Changelog

Source: NEWS.md

arframe 0.1.0.9000 (development version)

New features

• fr_wide_ard() converts ARD (Analysis Results Data) from cards/cardx into wide summary data frames for fr_table(). Supports multi-row continuous output (named vector format specs), per-variable decimal precision, hierarchical SOC/PT structures, custom format functions (including p.value < 0.001 threshold), and all 60+ cards/cardx stat types. Zero dependency on cards — duck-types on column names.

• fr_rows() refactored from 14 params down to 9. group_label moves into group_by list form: group_by = list(cols = "variable", label = "stat_label"). page_by_visible moves into page_by list form: page_by = list(cols = "PARAM", visible = FALSE). page_by_bold, page_by_align, group_bold removed — style via fr_styles() instead.

Breaking changes

• fr_rows(group_by) no longer auto-implies blank_after. To get blank rows between groups, set blank_after explicitly: fr_rows(group_by = "col", blank_after = "col").

Bug fixes

• Fixed decimal alignment across pages: page 2+ was showing page 1’s values because geometry indices used per-page row numbers instead of global indices. All three backends (RTF, HTML, LaTeX) now correctly map page-local rows to global decimal geometry.

• Fixed decimal alignment for mixed integer + decimal columns: columns mixing pure integers (e.g., “143”) with decimal stats (e.g., “75.7 (8.19)”) now align correctly.

• Fixed n_only decimal alignment: widened integer zone for columns with wider n values (e.g., 3-digit counts).

• Added pharma percentage boundary rules for integer percentages (d=0): handles edge cases like “100%” and “0” alignment.

• Unified internal decimal parser: eliminated parallel implementations into a single stat-type detection + alignment engine.

• Injected group header rows and blank-after rows now inherit page_by column values, fixing incorrect page splits when page_by and group_by/blank_after are used together.

Font system overhaul

• Default font changed from Courier New to Times New Roman (FDA recommended for submission documents).
• Font resolution order: FDA-recommended → ARFRAME_FONT_DIR → Adobe open-source (Source Serif 4, Source Sans 3, Source Code Pro) → CSS generic.
• Latin Modern and Liberation fonts removed from fallback chain.
• New vignettes("fonts") documents the full font resolution order and FDA guidance.

HTML backend improvements

• HTML output now respects all fr_spacing() settings: titles_after, footnotes_before, pagehead_after, pagefoot_before, page_by_after.
• col_gap wired to HTML cell padding (matching RTF).
• pagehead/pagefoot font_size applied in HTML output.
• Knitr/pkgdown output uses web-optimized sans-serif font (Source Sans 3).
• Bootstrap table style reset prevents pkgdown from overriding arframe table borders.
• File output uses plain white background (matching R HTML widget preview).
• Fixed pkgdown “On this page” sidebar displacement by adding overflow: hidden to outer container.
• CSS font stack deduplication (no more repeated font names).

New features

HTML render backend

• New HTML output format: fr_render(spec, "output.html") produces a self-contained HTML document with paper simulation (orientation-aware page dimensions, margins, flexbox page footer push-down).

• print(spec) auto-previews in the RStudio/Positron Viewer panel via htmltools::browsable() — just type the spec to see it, like gt.

• knit_print.fr_spec() renders tables inline in R Markdown, Quarto, and pkgdown vignettes with scoped CSS (multiple tables on one page don’t conflict).

• All vignettes now show live rendered tables instead of static PNG screenshots.

Multi-page figures

• fr_figure() now accepts a list of plot objects and an optional meta data frame. Each column of meta becomes a {token} resolved in titles and footnotes dynamically per page, enabling figure programmes that loop over parameters (e.g. one page per biomarker or visit).

Group label injection

• fr_rows(group_by = list(cols = ..., label = ...)) auto-injects a header row at each group boundary using values from a specified column. This is the standard demographics pattern where the group variable label appears as a row above indented summary statistics, without requiring pre-formatted data.

Multi-level indent_by

• fr_rows(indent_by = ) now accepts a named list specifying key, col, and levels for SOC/HLT/PT three-level hierarchies. Each row is indented to its natural depth, replacing the previous single-column form that only handled two levels.

