Skip to content

Instantly share code, notes, and snippets.

@lawlesst
Created November 19, 2012 14:29
Show Gist options
  • Star 12 You must be signed in to star a gist
  • Fork 5 You must be signed in to fork a gist
  • Save lawlesst/4110923 to your computer and use it in GitHub Desktop.
Save lawlesst/4110923 to your computer and use it in GitHub Desktop.
cleaning invalid characters from xml
def invalid_xml_remove(c):
#http://stackoverflow.com/questions/1707890/fast-way-to-filter-illegal-xml-unicode-chars-in-python
illegal_unichrs = [ (0x00, 0x08), (0x0B, 0x1F), (0x7F, 0x84), (0x86, 0x9F),
(0xD800, 0xDFFF), (0xFDD0, 0xFDDF), (0xFFFE, 0xFFFF),
(0x1FFFE, 0x1FFFF), (0x2FFFE, 0x2FFFF), (0x3FFFE, 0x3FFFF),
(0x4FFFE, 0x4FFFF), (0x5FFFE, 0x5FFFF), (0x6FFFE, 0x6FFFF),
(0x7FFFE, 0x7FFFF), (0x8FFFE, 0x8FFFF), (0x9FFFE, 0x9FFFF),
(0xAFFFE, 0xAFFFF), (0xBFFFE, 0xBFFFF), (0xCFFFE, 0xCFFFF),
(0xDFFFE, 0xDFFFF), (0xEFFFE, 0xEFFFF), (0xFFFFE, 0xFFFFF),
(0x10FFFE, 0x10FFFF) ]
illegal_ranges = ["%s-%s" % (unichr(low), unichr(high))
for (low, high) in illegal_unichrs
if low < sys.maxunicode]
illegal_xml_re = re.compile(u'[%s]' % u''.join(illegal_ranges))
if illegal_xml_re.search(c) is not None:
#Replace with space
return ' '
else:
return c
def scrub_literal(value):
"""
Scrubs control characters from the incoming values to remove
things like form feeds (\f) and line breaks (\n) which might
cause problems with Jena.
Data with these characters was found in the Backstage data.
"""
from curses import ascii
import unicodedata
if not value:
return
if (type(value) == long) or (type(value) == int):
return value
n = ''.join([c for c in value if not ascii.iscntrl(c)\
if not ascii.isctrl(c)])
#n = ''.join(new)
n = n.replace('"', '')
n = n.replace('\ufffd', '')
n = clean_text(n)
if type(n) != unicode:
n = unicode(n, errors='replace')
return n.strip()
def clean_char(char):
"""
Function for remove invalid XML characters from
incoming data.
"""
#Get rid of the ctrl characters first.
#http://stackoverflow.com/questions/1833873/python-regex-escape-characters
char = re.sub('\x1b[^m]*m', '', char)
#Clean up invalid xml
char = invalid_xml_remove(char)
replacements = [
(u'\u201c', '\"'),
(u'\u201d', '\"'),
(u"\u001B", ' '), #http://www.fileformat.info/info/unicode/char/1b/index.htm
(u"\u0019", ' '), #http://www.fileformat.info/info/unicode/char/19/index.htm
(u"\u0016", ' '), #http://www.fileformat.info/info/unicode/char/16/index.htm
(u"\u001C", ' '), #http://www.fileformat.info/info/unicode/char/1c/index.htm
(u"\u0003", ' '), #http://www.utf8-chartable.de/unicode-utf8-table.pl?utf8=0x
(u"\u000C", ' ')
]
for rep, new_char in replacements:
if char == rep:
#print ord(char), char.encode('ascii', 'ignore')
return new_char
return char
@AlmightyOatmeal
Copy link

Wouldn't it be easier to do something like:

''.join([c for c in xml_str if ord(c) < 127])

Or

re.sub('[^\x00-\x7F]', '', xml_str)

But I usually do:

data_str = data_str.encode('utf-8', 'xmlcharrefreplace')

Keep in mind that if you use lxml then using lxml.etree.tostring() with an encoding set to 'ascii' will force XML/XHTML/HTML-safe escapes, such as giving me &#10003; in place of \u2713 (aka: 0xe2).

This is what I typically use:

def sanitize_text(data):
    logger.debug('type(data): {}'.format(type(data)))
    replace_with = {
        u'\u2018': '\'',
        u'\u2019': '\'',
        u'\u201c': '"',
        u'\u201d': '"'
    }

    bad_chars = [c for c in data if ord(c) >= 127]
    if bad_chars:
        logger.warning('INVALID CHARACTERS: {}'.format(bad_chars))
    else:
        logger.debug('INVALID CHARACTERS: {}'.format(bad_chars))

    for uni_char in replace_with.keys():
        data = data.replace(uni_char, replace_with.get(uni_char))

    data = ''.join([c for c in data if ord(c) < 127])
    return data.encode('utf-8', 'xmlcharreplace')

Of course you can use other error handling in encode() to fit your needs if you're not using XML.

NOTE: When you do (u'\u201d', '\"'),, you don't need to escape the double-quote because it's already in a single quote.

@Jun711
Copy link

Jun711 commented Sep 24, 2018

Does this work?
data_str = data_str.encode('utf-8', 'xmlcharrefreplace')

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment