The section on spatial:dimensions says this:
The dimension names must match the dimension names defined in the array's shape.
But the dimension_names attribute is not provided in the examples, nor is it in the schema. Groups obviously don't have dimension_names but the arrays they contain must have the attribute. I would suggest that this is made explicit in the text and that the schema has a constraint for it when "node": "array".
Furthermore, the spec says:
For 2D spatial data, provide 2 elements in row-major order: ["y", "x"]
I am curious as to why row-major ordering is prescribed here. The array data has an ordering which is described in the transform codec, which could be any way, and the shape attribute gives dimension length in the order in which the data is stored. From my perspective (coming from the column-major world of R) it would make more sense to either record the spatial dimension names in the order in which they appear in the data, or not require any specific ordering at all (seeing that dimension names are being used, matching will presumably be on the name anyway).
The section on
spatial:dimensionssays this:But the
dimension_namesattribute is not provided in the examples, nor is it in the schema. Groups obviously don't havedimension_namesbut the arrays they contain must have the attribute. I would suggest that this is made explicit in the text and that the schema has a constraint for it when"node": "array".Furthermore, the spec says:
I am curious as to why row-major ordering is prescribed here. The array data has an ordering which is described in the
transformcodec, which could be any way, and theshapeattribute gives dimension length in the order in which the data is stored. From my perspective (coming from the column-major world of R) it would make more sense to either record the spatial dimension names in the order in which they appear in the data, or not require any specific ordering at all (seeing that dimension names are being used, matching will presumably be on the name anyway).