Iterating over rows in NumPy array, how do I properly take care of this 1D vs 2D case, where I want the 1D treated as an entire row for indexing?

Question

I have a series of csv files that I am using NumPy's genfromtxt function to turn into arrays. They each have the following format:

ColA, ColB, ColC
  -2, 10,   10
  -1, 10,   10
   0, 10,   10
   1, 10,   10
   2, 10,   10

In the case of the above csv , using genfromtxt(data.csv, delimeter=',', skip_header=1) will generate an array with shape (5, 3). I can then iterate over each row, ie:

data_array = np.genfromtxt(data.csv, delimeter=',', skip_header=1)

for row in data_array:
    some_function(row[0], row[1], row[2])

Where the function in the loop expects and requires three numbers. This breaks down when the csv only contains one line of data, ie:

ColA, ColB, ColC,
   0,   -4,   10, 

In this case, for row in data_array will be interpreted as looking at each number in the (1, 3)-shaped array individually, and so I cannot run the function on it. The data I have is a series of csv files and may or may not have more than one line of data. What is the proper way to handle this so that the some_function(x, y, z) can always be completed, regardless of the input array? I could do an if -statement to check if the array is 1D, but it seems like there is a better way to go about this.

Answer 1

您可以使用带有通配符（ -1 ）的np.reshape来强制数据为二维数组：

data_array = np.genfromtxt("data.csv", delimiter=',', skip_header=1).reshape(-1,3)