ExtendedCss

AdGuard's TypeScript library for non-standard element selecting and applying CSS styles with extended properties.

The idea of extended capabilities is an opportunity to match DOM elements with selectors based on their own representation (style, text content, etc.) or relations with other elements. There is also an opportunity to apply styles with non-standard CSS properties.

• Extended capabilities
  • Limitations
  • Pseudo-class :has()
  • Pseudo-class :contains()
  • Pseudo-class :matches-css()
  • Pseudo-class :matches-attr()
  • Pseudo-class :matches-property()
  • Pseudo-class :xpath()
  • Pseudo-class :nth-ancestor()
  • Pseudo-class :upward()
  • Pseudo-class :remove() and pseudo-property remove
  • Pseudo-class :is()
  • Pseudo-class :not()
  • Pseudo-class :if-not() (deprecated)
  • Selectors debugging mode
  • Backward compatible syntax
• How to build
• How to test
• Usage
  • API description
  • Constructor
  • init()
  • apply() and dispose()
  • query()
  • validate()
  • EXTENDED_CSS_VERSION
  • Debugging extended selectors
• Projects using ExtendedCss
• Browser compatibility
• Known issues

Extended capabilities

Some pseudo-classes does not require a selector before them. Still adding a universal selector * makes an extended selector easier to read, even though it has no effect on the matching behavior. So the selector #block :has(> .inner) works exactly like #block *:has(> .inner) but second one is more obvious.

Pseudo-class names are case-insensitive, e.g. :HAS() works as :has(). Still the lower-case names are used commonly.

Limitations

1. CSS comments and at-rules are not supported.

2. Specific pseudo-class may have its own limitations: :has(), :xpath(), :nth-ancestor(), :upward(), :is(), :not(), and :remove().

Pseudo-class :has()

Draft CSS 4.0 specification describes the :has() pseudo-class. Unfortunately, it is not yet supported by all popular browsers.

Rules with the :has() pseudo-class should use native implementation of :has() if they use ## marker and if it is possible, i.e. with no other extended selectors inside. To force applying ExtendedCss rules with :has(), use #?#/#$?# marker explicitly.

Synonym :-abp-has is supported by ExtendedCss for better compatibility.

:if() is no longer supported as a synonym for :has().

Syntax

[target]:has(selector)

• target — optional, standard or extended CSS selector, can be missed for checking any element
• selector — required, standard or extended CSS selector

The pseudo-class :has() selects the target elements that fit to the selector. Also the selector can start with a combinator.

A selector list can be set in selector as well. In this case all selectors in the list are being matched for now. It is one of the known issues and will be fixed for <forgiving-relative-selector-list> as argument.

Limitations and notes

Usage of the :has() pseudo-class is restricted for some cases (2, 3):

• disallow :has() inside the pseudos accepting only compound selectors;
• disallow :has() after regular pseudo-elements.

Native :has() pseudo-class does not allow :has(), :is(), :where() inside :has() argument to avoid increasing the :has() invalidation complexity (case 1). But ExtendedCss did not have such limitation earlier and filter lists already contain such rules, so we have not added this limitation to ExtendedCss and allow to use :has() inside :has() as it was possible before. To use it, just force ExtendedCss usage by setting #?#/#$?# rule marker.

Native implementation does not allow any usage of :scope inside :has() argument ([1], [2]). Still, there are some such rules in filter lists: div:has(:scope > a) which we continue to support by simply converting them to div:has(> a), as it used to be done previously.

Examples

div:has(.banner) selects all div elements which include an element with the banner class:

<!-- HTML code -->
<div>Not selected</div>
<div>Selected
  <span class="banner">inner element</span>
</div>

div:has(> .banner) selects all div elements which include an banner class element as a direct child of div:

<!-- HTML code -->
<div>Not selected</div>
<div>Selected
  <p class="banner">child element</p>
</div>

div:has(+ .banner) selects all div elements preceding banner class element which immediately follows the div and both are children of the same parent:

<!-- HTML code -->
<div>Not selected</div>
<div>Selected</div>
<p class="banner">adjacent sibling</p>
<span>Not selected</span>

div:has(~ .banner) selects all div elements preceding banner class element which follows the div but not necessarily immediately and both are children of the same parent:

<!-- HTML code -->
<div>Not selected</div>
<div>Selected</div>
<span>Not selected</span>
<p class="banner">general sibling</p>

div:has(span, .banner) selects all div elements which include both span element and banner class element:

<!-- HTML code -->
<div>Not selected</div>
<div>Selected
  <span>child span</span>
  <p class="banner">child .banner</p>
</div>

Backward compatible syntax for :has() is supported but not recommended.

