inkaR 0.7.0

New Features

• Maps from Eurostat GISCO with INKAR’s territorial status: plot_inkar() now draws boundaries from GISCO LAU 2024 instead of GADM 4.1. GADM uses pre-2016 boundaries, so merged districts (e.g. Göttingen) and about 1% of Gemeinden were missing. With GISCO every region at BND, BLD, KRE, GVB and GEM is drawn (10,978 Gemeinden, 4,594 Gemeindeverbände, 400 Kreise). Higher levels are dissolved from the Gemeinden; Gemeindeverbände use the Destatis Gemeindeverzeichnis (31.12.2023), stored as internal data. The file is downloaded once (about 43 MB) and cached. Maps carry the required “© EuroGeographics” attribution, and plot_inkar() warns when a region has no boundary.
• Indicator metadata synced with the live API: INKAR identifies indicators by the wizard key the API expects (M_ID) and by an official number from its indicator overview, which differ for 554 of 565 indicators. The bundled metadata held both as separate rows, so 400 indicators appeared twice: once usable but undocumented, once documented (short ID such as q_alo, source, English texts) but marked inactive because its number returns no data. These pairs are now merged: 565 active indicators, 411 with their official short ID, and only 2 genuinely discontinued ones. New columns IndID (official number) and Short_DE (official short name). get_inkar_data() accepts the short ID, the M_ID and the official number ("q_alo", "12" and "1101" give the same data), and a name typed in full ("Arbeitslosenquote") wins over longer names containing it. Theme now holds the 17 topic areas of the INKAR wizard, so get_themes() and the theme filters work. Kreise/Gemeinden show the years available at that level instead of the placeholders “Active”/“Unknown”. Description, algorithm and source are filled for all active indicators (English description for about three quarters); 29 more indicators got a unit taken from their name. The three commuter indicators (a_pend50, a_pend150, a_pend300, Gemeinden level) are now active. The metadata is rebuilt by data-raw/sync_indicators_api.R.
• get_inkar_geometry(): new exported function returning the boundaries as an sf object with a Kennziffer column that joins directly with get_inkar_data().
• digits parameter: get_inkar_data(), inkaR(), and inkar() now accept a digits argument to round numeric output values. The INKAR API uses single-precision floats, which can produce artifacts such as 8.44999... instead of 8.45. Setting digits = 2 corrects this transparently (e.g. inkar("12", digits = 2)).

Breaking Changes

• geodata is no longer used: automatic maps no longer depend on geodata/GADM.
• Multi-indicator wide columns are named by short ID: get_inkar_data(c("xbev", "bip"), format = "wide") now returns columns xbev and bip instead of names built from the truncated indicator title (e.g. Bruttoinlandsprodukt_BIP_absolut_in_Mill). The names are the same in lang = "de" and "en".
• Search and region filters match literally: search_indicators(), the name lookup in get_inkar_data(), compare_regions(), compare_districts(), inkar_trends(regions = ) and plot_inkar(highlight = ) no longer interpret their input as a regular expression.
• compare_districts() ID matching: IDs now match as a prefix ("09" keeps all of Bavaria) or exactly with leading zeros ignored (1001 matches "01001"), never in the middle of an ID.

