diff --git a/lib/soupsieve/__init__.py b/lib/soupsieve/__init__.py new file mode 100644 index 00000000..ebd3a4a4 --- /dev/null +++ b/lib/soupsieve/__init__.py @@ -0,0 +1,127 @@ +""" +Soup Sieve. + +A CSS selector filter for BeautifulSoup4. + +MIT License + +Copyright (c) 2018 Isaac Muse + +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. +""" +from __future__ import unicode_literals +from .__meta__ import __version__, __version_info__ # noqa: F401 +from . import css_parser as cp +from . import css_match as cm +from . import css_types as ct +from .util import DEBUG, deprecated, SelectorSyntaxError # noqa: F401 + +__all__ = ( + 'DEBUG', 'SelectorSyntaxError', 'SoupSieve', + 'closest', 'comments', 'compile', 'filter', 'icomments', + 'iselect', 'match', 'select', 'select_one' +) + +SoupSieve = cm.SoupSieve + + +def compile(pattern, namespaces=None, flags=0, **kwargs): # noqa: A001 + """Compile CSS pattern.""" + + if namespaces is not None: + namespaces = ct.Namespaces(**namespaces) + + custom = kwargs.get('custom') + if custom is not None: + custom = ct.CustomSelectors(**custom) + + if isinstance(pattern, SoupSieve): + if flags: + raise ValueError("Cannot process 'flags' argument on a compiled selector list") + elif namespaces is not None: + raise ValueError("Cannot process 'namespaces' argument on a compiled selector list") + elif custom is not None: + raise ValueError("Cannot process 'custom' argument on a compiled selector list") + return pattern + + return cp._cached_css_compile(pattern, namespaces, custom, flags) + + +def purge(): + """Purge cached patterns.""" + + cp._purge_cache() + + +def closest(select, tag, namespaces=None, flags=0, **kwargs): + """Match closest ancestor.""" + + return compile(select, namespaces, flags, **kwargs).closest(tag) + + +def match(select, tag, namespaces=None, flags=0, **kwargs): + """Match node.""" + + return compile(select, namespaces, flags, **kwargs).match(tag) + + +def filter(select, iterable, namespaces=None, flags=0, **kwargs): # noqa: A001 + """Filter list of nodes.""" + + return compile(select, namespaces, flags, **kwargs).filter(iterable) + + +@deprecated("'comments' is not related to CSS selectors and will be removed in the future.") +def comments(tag, limit=0, flags=0, **kwargs): + """Get comments only.""" + + return [comment for comment in cm.CommentsMatch(tag).get_comments(limit)] + + +@deprecated("'icomments' is not related to CSS selectors and will be removed in the future.") +def icomments(tag, limit=0, flags=0, **kwargs): + """Iterate comments only.""" + + for comment in cm.CommentsMatch(tag).get_comments(limit): + yield comment + + +def select_one(select, tag, namespaces=None, flags=0, **kwargs): + """Select a single tag.""" + + return compile(select, namespaces, flags, **kwargs).select_one(tag) + + +def select(select, tag, namespaces=None, limit=0, flags=0, **kwargs): + """Select the specified tags.""" + + return compile(select, namespaces, flags, **kwargs).select(tag, limit) + + +def iselect(select, tag, namespaces=None, limit=0, flags=0, **kwargs): + """Iterate the specified tags.""" + + for el in compile(select, namespaces, flags, **kwargs).iselect(tag, limit): + yield el + + +def escape(ident): + """Escape identifier.""" + + return cp.escape(ident) diff --git a/lib/soupsieve/__meta__.py b/lib/soupsieve/__meta__.py new file mode 100644 index 00000000..109e4733 --- /dev/null +++ b/lib/soupsieve/__meta__.py @@ -0,0 +1,190 @@ +"""Meta related things.""" +from __future__ import unicode_literals +from collections import namedtuple +import re + +RE_VER = re.compile( + r'''(?x) + (?P\d+)(?:\.(?P\d+))?(?:\.(?P\d+))? + (?:(?Pa|b|rc)(?P
\d+))?
+    (?:\.post(?P\d+))?
+    (?:\.dev(?P\d+))?
+    '''
+)
+
+REL_MAP = {
+    ".dev": "",
+    ".dev-alpha": "a",
+    ".dev-beta": "b",
+    ".dev-candidate": "rc",
+    "alpha": "a",
+    "beta": "b",
+    "candidate": "rc",
+    "final": ""
+}
+
+DEV_STATUS = {
+    ".dev": "2 - Pre-Alpha",
+    ".dev-alpha": "2 - Pre-Alpha",
+    ".dev-beta": "2 - Pre-Alpha",
+    ".dev-candidate": "2 - Pre-Alpha",
+    "alpha": "3 - Alpha",
+    "beta": "4 - Beta",
+    "candidate": "4 - Beta",
+    "final": "5 - Production/Stable"
+}
+
+PRE_REL_MAP = {"a": 'alpha', "b": 'beta', "rc": 'candidate'}
+
+
+class Version(namedtuple("Version", ["major", "minor", "micro", "release", "pre", "post", "dev"])):
+    """
+    Get the version (PEP 440).
+
+    A biased approach to the PEP 440 semantic version.
+
+    Provides a tuple structure which is sorted for comparisons `v1 > v2` etc.
+      (major, minor, micro, release type, pre-release build, post-release build, development release build)
+    Release types are named in is such a way they are comparable with ease.
+    Accessors to check if a development, pre-release, or post-release build. Also provides accessor to get
+    development status for setup files.
+
+    How it works (currently):
+
+    - You must specify a release type as either `final`, `alpha`, `beta`, or `candidate`.
+    - To define a development release, you can use either `.dev`, `.dev-alpha`, `.dev-beta`, or `.dev-candidate`.
+      The dot is used to ensure all development specifiers are sorted before `alpha`.
+      You can specify a `dev` number for development builds, but do not have to as implicit development releases
+      are allowed.
+    - You must specify a `pre` value greater than zero if using a prerelease as this project (not PEP 440) does not
+      allow implicit prereleases.
+    - You can optionally set `post` to a value greater than zero to make the build a post release. While post releases
+      are technically allowed in prereleases, it is strongly discouraged, so we are rejecting them. It should be
+      noted that we do not allow `post0` even though PEP 440 does not restrict this. This project specifically
+      does not allow implicit post releases.
+    - It should be noted that we do not support epochs `1!` or local versions `+some-custom.version-1`.
+
+    Acceptable version releases:
+
+    ```
+    Version(1, 0, 0, "final")                    1.0
+    Version(1, 2, 0, "final")                    1.2
+    Version(1, 2, 3, "final")                    1.2.3
+    Version(1, 2, 0, ".dev-alpha", pre=4)        1.2a4
+    Version(1, 2, 0, ".dev-beta", pre=4)         1.2b4
+    Version(1, 2, 0, ".dev-candidate", pre=4)    1.2rc4
+    Version(1, 2, 0, "final", post=1)            1.2.post1
+    Version(1, 2, 3, ".dev")                     1.2.3.dev0
+    Version(1, 2, 3, ".dev", dev=1)              1.2.3.dev1
+    ```
+
+    """
+
+    def __new__(cls, major, minor, micro, release="final", pre=0, post=0, dev=0):
+        """Validate version info."""
+
+        # Ensure all parts are positive integers.
+        for value in (major, minor, micro, pre, post):
+            if not (isinstance(value, int) and value >= 0):
+                raise ValueError("All version parts except 'release' should be integers.")
+
+        if release not in REL_MAP:
+            raise ValueError("'{}' is not a valid release type.".format(release))
+
+        # Ensure valid pre-release (we do not allow implicit pre-releases).
+        if ".dev-candidate" < release < "final":
+            if pre == 0:
+                raise ValueError("Implicit pre-releases not allowed.")
+            elif dev:
+                raise ValueError("Version is not a development release.")
+            elif post:
+                raise ValueError("Post-releases are not allowed with pre-releases.")
+
+        # Ensure valid development or development/pre release
+        elif release < "alpha":
+            if release > ".dev" and pre == 0:
+                raise ValueError("Implicit pre-release not allowed.")
+            elif post:
+                raise ValueError("Post-releases are not allowed with pre-releases.")
+
+        # Ensure a valid normal release
+        else:
+            if pre:
+                raise ValueError("Version is not a pre-release.")
+            elif dev:
+                raise ValueError("Version is not a development release.")
+
+        return super(Version, cls).__new__(cls, major, minor, micro, release, pre, post, dev)
+
+    def _is_pre(self):
+        """Is prerelease."""
+
+        return self.pre > 0
+
+    def _is_dev(self):
+        """Is development."""
+
+        return bool(self.release < "alpha")
+
+    def _is_post(self):
+        """Is post."""
+
+        return self.post > 0
+
+    def _get_dev_status(self):  # pragma: no cover
+        """Get development status string."""
+
+        return DEV_STATUS[self.release]
+
+    def _get_canonical(self):
+        """Get the canonical output string."""
+
+        # Assemble major, minor, micro version and append `pre`, `post`, or `dev` if needed..
+        if self.micro == 0:
+            ver = "{}.{}".format(self.major, self.minor)
+        else:
+            ver = "{}.{}.{}".format(self.major, self.minor, self.micro)
+        if self._is_pre():
+            ver += '{}{}'.format(REL_MAP[self.release], self.pre)
+        if self._is_post():
+            ver += ".post{}".format(self.post)
+        if self._is_dev():
+            ver += ".dev{}".format(self.dev)
+
+        return ver
+
+
+def parse_version(ver, pre=False):
+    """Parse version into a comparable Version tuple."""
+
+    m = RE_VER.match(ver)
+
+    # Handle major, minor, micro
+    major = int(m.group('major'))
+    minor = int(m.group('minor')) if m.group('minor') else 0
+    micro = int(m.group('micro')) if m.group('micro') else 0
+
+    # Handle pre releases
+    if m.group('type'):
+        release = PRE_REL_MAP[m.group('type')]
+        pre = int(m.group('pre'))
+    else:
+        release = "final"
+        pre = 0
+
+    # Handle development releases
+    dev = m.group('dev') if m.group('dev') else 0
+    if m.group('dev'):
+        dev = int(m.group('dev'))
+        release = '.dev-' + release if pre else '.dev'
+    else:
+        dev = 0
+
+    # Handle post
+    post = int(m.group('post')) if m.group('post') else 0
+
+    return Version(major, minor, micro, release, pre, post, dev)
+
+
+__version_info__ = Version(1, 9, 5, "final")
+__version__ = __version_info__._get_canonical()
diff --git a/lib/soupsieve/css_match.py b/lib/soupsieve/css_match.py
new file mode 100644
index 00000000..9ff2e88f
--- /dev/null
+++ b/lib/soupsieve/css_match.py
@@ -0,0 +1,1542 @@
+"""CSS matcher."""
+from __future__ import unicode_literals
+from datetime import datetime
+from . import util
+import re
+from .import css_types as ct
+import unicodedata
+
+# Empty tag pattern (whitespace okay)
+RE_NOT_EMPTY = re.compile('[^ \t\r\n\f]')
+
+RE_NOT_WS = re.compile('[^ \t\r\n\f]+')
+
+# Relationships
+REL_PARENT = ' '
+REL_CLOSE_PARENT = '>'
+REL_SIBLING = '~'
+REL_CLOSE_SIBLING = '+'
+
+# Relationships for :has() (forward looking)
+REL_HAS_PARENT = ': '
+REL_HAS_CLOSE_PARENT = ':>'
+REL_HAS_SIBLING = ':~'
+REL_HAS_CLOSE_SIBLING = ':+'
+
+NS_XHTML = 'http://www.w3.org/1999/xhtml'
+NS_XML = 'http://www.w3.org/XML/1998/namespace'
+
+DIR_FLAGS = ct.SEL_DIR_LTR | ct.SEL_DIR_RTL
+RANGES = ct.SEL_IN_RANGE | ct.SEL_OUT_OF_RANGE
+
+DIR_MAP = {
+    'ltr': ct.SEL_DIR_LTR,
+    'rtl': ct.SEL_DIR_RTL,
+    'auto': 0
+}
+
+RE_NUM = re.compile(r"^(?P-?(?:[0-9]{1,}(\.[0-9]+)?|\.[0-9]+))$")
+RE_TIME = re.compile(r'^(?P[0-9]{2}):(?P[0-9]{2})$')
+RE_MONTH = re.compile(r'^(?P[0-9]{4,})-(?P[0-9]{2})$')
+RE_WEEK = re.compile(r'^(?P[0-9]{4,})-W(?P[0-9]{2})$')
+RE_DATE = re.compile(r'^(?P[0-9]{4,})-(?P[0-9]{2})-(?P[0-9]{2})$')
+RE_DATETIME = re.compile(
+    r'^(?P[0-9]{4,})-(?P[0-9]{2})-(?P[0-9]{2})T(?P[0-9]{2}):(?P[0-9]{2})$'
+)
+RE_WILD_STRIP = re.compile(r'(?:(?:-\*-)(?:\*(?:-|$))*|-\*$)')
+
+MONTHS_30 = (4, 6, 9, 11)  # April, June, September, and November
+FEB = 2
+SHORT_MONTH = 30
+LONG_MONTH = 31
+FEB_MONTH = 28
+FEB_LEAP_MONTH = 29
+DAYS_IN_WEEK = 7
+
+
+class _FakeParent(object):
+    """
+    Fake parent class.
+
+    When we have a fragment with no `BeautifulSoup` document object,
+    we can't evaluate `nth` selectors properly.  Create a temporary
+    fake parent so we can traverse the root element as a child.
+    """
+
+    def __init__(self, element):
+        """Initialize."""
+
+        self.contents = [element]
+
+    def __len__(self):
+        """Length."""
+
+        return len(self.contents)
+
+
+class _DocumentNav(object):
+    """Navigate a Beautiful Soup document."""
+
+    @classmethod
+    def assert_valid_input(cls, tag):
+        """Check if valid input tag or document."""
+
+        # Fail on unexpected types.
+        if not cls.is_tag(tag):
+            raise TypeError("Expected a BeautifulSoup 'Tag', but instead recieved type {}".format(type(tag)))
+
+    @staticmethod
+    def is_doc(obj):
+        """Is `BeautifulSoup` object."""
+
+        import bs4
+        return isinstance(obj, bs4.BeautifulSoup)
+
+    @staticmethod
+    def is_tag(obj):
+        """Is tag."""
+
+        import bs4
+        return isinstance(obj, bs4.Tag)
+
+    @staticmethod
+    def is_comment(obj):
+        """Is comment."""
+
+        import bs4
+        return isinstance(obj, bs4.Comment)
+
+    @staticmethod
+    def is_declaration(obj):  # pragma: no cover
+        """Is declaration."""
+
+        import bs4
+        return isinstance(obj, bs4.Declaration)
+
+    @staticmethod
+    def is_cdata(obj):
+        """Is CDATA."""
+
+        import bs4
+        return isinstance(obj, bs4.CData)
+
+    @staticmethod
+    def is_processing_instruction(obj):  # pragma: no cover
+        """Is processing instruction."""
