W3C

CSS Grid Template Layout Module

Editor's Draft 16 April 2014

This version:
http://dev.w3.org/csswg/css3-layout/
Latest version:
http://www.w3.org/TR/css3-layout/
Previous version:
http://www.w3.org/TR/2011/WD-css3-layout-20111129/
Feedback:
www-style@w3.org with subject line “[css3-layout] … message topic …” (archives)
Editors:
Bert Bos (W3C)
César Acebal (University of Oviedo)

Abstract

Image: four elements move to four slots in a template

Above: a typical Web page. Below: the underlying grid template.

CSS is a simple, declarative language for creating style sheets that specify the rendering of HTML and other structured documents. This module defines a typographic grid for CSS. It has features to set up a grid template and to flow content into it. (A separate module, [CSS-GRID-1], uses the same template for absolutely positioning elements.)

The ‘grid’ property sets up a matrix of rows and columns and designates empty areas and areas for content (called “slots”). The ‘flow’ property specifies into which slot an element flows. A grid template is like a table in that the various slots are aligned in rows and columns and can automatically adapt to each other's size, and it is like absolute positioning in that elements can be put in arbitrary slots so that the visual order becomes independent of the document order.

The ::slot()’ pseudo-element allows to style the slots (backgrounds, borders, etc.)

Grid templates can also be associated with pages in paged media, to create page templates.

Slots can form “chains” to create non-rectangular regions. A chain of slots is like a series of pages or columns: content is distributed over the slots by breaking it at possible break points (see the ‘break-*’ properties from [CSS3-BREAK]).

The @region’ rule [CSS3-REGIONS] allows to select (parts of) elements based on whether they fall in a certain slot or not, and thus style content differently in different slots (region-based styling).

Status of this document

This is a public copy of the editors' draft. It is provided for discussion only and may change at any moment. Its publication here does not imply endorsement of its contents by W3C. Don't cite this document other than as work in progress.

The (archived) public mailing list www-style@w3.org (see instructions) is preferred for discussion of this specification. When sending e-mail, please put the text “css3-layout” in the subject, preferably like this: “[css3-layout] …summary of comment…

This document was produced by the CSS Working Group (part of the Style Activity).

This document was produced by a group operating under the 5 February 2004 W3C Patent Policy. W3C maintains a public list of any patent disclosures made in connection with the deliverables of the group; that page also includes instructions for disclosing a patent. An individual who has actual knowledge of a patent which the individual believes contains Essential Claim(s) must disclose the information in accordance with section 6 of the W3C Patent Policy.

This draft combines ideas from earlier drafts on Template Layout, Grid Positioning [CSS3GRID] and Grid Layout [CSS-GRID-1].

Issues on this draft are mentioned in the text itself and/or in the on-line Tracker under “Details on Product CSS3 Template Layout.”

This draft also used to contain a layout model based on the “stack of cards” idea, with all cards the same size and tabs for selecting which card is “on top” (i.e., visible). Where should that model be defined? (ISSUE-???)

Table of contents

1. Introduction

(This section is not normative.)

1.1. Summary and use cases

Image: four elements move to four slots in a template

Above: a typical Web page. Below: the underlying grid template. It has four slots, a, b, c and d and the grid template is defined by ‘grid: "a a a a a a a" "b c c c c d d"’.

The styling of a Web page, a form or a graphical user interface can roughly be divided into two parts: (1) defining the overall “grid” of the page or window and (2) specifying the fonts, indents, colors, etc., of the text and other objects. The two are not completely separate, of course, because indenting or coloring a text influences the perceived grid as well. Nevertheless, when one separates the parts of a style that should change when the window gets bigger from the parts that stay the same, one often finds that the grid changes (room for a sidebar, extra navigation bar, big margins, larger images…), while fonts, colors, indents, numbering styles, and many other things don't have to change, until the size of the window becomes extreme.

The properties in this specification associate a layout grid with an element. Rather than rendering all descendants of an element in a single flow, they give an element multiple flows, called slots, placed on an invisible grid by means of a grid template. The descendants are given a ‘flow’ property to designate the slot into which they flow.

Because layouts on the Web have to adapt to different window and paper sizes, the rows and columns of the grid can be made fixed or flexible in size.

In some typographical traditions, the slots are called “fields” or “spatial zones.” Traditionally, also, all columns have the same width, called the “module” or “unit.” The horizontal lines of a grid that define the rows are sometimes called “flow lines.” Some traditions make the space between pieces of content an explicit part of the grid, some consider the space part of the content: Both styles are possible in CSS, the latter by using the ‘margin’ property, the former by using the ‘.’ symbol in a grid template.

The typical use cases for these properties include:

Once a grid template has been defined, it can also be used to position elements without adding them to a flow. Setting ‘display: grid’ on a grid element makes all children into separate flows and automatically distributes them over the grid. This is explained in CSS Grid Layout [CSS-GRID-1].

Template-based positioning makes many traditional uses of absolute positioning unnecessary. Like absolute positioning, grid templates are especially useful for aligning elements that don't have simple relationships in the source (parent-child, ancestor-descendant, immediate sibling). But in contrast to absolute positioning, the elements are not positioned with the help of horizontal and vertical coordinates, but by flowing them into a template that is very much like a table: The size and alignment of elements are governed implicitly by the rows and columns of the template, and can automatically adapt to the overall size of the element and to the content in each row and column.

In this example, the four children of an element are assigned to four slots (called a, b, c and d) in a 2×2 template. (All mark-up examples in this specification are HTML fragments, unless otherwise stated.)

Image: sample rendering

Each element occupies one slot. In this template, all slots have the same size.

<style type="text/css">
  dl { grid: "a b"
             "c d" }
  #sym1 { flow: a }
  #lab1 { flow: b }
  #sym2 { flow: c }
  #lab2 { flow: d }
</style>
<dl>
  <dt id=sym1>A
  <dd id=lab1>A is een aapje
  <dt id=sym2>B
  <dd id=lab2>B is de bakker
</dl>

Try it out! (Uses a JavaScript emulation in browsers that do not implement grid templates natively.)

Templates can also help with device-independence. This example uses Media Queries [MEDIAQ] to change the overall layout of a page from 3-column layout for a wide screen to a 1-column layout for a narrow screen. It assumes the page has been marked-up with logical sections with IDs.

@media all
{
  body { grid: "a a a"
               "b c d" }
  #head { flow: a }
  #nav { flow: b }
  #adv { flow: c }
  #body { flow: d }
}
@media all and (max-width: 500px)
{
  body { grid: "a"
               "b"
               "c" }
  #head { flow: a }
  #nav { flow: b }
  #adv { display: none }
  #body { flow: c }
}

The slots in a grid do not overlap, unless with negative margins. Here is how the “zunflower” design of the CSS Zen Garden could be done:

#container {grid: "A B C"}
#container::slot(A) {box-shadow: 0.5em 0.5em 0.5em; z-index: 1}
#container::slot(B) {box-shadow: 0.5em 0.5em 0.5em; margin-left: -2em}
#intro {flow: A}
#supportingText {flow: B}
#linkList {flow: C}

Template-based positioning borrows some concepts from table layout, in particular the idea of aligning elements in rows and columns, so that they constrain one another's size. But there are also differences. This example shows some of them. Assume this document fragment:

<div class=group>
 <div>aa aa aa aa aa aa</div>
 <div>bbb</div>
 <div class=menu>ccccc</div>
</div>

We can lay it out as three columns, as the following illustrations show. The style sheet would contain the following.

.group {display: table}
.group > div {display: table-cell}

[Three unequal cells]

Example of rendering with a table.

We can also use a template, in which case the style sheet would contain this:

.group {grid: "a b c"}
.group > div {flow: a}
.group > div + div {flow: b}
.group > div + div + div {flow: c}

By default, the table is as wide as needed to fit its contents. To make sure it is as wide as its containing block, we need to add

.group {display: table; width: 100%}

That is not needed for the grid template, but, on the other hand, if we want the template to fit its contents, we would need to say so:

.group {grid: "a b c"; width: auto}

(See [CSS3BOX] for the definition of the ‘width’ property.) The columns of the template are by default all the same size. The columns of the table satisfy certain constraints, but the exact size is not defined. We can make them all the same by adding a rule (see [CSS3TBL]):

.group {display: table; width: 100%; table-layout: fixed}

[Three equal cells]

Example of rendering with equal columns.

In both styles, we can set a column to a certain size:

div.menu {width: 3em}

resp.,

.group {grid: * * 3em  "a b c"}

[Two equal cells, third is 3em wide]

Example of rendering with a fixed third column and the other two columns of equal width.

If there is an unknown number of columns (children of the div.group element), the style sheet for the table model will automatically take them into account. The style sheet for the template model, however, creates a template of exactly three columns and can't handle tables with an unknown number of columns. The extra elements will be added into the default slot (in this case the ‘a’ slot).

(However, setting ‘display’ to ‘grid’ on the grid element creates a different kind of grid, that does grow, see “Elements with ‘display’ set to ‘grid’ or ‘inline-grid’” below.)

In both models, elements can have borders, but only in the table model can borders be collapsed, which makes setting borders easier in the table model:

.group {display: table; border-collapse: collapse}
.group > div {border: solid}

resp.,

.group > div {border: solid; border-left: none}
.group > div:first-child {border-left: solid}

In the template model, the order of the elements is explicit, and thus it is possible to reverse the order of the columns:

.group > div {flow: c}
.group > div + div {flow: b}
.group > div + div + div {flow: a}

[Different contents for the cells]

Example of rendering with the contents of the three columns reversed: the third element is shown in the first slot and the first element in the third slot.

In the table model, the order of the rows and columns is given by the document source and thus can't be changed.

This example shows a way to move notes to the end of a section. “Notes” in this example refers to elements in HTML with a class of “note”. A fragment of HTML such as

<div class=section>
  <p>The balubious coster of the fifth secter<span
    class=note> The sixth secter coster is a difter
    manon.</span> of The Rollow Carpug mentizes a costernica.
  <p>…
</div>

with this style sheet

div.section {
    grid: "*"
          "F"}
.note {
    flow: F;
    content: counter(note) ".\A0" contents;
    counter-increment: note;
    font-size: smaller}
.note::before {
    content: counter(note);
    vertical-align: super;
    font-size: larger}

results in a rendering similar to this:

Same text, with the SPAN replaced by “(1)” and its content moved to the end.

Rendering of a text with footnotes.

The “Generated content for paged media” module [CSS3GCPM] is expected to define specific features for footnotes that are both easier and more powerful than this example.

This example shows the use of chained regions: text from region 1 continues in region 2, 3 and 4. And it shows how to use pseudo-elements to style text in a particular region: text in region 1 is bigger than in other regions.

Image of a complex, 3-column layout

Example rendering

We assume a document fragment similar to this:

<div id=article>
  <h1>Introduction</h1>
  <p><img src="sunset" alt="">
  <p>This is an example…

  <h1>More Details</h1>
  <p>This illustrates…
  <p>Then, the example…
  <p>Finally, this…
</div>

The style sheet makes the DIV into a grid element with five regions, called A, *, b, c and d. The regions are grouped into two chains: region A on its own, and the chain consisting of *, b, c and d.

#article {
  grid: "A A c"
        "A A c"
        "* * c"
        "a b c";
  chains: * a b c }

The ‘*’ is a special name. It indicates the default slot, the slot where content goes if it doesn't have an explicit ‘flow’ property. We could have used a normal letter and added an explicit ‘flow’ property to the children of the DIV, but by using * we don't have to. All children thus go into region * (and continue in a, b and c as needed). The IMG element goes into region A. We assume for simplicity that there is only one IMG element:

#article img { flow: A }

The text in region * is bolder, larger and a different color than in the other regions. Also, an H1 that falls into this region is rendered differently from other H1s:

@region ::slot(*)
{
  * {
    font-weight: bold;
    color: #0C3D5F;
    font-size: larger }
  h1 {
    color: crimson;
    display: run-in }
}

(For brevity, the example doesn't show the style rules to set the color and background, to make the text justified, add the left border to the second H1, etc.)

As the image shows, the text of the last paragraph isn't complete and presumably continues somewhere else, but that part is outside the scope of this example. In paged media, we could attach the grid template to a page instead of an element and let the text continue on the next page; we could also make slot c into a multi-column element so that the text forms additional columns to the right of current image (see [CSS3GCPM] for some ideas), or we could allow a scrollbar on slot c.

This example shows two [three] ways to align the contents of a fixed-height paragraph to the bottom of the box:

Method 1
P {height: 20em; grid: "." "*" fit-content}
Method 2
P {height: 20em; grid: "*"}
P::slot(*) {vertical-align: bottom}
[Method 3]
[If vertical-align is generalized to apply to all blocks, rather than only to table cells, margin boxes and slots:]
P {height: 20em; vertical-align: bottom}

However, the styles differ in case the content is taller than 20 em and overflows: the first style has a baseline-aligned content which will thus overflow below the element's box; the second example has bottom-aligned content, which will thus overflow above the element's box.

This example (based on a challenge by Andrew Fedoniouk) shows how a grid template can be used to display six items in two rows of three. Assume the following document:

<p>
  <span>...</span>
  <span>...</span>
  <span>...</span>
  <span>...</span>
  <span>...</span>
  <span>...</span>
</p>

The simplest way to break this into two lines (or more) is to add a hard line break after the third SPAN: ‘span:nth-child(3)::after {content: "\A"; white-space: pre}’. That works fine, unless the SPANs are inline blocks (‘span {display: inline-block}’). Another way is to float all SPANs and add a ‘clear’ property on the fourth: ‘span {float: left} span:nth-child(4) {clear: left}’. But if the SPANs have different heights or baselines, their vertical alignment will be different than if they were on the same line.

