utils.py

# -*- coding: utf-8 -*-

# Copyright © 2012-2015 Roberto Alsina and others.

# Permission is hereby granted, free of charge, to any
# person obtaining a copy of this software and associated
# documentation files (the "Software"), to deal in the
# Software without restriction, including without limitation
# the rights to use, copy, modify, merge, publish,
# distribute, sublicense, and/or sell copies of the
# Software, and to permit persons to whom the Software is
# furnished to do so, subject to the following conditions:
#
# The above copyright notice and this permission notice
# shall be included in all copies or substantial portions of
# the Software.
#
# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY
# KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE
# WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR
# PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS
# OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
# OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
# OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
# SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

"""Utility functions."""

from __future__ import print_function, unicode_literals, absolute_import
import calendar
import datetime
import dateutil.tz
import hashlib
import io
import locale
import logging
import natsort
import os
import re
import json
import shutil
import subprocess
import sys
import dateutil.parser
import dateutil.tz
import logbook
import warnings
import PyRSS2Gen as rss
from collections import defaultdict, Callable
from logbook.compat import redirect_logging
from logbook.more import ExceptionHandler, ColorizedStderrHandler
from pygments.formatters import HtmlFormatter
from zipfile import ZipFile as zipf
from doit import tools
from unidecode import unidecode
from pkg_resources import resource_filename
from doit.cmdparse import CmdParse

from nikola import DEBUG

__all__ = ['CustomEncoder', 'get_theme_path', 'get_theme_chain', 'load_messages', 'copy_tree',
           'copy_file', 'slugify', 'unslugify', 'to_datetime', 'apply_filters',
           'config_changed', 'get_crumbs', 'get_tzname', 'get_asset_path',
           '_reload', 'unicode_str', 'bytes_str', 'unichr', 'Functionary',
           'TranslatableSetting', 'TemplateHookRegistry', 'LocaleBorg',
           'sys_encode', 'sys_decode', 'makedirs', 'get_parent_theme_name',
           'demote_headers', 'get_translation_candidate', 'write_metadata',
           'ask', 'ask_yesno', 'options2docstring', 'os_path_split',
           'get_displayed_page_number', 'adjust_name_for_index_path_list',
           'adjust_name_for_index_path', 'adjust_name_for_index_link',
           'NikolaPygmentsHTML', 'create_redirect', 'TreeNode',
           'flatten_tree_structure', 'parse_escaped_hierarchical_category_name',
           'join_hierarchical_category_path', 'indent']

# Are you looking for 'generic_rss_renderer'?
# It's defined in nikola.nikola.Nikola (the site object).

if sys.version_info[0] == 3:
    # Python 3
    bytes_str = bytes
    unicode_str = str
    unichr = chr
    raw_input = input
    from imp import reload as _reload
else:
    bytes_str = str
    unicode_str = unicode  # NOQA
    _reload = reload  # NOQA
    unichr = unichr


class ApplicationWarning(Exception):
    pass


class ColorfulStderrHandler(ColorizedStderrHandler):
    """Stream handler with colors."""
    _colorful = False

    def should_colorize(self, record):
        """Inform about colorization using the value obtained from Nikola."""
        return self._colorful


def get_logger(name, handlers):
    """Get a logger with handlers attached."""
    l = logbook.Logger(name)
    for h in handlers:
        if isinstance(h, list):
            l.handlers += h
        else:
            l.handlers.append(h)
    return l


STDERR_HANDLER = [ColorfulStderrHandler(
    level=logbook.INFO if not DEBUG else logbook.DEBUG,
    format_string=u'[{record.time:%Y-%m-%dT%H:%M:%SZ}] {record.level_name}: {record.channel}: {record.message}'
)]


LOGGER = get_logger('Nikola', STDERR_HANDLER)
STRICT_HANDLER = ExceptionHandler(ApplicationWarning, level='WARNING')

USE_SLUGIFY = True

redirect_logging()

if DEBUG:
    logging.basicConfig(level=logging.DEBUG)
else:
    logging.basicConfig(level=logging.INFO)


def showwarning(message, category, filename, lineno, file=None, line=None):
    """Show a warning (from the warnings subsystem) to the user."""
    try:
        n = category.__name__
    except AttributeError:
        n = str(category)
    get_logger(n, STDERR_HANDLER).warn('{0}:{1}: {2}'.format(filename, lineno, message))

warnings.showwarning = showwarning


def req_missing(names, purpose, python=True, optional=False):
    """Log that we are missing some requirements.

    `names` is a list/tuple/set of missing things.
    `purpose` is a string, specifying the use of the missing things.
              It completes the sentence:
                  In order to {purpose}, you must install ...
    `python` specifies whether the requirements are Python packages
                               or other software.
    `optional` specifies whether the things are required
                                 (this is an error and we exit with code 5)
                                 or not (this is just a warning).

    Returns the message shown to the user (which you can usually discard).
    If no names are specified, False is returned and nothing is shown
    to the user.

    """
    if not (isinstance(names, tuple) or isinstance(names, list) or isinstance(names, set)):
        names = (names,)
    if not names:
        return False
    if python:
        whatarethey_s = 'Python package'
        whatarethey_p = 'Python packages'
    else:
        whatarethey_s = whatarethey_p = 'software'
    if len(names) == 1:
        msg = 'In order to {0}, you must install the "{1}" {2}.'.format(
            purpose, names[0], whatarethey_s)
    else:
        most = '", "'.join(names[:-1])
        pnames = most + '" and "' + names[-1]
        msg = 'In order to {0}, you must install the "{1}" {2}.'.format(
            purpose, pnames, whatarethey_p)

    if optional:
        LOGGER.warn(msg)
    else:
        LOGGER.error(msg)
        LOGGER.error('Exiting due to missing dependencies.')
        sys.exit(5)

    return msg


from nikola import filters as task_filters  # NOQA
ENCODING = sys.getfilesystemencoding() or sys.stdin.encoding


def sys_encode(thing):
    """Return bytes encoded in the system's encoding."""
    if isinstance(thing, unicode_str):
        return thing.encode(ENCODING)
    return thing


def sys_decode(thing):
    """Returns unicode."""
    if isinstance(thing, bytes_str):
        return thing.decode(ENCODING)
    return thing


def makedirs(path):
    """Create a folder."""
    if not path:
        return
    if os.path.exists(path):
        if not os.path.isdir(path):
            raise OSError('Path {0} already exists and is not a folder.'.format(path))
        else:
            return
    try:
        os.makedirs(path)
        return
    except Exception:
        if os.path.isdir(path):
            return
        raise


class Functionary(defaultdict):

    """Class that looks like a function, but is a defaultdict."""

    def __init__(self, default, default_lang):
        super(Functionary, self).__init__(default)
        self.default_lang = default_lang

    def __call__(self, key, lang=None):
        """When called as a function, take an optional lang
        and return self[lang][key]."""

        if lang is None:
            lang = LocaleBorg().current_lang
        return self[lang][key]


class TranslatableSetting(object):

    """
    A setting that can be translated.

