that way all the geo algorithms are "at hand" and you don't need to worry about round tripping everything all the time, but you can still add your own trait impls.

So.. pragmatically speaking, I'd guess the most productive way forward would be to start with Option 1, but that doesn't preclude working toward Option 2A in the longer term.

regard cavalier_contours et. al - I haven't worked with it, but it seems plausible!

from osm2streets.

I have tweaked the title of this issue, because we have started towards the extraction (in all its complexities) and testing elsewhere, while there are valuable ideas about the geom libraries, that I think we should keep working on here.

from osm2streets.

That said, I expect there'd be dividends in eventually making the types in geom a "New Type" wrapper for geo, that way all the geo algorithms are "at hand" and you don't need to worry about round tripping everything all the time, but you can still add your own trait impls.

I'm leaning more towards this now. If I had to categorize the parts of geom currently:

• could just wrap a geo object: Bounds and GPSBounds (probably Rect), LonLat and Pt2D, Line, Polygon, PolyLine, Ring
• other geometric with no equivalent: Angle, Circle, Distance
• unrelated to geometry: Duration, Percent, Speed, Statistic, Time

I'm still not sure what the end-game looks like, but it could be very incremental work to go through that first list, make each type wrap a geo equivalent, and rethink the current API exposed. For example, abst Polygon immediately does ear-cutting to triangulate and stores the result. That was a very legacy decision I'd love to revisit.

from osm2streets.