Bug Fixes

• BND and BLD levels returned no data: the INKAR API expects the codes BU (Bund) and BL (Laender). BND/BLD stay the public codes and are translated internally, so get_inkar_data(), get_geographies(), select_level() and select_years() now return data for both levels. The misleading “Most INKAR indicators are only available at KRE or GEM level” message has been removed.
• Missing indicator names in downloads: Indikator/indicator_name was NA for every indicator whose API key differs from its official number (554 of 565), because the name was joined on the official number. It is now joined on the API key.
• update_indicators() never worked: it sent an empty indicator list to Wizard/GetMöglich, which the API answers with HTTP 500, so it always reported “Could not reach INKAR API”. It now reads the wizard’s topic areas (Wizard/GetBereiche) and their indicators (Wizard/GetIndikatorenZuBereich), lists indicators missing from the local metadata and attaches them as attribute "new_indicators"; active local indicators the API dropped are reported as "removed_indicators". The API’s European (NUTS-2) indicators are ignored, since the package supports German levels only.
• plot_inkar(breaks = "quantile"): quantiles were only used as legend ticks on a continuous scale, so colors stayed linear and the labels overlapped. The map now uses five quantile classes with one color step each. Legend values are no longer shown in scientific notation (3e+06).
• plot_inkar() level detection: "Bundesländer" data was detected as Bund and drawn as a single national polygon; level detection now matches whole labels. GEM maps matched no region, because they were joined against GADM’s Gemeindeverband layer.
• Regex metacharacters: search_indicators("Bevölkerung (mit") and the same text in get_inkar_data() failed with “invalid regular expression”; compare_regions(df, "Frankfurt (Oder)") silently returned no rows.
• Name lookup prefers active indicators: a name that matches active and inactive indicators resolves to the active one. The inactive warning now points to search_indicators().
• Wizard input: in the terminal wizard, a multi-word name such as Total population was split into invalid IDs, and a short input such as 1 silently selected indicator 12. Names are now matched as a whole, several matches narrow the list, and inputs under 3 characters are not fuzzy-matched.
• Skipped indicators in multi-indicator downloads: an indicator that is ambiguous or has no data now triggers a warning naming it. Previously a NULL result deleted its slot in the result list, so it was dropped silently.
• Misleading warnings after a failed request: when the API cannot be reached, get_inkar_data() no longer adds “API response missing expected columns” and “No data found for this variable/level combination” after the connection message.
• Input validation: get_inkar_data() stops with a clear error for NA, "" or character(0); get_geographies() rejects unknown level codes.
• Umlauts in labels: the Kürzel column of view_indicators("de") and the level menu of select_level() showed <U+00FC>-style escapes when the package was installed in a non-UTF-8 locale.
• Bundled indicators table not found after library(inkaR): with LazyData: true the dataset lives in the lazy-data environment, so exists("indicators", envir = asNamespace("inkaR")) was always FALSE outside devtools::load_all() and testthat. As a result get_indicators() returned 0 rows, search_indicators() errored, get_themes() was empty, select_level() / select_years() skipped ID resolution, and get_inkar_data() sent short IDs such as "bev_korr" to the API unresolved. All sites now use an internal inkar_indicators() helper. A regression test runs in a fresh R process via callr (added to Suggests).
• Unresolvable indicator IDs: get_inkar_data() now stops with a clear error (and suggestions when stringdist is installed) instead of passing an unknown non-numeric ID to the API and returning an empty result. Numeric M_IDs not in local metadata are still passed on.
• Clearer SSL error: when certificate verification fails, the message now explains that www.inkar.de does not send its intermediate certificate and how to supply a CA bundle via CURL_CA_BUNDLE.
• view_indicators() DE: dplyr::any_of() was incorrectly used as a rename target inside dplyr::select(). The Statistische Grundlagen column is now normalized with an explicit dplyr::rename() before the select() call, removing the silent column-drop.
• utils::adist() invalid argument: method = "osa" was passed to utils::adist(), which does not accept a method parameter. The argument has been removed from all call sites in search_indicators() and select_indicator().
• dplyr::pick() outside data-masking context: Description/unit lookup in get_inkar_data() called dplyr::pick() inside an if() condition, which is only valid inside mutate() / summarise(). Replaced with pre-computed has_desc_en / has_unit_en logical flags evaluated before the mutate().
• Duplicate M_ID column after EN name join: The English indicator name lookup joined a table that included M_ID, creating a spurious M_ID.y column. The column is now dropped from the lookup before the join.
• get_time_references() schema guard: Added an explicit check for required API response columns (Gruppe, IndID, RaumID, ZeitID) before calling dplyr::select(), producing an informative warning instead of a cryptic error when the API changes its response shape.
• api.R redundant requireNamespace("httr2"): Removed the runtime check; httr2 is a hard Imports dependency and is always available.
• Content-type check: Changed != "application/json" to !startsWith(..., "application/json") so that responses with a charset=utf-8 suffix no longer trigger a spurious warning.
• globalVariables false entry: select_level is an exported function, not a free variable; removed from the globalVariables() declaration.
• data.R Roxygen syntax: Fixed a stray blank line before the closing } of the \describe{} block in the indicators dataset documentation.

