Skip to contents

Changelog

Source: NEWS.md

fiber (development version)

fiber 0.2.0

CRAN release: 2026-07-04

Breaking changes

  • streamline: @point_data holds numeric-only attributes (e.g. fractional anisotropy).

  • bundle: @bundle_data is now restricted to named lists of scalars (length-1 values, any type). Previously arbitrary R objects were accepted. Non-scalar values (e.g. affine matrices stored as a single @bundle_data entry) must be split into individual scalar entries or moved elsewhere.

  • bundle_set: @set_data is now restricted to named lists of scalars (length-1 values, any type), consistent with @bundle_data at the bundle level.

Improvements

New features

  • bundle gains a @streamline_data slot — a named list of vectors of length SS (one entry per streamline) — for per-streamline attributes aggregated at the bundle level. add_shape_descriptors() now stores scalar descriptors (euclidean_length, curvilinear_length, sinuosity) directly in bundle@streamline_data rather than in each individual streamline.

  • bundle_set gains a @bundle_data slot — a named list of vectors of length BB (one entry per bundle) — for per-bundle attributes aggregated at the set level.

  • Automatic attribute lifting: when constructing a bundle, any @streamline_data key present in all supplied streamlines is automatically copied into bundle@streamline_data as a length-SS vector. The same happens one level up: common @bundle_data keys are lifted into bundle_set@bundle_data when constructing a bundle_set. Individual child slots are dropped; parent-level values take precedence on conflict.

  • Subsetting push-down: bundle[[i]] now pushes the corresponding bundle@streamline_data values back into the extracted streamlines’ @streamline_data. Likewise bundle_set[[i]] and pushes bundle_set@bundle_data values into the extracted bundles’ @bundle_data.

  • bundle_set: @bundles no longer needs to be a named list. Names are used to populate a bundle attribute vector called id_from_input_list and names are then dropped from the input list.

  • bundle() constructor gains a streamline_data argument for supplying per-streamline data explicitly (overrides automatic lifting).

  • bind_bundles() gains a streamline_data argument.

  • bind_bundle_sets() gains a bundle_data argument. Bare bundle arguments no longer need to be named.

  • as_bundle_set() method for bundle: the name argument is now optional (default NULL); omitting it creates an unnamed single-element bundle_set.

fiber 0.1.2

CRAN release: 2026-05-30

  • Added bundle_set S7 class — a named collection of bundle objects for multi-subject or multi-session studies.
  • Added is_bundle_set() predicate, format()/print()/length()/names()/[[/[ methods for bundle_set.
  • Added as_bundle_set() generic with methods for bundle_set (identity) and bundle (wrap).
  • Added bind_bundle_sets() to combine named bundle objects and/or bundle_set objects into a single bundle_set.
  • Relaxed the streamline validator: @point_data entries no longer need to be numeric (any vector of the correct length is accepted); @streamline_data entries no longer need to be numeric either (any scalar is accepted). Non-numeric @point_data entries are dropped with a warning when reparametrize() is called, since they have no natural arc-length interpolant.
  • Added missing \value documentation to the compute_hausdorff_distance() catch-all method (#CRAN).
  • Replaced \dontrun{} with if (requireNamespace("dti", quietly = TRUE)) {} in the as_dwifiber() example (#CRAN).

fiber 0.1.1

  • Improved documentation of S7 classes, generics and methods following the release of roxygen2 8.0.0.
  • Removed redundant new_streamline() and new_bundle() constructors.
  • Added coercers from and to the dwiFiber S4 class of the dti package.
  • Refactored shape descriptors API.
  • Rewrote pairwise Hausdorff distance matrix using C++ via the cpp11 package.
  • Created the tractoverse GitHub organization and transferred fiber repository to it.

fiber 0.1.0

Breaking changes

  • The package has been rebuilt from scratch around two S7 classes: streamline and bundle. The previous tibble-based streamline and list-based tract are removed.

New features

Dependencies

  • Runtime dependencies reduced to S7, rlang, and cli only.