    You can access it via: SETTING(lang).  You can omit lang, in which
    case Nikola will ask LocaleBorg, unless you set SETTING.lang,
    which overrides that call.

    You can also stringify the setting and you will get something
    sensible (in what LocaleBorg claims the language is, can also be
    overriden by SETTING.lang). Note that this second method is
    deprecated.  It is kept for backwards compatibility and
    safety.  It is not guaranteed.

    The underlying structure is a defaultdict.  The language that
    is the default value of the dict is provided with __init__().
    If you need access the underlying dict (you generally don’t,
    """

    # WARNING: This is generally not used and replaced with a call to
    #          LocaleBorg().  Set this to a truthy value to override that.
    lang = None

    # Note that this setting is global.  DO NOT set on a per-instance basis!
    default_lang = 'en'

    def __getattribute__(self, attr):
        """Return attributes, falling back to string attributes."""
        try:
            return super(TranslatableSetting, self).__getattribute__(attr)
        except AttributeError:
            return self().__getattribute__(attr)

    def __dir__(self):
        return list(set(self.__dict__).union(set(dir(str))))

    def __init__(self, name, inp, translations):
        """Initialize a translated setting.

        Valid inputs include:

        * a string               -- the same will be used for all languages
        * a dict ({lang: value}) -- each language will use the value specified;
                                    if there is none, default_lang is used.

        """
        self.name = name
        self._inp = inp
        self.translations = translations
        self.overriden_default = False
        self.values = defaultdict()

        if isinstance(inp, dict):
            self.translated = True
            self.values.update(inp)
            if self.default_lang not in self.values.keys():
                self.default_lang = list(self.values.keys())[0]
                self.overridden_default = True
            self.values.default_factory = lambda: self.values[self.default_lang]
            for k in translations.keys():
                if k not in self.values.keys():
                    self.values[k] = inp[self.default_lang]
        else:
            self.translated = False
            self.values[self.default_lang] = inp
            self.values.default_factory = lambda: inp

    def get_lang(self):
        """Return the language that should be used to retrieve settings."""
        if self.lang:
            return self.lang
        elif not self.translated:
            return self.default_lang
        else:
            try:
                return LocaleBorg().current_lang
            except AttributeError:
                return self.default_lang

    def __call__(self, lang=None):
        """
        Return the value in the requested language.

        While lang is None, self.lang (currently set language) is used.
        Otherwise, the standard algorithm is used (see above).

        """
        if lang is None:
            return self.values[self.get_lang()]
        else:
            return self.values[lang]

    def __str__(self):
        """Return the value in the currently set language.  (deprecated)"""
        return self.values[self.get_lang()]

    def __unicode__(self):
        """Return the value in the currently set language.  (deprecated)"""
        return self.values[self.get_lang()]

    def __repr__(self):
        """Provide a representation for programmers."""
        return '<TranslatableSetting: {0!r}>'.format(self.name)

    def format(self, *args, **kwargs):
        """Format ALL the values in the setting the same way."""
        for l in self.values:
            self.values[l] = self.values[l].format(*args, **kwargs)
        self.values.default_factory = lambda: self.values[self.default_lang]
        return self

    def langformat(self, formats):
        """Format ALL the values in the setting, on a per-language basis."""
        if not formats:
            # Input is empty.
            return self
        else:
            # This is a little tricky.
            # Basically, we have some things that may very well be dicts.  Or
            # actually, TranslatableSettings in the original unprocessed dict
            # form.  We need to detect them.

            # First off, we need to check what languages we have and what
            # should we use as the default.
            keys = list(formats)
            if self.default_lang in keys:
                d = formats[self.default_lang]
            else:
                d = formats[keys[0]]
            # Discovering languages of the settings here.
            langkeys = []
            for f in formats.values():
                for a in f[0] + tuple(f[1].values()):
                    if isinstance(a, dict):
                        langkeys += list(a)

            # Now that we know all this, we go through all the languages we have.
            allvalues = set(keys + langkeys + list(self.values))
            self.values['__orig__'] = self.values[self.default_lang]
            for l in allvalues:
                if l in keys:
                    oargs, okwargs = formats[l]
                else:
                    oargs, okwargs = d

                args = []
                kwargs = {}

                for a in oargs:
                    # We create temporary TranslatableSettings and replace the
                    # values with them.
                    if isinstance(a, dict):
                        a = TranslatableSetting('NULL', a, self.translations)
                        args.append(a(l))
                    else:
                        args.append(a)

                for k, v in okwargs.items():
                    if isinstance(v, dict):
                        v = TranslatableSetting('NULL', v, self.translations)
                        kwargs.update({k: v(l)})
                    else:
                        kwargs.update({k: v})

                if l in self.values:
                    self.values[l] = self.values[l].format(*args, **kwargs)
                else:
                    self.values[l] = self.values['__orig__'].format(*args, **kwargs)
                self.values.default_factory = lambda: self.values[self.default_lang]

        return self

    def __getitem__(self, key):
        """Provide an alternate interface via __getitem__."""
        return self.values[key]

    def __setitem__(self, key, value):
        """Set values for translations."""
        self.values[key] = value

    def __eq__(self, other):
        """Test whether two TranslatableSettings are equal."""
        return self.values == other.values

    def __ne__(self, other):
        """Test whether two TranslatableSettings are inequal."""
        return self.values != other.values


class TemplateHookRegistry(object):

    """
    A registry for template hooks.

