neu_glance.layers

Local annotation layers: authored from coordinates, or from a volume’s occupancy boxes.

Pure: plain data in, plain data out, no store access. The occupancy analysis itself is neu_vol.ops.occupancy.labeled_regions.

Local annotations, carried inline in the state, appear in the viewer’s Annotations tab, where they can be clicked through and stepped with [ / ]. A precomputed annotation source renders but lists nothing there (its frontend class iterates as empty), so it suits sets too large to carry inline (neu_glance.sources), and a local layer suits a subset to step through.

Coordinates are zyx throughout and flipped to xyz only in build_annotation(); a wrong flip gives a valid annotation in the wrong place.

Attributes

ANNOTATION_TYPES
CSV_COLUMNS
DEFAULT_RELATIONSHIP
KINDS
OPTIONAL_COLUMNS
POSITION_GROUPS
RELATIONSHIP_PREFIX

Functions

boxes_layer(regions, dims, *[, name, color, kind, label]) → dict	The occupancy layer: one annotation per region of labeled_regions.
build_annotation(record, ident[, relationships]) → dict	One annotation object, zyx in and xyz out.
local_layer(annotations, dims, *[, name, color, links]) → dict	The layer envelope for inline (local://annotations) annotations.
output_dimensions(voxel_size_zyx, units) → tuple[dict, str | None]	{"x": [scale, "m"], ...} for the layer, and a warning if units are unknown.
positions(record)	The position tuples of a record, skipping any that are extents (zyx).
read_annotation_csv(text, kind, *[, source]) → list[dict]	Rows of a CSV as annotation records: {"kind", "coords", id?, ...}.
record_relationships(records) → list[str]	Every relationship the records name, in first-seen order. segments stands for
render(obj) → str	json.dumps with each annotation kept on a single line.
rescale(records, factor_zyx) → list[dict]	Multiply every coordinate of every record by factor_zyx (per axis).

Module Contents

neu_glance.layers.boxes_layer(regions: list[dict], dims: dict, *, name: str = 'regions', color: str = '#ffee00', kind: str = 'box', label: str = 'r') → dict

The occupancy layer: one annotation per region of labeled_regions.

regions are zyx {"lo", "hi"}. (neu_glance.sources.annotation_layer() is the unrelated layer for a precomputed annotation source.)

neu_glance.layers.build_annotation(record: dict, ident: str, relationships: Sequence[str] | None = None) → dict

One annotation object, zyx in and xyz out.

A box’s corners are sorted per axis, since a reversed pair renders as nothing; a line’s order is its direction and is kept.

relationships is the layer’s declared order, and segments gets one id list per relationship in that order: neuroglancer matches them by position, not name. Without it, only the default relationship is written.

neu_glance.layers.local_layer(annotations: list[dict], dims: dict, *, name: str = 'annotations', color: str = '#ffee00', links: dict[str, str] | None = None) → dict

The layer envelope for inline (local://annotations) annotations.

links maps each relationship name to the segmentation layer it refers to. Without it the segment ids are stored but inert: neuroglancer does not know which layer they belong to. Its order must match the relationships given to build_annotation().

neu_glance.layers.output_dimensions(voxel_size_zyx, units: str | None) → tuple[dict, str | None]

{"x": [scale, "m"], ...} for the layer, and a warning if units are unknown.

Declared on the layer, so it can be pasted into any state of the same volume.

neu_glance.layers.positions(record: dict)

The position tuples of a record, skipping any that are extents (zyx).

neu_glance.layers.read_annotation_csv(text: str, kind: str, *, source: str = '<csv>') → list[dict]

Rows of a CSV as annotation records: {"kind", "coords", id?, ...}.

coords is one zyx tuple per geometry field of kind (one for a point, two for a box), in the file’s units; rescale() converts them. Columns are found by name, not position, and unknown columns are ignored.

neu_glance.layers.record_relationships(records: Sequence[dict]) → list[str]

Every relationship the records name, in first-seen order. segments stands for the plain column.

neu_glance.layers.render(obj: Any) → str

json.dumps with each annotation kept on a single line.

For pasting into neuroglancer’s JSON editor, where a person reads it.

neu_glance.layers.rescale(records: list[dict], factor_zyx: Sequence[float]) → list[dict]

Multiply every coordinate of every record by factor_zyx (per axis).

Covers voxels at another level to level-0 voxels, and nm to level-0 voxels. Radii scale like positions.

neu_glance.layers.ANNOTATION_TYPES = ('point', 'line', 'axis_aligned_bounding_box', 'ellipsoid', 'polyline')

neu_glance.layers.CSV_COLUMNS: dict[str, tuple[tuple[str, ...], ...]]

neu_glance.layers.DEFAULT_RELATIONSHIP = 'segments'

neu_glance.layers.KINDS: dict[str, tuple[str, tuple[str, ...]]]

neu_glance.layers.OPTIONAL_COLUMNS = ('id', 'description', 'segments')

neu_glance.layers.POSITION_GROUPS

neu_glance.layers.RELATIONSHIP_PREFIX = 'segments:'