Chapter 3: Rendering Model

SVG 2 Requirement: Support the z-index.

Resolution: We will add Jonathan Watt's z-index proposal to SVG 2.

Auckland 2011 F2F day 5.

Purpose: Allow reordering (such as when a planet orbits the sun). Reordering without script support (e.g. CSS :hover).

Owner: Jonathan (Action 3002).

Status: Done.

The SVG 2 rendering model will follow the rules defined by the Compositing and Blending specification.

Resolution: Seattle/Paris 2012 F2F day 3.

Owner: Nikos (Action 3332).

Status: Done.

3.1. Introduction

Implementations of SVG must implement the rendering model as described in this chapter, as modified in the appendix on conformance requirementswhich describes situations where an implementation may deviate. In practice variability is allowed based on limitations of the output device (e.g. only a limited range of colors might be supported) and because of practical limitations in implementing a precise mathematical model (e.g. for realistic performance curves are approximated by straight lines, the approximation need only be sufficiently precise to match the conformance requirements).

The appendix on conformance requirements describes the extent to which an actual implementation may deviate from this description. In practice an actual implementation may deviate slightly because of limitations of the output device (e.g. only a limited range of colors might be supported) and because of practical limitations in implementing a precise mathematical model (e.g. for realistic performance curves are approximated by straight lines, the approximation need only be sufficiently precise to match the conformance requirements).

3.2. The rendering tree

The components of the final rendered representation of an SVG document do not have a one-to-one relationship with the underlying elements in the document model. The appearance of the graphic instead reflects a parallel structure, the rendering tree, in which some elements are excluded and others are repeated.

Many elements in the SVG namespace do not directly represent a component of the graphical document. Instead, they define graphical effects, metadata, content to be used to represent other elements, or alternatives to be displayed under certain conditions. In dynamic documents, certain components of the graphic may be rendered or not, depending on interaction or animation. These non-rendered elements are not directly included in the rendering tree.

Because SVG supports the reuse of graphical sub-components, some elements are rendered multiple times. The individual renderings may have context-dependent styling and may be rasterized at different scales or transformations.

3.2.1. Definitions

rendering tree

The rendering tree is the set of elements being rendered in an SVG document fragment. It is generated from the document tree by excluding non-rendered elements and inserting additional fragments for re-used graphics. Graphics are painted and composited in rendering-tree order, subject to stacking and re-ordering based on the z-index and paint-order properties. Note that elements that have no visual paint may still be in the rendering tree.

rendered element

An element that has a direct representation in the rendering tree for the current document. Includes a rendered instance of an element in a use-element shadow tree. Does not include elements that affect rendering as the source definition of re-used graphics but are not directly rendered themselves. See Rendered versus non-rendered elements

non-rendered element

An element that does not have a direct representation in the rendering tree for the current document. It may nonetheless affect the rendering tree as re-used graphics or graphical effects. See Rendered versus non-rendered elements.

re-used graphics

Graphical components that are included in the rendering tree but do not have a single direct equivalent element in the document model. They may be represented through shadow DOM elements (as in graphics re-used with a use element), or as image fragments generated as part of a graphical effect (as in patterns or masks).

never-rendered element

Any element type that is never directly rendered, regardless of context or display style value. It includes the following elements: clipPath, defs, desc, hatch, linearGradient, marker, mask, meshgradient, metadata, pattern, radialGradient, script, style and title; it also includes a symbol element that is not the instance root of a use-element shadow tree.

renderable element

Any element type that can have a direct representation in the rendering tree, as a graphic, container, text, audio, or animation. It includes the following elements: a, audio, canvas, circle, ellipse, foreignObject, g, iframe, image, line, mesh, path, polygon, polyline, rect, svg, switch, text, textPath, tspan, unknown, use and video; it also includes a symbol element that is the instance root of a use-element shadow tree.

A renderable element may or may not be rendered in a given document or point in time.

3.2.2. Rendered versus non-rendered elements

At any given time, every SVG element (or element instance in a use-element shadow tree) is either rendered or non-rendered. Whether an element is currently rendered or not affects not only its visual display but also interactivity and geometric calculations.

An element is not rendered in any of these four situations:

Non-rendered elements:

Non-rendered elements are not represented in the document accessibility tree. Nonetheless, they remain part of the document model, and participate in style inheritance and cascade.

