superfences.py

"""
SuperFences.

pymdownx.superfences
Nested Fenced Code Blocks

This is a modification of the original Fenced Code Extension.
Algorithm has been rewritten to allow for fenced blocks in blockquotes,
lists, etc.  And also , allow for special UML fences like 'flow' for flowcharts
and `sequence` for sequence diagrams.

Modified: 2014 - 2017 Isaac Muse <isaacmuse@gmail.com>
---

Fenced Code Extension for Python Markdown
=========================================

This extension adds Fenced Code Blocks to Python-Markdown.

See <https://pythonhosted.org/Markdown/extensions/fenced_code_blocks.html>
for documentation.

Original code Copyright 2007-2008 [Waylan Limberg](http://achinghead.com/).


All changes Copyright 2008-2014 The Python Markdown Project

License: [BSD](http://www.opensource.org/licenses/bsd-license.php)
"""

from markdown.extensions import Extension
from markdown.preprocessors import Preprocessor
from markdown.blockprocessors import CodeBlockProcessor
from markdown.extensions.attr_list import get_attrs
from markdown import util as md_util
import functools
import re

SOH = '\u0001'  # start
EOT = '\u0004'  # end

PREFIX_CHARS = ('>', ' ', '\t')

RE_NESTED_FENCE_START = re.compile(
    r'''(?x)
    (?P<fence>~{3,}|`{3,})[ \t]*                                                    # Fence opening
    (?:(\{(?P<attrs>[^\}\n]*)\})?|                                                # Optional attributes or
        (?:\.?(?P<lang>[\w#.+-]*))?[ \t]*                                           # Language
        (?P<options>
            (?:
                (?:\b[a-zA-Z][a-zA-Z0-9_]*(?:=(?P<quot>"|').*?(?P=quot))?[ \t]*) |  # Options
            )*
        )
    )[ \t]*$
    '''
)

RE_HL_LINES = re.compile(r'^(?P<hl_lines>\d+(?:-\d+)?(?:[ \t]+\d+(?:-\d+)?)*)$')
RE_LINENUMS = re.compile(r'(?P<linestart>[\d]+)(?:[ \t]+(?P<linestep>[\d]+))?(?:[ \t]+(?P<linespecial>[\d]+))?')
RE_OPTIONS = re.compile(
    r'''(?x)
    (?:
        (?P<key>[a-zA-Z][a-zA-Z0-9_]*)(?:=(?P<quot>"|')(?P<value>.*?)(?P=quot))?
    )
    '''
)

NESTED_FENCE_END = r'%s[ \t]*$'

FENCED_BLOCK_RE = re.compile(
    r'^([\> ]*)%s(%s)%s$' % (
        md_util.HTML_PLACEHOLDER[0],
        md_util.HTML_PLACEHOLDER[1:-1] % r'([0-9]+)',
        md_util.HTML_PLACEHOLDER[-1]
    )
)


def _escape(txt):
    """Basic html escaping."""

    txt = txt.replace('&', '&amp;')
    txt = txt.replace('<', '&lt;')
    txt = txt.replace('>', '&gt;')
    return txt


class CodeStash(object):
    """
    Stash code for later retrieval.

    Store original fenced code here in case we were
    too greedy and need to restore in an indented code
    block.
    """

    def __init__(self):
        """Initialize."""

        self.stash = {}

    def __len__(self):  # pragma: no cover
        """Length of stash."""

        return len(self.stash)

    def get(self, key, default=None):
        """Get the code from the key."""

        code = self.stash.get(key, default)
        return code

    def remove(self, key):
        """Remove the stashed code."""

        del self.stash[key]

    def store(self, key, code, indent_level):
        """Store the code in the stash."""

        self.stash[key] = (code, indent_level)

    def clear_stash(self):
        """Clear the stash."""

        self.stash = {}


def fence_code_format(source, language, class_name, options, md, **kwargs):
    """Format source as code blocks."""

