NEWS
wordbankr 2.0.0
- In a non-interactive session with no Redivis credentials (no
REDIVIS_API_TOKEN, no cached sign-in), data functions now return NULL
with a message immediately instead of waiting for a browser sign-in.
get_crossling_items() now fails gracefully (message + NULL) when the
database is unreachable, like every other data-access function.
- All
get_* functions (and summarise_items(), wb_dataset()) gain a
version argument for pinning a data release (e.g. version = "v2.0";
the default "current" is the most recent release).
- The
db_args argument is removed from all get_* functions.
- In
get_administration_data(), the race level "Other" is now
"Other/Mixed" (for all dataset versions), and birth_order gains the
levels "Ninth" to "Twelfth".
- New functions
get_aoa() and get_embeddings() return precomputed ages
of acquisition and multilingual word embeddings for every word item.
- All
get_* functions now record the data's dataset_version in a column
of their output; when version = "current" (the default), this is
resolved to the actual current version tag (e.g. "v1.5") rather than
the literal string "current".
- Breaking (with dataset v2.0+): in
get_administration_data(),
date_of_test is now a Date (previously a string), and the nested
language_exposures column exposure_proportion is renamed
exposure_percentage (its values were always percentages, 0-100). ASL
CDITwo item ids are normalized from "Item_N" to "item_N".
- The Redivis dataset (v2.0+) uses a normalized schema — child-level
variables live in the
children table and instrument-level variables in
instruments — but get_administration_data() joins these back together,
so its flat output shape is unchanged.
- Data now come from the versioned Wordbank dataset on Redivis
(https://stanford.redivis.com/datasets/627v-9ewzpdvz0) instead of the MySQL
database. All
get_* functions keep their signatures and return the same
data.
get_administration_data(filter_age = FALSE) now returns administrations
outside the instrument age range flagged by the source data (previously
these were only reachable through the age filter's absence).
connect_to_wordbank() and get_wordbank_args() are deprecated and
ignored.
get_crossling_items() now returns a single uni_lemma column derived
from item mappings; the internal database id column and uni-lemmas
unattached to any item are no longer included.
- The
redivis client is a suggested (not imported) dependency, installable
from https://langcog.r-universe.dev; wordbankr prompts with the install
command if it is missing.
- All network access retries transient failures and then fails gracefully
(message +
NULL), and no test or example requires network access on CRAN.