    Usage:

    >>> r = TemplateHookRegistry('foo', None)
    >>> r.append('Hello!')
    >>> r.append(lambda x: 'Hello ' + x + '!', False, 'world')
    >>> str(r())  # str() call is not recommended in real use
    'Hello!\\nHello world!'
    >>>
    """

    def __init__(self, name, site):
        """Initialize a hook registry."""
        self._items = []
        self.name = name
        self.site = site
        self.context = None

    def generate(self):
        """Generate items."""
        for c, inp, site, args, kwargs in self._items:
            if c:
                if site:
                    kwargs['site'] = self.site
                    kwargs['context'] = self.context
                yield inp(*args, **kwargs)
            else:
                yield inp

    def __call__(self):
        """Return items, in a string, separated by newlines."""
        return '\n'.join(self.generate())

    def append(self, inp, wants_site_and_context=False, *args, **kwargs):
        """
        Register an item.

        `inp` can be a string or a callable returning one.
        `wants_site` tells whether there should be a `site` keyword
                     argument provided, for accessing the site.

        Further positional and keyword arguments are passed as-is to the
        callable.

        `wants_site`, args and kwargs are ignored (but saved!) if `inp`
        is not callable.  Callability of `inp` is determined only once.
        """
        c = callable(inp)
        self._items.append((c, inp, wants_site_and_context, args, kwargs))

    def __hash__(self):
        return hash(config_changed({self.name: self._items})._calc_digest())

    def __str__(self):
        return '<TemplateHookRegistry: {0}>'.format(self._items)


class CustomEncoder(json.JSONEncoder):
    def default(self, obj):
        try:
            return super(CustomEncoder, self).default(obj)
        except TypeError:
            s = repr(obj).split('0x', 1)[0]
            return s


class config_changed(tools.config_changed):
    """ A copy of doit's but using pickle instead of serializing manually."""

    def __init__(self, config, identifier=None):
        super(config_changed, self).__init__(config)
        self.identifier = '_config_changed'
        if identifier is not None:
            self.identifier += ':' + identifier

    def _calc_digest(self):
        if isinstance(self.config, str):
            return self.config
        elif isinstance(self.config, dict):
            data = json.dumps(self.config, cls=CustomEncoder, sort_keys=True)
            if isinstance(data, str):  # pragma: no cover # python3
                byte_data = data.encode("utf-8")
            else:
                byte_data = data
            digest = hashlib.md5(byte_data).hexdigest()
            # LOGGER.debug('{{"{0}": {1}}}'.format(digest, byte_data))
            return digest
        else:
            raise Exception('Invalid type of config_changed parameter -- got '
                            '{0}, must be string or dict'.format(type(
                                self.config)))

    def configure_task(self, task):
        task.value_savers.append(lambda: {self.identifier: self._calc_digest()})

    def __call__(self, task, values):
        """Return True if config values are unchanged."""
        last_success = values.get(self.identifier)
        if last_success is None:
            return False
        return (last_success == self._calc_digest())

    def __repr__(self):
        return "Change with config: {0}".format(json.dumps(self.config,
                                                           cls=CustomEncoder))


def get_theme_path(theme, _themes_dir='themes'):
    """Given a theme name, returns the path where its files are located.

    Looks in ./themes and in the place where themes go when installed.
    """
    dir_name = os.path.join(_themes_dir, theme)
    if os.path.isdir(dir_name):
        return dir_name
    dir_name = resource_filename('nikola', os.path.join('data', 'themes', theme))
    if os.path.isdir(dir_name):
        return dir_name
    raise Exception("Can't find theme '{0}'".format(theme))


def get_template_engine(themes, _themes_dir='themes'):
    for theme_name in themes:
        engine_path = os.path.join(get_theme_path(theme_name, _themes_dir), 'engine')
        if os.path.isfile(engine_path):
            with open(engine_path) as fd:
                return fd.readlines()[0].strip()
    # default
    return 'mako'


def get_parent_theme_name(theme_name, _themes_dir='themes'):
    parent_path = os.path.join(get_theme_path(theme_name, _themes_dir), 'parent')
    if os.path.isfile(parent_path):
        with open(parent_path) as fd:
            return fd.readlines()[0].strip()
    return None


def get_theme_chain(theme, _themes_dir='themes'):
    """Create the full theme inheritance chain."""
    themes = [theme]

    while True:
        parent = get_parent_theme_name(themes[-1], _themes_dir)
        # Avoid silly loops
        if parent is None or parent in themes:
            break
        themes.append(parent)
    return themes


warned = []


class LanguageNotFoundError(Exception):
    def __init__(self, lang, orig):
        self.lang = lang
        self.orig = orig

    def __str__(self):
        return 'cannot find language {0}'.format(self.lang)


def load_messages(themes, translations, default_lang):
    """ Load theme's messages into context.

    All the messages from parent themes are loaded,
    and "younger" themes have priority.
    """
    messages = Functionary(dict, default_lang)
    oldpath = list(sys.path)
    for theme_name in themes[::-1]:
        msg_folder = os.path.join(get_theme_path(theme_name), 'messages')
        default_folder = os.path.join(get_theme_path('base'), 'messages')
        sys.path.insert(0, default_folder)
        sys.path.insert(0, msg_folder)
        english = __import__('messages_en')
        for lang in list(translations.keys()):
            try:
                translation = __import__('messages_' + lang)
                # If we don't do the reload, the module is cached
                _reload(translation)
                if sorted(translation.MESSAGES.keys()) !=\
                        sorted(english.MESSAGES.keys()) and \
                        lang not in warned:
                    warned.append(lang)
                    LOGGER.warn("Incomplete translation for language "
                                "'{0}'.".format(lang))
                messages[lang].update(english.MESSAGES)
                for k, v in translation.MESSAGES.items():
                    if v:
                        messages[lang][k] = v
                del(translation)
            except ImportError as orig:
                raise LanguageNotFoundError(lang, orig)
    sys.path = oldpath
    return messages


def copy_tree(src, dst, link_cutoff=None):
    """Copy a src tree to the dst folder.