    classes = kwargs['classes']
    id_value = kwargs['id_value']
    attrs = kwargs['attrs']

    if class_name:
        classes.insert(0, class_name)

    id_value = ' id="{}"'.format(id_value) if id_value else ''
    classes = ' class="{}"'.format(' '.join(classes)) if classes else ''
    attrs = ' ' + ' '.join('{k}="{v}"'.format(k=k, v=v) for k, v in attrs.items()) if attrs else ''

    return '<pre%s%s%s><code>%s</code></pre>' % (id_value, classes, attrs, _escape(source))


def fence_div_format(source, language, class_name, options, md, **kwargs):
    """Format source as div."""

    classes = kwargs['classes']
    id_value = kwargs['id_value']
    attrs = kwargs['attrs']

    if class_name:
        classes.insert(0, class_name)

    id_value = ' id="{}"'.format(id_value) if id_value else ''
    classes = ' class="{}"'.format(' '.join(classes)) if classes else ''
    attrs = ' ' + ' '.join('{k}="{v}"'.format(k=k, v=v) for k, v in attrs.items()) if attrs else ''

    return '<div%s%s%s>%s</div>' % (id_value, classes, attrs, _escape(source))


def highlight_validator(language, inputs, options, attrs, md):
    """Highlight validator."""

    use_pygments = md.preprocessors['fenced_code_block'].use_pygments

    for k, v in inputs.items():
        matched = False
        if use_pygments:
            if k.startswith('data-'):
                attrs[k] = v
                continue
            for opt, validator in (('hl_lines', RE_HL_LINES), ('linenums', RE_LINENUMS), ('title', None)):
                if k == opt:
                    if v is not True and (validator is None or validator.match(v) is not None):
                        options[k] = v
                        matched = True
                        break
        if not matched:
            attrs[k] = v

    return True


def default_validator(language, inputs, options, attrs, md):
    """Default validator."""

    for k, v in inputs.items():
        attrs[k] = v
    return True


def _validator(language, inputs, options, attrs, md, validator=None):
    """Validator wrapper."""

    md.preprocessors['fenced_code_block'].get_hl_settings()
    return validator(language, inputs, options, attrs, md)


def _formatter(src='', language='', options=None, md=None, class_name="", _fmt=None, **kwargs):
    """Formatter wrapper."""

    return _fmt(src, language, class_name, options, md, **kwargs)


def _test(language, test_language=None):
    """Test language."""

    return test_language is None or test_language == "*" or language == test_language


class SuperFencesCodeExtension(Extension):
    """SuperFences code block extension."""

    def __init__(self, *args, **kwargs):
        """Initialize."""

        self.superfences = []
        self.config = {
            'disable_indented_code_blocks': [False, "Disable indented code blocks - Default: False"],
            'custom_fences': [[], 'Specify custom fences. Default: See documentation.'],
            'css_class': [
                '',
                "Set class name for wrapper element. The default of CodeHilite or Highlight will be used"
                "if nothing is set. - "
                "Default: ''"
            ],
            'preserve_tabs': [False, "Preserve tabs in fences - Default: False"]
        }
        super().__init__(*args, **kwargs)

    def extend_super_fences(self, name, formatter, validator):
        """Extend SuperFences with the given name, language, and formatter."""

        obj = {
            "name": name,
            "test": functools.partial(_test, test_language=name),
            "formatter": formatter,
            "validator": validator
        }

        if name == '*':
            self.superfences[0] = obj
        else:
            self.superfences.append(obj)

    def extendMarkdown(self, md):
        """Add fenced block preprocessor to the Markdown instance."""

        # Not super yet, so let's make it super
        md.registerExtension(self)
        config = self.getConfigs()

        # Default fenced blocks
        self.superfences.insert(
            0,
            {
                "name": "superfences",
                "test": _test,
                "formatter": None,
                "validator": functools.partial(_validator, validator=highlight_validator)
            }
        )

        # Custom Fences
        custom_fences = config.get('custom_fences', [])
        for custom in custom_fences:
            name = custom.get('name')
            class_name = custom.get('class')
            fence_format = custom.get('format', fence_code_format)
            validator = custom.get('validator', default_validator)
            if name is not None and class_name is not None:
                self.extend_super_fences(
                    name,
                    functools.partial(_formatter, class_name=class_name, _fmt=fence_format),
                    functools.partial(_validator, validator=validator)
                )

        self.md = md
        self.patch_fenced_rule()
        self.stash = CodeStash()

    def patch_fenced_rule(self):
        """
        Patch Python Markdown with our own fenced block extension.