+
+        import bs4
+        return isinstance(obj, bs4.ProcessingInstruction)
+
+    @staticmethod
+    def is_navigable_string(obj):
+        """Is navigable string."""
+
+        import bs4
+        return isinstance(obj, bs4.NavigableString)
+
+    @staticmethod
+    def is_special_string(obj):
+        """Is special string."""
+
+        import bs4
+        return isinstance(obj, (bs4.Comment, bs4.Declaration, bs4.CData, bs4.ProcessingInstruction, bs4.Doctype))
+
+    @classmethod
+    def is_content_string(cls, obj):
+        """Check if node is content string."""
+
+        return cls.is_navigable_string(obj) and not cls.is_special_string(obj)
+
+    @staticmethod
+    def create_fake_parent(el):
+        """Create fake parent for a given element."""
+
+        return _FakeParent(el)
+
+    @staticmethod
+    def is_xml_tree(el):
+        """Check if element (or document) is from a XML tree."""
+
+        return el._is_xml
+
+    def is_iframe(self, el):
+        """Check if element is an `iframe`."""
+
+        return ((el.name if self.is_xml_tree(el) else util.lower(el.name)) == 'iframe') and self.is_html_tag(el)
+
+    def is_root(self, el):
+        """
+        Return whether element is a root element.
+
+        We check that the element is the root of the tree (which we have already pre-calculated),
+        and we check if it is the root element under an `iframe`.
+        """
+
+        root = self.root and self.root is el
+        if not root:
+            parent = self.get_parent(el)
+            root = parent is not None and self.is_html and self.is_iframe(parent)
+        return root
+
+    def get_contents(self, el, no_iframe=False):
+        """Get contents or contents in reverse."""
+        if not no_iframe or not self.is_iframe(el):
+            for content in el.contents:
+                yield content
+
+    def get_children(self, el, start=None, reverse=False, tags=True, no_iframe=False):
+        """Get children."""
+
+        if not no_iframe or not self.is_iframe(el):
+            last = len(el.contents) - 1
+            if start is None:
+                index = last if reverse else 0
+            else:
+                index = start
+            end = -1 if reverse else last + 1
+            incr = -1 if reverse else 1
+
+            if 0 <= index <= last:
+                while index != end:
+                    node = el.contents[index]
+                    index += incr
+                    if not tags or self.is_tag(node):
+                        yield node
+
+    def get_descendants(self, el, tags=True, no_iframe=False):
+        """Get descendants."""
+
+        if not no_iframe or not self.is_iframe(el):
+            next_good = None
+            for child in el.descendants:
+
+                if next_good is not None:
+                    if child is not next_good:
+                        continue
+                    next_good = None
+
+                is_tag = self.is_tag(child)
+
+                if no_iframe and is_tag and self.is_iframe(child):
+                    if child.next_sibling is not None:
+                        next_good = child.next_sibling
+                    else:
+                        last_child = child
+                        while self.is_tag(last_child) and last_child.contents:
+                            last_child = last_child.contents[-1]
+                        next_good = last_child.next_element
+                    yield child
+                    if next_good is None:
+                        break
+                    # Coverage isn't seeing this even though it's executed
+                    continue  # pragma: no cover
+
+                if not tags or is_tag:
+                    yield child
+
+    def get_parent(self, el, no_iframe=False):
+        """Get parent."""
+
+        parent = el.parent
+        if no_iframe and parent is not None and self.is_iframe(parent):
+            parent = None
+        return parent
+
+    @staticmethod
+    def get_tag_name(el):
+        """Get tag."""
+
+        return el.name
+
+    @staticmethod
+    def get_prefix_name(el):
+        """Get prefix."""
+
+        return el.prefix
+
+    @staticmethod
+    def get_uri(el):
+        """Get namespace `URI`."""
+
+        return el.namespace
+
+    @classmethod
+    def get_next(cls, el, tags=True):
+        """Get next sibling tag."""
+
+        sibling = el.next_sibling
+        while tags and not cls.is_tag(sibling) and sibling is not None:
+            sibling = sibling.next_sibling
+        return sibling
+
+    @classmethod
+    def get_previous(cls, el, tags=True):
+        """Get previous sibling tag."""
+
+        sibling = el.previous_sibling
+        while tags and not cls.is_tag(sibling) and sibling is not None:
+            sibling = sibling.previous_sibling
+        return sibling
+
+    @staticmethod
+    def has_html_ns(el):
+        """
+        Check if element has an HTML namespace.
+
+        This is a bit different than whether a element is treated as having an HTML namespace,
+        like we do in the case of `is_html_tag`.
+        """
+
+        ns = getattr(el, 'namespace') if el else None
+        return ns and ns == NS_XHTML
+
+    @staticmethod
+    def split_namespace(el, attr_name):
+        """Return namespace and attribute name without the prefix."""
+
+        return getattr(attr_name, 'namespace', None), getattr(attr_name, 'name', None)
+
+    @staticmethod
+    def get_attribute_by_name(el, name, default=None):
+        """Get attribute by name."""
+
+        value = default
+        if el._is_xml:
+            try:
+                value = el.attrs[name]
+            except KeyError:
+                pass
+        else:
+            for k, v in el.attrs.items():
+                if util.lower(k) == name:
+                    value = v
+                    break
+        return value
+
+    @staticmethod
+    def iter_attributes(el):
+        """Iterate attributes."""
+
+        for k, v in el.attrs.items():
+            yield k, v
+
+    @classmethod
+    def get_classes(cls, el):
+        """Get classes."""
+
+        classes = cls.get_attribute_by_name(el, 'class', [])
+        if isinstance(classes, util.ustr):
+            classes = RE_NOT_WS.findall(classes)
+        return classes
+
+    def get_text(self, el, no_iframe=False):
+        """Get text."""
+
+        return ''.join(
+            [node for node in self.get_descendants(el, tags=False, no_iframe=no_iframe) if self.is_content_string(node)]
+        )
+
+
+class Inputs(object):
+    """Class for parsing and validating input items."""
+
+    @staticmethod
+    def validate_day(year, month, day):
+        """Validate day."""
+
+        max_days = LONG_MONTH
+        if month == FEB:
+            max_days = FEB_LEAP_MONTH if ((year % 4 == 0) and (year % 100 != 0)) or (year % 400 == 0) else FEB_MONTH
+        elif month in MONTHS_30:
+            max_days = SHORT_MONTH
+        return 1 <= day <= max_days
+
+    @staticmethod
+    def validate_week(year, week):
+        """Validate week."""
+
+        max_week = datetime.strptime("{}-{}-{}".format(12, 31, year), "%m-%d-%Y").isocalendar()[1]
+        if max_week == 1:
+            max_week = 53
+        return 1 <= week <= max_week
+
+    @staticmethod
+    def validate_month(month):
+        """Validate month."""
+
+        return 1 <= month <= 12
+
+    @staticmethod
+    def validate_year(year):
+        """Validate year."""
+
+        return 1 <= year
+
+    @staticmethod
+    def validate_hour(hour):
+        """Validate hour."""
+
+        return 0 <= hour <= 23
+
+    @staticmethod
+    def validate_minutes(minutes):
+        """Validate minutes."""
+
+        return 0 <= minutes <= 59
+
+    @classmethod
+    def parse_value(cls, itype, value):
+        """Parse the input value."""
+
+        parsed = None
+        if itype == "date":
+            m = RE_DATE.match(value)
+            if m:
+                year = int(m.group('year'), 10)
+                month = int(m.group('month'), 10)
+                day = int(m.group('day'), 10)
+                if cls.validate_year(year) and cls.validate_month(month) and cls.validate_day(year, month, day):
+                    parsed = (year, month, day)
+        elif itype == "month":
+            m = RE_MONTH.match(value)
+            if m:
+                year = int(m.group('year'), 10)
+                month = int(m.group('month'), 10)
+                if cls.validate_year(year) and cls.validate_month(month):
+                    parsed = (year, month)
+        elif itype == "week":
+            m = RE_WEEK.match(value)
+            if m:
+                year = int(m.group('year'), 10)
+                week = int(m.group('week'), 10)
+                if cls.validate_year(year) and cls.validate_week(year, week):
+                    parsed = (year, week)
+        elif itype == "time":
+            m = RE_TIME.match(value)
+            if m:
+                hour = int(m.group('hour'), 10)
+                minutes = int(m.group('minutes'), 10)
+                if cls.validate_hour(hour) and cls.validate_minutes(minutes):
+                    parsed = (hour, minutes)
+        elif itype == "datetime-local":
+            m = RE_DATETIME.match(value)
+            if m:
+                year = int(m.group('year'), 10)
+                month = int(m.group('month'), 10)
+                day = int(m.group('day'), 10)
+                hour = int(m.group('hour'), 10)
+                minutes = int(m.group('minutes'), 10)
+                if (
+                    cls.validate_year(year) and cls.validate_month(month) and cls.validate_day(year, month, day) and
+                    cls.validate_hour(hour) and cls.validate_minutes(minutes)
+                ):
+                    parsed = (year, month, day, hour, minutes)
+        elif itype in ("number", "range"):
+            m = RE_NUM.match(value)
+            if m:
+                parsed = float(m.group('value'))
+        return parsed
+
+
+class _Match(object):
+    """Perform CSS matching."""
+
+    def __init__(self, selectors, scope, namespaces, flags):
+        """Initialize."""
+
+        self.assert_valid_input(scope)
+        self.tag = scope
+        self.cached_meta_lang = []
+        self.cached_default_forms = []
+        self.cached_indeterminate_forms = []
+        self.selectors = selectors
+        self.namespaces = {} if namespaces is None else namespaces
+        self.flags = flags
+        self.iframe_restrict = False
+
+        # Find the root element for the whole tree
+        doc = scope
+        parent = self.get_parent(doc)
+        while parent:
+            doc = parent
+            parent = self.get_parent(doc)
+        root = None
+        if not self.is_doc(doc):
+            root = doc
+        else:
+            for child in self.get_children(doc):
+                root = child
+                break
+
+        self.root = root
+        self.scope = scope if scope is not doc else root
+        self.has_html_namespace = self.has_html_ns(root)
+
+        # A document can be both XML and HTML (XHTML)
+        self.is_xml = self.is_xml_tree(doc)
+        self.is_html = not self.is_xml or self.has_html_namespace
+
+    def supports_namespaces(self):
+        """Check if namespaces are supported in the HTML type."""
+
+        return self.is_xml or self.has_html_namespace
+
+    def get_tag_ns(self, el):
+        """Get tag namespace."""
+
+        if self.supports_namespaces():
+            namespace = ''
+            ns = self.get_uri(el)
+            if ns:
+                namespace = ns
+        else:
+            namespace = NS_XHTML
+        return namespace
+
+    def is_html_tag(self, el):
+        """Check if tag is in HTML namespace."""
+
+        return self.get_tag_ns(el) == NS_XHTML
+
+    def get_tag(self, el):
+        """Get tag."""
+
+        name = self.get_tag_name(el)
+        return util.lower(name) if name is not None and not self.is_xml else name
+
+    def get_prefix(self, el):
+        """Get prefix."""
+
+        prefix = self.get_prefix_name(el)
+        return util.lower(prefix) if prefix is not None and not self.is_xml else prefix
+
+    def find_bidi(self, el):
+        """Get directionality from element text."""
+
+        for node in self.get_children(el, tags=False):
+
+            # Analyze child text nodes
+            if self.is_tag(node):
+
+                # Avoid analyzing certain elements specified in the specification.
+                direction = DIR_MAP.get(util.lower(self.get_attribute_by_name(node, 'dir', '')), None)
+                if (
+                    self.get_tag(node) in ('bdi', 'script', 'style', 'textarea', 'iframe') or
+                    not self.is_html_tag(node) or
+                    direction is not None
+                ):
+                    continue  # pragma: no cover
+
+                # Check directionality of this node's text
+                value = self.find_bidi(node)
+                if value is not None:
+                    return value
+
+                # Direction could not be determined
+                continue  # pragma: no cover
+
+            # Skip `doctype` comments, etc.
+            if self.is_special_string(node):
+                continue
+
+            # Analyze text nodes for directionality.
+            for c in node:
+                bidi = unicodedata.bidirectional(c)
+                if bidi in ('AL', 'R', 'L'):
+                    return ct.SEL_DIR_LTR if bidi == 'L' else ct.SEL_DIR_RTL
+        return None
+
+    def extended_language_filter(self, lang_range, lang_tag):
+        """Filter the language tags."""
+
+        match = True
+        lang_range = RE_WILD_STRIP.sub('-', lang_range).lower()
+        ranges = lang_range.split('-')
+        subtags = lang_tag.lower().split('-')
+        length = len(ranges)
+        rindex = 0
+        sindex = 0
+        r = ranges[rindex]
+        s = subtags[sindex]
+
+        # Primary tag needs to match
+        if r != '*' and r != s:
+            match = False
+
+        rindex += 1
+        sindex += 1
+
+        # Match until we run out of ranges
+        while match and rindex < length:
+            r = ranges[rindex]
+            try:
+                s = subtags[sindex]
+            except IndexError:
+                # Ran out of subtags,
+                # but we still have ranges
+                match = False
+                continue
+
+            # Empty range
+            if not r:
+                match = False
+                continue
+
+            # Matched range
+            elif s == r:
+                rindex += 1
+
+            # Implicit wildcard cannot match
+            # singletons
+            elif len(s) == 1:
+                match = False
+                continue
+
+            # Implicitly matched, so grab next subtag
+            sindex += 1
+
+        return match
+
+    def match_attribute_name(self, el, attr, prefix):
+        """Match attribute name and return value if it exists."""
+
+        value = None
+        if self.supports_namespaces():
+            value = None
+            # If we have not defined namespaces, we can't very well find them, so don't bother trying.
+            if prefix:
+                ns = self.namespaces.get(prefix)
+                if ns is None and prefix != '*':
+                    return None
+            else:
+                ns = None
+
+            for k, v in self.iter_attributes(el):
+
+                # Get attribute parts
+                namespace, name = self.split_namespace(el, k)
+
+                # Can't match a prefix attribute as we haven't specified one to match
+                # Try to match it normally as a whole `p:a` as selector may be trying `p\:a`.
+                if ns is None:
+                    if (self.is_xml and attr == k) or (not self.is_xml and util.lower(attr) == util.lower(k)):
+                        value = v
+                        break
+                    # Coverage is not finding this even though it is executed.
+                    # Adding a print statement before this (and erasing coverage) causes coverage to find the line.
+                    # Ignore the false positive message.
+                    continue  # pragma: no cover
+
+                # We can't match our desired prefix attribute as the attribute doesn't have a prefix
+                if namespace is None or ns != namespace and prefix != '*':
+                    continue
+
+                # The attribute doesn't match.
+                if (util.lower(attr) != util.lower(name)) if not self.is_xml else (attr != name):
+                    continue
+
+                value = v
+                break
+        else:
+            for k, v in self.iter_attributes(el):
+                if util.lower(attr) != util.lower(k):
+                    continue
+                value = v
+                break
+        return value
+
+    def match_namespace(self, el, tag):
+        """Match the namespace of the element."""