    Example:

    src = "themes/default/assets"
    dst = "output/assets"

    should copy "themes/defauts/assets/foo/bar" to
    "output/assets/foo/bar"

    if link_cutoff is set, then the links pointing at things
    *inside* that folder will stay as links, and links
    pointing *outside* that folder will be copied.
    """
    ignore = set(['.svn'])
    base_len = len(src.split(os.sep))
    for root, dirs, files in os.walk(src, followlinks=True):
        root_parts = root.split(os.sep)
        if set(root_parts) & ignore:
            continue
        dst_dir = os.path.join(dst, *root_parts[base_len:])
        makedirs(dst_dir)
        for src_name in files:
            if src_name in ('.DS_Store', 'Thumbs.db'):
                continue
            dst_file = os.path.join(dst_dir, src_name)
            src_file = os.path.join(root, src_name)
            yield {
                'name': dst_file,
                'file_dep': [src_file],
                'targets': [dst_file],
                'actions': [(copy_file, (src_file, dst_file, link_cutoff))],
                'clean': True,
            }


def copy_file(source, dest, cutoff=None):
    dst_dir = os.path.dirname(dest)
    makedirs(dst_dir)
    if os.path.islink(source):
        link_target = os.path.relpath(
            os.path.normpath(os.path.join(dst_dir, os.readlink(source))))
        # Now we have to decide if we copy the link target or the
        # link itself.
        if cutoff is None or not link_target.startswith(cutoff):
            # We copy
            shutil.copy2(source, dest)
        else:
            # We link
            if os.path.exists(dest) or os.path.islink(dest):
                os.unlink(dest)
            os.symlink(os.readlink(source), dest)
    else:
        shutil.copy2(source, dest)


def remove_file(source):
    if os.path.isdir(source):
        shutil.rmtree(source)
    elif os.path.isfile(source) or os.path.islink(source):
        os.remove(source)

# slugify is copied from
# http://code.activestate.com/recipes/
# 577257-slugify-make-a-string-usable-in-a-url-or-filename/
_slugify_strip_re = re.compile(r'[^+\w\s-]')
_slugify_hyphenate_re = re.compile(r'[-\s]+')


def slugify(value, force=False):
    """
    Normalizes string, converts to lowercase, removes non-alpha characters,
    and converts spaces to hyphens.

    From Django's "django/template/defaultfilters.py".

    >>> print(slugify('\xe1\xe9\xed.\xf3\xfa'))
    aeiou

    >>> print(slugify('foo/bar'))
    foobar

    >>> print(slugify('foo bar'))
    foo-bar

    """
    if not isinstance(value, unicode_str):
        raise ValueError("Not a unicode object: {0}".format(value))
    if USE_SLUGIFY or force:
        # This is the standard state of slugify, which actually does some work.
        # It is the preferred style, especially for Western languages.
        value = unicode_str(unidecode(value))
        value = _slugify_strip_re.sub('', value, re.UNICODE).strip().lower()
        return _slugify_hyphenate_re.sub('-', value, re.UNICODE)
    else:
        # This is the “disarmed” state of slugify, which lets the user
        # have any character they please (be it regular ASCII with spaces,
        # or another alphabet entirely).  This might be bad in some
        # environments, and as such, USE_SLUGIFY is better off being True!

        # We still replace some characters, though.  In particular, we need
        # to replace ? and #, which should not appear in URLs, and some
        # Windows-unsafe characters.  This list might be even longer.
        rc = '/\\?#"\'\r\n\t*:<>|"'

        for c in rc:
            value = value.replace(c, '-')
        return value


def unslugify(value, discard_numbers=True):
    """Given a slug string (as a filename), return a human readable string.

    If discard_numbers is True, numbers right at the beginning of input
    will be removed.
    """
    if discard_numbers:
        value = re.sub('^[0-9]+', '', value)
    value = re.sub('([_\-\.])', ' ', value)
    value = value.strip().capitalize()
    return value


# A very slightly safer version of zip.extractall that works on
# python < 2.6

class UnsafeZipException(Exception):
    pass


def extract_all(zipfile, path='themes'):
    pwd = os.getcwd()
    makedirs(path)
    os.chdir(path)
    z = zipf(zipfile)
    namelist = z.namelist()
    for f in namelist:
        if f.endswith('/') and '..' in f:
            raise UnsafeZipException('The zip file contains ".." and is '
                                     'not safe to expand.')
    for f in namelist:
        if f.endswith('/'):
            makedirs(f)
        else:
            z.extract(f)
    z.close()
    os.chdir(pwd)


def to_datetime(value, tzinfo=None):
    try:
        if not isinstance(value, datetime.datetime):
            # dateutil does bad things with TZs like UTC-03:00.
            dateregexp = re.compile(r' UTC([+-][0-9][0-9]:[0-9][0-9])')
            value = re.sub(dateregexp, r'\1', value)
            value = dateutil.parser.parse(value)
        if not value.tzinfo:
            value = value.replace(tzinfo=tzinfo)
        return value
    except Exception:
        raise ValueError('Unrecognized date/time: {0!r}'.format(value))


def get_tzname(dt):
    """
    Given a datetime value, find the name of the time zone.

    DEPRECATED: This thing returned basically the 1st random zone
    that matched the offset.
    """
    return dt.tzname()


def current_time(tzinfo=None):
    if tzinfo is not None:
        dt = datetime.datetime.now(tzinfo)
    else:
        dt = datetime.datetime.now(dateutil.tz.tzlocal())
    return dt


def apply_filters(task, filters, skip_ext=None):
    """
    Given a task, checks its targets.
    If any of the targets has a filter that matches,
    adds the filter commands to the commands of the task,
    and the filter itself to the uptodate of the task.
    """

    if '.php' in filters.keys():
        if task_filters.php_template_injection not in filters['.php']:
            filters['.php'].append(task_filters.php_template_injection)
    else:
        filters['.php'] = [task_filters.php_template_injection]

    def filter_matches(ext):
        for key, value in list(filters.items()):
            if isinstance(key, (tuple, list)):
                if ext in key:
                    return value
            elif isinstance(key, (bytes_str, unicode_str)):
                if ext == key:
                    return value
            else:
                assert False, key

    for target in task.get('targets', []):
        ext = os.path.splitext(target)[-1].lower()
        if skip_ext and ext in skip_ext:
            continue
        filter_ = filter_matches(ext)
        if filter_:
            for action in filter_:
                def unlessLink(action, target):
                    if not os.path.islink(target):
                        if isinstance(action, Callable):
                            action(target)
                        else:
                            subprocess.check_call(action % target, shell=True)

                task['actions'].append((unlessLink, (action, target)))
    return task


def get_crumbs(path, is_file=False, index_folder=None):
    """Create proper links for a crumb bar.
    index_folder is used if you want to use title from index file
    instead of folder name as breadcrumb text.