        We don't attempt to protect against a user loading the `fenced_code` extension with this.
        Most likely they will have issues, but they shouldn't have loaded them together in the first place :).
        """

        config = self.getConfigs()

        fenced = SuperFencesBlockPreprocessor(self.md)
        fenced.config = config
        fenced.extension = self
        if self.superfences[0]['name'] == "superfences":
            self.superfences[0]["formatter"] = fenced.highlight
        self.md.preprocessors.register(fenced, "fenced_code_block", 25)

        indented_code = SuperFencesCodeBlockProcessor(self.md.parser)
        indented_code.config = config
        indented_code.extension = self
        self.md.parser.blockprocessors.register(indented_code, "code", 80)

        if config["preserve_tabs"]:
            # Need to squeeze in right after critic.
            raw_fenced = SuperFencesRawBlockPreprocessor(self.md)
            raw_fenced.config = config
            raw_fenced.extension = self
            self.md.preprocessors.register(raw_fenced, "fenced_raw_block", 31.05)
            self.md.registerExtensions(["pymdownx._bypassnorm"], {})

        # Add the highlight extension, but do so in a disabled state so we can just retrieve default configurations
        self.md.registerExtensions(["pymdownx.highlight"], {"pymdownx.highlight": {"_enabled": False}})

    def reset(self):
        """Clear the stash."""

        self.stash.clear_stash()


class SuperFencesBlockPreprocessor(Preprocessor):
    """
    Preprocessor to find fenced code blocks.

    Because this is done as a preprocessor, it might be too greedy.
    We will stash the blocks code and restore if we mistakenly processed
    text from an indented code block.
    """

    CODE_WRAP = '<pre%s><code%s>%s</code></pre>'

    def __init__(self, md):
        """Initialize."""

        super().__init__(md)
        self.tab_len = self.md.tab_length
        self.checked_hl_settings = False
        self.codehilite_conf = {}

    def normalize_ws(self, text):
        """Normalize whitespace."""

        return text.expandtabs(self.tab_len)

    def rebuild_block(self, lines):
        """Dedent the fenced block lines."""

        return '\n'.join([line[self.ws_virtual_len:] for line in lines])

    def get_hl_settings(self):
        """Check for Highlight extension to get its configurations."""

        if not self.checked_hl_settings:
            self.checked_hl_settings = True

            config = None
            self.highlighter = None
            for ext in self.md.registeredExtensions:
                self.highlight_ext = ext
                try:
                    config = getattr(ext, "get_pymdownx_highlight_settings")()
                    self.highlighter = getattr(ext, "get_pymdownx_highlighter")()
                    break
                except AttributeError:
                    pass

            self.attr_list = 'attr_list' in self.md.treeprocessors

            css_class = self.config['css_class']
            self.css_class = css_class if css_class else config['css_class']

            self.extend_pygments_lang = config.get('extend_pygments_lang', None)
            self.guess_lang = config['guess_lang']
            self.pygments_style = config['pygments_style']
            self.use_pygments = config['use_pygments']
            self.noclasses = config['noclasses']
            self.linenums = config['linenums']
            self.linenums_style = config.get('linenums_style', 'table')
            self.linenums_class = config.get('linenums_class', 'linenums')
            self.linenums_special = config.get('linenums_special', -1)
            self.language_prefix = config.get('language_prefix', 'language-')
            self.code_attr_on_pre = config.get('code_attr_on_pre', False)
            self.auto_title = config.get('auto_title', False)
            self.auto_title_map = config.get('auto_title_map', {})
            self.line_spans = config.get('line_spans', '')
            self.line_anchors = config.get('line_anchors', '')
            self.anchor_linenums = config.get('anchor_linenums', False)

    def clear(self):
        """Reset the class variables."""