With a grid template, we can say that the P has not one, but two flows, a and b, the second one positioned below the first; and then we assign the fourth and subsequent SPANs to the “b” flow. (The others automatically go into the default, first flow.)

p {grid: "a" "b"}
span:nth-child(n+4) {flow: b}

This almost works. There is just a subtle issue with the spaces between the SPANs. We moved the last three SPANs to flow b, but we have no way to move the spaces between them. They remain in the default (first) slot. (They collapse, as there is nothing left between them, so we don't actually see them anymore.) To get spaces between the SPANs in slot b, we need to generate new spaces:

span:nth-child(n+4)::after {content: " "}

This space handling is not very intuitive. One might expect there to be a way to say “select everything from this element onward” or maybe a pseudo-element selector for the text between elements. Or a new property to specify different ways to handle the anonymous text inside a grid element. For the moment, there is no such thing. Suggestions are welcome.

This example uses a page-based template to create a running header that is not at the top of the page, but exactly in the middle. We need to define our own page template for this, because the predefined page template of [CSS3PAGE] only has margin boxes along the edge of the page.

@page {
  grid:
    "top-half"
    "running-header" 3em
    "bottom-half";
  chains: top-half bottom-half
}
::slot(running-header) {
  content: "My Magazine - page " counter(page) " - February 2013";
  text-align: center;
  vertical-align: middle;
  border: solid;
  margin: 1em 0;
}

Image: Example rendering

An example of a page rendered with the above page template. It has the “running header” with the page number right in the middle of the page.

1.2. Dependencies on other modules

This CSS3 module depends on the following other CSS3 modules:

Add to the above all the properties that are allowed on slots (see “Styling slots: the ‘::slot()’ pseudo-element”).

It has non-normative (informative) references to the following other CSS3 modules:

See section 1.4.2 of CSS level 2 [CSS21] for the grammar and other notations that this specification uses in property definitions.

1.3. Values

This specification follows the CSS property definition conventions from [CSS21]. Value types not defined in this specification are defined in CSS Level 2 Revision 1 [CSS21]. Other CSS modules may expand the definitions of these value types: for example [CSS3COLOR], when combined with this module, expands the definition of the <color> value type as used in this specification.

In addition to the property-specific values listed in their definitions, all properties defined in this specification also accept the inherit keyword as their property value. For readability it has not been repeated explicitly.

1.4. A note about accessibility

(This section is not normative.)

The facilities in this specification allow elements from a document to be displayed in a visual order that is to a large extent independent of the order in the document. That may have both positive and negative effects on accessibility. The positive aspect is that it allows the content in the document to be kept in logical order, so that the document is as functional as possible without the style sheet and on media where the style sheet doesn't apply. A negative aspect is that a document that has a defined tab order (the order in which elements get the focus when the tab-key is pressed) will show on the screen with a tab order unrelated to the visual order. It may be necessary to use the keyboard control features of the CSS Basic User Interface module [CSS3UI] to ensure that the tab navigation follows the visual order, or to refrain from positioning semantically related elements in different parts of a template.

The following two requirements from the Web Content Accessibility Guidelines (WCAG) 2.0 [WCAG20] are particularly relevant. See that specification for more information.

1.3.2 Meaningful Sequence: When the sequence in which content is presented affects its meaning, a correct reading sequence can be programmatically determined. (Level A)

2.4.3 Focus Order: If a Web page can be navigated sequentially and the navigation sequences affect meaning or operation, focusable components receive focus in an order that preserves meaning and operability. (Level A)

2. Declaring templates

Grid templates are declared with the ‘grid-template-areas’, ‘grid-template-rows’, ‘grid-template-columns’ and ‘grid’ properties, defined below.

An element that has a grid template is called a grid element. More precisely: an element is a grid element if (1) it generates a block container box [CSS21] and (2) at least one of ‘grid-template-areas’, ‘grid-template-rows’ or ‘grid-template-columns’ has a value other than the initial value.

Block containers are, informally, elements that can have block elements as children, such as blocks, inline blocks or table cells, but not inline elements or replaced elements.

An element's grid ancestor is the nearest ancestor that is a grid element.

Note that not all elements have a grid ancestor. E.g., the root element has none.

The number of columns of a grid element is the number of <col-width> values in ‘grid-template-columns’ or the number of columns in ‘grid-template-areas’, whichever is larger.

The number of rows is the number of <row-height> values in ‘grid-template-rows’ or the number of rows in ‘grid-template-areas’, whichever is larger.

Note that a grid element always has at least one row and one column. Those may, however, have zero size.

2.1. Declaring a template: ‘grid-template-areas

The ‘grid-template-areas’ property implicitly assigns a grid to an element and defines named slots in the grid. It can also designate a slot as the default slot for content without an explicit flow.

Name: grid-template-areas
Value: none | <string>+
Initial: none
Applies to: block container elements [CSS21] and grid container elements [CSS-GRID-1]
Inherited: no
Animatable: no
Percentages: N/A
Media: visual
Computed value: specified value
Canonical order: per grammar

Each string consist of one or more asterisks (“*”), identifiers, periods (“.”), spaces and tabs. Each string represents one row in the template, each symbol other than a space or tab represents one column in that row. Spaces and tabs have no meaning. They separate identifiers and may also be added for readability.

The symbols in the template have the following meaning

<identifier>
Named slot for content.
*
(Asterisk.) Default slot for content (explained below).
.
(Period.) Blank space.

Multiple identical identifiers in adjacent rows or columns form a single slot that spans those rows and columns. Ditto for multiple “*”s. Slot names are case-sensitive, so uppercase and lowercase letters are considered to be different letters.

Case-sensitivity should instead be defined by reference to the definition of identifiers, probably in css3-values.

Non-rectangular slots and multiple slots with the same name are illegal. A template without any slot name or “*” is illegal. A template with more than one “*” slot is illegal. Strings with different numbers of columns (before padding with “.”, see the next paragraph) are illegal. These errors cause the declaration to be ignored.

Note: non-rectangular and disconnected regions may be permitted in a future update of CSS.

If there are fewer symbols in a string than the number of columns in the element, the string is implicitly padded with “.” (period) symbols. Likewise, if there are fewer strings than the number of rows in the element, additional strings consisting of “.” (period) symbols are implicitly added.

Each slot (identifier or “*”) acts as a block element for its contents.

If the value is ‘none’, then no explicit slots are defined. If the element is a grid element, then the element instead has an implicit template consisting of a single ‘*’ slot.

Note that an element is a grid element in this case if ‘grid-template-columns’ and ‘grid-template-rows’ are not both ‘auto’.

For example, the following two sets of style rules are equivalent:

DIV {
  grid-template-areas: none;
  grid-template-rows: 10em 10em 10em;
  grid-template-columns: * * * * }

and

DIV {
  grid-template-areas: "****" "****" "****";
  grid-template-rows: 10em 10em 10em;
  grid=columns: * * * * }

Authors should be aware that naming slots with Hebrew or Arabic letters may cause some text editors to display the CSS rule with the slot name of the string on the right. However, the first symbol in each string corresponds to the first value in ‘grid-template-columns’ and thus represents the leftmost column in the rendering of the document.

Only a block container element can be a grid element. E.g., in the following document, the elements with ID “one,” “three” and “five” are grid elements, but the ‘grid’ property on “two” and “four” has no effect.

<html>
  <style>
   #one {grid: "a b c"}
   #two {grid: "a" "b"}
   #three {display: inline-block; grid: "a b" "c d"}
   #four {grid: "a @"}
   #five {grid: "a b"}
  </style>
  <table>
   <tr id=four><td>... <td id=five>...
  </table>
  <ol>
   <li id=one>
    ... <span id=three>... </span>...
   <li>
    ... <em id=two>... </em>...
  </ol>

2.2. Specifying the widths of columns: ‘grid-template-columns

Name: grid-template-columns
Value: auto | <col-width>+
Initial: auto
Applies to: block container elements and grid container elements
Inherited: no
Animatable: yes, between grids with the same number of tracks
Percentages: N/A
Media: visual
Computed value: specified value
Canonical order: per grammar

The ‘grid-template-columns’ property specifies the sizes of the columns of a grid.

Where

<col-width> = <length> | <percentage> | * | <fraction> |
  <minmax> | min-content | max-content | fit-content
<minmax> = minmax( <col-width> , <col-width> )

The definition of minmax isn't correct, it should only accepts certain values.

Each <col-width> sets the width of a column, the first value for the leftmost column, the second for the second column, etc.

If there are fewer <col-width> values than the number of columns in the element, or if the value is ‘auto’, the missing columns widths are all ‘*’.

Each <col-width> can be one of the following:

<length>
An explicit width for that column. Negative values are illegal.
<fraction>
(A non-negative number followed by ‘fr’.) A fraction of the remaining space, see Fraction values below.
*
(Asterisk.) All columns with a width of ‘*’ have the same width.
<percentage>
Expresses a size as a percentage of the element's a-priori content width. When the width of the grid element is dependent on content, the result is undefined.
max-content
min-content
The column's width is determined by its contents. See the algorithm below.
minmax(p,q)
The column's width is constrained to be greater than or equal to p and less than or equal to q. p and q stand for [ <length> | max-content | min-content | * ]. There may be white space around the p and q. If q < p, then q is ignored and ‘minmax(p,q)’ is treated as ‘minmax(p,p)’.
fit-content
Equivalent to ‘minmax(min-content, max-content)’.

2.3. Specifying the height of rows: ‘grid-template-rows

Name: grid-template-rows
Value: auto | <row-height>+
Initial: auto
Applies to: block container elements and grid container elements
Inherited: no
Animatable: yes, between grids with the same number of tracks
Percentages: N/A
Media: visual
Computed value: specified value
Canonical order: per grammar

The ‘grid-template-rows’ property specifies the heights of the rows of a grid.

<row-height> = <length> | <percentage> | * | <fraction> |
  <minmax> | min-content | max-content | fit-content
<minmax> = minmax( <row-height> , <row-height> )

The definition of minmax isn't correct, it should only accepts certain values.

Note that the syntax is the same syntax as for ‘grid-template-columns

Each <row-height> sets the height of a row, the first value for the topmost row, the second for the second row, etc.

If there are fewer <row-height> values than the number of rows in the element, or if the value is ‘auto’, the missing row heights are all ‘fit-content’.

Each <row-height> can be one of the following:

<length>
An explicit height for that row. Negative values are illegal.
<fraction>
(A non-negative number followed by ‘fr’.) A fraction of the remaining space, see Fraction values below.
* (asterisk)
All rows with a height of ‘*’ have the same height.
<percentage>
Expresses a size as a percentage of the element's a-priori content height.

When the height of the grid element is dependent on content, the result is undefined.

max-content
min-content
The row's height is determined by its contents. See the algorithm below.
minmax(p,q)
The row's height is constrained to be greater than or equal to p and less than or equal to q. p and q stand for [ <length> | max-content | min-content | * ]. There may be white space around the p and q. If q < p, then q is ignored and ‘minmax(p,q)’ is treated as ‘minmax(p,p)’.
fit-content
Equivalent to ‘minmax(min-content, max-content)’.

2.4. The ‘grid’ and ‘grid-template’ shorthand properties

Name: grid-template
Value: none | <col-width>* [ [ <string> <row-height>? ]+ | ‘/<row-height>+ ]
Initial: none
Applies to: block container elements and grid container elements
Inherited: no
Animatable: see individual properties
Percentages: see individual properties
Media: visual
Computed value: see individual properties
Canonical order: N/A
Name: grid
Value: <‘grid-template’> | <‘grid-auto-flow’> [ <‘grid-auto-columns’> [ / <‘grid-auto-rows’> ]? ]
Initial: See individual properties
Applies to: block container elements and grid container elements
Inherited: no
Animatable: see individual properties
Percentages: see individual properties
Media: visual
Computed value: see individual properties
Canonical order: N/A

The ‘grid-template’ property is a shorthand for ‘grid-template-areas’, ‘grid-template-columns’ and ‘grid-template-rows’ and is equivalent to setting those properties as follows:

The ‘grid’ property is a shorthand for setting ‘grid-template-areas’, ‘grid-template-rows’, ‘grid-template’columns', ‘grid-auto-rows’, ‘grid-auto-columns’, and ‘grid-auto-flow’ in a single declaration. See [CSS-GRID-1].

Note that ‘grid-auto-rows’, ‘grid-auto-columns’, and ‘grid-auto-flow’ only apply if ‘display’ is ‘grid’ or ‘inline-grid’.

For example, the rule

grid: * 10em "a b c" "a b d" 4em

is the same as

grid-template-columns: * 10em;
grid-template-areas: "a b c" "a b d";
grid-template-rows: auto 4em;

… and the values for ‘grid-auto-rows’, ‘grid-auto-columns’, and ‘grid-auto-flow’ are set to their default values. (Omitted for clarity.)

Note the ‘auto’ keyword that was implied in the shorthand, but must be made explicit in the ‘grid-template-rows’ property. Also note that the width of the third column is specified neither in the shorthand nor in ‘grid-template-columns’, which means it is implicitly set to ‘*’.

For example, the rule

grid: * 10em / 4em 5em

is the same as

grid-template-columns: * 10em;
grid-template-areas: "* *" "* *";
grid-template-rows: 4em 5em;

(And ‘grid-auto-rows’, ‘grid-auto-columns’, and ‘grid-auto-flow’ are set to their default values.)

2.5. Default slots

Every grid element has a default slot. If there is an asterisk (“*”) in the template, then that slot is the default. If there is no asterisk, then the first name in the template defines the default slot.

Note that a grid always has at least one slot, see ‘grid-template-areas’.

For example, if the template is defined by ‘grid-template-areas: "..." "..c" "a b b" "a b b"’, then “c” is the default slot.

If the grid has an implicit template (i.e., ‘grid-template-areas’ is ‘none’), then its single slot is the default slot.

All content of a grid element that is not inside another flow (i.e., not inside a float, not absolutely positioned, etc.) is flowed into the default slot. In particular, any text content of the grid element itself is in that default flow.

