Merge a91d9e1084 into e079ffbda6

* ignore matches within CDATA elements and comments

test/test_parsing.py

@@ -0,0 +1,359 @@
+import textwrap
+import unittest
+
+from yt_dlp.compat import compat_HTMLParseError
+from yt_dlp.parsing import (
+    MatchingElementParser,
+    HTMLIgnoreRanges,
+    HTMLTagParser,
+)
+
+extract_attributes = MatchingElementParser.extract_attributes
+get_element_by_attribute = MatchingElementParser.get_element_by_attribute
+get_element_by_class = MatchingElementParser.get_element_by_class
+get_element_html_by_attribute = MatchingElementParser.get_element_html_by_attribute
+get_element_html_by_class = MatchingElementParser.get_element_html_by_class
+get_element_text_and_html_by_tag = MatchingElementParser.get_element_text_and_html_by_tag
+get_elements_by_attribute = MatchingElementParser.get_elements_by_attribute
+get_elements_by_class = MatchingElementParser.get_elements_by_class
+get_elements_html_by_attribute = MatchingElementParser.get_elements_html_by_attribute
+get_elements_html_by_class = MatchingElementParser.get_elements_html_by_class
+get_elements_text_and_html_by_attribute = MatchingElementParser.get_elements_text_and_html_by_attribute
+get_elements_text_and_html_by_tag = MatchingElementParser.get_elements_text_and_html_by_tag
+
+
+class TestParsing(unittest.TestCase):
+    def test_extract_attributes(self):
+        self.assertEqual(extract_attributes('<e x="y">'), {'x': 'y'})
+        self.assertEqual(extract_attributes("<e x='y'>"), {'x': 'y'})
+        self.assertEqual(extract_attributes('<e x=y>'), {'x': 'y'})
+        self.assertEqual(extract_attributes('<e x="a \'b\' c">'), {'x': "a 'b' c"})
+        self.assertEqual(extract_attributes('<e x=\'a "b" c\'>'), {'x': 'a "b" c'})
+        self.assertEqual(extract_attributes('<e x="&#121;">'), {'x': 'y'})
+        self.assertEqual(extract_attributes('<e x="&#x79;">'), {'x': 'y'})
+        self.assertEqual(extract_attributes('<e x="&amp;">'), {'x': '&'})  # XML
+        self.assertEqual(extract_attributes('<e x="&quot;">'), {'x': '"'})
+        self.assertEqual(extract_attributes('<e x="&pound;">'), {'x': '£'})  # HTML 3.2
+        self.assertEqual(extract_attributes('<e x="&lambda;">'), {'x': 'λ'})  # HTML 4.0
+        self.assertEqual(extract_attributes('<e x="&foo">'), {'x': '&foo'})
+        self.assertEqual(extract_attributes('<e x="\'">'), {'x': "'"})
+        self.assertEqual(extract_attributes('<e x=\'"\'>'), {'x': '"'})
+        self.assertEqual(extract_attributes('<e x >'), {'x': None})
+        self.assertEqual(extract_attributes('<e x=y a>'), {'x': 'y', 'a': None})
+        self.assertEqual(extract_attributes('<e x= y>'), {'x': 'y'})
+        self.assertEqual(extract_attributes('<e x=1 y=2 x=3>'), {'y': '2', 'x': '3'})
+        self.assertEqual(extract_attributes('<e \nx=\ny\n>'), {'x': 'y'})
+        self.assertEqual(extract_attributes('<e \nx=\n"y"\n>'), {'x': 'y'})
+        self.assertEqual(extract_attributes("<e \nx=\n'y'\n>"), {'x': 'y'})
+        self.assertEqual(extract_attributes('<e \nx="\ny\n">'), {'x': '\ny\n'})
+        self.assertEqual(extract_attributes('<e CAPS=x>'), {'caps': 'x'})  # Names lowercased
+        self.assertEqual(extract_attributes('<e x=1 X=2>'), {'x': '2'})
+        self.assertEqual(extract_attributes('<e X=1 x=2>'), {'x': '2'})
+        self.assertEqual(extract_attributes('<e _:funny-name1=1>'), {'_:funny-name1': '1'})
+        self.assertEqual(extract_attributes('<e x="Fáilte 世界 \U0001f600">'), {'x': 'Fáilte 世界 \U0001f600'})
+        self.assertEqual(extract_attributes('<e x="décompose&#769;">'), {'x': 'décompose\u0301'})
+        # "Narrow" Python builds don't support unicode code points outside BMP.
