Spatial Data

Experimental With version 11 we release support for spatial data. This feature is not totally finished and polished yet. Spatial types is a community-driven feature. As the core team has little experience with spatial data, we need you feedback and decided to release what we have. It is important for use to deliver you the best experience, so reach out to us if you run into issues or have ideas to improve it. We try not to introduce breaking changes, but we save ourself the possibility to make changes to the api in future releases when we find flaws in the current design.

Spatial data describes locations or shapes in form of objects. Many database providers have support for storing this type of data. APIs often use GeoJSON to send spatial data over the network.

The most common library used for spatial data in .NET is NetTopologySuite. Entity Framework supports Spatial Data and uses NetToplogySuite as its data representation.

The package HotChocolate.Spatial integrates NetTopologySuite into HotChocolate. With this package your resolvers can return NetTopologySuite shapes and they will be transformed into GeoJSON.

Getting Started

You first need to add the package reference to your project. You can do this with the dotnet cli:

  dotnet add package HotChocolate.Spatial

To make the schema recognize the spatial types you need to register them on the schema builder.

services
    .AddGraphQLServer()
    .AddSpatialTypes();

All NetToplogySuite runtime types are now bound to the corresponding GeoJSON type.

public class Pub
{
    public int Id { get; set; }

    public string Name { get; set; }

    public Point Location { get; set; }
}

public class Query
{
    // we use ef in this example
    [UseDbContext(typeof(SomeDbContext))]
    public IQueryable<Pub> GetPubs([ScopedService] SomeDbContext someDbContext)
    {
        return someDbContext.Pubs;
    }
}

type Pub {
  id: Int!
  name: String!
  location: GeoJSONPointType!
}

type Query {
  pubs: [Pub!]!
}

{
  pubs {
    id
    location {
      __typename
      bbox
      coordinates
      crs
      type
    }
    name
  }
}

{
  "data": {
    "pubs": [
      {
        "id": 1,
        "location": {
          "__typename": "GeoJSONPointType",
          "bbox": [12, 12, 12, 12],
          "coordinates": [[12, 12]],
          "crs": 4326,
          "type": "Point"
        },
        "name": "The Winchester"
      },
      {
        "id": 2,
        "location": {
          "__typename": "GeoJSONPointType",
          "bbox": [43, 534, 43, 534],
          "coordinates": [[43, 534]],
          "crs": 4326,
          "type": "Point"
        },
        "name": "Fountains Head"
      }
    ]
  }
}

Spatial Types

HotChocolate supports GeoJSON input and output types. There is also a GeoJSON scalar to make generic inputs possible.

Output Types

The following mappings are available by default:

NetTopologySuite	GraphQL
Point	GeoJSONPointType
MultiPoint	GeoJSONMultiPointType
LineString	GeoJSONLineStringType
MultiLineString	GeoJSONMultiLineStringType
Polygon	GeoJSONPolygonType
MultiPolygon	GeoJSONMultiPolygonType
Geometry	GeoJSONInterface

All GeoJSON output types implement the following interface.

interface GeoJSONInterface {
  "The geometry type of the GeoJson object"
  type: GeoJSONGeometryType!
  "The minimum bounding box around the geometry object"
  bbox: [Float]
  "The coordinate reference system integer identifier"
  crs: Int
}

A NetTopologySuite.Gemeotry is mapped to this interface by default.

Input Types

For each output type there is a corresponding input type

NetTopologySuite	GraphQL
Point	GeoJSONPointInput
MultiPoint	GeoJSONMultiPointInput
LineString	GeoJSONLineStringInput
MultiLineString	GeoJSONMultiLineStringInput
Polygon	GeoJSONPolygonInput
MultiPolygon	GeoJSONMultiPolygonInput

Scalar

With interfaces or unions it is possible to have multiple possible return types. Input types do not yet have a way of defining multiple possibilities. As an addition to output and input types there is the Geometry scalar, which closes this gap. When a resolver expects any Geometry type as an input, you can use this scalar. This scalar should be used with caution. Input and output types are much more expressive than a custom scalar.

scalar Geometry

Projections

To project spatial types, a special handler is needed. This handler can be registered on the schema with .AddSpatialProjections()

    services
        .AddGraphQLServer()
        .AddProjections()
        .AddSpatialTypes()
        .AddSpatialProjections()

