Reading data from text file with missing values

Question

I want to read data from a file that has many missing values, as in this example:

1,2,3,4,5
6,,,7,8
,,9,10,11

I am using the numpy.loadtxt function:

data = numpy.loadtxt('test.data', delimiter=',')

The problem is that the missing values break loadtxt (I get a "ValueError: could not convert string to float:", no doubt because of the two or more consecutive delimiters).

Is there a way to do this automatically, with loadtxt or another function, or do I have to bite the bullet and parse each line manually?

Answer 1

I'd probably use genfromtxt:

>>> from numpy import genfromtxt
>>> genfromtxt("missing1.dat", delimiter=",")
array([[  1.,   2.,   3.,   4.,   5.],
       [  6.,  nan,  nan,   7.,   8.],
       [ nan,  nan,   9.,  10.,  11.]])

and then do whatever with the nans (change them to something, use a mask instead, etc.) Some of this could be done inline:

>>> genfromtxt("missing1.dat", delimiter=",", filling_values=99)
array([[  1.,   2.,   3.,   4.,   5.],
       [  6.,  99.,  99.,   7.,   8.],
       [ 99.,  99.,   9.,  10.,  11.]])

Answer 2

This is because the function expects to return a numpy array with all cells of the same type.

If you want a table with mixed strings and number, you should read it into a structured array instead, also you probably want to add skip_header=1 to skip the first line, ie in your case something like:

np.genfromtxt('upeak_names.txt', delimiter="\t", dtype="S10,S10,f4,S10,f4,S10,f4", 
names=["id", "name", "Distance", "name2", "Distance2", "name3", "Distance3], skip_header=1)

See also:

• Documentation for genfromtxt: https://docs.scipy.org/doc/numpy-1.15.1/reference/generated/numpy.genfromtxt.html

• Documentation for structured arrays in numpy:
  https://docs.scipy.org/doc/numpy-1.15.0/user/basics.rec.html