Awesome

ArithmeticEncodingPython

This project implements the lossless data compression technique called arithmetic encoding (AE). The project is simple and has just some basic features.

The project supports encoding the input as both a floating-point value and a binary code.

The project has a main module called pyae.py which contains a class called ArithmeticEncoding to encode and decode messages.

Usage Steps

To use the project, follow these steps:

1. Import pyae
2. Instantiate the ArithmeticEncoding Class
3. Prepare a Message
4. Encode the Message
5. Get the binary code of the encoded message.
6. Decode the Message

Import pyae

The first step is to import the pyae module.

import pyae

Instantiate the ArithmeticEncoding Class

Create an instance of the ArithmeticEncoding class. Its constructor accepts 2 arguments:

1. frequency_table: The frequency table as a dictionary where key is the symbol and value is the frequency.
2. save_stages: If True, then the intervals of each stage are saved in a list. Note that setting save_stages=True may cause memory overflow if the message is large

According to the following frequency table, the messages to be encoded/decoded must have only the 3 characters a, b, and c.

frequency_table = {"a": 2,
                   "b": 7,
                   "c": 1}

AE = pyae.ArithmeticEncoding(frequency_table=frequency_table,
                            save_stages=True)

Prepare a Message

Prepare the message to be compressed. All the characters in this message must exist in the frequency table.

original_msg = "abc"

Encode the Message

Encode the message using the encode() method. It accepts the message to be encoded and the probability table. It returns the encoded message (single double value) and the encoder stages.

encoded_msg, encoder , interval_min_value, interval_max_value = AE.encode(msg=original_msg, 
                                                                          probability_table=AE.probability_table)

Get the Binary Code of the Encoded Message

Convert the floating-point value returned from the AE.encode() function into a binary code using the AE.encode_binary() function.

binary_code, encoder_binary = AE.encode_binary(float_interval_min=interval_min_value,
                                               float_interval_max=interval_max_value)

Decode the Message

Decode the message using the decode() method. It accepts the encoded message, message length, and the probability table. It returns the decoded message and the decoder stages.

decoded_msg, decoder = AE.decode(encoded_msg=encoded_msg, 
                                 msg_length=len(original_msg),
                                 probability_table=AE.probability_table)

Note that the symbols in the decoded message are returned in a list. If the original message is a string, then consider converting the list into a string using join() function as follows.

decoded_msg = "".join(decoded_msg)

<u>IMPORTANT</u>: double Module

The floating-point numbers in Python are limited to a certain precision. Beyond it, Python cannot store any additional decimal numbers. This is why the project uses the double data type offered by the decimal module.

The decimal module has a class named Decimal that can use any precision. The precision can be changed using the prec attribute as follows:

getcontext().prec = 50

The precision defaults to 28. It is up to the user to set the precision to any value that serves the application. Note that the precision only affects the arithmetic operations.

For more information about the decimal module, check its documentation: https://docs.python.org/2/library/decimal.html

Example

The example.py script has an example that compresses the message abc using arithmetic encoding. The precision of the decimal data type is left to the default value 28 as it can encode the message abc without losing any information.

import pyae

# Example for encoding a simple text message using the PyAE module.
# This example returns the floating-point value in addition to its binary code that encodes the message. 

frequency_table = {"a": 2,
                   "b": 7,
                   "c": 1}

AE = pyae.ArithmeticEncoding(frequency_table=frequency_table,
                            save_stages=True)

original_msg = "abc"
print("Original Message: {msg}".format(msg=original_msg))

# Encode the message
encoded_msg, encoder , interval_min_value, interval_max_value = AE.encode(msg=original_msg, 
                                                                          probability_table=AE.probability_table)
print("Encoded Message: {msg}".format(msg=encoded_msg))

# Get the binary code out of the floating-point value
binary_code, encoder_binary = AE.encode_binary(float_interval_min=interval_min_value,
                                               float_interval_max=interval_max_value)
print("The binary code is: {binary_code}".format(binary_code=binary_code))

# Decode the message
decoded_msg, decoder = AE.decode(encoded_msg=encoded_msg, 
                                 msg_length=len(original_msg),
                                 probability_table=AE.probability_table)
decoded_msg = "".join(decoded_msg)
print("Decoded Message: {msg}".format(msg=decoded_msg))
print("Message Decoded Successfully? {result}".format(result=original_msg == decoded_msg))

The printed messages out of the code are:

Original Message: abc
Encoded Message: 0.1729999999999999989175325511
The binary code is: 0.0010110
Decoded Message: abc
Message Decoded Successfully? True

So, the message abc is encoded using the double number 0.173.

