setting type of numpy structure array

Question

I have a numpy structured array where I need to control the data types of the individual elements.

This is my example:

y = array([(True, 144.0), 
           (True, 86.0), 
           (True, 448.0), 
           (True, 76.0), 
           (True, 511.0)], dtype=object)

If I do:

print(y.dtype.fields)

I get back:

None

However, what I wanted was "bool" and "float". If I access the individual elements, such as y[0][0] and y[0][1] I surely see that they are indeed bool and float.
I am super confused with this. Any ideas?

I need this because I use the packge "sciki-survival gradient boosting":https://scikit-survival.readthedocs.io/en/latest/generated/sksurv.ensemble.GradientBoostingSurvivalAnalysis.html#sksurv.ensemble.GradientBoostingSurvivalAnalysis.fit Where the input needs to be a structured array of type "bool" and "float".

It gives dtype('O'), ie np.object_. This isn't a structured array — Eric
– Eric, Commented Jul 11, 2018 at 9:17
@Eric How do do I build a structured array then with the same content? — Kim O
– Kim O, Commented Jul 11, 2018 at 9:21
@KimO, Looks like you are using the wrong answer from your previous question (or didn't read my comment to that answer). Take one step at a time. Understand each step as you go along. This will help you resolve problems more quickly. — jpp
– jpp, Commented Jul 11, 2018 at 9:38

Eric · Accepted Answer · 2018-07-11 09:22:10Z

2

I have a numpy structured array

No you don't:

np.array(..., dtype=object)

You have a numpy object array, which contains tuples.

You can convert it to a structured array with y.astype([('b', bool), ('f', float)])

answered Jul 11, 2018 at 9:22

Eric

98.1k54 gold badges257 silver badges389 bronze badges

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

Comments

lhay86 · Accepted Answer · 2018-07-11 09:22:00Z

1

When you initialize the structured array, ensure that you specify the data types.

For example:

y = np.array([(True, 144.0), (True, 86.0), (True, 448.0)],
          dtype=[('col_1', 'bool'), ('col_2', 'f4')])

This should work and:

y.dtype.fields

shows as desired:

mappingproxy({'col_1': (dtype('bool'), 0), 'col_2': (dtype('float32'), 1)})

See the documentation here: https://docs.scipy.org/doc/numpy/user/basics.rec.html

answered Jul 11, 2018 at 9:22

lhay86

7262 gold badges5 silver badges19 bronze badges

Comments

juanpa.arrivillaga · Accepted Answer · 2018-07-11 09:22:20Z

To create a structured array, you must specify the dtype beforehand. If you just use numpy.array with a list-literal of pairs then you will get an array with object dtype. So, you need to do something like:

>>> mytype = np.dtype([('b', bool), ('f',float)])
>>> mytype
dtype([('b', '?'), ('f', '<f8')])

Then pass mytype to the array constructor:

>>> structured = np.array(
...    [(True, 144.0), (True, 86.0),
...     (True, 448.0), (True, 76.0),
...     (True, 511.0), (True, 393.0), 
...     (False, 466.0), (False, 470.0)], dtype=mytype)
>>>
>>> structured
array([( True,  144.), ( True,   86.), ( True,  448.), ( True,   76.),
       ( True,  511.), ( True,  393.), (False,  466.), (False,  470.)],
      dtype=[('b', '?'), ('f', '<f8')])

Collectives™ on Stack Overflow

setting type of numpy structure array

3 Answers 3

Comments

Comments

Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

3 Answers 3

Comments

Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related