current community

your communities

more stack exchange communities

Stack Exchange
tour help
stack overflow careers
Take the 2-minute tour ×
Stack Overflow is a question and answer site for professional and enthusiast programmers. It's 100% free, no registration required.
up vote 5 down vote favorite
1

I have a numpy array which I want to convert from an object to complex. If I take that array as dtype string and convert it, there is no problem:

In[22]: bane
Out[22]: array(['1.000027337501943-7.331085223659654E-6j',
       '1.0023086995640738-1.8228368353755985E-4j',
       '-0.017014515914781394-0.2820013864855318j'], 
       dtype='|S41')

In [23]: bane.astype(dtype=complex)
Out[23]: 
array([ 1.00002734 -7.33108522e-06j,  1.00230870 -1.82283684e-04j,
       -0.01701452 -2.82001386e-01j])

But when it is dtype object and I try to convert it, I get an error that a float is required. Why is this?

In [24]: bane.astype(dtype=object)
Out[24]: 
array(['1.000027337501943-7.331085223659654E-6j',
       '1.0023086995640738-1.8228368353755985E-4j',
       '-0.017014515914781394-0.2820013864855318j'], dtype=object)

In [25]: _.astype(dtype=complex)
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-25-f5d89c8cc46c> in <module>()
----> 1 _.astype(dtype=complex)

TypeError: a float is required

To convert it, I use two calls to the astype method which seems clumsy:

bane_obj
Out[27]: 
array(['1.000027337501943-7.331085223659654E-6j',
       '1.0023086995640738-1.8228368353755985E-4j',
       '-0.017014515914781394-0.2820013864855318j'], dtype=object)

In [28]: bane_obj.astype(dtype=str).astype(dtype=complex)
Out[28]: 
array([ 1.00002734 -7.33108522e-06j,  1.00230870 -1.82283684e-04j,
       -0.01701452 -2.82001386e-01j])
share|improve this question
    
Sounds like a NumPy bug, maybe. –  Ignacio Vazquez-Abrams Mar 13 at 22:22
    
It's hard to tell where in the numpy code the conversion is taking place. The string to complex conversion has probably worked for a long time, while the development of object dtypes is newer and ongoing. What is generating this sort of object array? –  hpaulj Mar 14 at 1:09
3  
An array of dtype=np.object could contain literally any kind of Python object, so I think it's understandable that the numpy developeres have not bothered to implement every conceivable type conversion rule here. Chaining together an .astype(np.str).astype(np.complex) actually seems seem like a quite reasonable workaround for what must be a very uncommon use-case. The real question in my mind is why you're having to deal with an np.object array containing complex numbers in string format in the first place! –  ali_m Mar 14 at 18:05
3  
I don't think that could be the full explanation. Concatenating floats and complex floats should safely cast the result to complex. I can't see how you would end up with anything involving strings unless those columns in the DataFrame were already strings to begin with. –  ali_m Mar 14 at 23:28
2  
They were lists of strings to begin with, representing floats and complex values... sorry for the confusion. I went back and used a record to cast them up front as floats and complex, which as you suggested eliminated the need to deal with dtype=object. –  mataleo99 Mar 17 at 13:32

1 Answer 1

up vote 0 down vote

I think you might want to do the following, depending on what's inside the object type and if there's no padding to worry about:

bane.view(np.complex64) or
bane.view(np.complex128)

However if that does not work, which it didn't for some small tuple example I tried, the following worked:

bane.astype(np.float).view(np.complex64)

Consider using numpy structures rather than objects for the base dtype, you may have an easier time over all.

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.