In 0.4.7, calling antimeridan.fix_item with an antimeridian-crossing polygon or antimerdian-touching multipolygon and strategy SPLIT would create a multipolygon and a "tight" bbox that crossed the antimeridian (e.g., had positive sw longitude and negative ne longitude). In https://github.com/stac-utils/stactools/pull/426 , this behavior was changed so that instead of calculating the longitude from the polygons within the multipolygon, the bounds attribute of a shapely shape is used. However, shapely calculates the bounds as a latitude-band-spanning bbox, not accounting for the fact that a tighter bbox can be found by crossing the antimerdian.
To reproduce
This geometry in an Item when called with antimeridan.fix_item will create a latitude-spanning bbox instead of a tight one.
Describe the bug
In 0.4.7, calling antimeridan.fix_item with an antimeridian-crossing polygon or antimerdian-touching multipolygon and strategy SPLIT would create a multipolygon and a "tight" bbox that crossed the antimeridian (e.g., had positive sw longitude and negative ne longitude). In https://github.com/stac-utils/stactools/pull/426 , this behavior was changed so that instead of calculating the longitude from the polygons within the multipolygon, the
bounds
attribute of a shapely shape is used. However, shapely calculates the bounds as a latitude-band-spanning bbox, not accounting for the fact that a tighter bbox can be found by crossing the antimerdian.To reproduce
This geometry in an Item when called with antimeridan.fix_item will create a latitude-spanning bbox instead of a tight one.
Expected behavior
antimeridian.fix_item creates a tight, antimeridian-crossing bbox.