Data hosting has been migrated from HuggingFace to Source Cooperative (source.coop) S3 storage. This change should resolve firewall and access issues that some users experienced with HuggingFace, particularly in institutional or corporate environments. The data is now served directly from AWS S3 (us-west-2.opendata.source.coop), providing more reliable access worldwide.
Data updated to 2025-04 release as well.
Fix the issue faced by users in China who can't access the URL https://huggingface.co directly. When the user's ip address is detected to be in China, it will be automatically switched to the mirror URL https://hf-mirror.com
Another streamlined re-design following new abilities for data hosting and access. This release relies on a HuggingFace datasets hosting for data and metadata hosting in parquet and schema.org.
Data access is simplified to use the simple HuggingFace datasets API instead of the previous contentid-based resolution. This allows metadata to be defined with directly alongside the data platform independent of the R package.
A simplified access protocol relies on duckdbfs for direct reads of tables.
Several functions previously used only to manage connections are now deprecated
or removed, along with a significant number of dependencies.
Core use still centers around the same package API using the fb_tbl() function,
with legacy helper functions for common tables like species() are still accessible and
can still optionally filter by species name where appropriate. As before, loading the
full tables and sub-setting manually is still recommended.
Historic helper functions like load_taxa() (combining the taxonomic classification from Species,
Genus, Family and Order tables), validate_names(), and common_to_sci() and
sci_to_common() should be in working order, all using table-based outputs.
duckdb httpfs on windows created segfaultcontentid now a fallback method.duckdb is upgraded or when multiple sessions attempt to access the directory. Remove the default
storage directory (given by db_dir()) after upgrading duckdb if using a local copy.Note that with hash-archive.org resolver is down, and software-heritage API has strict rate limits,
so direct URL access is preferable.
use collect() on taxa_tbl()
arrange()arkdb with duckdb as database backendvalidate_names() has been rewritten [#170]diet_items() [#158]v 3.0.0 accesses a new static API for fishbase with in-memory
memoization that significantly improves performance, particularly
for large queries.
(not released to CRAN, rolled into 3.0 release)
validate_names() (#121)faoareas() (#123)genetics() endpoint (#122)taxonomy() endpoint (#126)dontrun instead of \donttest in the documentation. Apparently CRAN has decided to run donttest blocks in the examples, which can fail on their test servers when transferring data files over a network connection (which is why they were marked donttest in the first place.)First official release of the new rfishbase package, a non-backwards compatible replacement to all earlier versions of FishBase. See the package vignette for a more detailed overview and introduction.