CSS Generated Content Module Level 3

Introduction

Authors sometimes want user agents to render content that does not come from the document tree. One familiar example of this is numbered headings; the author does not want to mark the numbers up explicitly, they want the user agent to generate them automatically. Counters and markers are used to achieve these effects.

h1::before { content: counter(section) ": "; }

Similarly, authors may want the user agent to insert the word "Figure" before the caption of a figure, or "Chapter 7" on a line before the seventh chapter title.

chapter { counter-increment: chapter; }
chapter > title::before { content: "Chapter " counter(chapter) "\A"; }

Another common effect is replacing elements with images or other multimedia content. Since not all user agents support all multimedia formats, fallbacks may have to be provided.

/* Replace <logo> elements with the site’s logo, using a format
 * supported by the UA */
logo { content: url(logo.mov), url(logo.mng), url(logo.png), none; }

/* Replace <figure> elements with the referenced document, or,
 * failing that, with either the contents of the alt attribute or the
 * contents of the element itself if there is no alt attribute */
figure[alt] { content: attr(href url), attr(alt); }
figure:not([alt]) { content: attr(href url), contents; }

Value Definitions

This specification follows the CSS property definition conventions from [CSS2] using the value definition syntax from [CSS-VALUES-3]. Value types not defined in this specification are defined in CSS Values & Units [CSS-VALUES-3]. Combination with other CSS modules may expand the definitions of these value types.

In addition to the property-specific values listed in their definitions, all properties defined in this specification also accept the CSS-wide keywords as their property value. For readability they have not been repeated explicitly.

1. Inserting and Replacing Content: the content property

User agents are expected to support this property on all media, including non-visual ones.

The content property dictates what is rendered inside an element or pseudo-element.

For elements, it has only one purpose: specifying that the element renders as normal, or replacing the element with an image (and possibly some associated "alt text").

For pseudo-elements and margin boxes, it is more powerful. It controls whether the element renders at all, can replace the element with an image, or replace it with arbitrary inline content (text and images).

normal

For an element or page margin box, this computes to contents.

For ::before and ::after, this computes to none.

For ::marker, ::placeholder, and ::file-selector-button, this computes to itself (normal).

none

On elements, this behaves as normal.

On pseudo-elements it inhibits the creation of the pseudo-element as if it had display: none.

In neither case does it prevent any pseudo-elements which have this element or pseudo-element as an originating element from being generated.

<content-replacement>

Equal to:

<image>

Makes the element or pseudo-element a replaced element, filled with the specified <image>. Its normal contents are suppressed and do not generate boxes, as if they were display: none.

If the <image> represents an invalid image, then it must be treated as instead representing an image with zero natural width and height, filled with transparent black.

The above invalid image behavior appears to be what Chrome is doing. Is this okay? Is there a better behavior we can/should use?

Note: Replaced elements use different layout rules than normal elements. (In effect, it becomes equivalent to an HTML img element.)

Note: Replaced elements do not have ::before or ::after pseudo-elements; the content property replaces their entire contents.

This value has historically been treated as <content-list> on ::before and ::after. Presumably there’s a Web-compat requirement on this, so these pseudo-elements might need an exception. [Issue #2889]

<content-list>

Replaces the element’s contents with one or more anonymous inline boxes corresponding to the specified values, in the order specified. Its normal contents are suppressed and do not generate boxes, as if they were display: none.

Each value contributes an inline box to the element’s contents. For <image>, this is an inline anonymous replaced element; for the others, it’s an anonymous inline run of text.

If an <image> represents an invalid image, the user agent must do one of the following:

• "Skip" the <image>, generating nothing for it.

• Display some indication that the image can’t be displayed in place of the <image>, such as a "broken image" icon.

This specification intentionally does not define which behavior a user agent must use, but it must use one or the other consistently.

Note: If the value of <content-list> is a single <image>, it must instead be interpreted as a <content-replacement>.

/ [ <string> | <counter> | <attr()> ]+

Specifies the "alt text" for the element. See § 1.2 Alternative Text for Accessibility for details. If omitted, the element has no "alt text".

