Changes in version 0.4.0 (2026-07-19)                  

  - region.tree() now validates its arguments (taxon, location mode)
    before checking that rgbif is installed, so argument errors (e.g.
    missing location) are reported even when rgbif isn't available,
    instead of being masked by the "requires the 'rgbif' package"
    message.

                        Changes in version 0.3.0                        

New features

  - region.tree() builds tree(s) for a clade within a geographic area,
    pulling the species list from GBIF (via rgbif) and handing off to
    taxa.tree(). Two modes:
      - Radius --- a circle of radius_km around (lat, lon); the WKT
        circle is built with geosphere and wound counter-clockwise as
        GBIF requires.
      - Admin area --- a GADM GID via gadm (worldwide), or a Canadian
        province/territory via province. All 13 Canadian provinces and
        territories are supported by name or postal code (AB, BC, MB,
        NB, NL, NT, NS, NU, ON, PE, QC, SK, YT).
  - GBIF species are found by occurrence faceting (no bulk download, no
    credentials); only georeferenced records without geospatial issues
    are counted. Note the result is occurrence-based and
    sampling-biased, not a curated checklist -- see ?region.tree.
  - New Suggests: rgbif, geosphere.

                        Changes in version 0.2.1                        

  - downto.tree()'s key argument now exports the NCBI Entrez key as the
    ENTREZ_KEY environment variable for the duration of the call, so
    both the get_uid and downstream requests use it (previously it
    reached only get_uid). The prior ENTREZ_KEY value is restored on
    exit. Setting ENTREZ_KEY yourself and leaving key = NULL works
    identically.

                        Changes in version 0.2.0                        

Major backend change to make the package installable and CRAN-eligible
again after its original dependencies were archived.

Breaking changes

  - Dropped brranching / Phylomatic backend. brranching was archived
    from CRAN (2023-02-24) and the Phylomatic web service it called
    (phylodiversity.net) was shut down. Both functions now build trees
    as induced subtrees of the Open Tree of Life synthetic tree via
    rotl.
  - Phylum safeguard. taxa.tree() groups taxa by phylum and returns one
    tree per phylum, so no induced tree ever spans above phylum.
    Mammals, angiosperms, arthropods, etc. come back as separate trees.
  - taxa.tree() and downto.tree() now return (invisibly) list(trees,
    unmatched), where trees is a list keyed by phylum and each element
    is list(tree, dist). (Earlier 0.2.0 drafts returned a single
    list(tree, dist, unmatched).)

New features

  - Fish backend. source = "fish" builds a dated chronogram from the
    Fish Tree of Life (fishtree, in Suggests) for ray-finned fishes.
    Unlike the topology-only Open Tree backend, its distance matrix is
    patristic (time). Requires install.packages("fishtree").

Improvements

  - Names that Open Tree cannot resolve, or that resolve but are absent
    from the synthetic tree, are reported in $unmatched instead of
    silently failing inside try().
  - downto.tree() resolves the taxon to a UID once (was queried
    repeatedly), runs non-interactively (ask = FALSE), and supports db =
    "gbif" / "itis" to avoid NCBI rate limits.
  - Because Open Tree subtrees are topology-only, unit branch lengths
    are assigned when absent so dist is well defined (edge counts).

Removed dependencies

  - brranching, phytools, xml2 are no longer used.