Assign column and row labels to merged numpy arrays?

Question

I have multiple arrays without column or row names, and I would like to combine them using something like numpy.vstack() or numpy.hstack().

Column and row labels can be assigned can be done when creating a structured array, but hstack and vstack don't seem to have this functionality.

import numpy as np
a1 = np.array([1,2,3,4])
a2 = np.array([5,6,7,8])
a3 = np.vstack([a1,a2],dtype=[('RowName1','double'),('RowName2','double')])

yielding:

TypeError: vstack() got an unexpected keyword argument 'dtype'

Any suggestions?

seberg · Accepted Answer · 2012-12-12 11:15:52Z

2

Something possible option is (since recfunctions are pretty hidden):

from numpy.lib import recfunctions
a1 = np.array([1,2,3,4]).astype(('RowName1',float))
a2 = np.array([5,6,7,8]).astype(('RowName2',float))
recfunctions.merge_arrays((a1, a2))

Had this, but this has a few problems to be careful with, because of how reinterpretation of memory works with view, its better to just create a new recarray with the concatenated array.

you could just turn around the logic:

import numpy    
a1 = np.array([1,2,3,4])
a2 = np.array([5,6,7,8])
# ok, not that beautiful. But if your arrays are the correct type to begin with
# you can skip that astype call. Using `np.c_[]` since it happens to concatenate right.
a3 = np.c_[v1,v2].astype(float).copy('C').view(dtype=[('RowName1',float),('RowName2',float)])

edited Dec 12, 2012 at 11:15

answered Dec 12, 2012 at 10:36

seberg

9,0352 gold badges34 silver badges30 bronze badges

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

4 Comments

tiago Over a year ago

I don't think that works as the OP wanted. First, it will view integer arrays as double (ie, garbled), and it will have the wrong shape because the wanted index is by row and not by column.

seberg Over a year ago

@tiango, right :/... I didn't check that vstack wasn't really right to begin with, and without explicit cast to the right type, the view is bad of course.

tiago Over a year ago

it still gives a wrong result. If you look at your a3['RowName1'], it gives a 2D array: array([[ 1., 5.], [ 3., 7.]]). I'm not trying to criticise your answers, just genuinely curious. Even if a1, a2 are of the same type and concatenated in the right direction, it doesn't seem possible to get a view per row -- the shape is wrong.

seberg Over a year ago

@tiango yeah sorry, it was not a good idea with view. The reason is that the concatenation array is not C-Contiguous and view only reinterprets the underlying memory, but because it is not C-Contiguous that is not the same as the wanted result... Its actually a good example why one needs to be careful with view...

durden2.0 · Accepted Answer · 2012-12-12 14:42:32Z

2

You might also consider looking at pandas. Pandas has a nice data frame data structure that might be good.

Of course, this requires you to add another dependency to your project. Luckily, if you are already using numpy then Pandas is pretty easy to get going.

answered Dec 12, 2012 at 14:42

durden2.0

9,58212 gold badges49 silver badges58 bronze badges

Comments

tiago · Accepted Answer · 2012-12-12 10:21:04Z

1

vstack doesn't work with structured arrays, but only with 'standard' numpy arrays that are contiguous in memory. The easiest way is for you to create an empty structured array and then fill it up with the rows that you want:

import numpy as np
a3 = np.empty(4, dtype=[('RowName1','double'),('RowName2','double')])
a3['RowName1'] = a1
a3['RowName2'] = a2

answered Dec 12, 2012 at 10:21

tiago

23.7k13 gold badges75 silver badges89 bronze badges

Collectives™ on Stack Overflow

Assign column and row labels to merged numpy arrays?

3 Answers 3

4 Comments

Comments

Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

3 Answers 3

4 Comments

Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related