neu_mark.annotations¶

Encoding neuroglancer_annotations_v1 LINE annotations.

One annotation per synaptic connection, a line from the T-bar to the PSD, as published synapse sources do. A shader can still draw either endpoint alone (setLineWidth(0.0); setEndpointMarkerSize(pre, 0.0) shows only T-bars).

Byte layout, little-endian, checked against a published male-CNS source. Per annotation: geometry (for LINE, two float32 xyz triples), then properties in descending size (4-, 2-, then 1-byte), then zero padding to 4 bytes. The single-annotation (by_id) encoding then adds, per relationship, a uint32 count and that many uint64 ids; the multi-annotation encoding instead starts with a uint64 count and ends with all the ids.

Property order matters: a reader infers offsets from the declared order, and a file written in any other order parses into misaligned values with no error, so sort_properties() fixes it. Properties cannot be uint64; full body ids go in relationships, with a truncated uint32 copy as a property if a shader needs one.

Attributes¶

`AT_TYPE`
`GEOMETRY_FLOATS`
`LINE`
`POINT`
`PROPERTY_TYPES`

Functions¶

`build_info`(*, lower_bound, upper_bound, voxel_size_xyz) → dict[str, Any]	The annotation source's `info`.
`decode_group`(raw, *[, annotation_type, properties]) → dict[str, Any]	Invert `encode_group()`. Returns `{"ids", "geometry", "values"}`.
`decode_single`(raw, *[, annotation_type, properties, ...]) → dict[str, Any]	Invert `encode_single()`. Returns `{"geometry", "values", "relationships"}`.
`encode_group`(geometry, ids, *[, annotation_type, ...]) → bytes	The multiple-annotation encoding: a spatial cell, or one relationship's annotations.
`encode_single`(geometry, *[, annotation_type, ...]) → bytes	The single-annotation encoding, for the `by_id` index.
`enum_property`(prop_id, labels, *[, dtype, description]) → dict[str, Any]	A property whose integer values index `labels`.
`record_size`(annotation_type, properties) → int	Bytes per annotation in a group: geometry + properties + padding to 4 bytes.
`sort_properties`(properties) → list[dict[str, Any]]	Properties in the order the format requires: descending size, the caller's order

Module Contents¶

neu_mark.annotations.build_info(*, lower_bound: Sequence[float], upper_bound: Sequence[float], voxel_size_xyz: Sequence[float], annotation_type: str = LINE, properties: Sequence[Mapping[str, Any]] = (), relationships: Sequence[Mapping[str, Any]] = (), by_id: Mapping[str, Any] | None = None, spatial: Sequence[Mapping[str, Any]] = ()) → dict[str, Any]¶

The annotation source’s info.

Every member is required (relationships and spatial may be empty). dimensions is written x, y, z, matching the stored geometry, in metres (8 nm is 8e-9).

neu_mark.annotations.decode_group(raw: bytes, *, annotation_type: str = LINE, properties: Sequence[Mapping[str, Any]] = ()) → dict[str, Any]¶

Invert encode_group(). Returns {"ids", "geometry", "values"}.

Used to check a written source against the table it came from.

neu_mark.annotations.decode_single(raw: bytes, *, annotation_type: str = LINE, properties: Sequence[Mapping[str, Any]] = (), n_relationships: int = 0) → dict[str, Any]¶: Invert encode_single(). Returns {"geometry", "values", "relationships"}.

neu_mark.annotations.encode_group(geometry: numpy.ndarray, ids: Sequence[int], *, annotation_type: str = LINE, properties: Sequence[Mapping[str, Any]] = (), values: Mapping[str, Any] | None = None) → bytes¶

The multiple-annotation encoding: a spatial cell, or one relationship’s annotations.

geometry is (n, floats) in xyz, from neu_mark.tables.positions_xyz(). Relationship ids are not part of this encoding; see encode_single().

neu_mark.annotations.encode_single(geometry: Sequence[float], *, annotation_type: str = LINE, properties: Sequence[Mapping[str, Any]] = (), values: Mapping[str, Any] | None = None, relationships: Sequence[Sequence[int]] = ()) → bytes¶

The single-annotation encoding, for the by_id index.

relationships is one sequence of segment ids per declared relationship, in the declared order.

neu_mark.annotations.enum_property(prop_id: str, labels: Sequence[str], *, dtype: str = 'int16', description: str | None = None) → dict[str, Any]¶

A property whose integer values index labels.

By convention index 0 means “none” (<unspecified>), so a shader can test for it.

neu_mark.annotations.record_size(annotation_type: str, properties: Sequence[Mapping[str, Any]]) → int¶: Bytes per annotation in a group: geometry + properties + padding to 4 bytes.

neu_mark.annotations.sort_properties(properties: Sequence[Mapping[str, Any]]) → list[dict[str, Any]]¶: Properties in the order the format requires: descending size, the caller’s order kept within a size.

neu_mark.annotations.AT_TYPE = 'neuroglancer_annotations_v1'¶

neu_mark.annotations.GEOMETRY_FLOATS¶

neu_mark.annotations.LINE = 'line'¶

neu_mark.annotations.POINT = 'point'¶

neu_mark.annotations.PROPERTY_TYPES: dict[str, tuple[int, str]]¶