+        try:
+            chr(0x10000)
+            supports_outside_bmp = True
+        except ValueError:
+            supports_outside_bmp = False
+        if supports_outside_bmp:
+            self.assertEqual(extract_attributes('<e x="Smile &#128512;!">'), {'x': 'Smile \U0001f600!'})
+        # Malformed HTML should not break attributes extraction on older Python
+        self.assertEqual(extract_attributes('<mal"formed/>'), {})
+
+    GET_ELEMENT_BY_CLASS_TEST_STRING = '''
+        <span class="foo bar">nice</span>
+        <div class="foo bar">also nice</div>
+    '''
+
+    def test_get_element_by_class(self):
+        html = self.GET_ELEMENT_BY_CLASS_TEST_STRING
+
+        self.assertEqual(get_element_by_class('foo', html), 'nice')
+        self.assertEqual(get_element_by_class('no-such-class', html), None)
+
+    def test_get_element_html_by_class(self):
+        html = self.GET_ELEMENT_BY_CLASS_TEST_STRING
+
+        self.assertEqual(get_element_html_by_class('foo', html),
+                         '<span class="foo bar">nice</span>')
+        self.assertEqual(get_element_by_class('no-such-class', html), None)
+
+    GET_ELEMENT_BY_ATTRIBUTE_TEST_STRING = '''
+        <div itemprop="author" itemscope>foo</div>
+    '''
+
+    def test_get_element_by_attribute(self):
+        html = self.GET_ELEMENT_BY_CLASS_TEST_STRING
+
+        self.assertEqual(get_element_by_attribute('class', 'foo bar', html), 'nice')
+        self.assertEqual(get_element_by_attribute('class', 'foo', html), None)
+        self.assertEqual(get_element_by_attribute('class', 'no-such-foo', html), None)
+        self.assertEqual(get_element_by_attribute('class', 'foo bar', html, tag='div'), 'also nice')
+
+        html = self.GET_ELEMENT_BY_ATTRIBUTE_TEST_STRING
+
+        self.assertEqual(get_element_by_attribute('itemprop', 'author', html), 'foo')
+
+    def test_get_element_html_by_attribute(self):
+        html = self.GET_ELEMENT_BY_CLASS_TEST_STRING
+
+        self.assertEqual(get_element_html_by_attribute('class', 'foo bar', html),
+                         '<span class="foo bar">nice</span>')
+        self.assertEqual(get_element_html_by_attribute('class', 'foo', html), None)
+        self.assertEqual(get_element_html_by_attribute('class', 'no-such-foo', html), None)
+
+        html = self.GET_ELEMENT_BY_ATTRIBUTE_TEST_STRING
+
+        self.assertEqual(get_element_html_by_attribute('itemprop', 'author', html), html.strip())
+
+    GET_ELEMENTS_BY_CLASS_TEST_STRING = '''
+        <span class="foo bar">nice</span>
+        <span class="foo bar">also nice</span>
+    '''
+    GET_ELEMENTS_BY_CLASS_RES = [
+        '<span class="foo bar">nice</span>',
+        '<span class="foo bar">also nice</span>'
+    ]
+
+    def test_get_elements_by_class(self):
+        html = self.GET_ELEMENTS_BY_CLASS_TEST_STRING
+
+        self.assertEqual(get_elements_by_class('foo', html), ['nice', 'also nice'])
+        self.assertEqual(get_elements_by_class('no-such-class', html), [])
+
+    def test_get_elements_html_by_class(self):
+        html = self.GET_ELEMENTS_BY_CLASS_TEST_STRING
+
+        self.assertEqual(get_elements_html_by_class('foo', html), self.GET_ELEMENTS_BY_CLASS_RES)
+        self.assertEqual(get_elements_html_by_class('no-such-class', html), [])
+
+    def test_get_elements_by_attribute(self):
+        html = self.GET_ELEMENTS_BY_CLASS_TEST_STRING
+
+        self.assertEqual(get_elements_by_attribute('class', 'foo bar', html), ['nice', 'also nice'])
+        self.assertEqual(get_elements_by_attribute('class', 'foo', html), [])
+        self.assertEqual(get_elements_by_attribute('class', 'no-such-foo', html), [])
+
+    def test_get_elements_html_by_attribute(self):
