neu_vol.backends.dvid¶

Read-only backend over a DVID labelmap instance.

Addressed by server/uuid/instance rather than a path (neu_vol.dvid.parse_url); addressing and version resolution are in neu_vol/dvid.py, and this module is the labelmap geometry and array view.

DVID omits blocks that hold no labels, so an empty region costs a round trip and no bandwidth, and no occupancy prefilter is needed. Reads request ?compression=blocks (DVID’s stored form) and inflate here with libdvid, on the dask worker. DVID is typically a shared service and answers overload with 503, which retry treats as transient.

Read-only: segmentation is never written back to DVID.

Attributes¶

TAG

Classes¶

DVIDBackend

Read-only (z, y, x) view over one scale of a DVID labelmap instance.

Functions¶

`geometry`(info, spec) → dict	Shape / chunks / voxel size / max scale, all zyx, from DVID's `/info`.
`level_shape`(shape0, scale) → tuple[int, ...]	Shape at a DVID downres level.
`provenance`(spec, node) → dict	What this labelmap pull came from — the shared node record plus label facts.

Module Contents¶

class neu_vol.backends.dvid.DVIDBackend(spec: Mapping[str, Any])¶

Read-only (z, y, x) view over one scale of a DVID labelmap instance.

classmethod open(spec: Mapping[str, Any]) → DVIDBackend¶

read_region(region: neu_vol.backends.base.Region) → numpy.ndarray¶

to_spec() → dict[str, Any]¶

write_region(region: neu_vol.backends.base.Region, data: numpy.ndarray) → None¶

property chunks: tuple[int, ...]¶

property dtype: numpy.dtype¶

property shape: tuple[int, ...]¶

voxel_size¶

neu_vol.backends.dvid.geometry(info: Mapping[str, Any], spec: Mapping[str, Any]) → dict¶

Shape / chunks / voxel size / max scale, all zyx, from DVID’s /info.

Presented origin-anchored: shape is MaxPoint + 1, so a voxel index means the same as in DVID. The empty margin below MinPoint costs a quick round trip per task and nothing on disk.

neu_vol.backends.dvid.level_shape(shape0, scale: int) → tuple[int, ...]¶

Shape at a DVID downres level.

DVID halves every axis per level, so here, unlike a general pyramid, the factor really is 2**scale. Do not copy this to another backend.

neu_vol.backends.dvid.provenance(spec: Mapping[str, Any], node: Mapping[str, Any]) → dict¶

What this labelmap pull came from — the shared node record plus label facts.

maxlabel advances as proofreading creates bodies, so with the node it characterises the snapshot; DVID’s mutation id and lastmod are per body, so not usable here.

neu_vol.backends.dvid.TAG = 'dvid'¶