France Departments winding order incompatible with Vega about ems-file-service HOT 5 OPEN

I'm all for using "standards" (even when standards were created after an established industry practice, and contradicts it). My only concern is that we must not break existing users -- once we introduce an automatic rewinding for Vega users (should be fairly easy to do), we should introduce a new version of the EMS file service that serves different set of files (same files but different winding order). This way existing users will continue using things as is, and eventually all will migrate to the new service.

from ems-file-service.

We should make this a test to find any other dataset that is breaking in Vega.

from ems-file-service.

I've been investigating this a bit more.

With geojsonhint I've checked and 61 of our current datasets have one or more warnings about the right-hand rule:

  1:1  warning  Polygons and MultiPolygons should follow the right-hand rule

As Elastic Maps and Landing Page, they render correctly in Vega as well. Just to take Kibana out of the equation you can test this code in the Vega Editor for Bolivia departments:

{
  "$schema": "https://vega.github.io/schema/vega-lite/v3.json",
  "width": 500,
  "height": 300,
  "data": {
    "url": "https://vector.maps.elastic.co/files/bolivia_departments_v1.geo.json?elastic_tile_service_tos=agree&my_app_name=ems-landing-page&my_app_version=7.11.0&license=643c1faf-80fc-4ab0-9323-4d9bd11f4bbc",
    "format": {"property": "features"}
  },
  "transform": [
    
  ],
  "projection": {"type": "mercator"},
  "mark": "geoshape"
}

And this file has warnings when validating it:

dist/vector/files on  master
➜ geojsonhint --format pretty bolivia_departments_v1.geo.json
  1:1  warning  Polygons and MultiPolygons should follow the right-hand rule
  1:1  warning  Polygons and MultiPolygons should follow the right-hand rule
  1:1  warning  Polygons and MultiPolygons should follow the right-hand rule
  1:1  warning  Polygons and MultiPolygons should follow the right-hand rule
  1:1  warning  Polygons and MultiPolygons should follow the right-hand rule
  1:1  warning  Polygons and MultiPolygons should follow the right-hand rule
  1:1  warning  Polygons and MultiPolygons should follow the right-hand rule
  1:1  warning  Polygons and MultiPolygons should follow the right-hand rule
  1:1  warning  Polygons and MultiPolygons should follow the right-hand rule

⚠ 9 warnings

Checking the previous version of the France departments there's a big wrong polygon in the South. geojsonhint only complains about the right-hand rule but maybe this is combined with something else.

I will continue debugging the file when I'm back from PTO.

from ems-file-service.

OK, I did a few more tests and the issue is as @nickpeihl stated the conformity of the France departments GeoJSON file to the RFC 7946 spec. This was already discussed in detail in 2018 on this issue with @thomasneirynck and @nyurik. I'm sorry I did not find that issue before 😞

I will explore adding a step in the build process to enforce a winding order with geojson-rewind, ensuring all our published GeoJSON files are consistent, one way or the other.

Still, the main point of the discussion remains. I'm more inclined to ensure all our GeoJSON files follow the RFC 7946 standard, finding a way for our Vega users to get them properly rendered by d3-geo.

from ems-file-service.

After sync discussion with @nickpeihl and @thomasneirynck we agreed that EMS data needs to follow industry standards so we need to

• Upgrade all data in EMS File Service that is not yet following RFC7946 using the EMS data versioning mechanism, so we don't modify current datasets, potentially breaking user workflows. #240
• Introduce in Kibana a process to rewind the polygons so they work as expected by d3-geo and Vega elastic/kibana#122569

from ems-file-service.