Should the contents keyword be replaced with content()?

1.1. Accessibility of Generated Content

Generated content should be searchable, selectable, and available to assistive technologies. The content property applies to speech and generated content must be rendered for speech output. [CSS3-SPEECH]

Start work on an AAM for CSS.

1.2. Alternative Text for Accessibility

Content intended for visual media sometimes needs alternative text for speech output or other non-visual mediums. The content property thus accepts alternative text to be specified after a slash (/) after the last <content-list>. If such alternative text is provided, it must be used for speech output instead.

This allows, for example, purely decorative text to be elided in speech output (by providing the empty string as alternative text), and allows authors to provide more readable alternatives to images, icons, or text-encoded symbols.

.new::before {
 content: url(./img/star.png) / "New!";
  /* or a localized attribute from the DOM: attr("data-alt") */
}

.expandable::before {
 content: "\25BA" / "";
/* a.k.a. ► */
 /* aria-expanded="false" already in DOM,
   so this pseudo-element is decorative */
}

2. Generated Content Values: the <content-list> type

The <content-list> value is used in content to fill an element with one or more anonymous inline boxes, including images, strings, the values of counters, and the text value of elements. In this section we enumerate the possibilities.

The syntax of <content-list> is defined as:

<content-list> = [ <string> | <image> | <attr()> | contents | <quote> | <leader()> | <target> | <string()> | <content()> | <counter> ]+

2.1. Basic Strings: <string> and <attr()> values

<string>

Represents an anonymous inline box filled with the specified text.

Note: White space in the string is handled the same as in literal text, and controlled by the properties in [CSS-TEXT-3] and elsewhere. In particular, white space character can collapse, even across multiple strings, such as in content: "First " " Second";, which by default will render similar to "First Second" (with a single visible space between the two words).

<attr()>

The attr() functional notation represents the string stored as the specified attribute’s value. Its argument is a CSS qualified name (qname) representing the attribute name and namespace, if any. See [CSS3-NAMESPACE].

Note: As in attribute selectors, attribute names without an explicit namespace do not associate with any default namespace.

2.2. 2D Images: <image> values

<image>

Represents an anonymous inline replaced element filled with the specified <image>.

If the <image> represents an invalid image, this value instead represents nothing. (No inline content is added to the element, as if this value were "skipped".)

CSS2.1 explicitly allowed the UA to substitute a broken image icon if the image was invalid. However, no browser appears to do this. Is this removal okay?

2.3. Elemental Content: the contents keyword

contents

The element’s descendants. Since this can only be used once per element (you can’t duplicate the children if, e.g., one is a plugin or form control), it is handled as follows:

If set on the element:

Always honoured. Note that this is the default, since the initial value of content is normal and normal computes to contents on an element.

If set on one of the element’s other pseudo-elements:

Check to see that it is not set on a "previous" pseudo-element, in the following order, depth first:

1. the element itself

2. ::before

3. ::after

Should this behave as an empty string on pseudo-elements?

If it is already used, then it evaluates to nothing (like none). Only pseudo-elements that are actually generated are checked.

foo { content: normal; }  /* this is the initial value */
foo::after { content: contents; }

foo { content: none; }
foo::after { content: contents; }

Use cases for suppressing the content on the element and using it in a pseudo-element would be welcome.

Note: While it is useless to include contents twice in a single content property, that is not a parse error. The second occurrence simply has no effect, as it has already been used. It is also not a parse error to use it on a ::marker pseudo-element, it is only during the rendering stage that it gets treated like none.

Do we need the statement about marker pseudo-elements here? Or is this legacy from the old version of the spec?

2.4. Quotation Marks

The quotes property, which defines the quotation mark system of an element, in conjunction with the various *-quote values of the content property, which insert corresponding quotation marks as generated content, can be used to automatically insert opening and closing quotation marks, such as for the HTML q element.

2.4.1. Quotation Mark System: the quotes property

User agents are expected to support this property on all media, including non-visual ones.

