Plot with two DataFrames

sehHeiden

I am searching for version to plot two dataframes with the same colon name into a layer each. I tried it in VegLite Json:

{
  "$schema": "https://vega.github.io/schema/vega-lite/v5.json",
  "data": {
    "values": [
      {"x": 1, "y": 10},
      {"x": 2, "y": 20},
      {"x": 3, "y": 15}
    ],
    "name": "data1"
  },
  "layer": [
    {
      "data": {
        "values": [
          {"x": 2, "y": 30},
          {"x": 3, "y": 25},
          {"x": 4, "y": 35}
        ],
        "name": "data2"
      },
      "mark": "line",
      "encoding": {
        "x": {"field": "x", "type": "quantitative"},
        "y": {"field": "y", "type": "quantitative"},
        "color": {"value": "blue"}
      }
    },
    {
      "data": {"name": "data1"},
      "mark": "line",
      "transform": [
        {"filter": {"field": "x", "range": [2, 3]}}
      ],
      "encoding": {
        "x": {"field": "x", "type": "quantitative"},
        "y": {"field": "y", "type": "quantitative"},
        "color": {"value": "red"}
      }
    }
  ],
  "resolve": {
    "scale": {
      "x": "shared",
      "y": "independent"
    }
  }
}

What I came up with in Elixir draws the seconds DataFrame so that, it is WWWWAY to high over the correct place and the y range is evenly wrongly scaled:

  def create_compare_graph(data_a, data_b, title, col) do
    Vl.new(width: 500, height: 300, title: title)
    |> Vl.data_from_values(data_a, 
                          only: ["day", col])
    |> Vl.layers([
      Vl.new()
      |> Vl.mark(:line, color: "blue")
      |> Vl.encode_field(:x, "day",
                         type: :quantitative,
                         title: "Day of the year"
        )
      |> Vl.encode_field(:y, col, 
                        type: :quantitative,
                        scale: [domain: [0, 1]], 
                        title: "DF1"),
      Vl.new()
      |>Vl.data_from_values(data_b,
                            only: ["day", col])
      |> Vl.mark(:line, color: "red")
      |> Vl.encode_field(:x, "day", 
                         type: :quantitative, 
                         title: "Day of the year")
      |> Vl.encode_field(:y, party,
                         type: :quantitative,
                         scale: [domain: [0, 1]],
                         title: "DF2")
    ])
    |> Vl.resolve(:scale, y: :independent)
  end
end

Example Data:

DF1:
day value
| 250 | 0.125
| 251 | 0.13
| 252 | 0.13
| 256 | 0.14
| 257 | 0.14

DF2:
day value
| 250 | 0.36
| 251 | 0.36
| 252 | 0.36
| 256 | 0.36
| 257 | 0.36

If possible I would prefer to do it, in Tucan as the code locks much cleaner with it.

4 comments

/nx #vegalite #tucan

0 570 4

2023-11-05 15:59:10 UTC

First 4 of 4 Posts

sehHeiden

A reply to myself.
VegaLite-datasets_from_values allows several sources.
I tried to apply it with:

def create_compare_graph(data_a, data_b, title, col) do
    Vl.new(width: 500, height: 300, title: title)
    |> Vl.data_from_values(a: data_a, 
                           b: data_b,
                           only: ["day", party])
    |> Vl.layers([
      Vl.new()
      |> Vl.data(name: "a")
      |> Vl.mark(:line, color: "blue")
      |> Vl.encode_field(:x, "day",
                         type: :quantitative,
                         title: "Day of the year"
        )
      |> Vl.encode_field(:y, party, 
                        type: :quantitative,
                        scale: [domain: [0, 1]], 
                        title: "Sentiment"),
      Vl.new()
      |> Vl.data(name: "b")
      |> Vl.mark(:line, color: "red")
      |> Vl.encode_field(:x, "day", 
                         type: :quantitative, 
                         title: "Day of the year")
      |> Vl.encode_field(:y, party,
                         type: :quantitative,
                         scale: [domain: [0, 1]],
                         title: "Polls")
    ])
    |> Vl.resolve(:scale, y: :independent)
  end
end

Error:

expected valid tabular data, but got ... (*Desribes the dataframe*)

Edited: the error Because chcnage Vl.Data → Vl.data

Post #1

awerment

Are the domains in the real data significantly different? If so, does setting the scale resolution to :shared help? (|> Vl.resolve(:scale, y: :shared)

Without a legend, independent scales can look confusing, although the chart is correct:

With :shared:

data_a = DF.new(day: 251..260, value: 1..10)
data_b = DF.new(day: 251..260, value: 4..13)

Post #2

sehHeiden

Sorry,
I did have a data error. And did not change the varible. Therefore they ware out of scale.

I also applied Tucan. Which greatly reduces the amount of code.

def create_compare_graph(sentiment_data_source, poll_data_source, title, party) do
  Tucan.layers([
  Tucan.lineplot(poll_data_source, "day", party, line_color: "red"),
  Tucan.lineplot(sentiment_data_source, "day", party, line_color: "green")
])
|> Tucan.set_width(500)
|> Tucan.set_title(title)
|> Tucan.Axes.set_x_title("Day of the year")
|> Tucan.Axes.set_y_title("Sentiments, Polls")
|> Tucan.Scale.set_y_domain(0,1)
end

The only thing I would like to add is a Legend.

Post #3