[joethorley]

Releases ssdtools 2.7.0. Closes #482.

The 2.7.0 version bump and full CRAN release prep were added in poissonconsulting#192, so this PR now carries the whole release rather than the ggtext change alone.

Why this is time sensitive

On 22 August the CRAN team gave notice that ggtext is scheduled for archival on 2026-09-12, and that archiving it would necessitate archiving its strong reverse dependencies. ssdtools 2.6.0, the version currently on CRAN, has ggtext in Imports. So 2.7.0 needs to be accepted, not merely submitted, before that date.

Removing ggtext

ggtext had a single call site, axis.text.x = ggtext::element_markdown() in ssd_plot(), needed only because ssd_label_comma_hc() emitted markdown (<br>**x**) to put the hazard concentration break on its own line in bold.

• ssd_label_comma_hc() now prefixes the hazard concentration label with a newline rather than marking it up as markdown.
• ssd_element_text_hc() is a new exported theme element that draws multi-line labels in bold. ssd_plot() uses it for axis.text.x.
• ggtext removed from Imports, with nothing added in its place.

The element is fifteen lines. It receives the whole label vector at draw time, computes the fontface vector from the labels, and hands it to ggplot2's own element_text() path, where gpar recycles fontface across labels. Bolding the whole label is harmless because its first line is empty. All margin, justification and rotation handling stays in ggplot2.

ssd_element_text_hc() is exported so that callers of ssd_label_comma_hc() in their own plots have a replacement for the role ggtext::element_markdown() played.

Bug fixed along the way

ssd_label_comma_hc() compared a formatted string against a coerced number, so any big.mark or decimal.mark that altered the formatting stopped the hazard concentration from being matched, leaving it unbolded and inline. It is now formatted with the same labeller before comparison.

This was visible in a checked-in snapshot: _snaps/ssd-plot/boron_decimark.png previously rendered 1-26 plain and inline, colliding with the neighbouring 0-1 label.

Appearance

Measured against the previous ggtext output, in the 400x400 snapshot:

	ggtext	this
hazard concentration label	rows 375-380, cols 167-182	identical
ordinary label row	rows 359-364	rows 358-363
bold on quartz png()	yes	yes
bold on cairo and ragg	yes	yes

18 of the 36 plot snapshots are byte-identical to the ggtext originals and so do not appear in the diff: all 6 autoplot, all 10 plot-cdf, match-moments/cdf and ssd-plot/boron_nohc. The 18 that changed are exactly the plots carrying a hazard concentration label, and differ only by a one pixel shift of the ordinary label row.

Alternatives considered

marquee was tried and rejected: it ignores <br>, collapses single newlines and drops a leading empty paragraph, so the label collides with its neighbour unless a visible placeholder holds the first line open, and bold renders only on cairo and ragg, not on the macOS quartz device that grDevices::png() defaults to. Vendoring ggtext was also rejected: it delegates to gridtext::richtext_grob(), so it would swap ggtext for gridtext rather than removing the dependency, and would mean copying about 157 lines including two ggplot2-private helpers plus MIT attribution into an Apache-2.0 package.

Caveat

The element relies on element_grob S3 dispatch over ggplot2's S7 element classes, the same compatibility shim ggtext itself uses. If ggplot2 removes it this breaks, but it is fifteen lines under our own control.

Release prep

Version bumped to 2.7.0 and the four development NEWS sections consolidated into one. CRAN submission fixes:

Item	Change
Remotes: open-AIMS/ssddata	Removed. ssddata 2.0.0 is on CRAN, and the submitted 2.6.0 DESCRIPTION carried no Remotes field either
Config/roxygen2/version: 8.0.0.9000	Re-documented with the released roxygen2 8.1.0. NAMESPACE unchanged
Description:	Posthuma et al.(2001) -> et al. (2001)
tests/testthat/test-hc.R	structure(list(), .Names = ...) -> names =. R-devel notes the deprecated special name; the two forms are identical()
vignettes/references.bib	One http URL -> https
cran-comments.md	Was stale at 2.3.0. Rewritten for 2.7.0

VGAM also moved from Suggests to Imports. In 2.6.0 sgompertz() opened with rlang::check_installed("VGAM"), which prompts interactively; that guard is gone, so VGAM is now a hard dependency and fitting the Gompertz distribution no longer prompts.

Checks

• win-builder R-devel (r90452 ucrt): Status: OK, 0 notes
• CI: 9/9 pass, covering ubuntu devel / release / oldrel-1, windows, macOS, plus check-no-suggests
• R CMD check --as-cran: 0 errors, 0 warnings, 0 notes
• revdepcheck::revdep_check() against CRAN 2.6.0: GenHMM1d 0.2.6 and serocalculator 1.4.1 both clean. 0 new problems, 0 failures
• urlchecker::url_check(): all 67 URLs pass

🤖 Generated with Claude Code