+
+        match = True
+        namespace = self.get_tag_ns(el)
+        default_namespace = self.namespaces.get('')
+        tag_ns = '' if tag.prefix is None else self.namespaces.get(tag.prefix, None)
+        # We must match the default namespace if one is not provided
+        if tag.prefix is None and (default_namespace is not None and namespace != default_namespace):
+            match = False
+        # If we specified `|tag`, we must not have a namespace.
+        elif (tag.prefix is not None and tag.prefix == '' and namespace):
+            match = False
+        # Verify prefix matches
+        elif (
+            tag.prefix and
+            tag.prefix != '*' and (tag_ns is None or namespace != tag_ns)
+        ):
+            match = False
+        return match
+
+    def match_attributes(self, el, attributes):
+        """Match attributes."""
+
+        match = True
+        if attributes:
+            for a in attributes:
+                value = self.match_attribute_name(el, a.attribute, a.prefix)
+                pattern = a.xml_type_pattern if self.is_xml and a.xml_type_pattern else a.pattern
+                if isinstance(value, list):
+                    value = ' '.join(value)
+                if value is None:
+                    match = False
+                    break
+                elif pattern is None:
+                    continue
+                elif pattern.match(value) is None:
+                    match = False
+                    break
+        return match
+
+    def match_tagname(self, el, tag):
+        """Match tag name."""
+
+        name = (util.lower(tag.name) if not self.is_xml and tag.name is not None else tag.name)
+        return not (
+            name is not None and
+            name not in (self.get_tag(el), '*')
+        )
+
+    def match_tag(self, el, tag):
+        """Match the tag."""
+
+        match = True
+        if tag is not None:
+            # Verify namespace
+            if not self.match_namespace(el, tag):
+                match = False
+            if not self.match_tagname(el, tag):
+                match = False
+        return match
+
+    def match_past_relations(self, el, relation):
+        """Match past relationship."""
+
+        found = False
+        if relation[0].rel_type == REL_PARENT:
+            parent = self.get_parent(el, no_iframe=self.iframe_restrict)
+            while not found and parent:
+                found = self.match_selectors(parent, relation)
+                parent = self.get_parent(parent, no_iframe=self.iframe_restrict)
+        elif relation[0].rel_type == REL_CLOSE_PARENT:
+            parent = self.get_parent(el, no_iframe=self.iframe_restrict)
+            if parent:
+                found = self.match_selectors(parent, relation)
+        elif relation[0].rel_type == REL_SIBLING:
+            sibling = self.get_previous(el)
+            while not found and sibling:
+                found = self.match_selectors(sibling, relation)
+                sibling = self.get_previous(sibling)
+        elif relation[0].rel_type == REL_CLOSE_SIBLING:
+            sibling = self.get_previous(el)
+            if sibling and self.is_tag(sibling):
+                found = self.match_selectors(sibling, relation)
+        return found
+
+    def match_future_child(self, parent, relation, recursive=False):
+        """Match future child."""
+
+        match = False
+        children = self.get_descendants if recursive else self.get_children
+        for child in children(parent, no_iframe=self.iframe_restrict):
+            match = self.match_selectors(child, relation)
+            if match:
+                break
+        return match
+
+    def match_future_relations(self, el, relation):
+        """Match future relationship."""
+
+        found = False
+        if relation[0].rel_type == REL_HAS_PARENT:
+            found = self.match_future_child(el, relation, True)
+        elif relation[0].rel_type == REL_HAS_CLOSE_PARENT:
+            found = self.match_future_child(el, relation)
+        elif relation[0].rel_type == REL_HAS_SIBLING:
+            sibling = self.get_next(el)
+            while not found and sibling:
+                found = self.match_selectors(sibling, relation)
+                sibling = self.get_next(sibling)
+        elif relation[0].rel_type == REL_HAS_CLOSE_SIBLING:
+            sibling = self.get_next(el)
+            if sibling and self.is_tag(sibling):
+                found = self.match_selectors(sibling, relation)
+        return found
+
+    def match_relations(self, el, relation):
+        """Match relationship to other elements."""
+
+        found = False
+
+        if relation[0].rel_type.startswith(':'):
+            found = self.match_future_relations(el, relation)
+        else:
+            found = self.match_past_relations(el, relation)
+
+        return found
+
+    def match_id(self, el, ids):
+        """Match element's ID."""
+
+        found = True
+        for i in ids:
+            if i != self.get_attribute_by_name(el, 'id', ''):
+                found = False
+                break
+        return found
+
+    def match_classes(self, el, classes):
+        """Match element's classes."""
+
+        current_classes = self.get_classes(el)
+        found = True
+        for c in classes:
+            if c not in current_classes:
+                found = False
+                break
+        return found
+
+    def match_root(self, el):
+        """Match element as root."""
+
+        is_root = self.is_root(el)
+        if is_root:
+            sibling = self.get_previous(el, tags=False)
+            while is_root and sibling is not None:
+                if (
+                    self.is_tag(sibling) or (self.is_content_string(sibling) and sibling.strip()) or
+                    self.is_cdata(sibling)
+                ):
+                    is_root = False
+                else:
+                    sibling = self.get_previous(sibling, tags=False)
+        if is_root:
+            sibling = self.get_next(el, tags=False)
+            while is_root and sibling is not None:
+                if (
+                    self.is_tag(sibling) or (self.is_content_string(sibling) and sibling.strip()) or
+                    self.is_cdata(sibling)
+                ):
+                    is_root = False
+                else:
+                    sibling = self.get_next(sibling, tags=False)
+        return is_root
+
+    def match_scope(self, el):
+        """Match element as scope."""
+
+        return self.scope is el
+
+    def match_nth_tag_type(self, el, child):
+        """Match tag type for `nth` matches."""
+
+        return(
+            (self.get_tag(child) == self.get_tag(el)) and
+            (self.get_tag_ns(child) == self.get_tag_ns(el))
+        )
+
+    def match_nth(self, el, nth):
+        """Match `nth` elements."""
+
+        matched = True
+
+        for n in nth:
+            matched = False
+            if n.selectors and not self.match_selectors(el, n.selectors):
+                break
+            parent = self.get_parent(el)
+            if parent is None:
+                parent = self.create_fake_parent(el)
+            last = n.last
+            last_index = len(parent) - 1
+            index = last_index if last else 0
+            relative_index = 0
+            a = n.a
+            b = n.b
+            var = n.n
+            count = 0
+            count_incr = 1
+            factor = -1 if last else 1
+            idx = last_idx = a * count + b if var else a
+
+            # We can only adjust bounds within a variable index
+            if var:
+                # Abort if our nth index is out of bounds and only getting further out of bounds as we increment.
+                # Otherwise, increment to try to get in bounds.
+                adjust = None
+                while idx < 1 or idx > last_index:
+                    if idx < 0:
+                        diff_low = 0 - idx
+                        if adjust is not None and adjust == 1:
+                            break
+                        adjust = -1
+                        count += count_incr
+                        idx = last_idx = a * count + b if var else a
+                        diff = 0 - idx
+                        if diff >= diff_low:
+                            break
+                    else:
+                        diff_high = idx - last_index
+                        if adjust is not None and adjust == -1:
+                            break
+                        adjust = 1
+                        count += count_incr
+                        idx = last_idx = a * count + b if var else a
+                        diff = idx - last_index
+                        if diff >= diff_high:
+                            break
+                        diff_high = diff
+
+                # If a < 0, our count is working backwards, so floor the index by increasing the count.
+                # Find the count that yields the lowest, in bound value and use that.
+                # Lastly reverse count increment so that we'll increase our index.
+                lowest = count
+                if a < 0:
+                    while idx >= 1:
+                        lowest = count
+                        count += count_incr
+                        idx = last_idx = a * count + b if var else a
+                    count_incr = -1
+                count = lowest
+                idx = last_idx = a * count + b if var else a
+
+            # Evaluate elements while our calculated nth index is still in range
+            while 1 <= idx <= last_index + 1:
+                child = None
+                # Evaluate while our child index is still in range.
+                for child in self.get_children(parent, start=index, reverse=factor < 0, tags=False):
+                    index += factor
+                    if not self.is_tag(child):
+                        continue
+                    # Handle `of S` in `nth-child`
+                    if n.selectors and not self.match_selectors(child, n.selectors):
+                        continue
+                    # Handle `of-type`
+                    if n.of_type and not self.match_nth_tag_type(el, child):
+                        continue
+                    relative_index += 1
+                    if relative_index == idx:
+                        if child is el:
+                            matched = True
+                        else:
+                            break
+                    if child is el:
+                        break
+                if child is el:
+                    break
+                last_idx = idx
+                count += count_incr
+                if count < 0:
+                    # Count is counting down and has now ventured into invalid territory.
+                    break
+                idx = a * count + b if var else a
+                if last_idx == idx:
+                    break
+            if not matched:
+                break
+        return matched
+
+    def match_empty(self, el):
+        """Check if element is empty (if requested)."""
+
+        is_empty = True
+        for child in self.get_children(el, tags=False):
+            if self.is_tag(child):
+                is_empty = False
+                break
+            elif self.is_content_string(child) and RE_NOT_EMPTY.search(child):
+                is_empty = False
+                break
+        return is_empty
+
+    def match_subselectors(self, el, selectors):
+        """Match selectors."""
+
+        match = True
+        for sel in selectors:
+            if not self.match_selectors(el, sel):
+                match = False
+        return match
+
+    def match_contains(self, el, contains):
+        """Match element if it contains text."""
+
+        match = True
+        content = None
+        for contain_list in contains:
+            if content is None:
+                content = self.get_text(el, no_iframe=self.is_html)
+            found = False
+            for text in contain_list.text:
+                if text in content:
+                    found = True
+                    break
+            if not found:
+                match = False
+        return match
+
+    def match_default(self, el):
+        """Match default."""
+
+        match = False
+
+        # Find this input's form
+        form = None
+        parent = self.get_parent(el, no_iframe=True)
+        while parent and form is None:
+            if self.get_tag(parent) == 'form' and self.is_html_tag(parent):
+                form = parent
+            else:
+                parent = self.get_parent(parent, no_iframe=True)
+
+        # Look in form cache to see if we've already located its default button
+        found_form = False
+        for f, t in self.cached_default_forms:
+            if f is form:
+                found_form = True
+                if t is el:
+                    match = True
+                break
+
+        # We didn't have the form cached, so look for its default button
+        if not found_form:
+            for child in self.get_descendants(form, no_iframe=True):
+                name = self.get_tag(child)
+                # Can't do nested forms (haven't figured out why we never hit this)
+                if name == 'form':  # pragma: no cover
+                    break
+                if name in ('input', 'button'):
+                    v = self.get_attribute_by_name(child, 'type', '')
+                    if v and util.lower(v) == 'submit':
+                        self.cached_default_forms.append([form, child])
+                        if el is child:
+                            match = True
+                        break
+        return match
+
+    def match_indeterminate(self, el):
+        """Match default."""
+
+        match = False
+        name = self.get_attribute_by_name(el, 'name')
+
+        def get_parent_form(el):
+            """Find this input's form."""
+            form = None
+            parent = self.get_parent(el, no_iframe=True)
+            while form is None:
+                if self.get_tag(parent) == 'form' and self.is_html_tag(parent):
+                    form = parent
+                    break
+                last_parent = parent
+                parent = self.get_parent(parent, no_iframe=True)
+                if parent is None:
+                    form = last_parent
+                    break
+            return form
+
+        form = get_parent_form(el)
+
+        # Look in form cache to see if we've already evaluated that its fellow radio buttons are indeterminate
+        found_form = False
+        for f, n, i in self.cached_indeterminate_forms:
+            if f is form and n == name:
+                found_form = True
+                if i is True:
+                    match = True
+                break
+
+        # We didn't have the form cached, so validate that the radio button is indeterminate
+        if not found_form:
+            checked = False
+            for child in self.get_descendants(form, no_iframe=True):
+                if child is el:
+                    continue
+                tag_name = self.get_tag(child)
+                if tag_name == 'input':
+                    is_radio = False
+                    check = False
+                    has_name = False
+                    for k, v in self.iter_attributes(child):
+                        if util.lower(k) == 'type' and util.lower(v) == 'radio':
+                            is_radio = True
+                        elif util.lower(k) == 'name' and v == name:
+                            has_name = True
+                        elif util.lower(k) == 'checked':
+                            check = True
+                        if is_radio and check and has_name and get_parent_form(child) is form:
+                            checked = True
+                            break
+                if checked:
+                    break
+            if not checked:
+                match = True
+            self.cached_indeterminate_forms.append([form, name, match])
+
+        return match
+
+    def match_lang(self, el, langs):
+        """Match languages."""
+
+        match = False
+        has_ns = self.supports_namespaces()
+        root = self.root
+        has_html_namespace = self.has_html_namespace
+
+        # Walk parents looking for `lang` (HTML) or `xml:lang` XML property.
+        parent = el
+        found_lang = None
+        last = None
+        while not found_lang:
+            has_html_ns = self.has_html_ns(parent)
+            for k, v in self.iter_attributes(parent):
+                attr_ns, attr = self.split_namespace(parent, k)
+                if (
+                    ((not has_ns or has_html_ns) and (util.lower(k) if not self.is_xml else k) == 'lang') or
+                    (
+                        has_ns and not has_html_ns and attr_ns == NS_XML and
+                        (util.lower(attr) if not self.is_xml and attr is not None else attr) == 'lang'
+                    )
+                ):
+                    found_lang = v
+                    break
+            last = parent
+            parent = self.get_parent(parent, no_iframe=self.is_html)
+
+            if parent is None:
+                root = last
+                has_html_namespace = self.has_html_ns(root)
+                parent = last
+                break
+
+        # Use cached meta language.
+        if not found_lang and self.cached_meta_lang:
+            for cache in self.cached_meta_lang:
+                if root is cache[0]:
+                    found_lang = cache[1]
+
+        # If we couldn't find a language, and the document is HTML, look to meta to determine language.
+        if found_lang is None and (not self.is_xml or (has_html_namespace and root.name == 'html')):
+            # Find head
+            found = False
+            for tag in ('html', 'head'):
+                found = False
+                for child in self.get_children(parent, no_iframe=self.is_html):
+                    if self.get_tag(child) == tag and self.is_html_tag(child):
+                        found = True
+                        parent = child
+                        break
+                if not found:  # pragma: no cover
+                    break
+
+            # Search meta tags
+            if found:
+                for child in parent:
+                    if self.is_tag(child) and self.get_tag(child) == 'meta' and self.is_html_tag(parent):
+                        c_lang = False
+                        content = None
+                        for k, v in self.iter_attributes(child):
+                            if util.lower(k) == 'http-equiv' and util.lower(v) == 'content-language':
+                                c_lang = True
+                            if util.lower(k) == 'content':
+                                content = v
+                            if c_lang and content:
+                                found_lang = content
+                                self.cached_meta_lang.append((root, found_lang))
+                                break
+                    if found_lang:
+                        break
+                if not found_lang:
+                    self.cached_meta_lang.append((root, False))
+
+        # If we determined a language, compare.