    >>> crumbs = get_crumbs('galleries')
    >>> len(crumbs)
    1
    >>> print('|'.join(crumbs[0]))
    #|galleries

    >>> crumbs = get_crumbs(os.path.join('galleries','demo'))
    >>> len(crumbs)
    2
    >>> print('|'.join(crumbs[0]))
    ..|galleries
    >>> print('|'.join(crumbs[1]))
    #|demo

    >>> crumbs = get_crumbs(os.path.join('listings','foo','bar'), is_file=True)
    >>> len(crumbs)
    3
    >>> print('|'.join(crumbs[0]))
    ..|listings
    >>> print('|'.join(crumbs[1]))
    .|foo
    >>> print('|'.join(crumbs[2]))
    #|bar
    """

    crumbs = path.split(os.sep)
    _crumbs = []
    if is_file:
        for i, crumb in enumerate(crumbs[-3::-1]):  # Up to parent folder only
            _path = '/'.join(['..'] * (i + 1))
            _crumbs.append([_path, crumb])
        _crumbs.insert(0, ['.', crumbs[-2]])  # file's folder
        _crumbs.insert(0, ['#', crumbs[-1]])  # file itself
    else:
        for i, crumb in enumerate(crumbs[::-1]):
            _path = '/'.join(['..'] * i) or '#'
            _crumbs.append([_path, crumb])
    if index_folder and hasattr(index_folder, 'parse_index'):
        folder = path
        for i, crumb in enumerate(crumbs[::-1]):
            if folder[-1] == os.sep:
                folder = folder[:-1]
            # We don't care about the created Post() object except for its title;
            # hence, the input_folder and output_folder given to
            # index_folder.parse_index() don't matter
            index_post = index_folder.parse_index(folder, '', '')
            folder = folder.replace(crumb, '')
            if index_post:
                crumb = index_post.title() or crumb
            _crumbs[i][1] = crumb
    return list(reversed(_crumbs))


def get_asset_path(path, themes, files_folders={'files': ''}, _themes_dir='themes'):
    """
    .. versionchanged:: 6.1.0

    Checks which theme provides the path with the given asset,
    and returns the "real", absolute path to the asset.

    If the asset is not provided by a theme, then it will be checked for
    in the FILES_FOLDERS

    >>> print(get_asset_path('assets/css/rst.css', ['bootstrap3', 'base']))
    /.../nikola/data/themes/base/assets/css/rst.css

    >>> print(get_asset_path('assets/css/theme.css', ['bootstrap3', 'base']))
    /.../nikola/data/themes/bootstrap3/assets/css/theme.css

    >>> print(get_asset_path('nikola.py', ['bootstrap3', 'base'], {'nikola': ''}))
    /.../nikola/nikola.py

    >>> print(get_asset_path('nikola/nikola.py', ['bootstrap3', 'base'], {'nikola':'nikola'}))
    None

    """
    for theme_name in themes:
        candidate = os.path.join(
            get_theme_path(theme_name, _themes_dir),
            path
        )
        if os.path.isfile(candidate):
            return candidate
    for src, rel_dst in files_folders.items():
        candidate = os.path.abspath(os.path.join(src, path))
        if os.path.isfile(candidate):
            return candidate

    # whatever!
    return None


class LocaleBorgUninitializedException(Exception):
    def __init__(self):
        super(LocaleBorgUninitializedException, self).__init__("Attempt to use LocaleBorg before initialization")


class LocaleBorg(object):
    """
    Provides locale related services and autoritative current_lang,
    where current_lang is the last lang for which the locale was set.

    current_lang is meant to be set only by LocaleBorg.set_locale

    python's locale code should not be directly called from code outside of
    LocaleBorg, they are compatibilty issues with py version and OS support
    better handled at one central point, LocaleBorg.

    In particular, don't call locale.setlocale outside of LocaleBorg.

    Assumptions:
        We need locales only for the languages there is a nikola translation.
        We don't need to support current_lang through nested contexts

    Usage:
        # early in cmd or test execution
        LocaleBorg.initialize(...)

        # any time later
        lang = LocaleBorg().<service>

    Available services:
        .current_lang : autoritative current_lang , the last seen in set_locale
        .set_locale(lang) : sets current_lang and sets the locale for lang
        .get_month_name(month_no, lang) : returns the localized month name

    NOTE: never use locale.getlocale() , it can return values that
    locale.setlocale will not accept in Windows XP, 7 and pythons 2.6, 2.7, 3.3
    Examples: "Spanish", "French" can't do the full circle set / get / set
    That used to break calendar, but now seems is not the case, with month at least
    """

    initialized = False

    @classmethod
    def initialize(cls, locales, initial_lang):
        """
        locales : dict with lang: locale_n
            the same keys as in nikola's TRANSLATIONS
            locale_n a sanitized locale, meaning
                locale.setlocale(locale.LC_ALL, locale_n) will succeed
                locale_n expressed in the string form, like "en.utf8"
        """
        assert initial_lang is not None and initial_lang in locales
        cls.reset()
        cls.locales = locales

        # needed to decode some localized output in py2x
        encodings = {}
        for lang in locales:
            locale.setlocale(locale.LC_ALL, locales[lang])
            loc, encoding = locale.getlocale()
            encodings[lang] = encoding

        cls.encodings = encodings
        cls.__shared_state['current_lang'] = initial_lang
        cls.initialized = True

    @classmethod
    def reset(cls):
        """used in testing to not leak state between tests"""
        cls.locales = {}
        cls.encodings = {}
        cls.__shared_state = {'current_lang': None}
        cls.initialized = False

    def __init__(self):
        if not self.initialized:
            raise LocaleBorgUninitializedException()
        self.__dict__ = self.__shared_state

    def set_locale(self, lang):
        """Sets the locale for language lang, returns ''

        in linux the locale encoding is set to utf8,
        in windows that cannot be guaranted.
        In either case, the locale encoding is available in cls.encodings[lang]
        """
        # intentional non try-except: templates must ask locales with a lang,
        # let the code explode here and not hide the point of failure
        # Also, not guarded with an if lang==current_lang because calendar may
        # put that out of sync
        locale_n = self.locales[lang]
        self.__shared_state['current_lang'] = lang
        locale.setlocale(locale.LC_ALL, locale_n)
        return ''

    def get_month_name(self, month_no, lang):
        """returns localized month name in an unicode string"""
        if sys.version_info[0] == 3:  # Python 3
            with calendar.different_locale(self.locales[lang]):
                s = calendar.month_name[month_no]
            # for py3 s is unicode
        else:  # Python 2
            with calendar.TimeEncoding(self.locales[lang]):
                s = calendar.month_name[month_no]
            enc = self.encodings[lang]
            if not enc:
                enc = 'UTF-8'