        self.ws = None
        self.ws_len = 0
        self.ws_virtual_len = 0
        self.fence = None
        self.lang = None
        self.quote_level = 0
        self.code = []
        self.empty_lines = 0
        self.fence_end = None
        self.options = {}
        self.classes = []
        self.id = ''
        self.attrs = {}
        self.formatter = None

    def eval_fence(self, ws, content, start, end):
        """Evaluate a normal fence."""

        if (ws + content).strip() == '':
            # Empty line is okay
            self.empty_lines += 1
            self.code.append(ws + content)
        elif len(ws) != self.ws_virtual_len and content != '':
            # Not indented enough
            self.clear()
        elif self.fence_end.match(content) is not None and not content.startswith((' ', '\t')):
            # End of fence
            try:
                self.process_nested_block(ws, content, start, end)
            except Exception:
                self.clear()
        else:
            # Content line
            self.empty_lines = 0
            self.code.append(ws + content)

    def eval_quoted(self, ws, content, quote_level, start, end):
        """Evaluate fence inside a blockquote."""

        if quote_level > self.quote_level:
            # Quote level exceeds the starting quote level
            self.clear()
        elif quote_level <= self.quote_level:
            if content == '':
                # Empty line is okay
                self.code.append(ws + content)
                self.empty_lines += 1
            elif len(ws) < self.ws_len:
                # Not indented enough
                self.clear()
            elif self.empty_lines and quote_level < self.quote_level:
                # Quote levels don't match and we are signified
                # the end of the block with an empty line
                self.clear()
            elif self.fence_end.match(content) is not None:
                # End of fence
                self.process_nested_block(ws, content, start, end)
            else:
                # Content line
                self.empty_lines = 0
                self.code.append(ws + content)

    def process_nested_block(self, ws, content, start, end):
        """Process the contents of the nested block."""

        self.last = ws + self.normalize_ws(content)
        code = None
        if self.formatter is not None:
            self.line_count = end - start - 2

            code = self.formatter(
                src=self.rebuild_block(self.code),
                language=self.lang,
                md=self.md,
                options=self.options,
                classes=self.classes,
                id_value=self.id,
                attrs=self.attrs if self.attr_list else {}
            )

        if code is not None:
            self._store(self.normalize_ws('\n'.join(self.code)) + '\n', code, start, end)
        self.clear()

    def normalize_hl_line(self, number):
        """
        Normalize highlight line number.

        Clamp outrages numbers. Numbers out of range will be only one increment out range.
        This prevents people from create massive buffers of line numbers that exceed real
        number of code lines.
        """

        number = int(number)
        if number < 1:
            number = 0
        elif number > self.line_count:
            number = self.line_count + 1
        return number

    def parse_hl_lines(self, hl_lines):
        """Parse the lines to highlight."""

        lines = []
        if hl_lines:
            for entry in hl_lines.split():
                line_range = [self.normalize_hl_line(e) for e in entry.split('-')]
                if len(line_range) > 1:
                    if line_range[0] <= line_range[1]:
                        lines.extend(list(range(line_range[0], line_range[1] + 1)))
                elif 1 <= line_range[0] <= self.line_count:
                    lines.extend(line_range)
        return lines

    def parse_line_start(self, linestart):
        """Parse line start."""

        return int(linestart) if linestart else -1

    def parse_line_step(self, linestep):
        """Parse line start."""

        step = int(linestep) if linestep else -1

        return step if step > 1 else -1

    def parse_line_special(self, linespecial):
        """Parse line start."""

        return int(linespecial) if linespecial else -1

    def parse_fence_line(self, line):
        """Parse fence line."""

        ws_len = 0
        ws_virtual_len = 0
        ws = []
        index = 0
        for c in line:
            if ws_virtual_len >= self.ws_virtual_len:
                break
            if c not in PREFIX_CHARS:
                break
            ws_len += 1
            if c == '\t':
                tab_size = self.tab_len - (index % self.tab_len)
                ws_virtual_len += tab_size
                ws.append(' ' * tab_size)
            else:
                tab_size = 1
                ws_virtual_len += 1
                ws.append(c)
            index += tab_size

        return ''.join(ws), line[ws_len:]

    def parse_whitespace(self, line):
        """Parse the whitespace (blockquote syntax is counted as well)."""