The projection middleware will use this handler to project the spatial data directly to the database

[UseDbContext(typeof(SomeDbContext))]
[UseProjection]
public IQueryable<Pub> GetPubs([ScopedService] SomeDbContext someDbContext)
{
    return someDbContext.Pubs;
}

{
  pubs {
    id
    location {
      __typename
      bbox
      coordinates
      crs
      type
    }
    name
  }
}

SELECT p."Id", p."Location", p."Name"
FROM "Pubs" AS p

Filtering

Entity framework supports filtering on NetTopologySuite objects. HotChocolate.Spatial provides handlers for filtering spatial types on IQueryable. These handlers can be registered on the schema with .AddSpatialFiltering()

    services
        .AddGraphQLServer()
        .AddProjections()
        .AddSpatialTypes()
        .AddSpatialFiltering()

After the registration of the handlers UseFiltering() will infer the possible filter types for all Geometry based types.

[UseDbContext(typeof(SomeDbContext))]
[UseFiltering]
public IQueryable<Pub> GetPubs([ScopedService] SomeDbContext someDbContext)
{
    return someDbContext.Pubs;
}

type Query {
  pubs(where: PubFilterInput): [Pub!]!
}

input PubFilterInput {
  and: [PubFilterInput!]
  or: [PubFilterInput!]
  id: ComparableInt32OperationFilterInput
  name: StringOperationFilterInput
  location: PointFilterInput}

input PointFilterInput {
  and: [PointFilterInput!]
  or: [PointFilterInput!]
  m: ComparableDoubleOperationFilterInput
  x: ComparableDoubleOperationFilterInput
  y: ComparableDoubleOperationFilterInput
  z: ComparableDoubleOperationFilterInput
  area: ComparableDoubleOperationFilterInput
  boundary: GeometryFilterInput
  centroid: PointFilterInput
  dimension: DimensionOperationFilterInput
  envelope: GeometryFilterInput
  geometryType: StringOperationFilterInput
  interiorPoint: PointFilterInput
  isSimple: BooleanOperationFilterInput
  isValid: BooleanOperationFilterInput
  length: ComparableDoubleOperationFilterInput
  numPoints: ComparableInt32OperationFilterInput
  ogcGeometryType: OgcGeometryTypeOperationFilterInput
  pointOnSurface: PointFilterInput
  srid: ComparableInt32OperationFilterInput
  contains: GeometryContainsOperationFilterInput
  distance: GeometryDistanceOperationFilterInput
  intersects: GeometryIntersectsOperationFilterInput
  overlaps: GeometryOverlapsOperationFilterInput
  touches: GeometryTouchesOperationFilterInput
  within: GeometryWithinOperationFilterInput
  ncontains: GeometryContainsOperationFilterInput
  ndistance: GeometryDistanceOperationFilterInput
  nintersects: GeometryIntersectsOperationFilterInput
  noverlaps: GeometryOverlapsOperationFilterInput
  ntouches: GeometryTouchesOperationFilterInput
  nwithin: GeometryWithinOperationFilterInput
}

Distance

The distance filter is an implementation of Geometry.Within

The filter requires an input geometry. You can optionally buffer this geometry with the input field buffer. The filter also has all comparable filters.

input GeometryDistanceOperationFilterInput {
  geometry: Geometry!
  buffer: Float
  eq: Float
  neq: Float
  in: [Float!]
  nin: [Float!]
  gt: Float
  ngt: Float
  gte: Float
  ngte: Float
  lt: Float
  nlt: Float
  lte: Float
  nlte: Float
}

{
  pubs(
    where: {
      location: {
        within: { geometry: { type: Point, coordinates: [1, 1] }, lt: 120 }
      }
    }
  ) {
    id
    name
    location
  }
}

SELECT c."Id", c."Name", c."Area"
FROM "Counties" AS c
WHERE ST_Within(c."Area", @__p_0)

The negation of this operation is nwithin

SELECT c."Id", c."Name", c."Area"
FROM "Counties" AS c
WHERE NOT ST_Within(c."Area", @__p_0)

