How to control the buffer size for numpy array elements?

Question

The idea here is to dynamically increase the size (rows X columns) of an array (using hstack and vstack). However, I do not know the length of the strings that are about to be written at a specific position of the array at the moment I have to increase the size. Thus, the unknown elements are initialized with e.g. ' ' or 0 or something similar.

This leads to an error, if the placeholders are too short:

x = array([["1;", "2;"],["3;", "4;"]])
x[0][0] = "1234567890;"
print x

delivers:

[['12' '2;']
 ['3;' '4;']]

Many characters are missing at the position [0][0], whereas

x = array([["1;", "2;"],["abcdefghij;", "4;"]])
x[0][0] = "1234567890;"
print x

delivers the desired result, i.e.:

[['1234567890;' '2;']
 ['abcdefghij;' '4;']]

How can we handle this in Python? Many thanks in advance!

I did not think about using nested lists because I did not expect any difficulties here. Would you recommend the nested lists way? If yes, why? Thanks! — user3224270
– user3224270, Commented Jan 22, 2014 at 16:28
That is probably due to my lacking programming skill. The main reasons for choosing numpy arrays are the vstack and hstack methods. You could do it without them, though. — user3224270
– user3224270, Commented Jan 22, 2014 at 16:31
Its the thingy I myself often miss, but telling, what you finally want to achieve (the basic problem behind it) will let you receive help in the most precise way — embert
– embert, Commented Jan 22, 2014 at 16:36

M4rtini · Accepted Answer · 2014-01-22 17:08:54Z

1

Numpy arrays is not that good for dealing with non numerical types i believe, but if you still want to use it this is what you can do.

You can manually set the dtype of the array to either object or 'S#' where # will be the maximum number of characters.

In [19]: x = np.array([["1;", "2;"],["3;", "4;"]], dtype=object)

In [20]: x
Out[20]:
array([['1;', '2;'],
       ['3;', '4;']], dtype=object)

In [21]: x[0,0] = "1234567890;"

In [22]: x
Out[22]:
array([['1234567890;', '2;'],
       ['3;', '4;']], dtype=object)


In [24]: x = np.array([["1;", "2;"],["3;", "4;"]], dtype='S64')

In [25]: x
Out[25]:
array([['1;', '2;'],
       ['3;', '4;']],
      dtype='|S64')

In [26]: x[0,0] = "1234567890;"

In [27]: x
Out[27]:
array([['1234567890;', '2;'],
       ['3;', '4;']],
      dtype='|S64')

answered Jan 22, 2014 at 17:08

M4rtini

13.6k4 gold badges37 silver badges42 bronze badges

Sign up to request clarification or add additional context in comments.

1 Comment

user3224270 Over a year ago

That looks fine, thank you very much! I am going to try this tomorrow. However, as you also stated that numpy arrays are not the best choice for dealing with non-numerical types, would you recommend using nested lists or something else instead?

Collectives™ on Stack Overflow

How to control the buffer size for numpy array elements?

1 Answer 1

1 Comment

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

1 Comment

Your Answer

Sign up or log in

Post as a guest

Related