mirror of
https://github.com/TheAlgorithms/Python.git
synced 2024-12-25 04:30:15 +00:00
126 lines
3.5 KiB
Python
126 lines
3.5 KiB
Python
|
"""
|
|||
|
One of the several implementations of Lempel–Ziv–Welch compression algorithm
|
|||
|
https://en.wikipedia.org/wiki/Lempel%E2%80%93Ziv%E2%80%93Welch
|
|||
|
"""
|
|||
|
|
|||
|
import math
|
|||
|
import os
|
|||
|
import sys
|
|||
|
|
|||
|
|
|||
|
def read_file_binary(file_path: str) -> str:
|
|||
|
"""
|
|||
|
Reads given file as bytes and returns them as a long string
|
|||
|
"""
|
|||
|
result = ""
|
|||
|
try:
|
|||
|
with open(file_path, "rb") as binary_file:
|
|||
|
data = binary_file.read()
|
|||
|
for dat in data:
|
|||
|
curr_byte = "{0:08b}".format(dat)
|
|||
|
result += curr_byte
|
|||
|
return result
|
|||
|
except IOError:
|
|||
|
print("File not accessible")
|
|||
|
sys.exit()
|
|||
|
|
|||
|
|
|||
|
def add_key_to_lexicon(
|
|||
|
lexicon: dict, curr_string: str, index: int, last_match_id: int
|
|||
|
) -> None:
|
|||
|
"""
|
|||
|
Adds new strings (curr_string + "0", curr_string + "1") to the lexicon
|
|||
|
"""
|
|||
|
lexicon.pop(curr_string)
|
|||
|
lexicon[curr_string + "0"] = last_match_id
|
|||
|
|
|||
|
if math.log2(index).is_integer():
|
|||
|
for curr_key in lexicon:
|
|||
|
lexicon[curr_key] = "0" + lexicon[curr_key]
|
|||
|
|
|||
|
lexicon[curr_string + "1"] = bin(index)[2:]
|
|||
|
|
|||
|
|
|||
|
def compress_data(data_bits: str) -> str:
|
|||
|
"""
|
|||
|
Compresses given data_bits using Lempel–Ziv–Welch compression algorithm
|
|||
|
and returns the result as a string
|
|||
|
"""
|
|||
|
lexicon = {"0": "0", "1": "1"}
|
|||
|
result, curr_string = "", ""
|
|||
|
index = len(lexicon)
|
|||
|
|
|||
|
for i in range(len(data_bits)):
|
|||
|
curr_string += data_bits[i]
|
|||
|
if curr_string not in lexicon:
|
|||
|
continue
|
|||
|
|
|||
|
last_match_id = lexicon[curr_string]
|
|||
|
result += last_match_id
|
|||
|
add_key_to_lexicon(lexicon, curr_string, index, last_match_id)
|
|||
|
index += 1
|
|||
|
curr_string = ""
|
|||
|
|
|||
|
while curr_string != "" and curr_string not in lexicon:
|
|||
|
curr_string += "0"
|
|||
|
|
|||
|
if curr_string != "":
|
|||
|
last_match_id = lexicon[curr_string]
|
|||
|
result += last_match_id
|
|||
|
|
|||
|
return result
|
|||
|
|
|||
|
|
|||
|
def add_file_length(source_path: str, compressed: str) -> str:
|
|||
|
"""
|
|||
|
Adds given file's length in front (using Elias gamma coding) of the compressed
|
|||
|
string
|
|||
|
"""
|
|||
|
file_length = os.path.getsize(source_path)
|
|||
|
file_length_binary = bin(file_length)[2:]
|
|||
|
length_length = len(file_length_binary)
|
|||
|
|
|||
|
return "0" * (length_length - 1) + file_length_binary + compressed
|
|||
|
|
|||
|
|
|||
|
def write_file_binary(file_path: str, to_write: str) -> None:
|
|||
|
"""
|
|||
|
Writes given to_write string (should only consist of 0's and 1's) as bytes in the
|
|||
|
file
|
|||
|
"""
|
|||
|
byte_length = 8
|
|||
|
try:
|
|||
|
with open(file_path, "wb") as opened_file:
|
|||
|
result_byte_array = [
|
|||
|
to_write[i : i + byte_length]
|
|||
|
for i in range(0, len(to_write), byte_length)
|
|||
|
]
|
|||
|
|
|||
|
if len(result_byte_array[-1]) % byte_length == 0:
|
|||
|
result_byte_array.append("10000000")
|
|||
|
else:
|
|||
|
result_byte_array[-1] += "1" + "0" * (
|
|||
|
byte_length - len(result_byte_array[-1]) - 1
|
|||
|
)
|
|||
|
|
|||
|
for elem in result_byte_array:
|
|||
|
opened_file.write(int(elem, 2).to_bytes(1, byteorder="big"))
|
|||
|
except IOError:
|
|||
|
print("File not accessible")
|
|||
|
sys.exit()
|
|||
|
|
|||
|
|
|||
|
def compress(source_path, destination_path: str) -> None:
|
|||
|
"""
|
|||
|
Reads source file, compresses it and writes the compressed result in destination
|
|||
|
file
|
|||
|
"""
|
|||
|
data_bits = read_file_binary(source_path)
|
|||
|
compressed = compress_data(data_bits)
|
|||
|
compressed = add_file_length(source_path, compressed)
|
|||
|
write_file_binary(destination_path, compressed)
|
|||
|
|
|||
|
|
|||
|
if __name__ == "__main__":
|
|||
|
compress(sys.argv[1], sys.argv[2])
|