3.2.3. Controlling visibility: the effect of the ‘display’ and ‘visibility’ properties

SVG uses two properties to toggle the visible display of elements that are normally rendered: display and visibility. Although they have a similar visible effect in static documents, they are conceptually distinct.

See the CSS 2.1 specification for the definitions of display and visibility. [CSS2]

Setting display to none results in the element not being rendered. When applied to graphics elements, text content elements, and container elements that are normally rendered, setting display to none results in the element (and all its descendents) not becoming part of the rendering tree. Note that display is not an inherited property.

Elements that have any other display value than none are rendered as normal.

The display property only applies to renderable elements. Setting display: none on an element that is never directly rendered or not rendered based on conditional processing has no effect.

The display property affects the direct processing of a given element, but it does not prevent it from being referenced by other elements. For example, setting display: none on a path element will prevent that element from getting rendered directly onto the canvas, but the path element can still be referenced by a textPath element and its geometry will be used in text-on-a-path processing.

When applied to a graphics element or use element, setting visibility to hidden or collapse results in the element not being painted. It is, however, still part of the rendering tree. It may be sensitive to pointer events (depending on the value of pointer-events), may receive focus (depending on the value of tabindex), contributes to bounding box calculations and clipping paths, and does affect text layout.

The visibility property only directly applies to graphics elements, text content elements (including the a element when it is a child of text content element) and use elements. Since visibility is an inherited property, however, although it has no effect on a use element or container element itself, its inherited value can affect descendant elements.

3.2.4. Re-used graphics

Graphical content defined in one part of the document (or in another document) may be used to render other elements. There are two types of re-used graphics from a rendering perspective:

Shadow DOM elements are rendered in the same way as normal elements, as if the host element (e.g., the use element) was a container and the shadow content was its descendents. Style and geometry properties on the shadow DOM elements are resolved independently from those on their corresponding element in the source document. The display property has its normal effect on shadow elements, except for special rules that apply to the symbol element.

Unless properties on the host element create a stacking context, z-index values on shadow DOM elements can change their rendering order relative to elements in the main DOM. For blending purposes, the use element forms a non-isolated group.

In contrast, graphical effects elements generate a self-contained SVG fragment which is rendered independently as a stacking context and an isolated group. The canvas for this fragment is scaled The graphical effect element's child content is rendered and composited into this canvas. The flattened canvas as a whole is treated as a vector image when compositing and blending with other paint layers

The display property on any child content of a graphical effects element has its normal effect when set to none, excluding that subtree from being used in rendering. However, the graphical effect is not altered by a value of display: none on the graphical effect element or an ancestor.

3.3. The painters model

SVG uses a "painters model" of rendering. Paint is applied in successive operations to the output device such that each operation paints onto some area of the output device, possibly obscuring paint that has previously been layed down. After each object or group is painted, it becomes part of the background for the next painting operation. SVG 2 supports advanced blending modes and compositing operations that control how each painting operation interacts with the background. The rules governing these painting operations are outlined in the Compositing and Blending Specification.

3.4. Rendering order

Elements in SVG are positioned in three dimensions. In addition to their position on the x and y axis of the SVG viewport, SVG elements are also positioned on the z axis. The position on the z-axis defines the order that they are painted.

Along the z axis, elements are grouped into stacking contexts, each stacking context has an associated stack level. A stack level may contain one or more child nodes - either child stack levels, graphics elements, or g elements. graphics elements and g elements within single stack level are painted in document order - that is, they are painted in the order that they are defined in the document.

Each stack level is assigned an integer value that defines it's position on the z axis relative to other stack levels within the same stacking context. Lower values are painted first, and so elements in a stack level with a higher value will paint over one with a lower value.

By default, everything is placed in stack level zero.

3.4.1. Controlling element rendering order: the ‘z-index’ property

This feature is at risk.

See the CSS 2.1 specification for the definition of z-index. [CSS2]

The z-index property allows an element to be assigned to a stack level.

The rules governing behavior for SVG elements with the z-index property specified are outlined below:

CSS specifies a property named z-index. The CSS rules that define the effect of the ‘z-index’ property were written specifically for the CSS box model, and those rules do not make sense as they stand for most SVG elements (most SVG elements do not participate in or establish a CSS box model layout). This section specifies how implementations must handle the z-index property on elements in the SVG namespace.