For example, in this document fragment

<DIV STYLE="grid: 'a b *'">
 <IMG STYLE="flow: a" SRC="image.png" ALT="Foo">
 This is an
 <EM STYLE="flow: *">emphasized</EM>
 <EM STYLE="flow: b">(note well!)</EM>
 sentence.
 <
</DIV>

The three slots of the grid contain:

Slot a
The image
Slot b
The text “(note well!)”
Default slot
The text “This is an emphasized sentence.”

(The ‘flow’ property is defined below.)

2.6. Interaction of ‘grid’ and ‘columns

The ‘columns’ property of a grid element applies to the default slot of the template.

For example, with this style:

BODY {grid: "a @ b" "c c c"; columns: 2}

The @-slot will have two columns.

Note that this is the same as setting columns on the default slot: ‘BODY::slot(*) {columns: 2}’ apart from differences in the specificity of the selector.

2.7. Calculating the size of the grid

For the purpose of the calculations below, each slot (identifier or “*”) in a grid has four dimensions associated with it, called MINW (“minimum width”), PREFW (“preferred width”), MINH (“minimum height”) and PREFH (“preferred height”). We can think of them as the minimum and preferred dimensions of the slot in isolation, i.e., if it wasn't constrained by other slots in the same row or column. They are defined as follows:

MINW
PREFW
MINH
PREFH

For example, the MINW values of the slots in this grid

grid:  auto * 10em
      "  a  a  .  "
      "  b  .  c  "

are as follows:

a
min-content’ (because the slot spans at least one column with a width specified as ‘min-content’ or ‘auto’).
b
min-content’ (ditto).
c
0 (because the slot spans only columns with a width that is independent of the content).

The UA must choose the widths and heights of all columns and rows such that the following constraints are satisfied.

  1. If the element has an a-priori known content height, then the sum of the heights of all rows must be equal to the element's height.

    For example, the two rows in this grid must be 8em each so that the total height matches the height of the element:

    div {
      height: 16em;
      grid: "a . b"  *
            "a c ."  *
    
  2. If the grid element has an a-priori known content width, then the sum of the widths of all columns must be equal to the element's width.

    For example, the three columns in this grid must be 20em each:

    div {width: 60em; grid: "a b c"}
  3. Each row with a height specified as a <length> must have exactly that height.
  4. Each column with a width specified as a <length> must have exactly that width.
  5. All rows with a height specified as ‘*’ must have the same height.

    If we have both ‘*’ and ‘fr’, then ‘*’ will be defined as ‘1fr’ and we can drop this rule and the next. Otherwise either these two rules or the next two must be removed.

  6. All columns with a width specified as ‘*’ must have the same width.
  7. For any pair (i,j) of rows whose <row-height> values hi and hj are both specified in ‘fr’, the computed heights Hi and Hj must be such that Hi * hj = Hj * hi. (I.e., their heights are proportional to their number of ‘fr’ units.)
  8. For any pair (i,j) of columns whose <col-width> values wi and wj are both specified in ‘fr’, the computed width Wi and Wj must be such that Wi * wj = Wj * Wi. (I.e., their widths are proportional to their number of ‘fr’ units.)
  9. Each row that contains slots that span only one row and no slots that span more than one row, must not be higher than the largest PREFH of all slots in the row plus the amount needed to vertically align the horizontal slots that have a ‘vertical-align’ of ‘baseline’.

    For example, the second row in this grid

    grid: "a . c . ."  5em
          ". . . . b"  auto
    

    must not be taller than the height (block dimension) of slot b. The first row contains a slot that spans two rows (slot a), so this rule does not apply to that row.

  10. Each column that contains slots that span only one column and no slots that span more than one column, must not be wider than the largest PREFW of all slots in the column plus the amount needed to align the vertical slots that have a ‘vertical-align’ of ‘baseline’.

    For example, in this grid

    grid: "a c"
          "a b"
          auto *
    

    the first column must not be wider than the PREFW of slot a. Both slots in the second column have an infinite PREFW, so this rule effectively puts no constraint on that column.

  11. Each slot must be at least as wide as its MINW.
  12. Each slot must be at least as high as its MINH.

If it is impossible to choose such widths and heights, then try without constraint 1. If it is still impossible, try without constraint 2 instead. And, finally, try with both 1 and 2 dropped.

For example, the sum of the row heights in this example can never be the same as the height of the element:

div {
  height: 20em;
  grid: "a b c"  7em
        "a . c"  7em }

The first constraint is therefore ignored, the rows are 7em each and 6em of space below the grid remains empty.

If there are multiple solutions, and constraint 1 was dropped or did not apply, then the sizes must additionally be chosen such that the sum of the heights of the rows is minimized.

If there are still multiple solutions, and constraint 2 was dropped or did not apply, then the sizes must additionally be chosen such that the sum of the widths of the columns is minimized.

The calculation is intended to be symmetric in width and height (except for the case when both width and height are fixed and the grid is over-constrained, when a solution that ignores the height is tried before trying one that ignores the width).

For example, there would be multiple solutions for this grid:

<DIV STYLE="float: left; grid: 'a . b'">
 <P STYLE="flow: a">Two words
 <P STYLE="flow: b">Only three words
</DIV>

The columns must have equal width, but there is no other constraint on the width. They could be narrow:

Two  
words
     
     
     
     
Only 
three
words

or wider:

Two words                       Only three words

or even wider still, e.g.:

Two words                             Only three words   

The rule to minimize height excludes the first, narrow solution. The rule to minimize width excludes the third and all wider solutions. So the second layout, the narrowest that has all words on one line, is the correct one.

(This example assumes the width of the floating DIV's containing block is large enough. The default width of a float is actually ‘fit-content’, and thus if the containing block is too narrow, the result will be narrower, too, because the calculation will have to be redone using that width as the a-priori width for the DIV.)

The width isn't known a-priori, if, e.g., ‘width’ is ‘auto’ and the element is floating, absolutely positioned, inline-block or a child of a block with vertical writing mode.

An extra step may be necessary in paged media if a page break occurs inside a template (only in the case of an element-based template, see below). If the template, after computing the width and height, is too big to fit on the current page, and if a suitable break point exists, the part of the template after that break point is put on the next page. The width of the containing block on that page may be different if that page has different margins from the current page (see [CSS3PAGE]) and thus the width and height of that part of the template must be recalculated in the new context.

Note that the widths of the columns can be completely determined before laying out any of the contents as long as there are no columns with a <col-width> of ‘min-content’ or ‘max-content’.

Do we define restrictions or approximations for slots that are part of a chain to avoid complicated optimization algorithms?

Note: In a future update of CSS, rows might get a property to specify how the height of that row is adjusted in case the above calculation yields a template that is less tall than the element itself.

The height of a slot is measured as if the slot had one anonymous block as a child that contains all the slot's contents and the anonymous block is a flow root (see [CSS3BOX]).

This example divides the window in three rows and three columns, separated by 1em of white space. The middle row and the middle column are flexible, the others are fixed at a specific size. The first column is 5em wide, the last one 10em.

<style type="text/css">
  body {
    height: 100%;
    grid: 5em 1em  *  1em 10em
          "a   .   b   .   c"  2em
          ".   .   .   .   ."  1em
          "d   .   e   .   f"
          ".   .   .   .   ."  1em
          "g   .   h   .   i"  2em}
  #logo {flow: a}
  #motto {flow: b}
  #date {flow: c}
  #main {flow: e}
  #adv {flow: f}
  #copy {flow: g}
  #about {flow: h}
</style>
<p id=logo><img src=...
<p id=motto>Making Web pages since 1862
<p id=date>August 2, 2004
...

[Add example with three columns, first two as narrow as possible, third one taking up all remaining space.]

3. Flowing content into slots: ‘flow

The ‘flow’ property adds an element to a slot.

Name: flow
Value: auto | <identifier> | <string> | ‘*’ | same
Initial: auto
Applies to: elements that have a grid ancestor and whose ‘position’ is ‘static’ or ‘relative
Inherited: no
Animatable: no
Percentages: N/A
Media: visual
Computed value: specified value
Canonical order: per grammar

If the value is not ‘auto’, the element is added to the flow of the given slot, instead of to the flow of its parent.

Note that the content of a slot depends on its ‘content’ property. Content flowed into the slot is thus not necessarily rendered. See “Slots and the ‘content’ property.”

The name of the slot may be quoted (<string>) or unquoted (<identifier>), but if the name is equal to one of the following, it must be quoted: ‘initial’, ‘inherit’, ‘default’ or ‘same’. This is to avoid ambiguity with the keywords of the same name.

For example, the following two style sheets are the same:

div {grid: "p1 p2"}
h2 {flow: p1}

and

div {grid: "p1 p2"}
h2 {flow: "p1"}

But the following two are not:

div {grid: "initial final"}
h2 {flow: "initial"}

and

div {grid: "initial final"}
h2 {flow: initial}

The last ‘initial’ is interpreted as a reserved keyword and not as the name of the first slot in the grid.

If the element has no grid ancestor, or that grid ancestor has no slot of the given name, the property is treated as if it were ‘auto’.

A value of ‘same’ means the element is put into the same slot as the nearest preceding element in document order that (1) has the same grid ancestor, (2) to which the ‘flow’ property applies and (3) whose ‘flow’ property is not ‘auto’.

If the name refers to a slot that doesn't exist in the element's grid ancestor (or there is no grid ancestor) there are several possibilities:

  1. The element is not positioned (i.e., it is positioned in the current flow).
  2. The letter is looked for in the grid ancestor's own grid ancestor, etc., recursively. If no slot of that name is found in any of them, then the element is not positioned.
  3. The element is positioned in the default slot (i.e., as if ‘flow: *’ had been specified); or not positioned at all, in case there is no grid ancestor.
  4. The name is looked for in the grid ancestor's own grid ancestor, etc., recursively. If no slot of that name is found in any of them, then the element is positioned in the default slot.

All content flowed into the same slot, whether explicitly with ‘flow’ or implicitly by flowing into the default slot, forms a single flow, with content in document order. The slot establishes a block formatting context and becomes the containing block of the resulting content flow. The boxes of elements flowed into the same slot explicitly (by means of ‘flow’) are each other's siblings in the slot.

For example, the style sheet

BODY {grid: "a ." ". b"}
.paris {flow: a}
.london {flow: b}

with this document

<DIV CLASS=london>
 <P>The...
 <DIV CLASS=paris>
  <P>The...
  <DIV CLASS=london>
   <P>The...
  </DIV>
 </DIV>
</DIV>

causes the second and third DIVs to be taken out of their parents. The second DIV becomes the first child of slot a (i.e., of the pseudo-element called ‘::slot(a)’). The third DIV becomes the sibling of the first DIV, because both are added to slot b independently.

Here is another example. In this document

<P STYLE="grid: 'a b'">
 <SPAN STYLE="flow: a">First text <SPAN
  STYLE="flow: a">Second text</SPAN> Third
  text</SPAN>
 <SPAN STYLE="flow: b">Fourth text <SPAN
  STYLE="flow: auto">Fifth text</SPAN> Sixth
  text</SPAN>
</P>

the second SPAN has a non-auto ‘flow’ and is thus taken out of its parent's flow and added at the end of slot ‘a’. As a result, slot ‘a’ contains text in this order: “First text Third Text Second text.” This is in contrast to the fourth SPAN, which has ‘flow: auto’ and thus remains inside its parent's flow. Slot ‘b’ has this text: “Fourth text Fifth text Sixth text.”

The content flowed into a slot does not inherit properties from the slot.

Note that ‘flow’ applies to floating elements: they are floated relative to their containing block, and if their ‘flow’ property indicates a slot in a grid, that slot is their containing block. See also “Floating elements inside templates” below.

A common markup in HTML for illustrations with captions is as follows:

<div class=figure>
 <p><img src="paul.jpg" alt="...">
 <p class=caption>A pond in a playground in Amsterdam
</div>

The caption can be put above the image by using a template as follows:

div.figure {grid: * min-content *
                  "a a a"
                  ". b ."}

div.figure p {flow: b}
div.figure p.caption {flow: a; text-align: center}

The caption can be wider than the image and the image will be centered.

When the figure is floating, it is probably better to not let the caption become wider than the image (unless the caption cannot be made narrow enough):

div.figure {float: right; grid: "a" "b" min-content}
div.figure p {flow: b}
div.figure p.caption {flow: a; text-align: center}

In this example, a form is laid out on a grid, with two labels and two input boxes and a submit and a reset button:

form {
  border: thin solid;
  grid: "a a a a . b b b b"
        ". . . . . . . . ." 1em
        "c c c c . d d d d"
        ". . . . . . . . ." 1em
        ". . . e e . . f f" }
label[for=minv] { flow: a }
input#minv { flow: b; display: block }
label[for=maxv] { flow: c }
input#maxv { flow: d; display: block }
input[type=submit] { flow: e; display: block }
input[type=reset] { flow: f; display: block }

Here is the fragment of HTML that the style is applied to:

<form action="./">
  <label for=minv>Enter minimum value:</label>
  <input id=minv name=minv>
  <label for=maxv>Enter maximum value:</label>
  <input id=maxv name=maxv>
  <input type=submit value="OK">
  <input type=reset value="Reset">
</form>

The addition of ‘display: block’ causes the form controls to use the width computation of blocks, in other words: they will be as wide as their containing block, which in this case means that they will be as wide as the slot they are assigned to. Without it, they would be inline elements and just be left-aligned in their slots.

Image simulating the layout of the example

Possible rendering of the example.

This example shows that templates can be nested. The body has two columns. The #content element that goes into the second column has itself another template, into which the various “modules” are placed.

[Screendump with nested templates]

Possible rendering of the nested templates. (The borders are added for clarity, they don't occur in the style rules below. The red border is the inner template.)

For clarity, the inner template uses different letters for the slots than the outer template. This is not required.

