5 from collect.bidict import bidict
7 special_characters = bidict(
18 def compress(uncompressed: str) -> bytes:
19 """Compress a word sequence into a stream of bytes. The compressed
20 form will be 5/8th the size of the original. Words can be lower
21 case letters or special_characters (above).
24 >>> binascii.hexlify(compress('this is a test'))
27 >>> binascii.hexlify(compress('scot'))
30 >>> binascii.hexlify(compress('scott')) # Note the last byte
34 compressed = bitstring.BitArray()
35 for (n, letter) in enumerate(uncompressed):
36 if 'a' <= letter <= 'z':
37 bits = ord(letter) - ord('a') + 1 # 1..26
39 if letter not in special_characters:
40 raise Exception(f'"{uncompressed}" contains uncompressable char="{letter}"')
41 bits = special_characters[letter]
42 compressed.append(f"uint:5={bits}")
43 while len(compressed) % 8 != 0:
44 compressed.append("uint:1=0")
45 return compressed.bytes
48 def decompress(kompressed: bytes) -> str:
50 Decompress a previously compressed stream of bytes back into
54 >>> decompress(binascii.unhexlify(b'a2133da67b0ee859d0'))
57 >>> decompress(binascii.unhexlify(b'98df4a00'))
62 compressed = bitstring.BitArray(kompressed)
64 # There are compressed messages that legitimately end with the
65 # byte 0x00. The message "scott" is an example; compressed it is
66 # 0x98df4a00. It's 5 characters long which means there are 5 x 5
67 # bits of compressed info (25 bits, just over 3 bytes). The last
68 # (25th) bit in the steam happens to be a zero. The compress code
69 # padded out the compressed message by adding seven more zeros to
70 # complete the partial 4th byte. In the 4th byte, however, one
71 # bit is information and seven are padding.
73 # It's likely that this APIs client code will treat a zero byte as
74 # a termination character and not regard it as part of the
75 # message. This is a bug in the client code.
77 # However, it's a bug we can work around:
79 # Here, I'm appending an extra 0x00 byte to the compressed message
80 # passed in. If the client code dropped the last 0x00 byte (and,
81 # with it, some of the legitimate message bits) by treating it as
82 # a termination mark, this 0x00 will replace it (and the missing
83 # message bits). If the client code didn't drop the last 0x00 (or
84 # if the compressed message didn't end in 0x00), adding an extra
85 # 0x00 is a no op because the codepoint 0b00000 is a "stop" message
86 # so we'll ignore the extras.
87 compressed.append("uint:8=0")
89 for chunk in compressed.cut(5):
93 elif 1 <= chunk <= 26:
94 letter = chr(chunk - 1 + ord('a'))
96 letter = special_characters.inverse[chunk][0]
97 decompressed += letter
101 if __name__ == '__main__':