Is there a way to create a categorical heatmap?

[JoshMcRose]

I want to create something like the bistro.corr graphs, but where two columns in my data frame are defined as x and y and the fill color is defined as the count between them. Is this possible, or is there a way to go about making this work with an existing plot type?

[alshan]

Try to use geomTile().
See an example in cell [4] here.

[JoshMcRose]

This ended up being exactly the answer. Thank you again!

[JoshMcRose]

Hey Alshan,

Thanks for getting back to me!

I've been trying geomTile, but I've run into a few issues with getting it to work. I've come back to it this evening, and I'm getting an interesting error (attached) that I really don't understand.

A snippet of my code:

  val z = data["questionText"].values().zip(data["answerOptions"].values())

  letsPlot(data.toMap()) {
      x = "questionText"
      y = "answerOptions"
      fill = z
  }
      .plus(geomTile())
      .plus(scaleFillHue())
      .show()

data["questionText"] and data["answerOptions"] are both columns of strings. To get my z values, I've zipped them into a list of pairs. This might be where I'm going wrong, but I'm not sure how else to make fill represent the count of the relationship between x and y.

If you have any advice, I'd be all ears.

[alshan]

Interesting issue. Try to add .map { it.toString() } to the z expression to make it a list of strings.

[JoshMcRose]

Like this? : val z = data["questionText"].values().zip(data["answerOptions"].values()).map { it.toString() }

It returned the same error.

[alshan]

It seems there is an issue related to geomTile() and categorical x/y. I'll need to take closer look.

[JoshMcRose]

Sure, thank you for looking into it!

[alshan]

Hi, it appears I'm not able to reproduce this exact issue :( Could you provide a minimal code to reproduce?

[JoshMcRose]

Hey Alshan,

I was putting together my data in redacted form to share with you along with my code, and the issue disappeared. It seems that the the string data in the "questionText" and "answerOptions" columns was too long. By adding + ggsize(width = 1200, height = 750), the error went away.

Thank you for the help!

[alshan]

I see, I'd been experimenting with tiles+small ggsize too, but in a jupyter notebook which is likely not the setup you have. I'll try in Swing/Batik later.
This issue is something that certainly must be fixed.
Thanx for reporting it!