How to change the data types of specific data type records in a NumPy structured array from a list of corresponding indices? (Python)

Question

I have NumPy structured array data:

data_tup = [tuple(ele) for ele in array]
data = np.array(
    data_tup, 
    dtype = [("field_1", "<U30"), ("field_2", "<U30"),
             ("field_3", "<U30"), ("field_4", "<U30"),
             ("field_5", "<U30"), ("field_6", "<U30"), 
             ("field_7", "<U30"), ("field_8", "<U30"), 
             ("field_9", "<U30")])       

I want to change the data types of all fields in data to float, except for fields specified by a list indices which holds the index values of the fields in dtype that should not change data types.

For example:

indices = [0, 4, 5, 7]
data = np.array(
data_tup, 
dtype = [("field_1", "<U30"), ("field_2", "<U30"),
         ("field_3", "<U30"), ("field_4", "<U30"),
         ("field_5", "<U30"), ("field_6", "<U30"), 
         ("field_7", "<U30"), ("field_8", "<U30"), 
         ("field_9", "<U30")])     

New dtype in data should now be:

 dtype = [("field_1", "<U30"), ("field_2", float),
         ("field_3", float), ("field_4", float),
         ("field_5", "<U30"), ("field_6", "<U30"), 
         ("field_7", float), ("field_8", "<U30"), 
         ("field_9", float)])     

Answer 1

A list of tuples is the easiest way to convert between dtypes like this:

In [63]: data = [('123','456')]
In [64]: np.array(data, dtype='U10,U10')
Out[64]: array([('123', '456')], dtype=[('f0', '<U10'), ('f1', '<U10')])
In [65]: np.array(data, dtype='U10,float')
Out[65]: array([('123', 456.)], dtype=[('f0', '<U10'), ('f1', '<f8')])

If starting with the all-string dtype array:

In [77]: np.array(data, dtype='U10,U10').tolist()
Out[77]: [('123', '456')]