multi-histogram plot with Vegalite

Question

I would like to create single visual showing multiple histograms on it. I have simple arrays of values, like so:

"data": {"values": {"foo": [0,0,0,1,1,1,2,2,2], "baz": [2,2,2,3,3,3,4,4,4]}}

I want to use different color bars to show the spread of values for "foo" and "baz". I am able to make a single histogram for "foo" like so:

{
  "data": {"values": {"foo": [0,0,0,1,1,1,2,2,2]}},
  "mark": "bar",
  "transform": [{"flatten": ["foo"]}],
  "encoding": {
    "x": {"field": "foo", "type": "quantitative"},
    "y": {"field": "foo", "type": "quantitative", "aggregate": "count"}
  }
}

However, I cannot find the correct way to flatten out the arrays. This doesn't work:

{
  "data": {"values": {"foo": [0,0,0,1,1,1,2,2,2], "bar": [0,0,0,1,1,1,2,2,2]}},
  "mark": "bar",
  "transform": [{"flatten": ["foo", "baz"]}],
  "encoding": {
    "x": {"field": "foo", "type": "quantitative"},
    "y": {"field": "foo", "type": "quantitative", "aggregate": "count"}
  },
  "layer": [{
    "mark": "bar",
    "encoding": {
      "y": {"field": "baz", "type": "quantitative", "aggregate": "count"}
    }
  }]
}

https://vega.github.io/editor/#/url/vega-lite/N4IgJghgLhIFygG4QDYFcCmBneoBmA9gfANoAMANJZQIwV10BMFzjAuhSAEYQBep1KvWFMWLNgF8JnALYQATgGt43BSE5R5EAHZZC8maXwpoUDNtIhCxTj36SOIcwGMCYAJbaA5rhAAPXzx3DBQwFWt1ECgATwAHDBUARzQdKHcYNMQE6RBowODQ8KJImPiklO00jPcsyIgvL3kML2gEuBBXNEqQKU4TaIx5IxA5JRUeIc4XN08fBFz8kLD2uxK4tpBk1PToGoTOesbm1pVO7qkJSSA

Inspecting data_0, there is are columns for foo and its counts, but nothing for baz.

This doesn't work, either:

{
  "data": {
    "values": {
      "foo": [0, 0, 0, 1, 1, 1, 2, 2, 2],
      "baz": [0, 0, 0, 1, 1, 1, 2, 2, 2]
    }
  },
  "mark": "bar",
  "transform": [{"flatten": ["foo"]},{"flatten": ["baz"]}],
  "encoding": {
    "x": {"field": "foo", "type": "quantitative"},
    "y": {"field": "foo", "type": "quantitative", "aggregate": "count"}
  },
  "layer": [
    {
      "mark": "bar",
      "encoding": {
        "y": {"field": "baz", "type": "quantitative", "aggregate": "count"}
      }
    }
  ]
}

https://vega.github.io/editor/#/url/vega-lite/N4IgJghgLhIFygG4QDYFcCmBneoBmA9gfANoAMANJZQIwV10BMFzjAuhSAEYQBep1KvWFMWLNgF8JnALYQATgGt43BSE5R5EAHZZC8maXwpoUDNtIhCxSRWOnzlnv0kcQ5gMYEwAS20BzXBAADyC8HwwUMBVrdRAoAE8ABwwVAEc0HSgfGGzEVOkQBLCIqJiiOMSU9MztbNyffLiIf395DH9oVLgQLzQ6kClOEwSMeSMQOSUVHnHOT28-QIQiksjonudK5O6QDKyc6EbUzha2jq6VPoGpCUkgA

That still only gives columns for foo and its count, but now the count is 27 for each bucket!

How can I accomplish a multi-histogram graphic starting with array data?

Answer 1

You can do this using a flatten transform followed by a fold transform, and then use a color encoding to separate the two datasets. For example (open in editor):

{
  "data": {
    "values": {
      "foo": [0, 0, 1, 1, 1, 1, 2, 2, 2],
      "baz": [4, 4, 5, 5, 6, 6, 6, 6, 7]
    }
  },
  "transform": [{"flatten": ["foo", "baz"]}, {"fold": ["foo", "baz"]}],
  "mark": "bar",
  "encoding": {
    "x": {"field": "value", "type": "quantitative"},
    "y": {
      "field": "value",
      "type": "quantitative",
      "aggregate": "count",
      "stack": null
    },
    "color": {"field": "key", "type": "nominal"}
  }
}

As an aside, your layer approach also works if you put the encodings in separate layers, so that the outer foo aggregate doesn't clobber the baz data, but it's a bit more verbose than the approach based on fold:

{
  "data": {
    "values": {
      "foo": [0, 0, 1, 1, 1, 1, 2, 2, 2],
      "baz": [4, 4, 5, 5, 6, 6, 6, 6, 7]
    }
  },
  "transform": [{"flatten": ["foo", "baz"]}],
  "layer": [
    {
      "mark": {"type": "bar", "color": "orange"},
      "encoding": {
        "x": {"field": "foo", "type": "quantitative"},
        "y": {"field": "foo", "type": "quantitative", "aggregate": "count"}
      }
    },
    {
      "mark": "bar",
      "encoding": {
        "x": {"field": "baz", "type": "quantitative"},
        "y": {"field": "baz", "type": "quantitative", "aggregate": "count"}
      }
    }
  ]
}