Contains

The contains filter is an implementation of Geometry.Contains

The filter requires an input geometry. You can optionally buffer this geometry with the input field buffer.

input GeometryContainsOperationFilterInput {
  geometry: Geometry!
  buffer: Float
}

{
  counties(
    where: {
      area: { contains: { geometry: { type: Point, coordinates: [1, 1] } } }
    }
  ) {
    id
    name
    area
  }
}

SELECT c."Id", c."Name", c."Area"
FROM "Counties" AS c
WHERE ST_Contains(c."Area", @__p_0)

The negation of this operation is ncontains

SELECT c."Id", c."Name", c."Area"
FROM "Counties" AS c
WHERE NOT ST_Contains(c."Area", @__p_0)

Touches

The touches filter is an implementation of Geometry.Touches

The filter requires an input geometry. You can optionally buffer this geometry with the input field buffer.

input GeometryTouchesOperationFilterInput {
  geometry: Geometry!
  buffer: Float
}

{
  counties(
    where: {
      area: {
        touches: {
          geometry: {
            type: Polygon,
            coordinates: [[1, 1], ....]
          }
        }
      }
    }){
      id
      name
      area
    }
}

SELECT c."Id", c."Name", c."Area"
FROM "Counties" AS c
WHERE ST_Touches(c."Area", @__p_0)

The negation of this operation is ntouches

SELECT c."Id", c."Name", c."Area"
FROM "Counties" AS c
WHERE NOT ST_Touches(c."Area", @__p_0)

Intersects

The intersects filter is an implementation of Geometry.Intersects

The filter requires an input geometry. You can optionally buffer this geometry with the input field buffer.

input GeometryIntersectsOperationFilterInput {
  geometry: Geometry!
  buffer: Float
}

{
  roads(
    where: {
      road: {
        intersects: {
          geometry: {
            type: LineString,
            coordinates: [[1, 1], ....]
          }
        }
      }
    }){
      id
      name
      road
    }
}

SELECT r."Id", r."Name", r."Road"
FROM "Roads" AS r
WHERE ST_Intersects(r."Road", @__p_0)

The negation of this operation is nintersects

SELECT r."Id", r."Name", r."Road"
FROM "Roads" AS r
WHERE NOT ST_Intersects(r."Road", @__p_0)

Overlaps

The overlaps filter is an implementation of Geometry.Overlaps

input GeometryOverlapsOperationFilterInput {
  geometry: Geometry!
  buffer: Float
}

{
  county(
    where: {
      area: {
        overlaps: {
          geometry: {
            type: Polygon,
            coordinates: [[1, 1], ....]
          }
        }
      }
    }){
      id
      name
      area
    }
}

SELECT c."Id", c."Name", c."Area"
FROM "Counties" AS c
WHERE ST_Overlaps(c."Area", @__p_0)

The negation of this operation is noverlaps

SELECT c."Id", c."Name", c."Area"
FROM "Counties" AS c
WHERE NOT ST_Overlaps(c."Area", @__p_0)

Within

The within filter is an implementation of Geometry.Within

input GeometryWithinOperationFilterInput {
  geometry: Geometry!
  buffer: Float
}

{
  pubs(
    where: {
      location: {
        within: { geometry: { type: Point, coordinates: [1, 1] }, buffer: 200 }
      }
    }
  ) {
    id
    name
    location
  }
}

SELECT c."Id", c."Name", c."Area"
FROM "Counties" AS c
WHERE ST_Within(c."Area", @__p_0)

The negation of this operation is nwithin

SELECT c."Id", c."Name", c."Area"
FROM "Counties" AS c
WHERE NOT ST_Within(c."Area", @__p_0)

What's next?

In upcoming releases spatial data will get reprojection features and sorting capabilities.

Reprojection

At the moment the coordinate reference system (crs) is fixed. The user has to know the crs of the backend to do spatial filtering. The API will furthermore always return the data in the crs it was stored in the database.

We want to improve this. The user should be able to send data to the backend without knowing what the crs. The backend should reproject the incoming data automatically to the correct crs.

Additionally we want to provide a way for users, to specify in what CRS they want to receive the data.

Sorting

Currently we only support filtering for spatial data. We also want to provide a way for users to sort results. This can e.g. be used to find the nearest result for a given point.