Decode MySQL filenames to unicode

filename_to_tablename.py

import binascii
import collections
import re
import struct

CodeRange = collections.namedtuple('CodeRange', 'start end')

class EncodingRule(object):
    def __init__(self, coderange, pattern, name, initial_code):
        self.coderange = coderange
        self.pattern = pattern
        self.name = name
        self.initial_code = initial_code

    def __call__(self, matchobj):
        value = matchobj.group()
        ivalue, = struct.unpack_from('>H', value, 1)
        initial_code, = struct.unpack_from('>H', self.initial_code, 1)
        uni_ord = self.coderange.start + (ivalue - initial_code)
        return unichr(uni_ord)

class Ucs2EncodingRule(object):
    def __init__(self):
        self.pattern = r'@[0-9a-f]{4}'

    def __call__(self, matchobj):
        value = matchobj.group()
        uni_ord, = struct.unpack('>H', binascii.unhexlify(value[1:]))
        return unichr(uni_ord)

class WindowsDeviceNameEncodingRule(object):
    def __init__(self):
        self.pattern = r'[@]{3}'

    def __call__(self, matchobj):
        return ''

encoding_rules = [
    EncodingRule(coderange=CodeRange(0x00C0, 0x017F),
         pattern=r'[@][0-4][g-zG-Z]',
         name='Latin-1 Supplement + Latin Extended-A',
         initial_code='@0G'),

    Ucs2EncodingRule(),

    WindowsDeviceNameEncodingRule(),

    EncodingRule(coderange=CodeRange(0x0370, 0x03FF),
         pattern=r'[@][5-9][g-zG-Z]',
         name='Greek and Coptic',
         initial_code='@5G'),

    EncodingRule(coderange=CodeRange(0x0400, 0x052F),
         pattern=r'[@][g-zG-Z][0-6]',
         name='Cyrillic + Cyrillic Supplement',
         initial_code='@G0'),

    EncodingRule(coderange=CodeRange(0x0530, 0x058F),
         pattern=r'[@][G-Zg-z][7-8]',
         name='Armenian',
         initial_code='@G7'),

    EncodingRule(coderange=CodeRange(0x2160, 0x217F),
         pattern=r'[@][G-Zg-z][9]',
         name='Number Forms',
         initial_code='@G9'),

    EncodingRule(coderange=CodeRange(0x0180, 0x02AF),
         pattern=r'[@][G-Zg-z][A-Ka-k]',
         name='Latin Extended-B + IPA Extensions',
         initial_code='@GA'),

    EncodingRule(coderange=CodeRange(0x1E00, 0x1EFF),
         pattern=r'[@][G-Zg-z][L-Rl-r]]',
         name='Latin Extended Additional',
         initial_code='@GL'),

    EncodingRule(coderange=CodeRange(0x1F00, 0x1FFF),
         pattern=r'[@][G-Zg-z][S-Zs-z]',
         name='Greek Extended',
         initial_code='@GS'),

    EncodingRule(coderange=CodeRange(0x24B6, 0x24E9),
         pattern=r'[@][@][A-Za-z]',
         name='Enclosed Alphanumerics',
         initial_code='@@A'),

    EncodingRule(coderange=CodeRange(0xFF21, 0xFF5A),
         pattern=r'[@][A-Za-z][@]',
         name=r'Halfwidth and Fullwidth forms',
         initial_code='@A@'),
]


def decode_filename(path):
    for rule in encoding_rules:
        match = re.search(rule.pattern, path)
        path = re.sub(rule.pattern, rule, path)
    return path

if __name__ == '__main__':
    import sys
    for name in sys.argv[1:]:
        print name, ":", decode_filename(name)