        self.ws_len = 0
        self.ws_virtual_len = 0
        ws = []
        for c in line:
            if c not in PREFIX_CHARS:
                break
            self.ws_len += 1
            ws.append(c)

        ws = self.normalize_ws(''.join(ws))
        self.ws_virtual_len = len(ws)

        return ws

    def parse_options(self, m):
        """Get options."""

        okay = False

        if m.group('lang'):
            self.lang = m.group('lang')

        string = m.group('options')

        self.options = {}
        self.attrs = {}
        self.formatter = None
        values = {}
        if string:
            for m in RE_OPTIONS.finditer(string):
                key = m.group('key')
                value = m.group('value')
                if value is None:
                    value = key
                values[key] = value

        # Run per language validator
        for entry in reversed(self.extension.superfences):
            if entry["test"](self.lang):
                options = {}
                attrs = {}
                validator = entry.get("validator", functools.partial(_validator, validator=default_validator))
                try:
                    okay = validator(self.lang, values, options, attrs, self.md)
                except Exception:
                    pass
                if attrs:
                    okay = False
                if okay:
                    self.formatter = entry.get("formatter")
                    self.options = options
                    break

        return okay

    def handle_attrs(self, m):
        """Handle attribute list."""

        okay = False
        attributes = get_attrs(m.group('attrs').replace('\t', ' ' * self.tab_len))

        self.options = {}
        self.attrs = {}
        self.formatter = None
        values = {}
        for k, v in attributes:
            if k == 'id':
                self.id = v
            elif k == '.':
                self.classes.append(v)
            else:
                values[k] = v

        self.lang = self.classes.pop(0) if self.classes else ''

        # Run per language validator
        for entry in reversed(self.extension.superfences):
            if entry["test"](self.lang):
                options = {}
                attrs = {}
                validator = entry.get("validator", functools.partial(_validator, validator=default_validator))
                try:
                    okay = validator(self.lang, values, options, attrs, self.md)
                except Exception:
                    pass
                if okay:
                    self.formatter = entry.get("formatter")
                    self.options = options
                    if self.attr_list:
                        self.attrs = attrs
                    break

        return okay

    def search_nested(self, lines):
        """Search for nested fenced blocks."""

        count = 0
        for line in lines:
            # Strip carriage returns if the lines end with them.
            # This is necessary since we are handling preserved tabs
            # Before whitespace normalization.
            line = line.rstrip('\r')
            if self.fence is None:
                ws = self.parse_whitespace(line)

                # Found the start of a fenced block.
                m = RE_NESTED_FENCE_START.match(line, self.ws_len)
                if m is not None:

                    # Parse options
                    if m.group('attrs'):
                        okay = self.handle_attrs(m)
                    else:
                        okay = self.parse_options(m)

                    if okay:
                        # Valid fence options, handle fence
                        start = count
                        self.first = ws + self.normalize_ws(m.group(0))
                        self.ws = ws
                        self.quote_level = self.ws.count(">")
                        self.empty_lines = 0
                        self.fence = m.group('fence')
                        self.fence_end = re.compile(NESTED_FENCE_END % self.fence)
                    else:
                        # Option parsing failed, abandon fence
                        self.clear()
            else:
                # Evaluate lines
                # - Determine if it is the ending line or content line
                # - If is a content line, make sure it is all indented
                #   with the opening and closing lines (lines with just
                #   whitespace will be stripped so those don't matter).
                # - When content lines are inside blockquotes, make sure
                #   the nested block quote levels make sense according to
                #   blockquote rules.
                ws, content = self.parse_fence_line(line)

                end = count + 1
                quote_level = ws.count(">")

                if self.quote_level:
                    # Handle blockquotes
                    self.eval_quoted(ws, content, quote_level, start, end)
                elif quote_level == 0:
                    # Handle all other cases
                    self.eval_fence(ws, content, start, end)
                else:
                    # Looks like we got a blockquote line
                    # when not in a blockquote.
                    self.clear()

            count += 1

        return self.reassemble(lines)

    def reassemble(self, lines):
        """Reassemble text."""

