在嵌套的for循环中附加DataFrame而不附加NaN值

Question

我有一个嵌套的for循环，它遍历一个字典，取出特定的数据切片，并将这个切片附加到一个新的字典中。我不确定这是否与append上的索引或循环的数量有关，或者两者都没有。代码如下所示。

def combine_timesteps(channels, startpoint = 5):
    startpoint_start = startpoint
    endpoint = None
    
    timestep_dataset = pd.DataFrame({'Discharge': discharge_values_at_strain_times[startpoint:endpoint]}).transpose()

    for channel in channels.items():
        
        for i, time_data in enumerate(channel[1]):
                        
            data = channel[1][startpoint:endpoint]
            time_int_string = str(channel[0]) + ' t-' + str(i)
            df_to_append = pd.DataFrame({time_int_string:data})

            timestep_dataset = timestep_dataset.append(df_to_append[time_int_string])
            
            print(timestep_dataset)
            
            new_start = startpoint - 1
            startpoint = new_start
            
            if endpoint == None:
                endpoint = 0
            new_end = endpoint - 1
            endpoint = new_end
            
            if startpoint == -1: 
                startpoint = startpoint_start
                endpoint = None
                break
    
    timestep_dataset_formatted = timestep_dataset.transpose()      
    
    return timestep_dataset_formatted

但是，每当发生新的追加时，它都会向下滑动附加了amount NaN值的数据，该值等于前面的起始点整数，如下所示。

Discharge           9.913190    9.908002    9.902636    9.897269    9.891902   
Channel 750 t-0          NaN         NaN         NaN         NaN         NaN   
Channel 750 t-1          NaN         NaN         NaN         NaN  259.839939   
Channel 750 t-2          NaN         NaN         NaN  274.439453  259.839939   
Channel 750 t-3          NaN         NaN  269.095527  274.439453  259.839939   
Channel 750 t-4          NaN  363.006610  269.095527  274.439453  259.839939   
Channel 750 t-5   977.719896  363.006610  269.095527  274.439453  259.839939   
Channel 1300 t-0         NaN         NaN         NaN         NaN         NaN   
Channel 1300 t-1         NaN         NaN         NaN         NaN  325.599363   
Channel 1300 t-2         NaN         NaN         NaN  420.957686  325.599363   
Channel 1300 t-3         NaN         NaN  376.701738  420.957686  325.599363   
Channel 1300 t-4         NaN  444.503183  376.701738  420.957686  325.599363   
Channel 1300 t-5  964.735686  444.503183  376.701738  420.957686  325.599363   

所需的输出格式是将t-0列中的第一个值作为第一个非NaN值(基本上是将每一列向上滑动空格，直到没有NaN值)，依此类推。我不确定为什么首先抓取并追加的数据帧前面有NaN值。

Discharge           9.913190    9.908002    9.902636    9.897269    9.891902   
Channel 750 t-0   287.678547  264.190236  182.871974  208.402388  246.174055
Channel 750 t-1   259.839939  287.678547  264.190236  182.871974  208.402388 
Channel 750 t-2   274.439453  259.839939  287.678547  264.190236  182.871974 
Channel 750 t-3   269.095527  274.439453  259.839939  287.678547  264.190236 
Channel 750 t-4   363.006610  269.095527  274.439453  259.839939  287.678547
Channel 750 t-5   977.719896  363.006610  269.095527  274.439453  259.839939   
Channel 1300 t-0  362.181147  403.321962  423.858839  341.982210  237.443283   
Channel 1300 t-1  325.599363  362.181147  403.321962  423.858839  341.982210 
Channel 1300 t-2  420.957686  325.599363  362.181147  403.321962  423.858839 
Channel 1300 t-3  376.701738  420.957686  325.599363  362.181147  403.321962 
Channel 1300 t-4  444.503183  376.701738  420.957686  325.599363  362.181147 
Channel 1300 t-5  964.735686  444.503183  376.701738  420.957686  325.599363   

谢谢你的帮助。

Answer 1

好了，我想通了。我必须确保在每个循环上重置索引。我可以通过使用.reset_index(drop = True)方法来实现这一点，这样新行就会显示为。。。

df_to_append = pd.DataFrame({time_int_string:data}).reset_index(drop = True)

这将确保数据被追加到正确的位置。