Memory Views

Introduction

Memoryview provides a view into a mutable buffer without copying data. It enables efficient slicing and type-casting of binary data.

Basic memoryview

# Create bytes-like object
data = bytearray(b"Hello World")

# Create memory view
mv = memoryview(data)

print(mv[0])        # 72 (ASCII 'H')
print(mv[0:5])      # memoryview of 'Hello'
print(bytes(mv[6:]))  # b'World'

# Modify through view
mv[0] = ord('J')
print(bytes(data))  # b"Jello World"

Slice Assignment

data = bytearray(10)
mv = memoryview(data)

# Assign single byte
mv[0] = 65

# Assign slice
mv[1:4] = b"ABC"

# Assign with different length (must match)
mv[5:8] = bytes([1, 2, 3])

print(bytes(data))  # b'ABC\x00\x01\x02\x03\x00'

Casting to Different Types

import struct

data = struct.pack("4I", 1, 2, 3, 4)
mv = memoryview(data)

# Cast to unsigned integers
int_view = mv.cast("I")
print(list(int_view))  # [1, 2, 3, 4]

# Cast to bytes
bytes_view = mv.cast("B")
print(list(bytes_view))  # [1, 0, 0, 0, 2, 0, 0, 0, ...]

Tobytes and Tolist

data = bytearray([1, 2, 3, 4, 5])
mv = memoryview(data)

# Convert to bytes (copy)
copy = mv.tobytes()
print(copy)  # b'\x01\x02\x03\x04\x05'

# Convert to list
lst = mv.tolist()
print(lst)  # [1, 2, 3, 4, 5]

# Convert to readonly
readonly = mv.toreadonly()

Use with NumPy

import numpy as np

# NumPy array to memoryview
arr = np.array([1, 2, 3, 4, 5], dtype=np.int32)
mv = memoryview(arr)

print(mv[0])  # 1
print(mv.shape)  # (5,)

# Modify through memoryview
mv[0] = 10
print(arr[0])  # 10

Practice Problems

1. Create memoryview from bytearray
2. Modify data through memoryview
3. Cast to different types
4. Use slice assignment
5. Share memory with NumPy