        # Now that we are done iterating the lines,
        # let's replace the original content with the
        # fenced blocks.
        while len(self.stack):
            fenced, start, end = self.stack.pop()
            lines = lines[:start] + [fenced] + lines[end:]
        return lines

    def highlight(self, src="", language="", options=None, md=None, **kwargs):
        """
        Syntax highlight the code block.

        If configuration is not empty, then the CodeHilite extension
        is enabled, so we call into it to highlight the code.
        """

        classes = kwargs['classes']
        id_value = kwargs['id_value']
        attrs = kwargs['attrs']

        if classes is None:  # pragma: no cover
            classes = []

        # Default format options
        linestep = None
        linestart = None
        linespecial = None
        hl_lines = None
        title = None

        if self.use_pygments:
            if 'hl_lines' in options:
                m = RE_HL_LINES.match(options['hl_lines'])
                hl_lines = m.group('hl_lines')
                del options['hl_lines']
            if 'linenums' in options:
                m = RE_LINENUMS.match(options['linenums'])
                linestart = m.group('linestart')
                linestep = m.group('linestep')
                linespecial = m.group('linespecial')
                del options['linenums']
            if 'title' in options:
                title = options['title']
                del options['title']

        linestep = self.parse_line_step(linestep)
        linestart = self.parse_line_start(linestart)
        linespecial = self.parse_line_special(linespecial)
        hl_lines = self.parse_hl_lines(hl_lines)

        self.highlight_ext.pygments_code_block += 1

        el = self.highlighter(
            guess_lang=self.guess_lang,
            pygments_style=self.pygments_style,
            use_pygments=self.use_pygments,
            noclasses=self.noclasses,
            linenums=self.linenums,
            linenums_style=self.linenums_style,
            linenums_special=self.linenums_special,
            linenums_class=self.linenums_class,
            extend_pygments_lang=self.extend_pygments_lang,
            language_prefix=self.language_prefix,
            code_attr_on_pre=self.code_attr_on_pre,
            auto_title=self.auto_title,
            auto_title_map=self.auto_title_map,
            line_spans=self.line_spans,
            line_anchors=self.line_anchors,
            anchor_linenums=self.anchor_linenums
        ).highlight(
            src,
            language,
            self.css_class,
            hl_lines=hl_lines,
            linestart=linestart,
            linestep=linestep,
            linespecial=linespecial,
            classes=classes,
            id_value=id_value,
            attrs=attrs,
            title=title,
            code_block_count=self.highlight_ext.pygments_code_block
        )

        return el

    def _store(self, source, code, start, end):
        """
        Store the fenced blocks in the stack to be replaced when done iterating.

        Store the original text in case we need to restore if we are too greedy.
        """
        # Save the fenced blocks to add once we are done iterating the lines
        placeholder = self.md.htmlStash.store(code)
        self.stack.append(('%s%s' % (self.ws, placeholder), start, end))
        if not self.disabled_indented:
            # If an indented block consumes this placeholder,
            # we can restore the original source
            self.extension.stash.store(
                placeholder[1:-1],
                "%s\n%s%s" % (self.first, self.normalize_ws(source), self.last),
                self.ws_virtual_len
            )

    def reindent(self, text, pos, level):
        """Reindent the code to where it is supposed to be."""

        indented = []
        for line in text.split('\n'):
            index = pos - level
            indented.append(line[index:])
        return indented

    def restore_raw_text(self, lines):
        """Revert a prematurely converted fenced block."""

        new_lines = []
        for line in lines:
            m = FENCED_BLOCK_RE.match(line)
            if m:
                key = m.group(2)
                indent_level = len(m.group(1))
                original = None
                original, pos = self.extension.stash.get(key, (None, None))
                if original is not None:
                    code = self.reindent(original, pos, indent_level)
                    new_lines.extend(code)
                    self.extension.stash.remove(key)
                if original is None:  # pragma: no cover
                    # Too much work to test this. This is just a fall back in case
                    # we find a placeholder, and we went to revert it and it wasn't in our stash.
                    # Most likely this would be caused by someone else. We just want to put it
                    # back in the block if we can't revert it.  Maybe we can do a more directed
                    # unit test in the future.
                    new_lines.append(line)
            else:
                new_lines.append(line)
        return new_lines

    def run(self, lines):
        """Search for fenced blocks."""

