trac/util/text.py

# -*- coding: utf-8 -*-
#
# Copyright (C) 2003-2021 Edgewall Software
# Copyright (C) 2003-2004 Jonas Borgström <jonas@edgewall.com>
# Copyright (C) 2006 Matthew Good <trac@matt-good.net>
# Copyright (C) 2005-2006 Christian Boos <cboos@edgewall.org>
# All rights reserved.
#
# This software is licensed as described in the file COPYING, which
# you should have received as part of this distribution. The terms
# are also available at https://trac.edgewall.org/wiki/TracLicense.
#
# This software consists of voluntary contributions made by many
# individuals. For the exact contribution history, see the revision
# history and logs, available at https://trac.edgewall.org/log/.
#
# Author: Jonas Borgström <jonas@edgewall.com>
#         Matthew Good <trac@matt-good.net>
#         Christian Boos <cboos@edgewall.org>

import base64
import configparser
import locale
import os
import pkg_resources
import re
import sys
import textwrap
from urllib.parse import quote, quote_plus, unquote
from unicodedata import east_asian_width

import jinja2

CRLF = '\r\n'

class Empty(str):
    """A special tag object evaluating to the empty string"""
    __slots__ = []

empty = Empty()

del Empty # shouldn't be used outside of Trac core


# -- Jinja2

_jinja2_ver = pkg_resources.parse_version(jinja2.__version__)
_jinja2_exts = ['jinja2.ext.do', 'jinja2.ext.i18n']
if _jinja2_ver < pkg_resources.parse_version('3'):
    _jinja2_exts.append('jinja2.ext.with_')

def jinja2env(**kwargs):
    """Creates a Jinja2 ``Environment`` configured with Trac conventions.

    All default parameters can optionally be overriden. The ``loader``
    parameter is not set by default, so unless it is set by the
    caller, only inline templates can be created from the environment.

    :rtype: `jinja.Environment`

    """
    exts = ('html', 'rss', 'xml')
    def filterout_none(v):
        return '' if v is None else v
    def autoescape_extensions(template):
        return template and template.rsplit('.', 1)[1] in exts
    defaults = dict(
        variable_start_string='${',
        variable_end_string='}',
        line_statement_prefix='#',
        line_comment_prefix='##',
        trim_blocks=True,
        lstrip_blocks=True,
        extensions=list(_jinja2_exts),
        finalize=filterout_none,
        autoescape=autoescape_extensions,
    )
    defaults.update(kwargs)
    jenv = jinja2.Environment(**defaults)
    jenv.globals.update(
        len=len,
    )
    return jenv

def jinja2template(template, text=False, **kwargs):
    """Creates a Jinja2 ``Template`` from inlined source.

    :param template: the template content
    :param text: if set to `False`, the result of the variable
                 expansion will be XML/HTML escaped
    :param kwargs: additional arguments to pass to `jinja2env`. See
                   `jinja2.Environment` for supported arguments.
    """
    return jinja2env(autoescape=not text, **kwargs).from_string(template)


# -- Unicode

def to_unicode(text, charset=None):
    """Convert input to a `str` object.

    For a `bytes` object, we'll first try to decode the bytes using the given
    `charset` encoding (or UTF-8 if none is specified), then we fall back to
    the latin1 encoding which might be correct or not, but at least preserves
    the original byte sequence by mapping each byte to the corresponding
    unicode code point in the range U+0000 to U+00FF.