Enhancements

• get_geographies() caching: Region lists fetched from the API are now cached for 7 days, eliminating a redundant network round-trip on every get_inkar_data() call.
• plot_inkar() year type coercion: The year parameter is now coerced with as.integer() before filtering, so year = 2021L, year = 2021, and year = "2021" all produce identical results.
• inkar_trends() wide-format guard: Passing wide-format data (year columns) to inkar_trends() now raises an informative error suggesting format = "long" instead of a generic column-not-found error.
• Numeric format heuristic: The value parser in parse_inkar_json() now distinguishes German locale formatting (1.234,56 → 1234.56) from English decimal notation (1.5) by checking for the co-occurrence of both . and , in the vector, preventing inadvertent corruption of English-formatted values.
• DESCRIPTION: Removed unused lifecycle entry from Suggests.

Testing

• Added 21 new unit tests covering: compare_districts() with German-language data, inkar_trends() with German-language data, wide-format error message, plot_inkar() with long-format English data and custom sf geometry, plot_inkar() year parameter type coercion, clear_inkar_cache() file-deletion verification, and the new digits rounding parameter.
• Replaced non-standard skip_on_ci() calls with skip_if(nzchar(Sys.getenv("CI"))).
• Wrapped search_indicators() and view_indicators() calls in capture.output() to suppress cat() noise during test runs.
• Total: 68 test blocks with 147 expectations; live API and download tests are skipped on CRAN, on CI and when the server is unreachable.

inkaR 0.6.6

New Functions

• compare_districts() / compare_district(): Filter downloaded datasets by district IDs (Kennziffer) or names.
• compare_region(): Singular alias for compare_regions().

Enhancements

• plot_inkar() upgrades:
  • Added support for BND (Germany level 0) and GEM (Municipalities level 3) boundaries.
  • Added geom parameter to plot data on custom sf boundaries (e.g. ROR spatial planning regions).
  • Added a viridis map output image to the introduction vignette.
• Base R Fuzzy Fallback: Added a fallback using built-in utils::adist() when the stringdist package is not available.
• CRAN Policy compliance: Relocated usage history tracking from ~/.inkaR_history to tools::R_user_dir("inkaR", which = "config") in interactive sessions, and disabled it entirely in non-interactive sessions.
• API stability: Added a null-response check to parallel request loops to handle timeouts gracefully.
• Documentation: Built a complete pkgdown package documentation website.

inkaR 0.6.5

New Functions

• inkar(): English-first shortcut alias for get_inkar_data() with lang = "en" and year = "latest" as defaults.
• compare_regions(): Filter a downloaded data frame to specific regions by partial name match.
• inkar_trends(): Plot time series as a ggplot2 line chart for selected regions.
• get_themes(): List unique indicator themes from the local metadata for use with theme filtering.