This property specifies the quotation mark system for the element, defining how the content property’s open-quote and close-quote values behave. See § 2.4.2 Inserting Quotation Marks: the *-quote keywords. Values have the following meanings:

none

The open-quote and close-quote values of the content property produce no quotations marks, as if they were no-open-quote and no-close-quote respectively.

auto

A typographically appropriate quotation mark system is automatically chosen by the UA based on the content language of the parent (or, if there is no parent, of the element itself).

Note: The Unicode Common Locale Data Repository [CLDR] maintains information on typographically appropriate quotation marks. UAs can use other sources of information as well, particularly as typographic preferences can vary; however it is encouraged to submit any improvements to Unicode so that the entire software ecosystem can benefit.

match-parent

Specifies the same quotation mark system as the parent. In general this is equivalent to inheriting the parent’s computed value, except that auto it resolves using the same content language that the parent used.

Two possible approaches here, currently speccing the latter: a) this computes to the relevant string values, and inherits as such. b) this value effectively inherits as a keyword + a language code, meaning auto, but with this language.

[ <string> <string> ]+

The quotation mark system is defined as the specified list of pairs of quotation marks (opening and closing). The first pair represents the outermost level of quotation, the second pair the next level of embedding, etc. Within a pair, the open-quote value refers to the first <string>, close-quote refers to the second.

2.4.2. Inserting Quotation Marks: the *-quote keywords

<quote> = open-quote | close-quote | no-open-quote | no-close-quote

open-quote

close-quote

These values are replaced by the appropriate string as defined by the quotes property, and increments (for no-open-quote) or decrements (for no-close-quote) the level of nesting for quotes.

no-open-quote

no-close-quote

Inserts nothing (as in none), but increments (for no-open-quote) or decrements (for no-close-quote) the level of nesting for quotes.

Quotation marks can be inserted in appropriate places in a document with the open-quote and close-quote values of the content property. Each occurrence of open-quote or close-quote is replaced by a quotation mark string as defined by the quotation mark system specified by the quotes property, based on the depth of nesting.

Which pair of quotes is used depends on the nesting level of quotes (the quote depth): the number of occurrences of open-quote in all generated text before the current occurrence, minus the number of occurrences of close-quote in all generated text before and including the current occurence. If the depth is 0, the first pair is used, if the depth is 1, the second pair is used, etc. If the depth is greater than the number of pairs, the last pair is repeated.

Note: Quote depth is independent of the nesting of the source document or the formatting structure. Also, like counter inheritance, it operates on the “flattened element tree” in the context of the [DOM].

A close-quote or no-close-quote that would make the quote depth negative is in error and is ignored (at rendering time): the depth stays at 0 and no quote mark is inserted (although the rest of the content property’s value is still inserted).

blockquote p::before { content: open-quote }
blockquote p::after { content: no-close-quote }
blockquote p:last-child::after { content: close-quote }

Note: If a quotation is in a different language than the surrounding text, it is customary to quote the text with the quote marks of the language of the surrounding text, not the language of the quotation itself.

:lang(fr) > * { quotes: "\00AB\2005" "\2005\00BB" "\2039\2005" "\2005\203A" }
:lang(en) > * { quotes: "\201C" "\201D" "\2018" "\2019" }

:lang(fr) > * { quotes: "« " " »" "‹ " " ›" }
:lang(en) > * { quotes: "“" "”" "‘" "’" }

/* Specify pairs of quotes for two levels in two languages */
:lang(en) > q { quotes: '"' '"' "'" "'" }
:lang(no) > q { quotes: "«" "»" "’" "’" }

/* Insert quotes before and after Q element content */
q::before { content: open-quote }
q::after  { content: close-quote }

<html lang="en">
  <head>
   <title>Quotes</title>
  </head>
  <body>
   <p><q>Quote me!</q></p>
  </body>
</html>

"Quote me!"

<html lang="no">
  <head>
   <title>Quotes</title>
  </head>
  <body>
   <p><q>Trøndere gråter når <q>Vinsjan på kaia</q> blir deklamert.</q></p>
  </body>
</html>

«Trøndere gråter når ’Vinsjan på kaia’ blir deklamert.»

2.5. Leaders

A leader, sometimes known as a tab leader or a dot leader, is a repeating pattern used to visually connect content across horizontal spaces. They are most commonly used in tables of contents, between titles and page numbers. The leader() function, as a value for the content property, is used to create leaders in CSS. This function takes a string (the leader string), which describes the repeating pattern for the leader.

2.5.1. The leader() function

leader( <leader-type> )

Inserts a leader. See the section on leaders for more information.

leader() = leader( <leader-type> )
<leader-type> = dotted | solid | space | <string>

Three keywords are shorthand values for common strings:

dotted

Equivalent to leader(".")

solid

Equivalent to leader("_")

space

Equivalent to leader(" ")

<string>

Issue: Define this.

ol.toc a::after {
  content: leader('.') target-counter(attr(href), page);
}

<h1>Table of Contents</h1>
<ol class="toc">
<li><a href="#chapter1">Loomings</a></li>
<li><a href="#chapter2">The Carpet-Bag</a></li>
<li><a href="#chapter3">The Spouter-Inn</a></li>
</ol>

Table of Contents

1. Loomings.....................1
2. The Carpet-Bag...............9
3. The Spouter-Inn.............13

Do leaders depend on the assumption that the content after the leader is right-aligned (end-aligned)?

2.5.2. Rendering leaders

Consider a line which contains the content before the leader (the “before content”), the leader, and the content after the leader (the “after content”). Leaders obey the following rules:

1. The leader string must appear in full at least once.

2. The leader should be as long as possible

3. Visible characters in leaders should vertically align with each other when possible.

4. Line break characters in the leader string must be ignored.

5. White space in the leader string follows normal CSS rules.

6. A leader only appears between the start content and the end content.

7. A leader only appears on a single line, even if the before content and after content are on different lines.

8. A leader can’t be the only thing on a line.

2.5.3. Procedure for rendering leaders

1. Lay out the before content, until reaching the line where the before content ends.

   BBBBBBBBBB
   BBB
   

2. The leader string consists of one or more glyphs, and is thus an inline box. A leader is a row of these boxes, drawn from the end edge to the start edge, where only those boxes not overlaid by the before or after content. On this line, draw the leader string, starting from the end edge, repeating as many times as possible until reaching the start edge.

   BBBBBBBBBB
   ..........
   

3. Draw the before and after content on top of the leader. If any part of the before content or after content overlaps a glyph in a leader string box, that glyph is not displayed.

   BBBBBBBBBB
   BBB....AAA
   

4. If one full copy of the leader string is not visible:

   BBBBBBB
   BBBBBBA
   

   Insert a line break after the before content, draw the leader on the next line, and draw the after content on top, and hide any leader strings that are not fully displayed.

   BBBBBBB
   BBBBBB
   ......A
   

what to do if after content is wider than the line box?

Leaders don’t quite work in table layouts. How can we fix this?

2.6. Cross References

Many documents contain internal references:

Three new values for the content property are used to automatically create these types of cross-references: target-counter(), target-counters(), and target-text(). Each of these displays information obtained from the target end of a link.

<target> = <target-counter()> | <target-counters()> | <target-text()>

See sections below for details on each of these.

2.6.1. The target-counter() function

target-counter() = target-counter( [ <string> | <url> ] , <custom-ident> , <counter-style>? )

The target-counter() function retrieves the value of the innermost counter with a given name. The required arguments are the url of the target and the name of the counter. An optional counter-style argument can be used to format the result.

These functions only take a fragment URL which points to a location in the current document. If there’s no fragment, if the ID referenced isn’t there, or if the URL points to an outside document, the user agent must treat that as an error.

what should error handling be?

restrict syntactically to local references for now.

…which will be discussed on page <a href="#chapter4_sec2"></a>.

a::after { content: target-counter(attr(href url), page) }

…which will be discussed on page 137.

