Skip to contents

Changelog

Source: NEWS.md

hdatools 0.5.0

  • Fixed an unwanted stroke outline on geom_col()/geom_bar() bars: brand themes (theme_hda(), theme_hfv(), theme_pha(), theme_vha()) now set colour = NA in their ggplot2::element_geom() default instead of the fill color, so column/bar charts no longer render a visible border (closes #21).

  • get_output_format() now distinguishes "typst" and "docx" output from "pdf" instead of folding all three into "pdf".

  • The “Using branded themes in hdatools” article no longer requires a Census API key or network access to build — it now uses a small bundled data table instead of a live tidycensus::get_acs() call.

  • Replaced the showtext/sysfonts font stack with systemfonts (ragg moved to Suggests, as a recommended rendering device — hdatools itself never calls it). register_hda_fonts() now registers the bundled Lato, Roboto Slab, Open Sans, Poppins, Noto Sans, and Montserrat faces via systemfonts::register_font(); the hdatools.fonts opt-out still works the same way. Consumers rendering with knitr/Quarto must add dev: "ragg_png" under knitr: opts_chunk: in _quarto.yml (see README.md) — the default Cairo device does not consult the systemfonts registry, so without it the bundled fonts won’t appear in rendered output. hdatools no longer sets knitr::opts_chunk$set(fig.showtext = TRUE) as a load-time global side effect.

  • Removed hda_pal_discrete(), hfv_pal_discrete(), pha_pal_discrete(), scale_color_gradient_hda(), scale_colour_gradient_hda(), scale_color_gradient_pha(), scale_colour_gradient_pha(), and scale_fill_gradient_pha() — all soft-deprecated in 0.3.0. Use the discrete scales (scale_color_hda() / scale_fill_hda() etc.) and continuous ramp scales (scale_color_hda_c() / scale_fill_hda_c() etc.) introduced in 0.3.0 and 0.4.0 respectively.

hdatools 0.4.0

  • New hda_focus_pal()/hfv_focus_pal()/pha_focus_pal()/vha_focus_pal() return a focus/emphasis palette for “highlight one series, mute the rest” charts: element 1 is the named brand color hex, elements 2–n are the brand’s neutral gray. Pass directly to scale_fill_manual(values = …) or scale_colour_manual(values = …).

  • New hda_span()/hfv_span()/pha_span()/vha_span() wrap a text string in <span style='color:#…'>…</span> using a named brand color, for use with ggtext::element_markdown() and ggtext::element_textbox(). Color is resolved through the per-brand .*_color() accessor, so invalid names error with the valid list.

  • New pkgdown article “CVD accessibility audit” (vignettes/articles/cvd-audit.Rmd) documents pairwise perceptual-distance (delta-E, CIE76) results for all four brand palettes under full-severity protanopia, deuteranopia, and tritanopia. Key findings: HDA Green vs Sea Green collapses to delta-E ≈ 6 under tritanopia (positions 2 and 6; use a secondary encoding if both must appear in a tritanopia-sensitive chart); HFV Sky vs Grass is borderline (~12 delta-E) under all CVD types; PHA Orange vs Red — flagged during design review — passes all CVD types (≥ 22). No palette order was changed (Q7, plans/DECISIONS.md, 2026-07-18). New tests/testthat/test-cvd.R adds colorspace::simulate_cvd()-based regression assertions that guard minimum pairwise delta-E for the first four slots of each brand palette.

  • Added VHA as a fourth first-class brand: theme_vha(), scale_color_vha()/ scale_fill_vha() (+ scale_colour_vha() alias), the full scale_*_vha_c()/scale_*_vha_b() continuous/binned matrix, and vha_colors/vha_color() — all generated from a single .brands$vha registry entry (R/brands.R), with no VHA-specific code in R/scales.R/ R/themes.R. Montserrat (OFL) is bundled alongside the existing fonts. See the new pkgdown article “Adding a new client brand to hdatools” (vignettes/articles/adding-a-brand.Rmd) for the general pattern this proves out. VHA’s diverging ramp is provisional — it pairs Dark Turq against Yellow, the palette’s only warm hue, but Yellow’s natural HCL lightness is too high to survive as a dark, saturated anchor, so that arm renders golden/olive rather than bright yellow (an sRGB gamut limit, not a tuning slip); candidate for a follow-up Ramp Lab pass, same as HDA’s diverging ramp.

