Plotly：排序多类别条形图

Question

我在排序多类别图表时遇到了一些问题。

一些示例代码。

import pandas as pd
import plotly.graph_objects as go

data = [
    [0, "Born", 4, "Rhino"],  # commenting this line will also reverse sub category sorting
    [0, "Died", -1, "Rhino"],
    [1, "Born", 4, "Lion"],
    [1, "Died", -1, "Lion"],
    [2, "Born", 12, "Rhino"],
    [2, "Died", -5, "Lion"],
]
z_data = list(zip(*data))

df = pd.DataFrame({
    "tick": z_data[0],
    "category": z_data[1],
    "value": z_data[2],
    "type": z_data[3],
})
df = df.sort_values(by=['tick', 'category', 'value', 'type'])
print(df)
fig = go.Figure()
for t in df.type.unique():
    plot_df = df[df.type == t]
    fig.add_trace(go.Bar(
        x=[plot_df.tick, plot_df.category],
        y=abs(plot_df.value),
        name=t,
    ))
fig.update_layout({
    'barmode': 'stack',
    'xaxis': {
        'title_text': "Tick",
        'tickangle': -90,
    },
    'yaxis': {
        'title_text': "Value",
    },
})
fig.write_html(str("./diagram.html"))

正如您所见，第 2 刻在第 1 刻之前。这是因为“Rhino”是类型列表中的第一个，这将创建第 0 和第 2 刻。在第 1 刻之后添加了狮子条。 但是我现在如何正确排序条形？

附言。 'barmode': 'stack' 是故意的。即使在这个测试示例中没有使用它。

Answer 1

我能够修正刻度，但不能修正出生/死亡顺序。我打算逐行绘制，所以我需要玩 showlegend

数据

import pandas as pd
import plotly.graph_objects as go
data = [
    [0, "Born", 4, "Rhino"],  # commenting this line will also reverse sub category sorting
    [0, "Died", -1, "Rhino"],
    [1, "Born", 4, "Lion"],
    [1, "Died", -1, "Lion"],
    [2, "Born", 12, "Rhino"],
    [2, "Died", -5, "Lion"],
]
# you don't really need to zip here
df = pd.DataFrame(data, columns=["tick", "category", "value", "type"])
df["value"] = df["value"].abs()

设置颜色

如果您有更多类型，这里有可以帮助您的答案。检查doc

color_diz = {"Rhino": "blue", "Lion": "red"}
df["color"] = df["type"].map(color_diz)

显示图例

这里我想展示每种类型第一次出现的图例

grp = df.groupby("type")\
        .apply(lambda x: x.index.min())\
        .reset_index(name="idx")

df = pd.merge(df, grp, on=["type"], how="left")

df["showlegend"] = df.index == df["idx"]

要绘制的数据

print(df)

   tick category  value   type color  idx  showlegend
0     0     Born      4  Rhino  blue    0        True
1     0     Died      1  Rhino  blue    0       False
2     1     Born      4   Lion   red    2        True
3     1     Died      1   Lion   red    2       False
4     2     Born     12  Rhino  blue    0       False
5     2     Died      5   Lion   red    2       False

情节

fig = go.Figure()
for i, row in df.iterrows():
    fig.add_trace(
        go.Bar(x=[[row["tick"]], [row["category"]]],
               y=[row["value"]],
               name=row["type"],
               marker_color=row["color"],
               showlegend=row["showlegend"],
               legendgroup=row["type"] # Fix legend
               ))
    
fig.update_layout({
    'barmode': 'stack',
    'xaxis': {
        'title_text': "Tick",
        'tickangle': -90,
    },
    'yaxis': {
        'title_text': "Value",
    },
})
fig.show()

编辑

如果您有更多type，您可以使用以下技巧。

首先我生成不同的类型

import string
import numpy as np
import pandas as pd
import plotly.express as px

df = pd.DataFrame({"type":np.random.choice(list(string.ascii_lowercase), 100)})

然后我从doc 中挑选一个颜色序列并将它们放入字典中

color_dict = {k:v for k,v in enumerate(px.colors.qualitative.Plotly)}

然后我将唯一的type 放在数据框上

df_col = pd.DataFrame({"type": df["type"].unique()})

我根据它们的索引为它们每个分配一个颜色

df_col["color"] = (df_col.index%len(color_dict)).map(color_dict)

最后我合并到原来的df

df = pd.merge(df, df_col, on=["type"], how="left")