<nav>
  <ol>
   <li class="frontmatter"><a href="#pref_01">Preface</a></li>
   <li class="frontmatter"><a href="#intr_01">Introduction</a></li>
   <li class="bodymatter"><a href="#chap_01">Chapter One</a></li>
  </ol>
</nav>

.frontmatter a::after { content: leader('.') target-counter(attr(href url), page, lower-roman) }
.bodymatter a::after { content: leader('.') target-counter(attr(href url), page, decimal) }

Preface.............vii
Introduction.........xi
Chapter One...........1

2.6.2. The target-counters() function

This functions fetches the value of all counters of a given name from the end of a link, and formats them by inserting a given string between the value of each nested counter.

target-counters() = target-counters( [ <string> | <url> ] , <custom-ident> , <string> , <counter-style>? )

I have not found a compelling example for target-counters() yet.

2.6.3. The target-text() function

The target-text() function retrieves the text value of the element referred to by the URL. An optional second argument specifies what content is retrieved, using the same values as the string-set property above.

target-text() = target-text( [ <string> | <url> ] , [ content | before | after | first-letter ]? )

A simpler syntax has been proposed by fantasai: http://lists.w3.org/Archives/Public/www-style/2012Feb/0745.html

…which will be discussed <a href="#chapter_h1_1">later</a>.

a::after { content: ", in the chapter entitled " target-text(attr(href url)) }

2.7. Named Strings

This section introduces named strings, which are the textual equivalent of counters and which have a distinct namespace from counters. Named strings follow the same nesting rules as counters. The string-set property accepts values similar to the content property, including the extraction of the current value of counters.

Named strings are a convenient way to pull metadata out of the document for insertion into headers and footers. In HTML, for example, META elements contained in the document HEAD can set the value of named strings. In conjunction with attribute selectors, this can be a powerful mechanism:

meta[author] { string-set: author attr(author); }
head > title { string-set: title contents; }
@page:left {
  @top {
   text-align: left;
   vertical-align: middle;
   content: string(title);
  }
}
@page:right {
  @top {
   text-align: right;
   vertical-align: middle;
   content: string(author);
  }
}

2.7.1. Naming Strings: the string-set property

User agents are expected to support this property on all media, including non-visual ones.

The string-set property copies the text content of an element into a named string, which functions as a variable. The text content of this named string can be retrieved using the string() function. Since these variables may change on a given page, an optional second value for the string() function allows authors to choose which value on a page is used.

none

The element does not set any named strings.

[ <custom-ident> <string>+ ]#

The element establishes one or more named strings, corresponding to each comma-separated entry in the list.

For each entry, the <custom-ident> gives the name of the named string. It’s followed by one or more <string> values, which are concatenated together to form the value of the named string.

If an element has style containment, the string-set property must have no effects on descendants of that element.

H1 { string-set: chapter contents; }

2.7.2. Inserting Named Strings: the string() function

string() = string( <custom-ident> , [ first | start | last | first-except ]? )

The string() function is used to copy the value of a named string to the document, via the content property. This function requires one argument, the name of the named string. Since the value of a named string may change several times on a page (as multiple elements defining the string can appear) an optional second argument indicates which value of the named string should be used.

The second argument of the string() function is one of the following keywords:

first

The value of the first assignment on the page is used. If there is no assignment on the page, the entry value is used. If no second argument is provided, this is the default value.

start

If the element is the first element on the page, the value of the first assignment is used. Otherwise the entry value is used. The entry value may be empty if the named string hasn’t yet appeared.

last

The exit value of the named string is used.

first-except

This is identical to first, except that the empty string is used on the page where the value is assigned.

we may need to kill the entire content string. Is this necessary?

The content values of named strings are assigned at the point when the content box of the element is first created (or would have been created if the element’s display value is none). The entry value for a page is the assignment in effect at the end of the previous page. The exit value for a page is the assignment in effect at the end of the current page.