Pseudo-class :contains()

This pseudo-class principle is very simple: it allows to select the elements that contain specified text or which content matches a specified regular expression. Regexp flags are supported.

Pseudo-class :contains() uses the textContent element property for matching, not the innerHTML.

Synonyms :-abp-contains and :has-text are supported for better compatibility.

Syntax

[target]:contains(match)

• target — optional, standard or extended CSS selector, can be missed for checking any element
• match — required, string or regular expression for matching element textContent

Regexp flags are supported for match.

Examples

For such DOM:

<!-- HTML code -->
<div>Not selected</div>
<div id="match">Selected as IT contains "banner"</div>
<div>Not selected <div class="banner"></div></div>

div#match can be selected by any on these extended selectors:

! plain text
div:contains(banner)

! regular expression
div:contains(/as .* banner/)

! regular expression with flags
div:contains(/it .* banner/gi)

Only the div with id=match is selected because the next element does not contain any text, and banner is a part of code, not a text.

Backward compatible syntax for :contains() is supported but not recommended.

Pseudo-class :matches-css()

Pseudo-class :matches-css() allows to match the element by its current style properties. The work of the pseudo-class is based on using the Window.getComputedStyle() method.

Syntax

[target]:matches-css([pseudo-element, ] property: pattern)

• target — optional, standard or extended CSS selector, can be missed for checking any element
• pseudo-element — optional, valid standard pseudo-element, e.g. before, after, first-line, etc.
• property — required, a name of CSS property to check the element for
• pattern — required, a value pattern that is using the same simple wildcard matching as in the basic url filtering rules OR a regular expression. For this type of matching, AdGuard always does matching in a case-insensitive manner. In the case of a regular expression, the pattern looks like /regexp/.

For non-regexp patterns (,),[,] must be unescaped, e.g. :matches-css(background-image:url(data:*)).

For regexp patterns \ should be escaped, e.g. :matches-css(background-image: /^url\\("data:image\\/gif;base64.+/).

Regexp patterns do not support flags.

Examples

For such DOM:

<!-- HTML code -->
<style type="text/css">
    #matched::before {
        content: "Block me"
    }
</style>
<div id="matched"></div>
<div id="not-matched"></div>

div elements with pseudo-element ::before with specified content property can be selected by any of these extended selectors:

! string pattern
div:matches-css(before, content: block me)

! string pattern with wildcard
div:matches-css(before, content: block*)

! regular expression pattern
div:matches-css(before, content: /block me/)

Obsolete pseudo-classes :matches-css-before() and :matches-css-after() are supported for better compatibility.

Backward compatible syntax for :matches-css() is supported but not recommended.

Pseudo-class :matches-attr()

Pseudo-class :matches-attr() allows to select an element by its attributes, especially if they are randomized.

Syntax

[target]:matches-attr("name"[="value"])

• target — optional, standard or extended CSS selector, can be missed for checking any element
• name — required, simple string or string with wildcard or regular expression for attribute name matching
• value — optional, simple string or string with wildcard or regular expression for attribute value matching

For regexp patterns " and \ should be escaped, e.g. div:matches-attr(class=/[\\w]{5}/).

Regexp patterns do not support flags.

Examples

div:matches-attr("ad-link") selects the element div#target1:

<!-- HTML code -->
<div id="target1" ad-link="1random23-banner_240x400"></div>

div:matches-attr("data-*"="adBanner") selects the element div#target2:

<!-- HTML code -->
<div id="target2" data-1random23="adBanner"></div>

div:matches-attr(*unit*=/^click$/) selects the element div#target3:

<!-- HTML code -->
<div id="target3" random123-unit094="click"></div>

*:matches-attr("/.{5,}delay$/"="/^[0-9]*$/") selects the element #target4:

<!-- HTML code -->
<div>
  <inner-random23 id="target4" nt4f5be90delay="1000"></inner-random23>
</div>

Pseudo-class :matches-property()

Pseudo-class :matches-property() allows to select an element by matching its properties.

