Memory-mapped files with np.memmap in NumPy - Practice Problems & Coding Challenges

Challenge - 5 Problems

❓ Predict Output

intermediate

Output of np.memmap array modification

What is the output of the following code snippet?

NumPy

import numpy as np
filename = 'data.dat'
# Create a memmap file with 5 integers
arr = np.memmap(filename, dtype='int32', mode='w+', shape=(5,))
arr[:] = np.arange(5)
arr.flush()
# Reopen in read mode
arr2 = np.memmap(filename, dtype='int32', mode='r', shape=(5,))
print(arr2[2])

A 2

B 0

C 3

D Raises a ValueError

💡 Hint

Remember that np.memmap writes data to disk and reopening in read mode reads the saved data.

The memmap file is created and filled with values 0 to 4. Reopening in read mode reads the saved data, so arr2[2] is 2.

❓ data_output

intermediate

Shape of np.memmap array after reshaping

Given this code, what is the shape of the memmap array after reshaping?

NumPy

import numpy as np
filename = 'data2.dat'
arr = np.memmap(filename, dtype='float64', mode='w+', shape=(12,))
arr[:] = np.arange(12)
arr.flush()
arr2 = np.memmap(filename, dtype='float64', mode='r+', shape=(3,4))
print(arr2.shape)

A (4, 3)

B (12,)

C (3, 4)

D Raises a TypeError

💡 Hint

The shape parameter defines how the data is viewed, not the original file size.

The memmap is reopened with shape (3,4), so arr2.shape is (3,4).

🔧 Debug

advanced

Identify the error in np.memmap usage

What error does this code raise?

NumPy

import numpy as np
filename = 'data3.dat'
arr = np.memmap(filename, dtype='int16', mode='w+', shape=(10,))
arr[:] = np.arange(10)
arr.flush()
# Reopen with wrong dtype
arr2 = np.memmap(filename, dtype='int32', mode='r', shape=(10,))
print(arr2[0])

A ValueError: cannot reshape array of size 20 into shape (10,)

B TypeError: dtype mismatch

C No error, prints 0

D ValueError: buffer size mismatch

💡 Hint

The file was saved with int16 (2 bytes per element), but reopened as int32 (4 bytes per element).

The memmap file size does not match the expected size for int32 with shape (10,), causing a buffer size mismatch error.

🚀 Application

advanced

Efficiently modifying large data with np.memmap

You have a large binary file with 1 million float64 numbers. You want to add 10 to every element without loading the entire file into memory. Which code snippet correctly does this?

A

arr = np.memmap('large.dat', dtype='float64', mode='r')
arr += 10
arr.flush()

B

arr = np.memmap('large.dat', dtype='float64', mode='r+')
for i in range(len(arr)):
    arr[i] += 10
arr.flush()

C

arr = np.memmap('large.dat', dtype='float64', mode='r+')
arr += 10
arr.flush()

D

arr = np.memmap('large.dat', dtype='float64', mode='w+')
arr[:] = 10
arr.flush()

💡 Hint

Mode 'r+' allows read and write. Vectorized operations may load all data into memory.

Option B modifies elements one by one, avoiding loading entire file into memory. Option B uses vectorized addition which loads all data, Option B is read-only mode, Option B overwrites all data with 10.

🧠 Conceptual

expert

Understanding np.memmap file persistence

Which statement about np.memmap file persistence is TRUE?

A Data changes in a memmap array are saved to disk only after calling flush()

B Data changes in a memmap array are immediately saved to disk without flush()

C Memmap arrays do not save data to disk; they only cache in memory

D Calling flush() deletes the memmap file from disk

💡 Hint

Think about when data is guaranteed to be written to disk.

flush() forces changes to be written to disk. Without flush(), changes may remain in OS cache and not be saved immediately.