Changelog • crantr

crantr 0.2.1

New crant_key_point() returns a “good” annotation point on a neuron: the principal branch point of its L2 skeleton, with the neuron optionally rerooted onto its furthest endpoint first. Shares the dataset-agnostic fafbseg::key_point_from_neuron() helper but reads via crant_read_l2skel() because the CRANT segmentation is hosted on a separate CAVE server that fafbseg::read_l2skel() cannot currently reach (requires fafbseg (>= 0.15.15)).

Metadata rework, now backed by fafbseg::cam_meta() (requires fafbseg (>= 0.15.12)):

crant_meta() is now a thin wrapper around fafbseg::cam_meta() pointed at the CRANTb_meta seatable. Column normalisation to the coconat/crantr schema is opt-in via normalise_colnames = TRUE, and translate_ids is exposed to keep returned root ids consistent with a requested version/timestamp.
crant_meta() gains drop_status (default "DUPLICATED") so CRANT’s known duplicate rows are filtered by default; pass drop_status = NULL to keep them.
crant_ids() gains na.rm (default FALSE) to drop missing root ids. The coconat id functions set this so CRANTb_meta’s incomplete rows no longer surface as spurious null-segment "0" ids.
Fixed crant_ids() query lookup and its integer64 default.
The legacy metadata cache API is deprecated in favour of the above.