    For anything else, a simple `str()` conversion is attempted,
    with special care taken with `Exception` objects.
    """
    if isinstance(text, bytes):
        try:
            return str(text, charset or 'utf-8')
        except UnicodeDecodeError:
            return str(text, 'latin1')
    if isinstance(text, Exception):
        # two possibilities for storing unicode strings in exception data:
        try:
            # custom __str__ method on the exception (e.g. PermissionError)
            result = str(text)
        except UnicodeError:
            # unicode arguments given to the exception (e.g. parse_date)
            return ' '.join(to_unicode(arg) for arg in text.args)
        if os.name == 'nt':
            # remove duplicated backslashes from filename in the message
            if isinstance(text, EnvironmentError) and text.filename:
                source = repr(text.filename)
            elif isinstance(text, configparser.ParsingError) and text.source:
                source = repr(text.source)
            else:
                source = None
            if source:
                result = result.replace(source, source.replace(r'\\', '\\'))
        return result
    return str(text)


def exception_to_unicode(e, traceback=False):
    """Convert an `Exception` to a `str` object.

    In addition to `to_unicode`, this representation of the exception
    also contains the class name and optionally the traceback.
    """
    message = '%s: %s' % (e.__class__.__name__, to_unicode(e))
    if traceback:
        from trac.util import get_last_traceback
        traceback_only = get_last_traceback().split('\n')[:-2]
        message = '\n%s\n%s' % (to_unicode('\n'.join(traceback_only)), message)
    return message


def path_to_unicode(path):
    """Convert a filesystem path to str, using the filesystem encoding."""
    if isinstance(path, bytes):
        try:
            return str(path, sys.getfilesystemencoding())
        except UnicodeDecodeError:
            return str(path, 'latin1')
    return str(path)


_ws_leading_re = re.compile('\\A[\\s\u200b]+', re.UNICODE)
_ws_trailing_re = re.compile('[\\s\u200b]+\\Z', re.UNICODE)

def stripws(text, leading=True, trailing=True):
    """Strips unicode white-spaces and ZWSPs from ``text``.

    :param leading: strips leading spaces from ``text`` unless ``leading`` is
                    `False`.
    :param trailing: strips trailing spaces from ``text`` unless ``trailing``
                     is `False`.
    """
    if leading:
        text = _ws_leading_re.sub('', text)
    if trailing:
        text = _ws_trailing_re.sub('', text)
    return text


def strip_line_ws(text, leading=True, trailing=True):
    """Strips unicode white-spaces and ZWSPs from each line of ``text``.

    :param leading: strips leading spaces from ``text`` unless ``leading`` is
                    `False`.
    :param trailing: strips trailing spaces from ``text`` unless ``trailing``
                     is `False`.
    """
    lines = re.compile(r'(\n|\r\n|\r)').split(text)
    if leading:
        lines[::2] = (_ws_leading_re.sub('', line) for line in lines[::2])
    if trailing:
        lines[::2] = (_ws_trailing_re.sub('', line) for line in lines[::2])
    return ''.join(lines)


_js_quote = {'\\': '\\\\', '"': '\\"', '\b': '\\b', '\f': '\\f',
             '\n': '\\n', '\r': '\\r', '\t': '\\t', "'": "\\'"}
for i in list(range(0x20)) + [ord(c) for c in '&<>\u2028\u2029']:
    _js_quote.setdefault(chr(i), '\\u%04x' % i)
_js_quote_re = re.compile(r'[\x00-\x1f\\"\b\f\n\r\t\'&<>' + '\u2028\u2029]')
_js_string_re = re.compile(r'[\x00-\x1f\\"\b\f\n\r\t&<>' + '\u2028\u2029]')


def javascript_quote(text):
    """Quote strings for inclusion in single or double quote delimited
    Javascript strings
    """
    if not text:
        return ''
    def replace(match):
        return _js_quote[match.group(0)]
    return _js_quote_re.sub(replace, text)


def to_js_string(text):
    """Embed the given string in a double quote delimited Javascript string
    (conform to the JSON spec)
    """
    if not text:
        return '""'
    def replace(match):
        return _js_quote[match.group(0)]
    return '"%s"' % _js_string_re.sub(replace, text)


def unicode_quote(value, safe='/'):
    """A unicode aware version of `urllib.quote`

    :param value: anything that converts to a `bytes`. If `str`
                  input is given, it will be UTF-8 encoded.
    :param safe: as in `quote`, the characters that would otherwise be
                 quoted but shouldn't here (defaults to '/')
    """
    return quote(value if isinstance(value, bytes) else str(value), safe)


def unicode_quote_plus(value, safe=''):
    """A unicode aware version of `urllib.quote_plus`.