Syntax

[target]:matches-property("name"[="value"])

• target — optional, standard or extended CSS selector, can be missed for checking any element
• name — required, simple string or string with wildcard or regular expression for element property name matching
• value — optional, simple string or string with wildcard or regular expression for element property value matching

For regexp patterns " and \ should be escaped, e.g. div:matches-property(prop=/[\\w]{4}/).

Regexp patterns are supported in name for any property in chain, e.g. prop./^unit[\\d]{4}$/.type.

Regexp patterns do not support flags.

Examples

An element with such properties:

divProperties = {
  id: 1,
  check: {
    track: true,
    unit_2random1: true,
  },
  memoizedProps: {
    key: null,
    tag: 12,
    _owner: {
      effectTag: 1,
      src: 'ad.com',
    },
  },
};

can be selected by any of these extended selectors:

div:matches-property(check.track)

div:matches-property("check./^unit_.{4,8}$/")

div:matches-property("check.unit_*"=true)

div:matches-property(memoizedProps.key="null")

div:matches-property(memoizedProps._owner.src=/ad/)

For filters maintainers: To check properties of a specific element, do the following:

1. Inspect the page element or select it in Elements tab of browser DevTools.
2. Run console.dir($0) in Console tab.

Pseudo-class :xpath()

The :xpath() pseudo-class allows to select an element by evaluating an XPath expression.

Syntax

[target]:xpath(expression)

• target- optional, standard or extended CSS selector
• expression — required, valid XPath expression

Limitations

target can be omitted so it is optional. For any other pseudo-class that would mean "apply to all DOM nodes", but in case of :xpath() it just means "apply to the whole document", and such applying slows elements selecting significantly. That's why rules like #?#:xpath(expression) are limited to looking inside the body tag. For example, rule #?#:xpath(//div[@data-st-area=\'Advert\']) is parsed as #?#body:xpath(//div[@data-st-area=\'Advert\']).

Extended selectors with defined target as any selector — *:xpath(expression) — can still be used but it is not recommended, so target should be specified instead.

Works properly only at the end of selector, except for pseudo-class :remove().

Examples

:xpath(//*[@class="banner"]) selects the element div#target1:

<!-- HTML code -->
<div id="target1" class="banner"></div>

:xpath(//*[@class="inner"]/..) selects the element div#target2:

<!-- HTML code -->
<div id="target2">
  <div class="inner"></div>
</div>

Pseudo-class :nth-ancestor()

The :nth-ancestor() pseudo-class allows to lookup the nth ancestor relative to the previously selected element.

Syntax

subject:nth-ancestor(n)

• subject — required, standard or extended CSS selector
• n — required, number >= 1 and < 256, distance to the needed ancestor from the element selected by subject

Limitations

Pseudo-class :nth-ancestor() is not supported inside the argument of the :not() pseudo-class. It is one of the known issues.

Examples

For such DOM:

<!-- HTML code -->
<div id="target1">
  <div class="child"></div>

  <div id="target2">
    <div>
      <div>
        <div class="inner"></div>
      </div>
    </div>
  </div>
</div>

.child:nth-ancestor(1) selects the element div#target1, div[class="inner"]:nth-ancestor(3) selects the element div#target2.

Pseudo-class :upward()

The :upward() pseudo-class allows to lookup the ancestor relative to the previously selected element.

Syntax

subject:upward(ancestor)

• subject — required, standard or extended CSS selector
• ancestor — required, specification for the ancestor of the element selected by subject, can be set as:
  • number >= 1 and < 256 for distance to the needed ancestor, same as :nth-ancestor()
  • standard CSS selector for matching closest ancestor

Limitations

Pseudo-class :upward() is not supported inside the argument of the :not() pseudo-class argument. It is one of the known issues.

Examples

For such DOM:

<!-- HTML code -->
<div id="target1" data="true">
  <div class="child"></div>

  <div id="target2">
    <div>
      <div>
        <div class="inner"></div>
      </div>
    </div>
  </div>
</div>

.inner:upward(div[data]) selects the element div#target1, .inner:upward(div[id]) selects the element div#target2, .child:upward(1) selects the element div#target1, .inner:upward(3) selects the element div#target2.

Pseudo-class :remove() and pseudo-property remove

Sometimes, it is necessary to remove a matching element instead of hiding it or applying custom styles. In order to do it, you can use the :remove() pseudo-class as well as the remove pseudo-property.