        self.get_hl_settings()
        self.clear()
        self.stack = []
        self.disabled_indented = self.config.get("disable_indented_code_blocks", False)
        self.preserve_tabs = self.config.get("preserve_tabs", False)

        if self.preserve_tabs:
            lines = self.restore_raw_text(lines)
        return self.search_nested(lines)


class SuperFencesRawBlockPreprocessor(SuperFencesBlockPreprocessor):
    """Special class for preserving tabs before normalizing whitespace."""

    def process_nested_block(self, ws, content, start, end):
        """Process the contents of the nested block."""

        self.last = ws + self.normalize_ws(content)
        code = '\n'.join(self.code)
        self._store(code + '\n', code, start, end)
        self.clear()

    def _store(self, source, code, start, end):
        """
        Store the fenced blocks in the stack to be replaced when done iterating.

        Store the original text in case we need to restore if we are too greedy.
        """
        # Just get a placeholder, we won't ever actually retrieve this source
        placeholder = self.md.htmlStash.store('')
        self.stack.append(('%s%s' % (self.ws, placeholder), start, end))
        # Here is the source we'll actually retrieve.
        self.extension.stash.store(
            placeholder[1:-1],
            "%s\n%s%s" % (self.first, source, self.last),
            self.ws_virtual_len
        )

    def reassemble(self, lines):
        """Reassemble text."""

        # Now that we are done iterating the lines,
        # let's replace the original content with the
        # fenced blocks.
        while len(self.stack):
            fenced, start, end = self.stack.pop()
            lines = lines[:start] + [fenced.replace(md_util.STX, SOH, 1)[:-1] + EOT] + lines[end:]
        return lines

    def run(self, lines):
        """Search for fenced blocks."""

        self.get_hl_settings()
        self.clear()
        self.stack = []
        self.disabled_indented = self.config.get("disable_indented_code_blocks", False)
        return self.search_nested(lines)


class SuperFencesCodeBlockProcessor(CodeBlockProcessor):
    """Process indented code blocks to see if we accidentally processed its content as a fenced block."""

    def test(self, parent, block):
        """Test method that is one day to be deprecated."""

        return True

    def reindent(self, text, pos, level):
        """Reindent the code to where it is supposed to be."""

        indented = []
        for line in text.split('\n'):
            index = pos - level
            indented.append(line[index:])
        return '\n'.join(indented)

    def revert_greedy_fences(self, block):
        """Revert a prematurely converted fenced block."""

        new_block = []
        for line in block.split('\n'):
            m = FENCED_BLOCK_RE.match(line)
            if m:
                key = m.group(2)
                indent_level = len(m.group(1))
                original = None
                original, pos = self.extension.stash.get(key)
                if original is not None:
                    code = self.reindent(original, pos, indent_level)
                    new_block.append(code)
                    self.extension.stash.remove(key)
                if original is None:  # pragma: no cover
                    # Too much work to test this. This is just a fall back in case
                    # we find a placeholder, and we went to revert it and it wasn't in our stash.
                    # Most likely this would be caused by someone else. We just want to put it
                    # back in the block if we can't revert it.  Maybe we can do a more directed
                    # unit test in the future.
                    new_block.append(line)
            else:
                new_block.append(line)
        return '\n'.join(new_block)

    def run(self, parent, blocks):
        """Look for and parse code block."""

        handled = False

        if not self.config.get("disable_indented_code_blocks", False):
            handled = CodeBlockProcessor.test(self, parent, blocks[0])
            if handled:
                if self.config.get("nested", True):
                    blocks[0] = self.revert_greedy_fences(blocks[0])
                handled = CodeBlockProcessor.run(self, parent, blocks) is not False
        return handled


def makeExtension(*args, **kwargs):
    """Return extension."""

    return SuperFencesCodeExtension(*args, **kwargs)