如何在Grakn中优雅建模多边形类型的物理对象？

Nice question! When it comes to modeling polygon-based physical objects in Grakn, you can build on your existing schema in a clean, extensible way that preserves consistency with your point/rectangle setup. Here's a practical, schema-compliant approach:

Extend Your Existing Schema #

We'll leverage Grakn's relational modeling capabilities to handle the ordered vertices that define a polygon, while reusing your existing attribute definitions.

First, add a few new elements to your schema:

define
  # Attribute to track vertex order (critical for valid polygon geometry)
  order sub attribute, value long;

  # Entity representing a single polygon vertex (reuses your position attributes)
  vertex sub entity, has lon, has lat, has elevation;

  # Relation linking a polygon object to its vertices with sequence metadata
  polygon_vertex_association sub relation,
    relates polygon_object,  # The physical object that is a polygon
    relates polygon_vertex, # A vertex belonging to the polygon
    has order;

  # Update your existing physical_object to participate in this relation
  physical_object sub entity,
    has lon, has lat, has elevation,
    has width, has depth, has height,
    relates polygon_object;

How This Works #

• Reusability: We reuse your existing lon, lat, and elevation attributes for vertices, keeping your schema consistent and avoiding redundant definitions.
• Order Preservation: The order attribute ensures vertices are sequenced correctly—this is essential for defining valid polygons, since vertex order determines the enclosed area and geometric validity.
• Flexibility: A polygon can have any number of vertices, so this setup supports simple shapes (triangles, quadrilaterals) and complex irregular polygons alike.
• Backward Compatibility: Your existing point and rectangle models still work as-is:
  • For points: Keep width, depth, height empty, and don't link any vertices.
  • For rectangles: You can either keep using the width/depth/height attributes, or model them as 4-vertex polygons (whichever fits your use case better).

Example: Inserting a Polygon #

Here's how you'd insert a triangular physical object using this schema:

insert
  # Create three vertices for the triangle
  $v1 isa vertex, has lon 0.0, has lat 0.0, has elevation 0.0;
  $v2 isa vertex, has lon 1.0, has lat 0.0, has elevation 0.0;
  $v3 isa vertex, has lon 0.5, has lat 1.0, has elevation 0.0;

  # Create the polygon physical object
  $triangular_object isa physical_object;

  # Link the object to its vertices with ordered positions
  (polygon_object: $triangular_object, polygon_vertex: $v1, order: 1) isa polygon_vertex_association;
  (polygon_object: $triangular_object, polygon_vertex: $v2, order: 2) isa polygon_vertex_association;
  (polygon_object: $triangular_object, polygon_vertex: $v3, order: 3) isa polygon_vertex_association;

Optional: Subtype for Clarity #

If you want to explicitly distinguish between point, rectangle, and polygon objects for easier querying, you can add subtypes:

define
  point_object sub physical_object;
  rectangle_object sub physical_object;
  polygon_object sub physical_object, relates polygon_object;

Then you can query all polygon objects directly:

match $poly isa polygon_object; get;

This approach keeps your schema modular, maintainable, and aligned with Grakn's core modeling principles.

内容的提问来源于stack exchange，提问作者John Doe