Expose shape argument in geom_curve()

[fmarotta]

Hi! there was some discussion in issue #5998 about exposing the shape argument in geom_curve(). This PR attempts to do that.

I've run some tests to understand the effect of the curveGrob() arguments, and it seems that elbow-style lines can indeed be achieved with the following combination: curvature = 1, ncp = 1, angle = 90, square = TRUE (see the reprex below).

library(grid)
library(gridExtra)
grobs <- list(
  `curvature=0.1` = curveGrob(x1 = 0.2, x2 = 0.8, y1 = 0.1, y2 = 0.6, shape = 0, curvature = 0.1),
  `curvature=1` = curveGrob(x1 = 0.2, x2 = 0.8, y1 = 0.1, y2 = 0.6, shape = 0, curvature = 1),
  `ncp=5` = curveGrob(x1 = 0.2, x2 = 0.8, y1 = 0.1, y2 = 0.6, shape = 0, curvature = 1, ncp = 5),
  `ncp=1` = curveGrob(x1 = 0.2, x2 = 0.8, y1 = 0.1, y2 = 0.6, shape = 0, curvature = 1, ncp = 1),
  `angle=60` = curveGrob(x1 = 0.2, x2 = 0.8, y1 = 0.1, y2 = 0.6, shape = 0, curvature = 1, ncp = 1, angle = 60),
  `angle=90` = curveGrob(x1 = 0.2, x2 = 0.8, y1 = 0.1, y2 = 0.6, shape = 0, curvature = 1, ncp = 1, angle = 90),
  `square=FALSE` = curveGrob(x1 = 0.2, x2 = 0.8, y1 = 0.1, y2 = 0.6, shape = 0, curvature = 1, ncp = 1, angle = 90, square = FALSE),
  `square=TRUE` = curveGrob(x1 = 0.2, x2 = 0.8, y1 = 0.1, y2 = 0.6, shape = 0, curvature = 1, ncp = 1, angle = 90, square = TRUE)
)
grobs_with_titles <- mapply(function(grob, title) {
  grobTree(
    grob,
    textGrob(title, x=0.5, y=0.7, gp=gpar(fontsize=10))
  )
}, grobs, names(grobs), SIMPLIFY = FALSE)
grid.newpage()
grid.arrange(grobs = grobs_with_titles, ncol = 2, nrow = 4)

^{Created on 2025-06-19 with reprex v2.1.1}

The PR adds the shape argument, but I also had to change the square argument. square is not exposed in geom_curve() but it is hard-coded to FALSE. The grid.curve() documentation says that

When ncp is 1 and angle is 90, this is typically TRUE, otherwise this should probably be set to FALSE.

so in this PR I follow such prescription.

Using square = FALSE results in "tilted" elbows, whereas when square = TRUE the lines are horizontal and vertical (see reprex above). The tilted effect can still be achieved by setting angle to something different than 90.

[teunbrand]

This looks good to me in principle. Would you mind adding an example to the docs using the new parameter (or integrating it in some example)?

We're currently in the process of preparing a release, so we've temporily halted adding new features. If all is well we can consider merging this after the release (and perhaps a hotfix) is out.

[teunbrand]

Can we make this 1 plot with 4 geom_curve() layers, each with something like aes(colour = "not square") that is indicative of the setting, so that it sort of becomes self-descriptive?

[teunbrand]

Also the tests indicate that the snapshot would need updating. Did you have an older svglite/vdiffr version when you generated the snapshots?