Zero null-masked bytes of a fixed-width array in PyArrow

zero_null-masked_bytes.py

import pyarrow as pa
import numpy as np
import pandas as pd

# Create an array of some fixed-width type containing nulls
a = pa.array(obj=pd.Series([1, 2, 3]), type=pa.int64(), mask=np.array([1, 0, 1], dtype=bool))

# Get the values buffer as a bytearray
b = a.buffers()
v = bytearray(b[1].to_pybytes())

# For each null-masked value...
for i in range(0, len(a)):
	if a.is_null()[i] == pa.scalar(True):
        
		# ...locate the associated bytes in the bytearray
		bytes_start = i * a.type.byte_width
		bytes_end = bytes_start + a.type.byte_width
        
		# Examine the bytes and notice they are not zeroed
		print('Original bytes ' + v[bytes_start:bytes_end].hex(), end=' ')
        
		# Replace them with zero bytes
		v[bytes_start:bytes_end] = bytearray(bytes_length)
		print('zeroed')

# Replace the values buffer and reassemble the array
b[1] = pa.py_buffer(v)
a_new = pa.IntegerArray.from_buffers(a.type, len(a), b)

# Notice PyArrow says the new array is the same as the original
a_new == a