<style type="text/css">
body {
     grid: 10em *
           "a   b";
}
#nav {
     flow: a;
}
#content {
     flow: b;
     grid:  *  1em  *  1em  *
           "c   .   d   .   e"
           ".   .   .   .   ." 1em
           ".   .   f   .   .";
}
.module.news {
     flow: c;
}
.module.sports {
     flow: d;
}
.module.personal {
     flow: e;
}
#foot {
     flow: f;
}
</style>
<body>
<ul id="nav">
     <li>navigation</li>
</ul>
<div id="content">
    <div class="module news">
         <h3>Weather</h3>
         <p>There will be weather</p>
    </div>
    <div class="module sports">
         <h3>Football</h3>
         <p>People like football.</p>
    </div>
    <div class="module sports">
         <h3>Chess</h3>
         <p>There was a brawl at the chess tournament</p>
    </div>
    <div class="module personal">
         <h3>Your Horoscope</h3>
         <p>You're going to die (eventually).</p>
    </div>
    <p id="foot">Copyright some folks</p>
</div>

</body>

This example shows the use of ‘same’ to put DD elements in the same slot as the preceding DT.

...
DL {grid: * 2em * 2em * "a . b . c"}
DT.mineral {flow: a}
DT.animal {flow: b}
DT.plant {flow: c}
DD {flow: same; margin-left: 1em}
...
<DL>
 <DT class=animal>falcon
 <DD>This bird of prey...
 <DT class=animal>rabbit
 <DD>Local restaurants...
 <DT class=mineral>granite
 <DD>This rock occurs...
 <DT class=plant>olive
 <DD>The fruit of...
 <DT class=mineral>limestone
 <DD>A rock composed of...
 <DT class=plant>pine
 <DD>The western half...
</DL>

screendump

Possible rendering of the DL list, with items sorted into three columns.

An alternative to ‘same’ may be to create selectors with regular-expression-like capabilities: ‘DT.plant + DD* + DD {flow: c}’ selects a DD that follows zero or more DDs that follow DT.plant.

Does a percentage ‘height’ work on an element that is flowed into a slot? If the grid template assigns a fixed height to the slot, it is obvious what a percentage means, but if the slot's height is ‘max-content’ and an element in the slot has ‘height: 110%’, there is a conflict…

Allowing multiple values on ‘flow’ could be an easy way to duplicate elements. E.g., this style

BODY {grid: "a . b . c ."}
H1 {flow: a, b, c}
#art1 {flow: a}
#art2 {flow: b}
#art3 {flow: c}

with a document like

...
<H1>Local news</H1>
<DIV ID=art1>...</DIV>
<DIV ID=art2>...</DIV>
<DIV ID=art3>...</DIV>

would repeat the H1 at the top of all three columns. With region-based styling, the three instances could even be styled differently, e.g., with generated content: ‘@region ::slot(b) { H1::before {content: "cont'd "} }

ISSUE-201: Terra Informatica's HTMLayout program proposes an interesting shortcut: if a slot's name is a number (instead of an identifier), it refers directly to the child with that number. That child is positioned in that slot and doesn't need ‘flow’ to be set. See the documentation by Andrew Fedoniouk and Ivan Goroun. E.g., <ul style="grid: '2 1'"><li>one <li>two </ul> puts “two” on the left and “one” on the right. It is thus also a shortcut for some applications of flex-order’.

4. Elements with ‘display’ set to ‘grid’ or ‘inline-grid

This section is not normative

The ‘grid’ and ‘inline-grid’ values of the ‘display’ property make an element into a grid container, (see [CSS-GRID-1]), which is an element that uses the ‘grid’ properties to define a grid, just like a grid element, but its chidren ignore the ‘flow’ property and are instead automatically made into separate flows and positioned in subsequent slots of the grid. (Although there are also properties to set their positions explicitly.)

Grid’ and ‘inline-grid’ also cause the template to be grow automatically with either extra rows or extra columns, if there are more children than predefined slots.

Note that the template of a grid container is also mirrored and/or rotated depending on the ‘writing-mode’ and ‘direction’ of the grid container, unlike the template of a grid element.

5. Styling slots

Slots have certain properties of their own, as defined below. If those properties are inherited, they inherit from the grid element. To set properties on the slots, two kinds of selectors are available: ‘::slot()’ and ‘::blank()’.

5.1. The ‘::slot()’ pseudo-element

The slots of a grid element can be individually addressed with the slot()’ pseudo-element.

For example, the following sets the background and vertical alignment of some of the slots in a template:

body { grid: "a a a"
             "b c d" }
body::slot(b) { background: #FF0 }
body::slot(c), body::slot(d) { vertical-align: bottom }

Only the following properties apply to the ‘slot()’ pseudo-element.

The following properties also apply to ‘::slot()’ pseudo elements, but only affect content inserted with the ‘content’ property other than what corresponds to the ‘contents’ keyword. (I.e., not to the content that is flowed into the slot via the ‘flow’ property or because the slot is the default slot. See “Styling the contents of slots with ‘@region’” for how to style that part of the content.)

ISSUE-37: Can a slot have a border and if so, where is it drawn?

The background of a slot is drawn immediately on top of the background of the element itself. E.g., the background set on ‘P::slot(a)’ is immediately in front of the background set on ‘P’.

Margins on slots do not collapse, neither with other slots, nor with elements inside the slot.

5.2. The ‘::blank()’ pseudo-element

The naming/syntax is an issue, because there is a ‘:blank’ pseudo-class in [CSS3PAGE] and a ‘:empty’ pseudo-class in [SELECT]. But ‘::slot(a):empty’ and ‘:slot(a):blank' are invalid syntax…

The 'blank()' pseudo-element selects named slots that have no content flowed into them. ’::slot(a)‘ and ’::blank(a)‘ select the same slot, but the latter matches only if the slot has no content flowed into it. Whether ’::blank()‘ matches is independent of the slot's 'content' property. (In other words: a “blank” slot is not necessarily empty and a slot that looks empty is not necessarily “blank,” although that is usually the case, given the initial value of 'content'.)

For example, slots a, b and c in the following document are blank, but slot * is not, even though slots b and c do not look empty, (because they have generated content) and slot d does look empty (because its 'content' property doesn't include the keyword ’contents'‘):

<style>
body { grid: "a b"
             "c *" }
body::slot(b) { content: "This is slot b" }
body::slot(c) { content: "Page " counter(page) }
body::slot(*) { content: "No content" }
</style>
<h1>The heading</h1>
<p>The first paragraph.

All content in this document is flowed into the default slot (*), so all other slots are blank.

This example shows how to put a border only on slots that have content:

div {grid: "a ."
           ". d"}
div::slot(a), div::slot(d) {border: solid}
div::blank(a), div::blank(d) {border: none}

An element that has no content, no background, no border, no padding, no margin and no fixed positive width or height has no influence on whether the slot it flows into is blank or not. In this case, the content of the element is the content that results from applying the ’content' property and also includes generated content of its ‘::before’ and ‘:after’ pseudo-elements.

For example, even though there is an address element flowed into slot a, slot a is still blank, because the address element happens to have no printable content:

<style>
body { grid: "a ."
             ". *" }
body::blank(a) { content: "No address defined" }
address { flow: a }
</style>
<h1>The heading</h1>
<p>The first paragraph.
<address></address>

A slot that is part of a chain (see ‘chains’) is blank if all content flowed into that chain fits into slots earlier in the chain.

For example, this document has very little content that easily fits into the first slot. Slot b, which is part of the same chain, thus remains blank.

<style>
body { grid: "a . b"; chains: a b; width: 60em }
body::blank(b) { background: silver }
</style>
<p>Very little content.

5.3. Slots and the ‘content’ property

The ‘content’ property applies to slot pseudo elements in a similar way as it applies to normal elements and other pseudo-elements. When applied to slots, values have the following meaning:

normal
Same as ‘contents(Note that ‘contents’ is also the computed value in this case.)
none
The content that is flowed into the slot (with the ‘flow’ property, or because the slot is the default slot) is not rendered.
inhibit
Same as ‘none’.
contents
Renders as the content that is flowed into the slot (with the ‘flow’ property, or because the slot is the default slot).

Other values are as defined in [CSS3GENCON] and in [CSS3PAGE]. See “Page-based grid templates” for an example of the use of ‘string()’ in the value of ‘content’.

[This should be moved to [CSS3GENCON].] Even if the value of ‘content’ does not include ‘contents’ and the content that is flowed into a slot is thus not rendered, any counters associated with that content are incremented normally.

Note: as defined in [CSS3GENCON], the keyword ‘contents’ may appear multiple times in the value of ‘content’, but the second and subsequent occurrences are ignored. (Those occurrences are still part of the computed value.)

6. Styling the contents of slots with ‘@region

The ‘@region’ at-rule [CSS3-REGIONS] provides a way to select an element, or a part of an element, based on the region it is rendered in. For this purpose, a slot is a particular kind of region.

For example, with a style sheet like this

BODY { grid: "a ." 10em ". b"; chains: a b }
P { flow: a; color: blue }
@region BODY::slot(b) { P { color: orange } }

and a document fragment like this:

<BODY>
 <P>The text of the first paragraph...
 <P>The text of the second paragraph...
 <P>The text of the third paragraph...
</BODY>

The three paragraphs will be rendered inside slot a and, if there is more content than fits in that slot, the rest will be rendered in slot b. Ps are blue by the second line in the style sheet, but the ‘@region’ rule says that any part of a P that is inside slot b will be orange.

Note that ‘@region BODY::slot(b) { P {…} }’ cannot be replaced by ‘BODY::slot(b) P {…}’, because pseudo-elements can only occur as the last component in a selector. Compare also ‘@region ::first-line { em {…} }’, which is a valid way to style (the part of) an EM that occurs on a first line, vs ‘::first-line em {…}’ which is an invalid selector.

To select a slot for use in an ‘@region’ rule, the selector after the ‘@region’ keyword must be a selector that ends with the pseudo-element ‘::slot(X)’.

See “Style the contents of slots (region-based styling)” below for a discussion of alternatives to ‘@region’.

7. Rendering of grid elements

Grid elements influence the stacking order, page breaks, the position of floats, etc. of the content inside them.

7.1. Vertical alignment of the contents of slots

Vertical-align’ applies to slots in a similar way as it applies to table cells [CSS3TBL] and margin boxes [CSS3PAGE].

Align-content’ in [CSS3-FLEXBOX] has ‘flex-start’, ‘flex-end’ and ‘center’, which are a bit similar to ‘top’, ‘bottom’ and ‘middle’. But it also has ‘space-between’ and ‘stretch’, which make little sense in a flow, and it lacks ‘baseline’. The similarity of slots to table cells and margin boxes argues for sticking with ‘vertical-align’.

For the purpose of this section we define the A edge and C edge of a box as a writing-mode-dependent edge as follows:

Value of ‘writing-modeMeaning of “A” Meaning of “C”
horizontal-tbtop bottom
vertical-rlright left
vertical-lrleft right

E.g., if a box is horizontal, the “A edge” is the top edge.

The ‘vertical-align’ property of a ‘::slot()’ pseudo-element can be used to align elements vertically in a slot (or horizontally, if the slot is vertical). The effect is as if the hypothetical anonymous block that contains the slot's contents is positioned as defined below.

bottom
The content of the slot is aligned to the C edge of the slot: the C margin edge of the anonymous block coincides with the C margin edge of the slot.
middle
The content of the slot is centered in the slot: the distance between the A margin edge of the anonymous block and the A margin edge of the slot is equal to the distance between the C margin edge of the anonymous block and the C margin edge of the slot.

(Note that if the content overflows the slot, it will overflow at both edges.)

baseline
The anonymous block that encloses the content is placed as far as possible in the direction against the block flow direction under two constraints:
  1. The A margin edge of the anonymous block may not be beyond the A margin edge of the slot.
  2. If the content has a relevant first line (see below), then the baseline of that line must align with the baselines of the relevant first lines in all other slots with the same ‘writing-mode’ and the same ‘vertical-align’ in the same row (if the slot is horizontal) or the same column (if the slot is vertical). A slot has a relevant first line, if the content has a first line (ignoring any lines inside floats) and that first line has the same ‘writing-mode’ as the slot itself.

For example, for a horizontal slot, this means that the first baseline must be aligned with the first baselines of all other horizontal slots in the row that also have ‘vertical-align: baseline’.

<percentage>
0%’ means the same as ‘bottom’, ‘100%’ means the same as ‘top’, other values are linear interpolations of these. Negative values and values over 100% are interpreted as 0% and 100% respectively.

Note that 100% minus the percentage corresponds to the initial position of the scrolling mechanism (if any) in case the content overflows.

For all other values, the content is aligned as for ‘baseline’.

Note that baseline alignment may cause a slot to overflow (if the slot is the last in a ‘chain’) or to remain empty (because the first line box is put in the next chained slot instead).

Example: Given a document like this

<BODY>
<P>...</P>
<FIGURE>...</FIGURE>
<P>...</P>
<FIGURE>...</FIGURE>
<P>...</P>
<FIGURE>...</FIGURE>
<P>...</P>
<FIGURE>...</FIGURE>
</BODY>

and a style like this:

BODY { grid: "a b" }
P { flow: a }
FIGURE { flow: b }

the paragraphs will be in the first slot (a) and the figures in the second (b). There are two ways to align the contents of the two slots to the bottom. One is with ‘vertical-align’:

BODY::slot(a), BODY::slot(b) { vertical-align: bottom }

The other is with flexible margins:

P:first-of-type, FIGURE:first-of-type { margin-top: fill }

Example: There is no value for ‘vertical-align’ to distribute elements vertically over the available space (similar to how ‘text-align: justify’ distributes words over the available space). But the effect can be achieved with stretchable margins. Assume the same document as in the previous example and this style sheet:

BODY { grid: "a b" }
P { flow: a; margin-top: fill; margin-bottom: fill }
FIGURE { flow: b }

Because there are stretchable margins between the Ps, the Ps will be equally distributed over the height of the slot (assuming the slot is taller than its contents, i.e., assuming the contents of slot b is taller than the contents of slot a).

