Panel仪表板切换Select控件时Seaborn图表不更新问题

我尝试搭建搭载Seaborn可视化图表的Panel仪表板，代码运行后可正常展示控件与初始渲染的图表，但选择不同年份或月份选项时，图表不会触发更新，已尝试.servable()、gridspec等不同实现方式，问题始终存在。

初始测试代码 #

import xarray as xr
import seaborn as sns
import panel as pn
import matplotlib.pyplot as plt
import numpy
import pandas as pd
pn.extension()

x = [numpy.datetime64('2011-01-01T00:00:00.000000000'),  numpy.datetime64('2011-01-02T00:00:00.000000000'),
  numpy.datetime64('2011-01-03T00:00:00.000000000'),  numpy.datetime64('2011-01-04T00:00:00.000000000'),
  numpy.datetime64('2011-01-05T00:00:00.000000000'),  numpy.datetime64('2011-01-06T00:00:00.000000000'),
  numpy.datetime64('2011-01-07T00:00:00.000000000'),  numpy.datetime64('2011-01-08T00:00:00.000000000'),
  numpy.datetime64('2011-01-09T00:00:00.000000000'),  numpy.datetime64('2011-01-10T00:00:00.000000000'),
  numpy.datetime64('2011-01-11T00:00:00.000000000'),  numpy.datetime64('2011-01-12T00:00:00.000000000'),
  numpy.datetime64('2011-01-13T00:00:00.000000000'),  numpy.datetime64('2011-01-14T00:00:00.000000000'),
  numpy.datetime64('2011-01-15T00:00:00.000000000'),  numpy.datetime64('2011-01-16T00:00:00.000000000'),
  numpy.datetime64('2011-01-17T00:00:00.000000000'),  numpy.datetime64('2011-01-18T00:00:00.000000000'),
  numpy.datetime64('2011-01-19T00:00:00.000000000'),  numpy.datetime64('2011-01-20T00:00:00.000000000'),
  numpy.datetime64('2011-01-21T00:00:00.000000000'),  numpy.datetime64('2011-01-22T00:00:00.000000000'),
  numpy.datetime64('2011-01-23T00:00:00.000000000'),  numpy.datetime64('2011-01-24T00:00:00.000000000'),
  numpy.datetime64('2011-01-25T00:00:00.000000000'),  numpy.datetime64('2011-01-26T00:00:00.000000000'),
  numpy.datetime64('2011-01-27T00:00:00.000000000'),  numpy.datetime64('2011-01-28T00:00:00.000000000'),
  numpy.datetime64('2011-01-29T00:00:00.000000000'),  numpy.datetime64('2011-01-30T00:00:00.000000000'),
  numpy.datetime64('2011-01-31T00:00:00.000000000'),  numpy.datetime64('2011-02-01T00:00:00.000000000'),
  numpy.datetime64('2011-02-02T00:00:00.000000000'),  numpy.datetime64('2011-02-03T00:00:00.000000000'),
  numpy.datetime64('2011-02-04T00:00:00.000000000'),  numpy.datetime64('2011-02-05T00:00:00.000000000'),
  numpy.datetime64('2011-02-06T00:00:00.000000000'),  numpy.datetime64('2011-02-07T00:00:00.000000000'),
  numpy.datetime64('2011-02-08T00:00:00.000000000'),  numpy.datetime64('2011-02-09T00:00:00.000000000'),
  numpy.datetime64('2011-02-10T00:00:00.000000000'),  numpy.datetime64('2011-02-11T00:00:00.000000000'),
  numpy.datetime64('2011-02-12T00:00:00.000000000'),  numpy.datetime64('2011-02-13T00:00:00.000000000'),
  numpy.datetime64('2011-02-14T00:00:00.000000000'),  numpy.datetime64('2011-02-15T00:00:00.000000000'),
  numpy.datetime64('2011-02-16T00:00:00.000000000'),  numpy.datetime64('2011-02-17T00:00:00.000000000'),
  numpy.datetime64('2011-02-18T00:00:00.000000000'),  numpy.datetime64('2011-02-19T00:00:00.000000000'),
  numpy.datetime64('2011-02-20T00:00:00.000000000'),  numpy.datetime64('2011-02-21T00:00:00.000000000'),
  numpy.datetime64('2011-02-22T00:00:00.000000000'),  numpy.datetime64('2011-02-23T00:00:00.000000000'),
  numpy.datetime64('2011-02-24T00:00:00.000000000'),  numpy.datetime64('2011-02-25T00:00:00.000000000'),
  numpy.datetime64('2011-02-26T00:00:00.000000000'),  numpy.datetime64('2011-02-27T00:00:00.000000000'),
  numpy.datetime64('2011-02-28T00:00:00.000000000'),  numpy.datetime64('2011-03-01T00:00:00.000000000')]
y =  [0.0,  0.0,  0.0,  0.0,  0.0,  0.0,  0.0,  0.0,  0.0, 0.0,  0.0,  0.0,  0.0,  0.0,  0.0,  0.0,  0.0,  0.0,  0.0,
  0.0,  0.0,  0.0,  0.0,  0.0,  0.0,  0.0,  0.0,  0.0,  3.0,  2.0,  0.0,  0.0,  0.0,  1.0,  6.0,  0.0,  0.0,  0.0,
  0.0,  0.0,  0.0,  0.0,  0.0,  0.0,  0.0,  0.0,  0.0,  0.0,  0.0,  0.0,  0.0,  0.0,  1.0,  3.0,  1.0,
  0.0,  1.0,  0.0,  0.0,  0.0]