+        if found_lang:
+            for patterns in langs:
+                match = False
+                for pattern in patterns:
+                    if self.extended_language_filter(pattern, found_lang):
+                        match = True
+                if not match:
+                    break
+
+        return match
+
+    def match_dir(self, el, directionality):
+        """Check directionality."""
+
+        # If we have to match both left and right, we can't match either.
+        if directionality & ct.SEL_DIR_LTR and directionality & ct.SEL_DIR_RTL:
+            return False
+
+        if el is None or not self.is_html_tag(el):
+            return False
+
+        # Element has defined direction of left to right or right to left
+        direction = DIR_MAP.get(util.lower(self.get_attribute_by_name(el, 'dir', '')), None)
+        if direction not in (None, 0):
+            return direction == directionality
+
+        # Element is the document element (the root) and no direction assigned, assume left to right.
+        is_root = self.is_root(el)
+        if is_root and direction is None:
+            return ct.SEL_DIR_LTR == directionality
+
+        # If `input[type=telephone]` and no direction is assigned, assume left to right.
+        name = self.get_tag(el)
+        is_input = name == 'input'
+        is_textarea = name == 'textarea'
+        is_bdi = name == 'bdi'
+        itype = util.lower(self.get_attribute_by_name(el, 'type', '')) if is_input else ''
+        if is_input and itype == 'tel' and direction is None:
+            return ct.SEL_DIR_LTR == directionality
+
+        # Auto handling for text inputs
+        if ((is_input and itype in ('text', 'search', 'tel', 'url', 'email')) or is_textarea) and direction == 0:
+            if is_textarea:
+                value = []
+                for node in self.get_contents(el, no_iframe=True):
+                    if self.is_content_string(node):
+                        value.append(node)
+                value = ''.join(value)
+            else:
+                value = self.get_attribute_by_name(el, 'value', '')
+            if value:
+                for c in value:
+                    bidi = unicodedata.bidirectional(c)
+                    if bidi in ('AL', 'R', 'L'):
+                        direction = ct.SEL_DIR_LTR if bidi == 'L' else ct.SEL_DIR_RTL
+                        return direction == directionality
+                # Assume left to right
+                return ct.SEL_DIR_LTR == directionality
+            elif is_root:
+                return ct.SEL_DIR_LTR == directionality
+            return self.match_dir(self.get_parent(el, no_iframe=True), directionality)
+
+        # Auto handling for `bdi` and other non text inputs.
+        if (is_bdi and direction is None) or direction == 0:
+            direction = self.find_bidi(el)
+            if direction is not None:
+                return direction == directionality
+            elif is_root:
+                return ct.SEL_DIR_LTR == directionality
+            return self.match_dir(self.get_parent(el, no_iframe=True), directionality)
+
+        # Match parents direction
+        return self.match_dir(self.get_parent(el, no_iframe=True), directionality)
+
+    def match_range(self, el, condition):
+        """
+        Match range.
+
+        Behavior is modeled after what we see in browsers. Browsers seem to evaluate
+        if the value is out of range, and if not, it is in range. So a missing value
+        will not evaluate out of range; therefore, value is in range. Personally, I
+        feel like this should evaluate as neither in or out of range.
+        """
+
+        out_of_range = False
+
+        itype = util.lower(self.get_attribute_by_name(el, 'type'))
+        mn = self.get_attribute_by_name(el, 'min', None)
+        if mn is not None:
+            mn = Inputs.parse_value(itype, mn)
+        mx = self.get_attribute_by_name(el, 'max', None)
+        if mx is not None:
+            mx = Inputs.parse_value(itype, mx)
+
+        # There is no valid min or max, so we cannot evaluate a range
+        if mn is None and mx is None:
+            return False
+
+        value = self.get_attribute_by_name(el, 'value', None)
+        if value is not None:
+            value = Inputs.parse_value(itype, value)
+        if value is not None:
+            if itype in ("date", "datetime-local", "month", "week", "number", "range"):
+                if mn is not None and value < mn:
+                    out_of_range = True
+                if not out_of_range and mx is not None and value > mx:
+                    out_of_range = True
+            elif itype == "time":
+                if mn is not None and mx is not None and mn > mx:
+                    # Time is periodic, so this is a reversed/discontinuous range
+                    if value < mn and value > mx:
+                        out_of_range = True
+                else:
+                    if mn is not None and value < mn:
+                        out_of_range = True
+                    if not out_of_range and mx is not None and value > mx:
+                        out_of_range = True
+
+        return not out_of_range if condition & ct.SEL_IN_RANGE else out_of_range
+
+    def match_defined(self, el):
+        """
+        Match defined.
+
+        `:defined` is related to custom elements in a browser.
+
+        - If the document is XML (not XHTML), all tags will match.
+        - Tags that are not custom (don't have a hyphen) are marked defined.
+        - If the tag has a prefix (without or without a namespace), it will not match.
+
+        This is of course requires the parser to provide us with the proper prefix and namespace info,
+        if it doesn't, there is nothing we can do.
+        """
+
+        name = self.get_tag(el)
+        return (
+            name.find('-') == -1 or
+            name.find(':') != -1 or
+            self.get_prefix(el) is not None
+        )
+
+    def match_placeholder_shown(self, el):
+        """
+        Match placeholder shown according to HTML spec.
+
+        - text area should be checked if they have content. A single newline does not count as content.
+
+        """
+
+        match = False
+        content = self.get_text(el)
+        if content in ('', '\n'):
+            match = True
+
+        return match
+
+    def match_selectors(self, el, selectors):
+        """Check if element matches one of the selectors."""
+
+        match = False
+        is_not = selectors.is_not
+        is_html = selectors.is_html
+
+        # Internal selector lists that use the HTML flag, will automatically get the `html` namespace.
+        if is_html:
+            namespaces = self.namespaces
+            iframe_restrict = self.iframe_restrict
+            self.namespaces = {'html': NS_XHTML}
+            self.iframe_restrict = True
+
+        if not is_html or self.is_html:
+            for selector in selectors:
+                match = is_not
+                # We have a un-matchable situation (like `:focus` as you can focus an element in this environment)
+                if isinstance(selector, ct.SelectorNull):
+                    continue
+                # Verify tag matches
+                if not self.match_tag(el, selector.tag):
+                    continue
+                # Verify tag is defined
+                if selector.flags & ct.SEL_DEFINED and not self.match_defined(el):
+                    continue
+                # Verify element is root
+                if selector.flags & ct.SEL_ROOT and not self.match_root(el):
+                    continue
+                # Verify element is scope
+                if selector.flags & ct.SEL_SCOPE and not self.match_scope(el):
+                    continue
+                # Verify element has placeholder shown
+                if selector.flags & ct.SEL_PLACEHOLDER_SHOWN and not self.match_placeholder_shown(el):
+                    continue
+                # Verify `nth` matches
+                if not self.match_nth(el, selector.nth):
+                    continue
+                if selector.flags & ct.SEL_EMPTY and not self.match_empty(el):
+                    continue
+                # Verify id matches
+                if selector.ids and not self.match_id(el, selector.ids):
+                    continue
+                # Verify classes match
+                if selector.classes and not self.match_classes(el, selector.classes):
+                    continue
+                # Verify attribute(s) match
+                if not self.match_attributes(el, selector.attributes):
+                    continue
+                # Verify ranges
+                if selector.flags & RANGES and not self.match_range(el, selector.flags & RANGES):
+                    continue
+                # Verify language patterns
+                if selector.lang and not self.match_lang(el, selector.lang):
+                    continue
+                # Verify pseudo selector patterns
+                if selector.selectors and not self.match_subselectors(el, selector.selectors):
+                    continue
+                # Verify relationship selectors
+                if selector.relation and not self.match_relations(el, selector.relation):
+                    continue
+                # Validate that the current default selector match corresponds to the first submit button in the form
+                if selector.flags & ct.SEL_DEFAULT and not self.match_default(el):
+                    continue
+                # Validate that the unset radio button is among radio buttons with the same name in a form that are
+                # also not set.
+                if selector.flags & ct.SEL_INDETERMINATE and not self.match_indeterminate(el):
+                    continue
+                # Validate element directionality
+                if selector.flags & DIR_FLAGS and not self.match_dir(el, selector.flags & DIR_FLAGS):
+                    continue
+                # Validate that the tag contains the specified text.
+                if not self.match_contains(el, selector.contains):
+                    continue
+                match = not is_not
+                break
+
+        # Restore actual namespaces being used for external selector lists
+        if is_html:
+            self.namespaces = namespaces
+            self.iframe_restrict = iframe_restrict
+
+        return match
+
+    def select(self, limit=0):
+        """Match all tags under the targeted tag."""
+
+        if limit < 1:
+            limit = None
+
+        for child in self.get_descendants(self.tag):
+            if self.match(child):
+                yield child
+                if limit is not None:
+                    limit -= 1
+                    if limit < 1:
+                        break
+
+    def closest(self):
+        """Match closest ancestor."""
+
+        current = self.tag
+        closest = None
+        while closest is None and current is not None:
+            if self.match(current):
+                closest = current
+            else:
+                current = self.get_parent(current)
+        return closest
+
+    def filter(self):  # noqa A001
+        """Filter tag's children."""
+
+        return [tag for tag in self.get_contents(self.tag) if not self.is_navigable_string(tag) and self.match(tag)]
+
+    def match(self, el):
+        """Match."""
+
+        return not self.is_doc(el) and self.is_tag(el) and self.match_selectors(el, self.selectors)
+
+
+class CSSMatch(_DocumentNav, _Match):
+    """The Beautiful Soup CSS match class."""
+
+
+class CommentsMatch(_DocumentNav):
+    """Comments matcher."""
+
+    def __init__(self, el):
+        """Initialize."""
+
+        self.assert_valid_input(el)
+        self.tag = el
+
+    def get_comments(self, limit=0):
+        """Get comments."""
+
+        if limit < 1:
+            limit = None
+
+        for child in self.get_descendants(self.tag, tags=False):
+            if self.is_comment(child):
+                yield child
+                if limit is not None:
+                    limit -= 1
+                    if limit < 1:
+                        break
+
+
+class SoupSieve(ct.Immutable):
+    """Compiled Soup Sieve selector matching object."""
+
+    __slots__ = ("pattern", "selectors", "namespaces", "custom", "flags", "_hash")
+
+    def __init__(self, pattern, selectors, namespaces, custom, flags):
+        """Initialize."""
+
+        super(SoupSieve, self).__init__(
+            pattern=pattern,
+            selectors=selectors,
+            namespaces=namespaces,
+            custom=custom,
+            flags=flags
+        )
+
+    def match(self, tag):
+        """Match."""
+
+        return CSSMatch(self.selectors, tag, self.namespaces, self.flags).match(tag)
+
+    def closest(self, tag):
+        """Match closest ancestor."""
+
+        return CSSMatch(self.selectors, tag, self.namespaces, self.flags).closest()
+
+    def filter(self, iterable):  # noqa A001
+        """
+        Filter.
+
+        `CSSMatch` can cache certain searches for tags of the same document,
+        so if we are given a tag, all tags are from the same document,
+        and we can take advantage of the optimization.
+
+        Any other kind of iterable could have tags from different documents or detached tags,
+        so for those, we use a new `CSSMatch` for each item in the iterable.
+        """
+
+        if CSSMatch.is_tag(iterable):
+            return CSSMatch(self.selectors, iterable, self.namespaces, self.flags).filter()
+        else:
+            return [node for node in iterable if not CSSMatch.is_navigable_string(node) and self.match(node)]
+
+    @util.deprecated("'comments' is not related to CSS selectors and will be removed in the future.")
+    def comments(self, tag, limit=0):
+        """Get comments only."""
+
+        return [comment for comment in CommentsMatch(tag).get_comments(limit)]
+
+    @util.deprecated("'icomments' is not related to CSS selectors and will be removed in the future.")
+    def icomments(self, tag, limit=0):
+        """Iterate comments only."""
+
+        for comment in CommentsMatch(tag).get_comments(limit):
+            yield comment
+
+    def select_one(self, tag):
+        """Select a single tag."""
+
+        tags = self.select(tag, limit=1)
+        return tags[0] if tags else None
+
+    def select(self, tag, limit=0):
+        """Select the specified tags."""
+
+        return list(self.iselect(tag, limit))
+
+    def iselect(self, tag, limit=0):
+        """Iterate the specified tags."""
+
+        for el in CSSMatch(self.selectors, tag, self.namespaces, self.flags).select(limit):
+            yield el
+
+    def __repr__(self):  # pragma: no cover
+        """Representation."""
+
+        return "SoupSieve(pattern={!r}, namespaces={!r}, custom={!r}, flags={!r})".format(
+            self.pattern,
+            self.namespaces,
+            self.custom,
+            self.flags
+        )
+
+    __str__ = __repr__
+
+
+ct.pickle_register(SoupSieve)
diff --git a/lib/soupsieve/css_parser.py b/lib/soupsieve/css_parser.py
new file mode 100644
index 00000000..e7c8833b
--- /dev/null
+++ b/lib/soupsieve/css_parser.py
@@ -0,0 +1,1213 @@
+"""CSS selector parser."""