Note that, if there are one or more margins in the contents of the slot with a ‘fill’ value (see [CSS3BOX]), the height of the content fits the height of the slot and thus the ‘top’, ‘middle’ and ‘bottom’ values are indistinguishable.

7.2. Breaking grid elements across pages or columns

Grid elements may be broken across pages, columns or similar regions (including chained slots of another grid element), subject to the ‘break-before’, ‘break-after’ and ‘break-inside’ properties. In addition to the break points listed in the Fragmentation module [CSS3-BREAK], a page break may occur between two rows in a template, if there is a possible break point at the same height or higher in all slots that span those two rows; and a page break may occur inside a row if there is a possible break point in all slots in that row.

Try to be more precise?

In the terminology of [CSS3-BREAK], a slot is a fragmenter of type ‘region’.

A forced break on an element in a slot causes the rest of the flow of that slot to continue in another page, column or slot (depending on the type of break). The following cases are special:

For example, this document fragment has a column element inside a grid element. Column breaks affect the content of a single slot, but do not affect the other slots of the grid element:

<DIV STYLE="grid: 'a b' 'c d'">
 <DIV STYLE="flow: a">
  I'm in slot a.
 </DIV>
 <DIV STYLE="flow: b; columns: 20em">
  <P>This text is in columns.
  <H2 STYLE="break-before: column">Heading at top of column</H2>
  <P>...
 </DIV
</DIV>

For example, this document fragment has a grid element inside a column element and thus the column break on the H2 is ignored:

<DIV STYLE="columns: 20em">
 <DIV STYLE="grid: 'a b' 'c d'">
  <P>I'm inside slot a inside some columns.
  <H2 STYLE="break-before: column">No break</H2>
  <P>...
 </DIV>
</DIV>

A slide presentation can be made with a template for each page (i.e., slide) and forced page break between the slides:

@page	{ grid: "a"  5em
	        "@"  *
	        "b"  auto }
h1	{ page-break-before: always;
	  flow: a }
p.note	{ flow: b }

With a document similar to this: (fragment)

<h1>Slide 1 title</h1>
<ul>
 <li>Topic one
</ul>
<h1>Slide 2 title</h1>
<ul>
 <li>More topics
</ul>
<p class=note>Note the note

The document in the example above doesn't have an element that corresponds to a slide; a slide simply starts at an H1 and ends before the next H1. But if there is a DIV around each slide (as is the case in many slide formats in practice), the same effect can also be achieved without page-based templates, by using the ‘vh’ unit [CSS3VAL]:

div.slide {height: 100vh; grid: "a" 5em "*" "b" intrinsic;
    page-break-before: always}
h1 {flow: a}
p.note {flow: b}

With a document similar to this: (fragment)

<div class=slide>
 <h1>Slide 1 title</h1>
 <ul>
  <li>Topic one
 </ul>
</div>
<div class=slide>
 <h1>Slide 2 title</h1>
 <ul>
  <li>More topics
 </ul>
 <p class=note>Note the note
</div>

7.3. Stacking order

Each slot generates a separate stacking context. The stacking order of these stacking contexts is given by their ‘z-index’ property, with ‘auto’ treated as ‘0’. Slots with the same ‘z-index’ are stacked in the order in which they are defined in the ‘grid-template-areas’ property, looking only at the first occurrence of the slot's name (i.e., the top left corner if the slot spans several rows or columns).

A slot thus has a very similar behavior in the stacking order as an element with ‘position: relative’ and ‘z-index: 0’ (or another value of ‘z-index’ other than ‘auto’).

For example, the stacking order of the slots in the following template is, from back to front, a, x, c, d, e, p, g.

grid: "a x x c"
      "d x x e"
      "p p g g"

The stacking order can be made visible, e.g., by giving slots opaque backgrounds and negative margins, so that they overlap. See the next example.

This example uses ‘z-index’ and negative margins to make the middle slot partly overlap the other slots:

body { grid: "a . b"
             ". c ."
             "d . e";
       height: 240px;
       width: 240px }
::slot(a) { background: #0C0 }
::slot(b) { background: #C00 }
::slot(c) { background: #FD0; margin: -20px; z-index: 1 }
::slot(d) { background: #00C }
::slot(e) { background: #A0A }

Five colored rectangles

Rendering of the above example.

This example has a heading (H1) that is wider than its slot and overlaps the image in the slot next to it:

body {grid: "a b"}
::slot(a) {z-index: 1}
h1 {flow: a; width: 200%}
p {flow: a}
img {flow: b}

The title overlaps the image

Rendering of the above example.

The ‘z-index’ is necessary, because the ‘a’ slot comes before the ‘b’ slot in the template and all its content would thus be rendered behind the ‘b’ slot with default ‘z-index’ values.

7.4. Floating elements inside templates

An element may be flowed into a slot and be a floating element at the same time. The following cases must be distinguished:

The ‘@footnote’ at-rule from [CSS3GCPM] needs to be extended to apply to slots: ‘@footnote :first::slot(a)’ is the footnote area of slot a on the first page.

8. Page-based grid templates

A template can also be attached to a page, rather than an element. Such a template is called a page-based template as opposed to an element-based template.

Here is an example of a template used to position two centered running headers with different styles. (This would be impossible to do with the predefined page template of [CSS3GCPM], because it has only one centered margin box at the top.)

@page {
 grid: "t1" 1.2em    /* space for 1st running header */
       "t2" 1.2em    /* space for 2nd running header */
       "." 2em
       "*"           /* page body */
}
::slot(t1) { content: string(chapter); color: red; text-align: center }
::slot(t2) { content: string(section); color: green; text-align: center }

h1 { string-set: chapter contents }
h2 { string-set: section contents }

See “Slots and the ‘content’ property” for the definition of the ‘content’ property on slots.

The syntax of a page-based template is the same as that of an element-based one, but the declaration appears in an ‘@page’ rule.

In a page-based template, the height and width are typically known (defined by the output media and the margin boxes, see [CSS3PAGE]). And content that overflows a slot typically is not clipped (or hidden behind a scrolling mechanism), but is continued on a subsequent page.

Because the grid template is not attached to an element but to a page, the slot names have global scope. But they can be hidden to elements that have a grid ancestor that uses the same slot names.

This modifies the rules for ‘flow’: if the property refers to a slot that is not found in a grid ancestor, it refers instead to a slot in a page template (if there is one and it has a slot of that name).

Should we simplify page-based templates to only allow <length> and * as row heights and column widths? (And treat illegal sizes as ‘*’?)

@page :first {grid: "a b c" "d e f"}
@page {grid: "d e f"}

body {flow: e}
h1 {flow: b}

A page that has a grid template does not have a footnote area [CSS3GCPM]. Instead each slot in the grid template has its own footnote area.

If a slot of a page-based template on non-interactive media has an ‘overflow’ property of ‘visible’, then content that overflows that slot in the block progression direction (i.e., below the slot in the case of horizontal text) causes a page break and is continued on the next page.

What happens in non-interactive media with an ‘overflow’ of ‘scroll’ or ‘auto’?

For page breaking purposes, each slot is considered as a page and the page break properties on the elements in that slot determine where the content for that slot is broken [CSS3-BREAK]. Content after the page break is put in the slot of the same name on the next page that has such a slot. If there is no such page, the UA should display the content on a separate page.

Or: the content after the page break is not displayed? displayed in the default slot?

Note that this may happen if the template for the first page (‘@page :first’) uses a slot name that occurs in no other @page rule. Possibly also if a page template is bound to a “named page” [CSS3GCPM] and that named page is not allowed to repeat. (At the time of writing, this is only a proposal in the GCPM Module.)

Note that an element A that appears later in the document than an element B may thus be displayed on an earlier page than element B, because their respective slots are broken across pages in different ways.

Because of limitations of a device (e.g., limited memory), it may be that a UA has to print a page (force page breaks) even though some slots aren't filled yet.

This example shows a document with text in two languages, which are to be shown side by side:

@page {grid: "a b"}
:lang(en) {flow: a}
:lang(fr) {flow: b}

This works with a document where the languages are separated like this:

<BODY>
 <DIV LANG=en>
  <H1>The blue house</H1>
  ...
 </DIV>
 <DIV LANG=fr>
  <H1>La maison bleue</H1>
  ...
 </DIV>
</BODY>

But also with a document where the languages are interleaved:

<BODY>
 <H1 LANG=en>
 <H1 LANG=fr>
 <P LANG=en>...
 <P LANG=fr>...
 ...
</BODY>

This example shows how the first page may have a different layout from the other pages. The slot ‘a’ only occurs on the first page. If the content for that slot (in this case: all H1 elements) is too long, the remaining content of that slot will not be displayed. The slot ‘@’ occurs on normal pages and all its content can thus be displayed by adding additional pages.

@page :first {grid: "a" "*"}
@page {grid: "*"}
h1 {flow: a}

Note that “page masters” (sequences of different templates for sequences of pages) can be made with the selectors defined in [not yet decided].

Both page-based and element-based templates can be used in the same document.

@page {grid: "a*"}
:lang(fr} {flow: a}
div.special {grid: "a b c" "a b d"}

Here is a page as one might find in a newspaper. It combines a layout template with multicolumn layout.

5-column newspaper page with some   blocks of text that span several columns

The front page of a newspaper, with the first parts of several stories that are continued on other pages and headings and images that span several columns.

@page :first {
  grid:  * 3em * 3em * 3em * 3em *
        "A  A  A  A  A  A  A  A  A"  5cm
	".  .  .  .  .  .  .  .  ."  0.25cm
	"B  .  C  C  C  C  C  C  C"  *
	"B  .  C  C  C  C  C  C  C"  *
	"B  .  C  C  C  C  C  C  C"  *
	"B  .  C  C  C  C  C  C  C"  *
	"B  .  C  C  C  C  C  C  C"  *
	"B  .  D  D  D  D  D  D  D"  *
	"B  .  D  D  D  D  D  D  D"  *
	"B  .  E  E  E  .  F  F  F"  *
	"B  .  E  E  E  .  F  F  F"  *
	"B  .  E  E  E  .  F  F  F"  *
}
h1 {flow: a; border-bottom: thick; margin-bottom: 1.5em}
#toc {flow: b; margin-right: -1.5em; border-right: thin;
  padding-right: 1.5em}
#leader {flow: c; columns: 4; column-gap: 3em}
#art1 {flow: d; columns: 4; column-gap: 3em; border-top: thin}
#art2 {flow: e; columns: 2; column-gap: 3em}
#art3 {flow: f; columns: 2; column-gap: 3em}

If a slot on a page is full and the content continues on the next page, it may be useful to insert something like “continued on page X.” This is useful at any page break, but more important if there are multiple “flows” of content on each page. Maybe a break-content property? ‘break-content: "▶ continued on p. " targetcounter(???, page)’ or extend text-overflow from [CSS-TEXT-3]?

How do you style the slots of page-based template? E.g., with

@page :first {
  grid: "a b"
        "c d" }
@page {
  grid: "a a a"
        "b c d" }

::slot(a) {background: silver}

the background could be set on all “a” slots on all pages. But how do you set a background only on the “a” slot of the first page? Maybe we need to extend the page selectors and allow something like this:

@page :first::slot(a) {background: silver}

A space between the pseudo-class and the pseudo-element could be optional.

9. Chaining slots: the ‘chains’ property

Slots must be rectangular, but the appearance of non-rectangular slots can be achieved to some extent by chaining slots together. Content that is positioned in the first slot of a chain is automatically continued in the second slot if the first slot is full, and then the third, etc.

Name: chains
Value: none | <identifier>+ [ , <identifier>+ ]*
Initial: none
Applies to: grid elements
Inherited: no
Animatable: no
Percentages: N/A
Media: visual
Computed value: specified value
Canonical order: per grammar

A value of ‘none’ means the element's template has no chains. Otherwise the value consists of one or more comma-separated lists of identifiers. No identifier may appear more than once in the value. Identifiers that do not occur in the template are ignored, but do not make the value invalid. A list with only ignored identifiers is itself ignored.

All the non-ignored identifiers in a list, except for the last one, must refer to slots whose size does not depend on their contents, otherwise the list is ignored. The size of a slot does not depend on its content if all the columns and all the rows that the slot spans have a width, respectively height, that is a <length> or ‘*’.

Each non-ignored list defines one chain.

Each chain is filled with content in an analogous way to the pages in paged media: all the content that is positioned to the first slot in the chain is flowed, in document order, into the first slot in the chain until the slot is full, the rest is flowed into the second slot until it is full, etc.

Content must only be split between slots at an allowed page break [CSS3PAGE]. As for page breaks, if a break occurs in the margin between blocks, all adjoining margins are set to zero.

CSS3 does not define what is the best break point to split content over slots. However, it is recommended to use the last possible break point that does not cause overflow. (If such a break point exists.)

Note: It is the author's responsibility to make the height of relevant slots an integral number of lines if he wants to ensure that the lines in chained slots are aligned. The height does not automatically “snap” to a multiple of the line height.