            s = s.decode(enc)
        # paranoid about calendar ending in the wrong locale (windows)
        self.set_locale(self.current_lang)
        return s


class ExtendedRSS2(rss.RSS2):
    xsl_stylesheet_href = None

    def publish(self, handler):
        if self.xsl_stylesheet_href:
            handler.processingInstruction("xml-stylesheet", 'type="text/xsl" href="{0}" media="all"'.format(self.xsl_stylesheet_href))
        # old-style class in py2
        rss.RSS2.publish(self, handler)

    def publish_extensions(self, handler):
        if self.self_url:
            handler.startElement("atom:link", {
                'href': self.self_url,
                'rel': "self",
                'type': "application/rss+xml"
            })
            handler.endElement("atom:link")


class ExtendedItem(rss.RSSItem):

    def __init__(self, **kw):
        self.creator = kw.pop('creator')
        # It's an old style class
        return rss.RSSItem.__init__(self, **kw)

    def publish_extensions(self, handler):
        if self.creator:
            handler.startElement("dc:creator", {})
            handler.characters(self.creator)
            handler.endElement("dc:creator")


# \x00 means the "<" was backslash-escaped
explicit_title_re = re.compile(r'^(.+?)\s*(?<!\x00)<(.*?)>$', re.DOTALL)


def split_explicit_title(text):
    """Split role content into title and target, if given.

       From Sphinx's "sphinx/util/nodes.py"
    """
    match = explicit_title_re.match(text)
    if match:
        return True, match.group(1), match.group(2)
    return False, text, text


def first_line(doc):
    """extract first non-blank line from text, to extract docstring title"""
    if doc is not None:
        for line in doc.splitlines():
            striped = line.strip()
            if striped:
                return striped
    return ''


def demote_headers(doc, level=1):
    """Demote <hN> elements by one."""
    if level == 0:
        return doc
    elif level > 0:
        r = range(1, 7 - level)
    elif level < 0:
        r = range(1 + level, 7)
    for i in reversed(r):
        # html headers go to 6, so we can’t “lower” beneath five
            elements = doc.xpath('//h' + str(i))
            for e in elements:
                e.tag = 'h' + str(i + level)


def get_root_dir():
    """Find root directory of nikola installation by looking for conf.py"""
    root = os.getcwd()

    if sys.version_info[0] == 2:
        confname = b'conf.py'
    else:
        confname = 'conf.py'

    while True:
        if os.path.exists(os.path.join(root, confname)):
            return root
        else:
            basedir = os.path.split(root)[0]
            # Top directory, already checked
            if basedir == root:
                break
            root = basedir

    return None


def get_translation_candidate(config, path, lang):
    """
    Return a possible path where we can find the translated version of some page
    based on the TRANSLATIONS_PATTERN configuration variable.

    >>> config = {'TRANSLATIONS_PATTERN': '{path}.{lang}.{ext}', 'DEFAULT_LANG': 'en', 'TRANSLATIONS': {'es':'1', 'en': 1}}
    >>> print(get_translation_candidate(config, '*.rst', 'es'))
    *.es.rst
    >>> print(get_translation_candidate(config, 'fancy.post.rst', 'es'))
    fancy.post.es.rst
    >>> print(get_translation_candidate(config, '*.es.rst', 'es'))
    *.es.rst
    >>> print(get_translation_candidate(config, '*.es.rst', 'en'))
    *.rst
    >>> print(get_translation_candidate(config, 'cache/posts/fancy.post.es.html', 'en'))
    cache/posts/fancy.post.html
    >>> print(get_translation_candidate(config, 'cache/posts/fancy.post.html', 'es'))
    cache/posts/fancy.post.es.html
    >>> print(get_translation_candidate(config, 'cache/stories/charts.html', 'es'))
    cache/stories/charts.es.html
    >>> print(get_translation_candidate(config, 'cache/stories/charts.html', 'en'))
    cache/stories/charts.html

    >>> config = {'TRANSLATIONS_PATTERN': '{path}.{ext}.{lang}', 'DEFAULT_LANG': 'en', 'TRANSLATIONS': {'es':'1', 'en': 1}}
    >>> print(get_translation_candidate(config, '*.rst', 'es'))
    *.rst.es
    >>> print(get_translation_candidate(config, '*.rst.es', 'es'))
    *.rst.es
    >>> print(get_translation_candidate(config, '*.rst.es', 'en'))
    *.rst
    >>> print(get_translation_candidate(config, 'cache/posts/fancy.post.html.es', 'en'))
    cache/posts/fancy.post.html
    >>> print(get_translation_candidate(config, 'cache/posts/fancy.post.html', 'es'))
    cache/posts/fancy.post.html.es

    """
    # FIXME: this is rather slow and this function is called A LOT
    # Convert the pattern into a regexp
    pattern = config['TRANSLATIONS_PATTERN']
    # This will still break if the user has ?*[]\ in the pattern. But WHY WOULD HE?
    pattern = pattern.replace('.', r'\.')
    pattern = pattern.replace('{path}', '(?P<path>.+?)')
    pattern = pattern.replace('{ext}', '(?P<ext>[^\./]+)')
    pattern = pattern.replace('{lang}', '(?P<lang>{0})'.format('|'.join(config['TRANSLATIONS'].keys())))
    m = re.match(pattern, path)
    if m and all(m.groups()):  # It's a translated path
        p, e, l = m.group('path'), m.group('ext'), m.group('lang')
        if l == lang:  # Nothing to do
            return path
        elif lang == config['DEFAULT_LANG']:  # Return untranslated path
            return '{0}.{1}'.format(p, e)
        else:  # Change lang and return
            return config['TRANSLATIONS_PATTERN'].format(path=p, ext=e, lang=lang)
    else:
        # It's a untranslated path, assume it's path.ext
        p, e = os.path.splitext(path)
        e = e[1:]  # No initial dot
        if lang == config['DEFAULT_LANG']:  # Nothing to do
            return path
        else:  # Change lang and return
            return config['TRANSLATIONS_PATTERN'].format(path=p, ext=e, lang=lang)


def write_metadata(data):
    """Write metadata."""
    order = ('title', 'slug', 'date', 'tags', 'category', 'link', 'description', 'type')
    f = '.. {0}: {1}'
    meta = []
    for k in order:
        try:
            meta.append(f.format(k, data.pop(k)))
        except KeyError:
            pass

