v2.10.0

For lack of official tags with previous releases, including all notes since 2.7.1:

2.10.0

• Fully supported 'between' as alternative to 'within' for filter operations.
• Allow passing labels through API directly.

2.9.2

• Fix bug in manually-assigned categorical color schemes involving the first color always being gray.

2.9.1

• Fix regression bug for log-scales on linear color schemes.

2.9.0

• Allow asynchronous transformations. This is an internal change that allows alteration of tiles using any external resources--for instance, fetching search results from the Web or running duckdb on wasm.
• Various changes resulting from that.
• Customizable options for foreground/background behavior passed to the API as 'background_options'.

2.8.0

• Add new 'foreground' aesthetic; when enabled, this moves points to the front of the screen and makes points behind it not clickable.
• Removed event listener that significantly slowed down map when clicking to drag locations.

2.7.1

• Bugfix: Supply missing functions to Aesthetic codes.

2.7.0

2.7.0

• Revamp a number of bad choices in the 'point_size' and 'alpha' parameters so that the units better correspond to screen pixels (for size) and alpha (on a scale of 1 to 100.) This unfortunately will requiring tweaking existing maps.
• Add auto-generated documentation.
• Allow dragging of labels around the screen for editing label collections.
• Make labels by default filter the underlying data if the geojson property name is in the
• Allow/restore custom color schemes for categorical data.
• Improve behavior of scales for temporal fields. (Note--deepscatter supports only Arrow timestamp fields, not Date32, Datetime64, or any of the other date/time implementations in Arrow.)

WebGL

Total rewrite of the original strategy around WebGL, Apache Arrow, and Webworkers.

Clean animations, arbitrary filters on individual dimensions.

Tiling requires a separate python module--see README. I'll probably allow input that isn't in the feather format, since it's reasonable to do this with, say, 500K points at once.

ES module is somewhat tested; umd module not.