Syntax

! pseudo-class
selector:remove()

! pseudo-property
selector { remove: true; }

• selector — required, standard or extended CSS selector

Limitations

The :remove() pseudo-class is limited to work properly only at the end of selector.

For applying the :remove() pseudo-class to any element universal selector * should be used. Otherwise such extended selector may be considered as invalid, e.g. .banner > :remove() is not valid for removing any child element of banner class element, so it should look like .banner > *:remove().

If the :remove() pseudo-class or the remove pseudo-property is used, all style properties are ignored except for the debug pseudo-property.

Examples

div.banner:remove()
div:has(> div[ad-attr]):remove()

div:contains(advertisement) { remove: true; }
div[class]:has(> a > img) { remove: true; }

Rules with the remove pseudo-property should use #$?# marker: $ for CSS style rules syntax, ? for ExtendedCss syntax.

Both :remove() pseudo-class and remove pseudo-property works the same, but we recommend to use the pseudo-property as it is related to an action which should be applied to element, since pseudo-classes is more about elements matching.

Pseudo-class :is()

The :is() pseudo-class allows to match any element that can be selected by any of selectors passed to it. Invalid selectors are skipped and the pseudo-class deals with valid ones with no error thrown. Our implementation of the native :is() pseudo-class.

Syntax

[target]:is(selectors)

• target — optional, standard or extended CSS selector, can be missed for checking any element
• selectors — forgiving selector list of standard or extended selectors. For extended selectors only compound selectors are supported, not complex.

Limitations

Rules with the :is() pseudo-class should use the native implementation of :is() if rules use ## marker and it is possible, i.e. with no other extended selectors inside. To force applying ExtendedCss rules with :is(), use #?#/#$?# marker explicitly.

If the :is() pseudo-class argument selectors is an extended selector, due to the way how the :is() pseudo-class is implemented in ExtendedCss v2.0, it is impossible to apply it to the top DOM node which is html, i.e. #?#html:is(<extended-selectors>) does not work. So if target is not defined or defined as an universal selector *, the extended pseudo-class applying is limited to html's children, e.g. rules #?#:is(...) and #?#*:is(...) are parsed as #?#html *:is(...). Please note that there is no such limitation for a standard selector argument, i.e. #?#html:is(.locked) works fine.

Complex selectors with extended pseudo-classes are not supported as selectors argument for :is() pseudo-class, only compound ones are allowed. It is one of the known issues. Check examples below for more details.

Examples

#container *:is(.inner, .footer) selects only the element div#target1:

<!-- HTML code -->
<div id="container">
  <div data="true">
    <div>
      <div id="target1" class="inner"></div>
    </div>
  </div>
</div>

Due to limitations :is(*:not([class]) > .banner)' does not work but :is(*:not([class]):has(> .banner)) can be used instead of it to select the element div#target2:

<!-- HTML code -->
<span class="span">text</span>
<div id="target2">
  <p class="banner">inner paragraph</p>
</div>

Pseudo-class :not()

The :not() pseudo-class allows to select elements which are not matched by selectors passed as argument. Invalid argument selectors are not allowed and error is to be thrown. Our implementation of the :not() pseudo-class.

Syntax

[target]:not(selectors)

• target — optional, standard or extended CSS selector, can be missed for checking any element
• selectors — list of standard or extended selectors

Limitations

Rules with the :not() pseudo-class should use the native implementation of :not() if rules use ## marker and it is possible, i.e. with no other extended selectors inside. To force applying ExtendedCss rules with :not(), use #?#/#$?# marker explicitly.

If the :not() pseudo-class argument selectors is an extended selector, due to the way how the :not() pseudo-class is implemented in ExtendedCss v2.0, it is impossible to apply it to the top DOM node which is html, i.e. #?#html:not(<extended-selectors>) does not work. So if target is not defined or defined as an universal selector *, the extended pseudo-class applying is limited to html's children, e.g. rules #?#:not(...) and #?#*:not(...) are parsed as #?#html *:not(...). Please note that there is no such limitation for a standard selector argument, i.e. #?#html:not(.locked) works fine.

The :not() is considered as a standard CSS pseudo-class inside argument of the :upward() pseudo-class because :upward() supports only standard selectors.

