testing that two ggplots are the same

[jgabry]

This is a follow up to r-lib/waldo#56, which made more sense to ask here than in the waldo repo.

Currently vdiffr::expect_doppelganger() accepts a single ggplot object as input and compares to an SVG, but is there a way (or would you consider adding a way) to use vdiffr to test that two ggplot objects correspond to the same looking plot? I guess that would require either comparing them both to the same saved SVG or comparing separate SVG files created for each plot to each other.

Here's one example of a use case: in the bayesplot package we want to test that two different plotting functions create the same ggplot in the special case where the input matrix has 1 row:

# with exactly 1 row in `yrep` matrix these should produce the same plot
plot1 <- ppc_scatter_avg(y, yrep[1,, drop=FALSE])
plot2 <- ppc_scatter(y, yrep[1,, drop = FALSE])

Is there a recommended way of using vdiffr to compare plot1 to plot2 or, if not, would you consider supporting this?

Thank you!

[lionel-]

I don't think this is in scope for vdiffr. The approach seems brittle because the graphics engine might take two plots that look exactly the same but draw them in different ways.

[jairav]

@jgabry did you come up with an alternate solution for this?

[shug0131]

A possible way to achieve this is by using the same title argument to expect_doppelganger

fig1<- plot_function1()
vdiffr::expect_doppelganger("plot", fig1)
fig2<- plot_function2()
vdiffr::expect_doppelganger("plot", fig2)

This seems to have the desired behaviour when I run the test. It passes if fig1 and fig2 are the same, and fails if they are different.