Broken gzip, whereas bzip2 is still stable.

broken_gzip.py

'''
When reading about gzip and bzip2, I found an interesting data pattern 
that terribly breaks gzip where the compressed file is larged than the 
uncompressed file. The reason is that the input data is incompressible 
to gzip, hence gzip has to store the data as uncompressed, adding file 
header and block headers, resulting in extra bytes. On the other hand, 
bzip is still quite stable and able to achieve a good compression ratio.

The format of data from the below script is as follows:
step 1, start 0: 0 1 2 3 ... 255 
step 2, start 0: 0 2 4 ... 254
step 2, start 1: 1 3 5 7 ... 255
...

Looking at the data, there is no 3 charater set appears more than once 
which means that gzip (which uses DEFLATE under the hood for compressed 
blocks) cannot compress anything. To make it worse, I can input reversed 
data since the reversed byte sequence is also incompressible. 

The below script generates 1343488 bytes. gzip -9 outputs a "compressed" 
file of size 1343921. Whereas bzip2 -9 outputs a compressed file of size 186546.
'''
data = []
f = open('broken.bin', 'wb')
for step in range(1, 129):
    for start in range(0, step):
        for b in range(start, 256, step):
            data.append(b)
b_data = bytes(data)
b_r_data = bytes(reversed(data))
for i in range(20):
    f.write(b_data)
    f.write(b_r_data)
f.write(b_data)
f.close()