Geek asks and answers

Fill arrays in python (numpy)?

The specified file is in the following format:

a a 0
a b 1
a c 1
b b 0
b a 1
b c 1
c c 0
c a 1
c b 1

The third column is the distance between the elements in the first and second columns. If I read such a file in pyton as a nested list, how to convert it to a symmetric matrix, i.e.

  a b c
a 0 1 1
b 1 0 1
b 1 1 0

? I also want to include column and row names.

I would rather use numpy to accomplish this task.

Any suggestions?

Thanks D.

+3
source share
2 answers
import numpy as np
from itertools import count

data = [line.split() for line in inputfile.readlines()]
rows = dict(zip(sorted(set(line[0] for line in data)), count()))
cols = dict(zip(sorted(set(line[1] for line in data)), count()))
array = np.zeros((len(rows), len(cols)))

for row, col, val in data:
    index = (rows[row], cols[col])
    array[index] = val

I don't know how to mark rows and columns in numpy, so I just made a dict matching the row label with the row index, and the other the same for the columns. If you need it, you can make a reverse map, as shown below, or you can make rows and columns of bidict .

rows_reverse = dict((v, k) for k, v in rows)
cols_reverse = dict((v, k) for k, v in cols)
+4
source

A slightly different approach:

import numpy as np
# Load "Row Col Value" text file
ar = np.loadtxt('file.txt', [('R','|S1'), ('C','|S1'), ('V','i')])
names = np.unique(np.row_stack((ar['R'], ar['C']))).tolist()
vf = np.vectorize(lambda x: names.index(x), otypes='i')
# load them in an output array
out = np.empty((len(names), len(names)), 'i')
out[vf(ar['R']), vf(ar['C'])] = ar['V']
+2
source

More articles:


All Articles