Save dtype=object numpy arrays in a csv file containing floats and strings

Question

I must save 3 strings and one float in each line of a CSV file. Here is what I did:

    filename = "results.txt"
    if os.path.exists(filename):
        append_write = "a"
    else:
        append_write = "w"
    
    #I am saving 3 strings and one float, they come from a numpy array 
    results = np.array(["a", "b", "c", 0.32],dtype=object)
    
    f = open(filename, append_write)
    np.savetxt(f, results, delimiter=",")
    f.close()

However, here is the error I get whenever I run such a code

  raise TypeError("Mismatch between array dtype ('%s') and "
  TypeError: Mismatch between array dtype ('object') and format specifier ('%.18e')

How can I save such mixed variables in a NumPy array to a CSV file?

Answer 1

You want to specify a fmt string to np.savetxt , like this:

results = np.array(["a", "b", "c", 0.32], dtype=object).reshape(1, -1)
fmt = '%s,%s,%s,%f'

with open(filename, append_write) as f:
    np.savetxt(f, results, delimiter=",", fmt=fmt)

NB It seems like you want to append your data row-wise, in which case, you need the reshape to get your data into columns (into one row). The fmt string should match the number of columns you want to specify.