numpy.genfromtxt 导入元组而不是数组

Question

我正在尝试学习Python和Numpy，请多多包涵。我使用numpy.genfromtxt将一个CSV文件导入一个矩阵中。CSV文件如下所示：

Time(min),Nm,Speed,Power,Distance,Rpm,Bpm,interval,Altitude,Rate,Incline,Temp,PowerBalance,LeftTorqueEffectiveness,RightTorqueEffectiveness,getLeftPedalSmoothness,getRightPedalSmoothness,getCombinedPedalSmoothness,THb,SmO2,km
0.016666668,,4.3555064,0,0.002,0,118,1,684.3,0.0,0.0,14.71,50,-1.0,-1.0,-1.0,-1.0,-1.0,311.72,311.72
0.033333335,,4.3555064,20,0.002,0,119,1,684.3,0.0,0.0,14.71,50,-1.0,-1.0,-1.0,-1.0,-1.0,311.72,311.72
0.05,,4.444291,13,0.004,0,119,1,684.3,0.0,0.0,14.71,50,-1.0,-1.0,-1.0,-1.0,-1.0,311.72,311.72

现在我运行：

matrixCsv = np.genfromtxt(open(csvFile, "rb"), delimiter=',', \
                          missing_values=0,skip_header=1,dtype=float,\
                          usecols=(0,2,3,4,5,6,7,8,9,10,11,17),names=True)

我得到：

[ (0.033333335, 4.3555064, 20.0, 0.002, 0.0, 119.0, 1.0, 684.3, 0.0, 0.0, 14.71, -1.0)
(0.05, 4.444291, 13.0, 0.004, 0.0, 119.0, 1.0, 684.3, 0.0, 0.0, 14.71, -1.0)
(0.06666667, 4.4781966, 16.0, 0.006, 0.0, 120.0, 1.0, 684.3, 0.0, 0.0, 14.71, -1.0)
...,

在我看来，这似乎是将元组封装成数组。但为什么要使用元组呢？我知道numpy数组/矩阵需要是同质的，并且numpy会将不同类型的数据转换为元组。但为什么我的数据是不同类型的呢？我不理解...

- aag

Answer 1

在使用 skip_header 和 names 时，您可能会感到困惑。正确的读取数据和使用第一行作为变量名称的方法是:

In [185]:

np.genfromtxt('temp.csv', delimiter=',', \
                          missing_values=0,skip_header=0,dtype=float,\
                          usecols=(0,2,3,4,5,6,7,8,9,10,11,17),names=True)
Out[185]:
array([ (0.016666668, 4.3555064, 0.0, 0.002, 0.0, 118.0, 1.0, 684.3, 0.0, 0.0, 14.71, -1.0),
       (0.033333335, 4.3555064, 20.0, 0.002, 0.0, 119.0, 1.0, 684.3, 0.0, 0.0, 14.71, -1.0),
       (0.05, 4.444291, 13.0, 0.004, 0.0, 119.0, 1.0, 684.3, 0.0, 0.0, 14.71, -1.0)], 
      dtype=[('Timemin', '<f8'), ('Speed', '<f8'), ('Power', '<f8'), ('Distance', '<f8'), ('Rpm', '<f8'), ('Bpm', '<f8'), ('interval', '<f8'), ('Altitude', '<f8'), ('Rate', '<f8'), ('Incline', '<f8'), ('Temp', '<f8'), ('getCombinedPedalSmoothness', '<f8')])

这不是一个元组数组，而是一个结构化数组。使用skip_header=1会将第一行数据用作名称，这可能不是您想要的（看看您是否缺少第一行数据？）。您还可以摆脱名称并将数据读入普通的numpyarray中。

In [186]:

np.genfromtxt('temp.csv', delimiter=',', \
                          missing_values=0,skip_header=1,dtype=float,\
                          usecols=(0,2,3,4,5,6,7,8,9,10,11,17))
Out[186]:
array([[  1.66666680e-02,   4.35550640e+00,   0.00000000e+00,
          2.00000000e-03,   0.00000000e+00,   1.18000000e+02,
          1.00000000e+00,   6.84300000e+02,   0.00000000e+00,
          0.00000000e+00,   1.47100000e+01,  -1.00000000e+00],
       [  3.33333350e-02,   4.35550640e+00,   2.00000000e+01,
          2.00000000e-03,   0.00000000e+00,   1.19000000e+02,
          1.00000000e+00,   6.84300000e+02,   0.00000000e+00,
          0.00000000e+00,   1.47100000e+01,  -1.00000000e+00],
       [  5.00000000e-02,   4.44429100e+00,   1.30000000e+01,
          4.00000000e-03,   0.00000000e+00,   1.19000000e+02,
          1.00000000e+00,   6.84300000e+02,   0.00000000e+00,
          0.00000000e+00,   1.47100000e+01,  -1.00000000e+00]])

- CT Zhu