The following template creates a double-staircase layout. In case the content is too long for the staircase, a slot of flexible height is added at the bottom.

div { grid:    "@ @ . . f f . . . ."  3.6em
               ". a a . . g g . . ."  3.6em
               ". . b b . . h h . ."  3.6em
               ". . . c c . . i i ."  3.6em
               ". . . . d d . . j j"  3.6em
               ". . . . . . . . . ."  0.6em
               "e e e e . . k k k k"  auto;
   chains: @ a b c d e, f g h i j k}
#first { position: @ }
#second { position: f }

This could be applied to a document fragment such as

<DIV>
  <P ID=first>...
  <P ID=second>...
</DIV>

Here is a paragraph shaped as a circle:

p { width: 12em;
    grid: ". . . . * * . . . ."  1.2em
	  ". . a a a a a a . ."  1.2em
	  ". b b b b b b b b ."  1.2em
	  ". b b b b b b b b ."  1.2em
	  "c c c c c c c c c c"  1.2em
	  "c c c c c c c c c c"  1.2em
	  ". d d d d d d d d ."  1.2em
	  ". d d d d d d d d ."  1.2em
	  ". . e e e e e e . ."  1.2em
	  ". . . . f f . . . ."  1.2em
	  "g g g g g g g g g g"  auto;
   chains: * a b c d e f g }

Here is a page-based template that creates a two-column layout with a “hole” in the center:

@page:first {
    grid:
    "a a a a a a a . d d d d d d d"
    "a a a a a a a . d d d d d d d"
    "a a a a a a a . d d d d d d d"
    "a a a a a a a . d d d d d d d"
    "b b b b . . . . . . . e e e e"
    "b b b b . g g g g g . e e e e"
    "b b b b . g g g g g . e e e e"
    "b b b b . g g g g g . e e e e"
    "b b b b . . . . . . . e e e e"
    "c c c c c c c . f f f f f f f"
    "c c c c c c c . f f f f f f f"
    "c c c c c c c . f f f f f f f"
    "c c c c c c c . f f f f f f f";
    chains: a b c d e f }
@page::slot(g) {
    vertical-align: middle }

body { flow: a }
h1 { flow: g }

Note: For more anaylysis of the possibilities and limits of non-rectangular slots, see the PhD thesis of César Acebal [ACEBAL2010].

Chaining of rectangular slots is not enough to create layouts with holes, e.g., an image in the middle of a text. Allowing non-rectangular, connected regions (in addition to chaining) would allow an example such as this:

grid: "A A A . . ."
      "A . A . . ."
      "A A A . . ."
      ". . . B B B"
      ". . . B . B"
      ". . . B B B";
chains: a b;

Such cutouts in the middle of text usually create text that is difficult to read, and that is why there is no ‘float: center’, e.g. But the CSS WG is considering a new property ‘wrap-flow’ for absolutely positioned elements that would allow, e.g.: ‘p {position: absolute; top: 1fr; left: 1fr; width: 1fr; height: 1fr; wrap-flow: both}’ to absolutely position a P element on top of a grid element and cause the content of the element under it to wrap around it as if it were a float [CSS3-EXCLUSIONS].

Note that a slot can have overflowing content even if it is part of a chain: it can have content that is too wide but cannot be broken.

The ‘break-before’, ‘break-after’ and ‘break-inside’ properties have values that control breaking of content between slots in a chain (in particular ‘region’ and ‘avoid-region’, see [CSS3-BREAK]).

Note that, as defined in “Breaking grid elements across pages or columns” above, a slot is a fragmenter of type region in the terminology of [CSS3-BREAK].

10. History

The following sections contain some of the use cases and design decision that led to this module and influenced its evolution.

10.1. Developing a model of declaring, using, styling and resizing grids

(This section is not normative.)

The following types of use cases were considered for template-based layout.

  1. Standard Web pages.

  2. Grids and other table-like layouts. This includes grid layouts, frame layouts and table-like subdivision of a rectangular area.

  3. A layout structure with “flex”ing information. The flexing is represented by constraints that specify how the cells are to relate to one another: which cells are to be allowed to grow or shrink and how much. There may also be a priority ordering, which determines, based on the size of the allowed display window, which cells shrink, which grow and under which conditions.

  4. Layout structures with absolutely positioned (fixed-size) elements; for example a block of text into which several illustrations intrude at fixed positions within the block. This is like a float with respect to tightly wrapping the text around the intrusion, but the position of the intrusion is determined by the layout structure, not the content flowed into that structure.

    An example of this is a multicolumn layout with one or more “absolutely positioned floats” that intrude on the columns (see figure).

    An image that partially overlaps two columns makes the text wrap around it on both sides.

    An image (or a “pull quote”) is placed centered on the page and intrudes on two areas.

  5. Multiple, disconnected, fixed-size areas on a page that are chained together, each one receiving the content that doesn't fit in the previous slot. In combination with page breaks, this may give a layout as often seen in newspapers: the first few lines of each story on the first page, the rest of the story in other areas on subsequent pages. (It will probably need a way to conditionally insert “continued on page 7” or similar text.)

For comparing proposals for template-based layouts, the working group identified four important aspects of each proposal:

  1. the physical layout structures – the way of structuring the “cells” (slots) into which content is flowed. This includes a way to identify the various layout containers.

  2. the binding mechanism – the way to specify that a given element (and its descendants) are to be placed in a given layout cell.

  3. the property distribution mechanism – the way to put properties onto the layout structure and the cells within it.

  4. the flexing mechanism – the way to describe how the layout structure should adapt itself to the higher level container (window) in which it is placed. This includes statements about which cells should grow and when they should grow.

In this specification, these aspects are as follows:

  1. A character matrix is used to show the layout structure and the cells are named by the character used to show where they are positioned.

  2. The binding of content to cells is handled by the ‘flow’ property which identifies a cell to which the content is bound.

  3. The shape, size and flexibility of the layout are specified with the character matrix. Some properties (background, border and vertical alignment) are attached to individual slots.

  4. There is limited “flexing” information. The choice is between fixed size, a fraction of the available size or the content's intrinsic size. (The latter is further subject to min/max sizes specified on that content.) It is not possible to say, e.g., that some column can only become wider if all other columns are at their maximum size.

10.2. Syntax, features and levels

(This section is not normative.)

The following is a partial list of design decisions and some arguments for and against each choice:

10.2.1. Named vs numbered slots

Named slots are very easy to understand and use. Experience with the draft showed that everybody who sees an example immediately understands what it means. Moreover, there is no need to use numbers or to count.

But it is different for absolutely positioned elements that use the grid. Those elements do not belong to a slot, they are merely placed on top of it, overlapping what is already there and each other.

Also, if grids are automatically extended with extra rows and columns (see below) based on content, then those extra rows and columns cannot have names and must (probably, see below) be referred to by number.

In this specification, named slots are used to create flows of content, while numbers are used for absolute positioning, although names can be used there too, if the author prefers.

10.2.2. Single letter names vs identifiers

The set of single letters is finite. And even if Unicode has many symbols, typing them may not be easy. Letters also do not permit to give a hint as to a slot's function (“heading,” “footer,” “advertisement”…).

On the other hand, any Unicode character can be typed as an escape, so if you run out of letters on the keyboard, you can always make a template like this:

DIV {grid: "\1001\1002\1003\1006"
           "\1001\1004\1005\1006"}
DIV H1 {flow: \1001}

In practice, it is also hard to come up with meaningful identifiers and so single letters require less thinking. And if you always have single letters, you don't need spaces between them either, which makes big templates easier to read.

This specification defines that a template consists of (space-separated) identifiers, but all the examples use single-letter identifiers.

10.2.3. Extend ‘display’ or add a grid property

Grids can be defined on elements such as table cells, blocks, inline blocks or list items, but not on elements such as tables or, inline elements. That can be expressed in the syntax by adding the grid to the ‘display’ property: ‘display: inline-block "abc" "abd"’ is valid, but ‘display: inline "abc" "abd"’ would not be. Or it can be expressed by an implicit dependency between properties: ‘display: inline-block; grid: "a b c" "a b d"’ has a grid, but ‘display: inline; grid: "a b c" "a b d"’ ignores the grid. (In this case it is actually the computed value of ‘display’ that applies, so, e.g., if the element floats, it would be a block and thus have a grid.)

The Multicol specification uses the latter model: ‘columns’ is a separate property and only applies if the element is a block container. Given that precedent, it seemed reasonable to do the same here.

10.2.4. Combining columns and grids

Style rules can specify both a grid and columns for the same element:

DIV {columns: 20em; grid: "a a b c c c"}

The specification could say that grids don't apply to column elements, or vice versa, or it could say that the columns apply to contents of the DIV, after it has been flowed into the default slot. This last option seemed the most useful. It is consistent with other elements that flow into a slots: they, too, can have columns.

10.2.5. Shorthand vs single property

It is convenient to specify the grid as the value of a single property. It puts the column and row sizes immediately next to the rows and columns they apply to. The typical column widths can either be omitted or are short values like ‘*’ and ‘1em’, and they fit easily in one line.

But if a grid has very many columns, or their sizes are complex (with many ‘minmax()’ and ‘calc()’ values, e.g.), it may be more readable to separate the slot names from the column sizes.

Also, separable properties for row and column sizes makes is easier to specify a transition/animation on them,, because you don't have to repeat the slot names (which cannot animate).

For those reasons, this specification defines the ‘grid’ property as a shorthand, at the cost for authors of having to learn four new properties instead of one (or zero, if the grid had been added to display (see above).

10.2.6. The number of rows and columns

If the grid template can also be specified with individual properties, how many columns & rows does it have if those properties contradict each other? Does the template (if not ‘none’) determine the number of columns or the maximum of the template and the list of sizes? E.g.:

Specified values Meaning if the template prevails Meaning if the maximum value is used
grid-template-areas: "a b c";
grid-template-columns: * * * * *
grid-template-areas: "a b c";
grid-template-columns: * * *
grid-template-areas: "a b c . .";
grid-template-columns: * * * *

The former is consistent with the model for ‘background-image’ and ‘background-size’. However, unlike for backgrounds, where a size without an image makes no sense, in this case a size without a slot name can make sense: it can represent empty space (‘.’) of the given size.

What is more likely: that people consciously specify empty rows and columns without putting dots in the template (to save typing or to avoid typing too many dots), or that people mistakenly add a size too many?

The specification currently sets the number of rows and columns to the maximum of the three properties.

10.2.7. *’ vs ‘fr

ISSUE-127: It is very common for the columns of a grid to have all the same width. That is because the slots in a grid, although of different widths themselves, typically are multiples of some fundamental measure. And thus equal columns is the default in the current syntax. And if it needs to be specified explicitly which columns have the same width (e.g., because there are also other columns in the grid) then the symbol for it is short: ‘*’. This is ideal for slots whose widths are small multiples of the fundamental measure. E.g., the widths of slots a, b and c in the following grid relate to each other as 1:3:2:

grid: "a b b b c c"
      "a . . . c c"

However, if the ratios of the slots are ratios of larger numbers, the notation may become long. E.g., to make slots a and b with ratio 10:11, you would have to make 21 columns:

grid: "a a a a a a a a a a b b b b b b b b b b b"

It needs to be investigated if such ratios are frequent. (They don't seem to occur in printed magazines.) And if among the ratios that do occur, even if infrequently, there are some that would be impractical to write in this way.

Assuming another notation is needed, it could be made by prefixing a number (e.g., ‘7* 10*’) or, to make the notation look more like a dimension, by creating a special unit ‘fr’ (‘e.g., ’3.5fr 5fr', which is the same as 7 + 10 stars).

If ‘fr’ is added, it could either be in addition to ‘*’ or instead of ‘*’. In the former case, ‘*’ = ‘1fr’.

Unless it is shown that the notation with only ‘*’ is not practical in real cases, it seems better to neither introduce numbers (‘7*’) nor units (‘3.5fr’). It would give too many different ways to write the same grid, causing authors and readers of style sheets to have to think longer about each grid. (E.g., ‘grid: "a b b b c c"’ could then also be written as ‘grid: 1* 3* 2* "a b c"’ or ‘grid: 0.5fr 0.5fr 0.5fr 0.5fr 1fr "a b b b c"’, etc.)

For the moment, the specification allows both the ‘fr’ and * notations.

10.2.8. Automatically add rows and columns

The main purpose of grids is to improve on absolute positioning for displaying elements in a different visual order. You typically know how many positions you need in advance.

If you don't need to change the visual order, then table layout can often align the elements sufficiently.

However, if table layout isn't possible (there are not enough elements in the source to create the necessary table rows, the control over the size of table cells is too limited, or you want a column-major table), it may be useful to use a grid in a similar manner to a table: the number of rows and columns is not specified in the style, but depends on the number of elements that are put in the grid.

E.g., you could transpose a table with an unknown number or rows like this:

<TABLE>
 <TR><TD>A1 <TD>A2 <TD>A3
 <TR><TD>B1 <TD>B2 <TD>B3
 ...
 <TR><TD>K1 <TD>K2 <TD>K3
 ...
</TABLE>

into a tabular display with an unknown number of columns somewhat like this:

A1 B1... K1...
A2 B2... K2...
A3 B3... K3...

with style rules like this:

TABLE {grid-template-columns: auto; grid-template-rows: auto}
TD:first-child {grid-area: next 1}
TD {position: grid; grid-area: same next}

This uses keywords instead of numbers, as described in “Automatic placement of elements” below.

What is the size of the added rows and columns? The example above assumes the size of the single, specified grid cell is simply repeated. There could also be a marker in the list of sizes to indicate which set of sizes is repeated. Or there could be a separate property with the list of sizes that is repeated for all added rows and columns.

It would probably be difficult to specify that any number of columns of size A can be added, but that the last column must have size B…

The current specification specifies that extra rows and columns are added when needed for absolute positioning (‘grid-area’). It is not possible to automatically create new, anonymous flows. (But sometimes you can use multi-column elements to do that, and possibly grid templates can be attached to columns, via a ‘::column’) pseudo-element, see [CSS3GCPM].)

10.2.9. Rectangular slots vs non-rectangular slots

Non-rectangular slots, such as

grid: "a a a"
      "a a ."
      "a . ."

are probably something for level 4. Or maybe the Exclusions module can be used instead.

10.2.10. Chained slots

Many layouts found in magazines can be handled with multi-columns [CSS3COL] (possibly extended with ideas in GCPM [CSS3GCPM]) or with multi-column elements inside a grid, but some more complicated designs appear to require chained slots (and some would be easier to specify that way than with columns).

grid: "A A E"
      "A A E"
      "B B E"
      "C D E";
chains: B C D E

Also, when grids are applied to pages (see “Page grids” below), it is essential that slots are connected to slots on the next or some subsequent page, so that the overflow can find a place.

