size_human started counting from kilobytes and then stepped up a unit
before reporting, so a 47 KB cache was reported as "47.1 MB". Sizes now
use the right unit.
fred_request(), which every data function goes through, let
httr2::req_perform() raise on any HTTP error and reported all of them
as "Failed to connect to the FRED API. Check your internet connection".
A mistyped series ID or an unregistered API key, both HTTP 400, sent
users to check their network, and the status-specific handling for 400,
404 and 429 below it was unreachable. HTTP errors now reach that
handling, and a 400 quotes FRED's own explanation ("The series does not
exist.", "The value for variable api_key is not registered.").
fred_series() cached each request under a key built from its arguments
and returned that copy on every later call, with no expiry. A call such as
fred_series("UNRATE") therefore kept returning the observations from the
first time it ran, without new releases or revisions, until the cache was
cleared or cache = FALSE was passed. Cached copies are now downloaded
again once they are 24 hours old. Set options(fred.cache_ttl = ) to a
number of hours to change this (Inf restores the old behaviour).
Resubmission of 0.3.0, which CRAN's incoming checks rejected on 2026-09-08. Two problems, both now fixed.
The donttest additional check ran fred_all_vintages() on a machine with
no FRED_API_KEY and it aborted rather than failing gracefully. 0.3.0
wrapped 14 example blocks in try() but missed 14 more, including every
vintage helper and fred_get_key() itself. All of them are wrapped now.
Verified by extracting every generated example with
tools::Rd2ex(commentDonttest = FALSE) and running all 35 with the API key
unset: none errors.
inst/CITATION carried doi = "10.5281/zenodo.fred", a placeholder that
was never replaced with a real deposit. It resolves to 404 on both CrossRef
and DataCite, which CRAN flagged as an invalid DOI. There is no Zenodo
record for this package, so the field has been removed rather than
corrected.
A research-grade upgrade in three logical blocks: (1) discoverability and reference data, (2) workflow utilities and the default plot method, and (3) reproducibility helpers. 35 exports + 4 S3 methods, 275 tests, three new vignettes.
Every \donttest{} example that reaches the FRED API is wrapped in try().
14 blocks were affected. CRAN runs these in its additional-issues
donttest check, on build machines the upstream host routinely refuses or
rate-limits, and an example that could not reach it was an ERROR rather
than a printed condition. The options(op) cache restore stays outside the
try() so it runs either way.
This is the CRAN Repository Policy requirement that a package using an internet resource fail gracefully when the resource is unavailable. It is the rule obr was archived under on 2026-08-22.
fred_catalogue() returns an offline curated catalogue of around 50
widely used FRED series (GDP, employment, inflation, rates, housing,
financial, money/credit, trade/FX, consumer, fiscal). Filter by category
or free-text. No API call required.fred_browse() pretty-prints the FRED category tree. With no
arguments, shows the eight top-level categories from a static reference
(no API call). Pass a category_id to drill into children.fred_recession_dates() returns NBER business-cycle reference dates
(peak, trough, duration) since 1857. Pass a vector of dates to flag to
get back per-observation in-recession indicators for use as a regression
covariate.fred_fomc_dates() returns FOMC scheduled meeting decision dates
2017 to 2026, including SEP-meeting flags. Selected unscheduled meetings
during stress periods are included.fred_tbl now threads through fred_search(), fred_category(),
fred_category_children(), fred_category_series(), fred_releases(),
fred_release_series(), fred_release_dates(), fred_sources(),
fred_source_releases(), fred_tags(), fred_related_tags(), and
fred_updates(). Their print headers now show the endpoint and search
query (where applicable).summary.fred_tbl() prints query metadata, dimensions, date range,
and value range before the standard summary.data.frame output.[.fred_tbl() preserves the fred_tbl class and fred_query
attribute when subsetting.fred_event_window() extracts data inside a c(before, after) day
window around event dates. Works on both long and wide format. Handy for
event studies around FOMC decisions, recession peaks, or release dates.fred_aggregate() aggregates long or wide format data to a coarser
calendar frequency (week / month / quarter / year) using mean, sum,
first, last, median, min, or max. Complements server-side
aggregation in fred_series(frequency = ...).fred_interpolate() fills NA values via last-observation-carry-
forward ("locf") or linear interpolation ("linear"). Useful for
mixed-frequency analysis.plot.fred_tbl() default plot method. Detects long or wide format,
draws one line per series, and shades NBER recession periods. Uses base
graphics: no ggplot2 dependency.fred_cite_series() produces a citation for a FRED series in BibTeX,
plain text, or bibentry form. Works offline (falls back to the series ID
as the title); pass fetch_metadata = TRUE to use the official series
title from fred_info(). Supports vintage-date pinning so cited data is
reproducible even after revisions.fred_manifest() snapshots one or more fred_tbl objects as a YAML
manifest with query metadata, dimensions, date range, and an MD5 hash of
each object. Saving the manifest alongside paper code lets reviewers
verify that the underlying data is unchanged.fred_vintage_revisions() returns per-observation revision summary
statistics (n_vintages, first/final value, total revision, mean/SD of
inter-vintage changes, days to final). Useful for choosing low-revision
series for nowcasting and real-time analysis.multi-series-workflows: fetch, transform, widen, plot.nowcasting-with-fred: pseudo-real-time GDP nowcasting
using monthly indicators, with vintage-aware backtesting (pairs with the
nowcast package).inflation-revisions: tracking core inflation revisions
using fred_real_time_panel() aligned to FOMC SEP meeting dates.CITATION.cff file at the repository root for the GitHub citation
widget and Zenodo deposit workflow.fred_series() gains a format argument. Pass format = "wide" to get
one row per date with a column per series, instead of the default long
layout.fred_series() gains a transform argument with readable aliases for
the FRED units codes: "level", "diff", "yoy_diff", "qoq_pct",
"yoy_pct", "annualised", "log", "log_diff", and more. The raw
units codes still work; transform and units are mutually exclusive.fred_as_of() returns a series as it appeared on a chosen vintage date.fred_first_release() returns only the initial release of each
observation, with no subsequent revisions.fred_all_vintages() returns the full revision history.fred_real_time_panel() returns the values that were available on
each of a chosen set of vintage dates.fred_cache_info() reports the cache directory, file count, total
size, and per-file metadata. Useful for debugging stale results.fred_tbl, a thin
data.frame subclass with a one-line provenance header showing the
query (series count, observation count, units, transform, frequency,
vintage). The header is informational; downstream code can keep
treating the result as a plain data frame.\donttest with tempdir() cache instead of \dontrun,
fixing CRAN policy compliance.options(fred.cache_dir = ...).fred_series() for fetching one or more FRED time series.