"Up-looking" pseudo-classes which are :nth-ancestor() and :upward() are not supported inside selectors argument for :not() pseudo-class. It is one of the known issues.

Examples

#container > *:not(h2, .text) selects only the element div#target1:

<!-- HTML code -->
<div id="container">
  <h2>Header</h2>
  <div id="target1"></div>
  <span class="text">text</span>
</div>

Pseudo-class :if-not() (deprecated)

The :if-not() pseudo-class is deprecated and is no longer supported. Rules with it are considered as invalid.

This pseudo-class was basically a shortcut for :not(:has()). It was supported by ExtendedCss for better compatibility with some filters subscriptions.

Selectors debugging mode

Sometimes, you might need to check the performance of a given selector or a stylesheet. In order to do it without interacting with JavaScript directly, you can use a special debug style property. When ExtendedCss meets this property, it enables the debugging mode either for a single selector or for all selectors, depending on the debug value.

Sometimes, you might need to check the performance of a given selector or a stylesheet. In order to do it without interacting with JavaScript directly, you can use a special debug style property. When ExtendedCss meets this property, it enables the debugging mode either for a single selector or for all selectors, depending on the debug value.

Open the browser console while on a web page to see the timing statistics for selector(s) that were applied there. Debugging mode displays the following stats as object where each of the debugged selectors are keys, and value is an object with such properties:

Always printed:

• selectorParsed — text of eventually parsed selector
• timings — list of DOM nodes matched by the selector
  • appliesCount — total number of times that the selector has been applied on the page
  • appliesTimings — time that it took to apply the selector on the page, for each of the instances that it has been applied (in milliseconds)
  • meanTiming — mean time that it took to apply the selector on the page
  • standardDeviation — standard deviation
  • timingsSum — total time it took to apply the selector on the page across all instances

Printed only for remove pseudos:

• removed — flag to signal if elements we removed

Printed if elements are not removed:

• matchedElements — list of DOM nodes matched by the selector
• styleApplied — parsed rule style declaration related to the selector

Examples

Debugging a single selector:

When the value of the debug property is true, only information about this selector will be shown in the browser console.

#$?#.banner { display: none; debug: true; }

Enabling global debug:

When the value of the debug property is global, the console will display information about all extended CSS selectors that have matches on the current page, for all the rules from any of the enabled filters.

#$?#.banner { display: none; debug: global; }

Global debugging mode also can be enabled by positive debug property in ExtCssConfiguration:

const extendedCss = new ExtendedCss({
styleSheet, // required, rules as string
debug,      // optional, boolean
});

Backward compatible syntax

Backward compatible syntax is supported but not recommended.

Old syntax for pseudo-class :has()

Syntax

target[-ext-has="selector"]

Examples

div[-ext-has=".banner"]

<!-- HTML code -->
<div>Not selected</div>
<div>Selected <span class="banner"></span></div>

Old syntax for pseudo-class :contains()

Syntax

// matching by plain text
target[-ext-contains="text"]

// matching by a regular expression
target[-ext-contains="/regex/"]

Examples

// matching by plain text
div[-ext-contains="banner"]

// matching by a regular expression
div[-ext-contains="/this .* banner/"]

<!-- HTML code -->
<div>Not selected</div>
<div id="selected">Selected as it contains "banner"</div>

Old syntax for pseudo-class :matches-css()

Syntax

target[-ext-matches-css="property: pattern"]
target[-ext-matches-css-after="property: pattern"]
target[-ext-matches-css-before="property: pattern"]

Examples

<!-- HTML code -->
<style type="text/css">
    #matched::before {
        content: "Block me"
    }
</style>
<div id="matched"></div>
<div id="not-matched"></div>

! string pattern
div[-ext-matches-css-before="content: block me"]

! regular expression pattern
div[-ext-matches-css-before="content: /block me/"]

How to build

Install dependencies

yarn install

And just run

yarn build

How to test

Install dependencies

yarn install

Run local node testing

yarn test local

Run performance tests which are not included in test local run and should be executed manually:

yarn test performance

Usage

You can import, require or copy IIFE module with ExtendedCss into your code, e.g.

import ExtendedCss from 'extended-css';

or

const ExtendedCss = require('extended-css');

IIFE module can be found by the following path ./dist/extended-css.js

After that you can use ExtendedCss as you wish.

