Findability of features at the dateline about cardboard

[sgillies]

Backing up: where would antimeridian-crossing features come from? I'm talking myself out of doing anything about this now.

Review a few possible sources.

mapbox.com

If you start at a line at 175 deg W and extend it 10 degrees westward, mapbox.com says 185 deg W. It doesn't cut at the antimeridian. Mapbox.com doesn't produce antimeridian-crossing features as such. Nor can it creat pole-crossers.

Mapbox.com will import lines that extend beyond 180dW and 180dE and does what you expect: draws them with no regard for the antimeridian.

There's definitely a search issue raised: given a bbox like [-190, 5, -185, 10](a 5x5 box W of the antimeridian) should we expect to find the point [187.5, 7.5]? Should carboard normalize input features or
do more splitting of its queries across the antimeridian?

geojson.io

Same.

uploaded data

Carefully prepared global GIS data using lng,lat WGS84 will be split at the antimeridian. NASA's Landsat footprints, for example, are split into multipolygons at 180dW. Their bounding tile should be the z0 tile and they should be searchable, if at an extra cost.

hand editing

There's nothing stopping anyone from typing {"coordinates": [[-175, 10], [175, 10]]} or {"coordinates": [[175, 10], [-175, 10]]} and uploading these. There's also no established way of knowing what the uploader's intent is: do they mean the antimerdian-crossing line or not?

Currently, mapbox.com calls these invalid and throws an error. Good for now, because it doesn't introduce problematic features.