+        html = self.GET_ELEMENTS_BY_CLASS_TEST_STRING
+
+        self.assertEqual(get_elements_html_by_attribute('class', 'foo bar', html),
+                         self.GET_ELEMENTS_BY_CLASS_RES)
+        self.assertEqual(get_elements_html_by_attribute('class', 'foo', html), [])
+        self.assertEqual(get_elements_html_by_attribute('class', 'no-such-foo', html), [])
+
+    def test_get_elements_text_and_html_by_attribute(self):
+        html = self.GET_ELEMENTS_BY_CLASS_TEST_STRING
+
+        self.assertEqual(
+            get_elements_text_and_html_by_attribute('class', 'foo bar', html),
+            list(zip(['nice', 'also nice'], self.GET_ELEMENTS_BY_CLASS_RES)))
+        self.assertEqual(get_elements_text_and_html_by_attribute('class', 'foo', html), [])
+        self.assertEqual(get_elements_text_and_html_by_attribute('class', 'no-such-foo', html), [])
+
+        self.assertEqual(get_elements_text_and_html_by_attribute(
+            'class', 'foo', '<a class="foo">nice</a><span class="foo">not nice</span>', tag='a'),
+            [('nice', '<a class="foo">nice</a>')])
+
+    def test_get_element_text_and_html_by_tag(self):
+        get_element_by_tag_test_string = '''
+        random text lorem ipsum</p>
+        <div>
+            this should be returned
+            <span>this should also be returned</span>
+            <div>
+                this should also be returned
+            </div>
+            closing tag above should not trick, so this should also be returned
+        </div>
+        but this text should not be returned
+        '''
+        html = textwrap.indent(textwrap.dedent(get_element_by_tag_test_string), ' ' * 4)
+        get_element_by_tag_res_outerdiv_html = html.strip()[32:276]
+        get_element_by_tag_res_outerdiv_text = get_element_by_tag_res_outerdiv_html[5:-6]
+        get_element_by_tag_res_innerspan_html = html.strip()[78:119]
+        get_element_by_tag_res_innerspan_text = get_element_by_tag_res_innerspan_html[6:-7]
+
+        self.assertEqual(
+            get_element_text_and_html_by_tag('div', html),
+            (get_element_by_tag_res_outerdiv_text, get_element_by_tag_res_outerdiv_html))
+        self.assertEqual(
+            get_element_text_and_html_by_tag('span', html),
+            (get_element_by_tag_res_innerspan_text, get_element_by_tag_res_innerspan_html))
+        self.assertIsNone(get_element_text_and_html_by_tag('article', html))
+
+    def test_get_elements_text_and_html_by_tag(self):
+        class StrictParser(MatchingElementParser):
+            STRICT = True
+
+        test_string = '''
+            <img src="a.png">
+            <img src="b.png" />
+            <span>ignore</span>
+        '''
+        items = get_elements_text_and_html_by_tag('img', test_string)
+        self.assertEqual(items, [('', '<img src="a.png">'), ('', '<img src="b.png" />')])
+
+        self.assertEqual(
+            StrictParser.get_element_text_and_html_by_tag('use', '<use><img></use>'),
+            ('<img>', '<use><img></use>'))
+
+    def test_get_element_text_and_html_by_tag_malformed(self):
+        inner_text = 'inner text'
+        malnested_elements = f'<malnested_a><malnested_b>{inner_text}</malnested_a></malnested_b>'
+        commented_html = '<!--<div>inner comment</div>-->'
+        outerdiv_html = f'<div>{malnested_elements}</div>'
+        html = f'{commented_html}{outerdiv_html}'
+
+        self.assertEqual(
+            get_element_text_and_html_by_tag('div', html), (malnested_elements, outerdiv_html))
+        self.assertEqual(
+            get_element_text_and_html_by_tag('malnested_a', html),
+            (f'<malnested_b>{inner_text}',
+             f'<malnested_a><malnested_b>{inner_text}</malnested_a>'))
+        self.assertEqual(
