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