Matplotlib交互导航工具栏导致图例选择器失效问题问询

首先，我注意到你的代码里有个明显的语法错误（entry.set_visi没写完），这会直接导致图例点击功能异常，更别说缩放平移后的问题了。另外，在Matplotlib 2.0.0这个旧版本中，缩放/平移操作会触发图表重绘，原来的图例条目和绘图元素的关联容易丢失，加上你的事件绑定和可见性控制逻辑有一些疏漏，才导致了失效的问题。

问题根源分析 #

• 语法错误：onpick函数中entry.set_visi未完成，应该是entry.set_visible(vis)
• 重绘导致的关联丢失：Matplotlib在缩放/平移后会重新渲染图表，原来绑定的图例条目和绘图元素的映射关系没有被正确保留
• 直方图可见性处理不当：直方图返回的是元组，你只保存了line3[2]（patches），但控制可见性时的遍历逻辑有问题
• 图例条目绑定逻辑不严谨：你用了临时创建的Line2D对象作为组图例条目，但没有正确关联到组内所有绘图元素的可见性状态

修复后的完整代码 #

import numpy as np
import random
import operator
import matplotlib.pyplot as plt
from matplotlib.lines import Line2D

# Generate some fake data
datalength = 60
datavariance = 25
xline = np.array(range(datalength))
y1 = np.array(range(datalength))
groups = 5  # Number of groups
members = 4  # Number of members per group
grouping = {}  # Nested dictionary to hold groups and member dictionaries

for g in range(groups):
    groupmembers = {}
    for m in range(members):
        groupmembers[f'SP_{g+1}_{m+1}'] = []
    grouping[f'SP_{g+1}'] = groupmembers

# Establish the figure and arrange the subplots
fig = plt.figure(figsize=(10,7))
ax1 = plt.subplot2grid((3,1), (0,0), rowspan=1, colspan=1, facecolor='white')
ax2 = plt.subplot2grid((3,1), (1,0), rowspan=1, colspan=1, facecolor='white')
ax3 = plt.subplot2grid((3,1), (2,0), rowspan=1, colspan=1, facecolor='white')

# Each member is plotted in each subplot and then the artist is added to the array in the nested dictionary.
for group in grouping:
    for member in grouping[group]:
        # Plot line
        line1, = ax1.plot(xline, y1+np.random.randint(low=0, high=datavariance, size=datalength), 
                         lw=1, label=member, marker='.')
        # Plot scatter
        line2 = ax2.scatter(np.random.randn(datalength)+4, y1+np.random.randint(low=0, high=datavariance, size=datalength), 
                           s=5, label=member)
        # Plot histogram
        hist_out = ax3.hist(np.random.randn(datalength)+4, 20, histtype='step', lw=1, label=member)
        # Save all relevant artists: line, scatter, histogram patches
        grouping[group][member].extend([line1, line2, hist_out[2]])

# Chart labeling
ax1.set_title('Process Line', fontsize=10, bbox=dict(facecolor='white'))
ax2.set_title('Process Scatter', fontsize=10, bbox=dict(facecolor='white'))
ax3.set_title('Process Histogram', fontsize=10, bbox=dict(facecolor='white'))
fig.suptitle('System Processes', fontsize=16, bbox=dict(facecolor='white'))

# Get handles and labels from subplot 1
leghandles, leglabels = ax1.get_legend_handles_labels() 
grouplines = {}  # Dictionary of lines by group
groupvisible = {}  # Dictionary for visibility control

# Add group entries to legend handles/labels
for group in grouping:
    # Collect all artists in the group
    group_artists = []
    for member in grouping[group]:
        group_artists.extend(grouping[group][member])
    grouplines[group] = group_artists
    groupvisible[group] = True
    # Add a dummy line for group legend entry
    leghandles.append(Line2D([], [], label=group, marker='_', markersize=7, lw=4, color='black'))
    leglabels.append(group)

# Sort legend by name
leghandles, leglabels = zip(*sorted(zip(leghandles, leglabels), key=operator.itemgetter(1)))
leghandlelist = list(leghandles)
leglabellist = list(leglabels)

# Add spacing between groups in legend
for i in range(groups):
    leghandlelist.insert((members + 2)*i, Line2D([], [], lw=0))
    leglabellist.insert((members + 2)*i, ' ')

leghandles = tuple(leghandlelist)
leglabels = tuple(leglabellist)

# Create legend
mylegend = fig.legend(leghandles, leglabels, fancybox=True, shadow=True, loc='upper left', ncol=1, title='Process Groups')

# Create a mapping from legend labels to their corresponding artists (members or groups)
legend_label_map = {}
# First map member labels to their artists
for group in grouping:
    for member in grouping[group]:
        legend_label_map[member] = grouping[group][member]
# Then map group labels to their artists
for group in grouping:
    legend_label_map[group] = grouplines[group]

# Set pickers on all legend entries
for entry in mylegend.get_lines():
    entry.set_picker(5)

def onpick(event):
    legend_entry = event.artist
    label = legend_entry.get_label()
    
    # Skip empty spacing entries
    if label.strip() == '':
        return
    
    # Toggle visibility state
    new_vis = not legend_entry.get_visible()
    legend_entry.set_visible(new_vis)
    
    # Update corresponding artists
    if label in legend_label_map:
        artists = legend_label_map[label]
        for artist in artists:
            # Handle different artist types
            if isinstance(artist, list):
                # For histogram patches (list of patches)
                for patch in artist:
                    patch.set_visible(new_vis)
            else:
                # For lines and scatter points
                artist.set_visible(new_vis)
    
    # Update group member legend entries if a group is toggled
    if label in grouping:
        for member in grouping[label]:
            for entry in mylegend.get_lines():
                if entry.get_label() == member:
                    entry.set_visible(new_vis)
    # Update group legend entry if a member is toggled (optional: sync group state)
    else:
        # Find which group this member belongs to
        member_group = None
        for group in grouping:
            if label in grouping[group]:
                member_group = group
                break
        if member_group:
            # Check if all members are visible/invisible to sync group legend
            all_vis = all(artist.get_visible() for artist in legend_label_map[label])
            for entry in mylegend.get_lines():
                if entry.get_label() == member_group:
                    entry.set_visible(all_vis)
    
    # Redraw the figure to apply changes
    fig.canvas.draw_idle()

# Bind the pick event
fig.canvas.mpl_connect('pick_event', onpick)

plt.subplots_adjust(left=0.2, right=0.95, top=0.90, bottom=.05, wspace=.2, hspace=.3)
plt.show()

关键修改说明 #

1. 修复语法错误：补全了onpick函数中未完成的entry.set_visible(vis)调用
2. 优化艺术家存储：直方图返回的元组中，正确保存了所有patches（hist_out[2]），并统一处理不同类型的艺术家（线条、散点、直方图补丁）
3. 更严谨的图例映射：创建了legend_label_map直接关联图例标签和对应的所有绘图艺术家，避免重绘后关联丢失
4. 同步组和成员的可见性：当点击组图例时，自动同步组内所有成员的图例条目可见性；点击成员时，可选同步组图例的状态（根据成员的可见性自动更新组的状态）
5. 强制重绘：在onpick函数末尾调用fig.canvas.draw_idle()确保可见性更改立即生效，即使在缩放/平移后也能正确更新
6. 跳过空条目：忽略图例中的空间隔条目，避免无效点击

经过这些修改后，即使使用缩放和平移工具，图例选择器也能正常工作，可见性切换不会失效。

内容的提问来源于stack exchange，提问作者GeorgeLPerkins