使用plt.hist()绘制多直方图时出现参数长度不匹配错误求助

我定义了check3_pt_bias(x)函数，用plt.hist()绘制多直方图，声明了4个数据集和4种对应颜色。但调用check3_pt_bias(2)时触发ValueError，提示'color'参数需为每个数据集分配一种颜色，但提供了914873个数据集和4种颜色。打印数据集信息显示biahist长度为4，每个元素长度都是914873，且其他x值调用函数能成功运行，求解决思路。

代码示例 #

def check3_pt_bias(x):
    bins=[[25,60,90,120,160,400],[25,40,60,400],[25,45,70,100,160,400],[25,35,55,400]]
    name = ['lead b-top','sub-lead b-top','lead b-Add','sub-lead b-Add']
    biahist=np.array([collection42[2][x],collection42[1][x],collection43[1][x],collection44[1][x]],dtype=object)
    for i in range(len(biahist)):
        print(len(biahist[i]))
    print(len(biahist))
    fig, axs = plt.subplots(2,1, gridspec_kw={'height_ratios': [2, 1]})
    (n, bins, patches)=axs[0].hist(biahist, bins[x],color=['r','g','b','y'], range =[0,400], 
                            density=True, histtype='step',label=['truth','All_var model','dR model','Mass model'])
    axs[0].legend()
    axs[0].set_ylabel('density')
    axs[0].set_yscale('log')
    axs[0].set_xscale('log')
    axs[0].set(xticklabels=[])
    ratio = [[],[],[]]
    div = [0,0,0]
    wdiv = [0,0,0]
    for i in range(len(n[0])):
        for j in range(3):
            ratio[j].append(n[j+1][i]/n[0][i])
            div[j] += (ratio[j][i]-1)**2
            wdiv[j] += ((ratio[j][i]-1)*n[0][i])**2
        axs[1].hlines(y = ratio[0][i], xmin = bins[i], xmax = bins[i+1],color = 'g',label='All_var model')
        axs[1].hlines(y = ratio[1][i], xmin = bins[i], xmax = bins[i+1],color = 'b',label='dR model')
        axs[1].hlines(y = ratio[2][i], xmin = bins[i], xmax = bins[i+1],color = 'y',label='Mass model')
    axs[1].hlines(y = 1, xmin = bins[0], xmax = bins[-1],color = 'r',label='truth')
    axs[1].set_xlabel('Pt ('+name[x]+'), GeV')
    axs[1].set_ylabel('model / truth')
    #plt.show()
    print("deviation of All_var model:",np.sqrt(div[0]))
    print("deviation of dR model:",np.sqrt(div[1]))
    print("deviation of Mass model:",np.sqrt(div[2]))
    print("weighted deviation of All_var model:",np.sqrt(wdiv[0]))
    print("weighted deviation of dR model:",np.sqrt(wdiv[1]))
    print("weighted deviation of Mass model:",np.sqrt(wdiv[2]))
check3_pt_bias(2)

运行输出 #

914873
914873
914873
914873
4
ValueError: The 'color' keyword argument must have one color per dataset, but 914873 datasets and 4 colors were provided

解决思路与修复方法 #

问题根源 #

plt.hist()对输入数据的维度解析逻辑导致：当你用np.array(..., dtype=object)包装4个长度相同的数组时，numpy会创建一个一维的object数组，但plt.hist()在配合x=2对应的bins参数时，会错误地将这个数组解析为914873个单独的数据集（每个元素一个），而不是4个数据集。其他x值能运行是因为对应的bins长度刚好让解析逻辑正常工作。

修复步骤 #

直接将biahist改为Python列表，避免使用numpy object数组：

# 原代码
# biahist=np.array([collection42[2][x],collection42[1][x],collection43[1][x],collection44[1][x]],dtype=object)
# 修改后
biahist = [collection42[2][x], collection42[1][x], collection43[1][x], collection44[1][x]]

原理说明 #

plt.hist()原生支持接收列表形式的多个数据集，每个列表元素对应一个直方图系列。改用Python列表后，函数会正确识别这是4个数据集，与你提供的4种颜色一一对应，不会出现解析错误。

内容的提问来源于stack exchange，提问作者GeoLouise