+            get_element_text_and_html_by_tag('malnested_b', html),
+            (f'{inner_text}</malnested_a>',
+             f'<malnested_b>{inner_text}</malnested_a></malnested_b>'))
+        self.assertEqual(
+            get_element_text_and_html_by_tag('orphan', f'<orphan>{html}'), ('', '<orphan>'))
+        self.assertIsNone(get_element_text_and_html_by_tag('orphan', f'{html}</orphan>'))
+
+        # ignore case on tags
+        ci_html = f'<SpAn>{html}</sPaN>'
+        self.assertEqual(get_element_text_and_html_by_tag('span', ci_html), (html, ci_html))
+
+    def test_strict_html_parsing(self):
+        class StrictTagParser(HTMLTagParser):
+            STRICT = True
+
+        parser = StrictTagParser()
+        with self.assertRaisesRegex(compat_HTMLParseError, "stray closing tag 'p'"):
+            parser.taglist('</p>', reset=True)
+        with self.assertRaisesRegex(compat_HTMLParseError, "unclosed tag 'p', 'div'"):
+            parser.taglist('<div><p>', reset=True)
+        with self.assertRaisesRegex(compat_HTMLParseError, "malnested closing tag 'div', expected after '</p>'"):
+            parser.taglist('<div><p></div></p>', reset=True)
+        with self.assertRaisesRegex(compat_HTMLParseError, "malnested closing tag 'div', expected after '</p>'"):
+            parser.taglist('<div><p>/p></div>', reset=True)
+        with self.assertRaisesRegex(compat_HTMLParseError, "malformed closing tag 'p<<'"):
+            parser.taglist('<div><p></p<< </div>', reset=True)
+        with self.assertRaisesRegex(compat_HTMLParseError, "stray closing tag 'img'"):
+            parser.taglist('<img>must be empty</img>', reset=True)
+
+    def test_relaxed_html_parsing(self):
+        Tag = HTMLTagParser.Tag
+        parser = HTMLTagParser()
+
+        self.assertEqual(parser.taglist('</p>', reset=True), [])
+
+        tags = parser.taglist('<div><p>', reset=True)
+        self.assertEqual(tags, [Tag('div'), Tag('p')])
+        self.assertEqual(tags[0].text_and_html(), ('', '<div>'))
+        self.assertEqual(tags[1].text_and_html(), ('', '<p>'))
+
+        tags = parser.taglist('<div><p></div></p>', reset=True)
+        self.assertEqual(tags, [Tag('div'), Tag('p')])
+        self.assertEqual(tags[0].text_and_html(), ('<p>', '<div><p></div>'))
+        self.assertEqual(tags[1].text_and_html(), ('</div>', '<p></div></p>'))
+
+        tags = parser.taglist('<div><p>/p></div>', reset=True)
+        self.assertEqual(tags, [Tag('div'), Tag('p')])
+        self.assertEqual(tags[0].text_and_html(), ('<p>/p>', '<div><p>/p></div>'))
+        self.assertEqual(tags[1].text_and_html(), ('', '<p>'))
+
+        tags = parser.taglist('<div><p>paragraph</p<ignored></div>', reset=True)
+        self.assertEqual(tags, [Tag('div'), Tag('p')])
+        self.assertEqual(tags[0].text_and_html(),
+                         ('<p>paragraph</p<ignored>', '<div><p>paragraph</p<ignored></div>'))
+        self.assertEqual(tags[1].text_and_html(), ('paragraph', '<p>paragraph</p<ignored>'))
+
+        tags = parser.taglist('<img width="300px">must be empty</img>', reset=True)
+        self.assertEqual(tags, [Tag('img')])
+        self.assertEqual(tags[0].text_and_html(), ('', '<img width="300px">'))
+
+    def test_compliant_html_parsing(self):
+        # certain elements don't need to be closed (see HTMLTagParser.VOID_TAGS)
+        Tag = HTMLTagParser.Tag
+        html = '''
+            no error without closing tag: <img>
+            self closing is ok: <img />
+        '''
+        parser = HTMLTagParser()
+        tags = parser.taglist(html, reset=True)
+        self.assertEqual(tags, [Tag('img'), Tag('img')])
+
+        # don't get fooled by '>' in attributes
+        html = '''<img greater_a='1>0' greater_b="1>0">'''