    # Leftover metadata (user-specified/non-default).
    for k in natsort.natsorted(list(data.keys()), alg=natsort.ns.F | natsort.ns.IC):
        meta.append(f.format(k, data[k]))

    meta.append('')

    return '\n'.join(meta)


def ask(query, default=None):
    """Ask a question."""
    if default:
        default_q = ' [{0}]'.format(default)
    else:
        default_q = ''
    if sys.version_info[0] == 3:
        inp = raw_input("{query}{default_q}: ".format(query=query, default_q=default_q)).strip()
    else:
        inp = raw_input("{query}{default_q}: ".format(query=query, default_q=default_q).encode('utf-8')).strip()
    if inp or default is None:
        return inp
    else:
        return default


def ask_yesno(query, default=None):
    """Ask a yes/no question."""
    if default is None:
        default_q = ' [y/n]'
    elif default is True:
        default_q = ' [Y/n]'
    elif default is False:
        default_q = ' [y/N]'
    if sys.version_info[0] == 3:
        inp = raw_input("{query}{default_q} ".format(query=query, default_q=default_q)).strip()
    else:
        inp = raw_input("{query}{default_q} ".format(query=query, default_q=default_q).encode('utf-8')).strip()
    if inp:
        return inp.lower().startswith('y')
    elif default is not None:
        return default
    else:
        # Loop if no answer and no default.
        return ask_yesno(query, default)


class CommandWrapper(object):
    """Converts commands into functions."""

    def __init__(self, cmd, commands_object):
        self.cmd = cmd
        self.commands_object = commands_object

    def __call__(self, *args, **kwargs):
        if args or (not args and not kwargs):
            self.commands_object._run([self.cmd] + list(args))
        else:
            # Here's where the keyword magic would have to go
            self.commands_object._run_with_kw(self.cmd, *args, **kwargs)


class Commands(object):

    """Nikola Commands.

    Sample usage:
    >>> commands.check('-l')                     # doctest: +SKIP

    Or, if you know the internal argument names:
    >>> commands.check(list=True)                # doctest: +SKIP
    """

    def __init__(self, main, config, doitargs):
        """Takes a main instance, works as wrapper for commands."""
        self._cmdnames = []
        self._main = main
        self._config = config
        self._doitargs = doitargs
        try:
            cmdict = self._doitargs['cmds'].to_dict()
        except AttributeError:  # not a doit PluginDict
            cmdict = self._doitargs['cmds']
        for k, v in cmdict.items():
            # cleanup: run is doit-only, init is useless in an existing site
            if k in ['run', 'init']:
                continue
            if sys.version_info[0] == 2:
                k2 = bytes(k)
            else:
                k2 = k

            self._cmdnames.append(k)

            try:
                # nikola command: already instantiated (singleton)
                opt = v.get_options()
            except TypeError:
                # doit command: needs some help
                opt = v(config=self._config, **self._doitargs).get_options()
            nc = type(
                k2,
                (CommandWrapper,),
                {
                    '__doc__': options2docstring(k, opt)
                })
            setattr(self, k, nc(k, self))

    def _run(self, cmd_args):
        self._main.run(cmd_args)

    def _run_with_kw(self, cmd, *a, **kw):
        # cyclic import hack
        from nikola.plugin_categories import Command
        try:
            cmd = self._doitargs['cmds'].get_plugin(cmd)
        except AttributeError:  # not a doit PluginDict
            cmd = self._doitargs['cmds'][cmd]
        try:
            opt = cmd.get_options()
        except TypeError:
            cmd = cmd(config=self._config, **self._doitargs)
            opt = cmd.get_options()

        options, _ = CmdParse(opt).parse([])
        options.update(kw)
        if isinstance(cmd, Command):
            cmd.execute(options=options, args=a)
        else:  # Doit command
            cmd.execute(options, a)

    def __repr__(self):
        """Return useful and verbose help."""

        return """\
<Nikola Commands>

    Sample usage:
    >>> commands.check('-l')

    Or, if you know the internal argument names:
    >>> commands.check(list=True)

Available commands: {0}.""".format(', '.join(self._cmdnames))


def options2docstring(name, options):
    result = ['Function wrapper for command %s' % name, 'arguments:']
    for opt in options:
        result.append('{0} type {1} default {2}'.format(opt.name, opt.type.__name__, opt.default))
    return '\n'.join(result)


class NikolaPygmentsHTML(HtmlFormatter):
    """A Nikola-specific modification of Pygments’ HtmlFormatter."""
    def __init__(self, anchor_ref, classes=None, linenos='table', linenostart=1):
        if classes is None:
            classes = ['code', 'literal-block']
        self.nclasses = classes
        super(NikolaPygmentsHTML, self).__init__(
            cssclass='code', linenos=linenos, linenostart=linenostart, nowrap=False,
            lineanchors=slugify(anchor_ref, force=True), anchorlinenos=True)

    def wrap(self, source, outfile):
        """
        Wrap the ``source``, which is a generator yielding
        individual lines, in custom generators.
        """

        style = []
        if self.prestyles:
            style.append(self.prestyles)
        if self.noclasses:
            style.append('line-height: 125%')
        style = '; '.join(style)
        classes = ' '.join(self.nclasses)

        yield 0, ('<pre class="{0}"'.format(classes) + (style and ' style="{0}"'.format(style)) + '>')
        for tup in source:
            yield tup
        yield 0, '</pre>'


def get_displayed_page_number(i, num_pages, site):
    if not i:
        i = 0
    if site.config["INDEXES_STATIC"]:
        return i if i > 0 else num_pages
    else:
        return i + 1 if site.config["INDEXES_PAGES_MAIN"] else i


def adjust_name_for_index_path_list(path_list, i, displayed_i, lang, site, force_addition=False, extension=None):
    index_file = site.config["INDEX_FILE"]
    if i or force_addition:
        path_list = list(path_list)
        if force_addition and not i:
            i = 0
        if not extension:
            _, extension = os.path.splitext(index_file)
        if len(path_list) > 0 and path_list[-1] == '':
            path_list[-1] = index_file
        elif len(path_list) == 0 or not path_list[-1].endswith(extension):
            path_list.append(index_file)
        if site.config["PRETTY_URLS"] and site.config["INDEXES_PRETTY_PAGE_URL"](lang) and path_list[-1] == index_file:
            path_schema = site.config["INDEXES_PRETTY_PAGE_URL"](lang)
            if isinstance(path_schema, (bytes_str, unicode_str)):
                path_schema = [path_schema]
        else:
            path_schema = None
        if path_schema is not None:
            del path_list[-1]
            for entry in path_schema:
                path_list.append(entry.format(number=displayed_i, old_number=i, index_file=index_file))
        else:
            path_list[-1] = '{0}-{1}{2}'.format(os.path.splitext(path_list[-1])[0], i, extension)
    return path_list