  • theme_hda()/theme_hfv()/theme_pha() alone now brand a plot with no scale_*() call required, via ggplot2 4.0’s theme-carried palettes: each theme sets palette.colour.discrete/palette.fill.discrete to the brand’s full discrete palette and palette.colour.continuous/ palette.fill.continuous to the brand’s sequential ramp (the same ramps behind scale_color_hda_c() and friends). An explicit scale_*() call (or a manual aes() value) always overrides these theme-carried defaults, so every existing plot that already sets its own scale is unaffected.

  • The same three themes now set a default geom fill/colour (via ggplot2::element_geom()), so a bare geom_bar()/geom_col()/ geom_point()/geom_line() with no fill/colour mapping at all renders in the brand’s first palette color instead of ggplot2’s stock grey/black.

  • Raised the ggplot2 dependency floor to >= 4.0.0 (from >= 3.5.0), the version required for the theme-carried palette/element_geom() features above.

  • New scale_color_hda_c()/scale_colour_hda_c()/scale_fill_hda_c() (and the matching hfv/pha versions) — a full continuous color/fill scale matrix (9 exports) built from six colorspace HCL sequential/diverging ramps tuned and CVD-checked (protanopia/deuteranopia/tritanopia) in the Ramp Lab review (plans/ramp-lab/REVIEW.md). Each takes palette = c("sequential", "diverging") to choose the ramp, plus direction, na.value, and guide. Sequential ramps default to higher value = darker color; diverging ramps default to the ramp’s own low-to-high arm order as reviewed. HDA’s diverging ramp (Blue vs Coral) is provisional — it’s a near-twin of PHA’s and is pending a follow-up Ramp Lab pass to differentiate it before final adoption (plans/DECISIONS.md, 2026-07-18).

  • New scale_color_hda_b()/scale_colour_hda_b()/scale_fill_hda_b() (and the matching hfv/pha versions) — the binned counterpart of the above (9 exports), defaulting to n.breaks = 7, the class count every ramp was tuned and CVD-checked against.

  • 7-class diverging maps built from any of the six ramps lose sign distinction in their innermost class pair under protanopia (structural to the shared cream center) — always pair a palette = "diverging" map with a legend or direct labels (documented on every new scale’s help page).

  • scale_color_gradient_hda(), scale_color_gradient_pha(), and scale_fill_gradient_pha()’s existing soft-deprecation notices now name their replacement (e.g. scale_color_hda_c()) via lifecycle’s use_instead, now that one exists.

  • "Leaf" (#6fb547) and "Cerulean" (#7fc7e0) join hfv_colors and .brands$hfv$palette as official HFV secondary colors — both HFV ramps are built on Leaf, and Cerulean was approved alongside it at Ramp Lab sign-off.

  • HFV gets its first na.value default (#d6dadd), needed now that it has continuous scales for the first time; HDA/PHA already had their own (#cfcfd0/#e2e4e3).

  • colorspace joins Imports (used only internally by the new ramp scales).

  • New hda_colors, hfv_colors, pha_colors — exported named character vectors of each brand’s palette hex values, sourced directly from the internal .brands registry. Downstream consumers can now reference brand hexes by name (e.g. hda_colors["Blue"]) or pass the full vector to scale_fill_manual(values = hda_colors) without hardcoding hex strings.

  • New hda_color(), hfv_color(), pha_color() — per-brand accessors that return the hex for a named color and error with the list of valid names on an unknown input.

hdatools 0.3.0

  • New scale_colour_hda(), scale_colour_hfv(), scale_colour_pha(), scale_colour_gradient_hda(), and scale_colour_gradient_pha() — British- spelling aliases of the existing scale_color_*() exports, for parity with ggplot2’s own colour/color aliasing. No change to any existing export.

  • hda_pal_discrete(), hfv_pal_discrete(), pha_pal_discrete(), scale_color_gradient_hda(), scale_color_gradient_pha(), and scale_fill_gradient_pha() are now soft-deprecated (lifecycle:: deprecate_soft()); they keep working exactly as before, but calling them directly will surface a one-time deprecation notice. lifecycle moves to Imports.

  • Internal-only: the three brands’ discrete palettes, gradient stops, NA colours, and select theme parameters (base_size, html_adjust, pdf_adjust, lineheight) now live in one internal registry (R/brands.R), and the 9 exported discrete/gradient scales are now thin wrappers over two shared internal constructors. No behavior change; every pre-refactor identity test asserting exact hex/parameter values passes unmodified.

  • Internal-only: theme_hda(), theme_hfv(), and theme_pha() are now thin wrappers over one shared internal builder (.brand_theme() in R/themes.R), reading the font/margin/size parameters that differ per brand from the R/brands.R registry. No behavior change; signatures and ... passthrough are unchanged, and every pre-refactor theme-identity test passes unmodified.

