Fix misalignment in geom_dotplot when stackratio != 1 and stackdir != "up"

[mjskay]

This PR fixes #4614.

The problem

For geom_dotplot() / the dotstack grob, the (1 - x$stackratio) / 2 adjustment on these lines...

ggplot2/R/grob-dotstack.r

Lines 34 to 42 in c89c265

	if (x$stackaxis == "x") {
	dotdiamm <- convertY(x$dotdia, "mm", valueOnly = TRUE)
	xpos <- xmm + dotdiamm * (x$stackposition * x$stackratio + (1 - x$stackratio) / 2)
	ypos <- ymm
	} else if (x$stackaxis == "y") {
	dotdiamm <- convertX(x$dotdia, "mm", valueOnly = TRUE)
	xpos <- xmm
	ypos <- ymm + dotdiamm * (x$stackposition * x$stackratio + (1 - x$stackratio) / 2)
	}

...is there because otherwise when stackratio != 1 and stackdir = "up", the base of the first dot in a stack will no longer touch the baseline. This is because stackratio will expand (or contract) the dot stack away (or towards) the origin.

However, currently this adjustment only works correctly for stackdir = "up". For example, here is stackdir = "down" (for other examples, see #4614):

ggplot(data.frame(x = c(rep(1, 3), rep(2, 2))), aes(x)) +
  geom_dotplot(binwidth = 0.5, alpha = 0.5, stackdir = "down", stackratio = 1.5) +
  coord_fixed() +
  ylim(-2, 2) +
  xlim(0, 6)

Notice how the top dot in each stack does not touch y = 0 as it should.

The solution

The solution is that the (1 - x$stackratio) / 2 adjustment needs to depend on stackdir:

• For "up", it should remain (1 - x$stackratio) / 2
• For "down", it should be reversed: -(1 - x$stackratio) / 2
• For "center" and "centerwhole", it is unnecessary, and should be 0.

This PR implements that change, and includes test cases to ensure the adjustment is correct when stackratio < 1 and when stackratio > 1.

[hadley]

Thanks for working on this 😄. As well addressing the comments, can you please add a bullet to the top of NEWS.md? It should briefly describe the change and end with (@yourname, #issuenumber).

[hadley]

Can you please make this a separate test? (The existing huge test unfortunately isn't a good pattern to follow). And can you please also create the smallest possible dataset needed to demonstrate the problem?

[mjskay]

Sure, done --- I pulled out the four tests for stackratio = 1.5 into a separate test with a single doppelganger on a simplified dataset that should be sensitive to problems with any of the four values of stackdir. Dunno if you would rather those be four separate small doppelgangers or one merged doppelganger as I have done.

[hadley]

I think one test with non-default stack ratio would be adequate. Visual tests are expensive to maintain so we really want to keep them as minimal as possible.

[mjskay]

Done.

[mjskay]

Happy to, thanks for the feedback! I've added a NEWS entry and updated the tests based on your comments.

[hadley]

Visual test is perfect, thanks!