Record data type for heterogeneous Numpy arrays

Question

I am working with numpy and to create a heterogeneous array it is necesary to use dtype() function. After read some python doc about this function I think I know how it works;

t = np.dtype([('name', str),('edad',int)]) <-- This tells to python that my array will have a new data type with a string named 'name' and an int named 'edad'.


R = np.array([('Rosendo',15)]) <-- And now everything I put with this other python will try to convert to str and int.

Is this the correct way to create heterogeneous arrays? My array items have to be always tuples? I saw some people code like this:

t = dtype([('name', str_, 40), ('numitems', int32), ('price',float32)])

But but this not gonna work, what about the "40" on ('name', str_, 40). Is there other ways to create heterogeneous arrays using dtype()?

Answer 1

To create a structured array you need to specify the compound dtype, and the data as a list of tuples:

In [98]: dt = np.dtype([('name', 'U10'),('edad',int)])                                               
In [99]: R = np.array([('Rosendo',15)], dtype=dt)                                                    
In [100]: R                                                                                          
Out[100]: array([('Rosendo', 15)], dtype=[('name', '<U10'), ('edad', '<i8')])
In [101]: R['name']                                                                                  
Out[101]: array(['Rosendo'], dtype='<U10')
                                

Often it is easier to specify data by field:

In [102]: R = np.zeros((3,), dt)                                                                     
In [103]: R                                                                                          
Out[103]: 
array([('', 0), ('', 0), ('', 0)],
      dtype=[('name', '<U10'), ('edad', '<i8')])
In [104]: R['name'] = ['one','two','three']                                                          
In [105]: R['edad'] = [101, 93, 3]                                                                   
In [106]: R                                                                                          
Out[106]: 
array([('one', 101), ('two',  93), ('three',   3)],
      dtype=[('name', '<U10'), ('edad', '<i8')])