+        tags = parser.taglist(html, reset=True)
+        self.assertEqual(tags[0].text_and_html(), ('', html))
+
+    def test_tag_return_order(self):
+        Tag = HTMLTagParser.Tag
+        html = '''
+        <t0>
+            <t1>
+                <t2>
+                    <t3 /> <t4 />
+                </t2>
+            </t1>
+            <t5>
+                <t6 />
+            </t5>
+        </t0>
+        <t7>
+            <t8 />
+        </t7>
+        '''
+        parser = HTMLTagParser()
+        tags = parser.taglist(html, reset=True)
+        self.assertEqual(
+            str(tags), str([Tag('t0'), Tag('t1'), Tag('t2'), Tag('t3'), Tag('t4'),
+                            Tag('t5'), Tag('t6'), Tag('t7'), Tag('t8')]))
+
+        tags = parser.taglist(html, reset=True, depth_first=True)
+        self.assertEqual(
+            str(tags), str([Tag('t3'), Tag('t4'), Tag('t2'), Tag('t1'), Tag('t6'),
+                            Tag('t5'), Tag('t0'), Tag('t8'), Tag('t7')]))
+
+        # return tags in nested order
+        tags = parser.taglist(html, reset=True, depth_first=None)
+        self.assertEqual(
+            str(tags), str([
+                [Tag('t0'),
+                 [Tag('t1'),
+                  [Tag('t2'), [Tag('t3')], [Tag('t4')]]],
+                 [Tag('t5'), [Tag('t6')]]],
+                [Tag('t7'), [Tag('t8')]]]))
+
+    def test_html_ignored_ranges(self):
+        def mark_comments(_string, char='^', nochar='-'):
+            cmts = HTMLIgnoreRanges(_string)
+            return "".join(char if _idx in cmts else nochar for _idx in range(len(_string)))
+
+        html_string = '''
+        no              comments         in            this              line
+        ---------------------------------------------------------------------
+        <!--                 whole line represents a comment              -->
+        ----^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^---
+        before <!--                      comment                  -->   after
+        -----------^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^-----------
+        this is a leftover comment -->     <!-- a new comment without closing
+        ^^^^^^^^^^^^^^^^^^^^^^^^^^^------------^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+        here   is   <!-- a comment -->   and   <!-- another comment --!>  end
+        ----------------^^^^^^^^^^^----------------^^^^^^^^^^^^^^^^^---------
+        <script> ignore here </script>            <SCRIPT> and here </SCRIPT>
+        --------^^^^^^^^^^^^^-----------------------------^^^^^^^^^^---------
+        '''
+
+        lines = textwrap.dedent(html_string).strip().splitlines()
+        for line, marker in zip(lines[0::2], lines[1::2]):
+            self.assertEqual((line, mark_comments(line)), (line, marker))
+
+        # yet we must be able to match script elements
+        test_string = '''<script type="text/javascript">var foo = 'bar';</script>'''
+        items = get_element_text_and_html_by_tag('script', test_string)
+        self.assertEqual(items, ("var foo = 'bar';", test_string))

yt_dlp/parsing.py

@@ -0,0 +1,348 @@
+import collections
+import contextlib
+import itertools
+import re
+from html.parser import HTMLParser
+
+from .compat import compat_HTMLParseError
+from .utils import orderedSet
+
+
+class HTMLIgnoreRanges:
+    """check if an offset is within CDATA content elements (script, style) or XML comments
+
+        note:
+            * given offsets must be in increasing order
+            * no detection of nested constructs (e.g. comments within script tags)
+
+        usage:
+            ranges = HTMLIgnoreRanges(html)
+            if offset in ranges:
+                ...