Contrary to the rules in CSS 2.1, the z-index property applies to all SVG elements regardless of the value of the position property, with one exception: as for boxes in CSS 2.1, outer svg elements must be positioned for z-index to apply to them.

The z-index property specifies:

  1. The stack level of the element in the current stacking context.
  2. Whether the element establishes a new local stacking context.

Values have the following meanings:

<integer>
This integer is the stack level of the element in the current stacking context. The element also establishes a new local stacking context for its descendants.
auto
The stack level of the element in the current stacking context is the same as its parent element, unless its parent established a new stacking context, in which case its stack level is 0. The element does not establish a new local stacking context.

Here is a simple example:

<svg xmlns="http://www.w3.org/2000/svg">
  <rect x="0"  width="100" height="100" style="fill: red;    z-index: -1;"/>
  <rect x="40" width="100" height="100" style="fill: lime;"/>
  <rect x="80" width="100" height="100" style="fill: blue;   z-index: 1;"/>
  <rect x="60" width="100" height="100" style="fill: aqua;"/>
  <rect x="20" width="100" height="100" style="fill: yellow; z-index: -1;"/>
</svg>

In this example there are three stack levels: -1, 0 (the default) and 1. The red and yellow rects are in stack level -1, the lime and aqua rects are in stack level 0 (the default), and the blue rect is in stack level 1. Going from lowest stack level to highest, and painting the elements in each stack level in document order, the painting order is: red, yellow, lime, aqua, blue.

3.4.2. Establishing a stacking context in SVG

A new stacking context must be established at an SVG element for its descendants if:

Stacking contexts and stack levels are conceptual tools used to describe the order in which elements must be painted one on top of the other when the document is rendered, and for determining which element is highest when determining the target of a pointer event. Stacking contexts and stack levels do not affect the position of elements in the DOM tree, and their presence or absence does not affect an element's position, size or orientation in the canvas' X-Y plane - only the order in which it is painted.

Stacking contexts can contain further stacking contexts. A stacking context is atomic from the point of view of its parent stacking context; elements in ancestor stacking contexts may not come between any of its elements.

Each element belongs to one stacking context. Each element in a given stacking context has an integer stack level. Elements with a higher stack level must be placed in front of elements with lower stack levels in the same stacking context. Elements may have negative stack levels. Elements with the same stack level in a stacking context must be stacked according to document order.

With the exception of the foreignObject element, the back to front stacking order for a stacking context created by an SVG element is:

  1. the background and borders of the element forming the stacking context, if any
  2. child stacking contexts created by descendants with negative stack levels, primarily ordered by most negative first, then by tree order
  3. descendants with 'z-index: auto' or 'z-index: 0', in tree order
  4. child stacking contexts created by descendants with positive stack levels, primarily ordered by lowest index first, then by tree order

Since the foreignObject element creates a "fixed position containing block" in CSS terms, the normative rules for the stacking order of the stacking context created by foreignObject elements are the rules in Appendix E of CSS 2.1.

In the following example, the z-index property on the g element is set to zero. This creates a new stacking context to contain the g element's children without moving the g to a different level in the document's root stacking context:

<svg xmlns="http://www.w3.org/2000/svg">
  <g style="z-index: 0;">
    <!-- this is a self contained graphic -->
    <rect x="40" width="100" height="100" style="fill: lime; z-index: 1;"/>
    <rect x="20" width="100" height="100" style="fill: yellow;"/>
  </g>
  <rect x="60" width="100" height="100" style="fill: aqua;"/>
  <rect x="0" width="100" height="100" style="fill: red; z-index: -1;"/>
</svg>

The example's root stacking context contains two stack levels: -1 and 0. The red rect is in level -1, and the g element and aqua rect are in level 0. Inside stack level 0, the g element's z-index property creates a new nested stacking context at the g for the g element's children. In this child stacking context there are two stack levels: 0 and 1. The yellow rect is in level 0 (the default), and the lime rect is in level 1.

Painting of this example starts with the stack levels of the root stacking context. First the red rect is painted in level -1, then in level 0 the g element is painted followed by the aqua rect. When the g element is painted, the child stacking context that its z-index created and all of that context's stack levels are also painted. In this child stacking context, first the yellow rect in level 0 is painted, followed by the lime rect in level 1. It's only after the g and the stacking context that it creates has been painted that the aqua rect is painted. Note that this means that although the z-index of 1 for the lime rect is a higher value than the (implicit) z-index of 0 for the aqua rect, the containment provided by the g's child stacking context results in the aqua rect painting over the lime rect. The painting order is therefore: red, yellow, lime, aqua.