  • pkgdown site rebuilt on Bootstrap 5 (template: bootstrap: 5); reference index now groups exports into Themes, Palettes & scales, Helpers, and Analysis utils. Fixed “depreciated” → “deprecated” wording in the space parameter docs for scale_color_gradient_hda(), scale_color_gradient_pha(), and scale_fill_gradient_pha(). No behavior change.

  • Internal cleanup only, no behavior changes: enabled Roxygen markdown so help pages and the pkgdown site render code/bold correctly, corrected the get_logo() @return description, dropped unused namespace imports, added flip_gridlines()/get_logo()/get_output_format()/adjust_base_size() to the README, removed the dead internal add_google_fonts() alias, and made the package startup message fire only when font registration actually succeeds.

hdatools 0.2.0

Breaking changes

  • add_reliability() gains a cv_col argument. Supplying it switches to a new classification path: tidy-eval column selection, a scale argument ("percent", the default, or "proportion"), and <= boundaries (a CV of exactly 15 percent is now “High”). Omitting cv_col keeps the original behavior byte-for-byte: auto-detect a single column ending in _cv, treat it as a 0-1 proportion, and use strict < boundaries.
  • flip_gridlines()’s size argument is deprecated in favor of linewidth, matching ggplot2’s own size → linewidth rename.
  • ggiraph moves from Imports to Suggests. publish_plot() now errors with an install hint if ggiraph isn’t installed, instead of failing to load the package at all.

New features

  • Fonts are bundled and registered offline. inst/fonts/ ships the exact static TTF faces theme_hda(), theme_hfv(), and theme_pha() use (Lato, Roboto Slab, Open Sans, Poppins, Noto Sans), registered by the new exported register_hda_fonts() — no more per-session network download from Google Fonts, and no more silent failure when a render happens offline. Opt out with options(hdatools.fonts = FALSE) or the HDATOOLS_NO_FONTS environment variable. See inst/fonts/LICENSES.md for the OFL and Apache 2.0 texts covering each face.
  • theme_pha() reaches parity with theme_hda()/theme_hfv(): new output_format, html_adjust/pdf_adjust, and ... passthrough arguments, and a strip.text element so markdown_wrap_gen() facet labels render correctly. html_adjust/pdf_adjust default to 0/0, so existing theme_pha() output is unchanged. theme_hda()/theme_hfv() also gain explicit html_adjust/pdf_adjust parameters (defaults match prior hardcoded behavior).
  • get_logo() now resolves its image path from the installed package (system.file()) instead of a hardcoded relative path, so it works correctly once the package is installed rather than only from source.

Modernization for ggplot2 4.0

  • Dropped the deprecated positional scale_name argument from all discrete and gradient scales.
  • Renamed every deprecated size= to linewidth= in the themes and in add_zero_line().
  • Raised the ggplot2 floor to >= 3.5.0 (required for the scale_name deprecation shape) and added floors to all other Imports, chosen below every known consumer’s lockfile.

Documentation & housekeeping

  • Fixed a broken add_reliability() README example (it passed 10 values into an 8-row data frame and relied on _cv auto-detection against a column literally named cv).
  • Documented that strip.text overrides on any theme must use ggtext::element_markdown(), not a raw ggplot2::element_text() — ggplot2 4.0 only merges theme elements of the same class.
  • Slimmed the namespace: replaced blanket @import package imports with qualified pkg:: calls throughout, keeping only %+replace% and rlang’s tidy-eval helpers as bare imports.
  • Fixed the Authors@R given/family name order for all three authors. Jonathan Knopf is now the maintainer; Kyle Walker moves to aut.
  • Added URL/BugReports, dropped Date/LazyData, and removed the legacy Author: field.
  • Removed the dead data-raw/DATASET.R stub and moved vignettes/branded-themes.Rmd to vignettes/articles/ (pkgdown builds it as a long-form article; it was never wired up as a real package vignette).

hdatools 0.1.7

  • See git history.

hdatools 0.1.6

  • See git history.

hdatools 0.1.5

  • See git history.

hdatools 0.1.4

  • See git history.

hdatools 0.1.3

  • See git history.

hdatools 0.1.2

  • See git history.

hdatools 0.1.1

  • See git history.

hdatools 0.1.0

  • Merges updates in jtk branch to original build-out in main.
  • Adds new color scale functions, including scale_color_gradient_*.
  • Adds new color palette (pha_pal_discrete()) for the Partnership for Housing Affordability (PHA).
  • Theme helper functions now load more Google Font options and include flip_gridlines() shortcut.
  • hda_pal_discrete() now includes repeat_pal option to loop colors for data with more than six variables.