TheDeveloperBlog.com

Home | Contact Us

C-Sharp | Java | Python | Swift | GO | WPF | Ruby | Scala | F# | JavaScript | SQL | PHP | Angular | HTML

Python Array Type Examples: Integer Performance

This Python page uses the array type to improve memory efficiency of numerical data. It provides examples and benchmarks.

Arrays. Computational processes often require lots of numbers.

A list in Python can store these values. But it uses excessive memory.

For efficiency, the array type in Python is better. We use arrays. We test them and measure performance. Code with arrays is more complex.

An example. Here we start. Please add the "from import" statement at the top of the program. This lets you access the array type.

Constructor: The array constructor receives two arguments here: the first is a code. The code "i" here signifies an integer array.

And: In this example the second argument to array() is a list. The array will be made of integers equal to those specified in the list.

Finally: We use a for-loop to enumerate, and print, the elements in the array. This is the same syntax used for a list.

Based on:

Python 3

Python program that creates int array

from array import array

# Create an int array of three elements.
a = array("i", [10, 20, 30])

# Display elements in array.
for value in a:
    print(value)

Output

10
20
30

String. Suppose you have a large string you want to store in an array. We can use the "u" type code. Here the string "python" is transformed into an array of six Unicode characters.

Tolist: We convert the array into a list with the tolist() method. We then join with an empty delimiter.

And: This yields a string. This approach can convert an array (or list) into a usable string.

String: Join

Python program that creates char array

from array import array

# Create a Unicode char array.
a = array("u", "python")

# Display letters in array.
for letter in a:
    print(letter)

# Convert array to a list.
# ... Then join it.
s = "".join(a.tolist())
print(s)

Output

p
y
t
h
o
n
python

Append, insert. Many helpful methods are available on arrays. For example, the append method adds an element at the end of the array. We can also insert, remove and count.

Note: We create an empty int array in the first part. The second argument to the array init method is optional.

Python program that uses append, insert, remove, count

from array import array

# New int array.
a = array("i")

# Append three integers.
a.append(100)
a.append(200)
a.append(300)

# Insert an integer at index 1.
a.insert(1, 900)

# Remove this element.
a.remove(200)

# Count elements with this value.
a.count(900)

# Print.
print(a)

Output

array('i', [100, 900, 300])

Memory. I tested two programs. The first version uses an array of ten million integers. And the second uses a list of those same integers. I measured the memory before the program exited.

Result: The array with 10 million integers required 43.8 MB of memory. The list version required 710.9 MB.

Memory of elements. We can estimate how much each element used. For 10 million ints, the array used about 43 MB—about 4 bytes per integer. The list required closer to 70 bytes per element.

Caution: These are just estimates. Many factors, including garbage collection overhead, are in play.

And: One megabyte is not exactly 1 million bytes. It is 1,048,576 bytes. So the math is lacking.

Python program that uses int array

from array import array

# Create an empty int array.
a = array("i")

# Append ten million ints.
for i in range(0, 10000000):
    a.append(i)

# Finished.
print("DONE")
input()

Python program that uses int list

# Create an empty list.
a = list()

# Append ints.
for i in range(0, 10000000):
    a.append(i)

# Finished.
print("DONE")
input()

Memory usage

 43.8 MB    Array
710.9 MB    List

Array benefits. Performance and memory usage were both vastly improved with the array. In testing this program, the speed of array is also much better.

Thus: Using an array instead of a list in a numerical program could make the program many times faster.

And: It could make a program that was previously unusable, usable. It is vastly superior in some contexts.

However: Avoid premature optimization. The performance benefits of arrays will not be measurable on smaller collections.

Speed. Is a small array faster than an equivalent list? Should we optimize programs by using more arrays? I found that looping over small arrays is actually slower than looping over lists.

Thus: Arrays are slower than lists for small collection sizes. We reduce speed by using arrays on small amounts of data.

But: For large amounts of data, an array will use less memory and this in turn improves performance. It improves memory locality.

Python program that uses int array, list

import time
from array import array

# Create an int array, and a list.
list = list(range(0, 50))
arr = array("i", list)

print(time.time())

# Version 1: loop over array.
for i in range(0, 1000000):
    sum = 0
    for value in arr:
	sum += value

print(time.time())

# Version 2: loop over list.
for i in range(0, 1000000):
    sum = 0
    for value in list:
	sum += value

print(time.time())

Output

1408894795.864125   [Python]
1408894803.765587   Array loop, 7.90 s
1408894811.303026   List loop,  7.53 s

1408894907.587      [PyPy3]
1408894908.066      Array loop, 0.48 s
1408894908.534      List loop,  0.47 s

Codes, bytes. There are many type codes available on arrays. Please consult the official Python documentation for a descriptive table. The number of bytes for each element type is shown.

Tip: A good developer should reference all helpful websites, not just ones with pictures.

Arrays are sequence types and behave very much like lists, except that the type of objects stored in them is constrained.

Efficient arrays of numeric values: Python.org

Essential. Arrays are essential in programs that compute large numerical data sets. We witnessed the memory improvement of an array over a list.

We used array methods, which are similar (but not identical) to lists. The types are complementary. For smaller data sets, lists are sufficient. And lists are friendlier in most ways.


Related Links

Adjectives Ado Ai Android Angular Antonyms Apache Articles Asp Autocad Automata Aws Azure Basic Binary Bitcoin Blockchain C Cassandra Change Coa Computer Control Cpp Create Creating C-Sharp Cyber Daa Data Dbms Deletion Devops Difference Discrete Es6 Ethical Examples Features Firebase Flutter Fs Git Go Hbase History Hive Hiveql How Html Idioms Insertion Installing Ios Java Joomla Js Kafka Kali Laravel Logical Machine Matlab Matrix Mongodb Mysql One Opencv Oracle Ordering Os Pandas Php Pig Pl Postgresql Powershell Prepositions Program Python React Ruby Scala Selecting Selenium Sentence Seo Sharepoint Software Spellings Spotting Spring Sql Sqlite Sqoop Svn Swift Synonyms Talend Testng Types Uml Unity Vbnet Verbal Webdriver What Wpf