+from __future__ import unicode_literals
+import re
+from . import util
+from . import css_match as cm
+from . import css_types as ct
+from .util import SelectorSyntaxError
+
+UNICODE_REPLACEMENT_CHAR = 0xFFFD
+
+# Simple pseudo classes that take no parameters
+PSEUDO_SIMPLE = {
+    ":any-link",
+    ":empty",
+    ":first-child",
+    ":first-of-type",
+    ":in-range",
+    ":out-of-range",
+    ":last-child",
+    ":last-of-type",
+    ":link",
+    ":only-child",
+    ":only-of-type",
+    ":root",
+    ':checked',
+    ':default',
+    ':disabled',
+    ':enabled',
+    ':indeterminate',
+    ':optional',
+    ':placeholder-shown',
+    ':read-only',
+    ':read-write',
+    ':required',
+    ':scope',
+    ':defined'
+}
+
+# Supported, simple pseudo classes that match nothing in the Soup Sieve environment
+PSEUDO_SIMPLE_NO_MATCH = {
+    ':active',
+    ':current',
+    ':focus',
+    ':focus-visible',
+    ':focus-within',
+    ':future',
+    ':host',
+    ':hover',
+    ':local-link',
+    ':past',
+    ':paused',
+    ':playing',
+    ':target',
+    ':target-within',
+    ':user-invalid',
+    ':visited'
+}
+
+# Complex pseudo classes that take selector lists
+PSEUDO_COMPLEX = {
+    ':contains',
+    ':has',
+    ':is',
+    ':matches',
+    ':not',
+    ':where'
+}
+
+PSEUDO_COMPLEX_NO_MATCH = {
+    ':current',
+    ':host',
+    ':host-context'
+}
+
+# Complex pseudo classes that take very specific parameters and are handled special
+PSEUDO_SPECIAL = {
+    ':dir',
+    ':lang',
+    ':nth-child',
+    ':nth-last-child',
+    ':nth-last-of-type',
+    ':nth-of-type'
+}
+
+PSEUDO_SUPPORTED = PSEUDO_SIMPLE | PSEUDO_SIMPLE_NO_MATCH | PSEUDO_COMPLEX | PSEUDO_COMPLEX_NO_MATCH | PSEUDO_SPECIAL
+
+# Sub-patterns parts
+# Whitespace
+NEWLINE = r'(?:\r\n|(?!\r\n)[\n\f\r])'
+WS = r'(?:[ \t]|{})'.format(NEWLINE)
+# Comments
+COMMENTS = r'(?:/\*[^*]*\*+(?:[^/*][^*]*\*+)*/)'
+# Whitespace with comments included
+WSC = r'(?:{ws}|{comments})'.format(ws=WS, comments=COMMENTS)
+# CSS escapes
+CSS_ESCAPES = r'(?:\\(?:[a-f0-9]{{1,6}}{ws}?|[^\r\n\f]|$))'.format(ws=WS)
+CSS_STRING_ESCAPES = r'(?:\\(?:[a-f0-9]{{1,6}}{ws}?|[^\r\n\f]|$|{nl}))'.format(ws=WS, nl=NEWLINE)
+# CSS Identifier
+IDENTIFIER = r'''
+(?:(?:-?(?:[^\x00-\x2f\x30-\x40\x5B-\x5E\x60\x7B-\x9f]|{esc})+|--)
+(?:[^\x00-\x2c\x2e\x2f\x3A-\x40\x5B-\x5E\x60\x7B-\x9f]|{esc})*)
+'''.format(esc=CSS_ESCAPES)
+# `nth` content
+NTH = r'(?:[-+])?(?:[0-9]+n?|n)(?:(?<=n){ws}*(?:[-+]){ws}*(?:[0-9]+))?'.format(ws=WSC)
+# Value: quoted string or identifier
+VALUE = r'''
+(?:"(?:\\(?:.|{nl})|[^\\"\r\n\f]+)*?"|'(?:\\(?:.|{nl})|[^\\'\r\n\f]+)*?'|{ident}+)
+'''.format(nl=NEWLINE, ident=IDENTIFIER)
+# Attribute value comparison. `!=` is handled special as it is non-standard.
+ATTR = r'''
+(?:{ws}*(?P[!~^|*$]?=){ws}*(?P{value})(?:{ws}+(?P[is]))?)?{ws}*\]
+'''.format(ws=WSC, value=VALUE)
+
+# Selector patterns
+# IDs (`#id`)
+PAT_ID = r'\#{ident}'.format(ident=IDENTIFIER)
+# Classes (`.class`)
+PAT_CLASS = r'\.{ident}'.format(ident=IDENTIFIER)
+# Prefix:Tag (`prefix|tag`)
+PAT_TAG = r'(?:(?:{ident}|\*)?\|)?(?:{ident}|\*)'.format(ident=IDENTIFIER)
+# Attributes (`[attr]`, `[attr=value]`, etc.)
+PAT_ATTR = r'\[{ws}*(?P(?:(?:{ident}|\*)?\|)?{ident}){attr}'.format(ws=WSC, ident=IDENTIFIER, attr=ATTR)
+# Pseudo class (`:pseudo-class`, `:pseudo-class(`)
+PAT_PSEUDO_CLASS = r'(?P:{ident})(?P\({ws}*)?'.format(ws=WSC, ident=IDENTIFIER)
+# Pseudo class special patterns. Matches `:pseudo-class(` for special case pseudo classes.
+PAT_PSEUDO_CLASS_SPECIAL = r'(?P:{ident})(?P\({ws}*)'.format(ws=WSC, ident=IDENTIFIER)
+# Custom pseudo class (`:--custom-pseudo`)
+PAT_PSEUDO_CLASS_CUSTOM = r'(?P:(?=--){ident})'.format(ident=IDENTIFIER)
+# Closing pseudo group (`)`)
+PAT_PSEUDO_CLOSE = r'{ws}*\)'.format(ws=WSC)
+# Pseudo element (`::pseudo-element`)
+PAT_PSEUDO_ELEMENT = r':{}'.format(PAT_PSEUDO_CLASS)
+# At rule (`@page`, etc.) (not supported)
+PAT_AT_RULE = r'@P{ident}'.format(ident=IDENTIFIER)
+# Pseudo class `nth-child` (`:nth-child(an+b [of S]?)`, `:first-child`, etc.)
+PAT_PSEUDO_NTH_CHILD = r'''
+(?P{name}
+(?P{nth}|even|odd))(?:{wsc}*\)|(?P{comments}*{ws}{wsc}*of{comments}*{ws}{wsc}*))
+'''.format(name=PAT_PSEUDO_CLASS_SPECIAL, wsc=WSC, comments=COMMENTS, ws=WS, nth=NTH)
+# Pseudo class `nth-of-type` (`:nth-of-type(an+b)`, `:first-of-type`, etc.)
+PAT_PSEUDO_NTH_TYPE = r'''
+(?P{name}
+(?P{nth}|even|odd)){ws}*\)
+'''.format(name=PAT_PSEUDO_CLASS_SPECIAL, ws=WSC, nth=NTH)
+# Pseudo class language (`:lang("*-de", en)`)
+PAT_PSEUDO_LANG = r'{name}(?P{value}(?:{ws}*,{ws}*{value})*){ws}*\)'.format(
+    name=PAT_PSEUDO_CLASS_SPECIAL, ws=WSC, value=VALUE
+)
+# Pseudo class direction (`:dir(ltr)`)
+PAT_PSEUDO_DIR = r'{name}(?Pltr|rtl){ws}*\)'.format(name=PAT_PSEUDO_CLASS_SPECIAL, ws=WSC)
+# Combining characters (`>`, `~`, ` `, `+`, `,`)
+PAT_COMBINE = r'{wsc}*?(?P[,+>~]|{ws}(?![,+>~])){wsc}*'.format(ws=WS, wsc=WSC)
+# Extra: Contains (`:contains(text)`)
+PAT_PSEUDO_CONTAINS = r'{name}(?P{value}(?:{ws}*,{ws}*{value})*){ws}*\)'.format(
+    name=PAT_PSEUDO_CLASS_SPECIAL, ws=WSC, value=VALUE
+)
+
+# Regular expressions
+# CSS escape pattern
+RE_CSS_ESC = re.compile(r'(?:(\\[a-f0-9]{{1,6}}{ws}?)|(\\[^\r\n\f])|(\\$))'.format(ws=WSC), re.I)
+RE_CSS_STR_ESC = re.compile(
+    r'(?:(\\[a-f0-9]{{1,6}}{ws}?)|(\\[^\r\n\f])|(\\$)|(\\{nl}))'.format(ws=WS, nl=NEWLINE), re.I
+)
+# Pattern to break up `nth` specifiers
+RE_NTH = re.compile(
+    r'(?P[-+])?(?P[0-9]+n?|n)(?:(?<=n){ws}*(?P[-+]){ws}*(?P[0-9]+))?'.format(ws=WSC),
+    re.I
+)
+# Pattern to iterate multiple values.
+RE_VALUES = re.compile(r'(?:(?P{value})|(?P{ws}*,{ws}*))'.format(ws=WSC, value=VALUE), re.X)
+# Whitespace checks
+RE_WS = re.compile(WS)
+RE_WS_BEGIN = re.compile('^{}*'.format(WSC))
+RE_WS_END = re.compile('{}*$'.format(WSC))
+RE_CUSTOM = re.compile(r'^{}$'.format(PAT_PSEUDO_CLASS_CUSTOM), re.X)
+
+# Constants
+# List split token
+COMMA_COMBINATOR = ','
+# Relation token for descendant
+WS_COMBINATOR = " "
+
+# Parse flags
+FLG_PSEUDO = 0x01
+FLG_NOT = 0x02
+FLG_RELATIVE = 0x04
+FLG_DEFAULT = 0x08
+FLG_HTML = 0x10
+FLG_INDETERMINATE = 0x20
+FLG_OPEN = 0x40
+FLG_IN_RANGE = 0x80
+FLG_OUT_OF_RANGE = 0x100
+FLG_PLACEHOLDER_SHOWN = 0x200
+
+# Maximum cached patterns to store
+_MAXCACHE = 500
+
+
+@util.lru_cache(maxsize=_MAXCACHE)
+def _cached_css_compile(pattern, namespaces, custom, flags):
+    """Cached CSS compile."""
+
+    custom_selectors = process_custom(custom)
+    return cm.SoupSieve(
+        pattern,
+        CSSParser(pattern, custom=custom_selectors, flags=flags).process_selectors(),
+        namespaces,
+        custom,
+        flags
+    )
+
+
+def _purge_cache():
+    """Purge the cache."""
+
+    _cached_css_compile.cache_clear()
+
+
+def process_custom(custom):
+    """Process custom."""
+
+    custom_selectors = {}
+    if custom is not None:
+        for key, value in custom.items():
+            name = util.lower(key)
+            if RE_CUSTOM.match(name) is None:
+                raise SelectorSyntaxError("The name '{}' is not a valid custom pseudo-class name".format(name))
+            if name in custom_selectors:
+                raise KeyError("The custom selector '{}' has already been registered".format(name))
+            custom_selectors[css_unescape(name)] = value
+    return custom_selectors
+
+
+def css_unescape(content, string=False):
+    """
+    Unescape CSS value.
+
+    Strings allow for spanning the value on multiple strings by escaping a new line.
+    """
+
+    def replace(m):
+        """Replace with the appropriate substitute."""
+
+        if m.group(1):
+            codepoint = int(m.group(1)[1:], 16)
+            if codepoint == 0:
+                codepoint = UNICODE_REPLACEMENT_CHAR
+            value = util.uchr(codepoint)
+        elif m.group(2):
+            value = m.group(2)[1:]
+        elif m.group(3):
+            value = '\ufffd'
+        else:
+            value = ''
+
+        return value
+
+    return (RE_CSS_ESC if not string else RE_CSS_STR_ESC).sub(replace, content)
+
+
+def escape(ident):
+    """Escape identifier."""
+
+    string = []
+    length = len(ident)
+    start_dash = length > 0 and ident[0] == '-'
+    if length == 1 and start_dash:
+        # Need to escape identifier that is a single `-` with no other characters
+        string.append('\\{}'.format(ident))
+    else:
+        for index, c in enumerate(ident):
+            codepoint = util.uord(c)
+            if codepoint == 0x00:
+                string.append('\ufffd')
+            elif (0x01 <= codepoint <= 0x1F) or codepoint == 0x7F:
+                string.append('\\{:x} '.format(codepoint))
+            elif (index == 0 or (start_dash and index == 1)) and (0x30 <= codepoint <= 0x39):
+                string.append('\\{:x} '.format(codepoint))
+            elif (
+                codepoint in (0x2D, 0x5F) or codepoint >= 0x80 or (0x30 <= codepoint <= 0x39) or
+                (0x30 <= codepoint <= 0x39) or (0x41 <= codepoint <= 0x5A) or (0x61 <= codepoint <= 0x7A)
+            ):
+                string.append(c)
+            else:
+                string.append('\\{}'.format(c))
+    return ''.join(string)
+
+
+class SelectorPattern(object):
+    """Selector pattern."""
+
+    def __init__(self, name, pattern):
+        """Initialize."""
+
+        self.name = name
+        self.re_pattern = re.compile(pattern, re.I | re.X | re.U)
+
+    def get_name(self):
+        """Get name."""
+
+        return self.name
+
+    def enabled(self, flags):
+        """Enabled."""
+
+        return True
+
+    def match(self, selector, index):
+        """Match the selector."""
+
+        return self.re_pattern.match(selector, index)
+
+
+class SpecialPseudoPattern(SelectorPattern):
+    """Selector pattern."""
+
+    def __init__(self, patterns):
+        """Initialize."""
+
+        self.patterns = {}
+        for p in patterns:
+            name = p[0]
+            pattern = SelectorPattern(name, p[2])
+            for pseudo in p[1]:
+                self.patterns[pseudo] = pattern
+
+        self.matched_name = None
+        self.re_pseudo_name = re.compile(PAT_PSEUDO_CLASS_SPECIAL, re.I | re.X | re.U)
+
+    def get_name(self):
+        """Get name."""
+
+        return self.matched_name.get_name()
+
+    def enabled(self, flags):
+        """Enabled."""
+
+        return True
+
+    def match(self, selector, index):
+        """Match the selector."""
+
+        pseudo = None
+        m = self.re_pseudo_name.match(selector, index)
+        if m:
+            name = util.lower(css_unescape(m.group('name')))
+            pattern = self.patterns.get(name)
+            if pattern:
+                pseudo = pattern.match(selector, index)
+                if pseudo:
+                    self.matched_name = pattern
+
+        return pseudo
+
+
+class _Selector(object):
+    """
+    Intermediate selector class.
+
+    This stores selector data for a compound selector as we are acquiring them.
+    Once we are done collecting the data for a compound selector, we freeze
+    the data in an object that can be pickled and hashed.
+    """
+
+    def __init__(self, **kwargs):
+        """Initialize."""
+
+        self.tag = kwargs.get('tag', None)
+        self.ids = kwargs.get('ids', [])
+        self.classes = kwargs.get('classes', [])
+        self.attributes = kwargs.get('attributes', [])
+        self.nth = kwargs.get('nth', [])
+        self.selectors = kwargs.get('selectors', [])
+        self.relations = kwargs.get('relations', [])
+        self.rel_type = kwargs.get('rel_type', None)
+        self.contains = kwargs.get('contains', [])
+        self.lang = kwargs.get('lang', [])
+        self.flags = kwargs.get('flags', 0)
+        self.no_match = kwargs.get('no_match', False)
+
+    def _freeze_relations(self, relations):
+        """Freeze relation."""
+
+        if relations:
+            sel = relations[0]
+            sel.relations.extend(relations[1:])
+            return ct.SelectorList([sel.freeze()])
+        else:
+            return ct.SelectorList()
+
+    def freeze(self):
+        """Freeze self."""
+
+        if self.no_match:
+            return ct.SelectorNull()
+        else:
+            return ct.Selector(
+                self.tag,
+                tuple(self.ids),
+                tuple(self.classes),
+                tuple(self.attributes),
+                tuple(self.nth),
+                tuple(self.selectors),
+                self._freeze_relations(self.relations),
+                self.rel_type,
+                tuple(self.contains),
+                tuple(self.lang),
+                self.flags
+            )
+
+    def __str__(self):  # pragma: no cover
+        """String representation."""
+
+        return (
+            '_Selector(tag={!r}, ids={!r}, classes={!r}, attributes={!r}, nth={!r}, selectors={!r}, '
+            'relations={!r}, rel_type={!r}, contains={!r}, lang={!r}, flags={!r}, no_match={!r})'
+        ).format(
+            self.tag, self.ids, self.classes, self.attributes, self.nth, self.selectors,
+            self.relations, self.rel_type, self.contains, self.lang, self.flags, self.no_match
+        )
+
+    __repr__ = __str__
+
+
+class CSSParser(object):
+    """Parse CSS selectors."""