    :param value: anything that converts to a `bytes`. If `str`
                  input is given, it will be UTF-8 encoded.
    :param safe: as in `quote_plus`, the characters that would
                 otherwise be quoted but shouldn't here (defaults to
                 '/')
    """
    return quote_plus(value if isinstance(value, bytes) else str(value), safe)


def unicode_unquote(value):
    """A unicode aware version of `urllib.unquote`.

    :param value: UTF-8 encoded `str` value (for example, as obtained by
                  `unicode_quote`).
    :rtype: `str`
    """
    if isinstance(value, bytes):
        value = value.decode('latin1')
    return unquote(value, encoding='utf-8', errors='strict')


def unicode_urlencode(params, safe=''):
    """A unicode aware version of `urllib.urlencode`.

    Values set to `empty` are converted to the key alone, without the
    equal sign.
    """
    if isinstance(params, dict):
        params = sorted(params.items(), key=lambda i: i[0])
    l = []
    for k, v in params:
        if v is empty:
            l.append(unicode_quote_plus(k, safe))
        else:
            l.append(unicode_quote_plus(k, safe) + '=' +
                     unicode_quote_plus(v, safe))
    return '&'.join(l)


_qs_quote_safe = ''.join(chr(c) for c in range(0x21, 0x7f))

def quote_query_string(text):
    """Quote strings for query string
    """
    return unicode_quote_plus(text, _qs_quote_safe)


def to_utf8(text, charset='latin1'):
    """Convert input to a UTF-8 `bytes` object.

    If the input is not an `str` object, we assume the encoding is
    already UTF-8, ISO Latin-1, or as specified by the optional
    *charset* parameter.
    """
    if isinstance(text, bytes):
        try:
            u = str(text, 'utf-8')
        except UnicodeError:
            try:
                # Use the user supplied charset if possible
                u = str(text, charset)
            except UnicodeError:
                # This should always work
                u = str(text, 'latin1')
        else:
            # Do nothing if it's already utf-8
            return text
    else:
        u = to_unicode(text)
    return u.encode('utf-8')


class unicode_passwd(str):
    """Conceal the actual content of the string when `repr` is called."""
    def __repr__(self):
        return '*******'


def stream_encoding(stream):
    """Return the appropriate encoding for the given stream."""
    encoding = getattr(stream, 'encoding', None)
    # Windows returns 'cp0' to indicate no encoding
    return encoding if encoding not in (None, 'cp0') else 'utf-8'


def console_print(out, *args, **kwargs):
    """Output the given arguments to the console, encoding the output
    as appropriate.

    :param kwargs: ``newline`` controls whether a newline will be appended
                   (defaults to `True`)
    """
    out.write(' '.join(to_unicode(a) for a in args))
    if kwargs.get('newline', True):
        out.write('\n')


def printout(*args, **kwargs):
    """Do a `console_print` on `sys.stdout`."""
    console_print(sys.stdout, *args, **kwargs)


def printerr(*args, **kwargs):
    """Do a `console_print` on `sys.stderr`."""
    console_print(sys.stderr, *args, **kwargs)


def printfout(message, *args, **kwargs):
    """Format `message`, do a `console.print` on `sys.stdout` and flush
    the buffer.
    """
    if args:
        message %= args
    printout(message, **kwargs)
    sys.stdout.flush()


def printferr(message, *args, **kwargs):
    """Format `message`, do a `console.print` on `sys.stderr` and flush
    the buffer.
    """
    if args:
        message %= args
    printerr(message, **kwargs)
    sys.stderr.flush()


def raw_input(prompt):
    """Input one line from the console and converts it to unicode as
    appropriate.
    """
    printout(prompt, newline=False)
    return to_unicode(input(), sys.stdin.encoding)


_preferredencoding = locale.getpreferredencoding()

def getpreferredencoding():
    """Return the encoding, which is retrieved on ahead, according to user
    preference.