data = pd.DataFrame({'time':x, 'value':y})
month_options = {'Januar' : 1, 'Februar' : 2}
month = pn.widgets.Select(name='Monat', options = month_options, value=1)
@pn.depends(month)
def dashboard(month):
    subset = data[data.time.dt.month == month] 
    def frosteindringtiefe(ds):
        sns.set_theme(style="whitegrid")
        f, ax = plt.subplots(figsize=(15, 6))
        bar = sns.barplot(x=ds.time, y=ds.value)
        ax.set_xticklabels(labels=range(len(ds.time)))
        plt.title(list(month_options.keys())[month -1 ])
        plt.ylabel('Frosteindringtiefe in cm')
        
        return bar    
    bar1 = frosteindringtiefe(subset)    
    return bar1

pn.panel(pn.Row(month,dashboard))

问题排查更新 #

已将代码替换为硬编码的测试数据，问题仍未解决，因此可以排除xarray模块相关的诱因，当前测试直接使用pandas DataFrame作为数据源。已尝试过如下几种启动方式：

• 方式1：

server = pn.panel(pn.Row(month,dashboard))
server.servable()

• 方式2：直接输出server变量

server

• 方式3：直接调用show方法启动

pn.panel(pn.Row(month,dashboard)).show()

上述所有方式均可正常加载初始页面，但切换月份控件选项后图表始终不会刷新更新；如果直接修改月份控件的初始value值（从1改为2），初始加载的图表会正确对应展示2月的数据。

核心错误有2个：

1. 返回值类型不被Panel识别为可更新的渲染对象：渲染函数最终返回的是seaborn的Axes对象（sns.barplot的返回值），Panel无法正确监听这个对象的重绘需求，只会在首次加载时渲染一次，后续控件值变更时不会触发刷新。
2. Matplotlib图表内存泄漏问题：每次触发更新时会新建plt.subplots()画布，但旧画布没有被关闭，多次切换后会产生大量冗余画布对象，也会干扰渲染逻辑。

修复后的可运行代码 #

import seaborn as sns
import panel as pn
import matplotlib.pyplot as plt
import numpy
import pandas as pd
pn.extension()

# 测试数据保持不变
x = [numpy.datetime64('2011-01-01T00:00:00.000000000'),  numpy.datetime64('2011-01-02T00:00:00.000000000'),
  numpy.datetime64('2011-01-03T00:00:00.000000000'),  numpy.datetime64('2011-01-04T00:00:00.000000000'),
  numpy.datetime64('2011-01-05T00:00:00.000000000'),  numpy.datetime64('2011-01-06T00:00:00.000000000'),
  numpy.datetime64('2011-01-07T00:00:00.000000000'),  numpy.datetime64('2011-01-08T00:00:00.000000000'),
  numpy.datetime64('2011-01-09T00:00:00.000000000'),  numpy.datetime64('2011-01-10T00:00:00.000000000'),
  numpy.datetime64('2011-01-11T00:00:00.000000000'),  numpy.datetime64('2011-01-12T00:00:00.000000000'),
  numpy.datetime64('2011-01-13T00:00:00.000000000'),  numpy.datetime64('2011-01-14T00:00:00.000000000'),
  numpy.datetime64('2011-01-15T00:00:00.000000000'),  numpy.datetime64('2011-01-16T00:00:00.000000000'),
  numpy.datetime64('2011-01-17T00:00:00.000000000'),  numpy.datetime64('2011-01-18T00:00:00.000000000'),
  numpy.datetime64('2011-01-19T00:00:00.000000000'),  numpy.datetime64('2011-01-20T00:00:00.000000000'),
  numpy.datetime64('2011-01-21T00:00:00.000000000'),  numpy.datetime64('2011-01-22T00:00:00.000000000'),
  numpy.datetime64('2011-01-23T0