+
+    css_tokens = (
+        SelectorPattern("pseudo_close", PAT_PSEUDO_CLOSE),
+        SpecialPseudoPattern(
+            (
+                ("pseudo_contains", (':contains',), PAT_PSEUDO_CONTAINS),
+                ("pseudo_nth_child", (':nth-child', ':nth-last-child'), PAT_PSEUDO_NTH_CHILD),
+                ("pseudo_nth_type", (':nth-of-type', ':nth-last-of-type'), PAT_PSEUDO_NTH_TYPE),
+                ("pseudo_lang", (':lang',), PAT_PSEUDO_LANG),
+                ("pseudo_dir", (':dir',), PAT_PSEUDO_DIR)
+            )
+        ),
+        SelectorPattern("pseudo_class_custom", PAT_PSEUDO_CLASS_CUSTOM),
+        SelectorPattern("pseudo_class", PAT_PSEUDO_CLASS),
+        SelectorPattern("pseudo_element", PAT_PSEUDO_ELEMENT),
+        SelectorPattern("at_rule", PAT_AT_RULE),
+        SelectorPattern("id", PAT_ID),
+        SelectorPattern("class", PAT_CLASS),
+        SelectorPattern("tag", PAT_TAG),
+        SelectorPattern("attribute", PAT_ATTR),
+        SelectorPattern("combine", PAT_COMBINE)
+    )
+
+    def __init__(self, selector, custom=None, flags=0):
+        """Initialize."""
+
+        self.pattern = selector.replace('\x00', '\ufffd')
+        self.flags = flags
+        self.debug = self.flags & util.DEBUG
+        self.custom = {} if custom is None else custom
+
+    def parse_attribute_selector(self, sel, m, has_selector):
+        """Create attribute selector from the returned regex match."""
+
+        inverse = False
+        op = m.group('cmp')
+        case = util.lower(m.group('case')) if m.group('case') else None
+        parts = [css_unescape(a) for a in m.group('ns_attr').split('|')]
+        ns = ''
+        is_type = False
+        pattern2 = None
+        if len(parts) > 1:
+            ns = parts[0]
+            attr = parts[1]
+        else:
+            attr = parts[0]
+        if case:
+            flags = re.I if case == 'i' else 0
+        elif util.lower(attr) == 'type':
+            flags = re.I
+            is_type = True
+        else:
+            flags = 0
+
+        if op:
+            if m.group('value').startswith(('"', "'")):
+                value = css_unescape(m.group('value')[1:-1], True)
+            else:
+                value = css_unescape(m.group('value'))
+        else:
+            value = None
+        if not op:
+            # Attribute name
+            pattern = None
+        elif op.startswith('^'):
+            # Value start with
+            pattern = re.compile(r'^%s.*' % re.escape(value), flags)
+        elif op.startswith('$'):
+            # Value ends with
+            pattern = re.compile(r'.*?%s$' % re.escape(value), flags)
+        elif op.startswith('*'):
+            # Value contains
+            pattern = re.compile(r'.*?%s.*' % re.escape(value), flags)
+        elif op.startswith('~'):
+            # Value contains word within space separated list
+            # `~=` should match nothing if it is empty or contains whitespace,
+            # so if either of these cases is present, use `[^\s\S]` which cannot be matched.
+            value = r'[^\s\S]' if not value or RE_WS.search(value) else re.escape(value)
+            pattern = re.compile(r'.*?(?:(?<=^)|(?<=[ \t\r\n\f]))%s(?=(?:[ \t\r\n\f]|$)).*' % value, flags)
+        elif op.startswith('|'):
+            # Value starts with word in dash separated list
+            pattern = re.compile(r'^%s(?:-.*)?$' % re.escape(value), flags)
+        else:
+            # Value matches
+            pattern = re.compile(r'^%s$' % re.escape(value), flags)
+            if op.startswith('!'):
+                # Equivalent to `:not([attr=value])`
+                inverse = True
+        if is_type and pattern:
+            pattern2 = re.compile(pattern.pattern)
+
+        # Append the attribute selector
+        sel_attr = ct.SelectorAttribute(attr, ns, pattern, pattern2)
+        if inverse:
+            # If we are using `!=`, we need to nest the pattern under a `:not()`.
+            sub_sel = _Selector()
+            sub_sel.attributes.append(sel_attr)
+            not_list = ct.SelectorList([sub_sel.freeze()], True, False)
+            sel.selectors.append(not_list)
+        else:
+            sel.attributes.append(sel_attr)
+
+        has_selector = True
+        return has_selector
+
+    def parse_tag_pattern(self, sel, m, has_selector):
+        """Parse tag pattern from regex match."""
+
+        parts = [css_unescape(x) for x in m.group(0).split('|')]
+        if len(parts) > 1:
+            prefix = parts[0]
+            tag = parts[1]
+        else:
+            tag = parts[0]
+            prefix = None
+        sel.tag = ct.SelectorTag(tag, prefix)
+        has_selector = True
+        return has_selector
+
+    def parse_pseudo_class_custom(self, sel, m, has_selector):
+        """
+        Parse custom pseudo class alias.
+
+        Compile custom selectors as we need them. When compiling a custom selector,
+        set it to `None` in the dictionary so we can avoid an infinite loop.
+        """
+
+        pseudo = util.lower(css_unescape(m.group('name')))
+        selector = self.custom.get(pseudo)
+        if selector is None:
+            raise SelectorSyntaxError(
+                "Undefined custom selector '{}' found at postion {}".format(pseudo, m.end(0)),
+                self.pattern,
+                m.end(0)
+            )
+
+        if not isinstance(selector, ct.SelectorList):
+            self.custom[pseudo] = None
+            selector = CSSParser(
+                selector, custom=self.custom, flags=self.flags
+            ).process_selectors(flags=FLG_PSEUDO)
+            self.custom[pseudo] = selector
+
+        sel.selectors.append(selector)
+        has_selector = True
+        return has_selector
+
+    def parse_pseudo_class(self, sel, m, has_selector, iselector, is_html):
+        """Parse pseudo class."""
+
+        complex_pseudo = False
+        pseudo = util.lower(css_unescape(m.group('name')))
+        if m.group('open'):
+            complex_pseudo = True
+        if complex_pseudo and pseudo in PSEUDO_COMPLEX:
+            has_selector = self.parse_pseudo_open(sel, pseudo, has_selector, iselector, m.end(0))
+        elif not complex_pseudo and pseudo in PSEUDO_SIMPLE:
+            if pseudo == ':root':
+                sel.flags |= ct.SEL_ROOT
+            elif pseudo == ':defined':
+                sel.flags |= ct.SEL_DEFINED
+                is_html = True
+            elif pseudo == ':scope':
+                sel.flags |= ct.SEL_SCOPE
+            elif pseudo == ':empty':
+                sel.flags |= ct.SEL_EMPTY
+            elif pseudo in (':link', ':any-link'):
+                sel.selectors.append(CSS_LINK)
+            elif pseudo == ':checked':
+                sel.selectors.append(CSS_CHECKED)
+            elif pseudo == ':default':
+                sel.selectors.append(CSS_DEFAULT)
+            elif pseudo == ':indeterminate':
+                sel.selectors.append(CSS_INDETERMINATE)
+            elif pseudo == ":disabled":
+                sel.selectors.append(CSS_DISABLED)
+            elif pseudo == ":enabled":
+                sel.selectors.append(CSS_ENABLED)
+            elif pseudo == ":required":
+                sel.selectors.append(CSS_REQUIRED)
+            elif pseudo == ":optional":
+                sel.selectors.append(CSS_OPTIONAL)
+            elif pseudo == ":read-only":
+                sel.selectors.append(CSS_READ_ONLY)
+            elif pseudo == ":read-write":
+                sel.selectors.append(CSS_READ_WRITE)
+            elif pseudo == ":in-range":
+                sel.selectors.append(CSS_IN_RANGE)
+            elif pseudo == ":out-of-range":
+                sel.selectors.append(CSS_OUT_OF_RANGE)
+            elif pseudo == ":placeholder-shown":
+                sel.selectors.append(CSS_PLACEHOLDER_SHOWN)
+            elif pseudo == ':first-child':
+                sel.nth.append(ct.SelectorNth(1, False, 0, False, False, ct.SelectorList()))
+            elif pseudo == ':last-child':
+                sel.nth.append(ct.SelectorNth(1, False, 0, False, True, ct.SelectorList()))
+            elif pseudo == ':first-of-type':
+                sel.nth.append(ct.SelectorNth(1, False, 0, True, False, ct.SelectorList()))
+            elif pseudo == ':last-of-type':
+                sel.nth.append(ct.SelectorNth(1, False, 0, True, True, ct.SelectorList()))
+            elif pseudo == ':only-child':
+                sel.nth.extend(
+                    [
+                        ct.SelectorNth(1, False, 0, False, False, ct.SelectorList()),
+                        ct.SelectorNth(1, False, 0, False, True, ct.SelectorList())
+                    ]
+                )
+            elif pseudo == ':only-of-type':
+                sel.nth.extend(
+                    [
+                        ct.SelectorNth(1, False, 0, True, False, ct.SelectorList()),
+                        ct.SelectorNth(1, False, 0, True, True, ct.SelectorList())
+                    ]
+                )
+            has_selector = True
+        elif complex_pseudo and pseudo in PSEUDO_COMPLEX_NO_MATCH:
+            self.parse_selectors(iselector, m.end(0), FLG_PSEUDO | FLG_OPEN)
+            sel.no_match = True
+            has_selector = True
+        elif not complex_pseudo and pseudo in PSEUDO_SIMPLE_NO_MATCH:
+            sel.no_match = True
+            has_selector = True
+        elif pseudo in PSEUDO_SUPPORTED:
+            raise SelectorSyntaxError(
+                "Invalid syntax for pseudo class '{}'".format(pseudo),
+                self.pattern,
+                m.start(0)
+            )
+        else:
+            raise NotImplementedError(
+                "'{}' pseudo-class is not implemented at this time".format(pseudo)
+            )
+
+        return has_selector, is_html
+
+    def parse_pseudo_nth(self, sel, m, has_selector, iselector):
+        """Parse `nth` pseudo."""
+
+        mdict = m.groupdict()
+        if mdict.get('pseudo_nth_child'):
+            postfix = '_child'
+        else:
+            postfix = '_type'
+        mdict['name'] = util.lower(css_unescape(mdict['name']))
+        content = util.lower(mdict.get('nth' + postfix))
+        if content == 'even':
+            # 2n
+            s1 = 2
+            s2 = 0
+            var = True
+        elif content == 'odd':
+            # 2n+1
+            s1 = 2
+            s2 = 1
+            var = True
+        else:
+            nth_parts = RE_NTH.match(content)
+            s1 = '-' if nth_parts.group('s1') and nth_parts.group('s1') == '-' else ''
+            a = nth_parts.group('a')
+            var = a.endswith('n')
+            if a.startswith('n'):
+                s1 += '1'
+            elif var:
+                s1 += a[:-1]
+            else:
+                s1 += a
+            s2 = '-' if nth_parts.group('s2') and nth_parts.group('s2') == '-' else ''
+            if nth_parts.group('b'):
+                s2 += nth_parts.group('b')
+            else:
+                s2 = '0'
+            s1 = int(s1, 10)
+            s2 = int(s2, 10)
+
+        pseudo_sel = mdict['name']
+        if postfix == '_child':
+            if m.group('of'):
+                # Parse the rest of `of S`.
+                nth_sel = self.parse_selectors(iselector, m.end(0), FLG_PSEUDO | FLG_OPEN)
+            else:
+                # Use default `*|*` for `of S`.
+                nth_sel = CSS_NTH_OF_S_DEFAULT
+            if pseudo_sel == ':nth-child':
+                sel.nth.append(ct.SelectorNth(s1, var, s2, False, False, nth_sel))
+            elif pseudo_sel == ':nth-last-child':
+                sel.nth.append(ct.SelectorNth(s1, var, s2, False, True, nth_sel))
+        else:
+            if pseudo_sel == ':nth-of-type':
+                sel.nth.append(ct.SelectorNth(s1, var, s2, True, False, ct.SelectorList()))
+            elif pseudo_sel == ':nth-last-of-type':
+                sel.nth.append(ct.SelectorNth(s1, var, s2, True, True, ct.SelectorList()))
+        has_selector = True
+        return has_selector
+
+    def parse_pseudo_open(self, sel, name, has_selector, iselector, index):
+        """Parse pseudo with opening bracket."""
+
+        flags = FLG_PSEUDO | FLG_OPEN
+        if name == ':not':
+            flags |= FLG_NOT
+        if name == ':has':
+            flags |= FLG_RELATIVE
+
+        sel.selectors.append(self.parse_selectors(iselector, index, flags))
+        has_selector = True
+        return has_selector
+
+    def parse_has_combinator(self, sel, m, has_selector, selectors, rel_type, index):
+        """Parse combinator tokens."""
+
+        combinator = m.group('relation').strip()
+        if not combinator:
+            combinator = WS_COMBINATOR
+        if combinator == COMMA_COMBINATOR:
+            if not has_selector:
+                # If we've not captured any selector parts, the comma is either at the beginning of the pattern
+                # or following another comma, both of which are unexpected. Commas must split selectors.
+                raise SelectorSyntaxError(
+                    "The combinator '{}' at postion {}, must have a selector before it".format(combinator, index),
+                    self.pattern,
+                    index
+                )
+            sel.rel_type = rel_type
+            selectors[-1].relations.append(sel)
+            rel_type = ":" + WS_COMBINATOR
+            selectors.append(_Selector())
+        else:
+            if has_selector:
+                # End the current selector and associate the leading combinator with this selector.
+                sel.rel_type = rel_type
+                selectors[-1].relations.append(sel)
+            elif rel_type[1:] != WS_COMBINATOR:
+                # It's impossible to have two whitespace combinators after each other as the patterns
+                # will gobble up trailing whitespace. It is also impossible to have a whitespace
+                # combinator after any other kind for the same reason. But we could have
+                # multiple non-whitespace combinators. So if the current combinator is not a whitespace,
+                # then we've hit the multiple combinator case, so we should fail.
+                raise SelectorSyntaxError(
+                    'The multiple combinators at position {}'.format(index),
+                    self.pattern,
+                    index
+                )
+            # Set the leading combinator for the next selector.
+            rel_type = ':' + combinator
+        sel = _Selector()
+
+        has_selector = False
+        return has_selector, sel, rel_type
+
+    def parse_combinator(self, sel, m, has_selector, selectors, relations, is_pseudo, index):
+        """Parse combinator tokens."""