    We should use this instead of `locale.getpreferredencoding()` which
    is not thread-safe."""
    return _preferredencoding


# -- Plain text formatting

def text_width(text, ambiwidth=1):
    """Determine the column width of `text` in Unicode characters.

    The characters in the East Asian Fullwidth (F) or East Asian Wide (W)
    have a column width of 2. The other characters in the East Asian
    Halfwidth (H) or East Asian Narrow (Na) have a column width of 1.

    That `ambiwidth` parameter is used for the column width of the East
    Asian Ambiguous (A). If `1`, the same width as characters in US-ASCII.
    This is expected by most users. If `2`, twice the width of US-ASCII
    characters. This is expected by CJK users.

    cf. http://www.unicode.org/reports/tr11/.
    """
    twice = 'FWA' if ambiwidth == 2 else 'FW'
    return sum([2 if east_asian_width(chr) in twice else 1
                for chr in to_unicode(text)])


def _get_default_ambiwidth():
    """Return width of East Asian Ambiguous based on locale environment
    variables or Windows codepage.
    """

    if os.name == 'nt':
        import ctypes
        codepage = ctypes.windll.kernel32.GetConsoleOutputCP()
        if codepage in (932,   # Japanese (Shift-JIS)
                        936,   # Chinese Simplified (GB2312)
                        949,   # Korean (Unified Hangul Code)
                        950):  # Chinese Traditional (Big5)
            return 2
    else:
        for name in ('LANGUAGE', 'LC_ALL', 'LC_MESSAGES', 'LANG'):
            value = os.environ.get(name) or ''
            if value:
                if name == 'LANGUAGE' and ':' in value:
                    value = value.split(':')[0]
                return 2 if value.lower().startswith(('zh', 'ja', 'ko')) else 1

    return 1


_default_ambiwidth = _get_default_ambiwidth()


def print_table(data, headers=None, sep='  ', out=None, ambiwidth=None):
    """Print data according to a tabular layout.

    :param data: a sequence of rows; assume all rows are of equal length.
    :param headers: an optional row containing column headers; must be of
                    the same length as each row in `data`.
    :param sep: column separator
    :param out: output file descriptor (`None` means use `sys.stdout`)
    :param ambiwidth: column width of the East Asian Ambiguous (A). If None,
                      detect ambiwidth with the locale settings. If others,
                      pass to the `ambiwidth` parameter of `text_width`.
    """
    if out is None:
        out = sys.stdout
    if ambiwidth is None:
        ambiwidth = _default_ambiwidth
    data = list(data)
    if headers:
        data.insert(0, headers)
    elif not data:
        return

    # Convert to a str object with `to_unicode`. If None, convert to a
    # empty string.
    def to_text(val):
        if val is None:
            return ''
        return to_unicode(val)

    def tw(text):
        return text_width(text, ambiwidth=ambiwidth)

    def to_lines(data):
        lines = []
        for row in data:
            row = [to_text(cell) for cell in row]
            if any('\n' in cell for cell in row):
                row = [cell.splitlines() for cell in row]
                max_lines = max(len(cell) for cell in row)
                for cell in row:
                    if len(cell) < max_lines:
                        cell += [''] * (max_lines - len(cell))
                lines.extend([cell[idx] for cell in row]
                             for idx in range(max_lines))
            else:
                lines.append(row)
        return lines

    data = to_lines(data)

    num_cols = len(data[0])
    col_width = [max(tw(row[idx]) for row in data)
                 for idx in range(num_cols)]

