Deck 3: Transformations

What is the precise rule about what a coordinate transformation connects?

Only an element to a coordinate system → never element-to-element or system-to-system.

What does Identity mean as a transformation?

The element's own coordinates are already exactly the target system's coordinates → no change needed.

Name the five real transformation classes shown in the docs.

Scale, MapAxis, Affine, Translation, Sequence

What does MapAxis({"x": "y", "y": "x"}) do?

Permutes/swaps the x and y axes.

What is notable about combining transformations with Sequence?

It can combine transformations defined over different axis sets without requiring matching axes.

What happens if you compose a transformation with its own .inverse() in a Sequence?

They cancel out → equivalent to no transformation.

Does setting a transformation on an element modify the underlying pixel/point data?

No → it's metadata only, until spatialdata.transform() is explicitly called.

Why is it generally discouraged to call transform() on raster data?

Raster data is heavy, so physically recomputing transformed pixel grids is expensive; it's preferred to transform the lightweight vector data instead, or inverse-transform a query region into the raster's pixel space.

What known bug affects transform(), rasterize(), and transform_to_data_extent()?

A half-pixel misalignment (tracked as GitHub issue #165).