But this could be a feature for level 4.

10.2.11. Page grids

Typically in magazines, each page has a slightly different layout of slots, although they are all based on the same set of columns, usually between 3 and 6. It is difficult to specify the grids of all pages as a single long grid on the BODY element; and indeed impossible if you don't know how many pages use each kind of layout.

Applying grids to ‘@page’ is an alternative. It has a few issues:

This is probably for level 4.

10.2.12. Style the slots (pseudo-elements)

Slots can have properties of their own: background, overflow, writing-mode, direction, box-shadow, margin, border, padding. What else?

For ‘vertical-align’, see “Vertical alignment inside slots” below.

Wrap-flow’ (from CSS Exclusions and Shapes) could be useful for slots with negative margins, too.

10.2.13. Style the contents of slots (region-based styling)

By means of a new pseudo-element, style could be applied to only the part of an element that is inside a certain slot (similar to how ‘::first-line’ applies style only to the part of an element in the first line of a paragraph):

P::flow(a) {font-size: large}

would be an alternative to

@region ::slot(a) { P {font-size: large} }

The notation with ‘::flow()’ is shorter, and more similar to ‘::first-line’ at first sight. However, ‘EM::first-line’ does not select the part of the EM inside a first line, while ‘EM::flow(a)would select the part of the EM inside slot a. Also, the existence of two pseudo-elements, ‘::slot()’ and ‘:flow()’ is confusing.

An alternative is to allow simple selectors after pseudo-elements:

::slot(a) P {font-size: large}

could select the parts of a P that are rendered inside slot a. That would mean that ‘P::first-line EM’ is also valid and selects the parts of an EM that are rendered on the first line of a P.

Another possibility might be to set styles on ‘::slot()’ itself and only rely on inheritance. (This is what is proposed by David Baron in [CSS-OVERFLOW-3].) Elements inside a grid element then inherit from the slot inside which they are, before they inherit from the grid element itself.

Yet another possibility is a generic ‘region()’ pseudo-element:

P::region(slot(c)) { color: yellow }
EM::region(first-line) { font-weight: normal }

Or without the nested parentheses:

P::region(slot c) { color: yellow }
EM::region(first-line) { font-weight: normal }

The name “region” may not be the best. Some alternatives are:

EM::part(first-line) { font-weight: normal }
EM::inside(first-line) { font-weight: normal }
EM::overlap(first-line) { font-weight: normal }

And [CSS3-REGIONS] proposes an at-rule, ‘@region’:

@region ::slot(a) {
 P {font-size: large}
}

Note that the specificity of the different alternative selectors would not be the same. [CSS3-REGIONS] defines that the selector immediately after ‘@region’ does not add specificity. In other words, ‘@region div::first-line { em {…} }’ has specificity 1, while ‘div::first-line em {...}’ has specificity 2. ‘em::region(div::first-line)’ would have specificity 1 as well.

Because the specificity of the selector after ‘@region’ is not taken into account (see Bug 15734), often only the order of the rules determines which rule is used. With a document like this:

<ul class="menu nav">
 <li>…
 <li>…
</ul>
<ul class="menu">
 <li>…
 <li>…
</ul>

and a style sheet like this:

ul.menu {grid: ...}
@region ul.menu.nav::slot(a) { li { color: green } }
@region ul.menu::slot(a) { li { color: orange } }

the two color rules apply to the same LIs, but the first, seemingly more specific, ‘@region’ rule has no effect. Is this a problem? Authors would have to remember to write the above like this instead:

@region ul.menu::slot(a) { ul.nav li { color: green } } /* spec = 12 */
@region ul.menu::slot(a) { li { color: orange } }       /* spec = 1 */

or

@region ul.menu::slot(a) {
  ul.nav li { color: green }  /* spec = 12 */
  li { color: orange }        /* spec = 1 */
}

or

@region ::slot(a) {
  ul.menu.nav li { color: green }  /* spec = 22 */
  ul.menu li { color: orange }     /* spec = 12 */
}

10.2.14. Orient the grid with writing mode vs orient the slots

A grid can contain vertical text, but the grid itself doesn't need to be rotated or flipped. Indeed, it would be confusing if it did. In other words, the following element has vertical text in the top right slot and an image in the bottom right slot:

<DIV STYLE="grid: "a *" "a b"; writing-mode: vertical-rl">
 Some text here...
 <IMG STYLE="flow: b"...>
 ...
</DIV>

In this specification, a grid template is always laid out with the first row at the top and the first column on the left. But the slots themselves can have a writing mode (by default the one “inherited” from the grid element).

This is different for container elements, i.e., elements with a ‘display’ of ‘grid’ or ‘inline-grid’, as those are designed for graphical user interfaces, where it is sometimes useful to swap a row of buttons, e.g., in a different language. Grids inside other elements (‘block’, ‘list-item’, ‘inline-block’, ‘table-cell’, etc.) are more likely to depend on the orientation of the page (landscape/portrait, recto/verso); but that has to be handled explicitly by the designer by means of media queries and selectors.

10.2.15. Indicating the default slot (‘*’ vs ‘@’)

The symbol ‘*’ seems a good choice, as the concept of default is somewhat similar to the concept of wildcard. The ‘*’ is also used to set the size of rows and columns, but the context is probably enough to avoid confusion.

10.2.16. Flowing vs absolutely positioning content in a grid

The most flexible model for positioning content in a grid is to consider the slot as a flow, which can contain zero, one or more elements and even anonymous content. This is the same concept as a table cell: the cell also is a separate flow that can contain several elements as well as anonymous content.

An element inside a slot thus has its size determined as an element in normal flow. E.g., if you set a background on it, the background covers the element only, not the whole slot. (But the slot has its own background property for that.)

But you may also want to use the grid as a coordinate system for absolutely positioned elements, which then overlap with the contents of the grid (above or below it, depending on ‘z-index’).

And, as “Automatically add rows and columns” above explains, you may want a table-like display of elements that are neither absolutely positioned nor flowed, but the table properties aren't powerful enough.

Those two ideas can be added to the grid template model: the declaration of the grid is reused, but instead of using ‘flow’ to position an element into it, the element is absolutely positioned on top of it. But unlike the absolute positioning in level 2, this positioning is allowed to influence the size of the grid element on which it is positioned.

There are two ideas for how to do this. One idea is to define a special kind of unit, ‘gr’, which is only defined for elements that have a grid ancestor, and which can be used on the ‘top’, ‘right’, ‘bottom’ and ‘left’, properties of absolutely positioned elements. ‘left: 2.25gr’ means the position is in the third column, one quarter of the way towards the fourth column.

Another way is to define a new property ‘grid-area’, which, if set, overrides ‘top’, ‘left’, etc. and which takes numbers or names of slots (up to four of them, to set the four sides).

Yet another way is to extend ‘top’, ‘left’, etc., with unitless values (meaning the n'th grid column or row) and letters (meaning the appropriate edge of the slot of that name).

This specification adds a ‘grid-area’ property, which overrides ‘top’, ‘right’, etc. When it is used, it also indicates that the element takes part in the size calculations of the rows and columns it is put on top of.

10.2.17. Use the pseudo-class to create additional slots

If you want a flow that overlaps with a grid element (i.e., absolutely positioned on top of it), but that flow doesn't correspond to an element in the document you could create a pseudo-element that is like an “absolutely positioned slot:”

DIV::slot(z) {position: grid; grid-area: 2 2}

Like other ‘::slot()’ pseudo-elements, it applies to grid elements only (the DIV in this example must be a grid element), but unlike other ‘::slot()’ pseudo-elements it doesn't style an existing slot, but implicitly creates a new one. (It is some kind of error if the name ‘z’ already exists in the grid template.)

Then you can flow other elements into this slot:

.note {flow: z}

This mechanism is difficult to understand. Experience shows it is also difficult to explain and in a style sheet it is difficult to recognize. What are the use cases (that can't be done with negative margins, absolute positioning, etc.)? This specification does not allow the creation of arbitrary, absolutely positioned pseudo-elements. If needed, it could be added in level 4.

10.2.18. Names for edges

If you position elements absolutely (with ‘top’, ‘left’, etc. or with ‘grid-area’, see “Flowing vs absolutely positioning content in a grid” above), you can use numbers to refer to grid lines: from ‘1’ (the left edge of the first column) to N+1 (the right edge of the last column) and ditto for rows. You can also refer to the edges of slots, for an extra level of indirection and to avoid the need for numbers.

It has been suggested to create a mechanism to assign arbitrary names to grid lines, even multiple aliases for the same grid line. You could then absolutely position an element at grid line ‘foo’ (e.g., ‘grid-area: foo 1’) and later assign that alias ‘foo’ to a different grid line without having to change the style rules that use it. E.g., the following could make the name ‘a-line’ an alias for the number 2 and ‘b-line’ an alias for the number 4:

/* Create a 7x2 grid template */
DIV {grid-template-columns: * a-line * * b-line * *; grid-template-rows: * *}
/* Put H2 at cell x=4, y=2 */
DIV H2 {position: grid; grid-area: b-line 2}

(This could also use strings instead of identifiers, for even more flexible names.)

This turns out to be very difficult to explain to users, partly because of the syntax, partly because of the indirection, which comes on top of the inherent indirection that CSS already requires (the style rules are in a style sheet, not on the elements they apply to) and the indirection of specifying a position indirectly by referring to an abstract grid.

Also, the number of usages of ‘grid-area’ (for a given grid) is not likely to be so big that a symbolic name for a grid line is necessary. Moreover, this is easy to handle with a macro processor.

This specification does not include a mechanism to assign aliases to grid lines.

10.2.19. Position from the right and bottom

If elements are absolutely positioned relative to a grid (with ‘top’, ‘left’ or ‘grid-area’, see “Flowing vs absolutely positioning content in a grid” above), and especially if the grid can automatically grow (see “Automatically add rows and columns”) it might be useful to position elements relative to the last row or column, independent of how many rows or columns the grid element has. That could be done with negative numbers:

grid-area: 1 -1

This specification does not allow positioning from the end. If uses cases are found, it could be added in level 4.

10.2.20. Syntax features to shorten long grid specifications

Grids are typically very regular (see also *’ vs ‘fr above). E.g., if you specify the margins between slots in the template (rather then with a ‘margin’ property on the slots), you often end up with a grid similar to this:

grid-template-areas: "a.b.c.d.e.f";
grid-template-columns: * 1em * 1em * 1em * 1em * 1em *

Even for the columns of broadsheet newspapers (6–8 columns), that still fits easily on one line, even with margins between the columns. But if the grid gets even longer, there may be a need for abbreviations, e.g.:

grid-template-columns: * repeat(1em *, 5)

or

grid-template-columns: * (1em *)[5]

or

grid-template-columns: * 5//1em *//

This makes the syntax harder to learn and read, though. So it needs some good use cases first. This can be postponed to level 4.

10.2.21. fit-content’ vs ‘auto’ vs…

Column sizes can have an exact size or a minimum and maximum size, by means of ‘minmax()’. A common case is probably ‘minmax(min-content, max-content)’. (Is that really true?) Should that common case have its own keyword? If so, should that keyword be ‘fit-content’?

fit-content’ in the Box Module is defined as something else, viz., min(max(available width, ‘min-content’), ‘max-content’)).

An alternative name could be ‘auto’.

10.2.22. Vertical alignment inside slots

The content of a slot can be vertically aligned with ‘vertical-align’, exactly like the content in table cells. (Despite the name, it would, of course, be horizontal alignment if the slot's writing mode was vertical.)

There could also be a property that applies exclusively to grids (grid-slot-align?), just like Flexbox proposes an alignment property that applies exclusively to flex items (‘flex-item-align’). But why should users learn a new property when ‘vertical-align’ fits so well?

Looking beyond tables, flex boxes and grids, there is also a need for vertical alignment of content in normal blocks (when those blocks have a height that does not depend on their contents). There might thus also be a new property that applies uniformly to all cases. E.g.: ‘block-align: top | bottom | middle | baseline | <percentage> | <length>’, where ‘baseline’ simply means ‘top’ everywhere outside grids and tables. (A separate question is if it would override ‘vertical-align’ in tables). Or ‘vertical-align’ itself could be generalized to apply to all block container boxes (although the meaning would be ambiguous on ‘inline-block’).

Given how little progress has been made on this generally applicable property, the present specification just reuses ‘vertical-align’. If a differently named property is developed after the grid template module becomes a REC, then whatever interaction it has with ‘vertical-align’ would apply to slots, too.

10.2.23. Horizontal alignment inside slots

Elements can be centered in, or aligned to a side of, their parent with ‘auto’ margins. But this doesn't always work: for many combinations of ‘width’ and ‘direction’, the auto margins are ignored. There may thus be a new property (‘block-align: center’?) or value (‘margin: any’?) in the future for correct alignment and centering.

Or should there be an alignment property that only applies to grid slots? ‘grid-slot-align’?

This specification defines nothing, in the expectation of a general solution in the future.

10.2.24. Auto margins on slots

Do ‘auto’ margins on slots (e.g., ‘::slot(a) {margin: auto}’) do anything? Or are they simply 0?

10.2.25. Page floats

[CSS3GCPM] defines “page floats” (which are slightly misnamed, because they cause floats to float not just to the top or bottom of a page, but also to the top or bottom of a column in certain contexts). What do they do in slots?

This specification treats a slot similar to a column, e.g., ‘float: top’ means the top of the slot, not of the page. Similarly, if slots are chained, then the float modifiers select the next slot, not the next page.

10.2.26. Break-* properties

The ‘break-before’, ‘break-after’ and ‘break-inside’ properties in the draft of February 2012 (see [CSS3-BREAK]) have keywords ‘region’ and ‘avoid-region’. Should they be called ‘slot’ and ‘avoid-slot’ instead?

