Hi

During our work on Positron and the graphics pane there we stumbled upon a peculiarity in tinyplot that may end up affecting how well it is supported in Positron. To give you a bit of background I'll briefly describe how plotting works in Positron:

When you open up the default device in Positron it will, under the hood open up a png() device. This device captures all the plotting that happens, but the rendering it does is not used for anything. Instead, the graphic commands that are sent to it are captured using recordPlot(). Once it is time to show something on the screen, or export a plot, the recorded plot is replayed (using replayPlot() on a graphics device that has been created with the correct dimensions etc. This is also how positron makes sure that plots continue to look good as the plot pane is resized etc.

This approach generally works well. Paul Murrell has done a lot of work to ensure the stability of recordPlot(). However, if the code that produces the plot makes any assumptions about the size/aspect ratio of the graphics device and hard codes this into the calls to graphics functions then the plot will look wrong if it is replayed on a device that doesn't match the dimensions of the device where it was recorded.

I found out that tinyplot does this, at least for the legend placement which means that the distance between the plot and the legend is aspect ratio dependent during replays. The lines responsible for this are

At least based on my understanding of the source code. There may be other places where this happens but the legend placement was by far what was most apparent.

I'm afraid that my poor understanding of the tinyplot codebase means that I can't provide a good fix for this. Perhaps wrapping it all in recordGraphics() will be enough.

I hope this report is useful