Bubblegum compiler

bubblegum.py

#!/usr/bin/env python3

# Script for turning ascii input into valid Bubblegum programs
# https://esolangs.org/wiki/Bubblegum

import hashlib, lzma, sys, zlib

def bb96encode(code):
    buf = []
    a = 0
    for byte in code:
        digit = 96 if byte == 10 else byte - 31
        assert 1 <= digit <= 96
        a = 96 * a + digit
    while a:
        r = a % 256
        buf.append(r)
        a //= 256
    return bytes(reversed(buf))

def deflate(data):
    compress = zlib.compressobj(9, wbits=-zlib.MAX_WBITS)
    deflated = compress.compress(data)
    deflated += compress.flush()
    return deflated

def do_lzma(data):
    return lzma.compress(
        code,
        format=lzma.FORMAT_RAW,
        filters=[{'id': lzma.FILTER_LZMA2, 'preset': 9 | lzma.PRESET_EXTREME}])

def dump(src, length=16):
    result=[]
    for i in range(0, len(src), length):
       s = src[i:i+length]
       hexa = ' '.join(["%02x" % x for x in s])
       result.append("%08x: %-*s  \n" % (i, length*3, hexa))
    return ''.join(result)

# TODO while unlikely to arise accidentally, to be complete we should ensure
# that the result of e.g. bbase96 isn't a valid deflate or lzma2 stream; if it
# were, we would need to use a less-efficient encoding to ensure the correct
# ouput.
codecs = {'DEFLATE':deflate, 'LZMA2':do_lzma, 'BBASE96':bb96encode}

code = sys.stdin.buffer.read()

best = None

print("Input length:", len(code), "bytes", file=sys.stderr)
for name, codec in codecs.items():
    opt = codec(code)
    print("%s:" % name, len(opt), file=sys.stderr)
    if best is None or len(opt) < len(best):
        best = opt

print(dump(best), end='')