tilezen / vector-datasource

Tilezen vector tile service - OpenStreetMap data in several formats
https://www.nextzen.org/
Other
507 stars 119 forks source link

Add more kind:* POV from Natural Earth for low zooms and places layer #1840

Closed nvkelso closed 5 years ago

nvkelso commented 5 years ago

Originally posted by @nvkelso in https://github.com/tilezen/vector-datasource/issues/1552#issuecomment-463516835

There's also work to add this in from the Natural Earth 1:50M and 1:110M scale sets (we've only added from 1:10M set in v1.7)... and from populated_places and states_provinces as some country and region capitals aren't recognized, and some regions aren't recognized.

This issue tracks integrating that data.

nvkelso commented 5 years ago

Sister issue to OSM data in #1810.

nvkelso commented 5 years ago

Nearterm, I'll prototype a sample data change in the 1:10m populated_placeas and region boundary lines, and same samples in the 1:110m and 1:50m scale sets to unblock Tilezen work. Then we'll do one other asset build once the final NE data is ready.

nvkelso commented 5 years ago

Prototype data:

Notes:

Flagging @zerebubuth for integration work.

nvkelso commented 5 years ago

Noting we should rename kind unrecognized to unrecognized_country so we can also support unrecognized_region.

Under our SemVer we can rename a recently added kind as a Patch release, but let's just bundle it in this minor update.

nvkelso commented 5 years ago

I've done more data work and need to post a new preview ZIP.

nvkelso commented 5 years ago

Here's v5.0.0-pre-4 with more detail filled out (not just samples) for the other themes, and some fixes in the 1:10m themes:

Still needs more QA, but is close.

nvkelso commented 5 years ago

We're missing data / modeling around Cyprus, followup here: https://github.com/tilezen/vector-datasource/issues/1898.

Cyprus low zoom from NE turns out to be styling problem: image

nvkelso commented 5 years ago

For Crimea we should also support FCLASS_UA (since Ukraine is primary claimant with Russia there). I'll update Natural Earth and provide new file here.

But kind:gb here should say unrecognized_country instead of just unrecognized.

image

nvkelso commented 5 years ago

I'm not sure why most of the India - Pakistan boundary is coming thru as kind:cn of line_of_control as the data I see in my source copy of NE says it should just be basic country boundary – that POV shouldn't be exported in tiles for that boundary as no one disputes it:

image

nvkelso commented 5 years ago

UPDATE: This is a data problem with the NE 50m set. New data forthcoming.

Similarly this disputed boundary is considered as country by Russia, China, and Pakistan, and is unrecognized_country by India in source Natural Earth data but those are all missing in this tile build - the red line between Kashmir and China:

image

nvkelso commented 5 years ago

Missing gaps between China and India, Bhutan at low zooms (fine and mid-zoom from NE 1:10m see 4/31.58/82.53). Data seems okay in my original of NE v5, wondering if the asset file didn't use latest ZIP?

image

nvkelso commented 5 years ago

I also need to plumb thru unrecognized_* versus the simpler unrecognized in NE and Tilezen logic needs to update.

zerebubuth commented 5 years ago

This is what I see in the ne_50m_admin_0_boundary_lines_land.shp shapefile from the ne_v5.0.0-pre4-boundaries-pov.zip NE pre-release:

image

Possibly it's just a typo? I don't see the same FCLASS_CN in the 1:10m version of the file.

nvkelso commented 5 years ago

Looks like POV for NE admin-1 region lines isn't fully wired up?

Here we see India's POV, but the Xinxang region line is still showing up inside of India's claim area:

image

image

nvkelso commented 5 years ago

Same with China / India border in the east:

image

image

This isn't a problem in Taiwan as those lines only show up zoom 8+ and would require further wiring up in https://github.com/tilezen/vector-datasource/issues/1900.

nvkelso commented 5 years ago

This needs a new NE data asset bundle and fixes for POV on region lines.

We should also investigage: https://github.com/tilezen/vector-datasource/issues/1840#issuecomment-497161486.

UPDATE: I suspect we're not seeing Taiwan not because the ETL needs work but because Taiwan region boundaries min_zoom is 8+ which never shows in mid-zoom from Natural Earth. But the ones mentioned above have been fixed in new data ZIP.

nvkelso commented 5 years ago

Here's Natural Earth v5.0.0-pre5:

nvkelso commented 5 years ago

@zerebubuth Can you look at https://github.com/tilezen/vector-datasource/issues/1840#issuecomment-497545789 again? I still don't see the POV wired up on NE sourced region lines.

nvkelso commented 5 years ago

There's also this oddness around zoom 2 where all the Kashmir lines go missing for some POV, here's India's (they're in zoom 3). But looking at the data it's because those are POV "assertion" lines that need a lower min_zoom when they are part of a POV, even though normally we wouldn't want to see them that early. The easiest fix now is to revise their min_zoom.

India:

Screen Shot 2019-06-06 at 18 20 43

China:

Screen Shot 2019-06-06 at 18 24 13

Pakistan:

Screen Shot 2019-06-06 at 18 24 21

Default:

Screen Shot 2019-06-06 at 18 24 28
nvkelso commented 5 years ago

I've verified https://github.com/tilezen/vector-datasource/pull/1904, which resolves all outstanding code change requirements.