CSS:

@page {
  size: 15cm 10cm;
  margin: 1.5cm;

  @top-left {
  content: "first: " string(heading, first);
  }
  @top-center {
  content: "start: " string(heading, start);
  }
   @top-right {
   content: "last: " string(heading, last);
  }
  }

h2 { string-set: heading content() }

The following figures show the first, start, and last assignments of the “heading” string on various pages.

2.7.3. The content() function

content() = content( [ text | before | after | first-letter | marker ]? )

text

The string value of the element. If no value is specified in content(), it acts as if text were specified.

before

The string value of the ::before pseudo-element.

after

The string value of the ::after pseudo-element.

first-letter

The first letter of the element, as defined for the ::first-letter pseudo-element

marker

The string value of the ::marker pseudo-element.

<h1>Loomings</h1>

h1::before { content: 'Chapter ' counter(chapter); }
h1 { string-set: header content(before) ':' content(text); }
h1::after { content: '.'; }

<section title="Loomings">

section { string-set: header attr(title) }

2.8. Automatic Counters and Numbering

See CSS Lists 3 § 4 Automatic Numbering With Counters.

Should this move back to CSS Content?

3. Bookmarks

Some document formats, most notably PDF, allow the use of bookmarks as an aid to navigation. Bookmarks provide a list of links to document elements, as well as text to label the links and a level value. A bookmark has three properties: bookmark-level, bookmark-label, and bookmark-state.

When a user activates a bookmark, the user agent must bring that reference point to the user’s attention, exactly as if navigating to that element by fragment URL. This will also trigger matching the :target pseudo-class.

If an element has style containment, the bookmark-level, bookmark-label, and bookmark-state properties must have no effect on descendants of the element.

3.1. Setting a Bookmark: the bookmark-level property

The bookmark-level property determines if a bookmark is created, and at what level. If this property is absent, or has value none, no bookmark should be generated, regardless of the values of bookmark-label or bookmark-state.

<integer [1,∞]>

defines the level of the bookmark, with the top level being 1 (negative and zero values are invalid).

none

no bookmark is generated.

section h1 { bookmark-level: 1; }
section section h1 { bookmark-level: 2; }
section section section h1 { bookmark-level: 3; }

Note: Bookmarks do not need to create a strict hierarchy of levels.

Should a bookmark be created for elements with display: none?

3.2. Labelling a Bookmark: the bookmark-label property

<content-list>

<content-list> is defined above, in the section on the string-set property. The value of <content-list> becomes the text content of the bookmark label.

<h1>Loomings</h1>

h1 {
bookmark-label: content(text);
bookmark-level: 1;
}

3.3. Initial Bookmark Toggle State: the bookmark-state property

The bookmark-state may be open or closed. The user must be able to toggle the bookmark state.

open

Subsequent bookmarks with bookmark-level greater than the given bookmark are displayed, until reaching another bookmark of the same level or lower. If one of subsequent bookmark is closed, apply the same test to determine if its subsequent bookmarks should be displayed.

closed

Subsequent bookmarks of bookmark-level greater than the given bookmark are not displayed, until reaching another bookmark of the same level or lower.

Is the initial bookmark state, or the bookmark state updated by the UA as appropriate?

4. Changes

Significant changes since the 2 August 2019 Working Draft consist primarily of:

• Updated auto to reference the parent content language rather than the content language of the element itself. (Issue 5478)
• Added match-parent value to quotes. (Issue 5478)
• Prevent negative quote depth by ignoring excess closing quotes, consistent with [CSS2]. (Issue 2034)
• Defined computed value of content: normal for ::placeholder and ::file-selector-button. (Issue 6124)
• Fixed an off-by-one error for the quote depth of closing quotes. (Issue 2506)
• Inlined the definition of attr(), since it has been dropped from [CSS-VALUES-4].
• Miscellaneous editorial spec clean up.

See also previous changes.

Acknowledgements

Stuart Ballard, David Baron, Bert Bos, Tantek Çelik, and James Craig provided invaluable suggestions used in this specification.

Privacy Considerations

No new privacy considerations have been reported on this specification.

Security Considerations

No new security considerations have been reported on this specification.