Pagination controls

• fr_rows(group_keep = FALSE) disables RTF \keepn for visual-only grouping (blank rows and indentation without page-keep constraints). Useful when groups are short and the default keep behaviour wastes space.
• fr_rows(page_by = list(cols = ..., visible = FALSE)) inserts page breaks at page_by boundaries while hiding the label row itself.
• fr_page(orphan_min = , widow_min = ) sets minimum body-row counts at the bottom and top of a page respectively, preventing isolated rows at page breaks.

Wrap parameter

• fr_rows(wrap = ) controls whether long cell content wraps within its column or is truncated, complementing the existing suppress behaviour.

Bug fixes

• measure_text_width_twips(): multi-byte UTF-8 characters (em dashes, arrows, etc.) now measured correctly via utf8ToInt() instead of misindexing the raw-byte lookup table.
• resolve_borders(): early return when the table body is empty prevents index-out-of-bounds crashes on zero-row tables.
• escape_and_resolve(): guard against malformed sentinel tokens with fewer than three capture groups (was a subscript-out-of-bounds crash).
• fr_rows(): error when group_by$label is set without group_by$cols.
• fr_rows(): warning when page_by and group_by share the same column.
• LaTeX hline dedup regex anchored with ^ to prevent row 10 matching row 100.
• Decimal alignment: round() replaces as.integer() for center-offset calculation, removing a systematic 1-twip leftward bias.
• paginate_rows(): all consecutive trailing blank rows captured, not just the first.

Pipeline API

• fr_table(), fr_listing(), fr_figure() entry points for tables, patient listings, and embedded figures.
• 10 pipeline verbs: fr_cols(), fr_titles(), fr_footnotes(), fr_header(), fr_rows(), fr_hlines(), fr_vlines(), fr_spans(), fr_styles(), fr_page().
• fr_render() produces RTF and PDF from the same spec.
• Verb order is irrelevant – all resolution deferred to render time.

Column system

• fr_col() constructor with width, alignment, visibility, and grouping.
• Width modes: fixed inches, percentages, "auto" (AFM font metrics), "fit", and "equal".
• Decimal alignment engine (15 stat-display types) for pharma summary tables.
• N-count injection via .n and .n_format in fr_cols().
• Column splitting (.split = TRUE) for wide tables exceeding page width.

Rendering

• Native RTF 1.9.1 backend with field codes for page numbering.
• LaTeX/tabularray backend compiled via XeLaTeX for PDF output.
• R-side pagination with group-aware page breaks.
• Token system ({thepage}, {total_pages}, {program}, {datetime}) for running headers and footers.
• Adobe open-source font fallback for PDF/RTF on Linux/Docker without Microsoft fonts.
• ARFRAME_FONT_DIR environment variable: point to a directory of .ttf/.otf files for project-local fonts without system-wide installation. Ideal for Docker/CI pipelines.

Styling

• fr_style(), fr_row_style(), fr_col_style() for cell-level formatting.
• fr_style_if() for data-driven conditional styles.
• fr_rows_matches() for pattern-based row selectors.
• fr_style_explain() for debugging style cascade resolution.
• Inline markup: fr_super(), fr_sub(), fr_bold(), fr_italic(), fr_dagger(), fr_emdash(), and more.

Configuration

• Four-tier defaults: package < _arframe.yml < fr_theme() < per-table verbs.
• fr_recipe() and fr_apply() for reusable pipeline fragments.
• c.fr_recipe() for composing recipes (company + study + table-type layers).

Validation

• fr_validate() pre-render checks for columns, widths, styles, spans, and fonts.
• fr_get_*() accessors for programmatic QC of spec internals.

Datasets

• Synthetic CDISC ADaM datasets (study TFRM-2024-001, 135 subjects): adsl, adae, adtte, adcm, advs.
• Pre-summarized TFL-ready tables: tbl_demog, tbl_ae_soc, tbl_ae_summary, tbl_disp, tbl_tte, tbl_cm, tbl_vs.