I have a list of strings string_list that I want to convert to a 1-element structured numpy array of pre-specified dtype:
string_list = ['123', '45', '6.7']
my_dtype = np.dtype([('a','i4'),('b','f4'),('c','f8')])
The output result I want is:
array([(123, 45., 6.7)], dtype=[('a', '<i4'), ('b', '<f4'), ('c', '<f8')])
Is there a np.astype syntax trick for this conversion?
Input for 'rows' of structure arrays is tuples:
In [464]: np.array([tuple(string_list)],dtype=my_dtype)
Out[464]:
array([(123, 45.0, 6.7)],
dtype=[('a', '<i4'), ('b', '<f4'), ('c', '<f8')])
Another useful trick is to create an empty array of the required dtype, and fill in fields, one at a time. Or fill row by row with tuples, or multiple rows with a list of tuples.
EDIT see @hpaulj answer. List has to be converted to tuple and list generates error.
with me this works, is that what you want?
string_list = ('123', '45.', '6.7')
b = np.array([string_list], dtype=[('a', '<i4'), ('b', '<f4'), ('c', '<f8')])
numpy version 1.8.2