3.5. How elements are rendered

Individual graphics elements are treated as if they are a non-isolated group, the components (fill, stroke, etc) that make up a graphic element (See Painting shapes and text) being members of that group. See How groups are rendered.

3.6. How groups are rendered

Grouping elements, such as the g element (see container elements ) create a compositing group. Similarly, a use element creates a compositing group for its shadow content. The Compositing and Blending specification normatively describes how to render compositing groups. In SVG, effects may be applied to a group. For example, opacity, filters or masking. These effects are applied to the rendered result of the group immediately before any transforms on the group are applied, which are applied immediately before the group is blended and composited with the group backdrop. Applying any such effects to a group makes that group isolated.

Thus, rendering a compositing group follows the following steps:
If the group is isolated:

  1. The initial backdrop is set to a new buffer initialised with rgba(0,0,0,0)
  2. The contents of the group that are graphics elements or g elements are rendered in order, onto the initial backdrop
  3. filters and other effects that modify the group canvas are applied

    To provide for high quality rendering, filter primitives and other bitmap effects must be applied in the operating coordinate space.

  4. Group transforms are applied
  5. The group canvas is blended and composited with the group backdrop
else (the group is not isolated):
  1. The initial backdrop is set to the group backdrop
  2. The contents of the group that are graphics elements or g elements are rendered in order, onto the initial backdrop. The group transforms are applied to each element as they are rendered.

3.6.1. Object and group opacity: the effect of the ‘opacity’ property

See the CSS Color Module Level 3 for the definition of opacity. [css3-color]

The opacity property specifies how opaque a given graphical element or container element will be when it is painted to the canvas. When applied to a container element, this is known as group opacity, and when applied to an individual rendering element, it is known as object opacity. The principle for these two operations however is the same.

There are several other opacity-related properties in SVG:

These four opacity properties are involved in intermediate rendering operations. Object and group opacity however can be thought of as a post-processing operation. Conceptually, the object or group to which opacity applies is rendered into an RGBA offscreen image. The offscreen image as whole is then blended into the canvas with the specified opacity value used uniformly across the offscreen image. Thus, the presence of opacity causes the group to be isolated.

The opacity property applies to the following SVG elements: svg, g, symbol, marker, a, switch, use, unknown and graphics elements.

The following example illustrates various usage of the opacity property on objects and groups.

<svg xmlns="http://www.w3.org/2000/svg"
     width="600" height="175" viewBox="0 0 1200 350">

  <!-- Background blue rectangle -->
  <rect x="100" y="100" width="1000" height="150" fill="blue"/>

  <!-- Red circles going from opaque to nearly transparent -->
  <circle cx="200" cy="100" r="50" fill="red" opacity="1"/>
  <circle cx="400" cy="100" r="50" fill="red" opacity=".8"/>
  <circle cx="600" cy="100" r="50" fill="red" opacity=".6"/>
  <circle cx="800" cy="100" r="50" fill="red" opacity=".4"/>
  <circle cx="1000" cy="100" r="50" fill="red" opacity=".2"/>

  <!-- Opaque group, opaque circles -->
  <g opacity="1">
    <circle cx="182.5" cy="250" r="50" fill="red" opacity="1"/>
    <circle cx="217.5" cy="250" r="50" fill="green" opacity="1"/>
  </g>
  <!-- Group opacity: .5, opacity circles -->
  <g opacity=".5">
    <circle cx="382.5" cy="250" r="50" fill="red" opacity="1"/>
    <circle cx="417.5" cy="250" r="50" fill="green" opacity="1"/>
  </g>
  <!-- Opaque group, semi-transparent green over red -->
  <g opacity="1">
    <circle cx="582.5" cy="250" r="50" fill="red" opacity=".5"/>
    <circle cx="617.5" cy="250" r="50" fill="green" opacity=".5"/>
  </g>
  <!-- Opaque group, semi-transparent red over green -->
  <g opacity="1">
    <circle cx="817.5" cy="250" r="50" fill="green" opacity=".5"/>
    <circle cx="782.5" cy="250" r="50" fill="red" opacity=".5"/>
  </g>
  <!-- Group opacity .5, semi-transparent green over red -->
  <g opacity=".5">
    <circle cx="982.5" cy="250" r="50" fill="red" opacity=".5"/>
    <circle cx="1017.5" cy="250" r="50" fill="green" opacity=".5"/>
  </g>
</svg>
Image showing different groups of circles blended into the background.

