[zandt-square]

Why

GeoLocation fields cannot currently be grouped in aggregation queries. This adds the geographic grouping API discussed in #1426.

What

Adds asGeoTile(zoomLevel: ...), with a default zoom level of 7. Each tile exposes zoomLevel, x, y, centerPoint, bounds, and corners.

How

Uses geotile_grid for both root and nested aggregations on Elasticsearch and OpenSearch. The response decodes the tile key into Web Mercator coordinates. Missing locations return a null tile, and root pagination compares tile coordinates in datastore order.

Standalone nested geo grids use conservative count bounds because the datastore does not provide their count error bounds.

Risk

This adds a new grouping API. Existing filtering and non-geographic grouping behavior is unchanged.

Testing

No manual testing.

Generated with Pi

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[myronmarston]

This is looking good! Left some suggestions. (Note: I haven't finished my initial review so I'll have more feedback at a later time, but I wanted to submit what I had so far).

[zandt-square]

This is looking good! Left some suggestions. (Note: I haven't finished my initial review so I'll have more feedback at a later time, but I wanted to submit what I had so far).

Thanks @myronmarston. I pushed the updates.

asGeoTile now has its own resolver, and the tile, bounds, and corners use ResolvableValue. GroupedBy no longer has geo-specific logic. I also moved sorting onto grouping.sort_value, added the shared zoom range, tightened integer parsing, and applied the naming and validation wording changes.

The grouping docs are split into subpages with an expandable nav. I clarified the boundary and corner descriptions too.

For centerPoint, I kept the projection math and clarified that it’s the latitude that differs from the average latitude bounds. The docs now include the 1/0/0 example: approximately 66.51° versus 42.53°.

[zandt-square]

@myronmarston I am closing this pr in favor of this one on the block org, which I recently gained access. #1438