Transforming morphology objects (deprecated)

Deprecated — moving to Ossify

The methods that transform MeshParty Skeleton and Meshwork objects in place — apply_skeleton, apply_meshwork_vertices, apply_meshwork_annotations, and the streamline transform_skeleton_vertices / transform_meshwork_vertices / transform_meshwork_annotations — are deprecated and will be removed in a future release. They emit a DeprecationWarning.

Transforming morphology objects is moving to Ossify. The design there is inverted: instead of a transform reaching into an object and mutating it, you pass the transformation to the object and it transforms its own linked representations (mesh, skeleton, graph, annotations) consistently.

What still works

standard_transform remains the right tool for coordinate arrays. Nothing about point/dataframe transforms is deprecated:

# Affine reorientation of a vertex array
verts_um = tform.apply(skel.vertices)

# Streamline straightening of a coordinate array (radial distance vs. depth)
verts_straight = sl.radial_points(soma_xyz, skel.vertices)
depths         = sl.depth_along(skel.vertices)

To transform a morphology object, compute the transformed coordinates this way and hand them to your object — or use Ossify, which manages the object and its linked layers for you.

Ossify integration

Ossify's Cell.transform takes a single-argument callable f(points) -> points. Streamline.transformer(anchor) returns exactly that — the streamline straightening with the anchor bound in — so no partial/lambda is needed:

cell.transform(sl.transformer(cell.s.root_location), inplace=True)

For the per-neuron field workflow, streamline_at_point remembers the soma, so the anchor can be omitted:

sl = v1dd_ds.streamline().streamline_at_point(soma_xyz)   # stores the anchor
cell.transform(sl.transformer(), inplace=True)

The transformer coerces its input to a float array, so it is robust to the object-dtype arrays some containers hand out.