C-like structure in Python

Introduction

A technique that can be used when you are in agony and want to use a C-like structure in Python. For example, if you want to create a structure like ↓.

c_struct.h

typedef struct{
    float pos[4];
    unsigned char class;
    float score;
} HOGE;

HOGE fuga[10];

Expression method in Python

Use Structured Data types

• https://numpy.org/doc/stable/user/basics.rec.html#structured-datatypes
• https://jakevdp.github.io/PythonDataScienceHandbook/02.09-structured-data-numpy.html

c_like_struct.py

import numpy as np

HOGE = np.dtype([("pos",    np.float32, (4,)),
                 ("class",  np.uint8,   (1,)),
                 ("score",  np.float32, (1,)),
                 ],
                align=True
                )
fuga = np.zeros(10, dtype=HOGE)

With this, a C language-like structure can be expressed. Padding is done by setting align = True above.

How to use ① (output for the time being)

processing

print(HOGE)
print(fuga)

result

{'names':['pos','class','score'], 'formats':[('<f4', (4,)),('u1', (1,)),('<f4', (1,))], 'offsets':[0,16,20], 'itemsize':24, 'aligned':True}
[([0., 0., 0., 0.], [0], [0.]) ([0., 0., 0., 0.], [0], [0.])
 ([0., 0., 0., 0.], [0], [0.]) ([0., 0., 0., 0.], [0], [0.])
 ([0., 0., 0., 0.], [0], [0.]) ([0., 0., 0., 0.], [0], [0.])
 ([0., 0., 0., 0.], [0], [0.]) ([0., 0., 0., 0.], [0], [0.])
 ([0., 0., 0., 0.], [0], [0.]) ([0., 0., 0., 0.], [0], [0.])]

How to use ② (dtype, shape)

processing

print("fuga.dtype: ",fuga.dtype)
print("fuga.shape: ",fuga.shape)
print("fuga.['pos'].shape: ",fuga['pos'].shape)
print("fuga.['pos'][0].shape: ",fuga['pos'][0].shape)
print("fuga['pos'][1]: ", fuga['pos'][1])

result

fuga.dtype:  {'names':['pos','class','score'], 'formats':[('<f4', (4,)),('u1', (1,)),('<f4', (1,))], 'offsets':[0,16,20], 'itemsize':24, 'aligned':True}
fuga.shape:  (10,)
fuga['pos'].shape:  (10, 4)
fuga['pos'][0].shape:  (4,)
fuga['pos'][1]:  [0. 0. 0. 0.]

How to use ③ (np.recarray)

processing

piyo = fuga.view(np.recarray)
print("piyo: ", piyo)
print("piyo.pos.shape: ", piyo.pos.shape)
print("piyo.pos[0].shape: ", piyo.pos[0].shape)
print("piyo.pos[1]: ",piyo.pos[1])

result

piyo:  [([0., 0., 0., 0.], [0], [0.]) ([0., 0., 0., 0.], [0], [0.])
        ([0., 0., 0., 0.], [0], [0.]) ([0., 0., 0., 0.], [0], [0.])
        ([0., 0., 0., 0.], [0], [0.]) ([0., 0., 0., 0.], [0], [0.])
        ([0., 0., 0., 0.], [0], [0.]) ([0., 0., 0., 0.], [0], [0.])
        ([0., 0., 0., 0.], [0], [0.]) ([0., 0., 0., 0.], [0], [0.])]
piyo.pos.shape:  (10, 4)
piyo.pos[0].shape:  (4,)
piyo.pos[1]:  [0. 0. 0. 0.]

Usage check ④ (assignment process)

processing

fuga['pos'][1] = [-1.0 , 2.3, 0.42, 0.0]
print("fuga['pos'][:3]: ",fuga['pos'][:3])

result

fuga['pos'][:3]:  [[ 0.    0.    0.    0.  ]
                   [-1.    2.3   0.42  0.  ]
                   [ 0.    0.    0.    0.  ]]