CODEWITHSUNDEEP
pythonarraysnumpyh5py
Master Bee
Master Bee

Reputation: 1099

How can I update arrays in h5py?

I need an array for saving values, but I also want to edit some of the values in the array anytime later.

I created an array with some random values and save it to disk. I can read it. Than I want to update it, an array slice with the value '23'. When I read it again it looks like it doesn't change.

How can I update this values?

import numpy as np
import h5py

x, y = 100,20

# create
a = np.random.random(size=(x, y))
h5f = h5py.File('data.h5', 'w')
h5f.create_dataset('dataset_1', data=a)
print a[1][0:5] # [ 0.77474947  0.3618912   0.16000164  0.47827977  0.93955235]
h5f.close()

# read
h5f = h5py.File('data.h5','r')
b = h5f['dataset_1'][:]
print b[1][0:5] #[ 0.77474947  0.3618912   0.16000164  0.47827977  0.93955235]
h5f.close()

# update
h5f = h5py.File('data.h5', 'r+')
b = h5f['dataset_1'][:]
b[1][0:5] = 23
print b[1][0:5] #[ 23.  23.  23.  23.  23.]
h5f.close()

# read again
h5f = h5py.File('data.h5','r')
b = h5f['dataset_1'][:]
print b[1][0:5] #[ 0.77474947  0.3618912   0.16000164  0.47827977  0.93955235]
h5f.close()

Upvotes: 2

Views: 4510

Answers (1)

trueter
trueter

Reputation: 199

Append mode works for me. Create file:

fh = h5py.File('dummy.h5', 'w')
fh.create_dataset('random', data=np.reshape(np.asarray([0, 1, 2, 3]), (2, 2)))
fh.close()

Open and edit in append mode ('a', default mode)..

fh = h5py.File('dummy.h5', 'a')
print fh['random'][:]
fh['random'][0, 0] = 1337
print fh['random'][:]
fh.close()

..and check again

fh = h5py.File('dummy.h5', 'r')
print fh['random'][:]
fh.close()

Write mode ('w') seems to clear the whole file. EDIT: It's important to directly access the dataset. As an earlier answer pointed out: In your problem description you assign the contents of 'dataset_1' to b, and then you edit b.

EDIT 1: 'r+' works as well for me, the problem might be somewhere else. Maybe your way to access the data set (by index and not by name) makes the difference.

EDIT 2: Works for 2-D as well. Added some info concerning the indexing

Upvotes: 4

Related Questions