+    """
+    REGEX = re.compile(r'<!--|--!?>|</?\s*(?:script|style)\b[^>]*>', flags=re.IGNORECASE)
+
+    def __init__(self, html):
+        self.html = html
+        self._last_match = None
+        self._final = False
+
+    def __contains__(self, offset):
+        assert isinstance(offset, int)
+
+        if not self._final and (self._last_match is None or offset >= self._last_match.end()):
+            match = self.REGEX.search(self.html, offset)
+            if match:
+                self._last_match = match
+            else:
+                self._final = True
+
+        if self._last_match is None:
+            return False
+        match_string = self._last_match.group()
+        if match_string.startswith('</') or match_string in ('-->', '--!>'):
+            return offset < self._last_match.start()
+        return offset >= self._last_match.end()
+
+
+class HTMLTagParser(HTMLParser):
+    """HTML parser which returns found elements as instances of 'Tag'
+    when STRICT=True can raise compat_HTMLParseError() on malformed HTML elements
+
+    usage:
+        parser = HTMLTagParser()
+        for tag_obj in parser.taglist(html):
+            tag_obj.text_and_html()
+
+    """
+
+    STRICT = False
+    ANY_TAG_REGEX = re.compile(r'''<(?:"[^"]*"|'[^']*'|[^"'>])*?>''')
+    VOID_TAGS = {
+        'area', 'base', 'br', 'col', 'embed', 'hr', 'img', 'input',
+        'keygen', 'link', 'meta', 'param', 'source', 'track', 'wbr',
+    }
+
+    class Tag:
+        __slots__ = 'name', 'string', 'attrs', '_openrange', '_closerange'
+
+        def __init__(self, name, *, string='', attrs=()):
+            self.name = name
+            self.string = string
+            self.attrs = tuple(attrs)
+            self._openrange = None
+            self._closerange = None
+
+        def __str__(self):
+            return self.name
+
+        def __repr__(self):
+            return f'{self.__class__.__name__}({str(self)!r})'
+
+        def __eq__(self, other):
+            return self.name == other
+
+        def openrange(self, offset, startlen=0):
+            if isinstance(offset, slice):
+                self._openrange = offset
+            else:
+                self._openrange = slice(offset, offset + startlen)
+
+        def closerange(self, offset, stoplen=0):
+            if isinstance(offset, slice):
+                self._closerange = offset
+            else:
+                self._closerange = slice(offset, offset + stoplen)
+
+        def opentag(self):
+            return self.string[self._openrange] if self._openrange else ''
+
+        def html(self):
+            if not self._openrange:
+                return ''
+            if self._closerange:
+                return self.string[self._openrange.start:self._closerange.stop]
+            return self.string[self._openrange]
+
+        def text(self):
+            if self._openrange and self._closerange:
+                return self.string[self._openrange.stop:self._closerange.start]
+            return ''
+
+        def text_and_html(self):
+            return self.text(), self.html()
+
+    class AbortException(Exception):
+        pass
+
+    def __init__(self):
+        self.tagstack = collections.deque()
+        self._nestedtags = [[]]
+        super().__init__()
+        self._offset = self.offset
+
+    def predicate(self, tag, attrs):
+        """ return True for every encountered opening tag that should be processed """
+        return True
+
+    def callback(self, tag_obj):
+        """ this will be called when the requested tag is closed """
+
+    def reset(self):
+        super().reset()
+        self.tagstack.clear()
+
+    def taglist(self, data, reset=True, depth_first=False):
+        """ parse data and return found tag objects
+        @param data:    html string
+        @param reset:   reset state
+        @param depth_first: return order: as opened (False), as closed (True), nested (None)
+        @return: list of Tag objects
+        """
+        def flatten(_list, first=True):
+            rlist = _list if first or not depth_first else itertools.chain(_list[1:], _list[:1])
+            for item in rlist:
+                if isinstance(item, list):
+                    yield from flatten(item, first=False)
+                else:
+                    yield item
+
+        if reset:
+            self.reset()
+        with contextlib.suppress(HTMLTagParser.AbortException):
+            self.feed(data)
