Upgrading from standard_transform 1.x¶

2.0 changes results, not the interface. Code written against 1.x keeps running — the only hard renames are behind deprecation warnings, not errors — but the numbers it produces change, and some of them change by design. This page lists what moved and how to reproduce 1.x output exactly.

TL;DR¶

Two things produce different numbers in 2.0: the v1dd transform (a revised pia point) and the streamlines for both datasets (hand-drawn curve → data-derived field).
Nothing breaks at the API level: the field is a drop-in subclass of Streamline, and the old *_nm / *_vx factory names still work (with a DeprecationWarning).
To reproduce 1.x results exactly, pin version="1.4" on both the transform and the streamline.

Reproducing 1.x results exactly¶

from standard_transform import v1dd_ds, minnie_ds

# v1dd: pin BOTH tracks to 1.4 (its affine changed in 2.0, see below)
tf = v1dd_ds.transform(version="1.4")
sl = v1dd_ds.streamline(version="1.4")

# minnie65: the affine is unchanged, so only the streamline needs pinning
tf_m = minnie_ds.transform()                 # identical to 1.x
sl_m = minnie_ds.streamline(version="1.4")   # the old hand-drawn curve

available_versions("v1dd") (or dataset.available_versions()) lists every version and which one is latest. See Versioning & reproducibility.

What changed in results¶

v1dd transform (affine) — new default frame¶

The v1dd affine advanced to 2.0: a revised pia point, which shifts oriented depth by ~39 µm. So v1dd_ds.transform() / v1dd_transform() now returns the 2.0 frame by default, and any depth or transformed coordinate differs from 1.x by that offset. Pin version="1.4" to recover the old frame. Minnie65's affine did not change — its transformed points are identical to 1.x.

Streamlines — now data-derived fields (both datasets)¶

v1dd_ds.streamline() and minnie_ds.streamline() now return a StreamlineField — a spatially-varying, curl-free field fit from thousands of apical dendrites — instead of the single hand-drawn curve. It is a subclass of Streamline, so every method (streamline_at, radial_distance, depth_along, radial_points, transformer) works unchanged; only the values differ (and are more accurate — see Streamline field: method). Pin version="1.4" for the old curve.

If you work with single neurons, note the intended field workflow: integrate the field once at the soma and apply that one curve across the arbor, rather than re-deriving the axis per vertex —

sl = v1dd_ds.streamline().streamline_at_point(soma_xyz)  # a plain Streamline
verts_straight = sl.radial_points(soma_xyz, skel.vertices)

See the per-neuron workflow.

Renamed factory API (`resolution=`)¶

The *_nm / *_vx factories and Dataset accessors are replaced by a single resolution= argument ("nm", "vx", or an explicit [x, y, z]). The old names still work but emit a DeprecationWarning and will be removed in a future release.

1.x	2.0
`minnie_transform_nm()`	`minnie_transform("nm")`
`minnie_transform_vx()`	`minnie_transform("vx")`
`v1dd_transform_nm()` / `_vx()`	`v1dd_transform("nm")` / `("vx")`
`minnie_streamline_nm()` / `_vx()`	`minnie_streamline("nm")` / `("vx")`
`v1dd_streamline_nm()` / `_vx()`	`v1dd_streamline("nm")` / `("vx")`
`minnie_ds.transform_nm` / `.transform_vx`	`minnie_ds.transform("nm")` / `("vx")`
`minnie_ds.transform_res(res)`	`minnie_ds.transform(res)`
`minnie_ds.streamline_nm` / `.streamline_vx`	`minnie_ds.streamline("nm")` / `("vx")`
`minnie_ds.streamline_res(res)`	`minnie_ds.streamline(res)`

An explicit voxel resolution still works: minnie_transform([8, 8, 40]).

Morphology object transforms are deprecated¶

Transforming MeshParty objects in place (tform.apply_skeleton, tform.apply_meshwork_vertices, streamline.transform_skeleton_vertices, transform_meshwork_annotations, …) is deprecated. That capability is moving to Ossify, where the transformation is passed to the object. standard_transform now focuses on coordinate arrays and dataframes — transform the vertex/annotation arrays directly (e.g. with radial_points) instead. The object methods still run but warn; see Morphology objects.

Recommended entry points¶

If you are still importing the module-level factories directly, the Dataset singletons are now the recommended surface and bundle a transform + streamline together:

from standard_transform import minnie_ds, v1dd_ds

pts_um = minnie_ds.transform().apply(xyz_nm)
d      = v1dd_ds.streamline().radial_distance(xyz0_nm, xyz1_nm)