Each group of red and green circles is first rendered to an offscreen image before being blended with the background blue rectangle as a whole, with the given opacity values.

In the example, the top row of circles have differing opacities, ranging from 1.0 to 0.2. The bottom row illustrates five g elements, each of which contains overlapping red and green circles, as follows:

3.7. Types of graphics elements

SVG supports three fundamental types of graphics elements that can be rendered onto the canvas:

3.7.1. Painting shapes and text

Shapes and text can be filled (i.e., apply paint to the interior of the shape) and stroked (i.e., apply paint along the outline of the shape).

For certain types of shapes, marker symbols (which themselves can consist of any combination of shapes, text and images) can be drawn at positions along the shape boundary. Each marker symbol is painted as if its graphical content were expanded into the SVG document tree just after the shape object which is using the given marker symbol. The graphical contents of a marker symbol are rendered using the same methods as graphics elements. Marker symbols are not applicable to text.

The order in which fill, stroke and markers are painted is determined by the paint-order property. The default is that fill is painted first, then the stroke, and then the marker symbols. The marker symbols are rendered in order along the outline of the shape, from the start of the shape to the end of the shape.

The fill and stroke operations are entirely independent; for instance, each fill or stroke operation has its own opacity setting.

SVG supports numerous built-in types of paint which can be used in fill and stroke operations. These are described in Paint Servers.

3.7.2. Painting raster images

When a raster image is rendered, the original samples are "resampled" using standard algorithms to produce samples at the positions required on the output device. Resampling requirements are discussed under conformance requirements.

As in HTML [HTML, 10.4.2], all animated images with the same absolute URL and the same image data are expected to be rendered synchronised to the same timeline as a group, with the timeline starting at the time of the least recent addition to the group.

When a user agent is to restart the animation for an img element showing an animated image, all animated images with the same absolute URL and the same image data in that img element's node document are expected to restart their animation from the beginning.

3.8. Filtering painted regions

SVG allows any painting operation to be filtered. (See Filter Effects.)

In this case the result must be as though the paint operations had been applied to an intermediate canvas initialized to transparent black, of a size determined by the rules given in Filter Effects then filtered by the processes defined in Filter Effects.

3.9. Clipping and masking

SVG supports the following clipping/masking features:

Both, clipping and masking, are specified in the module CSS Masking [css-masking-1].

3.10. Parent compositing

SVG document fragments can be semi-opaque.

In accordance with the Compositing and Blending specification, the svg element always creates an isolated group. When an SVG document is a top-level document, meaning it is not embedded in another document, the root svg element is considered to be the page group and is composited with a backdrop of white with 100% opacity. In all other cases, the SVG document or document fragment is composited into the parent document with opacity preserved.

3.11. The effect of the ‘overflow’ property

See the Cascading Style Sheets Level 2 Revision 1 (CSS 2.1) Specification [CSS2] for the definition of overflow.

A summary of the behavior of the overflow property in SVG.
element initial ua stylesheet auto visible hidden scroll
document root svgvisiblen/avisible | scrollvisiblehiddenscroll
other svgvisiblehiddenvisible | scrollvisiblehiddenscroll
meshvisiblen/avisiblevisiblehiddenhidden
textvisiblehiddenvisiblevisiblehiddenhidden
patternvisiblehiddenvisiblevisiblehiddenhidden
hatchesvisiblehiddenvisiblevisiblehiddenhidden
markervisiblehiddenvisiblevisiblehiddenhidden
symbolvisiblehiddenvisiblevisiblehiddenhidden
imagevisiblehiddenvisiblevisiblehiddenhidden
iframevisiblehiddenvisible | scrollvisiblehiddenscroll
foreignObjectvisiblehiddenvisible | scrollvisiblehiddenscroll

The overflow property has the same parameter values and has the same meaning as defined in CSS 2.1 ([CSS2], section 11.1.1); however, the following additional points apply:

Although the initial value for overflow is auto. In the User Agent style sheet, overflow is overriden for the svg element when it is not the root element of a stand-alone document, the pattern element, hatch element, and the marker element to be hidden by default.