Converting a Python List into a Numpy Structured array?

Question

I need to convert a Python list with mixed type (integers and floats) to a Numpy structured array with certain column names.

I have tried the following code but for some reason I cant see it doesn't work.

import numpy as np
lmtype = [('el','intp',1),  ('n1','intp',1),  ('n2','intp',1),  ('n3','float64',1),
          ('n4','float64',1),  ('n5','float64',1),  ('n6','float64',1),  ('n7','float64',1),
          ('n8','float64',1),  ('n9','float64',1),  ('n10','float64',1),  ('n11','float64',1)]
LAMI = np.zeros(5, dtype=lmtype)
linea = ['1', '2', '3', '0.0', '0.0', '0.0', '0.0', '0.0', '0.0', '0.0', '0.0', '0.0']
for idx, la in enumerate(LAMI):
    lineanum = ([ int(j) for j in linea[0:3] ] + [float(i) for i in linea[3:12] ] )
    print lineanum
    LAMI[idx] = np.array( lineanum )

The code runs, but look what LAMI has inside:

>>> LAMI[0]
(0, 1072693248, 0, 5.304989477e-315, 5.307579804e-315, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0)

Answer 1

Try:

LAMI[idx] = tuple( lineanum )

Tuples are the normal way to assign to a record (row) of a structured array. They can hold mixed types, just like the structured element.

np.array(lineanum) is all float . LAMI[idx] = np.array( lineanum ) just copies the buffer full of floats to a segment of the LAMI buffer. I'm a little surprised that it permits this copy; it must be doing some sort of 'copy only as much as fits'. LAMI.itemsize is 84 , while the total length of np.array(lineanum) is 12*8=96.

LAMI[0]=np.array(lineanum, dtype=int) # or better
LAMI[0]=np.array(lineanum[:3], type=int)

would also work, since the only nonzero values are those 1st 3 which are suppposed to be ints.