how to add headers to a numpy.ndarray

Question

I have a numpy.ndarray having dimensions of 23411 x 3. I would like to add headers to the top of the matrix called: "summary", "age", and "label". In that order.

In:

matrix.shape

Out:

(23411L, 3L)

In:

type(matrix)

Out:

numpy.ndarray

I tried using the numpy.recarray but it did not work. any suggestions??

how did you try recarray? I can construct a structured array, but the process isn't trivial. And result isn't just added headers. What would you do with the array and headers? — hpaulj
– hpaulj, Commented Mar 1, 2018 at 19:23

sarthak · Accepted Answer · 2018-03-01 19:21:27Z

4

You can achieve this with pandas.

import pandas as pd
matrix = [...] # your ndarray

matrix = pd.DataFrame(data=matrix, columns=["summary", "age", "label"])

answered Mar 1, 2018 at 19:21

sarthak

5089 silver badges19 bronze badges

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

Comments

Paul Panzer · Accepted Answer · 2018-03-01 20:23:59Z

4

You can fiddle the dtype:

>>> a = np.arange(12).reshape(4, 3)
>>> 
>>> dt = a.dtype
>>> 
>>> ahead = a.view(np.dtype([('summary', dt), ('age', dt), ('label', dt)]))
>>> 
>>> ahead
array([[(0,  1,  2)],
       [(3,  4,  5)],
       [(6,  7,  8)],
       [(9, 10, 11)]],
      dtype=[('summary', '<i8'), ('age', '<i8'), ('label', '<i8')])
>>> ahead['summary']
array([[0],
       [3],
       [6],
       [9]])

But be warned that those composite dtype arrays are not very useful as far as I can tell:

>>> ahead @ ahead.T
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
TypeError: invalid data type for einsum

to give just one example.

answered Mar 1, 2018 at 20:23

Paul Panzer

53.3k3 gold badges59 silver badges103 bronze badges

Comments

ColonelFazackerley · Accepted Answer · 2018-03-01 19:36:20Z

1

Create your own class with members including the numpy array and the headings as strings.

answered Mar 1, 2018 at 19:36

ColonelFazackerley

9227 silver badges17 bronze badges

Comments

G.Brown · Accepted Answer · 2018-03-01 19:18:52Z

0

I think there is no way to do this. Try matrix.dtype. This should give you something like dtype('int32'). Your headers would be strings.

As far as I know you can only store one datatype in a numpy array/matrix.

answered Mar 1, 2018 at 19:18

G.Brown

2131 gold badge2 silver badges14 bronze badges

2 Comments

Mr. T Over a year ago

a = np.asarray([2, 2.1, "2"], dtype = object) Now test the type of each element.

G.Brown Over a year ago

@MrT Good point. This way you could force the 0th element to be a string with the name.

Collectives™ on Stack Overflow

how to add headers to a numpy.ndarray

4 Answers 4

Comments

Comments

Comments

2 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

4 Answers 4

Comments

Comments

Comments

2 Comments

Your Answer

Sign up or log in

Post as a guest

Related