+
+        combinator = m.group('relation').strip()
+        if not combinator:
+            combinator = WS_COMBINATOR
+        if not has_selector:
+            raise SelectorSyntaxError(
+                "The combinator '{}' at postion {}, must have a selector before it".format(combinator, index),
+                self.pattern,
+                index
+            )
+
+        if combinator == COMMA_COMBINATOR:
+            if not sel.tag and not is_pseudo:
+                # Implied `*`
+                sel.tag = ct.SelectorTag('*', None)
+            sel.relations.extend(relations)
+            selectors.append(sel)
+            del relations[:]
+        else:
+            sel.relations.extend(relations)
+            sel.rel_type = combinator
+            del relations[:]
+            relations.append(sel)
+        sel = _Selector()
+
+        has_selector = False
+        return has_selector, sel
+
+    def parse_class_id(self, sel, m, has_selector):
+        """Parse HTML classes and ids."""
+
+        selector = m.group(0)
+        if selector.startswith('.'):
+            sel.classes.append(css_unescape(selector[1:]))
+        else:
+            sel.ids.append(css_unescape(selector[1:]))
+        has_selector = True
+        return has_selector
+
+    def parse_pseudo_contains(self, sel, m, has_selector):
+        """Parse contains."""
+
+        values = m.group('values')
+        patterns = []
+        for token in RE_VALUES.finditer(values):
+            if token.group('split'):
+                continue
+            value = token.group('value')
+            if value.startswith(("'", '"')):
+                value = css_unescape(value[1:-1], True)
+            else:
+                value = css_unescape(value)
+            patterns.append(value)
+        sel.contains.append(ct.SelectorContains(tuple(patterns)))
+        has_selector = True
+        return has_selector
+
+    def parse_pseudo_lang(self, sel, m, has_selector):
+        """Parse pseudo language."""
+
+        values = m.group('values')
+        patterns = []
+        for token in RE_VALUES.finditer(values):
+            if token.group('split'):
+                continue
+            value = token.group('value')
+            if value.startswith(('"', "'")):
+                value = css_unescape(value[1:-1], True)
+            else:
+                value = css_unescape(value)
+
+            patterns.append(value)
+
+        sel.lang.append(ct.SelectorLang(patterns))
+        has_selector = True
+
+        return has_selector
+
+    def parse_pseudo_dir(self, sel, m, has_selector):
+        """Parse pseudo direction."""
+
+        value = ct.SEL_DIR_LTR if util.lower(m.group('dir')) == 'ltr' else ct.SEL_DIR_RTL
+        sel.flags |= value
+        has_selector = True
+        return has_selector
+
+    def parse_selectors(self, iselector, index=0, flags=0):
+        """Parse selectors."""
+
+        sel = _Selector()
+        selectors = []
+        has_selector = False
+        closed = False
+        relations = []
+        rel_type = ":" + WS_COMBINATOR
+        is_open = bool(flags & FLG_OPEN)
+        is_pseudo = bool(flags & FLG_PSEUDO)
+        is_relative = bool(flags & FLG_RELATIVE)
+        is_not = bool(flags & FLG_NOT)
+        is_html = bool(flags & FLG_HTML)
+        is_default = bool(flags & FLG_DEFAULT)
+        is_indeterminate = bool(flags & FLG_INDETERMINATE)
+        is_in_range = bool(flags & FLG_IN_RANGE)
+        is_out_of_range = bool(flags & FLG_OUT_OF_RANGE)
+        is_placeholder_shown = bool(flags & FLG_PLACEHOLDER_SHOWN)
+
+        if self.debug:  # pragma: no cover
+            if is_pseudo:
+                print('    is_pseudo: True')
+            if is_open:
+                print('    is_open: True')
+            if is_relative:
+                print('    is_relative: True')
+            if is_not:
+                print('    is_not: True')
+            if is_html:
+                print('    is_html: True')
+            if is_default:
+                print('    is_default: True')
+            if is_indeterminate:
+                print('    is_indeterminate: True')
+            if is_in_range:
+                print('    is_in_range: True')
+            if is_out_of_range:
+                print('    is_out_of_range: True')
+            if is_placeholder_shown:
+                print('    is_placeholder_shown: True')
+
+        if is_relative:
+            selectors.append(_Selector())
+
+        try:
+            while True:
+                key, m = next(iselector)
+
+                # Handle parts
+                if key == "at_rule":
+                    raise NotImplementedError("At-rules found at position {}".format(m.start(0)))
+                elif key == 'pseudo_class_custom':
+                    has_selector = self.parse_pseudo_class_custom(sel, m, has_selector)
+                elif key == 'pseudo_class':
+                    has_selector, is_html = self.parse_pseudo_class(sel, m, has_selector, iselector, is_html)
+                elif key == 'pseudo_element':
+                    raise NotImplementedError("Psuedo-element found at position {}".format(m.start(0)))
+                elif key == 'pseudo_contains':
+                    has_selector = self.parse_pseudo_contains(sel, m, has_selector)
+                elif key in ('pseudo_nth_type', 'pseudo_nth_child'):
+                    has_selector = self.parse_pseudo_nth(sel, m, has_selector, iselector)
+                elif key == 'pseudo_lang':
+                    has_selector = self.parse_pseudo_lang(sel, m, has_selector)
+                elif key == 'pseudo_dir':
+                    has_selector = self.parse_pseudo_dir(sel, m, has_selector)
+                    # Currently only supports HTML
+                    is_html = True
+                elif key == 'pseudo_close':
+                    if not has_selector:
+                        raise SelectorSyntaxError(
+                            "Expected a selector at postion {}".format(m.start(0)),
+                            self.pattern,
+                            m.start(0)
+                        )
+                    if is_open:
+                        closed = True
+                        break
+                    else:
+                        raise SelectorSyntaxError(
+                            "Unmatched pseudo-class close at postion {}".format(m.start(0)),
+                            self.pattern,
+                            m.start(0)
+                        )
+                elif key == 'combine':
+                    if is_relative:
+                        has_selector, sel, rel_type = self.parse_has_combinator(
+                            sel, m, has_selector, selectors, rel_type, index
+                        )
+                    else:
+                        has_selector, sel = self.parse_combinator(
+                            sel, m, has_selector, selectors, relations, is_pseudo, index
+                        )
+                elif key == 'attribute':
+                    has_selector = self.parse_attribute_selector(sel, m, has_selector)
+                elif key == 'tag':
+                    if has_selector:
+                        raise SelectorSyntaxError(
+                            "Tag name found at position {} instead of at the start".format(m.start(0)),
+                            self.pattern,
+                            m.start(0)
+                        )
+                    has_selector = self.parse_tag_pattern(sel, m, has_selector)
+                elif key in ('class', 'id'):
+                    has_selector = self.parse_class_id(sel, m, has_selector)
+
+                index = m.end(0)
+        except StopIteration:
+            pass
+
+        if is_open and not closed:
+            raise SelectorSyntaxError(
+                "Unclosed pseudo-class at position {}".format(index),
+                self.pattern,
+                index
+            )
+
+        if has_selector:
+            if not sel.tag and not is_pseudo:
+                # Implied `*`
+                sel.tag = ct.SelectorTag('*', None)
+            if is_relative:
+                sel.rel_type = rel_type
+                selectors[-1].relations.append(sel)
+            else:
+                sel.relations.extend(relations)
+                del relations[:]
+                selectors.append(sel)
+        else:
+            # We will always need to finish a selector when `:has()` is used as it leads with combining.
+            raise SelectorSyntaxError(
+                'Expected a selector at position {}'.format(index),
+                self.pattern,
+                index
+            )
+
+        # Some patterns require additional logic, such as default. We try to make these the
+        # last pattern, and append the appropriate flag to that selector which communicates
+        # to the matcher what additional logic is required.
+        if is_default:
+            selectors[-1].flags = ct.SEL_DEFAULT
+        if is_indeterminate:
+            selectors[-1].flags = ct.SEL_INDETERMINATE
+        if is_in_range:
+            selectors[-1].flags = ct.SEL_IN_RANGE
+        if is_out_of_range:
+            selectors[-1].flags = ct.SEL_OUT_OF_RANGE
+        if is_placeholder_shown:
+            selectors[-1].flags = ct.SEL_PLACEHOLDER_SHOWN
+
+        return ct.SelectorList([s.freeze() for s in selectors], is_not, is_html)
+
+    def selector_iter(self, pattern):
+        """Iterate selector tokens."""
+
+        # Ignore whitespace and comments at start and end of pattern
+        m = RE_WS_BEGIN.search(pattern)
+        index = m.end(0) if m else 0
+        m = RE_WS_END.search(pattern)
+        end = (m.start(0) - 1) if m else (len(pattern) - 1)
+
+        if self.debug:  # pragma: no cover
+            print('## PARSING: {!r}'.format(pattern))
+        while index <= end:
+            m = None
+            for v in self.css_tokens:
+                if not v.enabled(self.flags):  # pragma: no cover
+                    continue
+                m = v.match(pattern, index)
+                if m:
+                    name = v.get_name()
+                    if self.debug:  # pragma: no cover
+                        print("TOKEN: '{}' --> {!r} at position {}".format(name, m.group(0), m.start(0)))
+                    index = m.end(0)
+                    yield name, m
+                    break
+            if m is None:
+                c = pattern[index]
+                # If the character represents the start of one of the known selector types,
+                # throw an exception mentioning that the known selector type is in error;
+                # otherwise, report the invalid character.
+                if c == '[':
+                    msg = "Malformed attribute selector at position {}".format(index)
+                elif c == '.':
+                    msg = "Malformed class selector at position {}".format(index)
+                elif c == '#':
+                    msg = "Malformed id selector at position {}".format(index)
+                elif c == ':':
+                    msg = "Malformed pseudo-class selector at position {}".format(index)
+                else:
+                    msg = "Invalid character {!r} position {}".format(c, index)
+                raise SelectorSyntaxError(msg, self.pattern, index)
+        if self.debug:  # pragma: no cover
+            print('## END PARSING')
+
+    def process_selectors(self, index=0, flags=0):
+        """Process selectors."""
+
+        return self.parse_selectors(self.selector_iter(self.pattern), index, flags)
+
+
+# Precompile CSS selector lists for pseudo-classes (additional logic may be required beyond the pattern)
+# A few patterns are order dependent as they use patterns previous compiled.
+
+# CSS pattern for `:link` and `:any-link`
+CSS_LINK = CSSParser(
+    'html|*:is(a, area, link)[href]'
+).process_selectors(flags=FLG_PSEUDO | FLG_HTML)
+# CSS pattern for `:checked`
+CSS_CHECKED = CSSParser(
+    '''
+    html|*:is(input[type=checkbox], input[type=radio])[checked], html|option[selected]
+    '''
+).process_selectors(flags=FLG_PSEUDO | FLG_HTML)
+# CSS pattern for `:default` (must compile CSS_CHECKED first)
+CSS_DEFAULT = CSSParser(
+    '''
+    :checked,
+
+    /*
+    This pattern must be at the end.
+    Special logic is applied to the last selector.
+    */
+    html|form html|*:is(button, input)[type="submit"]
+    '''
+).process_selectors(flags=FLG_PSEUDO | FLG_HTML | FLG_DEFAULT)
+# CSS pattern for `:indeterminate`
+CSS_INDETERMINATE = CSSParser(
+    '''
+    html|input[type="checkbox"][indeterminate],
+    html|input[type="radio"]:is(:not([name]), [name=""]):not([checked]),
+    html|progress:not([value]),
+
+    /*
+    This pattern must be at the end.
+    Special logic is applied to the last selector.
+    */
+    html|input[type="radio"][name][name!='']:not([checked])
+    '''
+).process_selectors(flags=FLG_PSEUDO | FLG_HTML | FLG_INDETERMINATE)
+# CSS pattern for `:disabled`
+CSS_DISABLED = CSSParser(
+    '''
+    html|*:is(input[type!=hidden], button, select, textarea, fieldset, optgroup, option, fieldset)[disabled],
+    html|optgroup[disabled] > html|option,
+    html|fieldset[disabled] > html|*:is(input[type!=hidden], button, select, textarea, fieldset),
+    html|fieldset[disabled] >
+        html|*:not(legend:nth-of-type(1)) html|*:is(input[type!=hidden], button, select, textarea, fieldset)
+    '''
+).process_selectors(flags=FLG_PSEUDO | FLG_HTML)
+# CSS pattern for `:enabled`
+CSS_ENABLED = CSSParser(
+    '''
+    html|*:is(input[type!=hidden], button, select, textarea, fieldset, optgroup, option, fieldset):not(:disabled)
+    '''
+).process_selectors(flags=FLG_PSEUDO | FLG_HTML)
+# CSS pattern for `:required`
+CSS_REQUIRED = CSSParser(
+    'html|*:is(input, textarea, select)[required]'
+).process_selectors(flags=FLG_PSEUDO | FLG_HTML)
+# CSS pattern for `:optional`
+CSS_OPTIONAL = CSSParser(
+    'html|*:is(input, textarea, select):not([required])'
+).process_selectors(flags=FLG_PSEUDO | FLG_HTML)
+# CSS pattern for `:placeholder-shown`
+CSS_PLACEHOLDER_SHOWN = CSSParser(
+    '''
+    html|input:is(
+        :not([type]),
+        [type=""],
+        [type=text],
+        [type=search],
+        [type=url],
+        [type=tel],
+        [type=email],
+        [type=password],
+        [type=number]
+    )[placeholder][placeholder!='']:is(:not([value]), [value=""]),
+    html|textarea[placeholder][placeholder!='']
+    '''
+).process_selectors(flags=FLG_PSEUDO | FLG_HTML | FLG_PLACEHOLDER_SHOWN)
+# CSS pattern default for `:nth-child` "of S" feature
+CSS_NTH_OF_S_DEFAULT = CSSParser(
+    '*|*'
+).process_selectors(flags=FLG_PSEUDO)
+# CSS pattern for `:read-write` (CSS_DISABLED must be compiled first)
+CSS_READ_WRITE = CSSParser(
+    '''
+    html|*:is(
+        textarea,
+        input:is(
+            :not([type]),
+            [type=""],
+            [type=text],
+            [type=search],
+            [type=url],
+            [type=tel],
+            [type=email],
+            [type=number],
+            [type=password],
+            [type=date],
+            [type=datetime-local],
+            [type=month],
+            [type=time],
+            [type=week]
+        )
+    ):not([readonly], :disabled),
+    html|*:is([contenteditable=""], [contenteditable="true" i])
+    '''
+).process_selectors(flags=FLG_PSEUDO | FLG_HTML)
+# CSS pattern for `:read-only`
+CSS_READ_ONLY = CSSParser(
+    '''
+    html|*:not(:read-write)
+    '''
+).process_selectors(flags=FLG_PSEUDO | FLG_HTML)
+# CSS pattern for `:in-range`
+CSS_IN_RANGE = CSSParser(
+    '''
+    html|input:is(
+        [type="date"],
+        [type="month"],
+        [type="week"],
+        [type="time"],
+        [type="datetime-local"],
+        [type="number"],
+        [type="range"]
+    ):is(
+        [min],
+        [max]
+    )
+    '''
+).process_selectors(flags=FLG_PSEUDO | FLG_IN_RANGE | FLG_HTML)
+# CSS pattern for `:out-of-range`
+CSS_OUT_OF_RANGE = CSSParser(
+    '''
+    html|input:is(
+        [type="date"],
+        [type="month"],
+        [type="week"],
+        [type="time"],
+        [type="datetime-local"],
+        [type="number"],
+        [type="range"]
+    ):is(
+        [min],
+        [max]
+    )
+    '''
+).process_selectors(flags=FLG_PSEUDO | FLG_OUT_OF_RANGE | FLG_HTML)
diff --git a/lib/soupsieve/css_types.py b/lib/soupsieve/css_types.py
new file mode 100644
index 00000000..e4baef37
--- /dev/null
+++ b/lib/soupsieve/css_types.py
@@ -0,0 +1,345 @@
+"""CSS selector structure items."""