    out.write('\n')
    for ridx, row in enumerate(data):
        for cidx, cell in enumerate(row):
            if cidx + 1 == num_cols:
                line = cell  # No separator after last column
            else:
                if headers and ridx == 0:
                    sp = ' ' * tw(sep)  # No separator in header
                else:
                    sp = sep
                line = '%-*s%s' % (col_width[cidx] - tw(cell) + len(cell),
                                    cell, sp)
            out.write(line)

        out.write('\n')
        if ridx == 0 and headers:
            out.write('-' * (tw(sep) * cidx + sum(col_width)))
            out.write('\n')
    out.write('\n')


def shorten_line(text, maxlen=75):
    """Truncates `text` to length less than or equal to `maxlen` characters.

    This tries to be (a bit) clever and attempts to find a proper word
    boundary for doing so.
    """
    if len(text or '') <= maxlen:
        return text
    suffix = ' ...'
    maxtextlen = maxlen - len(suffix)
    cut = max(text.rfind(' ', 0, maxtextlen), text.rfind('\n', 0, maxtextlen))
    if cut < 0:
        cut = maxtextlen
    return text[:cut] + suffix


class UnicodeTextWrapper(textwrap.TextWrapper):
    breakable_char_ranges = [
        (0x1100, 0x11FF),   # Hangul Jamo
        (0x2E80, 0x2EFF),   # CJK Radicals Supplement
        (0x3000, 0x303F),   # CJK Symbols and Punctuation
        (0x3040, 0x309F),   # Hiragana
        (0x30A0, 0x30FF),   # Katakana
        (0x3130, 0x318F),   # Hangul Compatibility Jamo
        (0x3190, 0x319F),   # Kanbun
        (0x31C0, 0x31EF),   # CJK Strokes
        (0x3200, 0x32FF),   # Enclosed CJK Letters and Months
        (0x3300, 0x33FF),   # CJK Compatibility
        (0x3400, 0x4DBF),   # CJK Unified Ideographs Extension A
        (0x4E00, 0x9FFF),   # CJK Unified Ideographs
        (0xA960, 0xA97F),   # Hangul Jamo Extended-A
        (0xAC00, 0xD7AF),   # Hangul Syllables
        (0xD7B0, 0xD7FF),   # Hangul Jamo Extended-B
        (0xF900, 0xFAFF),   # CJK Compatibility Ideographs
        (0xFE30, 0xFE4F),   # CJK Compatibility Forms
        (0xFF00, 0xFFEF),   # Halfwidth and Fullwidth Forms
        (0x20000, 0x2FFFF, '[\uD840-\uD87F][\uDC00-\uDFFF]'), # Plane 2
        (0x30000, 0x3FFFF, '[\uD880-\uD8BF][\uDC00-\uDFFF]'), # Plane 3
    ]

    split_re = None
    breakable_re = None

    @classmethod
    def _init_patterns(cls):
        char_ranges = []
        for val in cls.breakable_char_ranges:
            high = chr(val[0])
            low = chr(val[1])
            char_ranges.append('%s-%s' % (high, low))
        char_ranges = ''.join(char_ranges)
        pattern = '[%s]+' % char_ranges

        cls.split_re = re.compile(
            r'(\s+|' +                                  # any whitespace
            pattern + '|' +                             # breakable text
            r'[^\s\w]*\w+[^0-9\W]-(?=\w+[^0-9\W])|' +   # hyphenated words
            r'(?<=[\w\!\"\'\&\.\,\?])-{2,}(?=\w))',     # em-dash
            re.UNICODE)
        cls.breakable_re = re.compile(r'\A' + pattern, re.UNICODE)

    def __init__(self, cols, replace_whitespace=0, break_long_words=0,
                 initial_indent='', subsequent_indent='', ambiwidth=1):
        textwrap.TextWrapper.__init__(
                self, cols, replace_whitespace=0, break_long_words=0,
                initial_indent=initial_indent,
                subsequent_indent=subsequent_indent)
        self.ambiwidth = ambiwidth
        if self.split_re is None:
            self._init_patterns()

