使用Jupyter Widget值切片Pandas DataFrame触发step必须≥0错误

在Jupyter Notebook中使用Jupyter Widget与函数交互，通过下拉框切换不同图表，选项来自DataFrame。以下是可正常运行的伪代码：

def plot_option(x):
    plot(x)  
        
        
option = widgets.Dropdown(
               options=[ df[column].iloc[i] for i in range(10)])
        
        
interact(plot_option,x=option)

但实际实现中希望Dropdown的值为排序后DataFrame的索引i，修改代码后触发错误：

option = widgets.Dropdown(
                options=[i for i in range(10)])
            
            
interact(plot_option,x=df[column].iloc[option.value])

错误信息：

step must >= 0, not -2

疑问：option.value难道不是整数吗？为何不能在interact定义中进行切片操作？

回溯信息：

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
Cell In[332], line 11
      8 def plot_tiles_option(board,free_coords,coord):
      9     plot_tiles(board,free_coords,local_tiles(coord,2),coord=coord)
---> 11 interact(plot_tiles_option,board=fixed(Board),free_coords=fixed(Free_coords),coord=top10['coordinates'].iloc[int(option.value)])

File ~\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.13_qbz5n2kfra8p0\LocalCache\local-packages\Python313\site-packages\ipywidgets\widgets\interaction.py:536, in _InteractFactory.__call__(self, _InteractFactory__interact_f, **kwargs)
    528     return self
    530 # positional arg support in: https://gist.github.com/8851331
    531 # Handle the cases 1 and 2
    532 # 1. interact(f, **kwargs)
    533 # 2. @interact
    534 #    def f(*args, **kwargs):
    535 #        ...
---> 536 w = self.widget(f)
    537 try:
    538     f.widget = w

File ~\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.13_qbz5n2kfra8p0\LocalCache\local-packages\Python313\site-packages\ipywidgets\widgets\interaction.py:452, in _InteractFactory.widget(self, f)
    436 def widget(self, f):
    437     """
    438     Return an interactive function widget for the given function.
    439 
   (...)    450         The function to which the interactive widgets are tied.
    451     """
---> 452     return self.cls(f, self.opts, **self.kwargs)

File ~\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.13_qbz5n2kfra8p0\LocalCache\local-packages\Python313\site-packages\ipywidgets\widgets\interaction.py:187, in interactive.__init__(self, _interactive__interact_f, _interactive__options, **kwargs)
    185     getcallargs(f, **{n:v for n,v,_ in new_kwargs})
    186 # Now build the widgets from the abbreviations.
---> 187 self.kwargs_widgets = self.widgets_from_abbreviations(new_kwargs)
    189 # This has to be done as an assignment, not using self.children.append,
    190 # so that traitlets notices the update. We skip any objects (such as fixed) that
    191 # are not DOMWidgets.
    192 c = [w for w in self.kwargs_widgets if isinstance(w, DOMWidget)]

File ~\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.13_qbz5n2kfra8p0\LocalCache\local-packages\Python313\site-packages\ipywidgets\widgets\interaction.py:286, in interactive.widgets_from_abbreviations(self, seq)
    284 if isinstance(abbrev, Widget) and (not isinstance(abbrev, ValueWidget)):
    285     raise TypeError("{!r} is not a ValueWidget".format(abbrev))
---> 286 widget = self.widget_from_abbrev(abbrev, default)
    287 if widget is None:
    288     raise ValueError("{!r} cannot be transformed to a widget".format(abbrev))

File ~\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.13_qbz5n2kfra8p0\LocalCache\local-packages\Python313\site-packages\ipywidgets\widgets\interaction.py:302, in interactive.widget_from_abbrev(cls, abbrev, default)
    299     return abbrev
    301 if isinstance(abbrev, tuple):
---> 302     widget = cls.widget_from_tuple(abbrev)
    303     if default is not empty:
    304         try:

File ~\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.13_qbz5n2kfra8p0\LocalCache\local-packages\Python313\site-packages\ipywidgets\widgets\interaction.py:382, in interactive.widget_from_tuple(o)
    380 step = o[2]
    381 if step <= 0:
---> 382     raise ValueError("step must be >= 0, not %r" % step)
    383 min, max, value = _get_min_max_value(o[0], o[1], step=step)
    384 if all(isinstance(_, Integral) for _ in o):

ValueError: step must be >= 0, not -2

错误原因 #

你在interact里直接写x=df[column].iloc[option.value]时，interact会把这个值当成静态参数解析，而非绑定下拉框的动态值。如果df[column].iloc[option.value]返回的是类似元组的结构，interact会误将其识别为生成滑块组件的参数（比如(min, max, step)），当其中的step为负数时就会触发该错误。

另外，option.value本身是整数，但你在interact中直接计算出了最终值，这相当于给x传入一个固定值，而非绑定下拉框组件，完全失去了交互的意义。

正确写法 #

要实现通过下拉框选择索引、再用索引取DataFrame数据绘图的需求，应将下拉框组件传给interact，然后在绘图函数内部用传入的索引获取数据：

def plot_option(idx):
    # 在函数内部根据索引获取数据
    x = df[column].iloc[idx]
    plot(x)  

# 下拉框选项为索引0-9
option = widgets.Dropdown(options=[i for i in range(10)])

# 直接把下拉框组件传给interact的x参数
interact(plot_option, x=option)

或者更简洁的写法，无需提前定义Dropdown，直接给interact传选项列表，它会自动生成下拉框：

def plot_option(idx):
    x = df[column].iloc[idx]
    plot(x)  

# 直接传入选项范围，interact自动生成下拉框
interact(plot_option, x=list(range(10)))

这样就能实现选择索引后动态绘图的效果，同时避免interact误解析参数的问题。

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