+from __future__ import unicode_literals
+from . import util
+
+__all__ = (
+    'Selector',
+    'SelectorNull',
+    'SelectorTag',
+    'SelectorAttribute',
+    'SelectorContains',
+    'SelectorNth',
+    'SelectorLang',
+    'SelectorList',
+    'Namespaces',
+    'CustomSelectors'
+)
+
+
+SEL_EMPTY = 0x1
+SEL_ROOT = 0x2
+SEL_DEFAULT = 0x4
+SEL_INDETERMINATE = 0x8
+SEL_SCOPE = 0x10
+SEL_DIR_LTR = 0x20
+SEL_DIR_RTL = 0x40
+SEL_IN_RANGE = 0x80
+SEL_OUT_OF_RANGE = 0x100
+SEL_DEFINED = 0x200
+SEL_PLACEHOLDER_SHOWN = 0x400
+
+
+class Immutable(object):
+    """Immutable."""
+
+    __slots__ = ('_hash',)
+
+    def __init__(self, **kwargs):
+        """Initialize."""
+
+        temp = []
+        for k, v in kwargs.items():
+            temp.append(type(v))
+            temp.append(v)
+            super(Immutable, self).__setattr__(k, v)
+        super(Immutable, self).__setattr__('_hash', hash(tuple(temp)))
+
+    @classmethod
+    def __base__(cls):
+        """Get base class."""
+
+        return cls
+
+    def __eq__(self, other):
+        """Equal."""
+
+        return (
+            isinstance(other, self.__base__()) and
+            all([getattr(other, key) == getattr(self, key) for key in self.__slots__ if key != '_hash'])
+        )
+
+    def __ne__(self, other):
+        """Equal."""
+
+        return (
+            not isinstance(other, self.__base__()) or
+            any([getattr(other, key) != getattr(self, key) for key in self.__slots__ if key != '_hash'])
+        )
+
+    def __hash__(self):
+        """Hash."""
+
+        return self._hash
+
+    def __setattr__(self, name, value):
+        """Prevent mutability."""
+
+        raise AttributeError("'{}' is immutable".format(self.__class__.__name__))
+
+    def __repr__(self):  # pragma: no cover
+        """Representation."""
+
+        return "{}({})".format(
+            self.__base__(), ', '.join(["{}={!r}".format(k, getattr(self, k)) for k in self.__slots__[:-1]])
+        )
+
+    __str__ = __repr__
+
+
+class ImmutableDict(util.Mapping):
+    """Hashable, immutable dictionary."""
+
+    def __init__(self, *args, **kwargs):
+        """Initialize."""
+
+        arg = args[0] if args else kwargs
+        is_dict = isinstance(arg, dict)
+        if (
+            is_dict and not all([isinstance(v, util.Hashable) for v in arg.values()]) or
+            not is_dict and not all([isinstance(k, util.Hashable) and isinstance(v, util.Hashable) for k, v in arg])
+        ):
+            raise TypeError('All values must be hashable')
+
+        self._d = dict(*args, **kwargs)
+        self._hash = hash(tuple([(type(x), x, type(y), y) for x, y in sorted(self._d.items())]))
+
+    def __iter__(self):
+        """Iterator."""
+
+        return iter(self._d)
+
+    def __len__(self):
+        """Length."""
+
+        return len(self._d)
+
+    def __getitem__(self, key):
+        """Get item: `namespace['key']`."""
+        return self._d[key]
+
+    def __hash__(self):
+        """Hash."""
+
+        return self._hash
+
+    def __repr__(self):  # pragma: no cover
+        """Representation."""
+
+        return "{!r}".format(self._d)
+
+    __str__ = __repr__
+
+
+class Namespaces(ImmutableDict):
+    """Namespaces."""
+
+    def __init__(self, *args, **kwargs):
+        """Initialize."""
+
+        # If there are arguments, check the first index.
+        # `super` should fail if the user gave multiple arguments,
+        # so don't bother checking that.
+        arg = args[0] if args else kwargs
+        is_dict = isinstance(arg, dict)
+        if is_dict and not all([isinstance(k, util.string) and isinstance(v, util.string) for k, v in arg.items()]):
+            raise TypeError('Namespace keys and values must be Unicode strings')
+        elif not is_dict and not all([isinstance(k, util.string) and isinstance(v, util.string) for k, v in arg]):
+            raise TypeError('Namespace keys and values must be Unicode strings')
+
+        super(Namespaces, self).__init__(*args, **kwargs)
+
+
+class CustomSelectors(ImmutableDict):
+    """Custom selectors."""
+
+    def __init__(self, *args, **kwargs):
+        """Initialize."""
+
+        # If there are arguments, check the first index.
+        # `super` should fail if the user gave multiple arguments,
+        # so don't bother checking that.
+        arg = args[0] if args else kwargs
+        is_dict = isinstance(arg, dict)
+        if is_dict and not all([isinstance(k, util.string) and isinstance(v, util.string) for k, v in arg.items()]):
+            raise TypeError('CustomSelectors keys and values must be Unicode strings')
+        elif not is_dict and not all([isinstance(k, util.string) and isinstance(v, util.string) for k, v in arg]):
+            raise TypeError('CustomSelectors keys and values must be Unicode strings')
+
+        super(CustomSelectors, self).__init__(*args, **kwargs)
+
+
+class Selector(Immutable):
+    """Selector."""
+
+    __slots__ = (
+        'tag', 'ids', 'classes', 'attributes', 'nth', 'selectors',
+        'relation', 'rel_type', 'contains', 'lang', 'flags', '_hash'
+    )
+
+    def __init__(
+        self, tag, ids, classes, attributes, nth, selectors,
+        relation, rel_type, contains, lang, flags
+    ):
+        """Initialize."""
+
+        super(Selector, self).__init__(
+            tag=tag,
+            ids=ids,
+            classes=classes,
+            attributes=attributes,
+            nth=nth,
+            selectors=selectors,
+            relation=relation,
+            rel_type=rel_type,
+            contains=contains,
+            lang=lang,
+            flags=flags
+        )
+
+
+class SelectorNull(Immutable):
+    """Null Selector."""
+
+    def __init__(self):
+        """Initialize."""
+
+        super(SelectorNull, self).__init__()
+
+
+class SelectorTag(Immutable):
+    """Selector tag."""
+
+    __slots__ = ("name", "prefix", "_hash")
+
+    def __init__(self, name, prefix):
+        """Initialize."""
+
+        super(SelectorTag, self).__init__(
+            name=name,
+            prefix=prefix
+        )
+
+
+class SelectorAttribute(Immutable):
+    """Selector attribute rule."""
+
+    __slots__ = ("attribute", "prefix", "pattern", "xml_type_pattern", "_hash")
+
+    def __init__(self, attribute, prefix, pattern, xml_type_pattern):
+        """Initialize."""
+
+        super(SelectorAttribute, self).__init__(
+            attribute=attribute,
+            prefix=prefix,
+            pattern=pattern,
+            xml_type_pattern=xml_type_pattern
+        )
+
+
+class SelectorContains(Immutable):
+    """Selector contains rule."""
+
+    __slots__ = ("text", "_hash")
+
+    def __init__(self, text):
+        """Initialize."""
+
+        super(SelectorContains, self).__init__(
+            text=text
+        )
+
+
+class SelectorNth(Immutable):
+    """Selector nth type."""
+
+    __slots__ = ("a", "n", "b", "of_type", "last", "selectors", "_hash")
+
+    def __init__(self, a, n, b, of_type, last, selectors):
+        """Initialize."""
+
+        super(SelectorNth, self).__init__(
+            a=a,
+            n=n,
+            b=b,
+            of_type=of_type,
+            last=last,
+            selectors=selectors
+        )
+
+
+class SelectorLang(Immutable):
+    """Selector language rules."""
+
+    __slots__ = ("languages", "_hash",)
+
+    def __init__(self, languages):
+        """Initialize."""
+
+        super(SelectorLang, self).__init__(
+            languages=tuple(languages)
+        )
+
+    def __iter__(self):
+        """Iterator."""
+
+        return iter(self.languages)
+
+    def __len__(self):  # pragma: no cover
+        """Length."""
+
+        return len(self.languages)
+
+    def __getitem__(self, index):  # pragma: no cover
+        """Get item."""
+
+        return self.languages[index]
+
+
+class SelectorList(Immutable):
+    """Selector list."""
+
+    __slots__ = ("selectors", "is_not", "is_html", "_hash")
+
+    def __init__(self, selectors=tuple(), is_not=False, is_html=False):
+        """Initialize."""
+
+        super(SelectorList, self).__init__(
+            selectors=tuple(selectors),
+            is_not=is_not,
+            is_html=is_html
+        )
+
+    def __iter__(self):
+        """Iterator."""
+
+        return iter(self.selectors)
+
+    def __len__(self):
+        """Length."""
+
+        return len(self.selectors)
+
+    def __getitem__(self, index):
+        """Get item."""
+
+        return self.selectors[index]
+
+
+def _pickle(p):
+    return p.__base__(), tuple([getattr(p, s) for s in p.__slots__[:-1]])
+
+
+def pickle_register(obj):
+    """Allow object to be pickled."""
+
+    util.copyreg.pickle(obj, _pickle)
+
+
+pickle_register(Selector)
+pickle_register(SelectorNull)
+pickle_register(SelectorTag)
+pickle_register(SelectorAttribute)
+pickle_register(SelectorContains)
+pickle_register(SelectorNth)
+pickle_register(SelectorLang)
+pickle_register(SelectorList)
diff --git a/lib/soupsieve/util.py b/lib/soupsieve/util.py
new file mode 100644
index 00000000..6158367a
--- /dev/null
+++ b/lib/soupsieve/util.py
@@ -0,0 +1,170 @@
+"""Utility."""
+from __future__ import unicode_literals
+from functools import wraps
+import warnings
+import sys
+import struct
+import os
+import re
+MODULE = os.path.dirname(__file__)
+
+PY3 = sys.version_info >= (3, 0)
+PY35 = sys.version_info >= (3, 5)
+PY37 = sys.version_info >= (3, 7)
+
+if PY3:
+    from functools import lru_cache  # noqa F401
+    import copyreg  # noqa F401
+    from collections.abc import Hashable, Mapping  # noqa F401
+
+    ustr = str
+    bstr = bytes
+    unichar = chr
+    string = str
+else:
+    from backports.functools_lru_cache import lru_cache  # noqa F401
+    import copy_reg as copyreg  # noqa F401
+    from collections import Hashable, Mapping  # noqa F401
+
+    ustr = unicode  # noqa: F821
+    bstr = str
+    unichar = unichr  # noqa: F821
+    string = basestring  # noqa: F821
+
+DEBUG = 0x00001
+
+RE_PATTERN_LINE_SPLIT = re.compile(r'(?:\r\n|(?!\r\n)[\n\r])|$')
+
+LC_A = ord('a')
+LC_Z = ord('z')
+UC_A = ord('A')
+UC_Z = ord('Z')
+
+
+def lower(string):
+    """Lower."""
+
+    new_string = []
+    for c in string:
+        o = ord(c)
+        new_string.append(chr(o + 32) if UC_A <= o <= UC_Z else c)
+    return ''.join(new_string)
+
+
+def upper(string):  # pragma: no cover
+    """Lower."""
+
+    new_string = []
+    for c in string:
+        o = ord(c)
+        new_string.append(chr(o - 32) if LC_A <= o <= LC_Z else c)
+    return ''.join(new_string)
+
+
+def uchr(i):
+    """Allow getting Unicode character on narrow python builds."""
+
+    try:
+        return unichar(i)
+    except ValueError:  # pragma: no cover
+        return struct.pack('i', i).decode('utf-32')
+
+
+def uord(c):
+    """Get Unicode ordinal."""
+
+    if len(c) == 2:  # pragma: no cover
+        high, low = [ord(p) for p in c]
+        ordinal = (high - 0xD800) * 0x400 + low - 0xDC00 + 0x10000
+    else:
+        ordinal = ord(c)
+
+    return ordinal
+
+
+class SelectorSyntaxError(SyntaxError):
+    """Syntax error in a CSS selector."""
+
+    def __init__(self, msg, pattern=None, index=None):
+        """Initialize."""
+
+        self.line = None
+        self.col = None
+        self.context = None
+
+        if pattern is not None and index is not None:
+            # Format pattern to show line and column position
+            self.context, self.line, self.col = get_pattern_context(pattern, index)
+            msg = '{}\n  line {}:\n{}'.format(msg, self.line, self.context)
+
+        super(SelectorSyntaxError, self).__init__(msg)
+
+
+def deprecated(message, stacklevel=2):  # pragma: no cover
+    """
+    Raise a `DeprecationWarning` when wrapped function/method is called.
+
+    Borrowed from https://stackoverflow.com/a/48632082/866026
+    """
+
+    def _decorator(func):
+        @wraps(func)
+        def _func(*args, **kwargs):
+            warnings.warn(
+                "'{}' is deprecated. {}".format(func.__name__, message),
+                category=DeprecationWarning,
+                stacklevel=stacklevel
+            )
+            return func(*args, **kwargs)
+        return _func
+    return _decorator
+
+
+def warn_deprecated(message, stacklevel=2):  # pragma: no cover
+    """Warn deprecated."""
+
+    warnings.warn(
+        message,
+        category=DeprecationWarning,
+        stacklevel=stacklevel
+    )
+
+
+def get_pattern_context(pattern, index):
+    """Get the pattern context."""
+
+    last = 0
+    current_line = 1
+    col = 1
+    text = []
+    line = 1
+
+    # Split pattern by newline and handle the text before the newline
+    for m in RE_PATTERN_LINE_SPLIT.finditer(pattern):
+        linetext = pattern[last:m.start(0)]
+        if not len(m.group(0)) and not len(text):
+            indent = ''
+            offset = -1
+            col = index - last + 1
+        elif last <= index < m.end(0):
+            indent = '--> '
+            offset = (-1 if index > m.start(0) else 0) + 3
+            col = index - last + 1
+        else:
+            indent = '    '
+            offset = None
+        if len(text):
+            # Regardless of whether we are presented with `\r\n`, `\r`, or `\n`,
+            # we will render the output with just `\n`. We will still log the column
+            # correctly though.
+            text.append('\n')
+        text.append('{}{}'.format(indent, linetext))
+        if offset is not None:
+            text.append('\n')
+            text.append(' ' * (col + offset) + '^')
+            line = current_line
+
+        current_line += 1
+        last = m.end(0)
+
+    return ''.join(text), line, col