def os_path_split(path):
    result = []
    while True:
        previous_path = path
        path, tail = os.path.split(path)
        if path == previous_path and tail == '':
            result.insert(0, path)
            break
        result.insert(0, tail)
        if len(path) == 0:
            break
    return result


def adjust_name_for_index_path(name, i, displayed_i, lang, site, force_addition=False, extension=None):
    return os.path.join(*adjust_name_for_index_path_list(os_path_split(name), i, displayed_i, lang, site, force_addition, extension))


def adjust_name_for_index_link(name, i, displayed_i, lang, site, force_addition=False, extension=None):
    link = adjust_name_for_index_path_list(name.split('/'), i, displayed_i, lang, site, force_addition, extension)
    if not extension == ".atom":
        if len(link) > 0 and link[-1] == site.config["INDEX_FILE"] and site.config["STRIP_INDEXES"]:
            link[-1] = ''
    return '/'.join(link)


def create_redirect(src, dst):
    makedirs(os.path.dirname(src))
    with io.open(src, "w+", encoding="utf8") as fd:
        fd.write('<!DOCTYPE html>\n<head>\n<meta charset="utf-8">\n'
                 '<title>Redirecting...</title>\n<meta name="robots" '
                 'content="noindex">\n<meta http-equiv="refresh" content="0; '
                 'url={0}">\n</head>\n<body>\n<p>Page moved '
                 '<a href="{0}">here</a>.</p>\n</body>'.format(dst))


class TreeNode(object):
    indent_levels = None  # use for formatting comments as tree
    indent_change_before = 0  # use for formatting comments as tree
    indent_change_after = 0  # use for formatting comments as tree

    # The indent levels and changes allow to render a tree structure
    # without keeping track of all that information during rendering.
    #
    # The indent_change_before is the different between the current
    # comment's level and the previous comment's level; if the number
    # is positive, the current level is indented further in, and if it
    # is negative, it is indented further out. Positive values can be
    # used to open HTML tags for each opened level.
    #
    # The indent_change_after is the difference between the next
    # comment's level and the current comment's level. Negative values
    # can be used to close HTML tags for each closed level.
    #
    # The indent_levels list contains one entry (index, count) per
    # level, informing about the index of the current comment on that
    # level and the count of comments on that level (before a comment
    # of a higher level comes). This information can be used to render
    # tree indicators, for example to generate a tree such as:
    #
    # +--- [(0,3)]
    # +-+- [(1,3)]
    # | +--- [(1,3), (0,2)]
    # | +-+- [(1,3), (1,2)]
    # |   +--- [(1,3), (1,2), (0, 1)]
    # +-+- [(2,3)]
    #   +- [(2,3), (0,1)]
    #
    # (The lists used as labels represent the content of the
    # indent_levels property for that node.)

    def __init__(self, name, parent=None):
        self.name = name
        self.parent = parent
        self.children = []

    def get_path(self):
        path = []
        curr = self
        while curr is not None:
            path.append(curr)
            curr = curr.parent
        return reversed(path)

    def get_children(self):
        return self.children


def flatten_tree_structure(root_list):
    elements = []

    def generate(input_list, indent_levels_so_far):
        for index, element in enumerate(input_list):
            # add to destination
            elements.append(element)
            # compute and set indent levels
            indent_levels = indent_levels_so_far + [(index, len(input_list))]
            element.indent_levels = indent_levels
            # add children
            children = element.get_children()
            element.children_count = len(children)
            generate(children, indent_levels)

    generate(root_list, [])
    # Add indent change counters
    level = 0
    last_element = None
    for element in elements:
        new_level = len(element.indent_levels)
        # Compute level change before this element
        change = new_level - level
        if last_element is not None:
            last_element.indent_change_after = change
        element.indent_change_before = change
        # Update variables
        level = new_level
        last_element = element
    # Set level change after last element
    if last_element is not None:
        last_element.indent_change_after = -level
    return elements


def parse_escaped_hierarchical_category_name(category_name):
    result = []
    current = None
    index = 0
    next_backslash = category_name.find('\\', index)
    next_slash = category_name.find('/', index)
    while index < len(category_name):
        if next_backslash == -1 and next_slash == -1:
            current = (current if current else "") + category_name[index:]
            index = len(category_name)
        elif next_slash >= 0 and (next_backslash == -1 or next_backslash > next_slash):
            result.append((current if current else "") + category_name[index:next_slash])
            current = ''
            index = next_slash + 1
            next_slash = category_name.find('/', index)
        else:
            if len(category_name) == next_backslash + 1:
                raise Exception("Unexpected '\\' in '{0}' at last position!".format(category_name))
            esc_ch = category_name[next_backslash + 1]
            if esc_ch not in {'/', '\\'}:
                raise Exception("Unknown escape sequence '\\{0}' in '{1}'!".format(esc_ch, category_name))
            current = (current if current else "") + category_name[index:next_backslash] + esc_ch
            index = next_backslash + 2
            next_backslash = category_name.find('\\', index)
            if esc_ch == '/':
                next_slash = category_name.find('/', index)
    if current is not None:
        result.append(current)
    return result


def join_hierarchical_category_path(category_path):
    def escape(s):
        return s.replace('\\', '\\\\').replace('/', '\\/')

    return '/'.join([escape(p) for p in category_path])


# Stolen from textwrap in Python 3.4.3.
def indent(text, prefix, predicate=None):
    """Adds 'prefix' to the beginning of selected lines in 'text'.

    If 'predicate' is provided, 'prefix' will only be added to the lines
    where 'predicate(line)' is True. If 'predicate' is not provided,
    it will default to adding 'prefix' to all non-empty lines that do not
    consist solely of whitespace characters.
    """
    if predicate is None:
        def predicate(line):
            return line.strip()

    def prefixed_lines():
        for line in text.splitlines(True):
            yield (prefix + line if predicate(line) else line)
    return ''.join(prefixed_lines())