    def _split(self, text):
        chunks = self.split_re.split(to_unicode(text))
        return list(filter(None, chunks))

    def _text_width(self, text):
        return text_width(text, ambiwidth=self.ambiwidth)

    def _wrap_chunks(self, chunks):
        lines = []
        chunks.reverse()
        text_width = self._text_width

        while chunks:
            cur_line = []
            cur_width = 0

            if lines:
                indent = self.subsequent_indent
            else:
                indent = self.initial_indent
            width = self.width - text_width(indent)

            if chunks[-1].strip() == '' and lines:
                del chunks[-1]

            while chunks:
                chunk = chunks[-1]
                w = text_width(chunk)
                if cur_width + w <= width:
                    cur_line.append(chunks.pop())
                    cur_width += w
                elif self.breakable_re.match(chunk):
                    left_space = width - cur_width
                    for i in range(len(chunk)):
                        w = text_width(chunk[i])
                        if left_space < w:
                            break
                        left_space -= w
                    if i > 0:
                        cur_line.append(chunk[:i])
                        chunk = chunk[i:]
                        chunks[-1] = chunk
                    w = text_width(chunk)
                    break
                else:
                    break

            if chunks and w > width:
                self._handle_long_word(chunks, cur_line, cur_width, width)

            if cur_line and cur_line[-1].strip() == '':
                del cur_line[-1]

            if cur_line:
                lines.append(indent + ''.join(cur_line))

        return lines


def wrap(t, cols=75, initial_indent='', subsequent_indent='',
         linesep=os.linesep, ambiwidth=1):
    """Wraps the single paragraph in `t`, which contains unicode characters.
    The every line is at most `cols` characters long.

    That `ambiwidth` parameter is used for the column width of the East
    Asian Ambiguous (A). If `1`, the same width as characters in US-ASCII.
    This is expected by most users. If `2`, twice the width of US-ASCII
    characters. This is expected by CJK users.
    """
    t = t.strip().replace('\r\n', '\n').replace('\r', '\n')
    wrapper = UnicodeTextWrapper(cols, replace_whitespace=0,
                                 break_long_words=0,
                                 initial_indent=initial_indent,
                                 subsequent_indent=subsequent_indent,
                                 ambiwidth=ambiwidth)
    wrappedLines = []
    for line in t.split('\n'):
        wrappedLines += wrapper.wrap(line.rstrip()) or ['']
    return linesep.join(wrappedLines)


_obfuscation_char = '@\u2026'

def obfuscate_email_address(address):
    """Replace anything looking like an e-mail address (``'@something'``)
    with a trailing ellipsis (``'@…'``)
    """
    if address:
        at = address.find('@')
        if at != -1:
            return address[:at] + _obfuscation_char + \
                   ('>' if address[-1] == '>' else '')
    return address


def is_obfuscated(word):
    """Returns `True` if the `word` looks like an obfuscated e-mail
    address.

    :since: 1.2
    """
    return _obfuscation_char in word


def breakable_path(path):
    """Make a path breakable after path separators, and conversely, avoid
    breaking at spaces.
    """
    if not path:
        return path
    prefix = ''
    if path.startswith('/'):    # Avoid breaking after a leading /
        prefix = '/'
        path = path[1:]
    return prefix + path.replace('/', '/\u200b').replace('\\', '\\\u200b') \
                        .replace(' ', '\u00a0')


def normalize_whitespace(text, to_space='\u00a0', remove='\u200b'):
    """Normalize whitespace in a string, by replacing special spaces by normal
    spaces and removing zero-width spaces."""
    if not text:
        return text
    for each in to_space:
        text = text.replace(each, ' ')
    for each in remove:
        text = text.replace(each, '')
    return text


def unquote_label(txt):
    """Remove (one level of) enclosing single or double quotes.

