Geospatial Search

Globus Search provides features in support of geospatial searches.

Data being indexed may be described as geo_point or geo_shape data. These allow users to ingest data in the form of specific coordinates, or to describe shapes like GeoJSON Polygons.

When querying data, use the geo_bounding_box filter type to select for data which is within a specific geographic area. Bounding boxes are the four corner coordinates of a rectangular map area, and are considered matching if they intersect with geospatial data (e.g. if a point is inside the box).

You must have an index where you have write permissions. At least a passing familiarity with GeoJSON is also recommended.

API Methods

We will leverage these API methods:

Ingest API

Submit an Ingest Task

POST Query

Perform a complex query

It is assumed that you have a client (e.g. globus_sdk.SearchClient in python) which can be used to leverage these APIs.

Defining Geo Data in Ingest

Normally, Globus Search automatically deduces the types of fields based on the data seen for those fields. However, in the case of geo data, this is not feasible.

For this reason, the Ingest API supports additional data regarding geo_point and geo_shape fields. Use the field_mapping field of the GIngest document to specify that a field is a geo_point or geo_shape.

field_mapping is a JSON object whose keys are field names, and whose values are either geo_point or geo_shape.

Define a geo_point

The following GIngest document defines a single point, the approximate coordinates of the city of Chicago.

Chicago has a latitude of 41.9° N and a longitude of 87.6° W. Therefore, the document reads:

{
  "ingest_type": "GMetaEntry",
  "ingest_data": {
    "subject": "https://example.com/cities/chicago",
    "visible_to": ["public"],
    "content": {
      "city": {
        "name": "Chicago",
        "state_or_province": "Illinois",
        "coordinates": "41.9, -87.6"
      }
    }
  },
  "field_mapping": {
    "city.coordinates": "geo_point"
  }
}
Tip
Remember, North and East are positive, South and West are negative.

There are several important notes here:

  1. The field name of city.coordinates is dotted, to indicate the path to the field

    If a field contains dots in one of the components in its path, they must be backslash-escaped.

  2. The coordinates are given as latitute, longitude

    This applies whenever a geo_point is specified as a string.

Alternate geo_point Formats

Globus Search supports several alternative formats for describing points. These are the same as the document as above, but with the various different supported encodings of Chicago’s coordinates.

All of these documents are equally valid and encode the same data.

In GeoJSON, points are specified as two-element arrays, of the form [longitude, latitude]. Therefore:

{
  "ingest_type": "GMetaEntry",
  "ingest_data": {
    "subject": "https://example.com/cities/chicago",
    "visible_to": ["public"],
    "content": {
      "city": {
        "name": "Chicago",
        "state_or_province": "Illinois",
        "coordinates": [-87.6, 41.9]
      }
    }
  },
  "field_mapping": {
    "city.coordinates": "geo_point"
  }
}

Search also supports points specified as two-element objects, of the form {"lon": longitude, "lat": latitude}. Therefore:

{
  "ingest_type": "GMetaEntry",
  "ingest_data": {
    "subject": "https://example.com/cities/chicago",
    "visible_to": ["public"],
    "content": {
      "city": {
        "name": "Chicago",
        "state_or_province": "Illinois",
        "coordinates": {
          "lon": -87.6,
          "lat": 41.9
        }
      }
    }
  },
  "field_mapping": {
    "city.coordinates": "geo_point"
  }
}

Define a geo_shape

Globus Search supports a wide range of shapes, as specified in GeoJSON. The following shapes are suported:

  • Point

  • LineString

  • Polygon

  • MultiPoint

  • MultiLineString

  • MultiPolygon

  • GeometryCollection

For simplicity in this demonstration, we will define a rectangular polygon for the area of the continental United States, excluding Alaska. (We are aware that the national borders are not rectangular. Therefore, this polygon will describe areas outside of the United States as well.) This box corresponds to a viewport in a map interface which displays most of the US.

The following values will be used:

  • West (longitude): -124.9

  • South (latitude): 24.9

  • East (longitude): -67.1

  • North (latitude): 49.1

This area, expressed as a GeoJSON polygon, is as follows:

{
  "type": "Polygon",
  "coordinates": [
    [
      [-124.9, 24.9],
      [-67.1, 24.9],
      [-67.1, 49.1],
      [-124.9, 49.1],
      [-124.9, 24.9]
    ]
  ]
}
Warning

In Globus Search, polygons must be closed. If the last point had been omitted from the coordinates, the polygon would be treated as invalid.

We could store this data in a Search index by ingesting the following GIngest document:

{
  "ingest_type": "GMetaEntry",
  "ingest_data": {
    "subject": "https://example.com/countries/united-states",
    "visible_to": ["public"],
    "content": {
      "country": {
        "name": "United States",
        "country_code": "us",
        "default_viewport": {
          "type": "Polygon",
          "coordinates": [
            [
              [-124.9, 24.9],
              [-67.1, 24.9],
              [-67.1, 49.1],
              [-124.9, 49.1],
              [-124.9, 24.9]
            ]
          ]
        }
      }
    }
  },
  "field_mapping": {
    "country.default_viewport": "geo_shape"
  }
}

Querying for Geo Data

Globus Search only supports querying geo data by use of the geo_bounding_box filter type.

A bounding box is a polygon of four corners, defined by a top_left corner and a bottom_right corner. They are specified like so:

{
  "type": "geo_bounding_box",
  "field_name": "country.default_viewport",
  "top_left": {
    "lat": 101.0,
    "lon": 100.0
  },
  "bottom_right": {
    "lat": 100.0,
    "lon": 101.0
  }
}

Note that the top edge must be North of the bottom edge, and the right edge must be East of the left edge.

A geo_bounding_box is a type of GFilter document, which you can find documented as part of the POST Query documentation.

Given the coordinates above used for a "United States" view, the following query searches for cities within that area:

{
  "q": "*",
  "filters": [
    {
      "type": "geo_bounding_box",
      "field_name": "city.coordinates",
      "top_left": {
        "lat": 49.1,
        "lon": -124.9
      },
      "bottom_right": {
        "lat": 24.9,
        "lon": -67.1
      }
    }
  ]
}