API description

Constructor

/**
 * Creates an instance of ExtendedCss
 *
 * @param configuration — required
 */
constructor(configuration: ExtCssConfiguration)

where

interface ExtCssConfiguration {
  // css stylesheet — css rules combined in one string
  styleSheet?: string;

  // css rules — array of separated css rules
  cssRules?: string;

  // the callback that handles affected elements
  beforeStyleApplied?: BeforeStyleAppliedCallback;

  // flag for applied selectors logging; equals to `debug: global` in `styleSheet`
  debug?: boolean;
}

Both styleSheet and cssRules are optional but at least one of them should be set.

If both styleSheet and cssRules are set, both of them are to be applied.

/**
 * Needed for getting affected node elements and handle style properties before they are applied to them if it is necessary.
 *
 * Used by AdGuard Browser extension to display rules in Filtering log and `collect-hits-count` (via tsurlfilter's CssHitsCounter)
 */
type BeforeStyleAppliedCallback = (x:IAffectedElement) => IAffectedElement;

/**
 * Simplified just for representation.
 * Its optional property 'content' may contain the applied rule text
 */
interface IAffectedElement {
  rules: { style: { content?: string }}[]
  node: HTMLElement;
}

Public method init()

The init() public method initializes ExtendedCss on a page. It should be executed on page as soon as possible, even before the ExtendedCss instance is constructed, otherwise the :contains() pseudo-class may work incorrectly.

Public methods apply() and dispose()

After the instance of ExtendedCss is created, it can be applied on the page by the apply() method. Its applying also can be stopped and styles are to be restored by the dispose() method.

(function() {
  let styleSheet = 'div.wrapper > div:has(.banner) { display:none!important; }\n';
  styleSheet += 'div.wrapper > div:contains(ads) { background:none!important; }';
  const extendedCss = new ExtendedCss({ styleSheet });

  // apply styleSheet
  extendedCss.apply();

  // stop applying of this styleSheet
  setTimeout(function() {
    extendedCss.dispose();
  }, 10 * 1000);
})();

Public method query()

/**
 * Returns a list of the document's elements that match the specified selector
 *
 * @param {string} selector — selector text
 * @param {boolean} [noTiming=true] — optional, if true -- do not print the timing to the console
 *
 * @returns a list of elements found
 * @throws an error if the argument is not a valid selector
 */
public static query(selector: string, noTiming = true): HTMLElement[]

Public method validate()

/**
 * Validates selector
 * @param selector — selector text
 */
public static validate(selector: string): ValidationResult

where

type ValidationResult = {
    // true for valid selector, false for invalid one
    ok: boolean,
    // specified for invalid selector
    error: string | null,
};

Public property EXTENDED_CSS_VERSION

type: string

Current version of ExtendedCss.

Debugging extended selectors

ExtendedCss can be executed on any page without using any AdGuard product. In order to do that you should copy and execute the following code in a browser console:

!function(e,t,d){C=e.createElement(t),C.src=d,C.onload=function(){alert("ExtendedCss loaded successfully")},s=e.getElementsByTagName(t)[0],s?s.parentNode.insertBefore(C,s):(h=e.getElementsByTagName("head")[0],h.appendChild(C))}(document,"script","https://AdguardTeam.github.io/ExtendedCss/extended-css.min.js");

Alternatively, install the ExtendedCssDebugger userscript.

Now you can now use the ExtendedCss from global scope, and run its method query() as Document.querySelectorAll()

Examples

const selector = 'div.block:has(.header:matches-css(after, content: Ads))';

// array of HTMLElements matched the `selector` is to be returned
ExtendedCss.query(selector);

Projects using ExtendedCss

• CoreLibs — Content Script dist should be updated
• TSUrlFilter
• FiltersCompiler
• AdguardBrowserExtension — TSUrlFilter should be updated
• AdguardForSafari — adguard-resources should be updated
• AdguardForiOS — both ExtendedCss and TSUrlFilter should be updated in advanced-adblocker-web-extension

Browser compatibility

Browser	Version
Chrome	✅ 88
Firefox	✅ 84
Edge	✅ 88
Opera	✅ 80
Safari	✅ 14
Internet Explorer	❌

Known issues

• :has() pseudo-class should take <forgiving-relative-selector-list> as argument
• :nth-ancestor() and :upward() are not supported inside of :not() pseudo-class argument
• :is() pseudo-class does not support complex selectors