Enhancements

• year = "latest": get_inkar_data() now accepts year = "latest" to automatically download only the most recent available year.
• plot_inkar() improvements: New highlight, breaks (equal/quantile), and title parameters for richer map customization.
• search_indicators() / view_indicators(): New theme parameter to pre-filter by indicator domain (see get_themes()).
• pkgdown support: Added _pkgdown.yml for documentation website generation.

inkaR 0.6.4

New Features

• Smart ID resolver: get_inkar_data() now resolves indicator IDs through five fallback tiers (exact, numeric, normalized, partial name, fuzzy).
• Schema guarantee: All 9 expected output columns are always present, even when the API returns partial data.
• Readable wide-format column names: Multi-indicator wide-format output now uses indicator names (e.g., Gross_domestic_product) instead of opaque IDs (ind_11).
• Retry & timeout: API requests now include a 30-second timeout and automatic retry.
• Progress bar: Multi-indicator downloads show a cli progress bar.
• Improved fuzzy search: search_indicators() now compares against Name_EN only and uses a tighter Jaro-Winkler threshold.
• update_indicators() stability: Gracefully handles API 500 errors instead of crashing.
• BLD-level warning: Informative message when BLD-level data is unavailable for a given indicator.

inkaR 0.6.2

Bug Fixes

• CRAN Archive Fix: get_cache_dir() no longer writes to ~/.cache/R/inkaR/ during non-interactive sessions (e.g. R CMD check). In non-interactive environments the cache now uses tempdir() (cleaned up at session end), while interactive sessions continue to use the persistent tools::R_user_dir("inkaR", which = "cache") directory. This resolves the NOTE that caused the package to be archived on CRAN (checking for new files in some other directories).

inkaR 0.6.0

New Features

• Interactive Selection Wizard: Calling inkaR() without arguments now opens a professional CLI wizard to guide users through selecting an indicator, then an available spatial level, and finally a year or year range.
• Analytical Multi-Indicator Downloads: Users can now pass a vector of IDs (e.g., c("bip", "xbev")) or enter multiple IDs in the wizard (separated by spaces).
• Multi-Indicator Joins: Automatically handles the merging of multiple indicators.
  • format = "long" (default) stacks indicators in a tidy format.
  • format = "wide" pivots indicators into individual columns for direct comparison and mathematical operations.
• Fuzzy Search Engine: Indicator selection now supports error-tolerant search using the stringdist package (Jaro-Winkler method).

CLI Enhancements

• Professional Table Display: Rewrote indicator selection and search results using the cli package for high-end, paginated, and responsive terminal tables.
• Usage History & Favorites: The package now tracks frequently and recently used indicators, highlighting them in green and pinning them to the top of selection lists.
• Intelligent Year Selection: Implemented a professional 4-column layout for year selection with validated user input.

Professional Visualizations

• Premium Themes: Introduced theme_inkaR(), a high-end ggplot2 theme with dedicated light and dark modes.
• Upgraded Mapping: plot_inkar() now supports bilingual labels, automatic unit extraction, and professional color palettes (Viridis, Magma).

inkaR 0.4.4

• FIXED: Added graceful handling for SSL certificate verification issues on specific CRAN builders (e.g., Fedora).
• IMPROVED: Wrapped network-dependent examples in try() to ensure CRAN checks pass even with environment-specific network limitations.

inkaR 0.4.3

• FIXED: Addressed CRAN feedback regarding single quotes in DESCRIPTION.
• FIXED: Added missing \value tags to exported .Rd files (e.g., clear_inkar_cache()).
• FIXED: Replaced \dontrun{} with \donttest{} or if(interactive()) to ensure proper example testing.
• FIXED: Ensured CSV exporting examples do not write to the user home directory (tempdir()).

inkaR 0.4.2

Enhancements

• Added persistent disk-caching (tools::R_user_dir) to heavily improve GetGebieteZumRaumbezug lookup times across sessions.
• Implemented parallel API dispatching (httr2::req_perform_parallel) for exploring spatial levels, eliminating the sequential wait time.
• Introduced plot_inkar() for native geospatial mapping with ggplot2 and sf by directly downloading and matching GADM polygon data.

Bug Fixes

• Fixed missing mock payloads in testthat by completely regenerating httptest2 records.
• Avoided decimal conversion errors (NA) by implementing robust locale-agnostic numeric parsing.
• Addressed spell checking NOTEs caught by win-builder.