Skip to contents

Changelog

Source: NEWS.md

bancr 0.3.2

  • CAVE-table accessors now default to a public GCS snapshot (neuron_annotations/v888/*.parquet). Each of banc_cell_info(), banc_codex_annotations(), banc_nuclei(), banc_backbone_proofread(), banc_proofreading_notes(), banc_neck_connective_neurons() and banc_peripheral_nerves() gains:

    • source = c("gcs", "cave") — default "gcs". Reads the v888 parquet snapshot under gs://lee-lab_brain-and-nerve-cord-fly-connectome/neuron_annotations/v888/ (cached locally; no BANC authentication required).
    • fallback = TRUE — on primary-source failure, automatically retry against the other source with a warning. Pass fallback = FALSE to surface the original error instead.
    • banc_cell_info() and banc_proofreading_notes() now actually honour the rootids argument (previously declared but unused); filter applies in both GCS and CAVE branches.

  • On first call with a given source, the function emits a one-shot message describing where data is coming from and how to switch. Subsequent calls in the same session are silent (rlang’s .frequency = "once" mechanism).

  • Three new GCS-only accessors:

    • banc_metrics() — per-neuron cable / volume / synapse-count metrics from compiled_data/banc_888/banc_888_metrics.feather (~7.5 MB).
    • banc_edgelist_split(version = c("v2", "v3", "legacy")) — compartment-resolved edgelist (axon / dendrite / etc.) from compiled_data/banc_888/banc_888_edgelist_split_<version>.feather.
    • banc_synapses_enriched(version = c("v2", "v3")) — lazy arrow::open_dataset() handle over the per-synapse enriched parquet (~9.6 GB v2 / ~15 GB v3). Apply dplyr::filter() and dplyr::collect() to get a data frame.

  • safe_raw2nm_position() now accepts either the character form returned by CAVE or the arrow list<integer> form returned by arrow::read_parquet() on the GCS snapshots, so the GCS and CAVE branches share the same post-processing path.

  • New internal helpers in R/gcs_sources.R (banc_source_announce(), banc_with_fallback(), banc_gcs_annotation_parquet()) and a banc_gcs_compiled_path() factored out from banc_gcs_compiled_feather() so lazy parquet consumers (like banc_synapses_enriched()) can share the cached download path.

  • banc_edgelist() now defaults to source = "gcs", reading the pre-computed compiled_data/banc_888/banc_888_edgelist_simple_<version>.feather from the public bucket (no auth needed; ~285 MB for v2, ~336 MB for v3, cached locally). New version = c("v2", "v3") argument selects paper-version synapses (v2, default) vs the updated synapses_v3 edgelist. source = "cave" preserves the previous CAVE materialised view query for callers who need live data; the auto-derived view name is synapses_<version>_backbone_proofread_and_peripheral_nerves_counts, overridable via edgelist_view. Returned schema follows the source: pre, post, count, norm, post_count, pre_count from GCS; pre_pt_root_id, post_pt_root_id, n from CAVE.

  • banc_meta_create_cache() now defaults to source = "gcs", reading the public compiled meta feather at gs://lee-lab_brain-and-nerve-cord-fly-connectome/compiled_data/banc_888/banc_888_meta.feather. This needs no BANC authentication and avoids the slow CAVE union of banc_cell_info() + banc_codex_annotations().

    • source = "cave" keeps the previous CAVE-derived behaviour for callers who need labels fresher than the GCS snapshot.
    • source = "seatable" reads the draft banc_meta SeaTable and is restricted to the BANC production team. The dead SQL bug in this branch (renaming super_class despite never selecting it) is fixed.
    • The use_seatable argument is kept as a deprecated alias.

  • franken_meta() likewise defaults to source = "gcs", reading per-dataset compiled feathers from compiled_data/<slug>/<slug>_meta.feather (fafb_783, manc_121, hemibrain_121, malecns_09). source = "seatable" (formerly the "split" default) and source = "legacy" remain for production-team callers.

  • New internal helper banc_gcs_meta_feather() (downloads + caches a compiled meta feather under tools::R_user_dir("bancr", "cache")). Adds a soft dependency on the arrow package, already in Suggests.

  • README: new “Getting BANC meta data” section advertising the GCS feather as the main route for almost all users, and clarifying that the banctable_* family is for the BANC production team only.

bancr 0.3.1

  • Repository moved from flyconnectome/bancr to natverse/bancr. The old URL continues to redirect, but please update local clones with git remote set-url origin git@github.com:natverse/bancr.git and any Remotes: lines in downstream packages.
  • banc_all_synapses() retargeted to gs://lee-lab_brain-and-nerve-cord-fly-connectome/neuron_connectivity/v888/ with a new version argument (defaults to "v2"; "v1" is deprecated, "v3" is in testing). The dead rawcoords and min_size arguments have been dropped and the help page now documents each version’s CAVE coordinate space.
  • New inst/CITATION with the BANC paper (Bates et al. 2025, bioRxiv 2025.07.31.667571) and the natverse paper.
  • License field updated to GPL (>= 3) via usethis::use_gpl3_license().

bancr 0.3.0

  • new banc_lm_scene() for overlaying precomputed light-microscopy image layers on top of a public BANC Neuroglancer scene, with optional shortened-URL POST to nglstate/api/v1/post.
  • new banc_influence() family (banc_influence_arrow, banc_influence_duckdb, banc_influence_path) for connectome influence scoring against parquet snapshots in GCS.
  • read from BANC synapses_v3; with details = TRUE, banc_synapses() now returns neurotransmitter prediction scores.
  • bancsee() gains a clean_segments option and surfaces NBLAST scores alongside synapse data; assorted scene-building fixes.
  • read NBLAST match CAVE tables via banc_nblast_matches().
  • fix a Linux PATH_MAX crash in the Neuroglancer URL encoder when the scene JSON is long.
  • internal: rework CI to use system Python via actions/setup-python (the reticulate miniconda env no longer ships pip), and clean up R CMD check Rd cross-reference / \usage warnings.

Full Changelog: https://github.com/flyconnectome/bancr/compare/v0.2.1…v0.3.0

bancr 0.2.1

Full Changelog: https://github.com/flyconnectome/bancr/compare/v0.2.0…v0.2.1

bancr 0.2.0

  • first tagged release after BANC preprint

Full Changelog: https://github.com/flyconnectome/bancr/commits/v0.2.0