Plotly:如何从单条迹线制作堆积条形图?

问题描述

Is it possible to have two stacked bar charts side by side each of them coming from a single column?

Here's my df:

Field       Issue

Police      Budget cuts
Research    Budget cuts
Police      Time consuming
Banking     Lack of support
Healthcare  Lack of support
Research    Bureaucracy
Healthcare  Bureaucracy
Banking     Budget cuts

I want a stacked bar chart of Field next to a stacked bar chart of issues by field.

Thanks guys!

解决方案

For an example on how to make a stacked bar chart with the specific dataset provided, take a look at the suggestion in Plotly-Dash: Want two stacked bar charts side by side from single df column. The following answer addresses how you can not make a stacked bar chart with a single trace. To show you why, I'll start with what may seem like redundant details. But hopefully everything will be clear in the end.

---

In plotly terms, trace is used to describe a graph object such as Scatter or Bar like in the snippet below.

Snippet 1.1:

import plotly.graph_objs as go
fig = go.Figure(go.Bar(x=['CategoryA', 'CategoryB', 'CategoryC'],
                       y=[20, 14, 23]))
fig.update_layout(template = 'plotly_dark').show()

If you look at the snippet above, you'll see that that is a bar chart with a single trace in the form of go.Bar(x=['CategoryA',....

Plot 1.1:

Now, how can we add something to make it a stacked chart? If you start out by adding an element to x like 'CategoryD', then nothing happens. And that's a bit interesting since you'd might expect an error message instead.

Plot 1.2: Added x value

Snippet 1.2

import plotly.graph_objs as go
fig = go.Figure(go.Bar(x=['CategoryA', 'CategoryB', 'CategoryC', 'CategoryD'],
                       y=[20, 14, 23]))
fig.update_layout(template = 'plotly_dark').show()

But nothing happens before CategoryD has a corresponding y value like 15:

Plot 1.3: Added x and y value

Conclusion so far: Adding values to x and y will add another category on the x-axis and a value on the y-axis. As you can see, nothing is getting stacked here quite yet:

Snippet 1.3:

import plotly.graph_objs as go
fig = go.Figure(go.Bar(x=['CategoryA', 'CategoryB', 'CategoryC', 'CategoryD'],
                       y=[20, 14, 23, 15]))
fig.update_layout(template = 'plotly_dark').show()

But what if you add a layout term with barmode='stack'?

---

Snippet 2:

import plotly.graph_objs as go

fig = go.Figure(go.Bar(x=['CategoryA', 'CategoryB', 'CategoryC', 'CategoryD'],
                       y=[20, 14, 23, 15]))

fig.update_layout(barmode='stack',
                  title = 'Stacked bar chart',
                  template = 'plotly_dark').show()

Plot 2: barmode='stack'

I can understand anyone hoping that this would stack all data within a single trace, but plotly just isn't built that way. To get what you need, you'll have to add another trace using fig.add_trace(go.Bar()) or simply fig.add_bar().

Snippet 3:

import plotly.graph_objs as go
fig = go.Figure(go.Bar(x=['CategoryA', 'CategoryB', 'CategoryC', 'CategoryD'],
                       y=[20, 14, 23, 15]))
fig.add_bar(x=['CategoryA', 'CategoryB', 'CategoryC', 'CategoryD'],
            y=[5, 14, 4, 20])
fig.update_layout(barmode='stack',
                  title = 'Stacked bar chart!',
                  template = 'plotly_dark').show()

Plot 3: Add trace

Notice that neither of the go.Bar() objects have had any name assigned to them, and that plotly by default names them trace 0 and trace 1. So I guess it's more correct to say that a trace contains or 'shows' a plotly graph object rather than calling them the same thing. If you'd like to specify other names, you can do so with, for example, name = 'segment 1' like this:

Snippet 4:

import plotly.graph_objs as go
fig = go.Figure(go.Bar(x=['CategoryA', 'CategoryB', 'CategoryC', 'CategoryD'],
                       y=[20, 14, 23, 15],
                       name = 'segment 1'))
fig.add_bar(x=['CategoryA', 'CategoryB', 'CategoryC', 'CategoryD'],
            y=[5, 14, 4, 20],
            name = 'segment 2')
fig.update_layout(barmode='stack',
                  title = 'Stacked bar chart!',
                  template = 'plotly_dark').show()

Plot 4: Add named traces

If you'd like to 'unstack' your bars, just change barmode to 'group' like this:

fig.update_layout(barmode='group'),

Plot 5: