Skip to contents

Changelog

Source: NEWS.md

ggconsort 0.4.0

Breaking changes

  • ggconsort now uses the base pipe |> throughout and no longer imports or re-exports magrittr’s %>%. Code that relied on ggconsort providing %>% should attach magrittr (or dplyr) itself. ggconsort now requires R >= 4.1.

  • In a row/column layout, row gaps are equalized but capped at 2 lines and the diagram is vertically centered, so large devices produce a compact CONSORT-style figure instead of a stretched one. Set geom_consort(row_gap =) for explicit control.

  • consort_stage_add() badges default to a new "margin" column, resolved at plot time to one column left of the leftmost box — stage labels sit in the margin as in the official CONSORT and PRISMA templates, and never collide with boxes. Pass an explicit col for the previous behavior.

New features

  • consort_box_add() labels a box automatically when its name matches a cohort: the box text becomes the cohort’s label and count, as formatted by cohort_count_adorn(), so most boxes need only a name and a grid position.

  • New cohort_count_bullets() builds multi-line box labels — a header cohort followed by bulleted detail cohorts — replacing hand-written glue()/<br> blocks for CONSORT exclusion boxes and PRISMA reasons boxes.

  • geom_consort() gains styling parameters: family (font), linewidth (box borders, with arrows scaled in proportion), box_r (corner radius), box_padding, arrow_length (arrow head size), row_gap, and equal_columns (draw every box in a column at the width of the column’s widest box, matching the uniform boxes of the official templates).

  • consort_box_add() gains per-box fill, color, and text_color overrides for highlighting individual boxes.

  • consort_stage_add() accepts a length-2 col to span columns, e.g. col = c("main", "side") for a PRISMA 2020 header bar.

  • trial_data gains follow-up and analysis indicators (lost_to_followup, discontinued, not_analyzed) so a complete four-stage CONSORT diagram can be built from the bundled data.

Bug fixes

  • cohort_label() now errors informatively when a label is not a single (non-NA) string, instead of accepting values that fail later at plot time.

Documentation

  • New pkgdown article building a complete four-stage CONSORT diagram (Enrollment, Allocation, Follow-up, Analysis) with margin stage labels.

  • The PRISMA article now includes the official template’s spanning header bar and uniform column widths; README and reference examples use the automatic box labels and cohort_count_bullets().

ggconsort 0.3.0

New features

  • consort_box_add() gains a row/column layout: declare row and col ("main", "side", "left", or a number) instead of x/y coordinates, and the diagram is laid out at draw time from the measured box sizes — row gaps are equalized, columns are spread so nothing overlaps or clips, and the layout adapts to the device size. Explicit coordinates remain fully supported.

  • In a row/column layout, consort_arrow_add() needs only start and end box names: boxes in the same column connect vertically, boxes in the same row horizontally, boxes elsewhere by a branch off the start box’s column. A vector end (e.g. end = c("arm_a", "arm_b")) draws the classic T-split into study arms in one call.

  • New consort_stage_add() adds stage badges (“Allocation”, “Identification”, …) to a row/column layout, vertically centered on a row or a span of rows, with optional rotation (angle = 90) for PRISMA-style margin labels.

  • geom_consort() is now a single layer that measures each box on the open graphics device at draw time. Boxes are placed exactly at their (x, y) coordinates (centered by default, or per user-supplied hjust/vjust), and arrows start and end precisely at the measured box edges instead of being anchored at box centers. Diagrams no longer depend on justification tricks to fake edge attachment, and connections stay correct at any device size. The previous layers remain available as geom_consort_box() and geom_consort_arrow().

  • Boxes that receive an arrow are no longer shifted by inferred justification; a box entered from the left keeps its position and only left-aligns if you ask for hjust = 0. Set hjust/vjust explicitly where the old inference was load-bearing.

ggconsort 0.2.0

New features

  • consort_box_add() gains optional hjust and vjust arguments to set the text justification of an individual box, overriding the default inference from connected arrows (#24).

  • geom_consort_box() (and therefore geom_consort()) now applies its label_size, label_color, and label_height arguments, so box text can be resized and recolored, and forwards ... to ggtext::geom_richtext() (#18).

  • cohort_count_adorn() formats counts with comma separators by default, e.g. "Randomized (n = 5,932,291)". Use the .label_fn argument for custom formats (#23).

  • ggsave() now works directly on a ggconsort object via a new grid.draw() method (#26).

  • New pkgdown article showing how to build a PRISMA 2020 flow diagram (#14).

Bug fixes

  • create_consort_data() no longer errors on a diagram with no arrows or no lines (#21, #22, thanks @jrealgatius).

  • Arrows and lines use linewidth instead of the size aesthetic deprecated in ggplot2 3.4.0; ggconsort now requires ggplot2 >= 3.4.0 (#28).

  • plot() on a ggconsort object now passes its margin_h and margin_v arguments through to theme_consort().

  • A box with several incoming arrows is no longer drawn once per arrow; its text justification is inferred across all entry sides. create_consort_data() also gives an informative error when called on anything other than a ggconsort object.

Documentation

ggconsort 0.1.0

  • Initial GitHub release.