Allow specifying specific order of stack in stacked charts

[kanitw]

We currently do not have a way to specify manual order of stacking group (e.g., color or detail channel) in stacked charts.

Note: As seen in #1732 (comment), users might expect that color's scale domain also affect how stack is ordered.

For example, in the following spec.

{
  "data": {
    "url": "https://raw.githubusercontent.com/dhimmel/biorxiv-licenses/4ad9659e4fc823f4c491e13be4505248df5c1ab6/figure/license-vs-time/vega-lite-data.json"
  },
  "config": {
    "mark": {
      "stacked": "normalize"
    }
  },
  "encoding": {
    "color": {
      "field": "license",
      "scale": {
        "domain": [
          "CC BY",
          "CC BY-ND",
          "CC BY-NC",
          "CC BY-NC-ND",
          "None"
        ],
        "type": "ordinal"
      },
      "type": "nominal"
    },
    "x": {
      "axis": {
        "axisWidth": 0.0,
        "format": "%Y",
        "labelAngle": 0.0,
        "tickSize": 0.0
      },
      "field": "date",
      "scale": {
        "nice": "month"
      },
      "timeUnit": "yearmonth",
      "title": " ",
      "type": "temporal"
    },
    "y": {
      "aggregate": "count",
      "axis": {
        "axisWidth": 0.0,
        "grid": false,
        "labels": false,
        "ticks": 0.0
      },
      "field": "*",
      "title": "Proportion of Preprints",
      "type": "quantitative"
    }
  },
  "mark": "area"
}

The scale domain only determines how the domain values are mapped to first 5 color in the default color palette, but does not affect order of the stacking.

TODOs:

• design how we support manual order for stacking

  • Under the hood, one possible way to do this is to add a formula transform that add index for each domain value and use the index as sortby of stack transform.
  • syntax

• decide if we want to use color/size/detail's sort by default.

cc: @dhimmel

[kanitw]

Workaround

A workaround for now would be to use calculate to derive a new field and map the field to order channel. See https://vega.github.io/vega-lite/docs/stack.html#order

[kanitw]

Potential Syntax

One potential syntax would be to make stack grouping channel (color/opacity/size/detail) supports sort as array of values for custom ordering and then make stack order depends on the grouping channel's order too.

cc: @domoritz, @arvind

[samuelcolvin]

Trying to do this and still having trouble, stack transform isn't working for me.

Is there currently any way of achieving this?

[kanitw]

Have you tried the workaround above?

[samuelcolvin]

Hi, I'm afraid with a stacked area chart that gives us a (very pretty) but completely messed up chart:

[kanitw]

Please include a specification (or link to editor) so one can help you debug

[samuelcolvin]

That is a link to the editor...

[kanitw]

Oh i just realize your image has a link. Thanks!

(Normally people put link on a text so it’s clearer as the text will be blue)

[mattijn]

Took me a long time to figure out that there only exist a workaround for sorting area order.

Also sorting by encoded channel, where the channel eg. color has a custom scale domain would make sense (to me).

Such as:

    "y": {
      "aggregate": "sum",
      "field": "count",
      "type": "quantitative",
      "sort": {"encoding": "color"}
    },
    "color": {
      "field": "series",
      "type": "nominal",
      "scale": {
        "domain": [
          "Construction",
          "Government",
          "Mining and Extraction",
          "Manufacturing"
        ]
      }
    }

editor-spec (it starts with an order-channel included, but upon removal I hoped the order is preserved as it is defined in the color-channel)

[mattijn]

I found another way of doing this:

using color_license_sort_index for order.

{
  "data": {
    "url": "https://raw.githubusercontent.com/dhimmel/biorxiv-licenses/4ad9659e4fc823f4c491e13be4505248df5c1ab6/figure/license-vs-time/vega-lite-data.json"
  },
  "encoding": {
    "color": {
      "field": "license",
      "type": "nominal",
      "sort": [
        "CC BY",
        "CC BY-ND",
        "CC BY-NC",
        "CC BY-NC-ND",
        "None"
      ]
    },
    "x": {
      "axis": {
        "axisWidth": 0.0,
        "format": "%Y",
        "labelAngle": 0.0,
        "tickSize": 0.0
      },
      "field": "date",
      "scale": {
        "nice": "month"
      },
      "timeUnit": "yearmonth",
      "title": "",
      "type": "temporal"
    },
    "order": {
      "field": "color_license_sort_index",
      "type": "quantitative"
    },
    "y": {
      "aggregate": "count",
      "axis": null,
      "type": "quantitative",
      "stack": "normalize"
    }
  },
  "mark": "area"
}

editor-url

[marcprux]

Since the order of the color's domain affect the order of the legend, I think it would make sense for it to also affect the order of the stacking. Take this chart for example:

{
  "$schema": "https://vega.github.io/schema/vega-lite/v4.13.1.json",
  "description": "Stack bar chart.",
  "data": {
    "values": [
      {"a": "A", "b": 28, "c": "X"},
      {"a": "A", "b": 55, "c": "Y"},
      {"a": "B", "b": 55, "c": "Y"},
      {"a": "B", "b": 91, "c": "Z"}
    ]
  },
  "mark": "bar",
  "encoding": {
    "x": {"field": "a", "type": "ordinal"},
    "y": {"field": "b", "type": "quantitative"},
    "color": {
      "field": "c",
      "type": "nominal",
      "scale": {
        "domain": ["X", "Y", "Z"],
        "range": ["darkgray", "orange", "lightgray"]
      }
    }
  }
}

Since the orange bars represent the same value, I decide that I'd like the orange bars to line up at the bottom. Re-ordering the domain & range will make the orange item appear at the bottom of the legend:

{
  "$schema": "https://vega.github.io/schema/vega-lite/v4.13.1.json",
  "description": "Stack bar chart.",
  "data": {
    "values": [
      {"a": "A", "b": 28, "c": "X"},
      {"a": "A", "b": 55, "c": "Y"},
      {"a": "B", "b": 55, "c": "Y"},
      {"a": "B", "b": 91, "c": "Z"}
    ]
  },
  "mark": "bar",
  "encoding": {
    "x": {"field": "a", "type": "ordinal"},
    "y": {"field": "b", "type": "quantitative"},
    "color": {
      "field": "c",
      "type": "nominal",
      "scale": {
        "domain": ["X", "Z", "Y"],
        "range": ["darkgray", "lightgray", "orange"]
      }
    }
  }
}

It might be intuitive that this scale ordering also affect the order of where the orange bars appear in the stack as well, so that orange being bottom-most item in the legend also results in orange being the bottom-most mark in the stacked bars.

[joelostblom]

I agree with @marcprux , it would be intuitive if changing the legend order also changed the stacking order, maybe by automatically doing the calculate transform linked above? Ideally with an optional argument in case the color legend order sometimes needs to be changed independently of the stacking order.

[joelostblom]

Another good reason for automatically ordering according the a custom color sort, is that this is the current behavior if the color sort is set to 'ascending' or 'descending' and it would be intuitive to be consistent with this for custom color sorts. Vega Editor Example

[rben01]

@mattijn

using color_license_sort_index for order.

It took me embarrassingly long for me to realize that license is not a keyword, but rather the name of the color encoding's field