I use Atom and the flake8 plugin, and I was getting a barrage of flake8 UnicodeDecode errors. I couldn't figure out why, so I wrote this script to recursively scan all of the files in a directory and show me which ones had unexpected non-ascii characters in them based on a whitelist. This helped me track down the file causing the problem.

recursively_find_bad_characters.py

import os
import re

skip_paths = (
    '.git',
    '.DS_Store',
    '.bz2',
)
bad_chars_re = re.compile(
    r'[^a-zA-Z0-9_\(\),\.\s"%\':;=@\[\]`\{\}\-!#\*\+/<>\?\$&\\\^\|~]'
)


def recursively_find_bad_characters(entrypath):
    allbaduns = set()
    for (path, dirs, files) in os.walk(entrypath):
        for filename in files:
            filepath = os.path.join(path, filename)
            skipmatches = [skip for skip in skip_paths if skip in filepath]
            if skipmatches:
                continue
            s = open(filepath).read()
            baduns = set(bad_chars_re.findall(s))
            allbaduns.update(baduns)
            if baduns:
                print filepath, baduns
                for badun in baduns:
                    print badun, repr(badun)
                    index = s.index(badun)
                    print index, s[max(0, index - 10):index + 10]
                    print "on line %d" % (s[:index].count('\n') + 1)
    if allbaduns:
        print 'allbaduns = %r' % allbaduns

if __name__ == '__main__':
    import sys
    for arg in sys.argv[1:]:
        recursively_find_bad_characters(arg)

The problem turned out to be in the .flake8 configuration file itself, which had a rogue curly apostrophe in a comment.

Usage:

python recursively_find_bad_characters.py ~/path/to/my/project