abydos.compression module

abydos.compression.

The compression module defines compression and compression-related functions for use within Abydos, including implementations of the following:

  • arithmetic coding functions (ac_train, ac_encode, & ac_decode)
  • Burrows-Wheeler transform encoder/decoder (bwt_encode & bwt_decode)
  • Run-Length Encoding encoder/decoder (rle_encode & rle_decode)
abydos.compression.ac_decode(longval, nbits, probs)[source]

Decode the number to a string using the given statistics.

This is based on Andrew Dalke’s public domain implementation [Dal05]. It has been ported to use the fractions.Fraction class.

Parameters:
  • longval (int) – The first part of an encoded tuple from ac_encode
  • nbits (int) – The second part of an encoded tuple from ac_encode
  • probs (dict) – A probability statistics dictionary generated by ac_train
Returns:

The arithmetically decoded text
Return type:

str

 
>>> pr = ac_train('the quick brown fox jumped over the lazy dog')
>>> ac_decode(16720586181, 34, pr)
'align'
abydos.compression.ac_encode(text, probs)[source]

Encode a text using arithmetic coding with the provided probabilities.

Text and the 0-order probability statistics -> longval, nbits

The encoded number is Fraction(longval, 2**nbits)

This is based on Andrew Dalke’s public domain implementation [Dal05]. It has been ported to use the fractions.Fraction class.

Parameters:
  • text (str) – A string to encode
  • probs (dict) – A probability statistics dictionary generated by ac_train
Returns:

The arithmetically coded text
Return type:

tuple

 
>>> pr = ac_train('the quick brown fox jumped over the lazy dog')
>>> ac_encode('align', pr)
(16720586181, 34)
abydos.compression.ac_train(text)[source]

Generate a probability dict from the provided text.

Text -> 0-order probability statistics as a dict

This is based on Andrew Dalke’s public domain implementation [Dal05]. It has been ported to use the fractions.Fraction class.

Parameters:text (str) – The text data over which to calculate probability statistics. This must not contain the NUL (0x00) character because that’s used to indicate the end of data.
Returns:a probability dict
Return type:dict 
>>> ac_train('the quick brown fox jumped over the lazy dog')
{' ': (Fraction(0, 1), Fraction(8, 45)),
 'o': (Fraction(8, 45), Fraction(4, 15)),
 'e': (Fraction(4, 15), Fraction(16, 45)),
 'u': (Fraction(16, 45), Fraction(2, 5)),
 't': (Fraction(2, 5), Fraction(4, 9)),
 'r': (Fraction(4, 9), Fraction(22, 45)),
 'h': (Fraction(22, 45), Fraction(8, 15)),
 'd': (Fraction(8, 15), Fraction(26, 45)),
 'z': (Fraction(26, 45), Fraction(3, 5)),
 'y': (Fraction(3, 5), Fraction(28, 45)),
 'x': (Fraction(28, 45), Fraction(29, 45)),
 'w': (Fraction(29, 45), Fraction(2, 3)),
 'v': (Fraction(2, 3), Fraction(31, 45)),
 'q': (Fraction(31, 45), Fraction(32, 45)),
 'p': (Fraction(32, 45), Fraction(11, 15)),
 'n': (Fraction(11, 15), Fraction(34, 45)),
 'm': (Fraction(34, 45), Fraction(7, 9)),
 'l': (Fraction(7, 9), Fraction(4, 5)),
 'k': (Fraction(4, 5), Fraction(37, 45)),
 'j': (Fraction(37, 45), Fraction(38, 45)),
 'i': (Fraction(38, 45), Fraction(13, 15)),
 'g': (Fraction(13, 15), Fraction(8, 9)),
 'f': (Fraction(8, 9), Fraction(41, 45)),
 'c': (Fraction(41, 45), Fraction(14, 15)),
 'b': (Fraction(14, 15), Fraction(43, 45)),
 'a': (Fraction(43, 45), Fraction(44, 45)),
 '\x00': (Fraction(44, 45), Fraction(1, 1))}
abydos.compression.bwt_decode(code, terminator='\x00')[source]

Return a word decoded from BWT form.

The Burrows-Wheeler transform is an attempt at placing similar characters together to improve compression. This function reverses the transform. Cf. [BW94].

Parameters:
  • code (str) – the word to transform from BWT form
  • terminator (str) – a character added to word to signal the end of the string
Returns:

word decoded by BWT
Return type:

str

 
>>> bwt_decode('n\x00ilag')
'align'
>>> bwt_decode('annb\x00aa')
'banana'
>>> bwt_decode('annb@aa', '@')
'banana'
abydos.compression.bwt_encode(word, terminator='\x00')[source]

Return the Burrows-Wheeler transformed form of a word.

The Burrows-Wheeler transform is an attempt at placing similar characters together to improve compression. Cf. [BW94].

Parameters:
  • word (str) – the word to transform using BWT
  • terminator (str) – a character to add to word to signal the end of the string
Returns:

word encoded by BWT
Return type:

str

 
>>> bwt_encode('align')
'n\x00ilag'
>>> bwt_encode('banana')
'annb\x00aa'
>>> bwt_encode('banana', '@')
'annb@aa'
abydos.compression.rle_decode(text, use_bwt=True)[source]

Perform decoding of run-length-encoding (RLE).

Cf. [RC67].

Based on http://rosettacode.org/wiki/Run-length_encoding#Python [Cod18b]. This is licensed GFDL 1.2.

Digits 0-9 cannot have been in the original text.

Parameters:
  • text (str) – a text string to decode
  • use_bwt (bool) – boolean indicating whether to perform BWT decoding after RLE decoding
Returns:

word decoded by BWT
Return type:

str

 
>>> rle_decode('n\x00ilag')
'align'
>>> rle_decode('align', use_bwt=False)
'align'

>>> rle_decode('annb\x00aa')
'banana'
>>> rle_decode('banana', use_bwt=False)
'banana'

>>> rle_decode('ab\x00abbab5a')
'aaabaabababa'
>>> rle_decode('3abaabababa', False)
'aaabaabababa'
abydos.compression.rle_encode(text, use_bwt=True)[source]

Perform encoding of run-length-encoding (RLE).

Cf. [RC67].

Based on http://rosettacode.org/wiki/Run-length_encoding#Python [Cod18b]. This is licensed GFDL 1.2.

Digits 0-9 cannot be in text.

Parameters:
  • text (str) – a text string to encode
  • use_bwt (bool) – boolean indicating whether to perform BWT encoding before RLE encoding
Returns:

word decoded by BWT
Return type:

str

 
>>> rle_encode('align')
'n\x00ilag'
>>> rle_encode('align', use_bwt=False)
'align'

>>> rle_encode('banana')
'annb\x00aa'
>>> rle_encode('banana', use_bwt=False)
'banana'

>>> rle_encode('aaabaabababa')
'ab\x00abbab5a'
>>> rle_encode('aaabaabababa', False)
'3abaabababa'