Skip to contents

Changelog

dfdiffs 2.2.0

PROC COMPARE parity: dfdiffs now reports the headline summary statistics and variable-structure diffs that PROC COMPARE prints before its detailed value comparison, and closes a gap where the app’s download handler had drifted from the package functions it should have been calling.

  • New compare_columns(): reports which columns are only in base, only in compare, or common to both, with a print.dfdiffs_column_diff() console method.

  • New compare_summary(): row/column/value headline counts (common, new, and deleted rows; common and dataset-only columns; differing values and class mismatches), with a print.dfdiffs_summary() console method. Built entirely on compare_data() and compare_columns(), no new comparison logic.

  • compare_values() gains ignore_case/trim_ws arguments for character comparisons, and diff_values_by_key() now warns (not errors) when by doesn’t uniquely identify rows in base or compare – both threaded through create_changed_data(), create_modified_data(), and compare_data().

  • class_diffs (already computed by diff_values_by_key(), but previously discarded) is now surfaced through create_changed_data(), create_modified_data(), and compare_data() (as changed_class_diffs).

  • create_comparison_report() writes three new xlsx sheets – “Column Diffs”, “Class Diffs”, and “Summary” – alongside the existing six.

  • The vignettes were rewritten. The repeated walkthroughs of all ten by/by_col/cols combinations in create-new-data, create-deleted-data, and changed-modified-data are now one reference table plus three worked examples each, and modified-data was merged into changed-modified-data. multiple-comparisons now iterates compare_data() instead of arsenal/diffdf, and similar-work compares against them directly. Document-wide error = TRUE was removed, which had been hiding a missing dplyr import in similar-work and a stale file path in change-frequency-from-proc-compare.

  • The app now uses the “Golden Age Pulp Cover” palette (cream page, navy/red/amber accents), the same one as the pkgdown site, in place of “Atomic Age Space Race”. dfdiffs_fresh_theme(), the navbar, the results-tab icons, and every *_react_theme() (light and dark) were updated.

  • mod_compare.R’s download handler no longer reimplements create_comparison_report()’s logic by hand; it now calls create_comparison_report() directly, which also fixes the downloaded report being missing the “Base Data”/“Compare Data” sheets the package function already produced.

dfdiffs 2.1.0

The comparison engine (create_new_data(), create_deleted_data(), create_changed_data(), create_modified_data(), and the helpers behind them) is now built entirely on base R. dplyr, tidyr, tidyselect, and diffdf are no longer required to use the package; Imports drops from 16 packages to 11.

  • New base-R helpers replace the dplyr/diffdf call sites: select_cols(), anti_join_base(), left_join_base() (in R/join_helpers.R), and compare_values()/diff_values_by_key() (in R/compare_values.R), the direct replacement for diffdf::diffdf().

  • Value comparison is more accurate than before, not just equivalent: compare_values() does real numeric-tolerance comparison (ported from diffdf’s own formula), treats a Date and a POSIXct representing the same instant as equal (an explicit improvement, since diffdf itself treats that as a class mismatch), and compares factors by label. Previously, both create_changed_data() and create_modified_data() coerced every column to character before comparing, so none of this tolerance or type-awareness was ever actually exercised.

  • diffdf, dplyr, and tidyr moved to Suggests (a few vignettes demonstrate them directly as points of comparison, matching how arsenal was already handled); tidyselect and stringr were dropped entirely.

  • Fixed two latent cross-file NAMESPACE coupling bugs that this change surfaced: mod_select.R/mod_compare.R’s bare select() calls and launch_select_demo.R’s bare str_detect() call were only working because an unrelated file happened to declare the matching @importFrom tag.

  • Fixed a data.table compatibility bug in the new base-R helpers. Because data.table’s [ has different indexing semantics than a plain data.frame, every helper now coerces its input with as.data.frame() first.

dfdiffs 2.0.0

Major overhaul of the app-package: bslib UI, fewer dependencies, new example data, and a more complete test suite.

  • App migrated from bs4Dash to bslib (page_navbar(), input_dark_mode()), with modules renamed to mod_upload/mod_select/mod_compare and entry points renamed to launch_app()/launch_app_dev()/launch_*_demo(). A custom “Atomic Age Space Race” theme applies to both the app and the pkgdown site.

  • Every reactable table in mod_upload/mod_select/mod_compare now follows the navbar’s dark-mode toggle (react_themes.R’s themes became dark-aware functions).

  • The upload/select/compare workflow now auto-advances between steps as the user completes them, with a “Step X of 3” indicator in the navbar. fluidRow()/column() grids were replaced with bslib::layout_columns(), and the results tables can go full-screen.

  • mod_display (unwired, redundant with mod_select’s preview) was removed, along with its demo app and tests.

  • Dependencies cut from ~34 Imports to ~17: dropped unused packages entirely (hrbrthemes, RColorBrewer, vroom from Imports), rewrote small call sites to drop magrittr, forcats, and rlang, and moved vignette-only packages (flextable, kableExtra, gtsummary, gt, glue, labelled, readr, lubridate, vetr, waldo, vctrs, fs) to Suggests. Added ggplot2 to Imports (it was already used but undeclared). create_modified_data()/create_comparison_report() now build on diffdf::diffdf() instead of arsenal::comparedf(), so the package no longer needs both comparison engines; arsenal was dropped.

  • All 8 core example datasets (T1Data, T2Data, NewData, CompleteData, IncompleteData, DeletedData, InitialData, ChangedData) now carry a clinical-observation story instead of placeholder pangram text. The Sean Lahman baseball archive data (~96MB across data/ and inst/extdata/, undeclared and unattributed) was replaced by a synthetic, package-generated “site roster” dataset family (Roster2021-Roster2024) used across the compare-data, create-comparison-report, create-join-column, upload-data, and multiple-comparisons vignettes. 11 orphaned/unused datasets were removed.

  • Fixed: ggplot2 was used in mod_compare.R without being declared in DESCRIPTION. Fixed: R/_diffdf.R, a dead, unexported, broken vendored copy of diffdf::diffdf(), was removed. Fixed: a create-join-column.Rmd vignette chunk locally redefined create_join_column() with stale parameter names, shadowing the real function so the vignette never actually exercised it.

  • New tests for compare_data(), create_comparison_report(), the xlsx upload path in mod_upload_server(), and a build smoke test for every launch_*() entry point. Filled in missing @param/ @return documentation on the module UI/server functions.

  • Known issue (not fixed in this release): create_new_column() errors with sep must be a single string, not absent when compare_data()/create_comparison_report() are called with a by_col argument or a multi-column by.

dfdiffs 0.1.0

dfdiffs 0.0.0.9000

2022-07-16

2022-07-03

  • updated all functions to use base/comp instead of prev/curr and olddf/newdf.

  • new function create_changed_data() for changed/modified data based on the diffdf package.

  • uploadDataDemo deployed dev sheet

    • dev folder contains uploadData module

  • New vignettes:

    • getting started

    • create new data

    • create deleted data

    • create modified data

    • create changed data

    • similar work