It is possible to print the encoder to get information about the stages of the encoding process. The encoder is a list of dictionaries where each dictionary represents a stage.

print(encoder)

[{'a': [Decimal('0'), Decimal('0.6999999999999999555910790150')],
  'b': [Decimal('0.6999999999999999555910790150'),
   Decimal('0.7999999999999999611421941381')],
  'c': [Decimal('0.7999999999999999611421941381'),
   Decimal('0.9999999999999999722444243844')]},
 {'a': [Decimal('0'), Decimal('0.4899999999999999378275106210')],
  'b': [Decimal('0.4899999999999999378275106210'),
   Decimal('0.5599999999999999372723991087')],
  'c': [Decimal('0.5599999999999999372723991087'),
   Decimal('0.6999999999999999361621760841')]},
 {'a': [Decimal('0.4899999999999999378275106210'),
   Decimal('0.5389999999999999343303080934')],
  'b': [Decimal('0.5389999999999999343303080934'),
   Decimal('0.5459999999999999346633750008')],
  'c': [Decimal('0.5459999999999999346633750008'),
   Decimal('0.5599999999999999353295088156')]},
 {'a': [Decimal('0.5459999999999999346633750008'),
   Decimal('0.5557999999999999345079437774')],
  'b': [Decimal('0.5557999999999999345079437774'),
   Decimal('0.5571999999999999346522727706')],
  'c': [Decimal('0.5571999999999999346522727706'),
   Decimal('0.5599999999999999349409307570')]}]

Here is the binary encoder:

print(encoder_binary)

[{0: ['0.0', '0.1'], 1: ['0.1', '1.0']},
 {0: ['0.00', '0.01'], 1: ['0.01', '0.1']},
 {0: ['0.000', '0.001'], 1: ['0.001', '0.01']},
 {0: ['0.0010', '0.0011'], 1: ['0.0011', '0.01']},
 {0: ['0.00100', '0.00101'], 1: ['0.00101', '0.0011']},
 {0: ['0.001010', '0.001011'], 1: ['0.001011', '0.0011']},
 {0: ['0.0010110', '0.0010111'], 1: ['0.0010111', '0.0011']}]

Low Precision

Assume the message to be encoded is "abc"*20 (i.e. abc repeated 20 times) while using the default precision 28. The length of the message is 60.

original_msg = "abc"*20

Here is the code that uses this new message.

import pyae

frequency_table = {"a": 2,
                   "b": 7,
                   "c": 1}

AE = pyae.ArithmeticEncoding(frequency_table=frequency_table,
                            save_stages=True)

original_msg = "abc"*20
print("Original Message: {msg}".format(msg=original_msg))

encoded_msg, encoder , interval_min_value, interval_max_value = AE.encode(msg=original_msg, 
                                                                          probability_table=AE.probability_table)
print("Encoded Message: {msg}".format(msg=encoded_msg))

decoded_msg, decoder = AE.decode(encoded_msg=encoded_msg, 
                                 msg_length=len(original_msg),
                                 probability_table=AE.probability_table)
decoded_msg = "".join(decoded_msg)
print("Decoded Message: {msg}".format(msg=decoded_msg))
print("Message Decoded Successfully? {result}".format(result=original_msg == decoded_msg))

By running the previous code, here are the results of the print statements. The decoded message is different from the original message. The reason is that the current precision of 28 is not sufficient to encode a message of length 60.

Original Message: abcabcabcabcabcabcabcabcabcabcabcabcabcabcabcabcabcabcabcabc
Encoded Message: 0.1683569979716024329522342419
Decoded Message: abcabcabcabcabcabcabcabcabcabcabcabcabcabcabbcbbbbbbbbbbbbbb
Message Decoded Successfully? False

In this case, the precision should be increased. Here is how to change the precision to be 45:

from decimal import getcontext

getcontext().prec = 45

Here is the new code after increasing the precision of the Double data type:

import pyae
from decimal import getcontext

getcontext().prec = 45

frequency_table = {"a": 2,
                   "b": 7,
                   "c": 1}

AE = pyae.ArithmeticEncoding(frequency_table=frequency_table,
                            save_stages=True)

original_msg = "abc"*20
print("Original Message: {msg}".format(msg=original_msg))

encoded_msg, encoder , interval_min_value, interval_max_value = AE.encode(msg=original_msg, 
                                                                          probability_table=AE.probability_table)
print("Encoded Message: {msg}".format(msg=encoded_msg))

decoded_msg, decoder = AE.decode(encoded_msg=encoded_msg, 
                                 msg_length=len(original_msg),
                                 probability_table=AE.probability_table)
decoded_msg = "".join(decoded_msg)
print("Decoded Message: {msg}".format(msg=decoded_msg))
print("Message Decoded Successfully? {result}".format(result=original_msg == decoded_msg))

After running the code, here are the results where the original message is restored successfully:

Original Message: abcabcabcabcabcabcabcabcabcabcabcabcabcabcabcabcabcabcabcabc
Encoded Message: 0.168356997971602432952234241597600194030293262
Decoded Message: abcabcabcabcabcabcabcabcabcabcabcabcabcabcabcabcabcabcabcabc
Message Decoded Successfully? True

Contact Us

• E-mail: ahmed.f.gad@gmail.com
• LinkedIn
• Amazon Author Page
• Heartbeat
• Paperspace
• KDnuggets
• TowardsDataScience
• GitHub