Higher resolution / programmatic 3D plots

[trevorcampbell]

From @ttimbers : I CANNOT programmatically save a nice, high resolution view of a 3D plotly scatter plot. Current one in PDF is a screen shot. This might be helpful: https://stackoverflow.com/questions/63460719/plotly-how-to-set-manual-zoom-for-camera-view-in-a-3d-plot. If we can figure that out, we need to add orca to the Dockerfile so that we can save plotly plots to .png for PDF version.

[ttimbers]

This will change the plot3d object in classification1.Rmd camera angle to be nice:

scene = list(camera = list(eye = list(x=2, y=2, z = 1.5)))
plot_3d <- plot_3d  |> 
  layout(title = "changed zoom", scene = scene)

and then we can programmatically save it to a svg via:

orca(plot_3d, "3d-knn-classification.svg")

However, to do this we need to install orca and the processx R package... I figured out how to do this on my mac, will try for the container now...

[ttimbers]

OK, orca is depracted, we use reticulate + the python-kaleido package.

Install via:

install.packages('reticulate')
reticulate::install_miniconda()
reticulate::conda_install('r-reticulate', 'python-kaleido')
reticulate::conda_install('r-reticulate', 'plotly', channel = 'plotly')
reticulate::use_miniconda('r-reticulate')

Then use this R code:

save_image(plot_3d, "img/3d-knn-classification.svg")

[ttimbers]

So I think this code will do it:

# add this in setup chunk
reticulate::use_miniconda('r-reticulate')

# use this control flow in plot chunk to conditionally save 
# high resolution image if plotting in LaTeX and then load it 
if(!is_latex_output()){  
  plot_3d
} else {
  scene = list(camera = list(eye = list(x=2, y=2, z = 1.5)))
  plot_3d <- plot_3d  %>% layout(scene = scene)
  save_image(plot_3d, "img/test.png", scale = 10)
  knitr::include_graphics("img/plot3d_knn_classification.png")
}

[ttimbers]

Here's what it looks like now:

[ttimbers]

This is resolved for the regression plots sufficiently, however the marker size is too large via the method I am using for the plot in classification1.Rmd. So my workaround for now is a screenshot. Leaving this open until we fully resolve it...