10.2.27. Automatic placement of elements

There may be cases where elements are to be placed based on where the previous elements was placed, which isn't always expressible with a clever selector. An example is placing DD elements in the same slot as the preceding DT element. (Note that there may be more than one DD associated with a single DT, so a selector such as ‘DT + DD’ doesn't work. The selector ‘DT ~ DD’ doesn't work either, because it selects DDs that belong to the next DT, too.).

A keyword may help: ‘flow: same’.

In some cases, elements may need to be placed in the next column or row after the previous element. If there are cases where rows or columns need to be added automatically based on content, those are probably also the cases where element are to be placed in the “next” row or column w.r.t. to the preceding element.

E.g., with a keyword: ‘grid-area: next same’ (for “next column, same row”).

10.2.28. The names of the properties

This draft currently has ‘grid’ = ‘grid-template-areas’ + ‘grid-template-rows’ + ‘grid-template-columns’ for creating a grid template; a ‘grid-area’ property for absolute positioning; and ‘flow’ for flowing content into slots.

[CSS-GRID-1] proposes grid-template + grid-definition-row + grid-definition-columns for creating a grid template, but no shorthand; and grid-area = grid-template-row + grid-column, where grid-template-row = grid-before + grid-after and grid-column = grid-start + grid-end. There are also grid-auto-columns and grid-auto-rows properties for the sizes of automatically added columns and rows.

The first set clearly concentrates on making it easy to create templates, with short property names and a shorthand that allows compact, one-line declarations. The expectation is that most content will be positioned with ‘flow’, another short property name, and that the absolute positioning with ‘grid-area’ is rare. ‘grid-area’ can still become a shorthand later, but reserving short names seems unnecessary in this model.

The second set takes the opposite approach. Authors have a large choice of properties to absolutely position an element on top of a grid, with three shorthand properties and four individual properties. In contrast, creating a template is verbose, because there is no shorthand and the property names are long. They have to be long to avoid confusion with the short names used for positioning.

Where this draft has ‘flow’ for flowing content into a region, the Regions draft [CSS3-REGIONS] (version of August 2012) proposes ‘flow-into’. It's more descriptive, but longer.

11. Conformance

Document conventions

Conformance requirements are expressed with a combination of descriptive assertions and RFC 2119 terminology. The key words “MUST”, “MUST NOT”, “REQUIRED”, “SHALL”, “SHALL NOT”, “SHOULD”, “SHOULD NOT”, “RECOMMENDED”, “MAY”, and “OPTIONAL” in the normative parts of this document are to be interpreted as described in RFC 2119. However, for readability, these words do not appear in all uppercase letters in this specification.

All of the text of this specification is normative except sections explicitly marked as non-normative, examples, and notes. [RFC2119]

Examples in this specification are introduced with the words “for example” or are set apart from the normative text with class="example", like this:

This is an example of an informative example.

Informative notes begin with the word “Note” and are set apart from the normative text with class="note", like this:

Note, this is an informative note.

Conformance classes

Conformance to this specification is defined for three conformance classes:

style sheet
A CSS style sheet.
renderer
A UA that interprets the semantics of a style sheet and renders documents that use them.
authoring tool
A UA that writes a style sheet.

A style sheet is conformant to this specification if all of its statements that use syntax defined in this module are valid according to the generic CSS grammar and the individual grammars of each feature defined in this module.

A renderer is conformant to this specification if, in addition to interpreting the style sheet as defined by the appropriate specifications, it supports all the features defined by this specification by parsing them correctly and rendering the document accordingly. However, the inability of a UA to correctly render a document due to limitations of the device does not make the UA non-conformant. (For example, a UA is not required to render color on a monochrome monitor.)

An authoring tool is conformant to this specification if it writes style sheets that are syntactically correct according to the generic CSS grammar and the individual grammars of each feature in this module, and meet all other conformance requirements of style sheets as described in this module.

Partial implementations

So that authors can exploit the forward-compatible parsing rules to assign fallback values, CSS renderers must treat as invalid (and ignore as appropriate) any at-rules, properties, property values, keywords, and other syntactic constructs for which they have no usable level of support. In particular, user agents must not selectively ignore unsupported component values and honor supported values in a single multi-value property declaration: if any value is considered invalid (as unsupported values must be), CSS requires that the entire declaration be ignored.

Experimental implementations

To avoid clashes with future CSS features, the CSS 2.1 specification reserves a prefixed syntax for proprietary and experimental extensions to CSS.

Prior to a specification reaching the Candidate Recommendation stage in the W3C process, all implementations of a CSS feature are considered experimental. The CSS Working Group recommends that implementations use a vendor-prefixed syntax for such features, including those in W3C Working Drafts. This avoids incompatibilities with future changes in the draft.

Non-experimental implementations

Once a specification reaches the Candidate Recommendation stage, non-experimental implementations are possible, and implementors should release an unprefixed implementation of any CR-level feature they can demonstrate to be correctly implemented according to spec.

To establish and maintain the interoperability of CSS across implementations, the CSS Working Group requests that non-experimental CSS renderers submit an implementation report (and, if necessary, the testcases used for that implementation report) to the W3C before releasing an unprefixed implementation of any CSS features. Testcases submitted to W3C are subject to review and correction by the CSS Working Group.

Further information on submitting testcases and implementation reports can be found from on the CSS Working Group's website at http://www.w3.org/Style/CSS/Test/. Questions should be directed to the public-css-testsuite@w3.org mailing list.

12. Changes

Summary of major changes since the 29 November 2011 draft:

Acknowledgments

The first ideas for describing a template in CSS date from 1996 and are described in Frame-based layout via Style Sheets by Bert Bos, Dave Raggett and Håkon Wium Lie. The idea was revived in 2005 on the request of W3C's Device Independence Working Group and benefited especially from discussions with Rhys Lewis and Rotan Hanrahan from that group.

This specification was further influenced by ideas about form layout by Dave Raggett [member-only link] and an early write-up of the features of XUL by Ian Hickson [member-only link].

Andy Clarke, Jina Bolton and Kevin Lawver provided use cases, examples and requirements. The analysis in the History section is a slightly shortened version of work by Steve Zilles.

César Acebal built the first prototype, see [ACEBAL2012]. Andrew Fedoniouk built the second. A third prototype was made by Alexis Deveria. The fourth prototype, ALMcss3, was again made by César Acebal.

References

Normative references:

[CSS-GRID-1]
Tab Atkins Jr.; fantasai; Rossen Atanassov. CSS Grid Layout Module Level 1. 23 January 2014. W3C Working Draft. (Work in progress.) URL: http://www.w3.org/TR/2014/WD-css-grid-1-20140123/
[CSS-TEXT-3]
Elika J. Etemad; Koji Ishii. CSS Text Module Level 3. 10 October 2013. W3C Last Call Working Draft. (Work in progress.) URL: http://www.w3.org/TR/2013/WD-css-text-3-20131010/
[CSS-WRITING-MODES-3]
fantasai; Koji Ishii. CSS Writing Modes Level 3. 26 November 2013. W3C Last Call Working Draft. (Work in progress.) URL: http://www.w3.org/TR/2013/WD-css-writing-modes-3-20131126/
[CSS21]
Bert Bos; et al. Cascading Style Sheets Level 2 Revision 1 (CSS 2.1) Specification. 7 June 2011. W3C Recommendation. URL: http://www.w3.org/TR/2011/REC-CSS2-20110607
[CSS3-BREAK]
Rossen Atanassov; Elika J. Etemad. CSS Fragmentation Module Level 3. 16 January 2014. W3C Working Draft. (Work in progress.) URL: http://www.w3.org/TR/2014/WD-css3-break-20140116/
[CSS3-REGIONS]
Rossen Atanassov; Alan Stearns. CSS Regions Module Level 1. 18 February 2014. W3C Working Draft. (Work in progress.) URL: http://www.w3.org/TR/2014/WD-css3-regions-20140218/
[CSS3BOX]
Bert Bos. CSS basic box model. 9 August 2007. W3C Working Draft. (Work in progress.) URL: http://www.w3.org/TR/2007/WD-css3-box-20070809
[CSS3GENCON]
Ian Hickson. CSS3 Generated and Replaced Content Module. 14 May 2003. W3C Working Draft. (Work in progress.) URL: http://www.w3.org/TR/2003/WD-css3-content-20030514
[CSS3PAGE]
Melinda Grant; et al. CSS Paged Media Module Level 3. 14 March 2013. W3C Working Draft. (Work in progress.) URL: http://www.w3.org/TR/2013/WD-css3-page-20130314/
[CSS3POS]
Bert Bos. CSS3 Positioning Module. (forthcoming). W3C Working Draft. (Work in progress.)
[CSS3SYN]
Tab Atkins Jr.; Simon Sapin. CSS Syntax Module Level 3. 20 February 2014. W3C Candidate Recommendation. (Work in progress.) URL: http://www.w3.org/TR/2014/CR-css-syntax-3-20140220/
[CSS3TBL]
Bert Bos; David Hyatt. CSS3 Tables Module. (forthcoming). W3C Working Draft. (Work in progress.)
[CSS3VAL]
Håkon Wium Lie; Tab Atkins; Elika J. Etemad. CSS Values and Units Module Level 3. 30 July 2013. W3C Candidate Recommendation. (Work in progress.) URL: http://www.w3.org/TR/2013/CR-css3-values-20130730/
[RFC2119]
S. Bradner. Key words for use in RFCs to Indicate Requirement Levels. RFC 2119. URL: http://www.ietf.org/rfc/rfc2119.txt

Other references:

[ACEBAL2010]
César Fernández Acebal. ALMcss: Separación de estructura y presentación en la web mediante posicionamiento avanzado en CSS. 2010. Oviedo, Spain. PhD thesis URL: http://di002.edv.uniovi.es/~acebal/phd/dissertation.pdf
[ACEBAL2012]
César Acebal; et al. “ALMcss: a Javascript implementation of the CSS template layout module” in: Proceedings of the 2012 ACM symposium on Document Engineering. ACM. September 2012. New York, NY, USA. ISBN 978-1-4503-1116-8
[CSS-OVERFLOW-3]
L. David Baron. CSS Overflow Module Level 3. 18 April 2013. W3C First Public Working Draft. (Work in progress.) URL: http://www.w3.org/TR/2013/WD-css-overflow-3-20130418/
[CSS3-EXCLUSIONS]
Vincent Hardy; Rossen Atanassov; Alan Stearns. CSS Exclusions Module Level 1. 28 May 2013. W3C Working Draft. (Work in progress.) URL: http://www.w3.org/TR/2013/WD-css3-exclusions-20130528/
[CSS3-FLEXBOX]
Tab Atkins Jr.; Elika J. Etemad; Alex Mogilevsky. CSS Flexible Box Layout Module. 18 September 2012. W3C Candidate Recommendation. (Work in progress.) URL: http://www.w3.org/TR/2012/CR-css3-flexbox-20120918/
[CSS3BG]
Bert Bos; Elika J. Etemad; Brad Kemper. CSS Backgrounds and Borders Module Level 3. 4 February 2014. W3C Last Call Working Draft. (Work in progress.) URL: http://www.w3.org/TR/2014/WD-css3-background-20140204/
[CSS3COL]
Håkon Wium Lie. CSS Multi-column Layout Module. 12 April 2011. W3C Candidate Recommendation. (Work in progress.) URL: http://www.w3.org/TR/2011/CR-css3-multicol-20110412
[CSS3COLOR]
Tantek Çelik; Chris Lilley; L. David Baron. CSS Color Module Level 3. 7 June 2011. W3C Recommendation. URL: http://www.w3.org/TR/2011/REC-css3-color-20110607
[CSS3GCPM]
Håkon Wium Lie. CSS Generated Content for Paged Media Module. 29 November 2011. W3C Working Draft. (Work in progress.) URL: http://www.w3.org/TR/2011/WD-css3-gcpm-20111129/
[CSS3GRID]
Alex Mogilevsky; Markus Mielke. CSS Grid Positioning Module Level 3. 5 September 2007. W3C Working Draft. (Work in progress.) URL: http://www.w3.org/TR/2007/WD-css3-grid-20070905
[CSS3UI]
Tantek Çelik. CSS Basic User Interface Module Level 3 (CSS3 UI). 17 January 2012. W3C Working Draft. (Work in progress.) URL: http://www.w3.org/TR/2012/WD-css3-ui-20120117/
[MEDIAQ]
Florian Rivoal. Media Queries. 19 June 2012. W3C Recommendation. URL: http://www.w3.org/TR/2012/REC-css3-mediaqueries-20120619/
[SELECT]
Tantek Çelik; et al. Selectors Level 3. 29 September 2011. W3C Recommendation. URL: http://www.w3.org/TR/2011/REC-css3-selectors-20110929/
[WCAG20]
Ben Caldwell; et al. Web Content Accessibility Guidelines (WCAG) 2.0. 11 December 2008. W3C Recommendation. URL: http://www.w3.org/TR/2008/REC-WCAG20-20081211/

Index

Property index

Property Values Initial Applies to Inh. Percentages Media
chains none | <identifier>+ [ , <identifier>+ ]* none grid elements no N/A visual
flow auto | <identifier> | <string> | ‘*’ | same auto elements that have a grid ancestor and whose ‘position’ is ‘static’ or ‘relative’ no N/A visual
grid <‘grid-template’> | <‘grid-auto-flow’> [ <‘grid-auto-columns’> [ / <‘grid-auto-rows’> ]? ] See individual properties block container elements and grid container elements no see individual properties visual
grid-template none | <col-width>* [ [ <string> <row-height>? ]+ | ‘/’ <row-height>+ ] none block container elements and grid container elements no see individual properties visual
grid-template-areas none | <string>+ none block container elements [CSS21] and grid container elements [CSS-GRID-1] no N/A visual
grid-template-columns auto | <col-width>+ auto block container elements and grid container elements no N/A visual
grid-template-rows auto | <row-height>+ auto block container elements and grid container elements no N/A visual