rKolada 0.3.1

Breaking changes

magrittr pipe removed: The %>% pipe is no longer re-exported. Use the native R pipe |> instead (available since R 4.1.0, which is already the minimum version for rKolada). If you rely on %>% in your own code, add library(magrittr) or switch to |>.

kolada_cache_dir(): New exported function returning the persistent cache directory path (using tools::R_user_dir()).
kolada_clear_cache(): New exported function for clearing cached files.
Lightweight kolada_available(): Now performs a fast HTTP check (single request, 5s timeout) instead of downloading actual data. Much faster in examples and tests.
HTTP retry with backoff: API requests now automatically retry on HTTP 429 (rate limiting) with exponential backoff, up to 3 retries.

Improved quickstart vignette: explains the three-dimension data model, enriched metadata entity descriptions, clearer simplify = TRUE explanation, inline ggplot comments, and “K”-filter explanations.
Improved introduction vignette: prominent Swedish-only note, inline ggplot comments, clearer groups/unnest explanation, and ggplot2 reference link.

glue replaced by cli: All string interpolation now uses cli for user-facing messages and paste0() for URL construction. glue is no longer a dependency.
magrittr removed: All internal code uses the native pipe |>. magrittr is no longer a dependency.
Structured messaging: message(), warning(), and stop() replaced with cli::cli_inform(), cli::cli_warn(), and cli::cli_abort() throughout.
Roxygen2 markdown: Documentation now uses markdown syntax (e.g., [func()] instead of \code{\link{func}}).
kolada_get() internal helper: Centralised HTTP request logic with tryCatch(), timeout, and retry.
Removed legacy CRAN-RELEASE file.
Removed legacy test_api.R test stub.

KPI field rename: The auspices column returned by get_kpi() is now named auspice (matching the v3 API schema).
ou_publication_date removed: This field is no longer returned by the v3 API and has been removed from kpi_describe() output.
is_divided_by_gender: Now returns a logical (TRUE/FALSE) instead of integer (0/1).
remove_undocumented_columns removed: kpi_minimize() no longer accepts the remove_undocumented_columns parameter as the API has been updated with proper documentation for all data properties.

Kolada API v3: All API calls now use https://api.kolada.se/v3/ instead of http://api.kolada.se/v2/. The v2 API is being shut down on March 31, 2026.
get_municipality() gains a region_type parameter to filter by municipality type (e.g., "K" for municipalities, "L" for regions).
get_values() gains a from_date parameter to fetch only data updated after a specific date.
get_values() gains a keep_deleted parameter (default FALSE). Rows marked as deleted in the v3 API are filtered out by default.
Pagination now uses the v3 next_url field and supports up to 5000 results per page (up from 2000).
kolada_available() is now more resilient to v3 response shape changes.

httr replaced by httr2: The HTTP backend has been migrated from httr to httr2. This is an internal change with no user-facing API differences.
urltools removed: No longer a dependency.
Deprecated dplyr::filter_at() and dplyr::select_if() replaced with modern dplyr::if_any() / dplyr::all_of() equivalents.
New internal entity_search() helper eliminates code duplication across kpi_search(), municipality_search(), and ou_search().
New internal append_query_params() helper replaces urltools::param_set().
Deprecated .data$ pronoun usage in tidyselect contexts replaced with string column names.
Comprehensive test suite added: 140+ tests covering query composers, utilities, search functions, KPI/municipality/group operations, mocked API responses, and live integration tests.
Test infrastructure uses testthat edition 3.
get_values() and get_metadata() now automatically chunk oversized path parameters into batches of 25 to respect the Kolada v3 API limit. Previously, passing more than 25 KPIs, municipalities, or periods would return an error. Chunking is transparent — results are combined into a single tibble.

Add kolada_available() for programming with the rKolada package
Fix vignettes and examples so they don’t crash when the Kolada API is unavailable

add support for paging. This implies that queries that produce a result with more than 5,000 rows are no longer truncated
vignette polish
graceful fail on no internet connection (and other HTTP problems)
add max_results argument to all get_ functions