    .. versionadded :: 1.0
    """
    return txt[1:-1] if txt and txt[0] in "'\"" and txt[0] == txt[-1] else txt


def cleandoc(message):
    """Removes uniform indentation and leading/trailing whitespace."""
    from inspect import cleandoc
    return cleandoc(message).strip()


# -- Conversion

def pretty_size(size, format='%.1f'):
    """Pretty print content size information with appropriate unit.

    :param size: number of bytes
    :param format: can be used to adjust the precision shown
    """
    if size is None:
        return ''

    jump = 1024
    if size < jump:
        from trac.util.translation import ngettext
        return ngettext("%(num)d byte", "%(num)d bytes", num=size)

    units = ['KB', 'MB', 'GB', 'TB']
    i = 0
    while size >= jump and i < len(units):
        i += 1
        size /= 1024.

    return (format + ' %s') % (size, units[i - 1])


def expandtabs(s, tabstop=8, ignoring=None):
    """Expand tab characters `'\\\\t'` into spaces.

    :param tabstop: number of space characters per tab
                    (defaults to the canonical 8)

    :param ignoring: if not `None`, the expansion will be "smart" and
                     go from one tabstop to the next. In addition,
                     this parameter lists characters which can be
                     ignored when computing the indent.
    """
    if '\t' not in s:
        return s
    if ignoring is None:
        return s.expandtabs(tabstop)

    outlines = []
    for line in s.split('\n'):
        if '\t' not in line:
            outlines.append(line)
            continue
        p = 0
        s = []
        for c in line:
            if c == '\t':
                n = tabstop - p % tabstop
                s.append(' ' * n)
                p += n
            elif not ignoring or c not in ignoring:
                p += 1
                s.append(c)
            else:
                s.append(c)
        outlines.append(''.join(s))
    return '\n'.join(outlines)


def fix_eol(text, eol):
    """Fix end-of-lines in a text."""
    lines = text.splitlines()
    if isinstance(text, bytes):
        last = b''
        eol = eol.encode('utf-8')
    else:
        last = ''
    lines.append(last)
    return eol.join(lines)

def unicode_to_base64(text, strip_newlines=True):
    """Safe conversion of ``text`` to base64 representation using
    utf-8 bytes.

    Strips newlines from output unless ``strip_newlines`` is `False`.
    """
    text = to_unicode(text)
    text = text.encode('utf-8')
    if strip_newlines:
        rv = base64.b64encode(text)
    else:
        rv = base64.encodebytes(text)
    return str(rv, 'ascii')

def unicode_from_base64(text):
    """Safe conversion of ``text`` to str based on utf-8 bytes."""
    return str(base64.b64decode(text), 'utf-8')


def levenshtein_distance(lhs, rhs):
    """Return the Levenshtein distance between two strings."""
    if len(lhs) > len(rhs):
        rhs, lhs = lhs, rhs
    if not lhs:
        return len(rhs)

    prev = range(len(rhs) + 1)
    for lidx, lch in enumerate(lhs):
        curr = [lidx + 1]
        for ridx, rch in enumerate(rhs):
            cost = (lch != rch) * 2
            curr.append(min(prev[ridx + 1] + 1, # deletion
                            curr[ridx] + 1,     # insertion
                            prev[ridx] + cost)) # substitution
        prev = curr
    return prev[-1]


sub_vars_re = re.compile("[$]([A-Z_][A-Z0-9_]*)")

def sub_vars(text, args):
    """Substitute $XYZ-style variables in a string with provided values.

    :param text: string containing variables to substitute.
    :param args: dictionary with keys matching the variables to be substituted.
                 The keys should not be prefixed with the $ character."""
    def repl(match):
        key = match.group(1)
        return args[key] if key in args else '$' + key
    return sub_vars_re.sub(repl, text)