Skip to contents

Changelog

Source: NEWS.md

sube (development version)

  • Added run_sube_pipeline(), a one-call wrapper that chains import_suts() or read_figaro() → extract_domestic_block() → build_matrices() → compute_sube() (with opt-in estimate_elasticities()) for a single SUT path. Returns a structured sube_pipeline_result object carrying $results, $models, $diagnostics, and $call provenance. See vignette("pipeline-helpers").
  • Added batch_sube(), which loops the convenience pipeline over a pre-imported sube_suts table grouped by country, year, or country-year (default by = "country_year"), returning per-group results alongside merged tidy $summary, $tidy, and $diagnostics tables suitable for downstream analysis. Per-group errors never abort the batch. See vignette("pipeline-helpers").
  • Added the unified pipeline diagnostics layer: run_sube_pipeline() and batch_sube() surface four categories of silent data-quality issues - coerced-NA rows at import, country-years dropped by correspondence-map alignment, singular matrix branches from compute_sube(), and input-metric misalignments from build_matrices() model-data - through a unified $diagnostics data.table and a single summary warning() per run.
  • BREAKING (development contract, INFRA-02): resolve_wiod_root() no longer falls back to inst/extdata/wiod/ when SUBE_WIOD_DIR is unset. The gated replication test now skips cleanly in that case instead of silently using locally-mounted data (which previously caused a known ~4.4% multiplier divergence). Introduced resolve_figaro_root() with the same env-var-only contract (SUBE_FIGARO_DIR). Test helper renamed tests/testthat/helper-replication.R → tests/testthat/helper-gated-data.R.
  • Added end-to-end FIGARO validation coverage: a gated real-data test (SUBE_FIGARO_DIR, DE/FR/IT/NL × 2023) with a testthat golden snapshot, a synthetic-fixture contract test running on every CRAN build, and a new figaro-workflow vignette narrating the full pipeline from flatfile to multipliers. The shipped synthetic fixture under inst/extdata/figaro-sample/ is extended to 8 real FIGARO A*64 codes × 3 countries to exercise a non-degenerate Leontief inversion on every build.
  • Added read_figaro() for importing Eurostat FIGARO industry-by-industry supply and use flat-format CSV files into the canonical sube_suts long table. The importer auto-pairs one supply and one use file from a directory, strips the CPA_ prefix from product codes, filters SNA primary-input rows (B2A3G, D1, D21X31, D29X39, OP_RES, OP_NRES), and aggregates the five FIGARO final-demand codes (P3_S13, P3_S14, P3_S15, P51G, P5M) into a single VAR = "FU_bas" row per (REP, PAR, CPA) so the output feeds build_matrices() unchanged. FIGW1 (FIGARO rest-of-world 1) is preserved as a real country code.
  • Extended .coerce_map() to recognize NACE and NACE_R2 as synonyms for the industry-identifier column, so FIGARO-derived ind_map tables route correctly through build_matrices() without falling through to positional matching.
  • Added inst/extdata/figaro-sample/ with a synthetic 2-country FIGARO fixture and tests/testthat/test-figaro.R covering the new importer end-to-end.
  • Exported filter_paper_outliers() (formerly internal .apply_paper_filters()) with variables and apply_bounds arguments so researchers can apply the paper’s six-layer outlier treatment directly to SUBE comparison or results tables. See ?filter_paper_outliers for the full rule list drawn from the 2024 paper’s outlier treatment.
  • Added the paper-replication vignette: a nine-section walkthrough of the end-to-end reproduction of the 2024 paper’s raw supply, use, and net-supply matrices from WIOD data. Builds with eval = FALSE so it renders cleanly on CRAN.
  • Added the gated tests/testthat/test-replication.R suite (requires SUBE_WIOD_DIR; auto-skipped on CRAN and in CI) which asserts bit-level equality against the 2024 paper ground-truth matrices for AUS, DEU, USA, and JPN in 2005.
  • Fixed the legacy-wrapper subprocess test (test-workflow.R) to pass under R CMD check --as-cran by threading .libPaths() into the child Rscript process via the R_LIBS environment variable (INFRA-01). Previously, the subprocess could not find sube because R CMD check installs to a temporary library not on the child’s default search path.

sube 0.1.2

  • Added explicit Leontief matrix extraction in list, long, and wide formats.
  • Added paper-style comparison, plotting, and export helpers around package objects.
  • Tightened the package-first documentation story around the stabilized workflow surface.
  • Aligned the local release path with GitHub Actions and kept the legacy wrapper as a documented compatibility bridge.

sube 0.1.1

  • Bumped the development version for the next documentation-focused pass.
  • Next release direction: reframe the README around supply-use based econometrics more generally and position sube as a companion package to the Stehrer et al. paper.

sube 0.1.0

  • Converted the repository into a package-first layout with a stable exported workflow API.
  • Made plm a required dependency for pooled and between model estimation.
  • Added CRAN-oriented metadata, release notes, and documentation scaffolding.
  • Added sample-data-driven vignettes and pkgdown configuration.
  • Moved the 2024 paper scripts into a local ignored archive instead of packaging them.