Plotly：如何使用 python 创建子图？

Question

I am wondering what is best practice to create subplots using Python Plotly. Is it to use plotly.express or the standard plotly.graph_objects ?

I'm trying to create a figure with two subplots, which are stacked bar charts. The following code doesn't work. I didn't find anything useful in the official documentation. The classic Titanic dataset was imported as train_df here.

import plotly.express as px

train_df['Survived'] = train_df['Survived'].astype('category')
fig1 = px.bar(train_df, x="Pclass", y="Age", color='Survived')
fig2 = px.bar(train_df, x="Sex", y="Age", color='Survived')

trace1 = fig1['data'][0]
trace2 = fig2['data'][0]

fig = make_subplots(rows=1, cols=2, shared_xaxes=False)
fig.add_trace(trace1, row=1, col=1)
fig.add_trace(trace2, row=1, col=2)

fig.show()

I got the following figure:

What I expect is as follows:

Answer 1

From what I know, it's not possible to subplot stakedbar (because stacked bar are in facted figures and not traces)...

On behalf of fig.show(), you can put to check if the html file is okay for you (The plots are unfortunately one under the other...) :

with open('p_graph.html', 'a') as f:
    f.write(fig1.to_html(full_html=False, include_plotlyjs='cdn',default_height=500))
    f.write(fig2.to_html(full_html=False, include_plotlyjs='cdn',default_height=500))

try the code below to check if the html file generate can be okay for you:

import pandas as pd
import plotly.graph_objects as go

#Remove the .astype('category') to easily 
#train_df['Survived'] = train_df['Survived'].astype('category')
Pclass_pivot=pd.pivot_table(train_df,values='Age',index='Pclass',
                   columns='Survived',aggfunc=lambda x: len(x))
Sex_pivot=pd.pivot_table(train_df,values='Age',index='Sex',
                   columns='Survived',aggfunc=lambda x: len(x))

fig1 = go.Figure(data=[
    go.Bar(name='Survived', x=Pclass_pivot.index.values, y=Pclass_pivot[1]),
    go.Bar(name='NotSurvived', x=Pclass_pivot.index.values, y=Pclass_pivot[0])])

# Change the bar mode
fig1.update_layout(barmode='stack')


fig2 = go.Figure(data=[
    go.Bar(name='Survived', x=Sex_pivot.index.values, y=Sex_pivot[1]),
    go.Bar(name='NotSurvived', x=Sex_pivot.index.values, y=Sex_pivot[0])])
# Change the bar mode
fig2.update_layout(barmode='stack')

with open('p_graph.html', 'a') as f:
    f.write(fig1.to_html(full_html=False, include_plotlyjs='cdn',default_height=500))
    f.write(fig2.to_html(full_html=False, include_plotlyjs='cdn',default_height=500))

Answer 2

I'm hoping that the existing answer suits your needs, but I'd just like to note that the statement

it's not possible to subplot stakedbar (because stacked bar are in facted figures and not traces

is not entirely correct. It's possible to build a plotly subplot figure using stacked bar charts as long as you put it together correctly using add_trace() and go.Bar() . And this also answers your question regarding:

I am wondering what is best practice to create subplots using Python Plotly. Is it to use plotly.express or the standard plotly.graph_objects?

Use plotly.express ff you find a px approach that suits your needs. And like in your case where you do not find it; build your own subplots using plotly.graphobjects .

Below is an example that will show you one such possible approach using the titanic dataset. Note that the column names are noe the same as yours since there are no capital letters. The essence of this approav is that you use go.Bar() for each trace, and specify where to put those traces using the row and col arguments in go.Bar() . If you assign multiple traces to the same row and col , you will get stacked bar chart subplots if you specify barmode='stack' in fig.update_layout(). Using fig.update_layout(). Using px.colors.qualitative.Plotly[i]` will let you assign colors from the standard plotly color cycle sequentially.

Plot:

Code:

from plotly.subplots import make_subplots
import plotly.graph_objects as go
import plotly.express as px
import pandas as pd

url = "https://raw.github.com/mattdelhey/kaggle-titanic/master/Data/train.csv"
titanic = pd.read_csv(url)
#titanic.info()
train_df=titanic
train_df

# data for fig 1
df1=titanic.groupby(['sex', 'pclass'])['survived'].aggregate('mean').unstack()

# plotly setup for fig
fig = make_subplots(2,1)
fig.add_trace(go.Bar(x=df1.columns.astype('category'), y=df1.loc['female'],
                     name='female',
                     marker_color = px.colors.qualitative.Plotly[0]),
    row=1, col=1)


fig.add_trace(go.Bar(x=df1.columns.astype('category'), y=df1.loc['male'],
                     name='male',
                     marker_color = px.colors.qualitative.Plotly[1]),
    row=1, col=1)


# data for plot 2
age = pd.cut(titanic['age'], [0, 18, 80])
df2 = titanic.pivot_table('survived', [age], 'pclass')
groups=['(0, 18]', '(18, 80]']

fig.add_trace(go.Bar(x=df2.columns, y=df2.iloc[0],
                     name=groups[0],
                     marker_color = px.colors.qualitative.Plotly[3]),
    row=2, col=1)

fig.add_trace(go.Bar(x=df2.columns, y=df2.iloc[1],
                     name=groups[1],
                     marker_color = px.colors.qualitative.Plotly[4]),
    row=2, col=1)

fig.update_layout(title=dict(text='Titanic survivors by sex and age group'), barmode='stack', xaxis = dict(tickvals= df1.columns))
fig.show()

fig.show()

Answer 3

I managed to generate the subplots using the add_bar function.

Code:

from plotly.subplots import make_subplots

# plotly can only support one legend per graph at the moment.
fig = make_subplots(
    rows=1, cols=2,
    subplot_titles=("Pclass vs. Survived", "Sex vs. Survived")
)
fig.add_bar(
    x=train_df[train_df.Survived == 0].Pclass.value_counts().index,
    y=train_df[train_df.Survived == 0].Pclass.value_counts().values,
    text=train_df[train_df.Survived == 0].Pclass.value_counts().values,
    textposition='auto',
    name='Survived = 0',
    row=1, col=1
)
fig.add_bar(
    x=train_df[train_df.Survived == 1].Pclass.value_counts().index,
    y=train_df[train_df.Survived == 1].Pclass.value_counts().values,
    text=train_df[train_df.Survived == 1].Pclass.value_counts().values,
    textposition='auto',
    name='Survived = 1',
    row=1, col=1
)
fig.add_bar(
    x=train_df[train_df.Survived == 0].Sex.value_counts().index,
    y=train_df[train_df.Survived == 0].Sex.value_counts().values,
    text=train_df[train_df.Survived == 0].Sex.value_counts().values,
    textposition='auto',
    marker_color='#636EFA',
    showlegend=False,
    row=1, col=2
)
fig.add_bar(
    x=train_df[train_df.Survived == 1].Sex.value_counts().index,
    y=train_df[train_df.Survived == 1].Sex.value_counts().values,
    text=train_df[train_df.Survived == 1].Sex.value_counts().values,
    textposition='auto',
    marker_color='#EF553B',
    showlegend=False,
    row=1, col=2
)

fig.update_layout(
    barmode='stack',
    height=400, width=1200,
)
fig.update_xaxes(ticks="inside")
fig.update_yaxes(ticks="inside", col=1)
fig.show()

Resulting plot:

Hope this is helpful to the newbies of plotly like me.