+        if self.STRICT and self.tagstack:
+            orphans = ', '.join(map(repr, map(str, orderedSet(self.tagstack, lazy=True))))
+            raise compat_HTMLParseError(f'unclosed tag {orphans}')
+        taglist = self._nestedtags[0] if depth_first is None else list(flatten(self._nestedtags[0]))
+        self._nestedtags = [[]]
+        return taglist
+
+    def updatepos(self, i, j):
+        offset = self._offset = super().updatepos(i, j)
+        return offset
+
+    def handle_starttag(self, tag, attrs):
+        try:
+            # we use internal variable for performance reasons
+            tag_text = getattr(self, '_HTMLParser__starttag_text')
+        except AttributeError:
+            tag_text = HTMLTagParser.ANY_TAG_REGEX.match(self.rawdata[self._offset:]).group()
+
+        tag_obj = tag
+        tag_is_open = not (tag_text.endswith('/>') or tag in self.VOID_TAGS)
+        if self.predicate(tag, attrs):
+            tag_obj = self.Tag(tag, string=self.rawdata, attrs=attrs)
+            tag_obj.openrange(self._offset, len(tag_text))
+            nesting = [tag_obj]
+            self._nestedtags[-1].append(nesting)
+            if tag_is_open:
+                self._nestedtags.append(nesting)
+            else:
+                self.callback(tag_obj)
+        if tag_is_open:
+            self.tagstack.appendleft(tag_obj)
+
+    handle_startendtag = handle_starttag
+
+    def handle_endtag(self, tag):
+        if '<' in tag:
+            if self.STRICT:
+                raise compat_HTMLParseError(f'malformed closing tag {tag!r}')
+            tag = tag[:tag.index('<')]
+
+        try:
+            idx = self.tagstack.index(tag)
+            if self.STRICT and idx:
+                open_tags = ''.join(f'</{tag}>' for tag in itertools.islice(self.tagstack, idx))
+                raise compat_HTMLParseError(
+                    f'malnested closing tag {tag!r}, expected after {open_tags!r}')
+            tag_obj = self.tagstack[idx]
+            self.tagstack.remove(tag)
+            if isinstance(tag_obj, self.Tag):
+                tag_obj.closerange(slice(self._offset, self.rawdata.find('>', self._offset) + 1))
+                self._nestedtags.pop()
+                self.callback(tag_obj)
+        except ValueError as exc:
+            if isinstance(exc, compat_HTMLParseError):
+                raise
+            if self.STRICT:
+                raise compat_HTMLParseError(f'stray closing tag {tag!r}') from exc
+
+
+class MatchingElementParser(HTMLTagParser):
+    """ optimized version of HTMLTagParser
+    """
+    def __init__(self, matchfunc):
+        super().__init__()
+        self.matchfunc = matchfunc
+        self.found_none = True
+
+    def reset(self):
+        super().reset()
+        self.found_none = True
+
+    def callback(self, tag_obj):
+        raise self.AbortException()
+
+    def predicate(self, tag, attrs):
+        if self.found_none and self.matchfunc(tag, attrs):
+            self.found_none = False
+            return True
+        return False
+
+    @staticmethod
+    def class_value_regex(class_name):
+        return rf'[\w\s\-]*(?<![\w\-]){re.escape(class_name)}(?![\w\-])[\w\s\-]*'
+
+    @staticmethod
+    def matching_tag_regex(tag, attribute, value_regex, escape=True):
+        if isinstance(value_regex, re.Pattern):
+            value_regex = value_regex.pattern
+        elif escape:
+            value_regex = re.escape(value_regex)
+
+        return rf'''(?x)
+            <(?i:{tag})
+             (?:\s(?:[^>"'\\]|"[^"\\]*"|'[^'\\]*')*)?
+             \s{re.escape(attribute)}\s*=\s*(?P<_q>['"])(?-x:{value_regex})(?P=_q)
+            '''
+
+    @classmethod
+    def iter_tags(cls, regex, html, *, matchfunc):
+        ignored = HTMLIgnoreRanges(html)
+        parser = cls(matchfunc)
+        for match in re.finditer(regex, html):
+            if match.start() not in ignored:
+                yield from parser.taglist(html[match.start():], reset=True)
+
+    @classmethod
+    def tags_by_name(cls, tag, html):
+        def matchfunc(tag_str, _attrs):
+            return tag_str == tag
+
+        tag_regex = rf'''<\s*(?i:{re.escape(tag)})(?:\s(?:[^>"'\\]|"[^"\\]*"|'[^'\\]*')*)?>'''
+        yield from cls.iter_tags(tag_regex, html, matchfunc=matchfunc)
+
+    @classmethod
+    def tags_by_attribute(cls, attribute, value, html, *, tag=r'[\w:.-]+', escape_value=True):
+        def matchfunc(_tag_str, attrs):
