Compression of the same data, represented two ways, using BZ2 (similar results with LZMA)

test_compression.py

from __future__ import division
import struct
import bz2

# Generate data - round to nearest 1e11 so that no data is lost when formatting in ASCII
xx = range(10**7)
x = [round(ii*1e11)/(1e11*len(xx)) for ii in xx]
sx = struct.pack('d'*len(x), *x)

# Write it to file as a text and binary file
with open('test_ascii.tdata', 'w') as ft:           # Full size, 10**7 entries: 152 MB
    for val in x:
        ft.write('{:01.12f}\n'.format(val))

with open('test_binary.bdata', 'wb') as fb:         # Full size, 10**7 entries: 76.2 MB
    fb.write(sx)


# Compress both files using bz2 compression
buff_size = 4096
with open('test_ascii.tdata', 'r') as ft:
    with bz2.BZ2File('test_ascii.bz2', 'w', buff_size) as bz2f:
        for line in ft:
            bz2f.write(line)
        bz2f.close()                                # Compressed size: 11.3 MB

with open('test_binary.bdata', 'rb') as fb:
    with bz2.BZ2File('test_binary.bz2', 'w', buff_size) as bz2f:
        bd = fb.read()
        bz2f.write(bd)
        bz2f.close()                                # Compressed size: 39.5 MB