+            return any(attr == attribute and re.fullmatch(value, value_str)
+                       for attr, value_str in attrs)
+
+        tag_regex = cls.matching_tag_regex(tag, attribute, value, escape_value)
+        yield from cls.iter_tags(tag_regex, html, matchfunc=matchfunc)
+
+    @classmethod
+    def extract_attributes(cls, html):
+        attr_dict = {}
+
+        def matchfunc(_tag, attrs):
+            attr_dict.update(attrs)
+            raise cls.AbortException()
+
+        with contextlib.suppress(cls.AbortException):
+            cls(matchfunc).feed(html)
+
+        return attr_dict
+
+    @classmethod
+    def get_elements_text_and_html_by_tag(cls, tag, html):
+        return [tag.text_and_html() for tag in cls.tags_by_name(tag, html)]
+
+    @classmethod
+    def get_element_text_and_html_by_tag(cls, tag, html):
+        tag = next(cls.tags_by_name(tag, html), None)
+        return tag and tag.text_and_html()
+
+    @classmethod
+    def get_elements_text_and_html_by_attribute(cls, *args, **kwargs):
+        return [tag.text_and_html() for tag in cls.tags_by_attribute(*args, **kwargs)]
+
+    @classmethod
+    def get_elements_by_attribute(cls, *args, **kwargs):
+        return [tag.text() for tag in cls.tags_by_attribute(*args, **kwargs)]
+
+    @classmethod
+    def get_elements_html_by_attribute(cls, *args, **kwargs):
+        return [tag.html() for tag in cls.tags_by_attribute(*args, **kwargs)]
+
+    @classmethod
+    def get_element_by_attribute(cls, *args, **kwargs):
+        tag = next(cls.tags_by_attribute(*args, **kwargs), None)
+        return tag and tag.text()
+
+    @classmethod
+    def get_element_html_by_attribute(cls, *args, **kwargs):
+        tag = next(cls.tags_by_attribute(*args, **kwargs), None)
+        return tag and tag.html()
+
+    @classmethod
+    def get_elements_by_class(cls, class_name, html):
+        value = cls.class_value_regex(class_name)
+        return [tag.text() for tag
+                in cls.tags_by_attribute('class', value, html, escape_value=False)]
+
+    @classmethod
+    def get_elements_html_by_class(cls, class_name, html):
+        value = cls.class_value_regex(class_name)
+        return [tag.html() for tag
+                in cls.tags_by_attribute('class', value, html, escape_value=False)]
+
+    @classmethod
+    def get_elements_text_and_html_by_class(cls, class_name, html):
+        value = cls.class_value_regex(class_name)
+        return [tag.text_and_html() for tag
+                in cls.tags_by_attribute('class', value, html, escape_value=False)]
+
+    @classmethod
+    def get_element_html_by_class(cls, class_name, html):
+        value = cls.class_value_regex(class_name)
+        tag = next(cls.tags_by_attribute('class', value, html, escape_value=False), None)
+        return tag and tag.html()
+
+    @classmethod
+    def get_element_by_class(cls, class_name, html):
+        value = cls.class_value_regex(class_name)
+        tag = next(cls.tags_by_attribute('class', value, html, escape_value=False), None)
+        return tag and tag.text()

yt_dlp/utils/_utils.py

@@ -408,17 +408,13 @@ class HTMLBreakOnClosingTagParser(html.parser.HTMLParser):
         pass
 
     def handle_starttag(self, tag, _):
-        self.tagstack.append(tag)
+        self.tagstack.appendleft(tag)
 
     def handle_endtag(self, tag):
         if not self.tagstack:
             raise compat_HTMLParseError('no tags in the stack')
-        while self.tagstack:
-            inner_tag = self.tagstack.pop()
-            if inner_tag == tag:
-                break
-        else:
-            raise compat_HTMLParseError(f'matching opening tag for closing {tag} tag not found')
+        with contextlib.suppress(ValueError):
+            self.tagstack.remove(tag)
         if not self.tagstack:
             raise self.HTMLBreakOnClosingTagException
 
@@ -452,6 +448,8 @@ def get_element_text_and_html_by_tag(tag, html):
             next_closing_tag_end = next_closing_tag_start + len(closing_tag)
             try:
                 parser.feed(html[offset:offset + next_closing_tag_end])
+                if tag not in parser.tagstack:
+                    raise HTMLBreakOnClosingTagParser.HTMLBreakOnClosingTagException()
                 offset += next_closing_tag_end
             except HTMLBreakOnClosingTagParser.HTMLBreakOnClosingTagException:
                 return html[content_start:offset + next_closing_tag_start], \