index.bs

<pre class='metadata'>
Title: Paint Timing
Group: webperf
Shortname: paint-timing
Level: none
ED: https://w3c.github.io/paint-timing/
TR: https://www.w3.org/TR/paint-timing/
Status: ED
Editor: Ian Clelland, Google https://google.com, iclelland@chromium.org, w3cid 76841
        Noam Rosenthal, Invited Expert, noam@webkit.org, w3cid 121539
Former Editor: Shubhie Panicker, Google https://google.com, panicker@google.com, w3cid 92587
               Nicolás Peña Moreno, Google https://google.com, npm@chromium.org, w3cid 103755
Repository: w3c/paint-timing
Abstract: This document defines an API that can be used to capture a series of key moments (first paint, first contentful paint) during pageload which developers care about.
Default Highlight: js
</pre>
<pre class=link-defaults>
spec:dom; type:dfn; text:descendant
spec:dom; type:dfn; text:element
spec:css22; type:dfn; text:visibility
</pre>
<pre class=anchors>
urlPrefix: https://html.spec.whatwg.org/multipage/images.html
    type: dfn; text: available; url: #img-available;
    type: dfn; text: image; url: #images;
urlPrefix: https://www.w3.org/TR/SVG2/render.html; spec: CR-SVG2
    type: dfn; url: #Rendered-vs-NonRendered; text: svg element with rendered descendants;
urlPrefix: https://www.w3.org/TR/css-backgrounds-3/; spec: CSS-BACKGROUNDS-3;
    type: dfn; text: background-image; url: #propdef-background-image;
    type: dfn; text: background-size; url: #background-size;
urlPrefix: https://html.spec.whatwg.org/multipage/canvas.html; spec: HTML;
    type: dfn; text: context mode; url: #concept-canvas-context-mode;
urlPrefix: https://html.spec.whatwg.org/multipage/images.html; spec: HTML;
    type: dfn; text: completely available; url: #img-all;
urlPrefix: https://html.spec.whatwg.org/multipage/rendering.html; spec: HTML;
    type: dfn; text: being rendered; url: #being-rendered;
urlPrefix: https://w3c.github.io/IntersectionObserver/
    type: dfn; text: Intersection rect algorithm; url: #calculate-intersection-rect-algo
urlPrefix: https://html.spec.whatwg.org/multipage/dom.html
    type: dfn; text: represents; url: #represents;
urlPrefix: https://drafts.csswg.org/css-pseudo-4
    type: dfn; text: generated content pseudo-element; url: #generated-content;
    type: dfn; text: typographical pseudo-element; url: #typographic-pseudos;
urlPrefix: https://drafts.csswg.org/css2/zindex.html; spec: CSS;
    type: dfn; url:#painting-order; text: painting order;
urlPrefix: https://www.w3.org/TR/cssom-view
    type: dfn; text: scrolling area; url: #scrolling-area;
urlPrefix: https://www.w3.org/TR/css3-values/
    type: dfn; text: url valued; url: #url-value;
urlPrefix: https://drafts.fxtf.org/css-masking-1/
    type: dfn; text: clip-path; url: #the-clip-path;
urlPrefix: https://www.w3.org/TR/css-images-3/
    type: dfn; text: CSS image; url: #typedef-image;
urlPrefix: https://html.spec.whatwg.org/multipage/media.html
    type: dfn; text: poster frame; url: #poster-frame;
urlPrefix: https://html.spec.whatwg.org/multipage/browsers.html
    type: dfn; text: browsing context; url: #browsing-context;
    type: dfn; text: nested browsing context; url: #nested-browsing-context;
urlPrefix: https://fetch.spec.whatwg.org/; spec: FETCH;
    type: dfn; url:#concept-tao-check; text: timing allow check;
urlPrefix: https://wicg.github.io/largest-contentful-paint/; spec: ELEMENT-TIMING
    type: dfn; url:#report-largest-contentful-paint; text: Report largest contentful paint
urlPrefix: https://wicg.github.io/element-timing/; spec: ELEMENT-TIMING
    type: dfn; url:#report-element-timing; text: Report element timing
</pre>

Introduction {#intro}
=====================

<div class=non-normative>

<em>This section is non-normative.</em>

Much of the purpose of a web browser is to translate HTML, CSS and image
resources into pixels on a screen for users. Measuring the performance of a web
page often involves measuring the time it takes to perform these tasks - to
render content, whether text or image, to the screen. There are many different
ways to use this timing to make statemements about the performance of a page,
or about the user experience of loading, but fundamentally all of those ways
begin with a common means of measuring time.

This is a foundational document which specifies how to measure paint timing as a
general-purpose mechanism. That foundation is then used to define the First
Paint and First Contentful Paint metrics. Other specific instances of paint
measurement may be specified in other documents.

Specifically, this specification covers:
* Measuring the time when images are decoded and ready for painting
* Measuring the time when elements are painted
* Measuring the size of the painted elements
* Determining whether a painted element contains any visible content.

First Paint and First Contentful Paint {#first-paint-and-first-contentful-paint}
--------------------------------------------------------------------------------

Load is not a single moment in time — it's an experience that no one metric can fully capture. There are multiple moments during the load experience that can affect whether a user perceives it as "fast" or "slow".

First paint (FP) is the first of these key moments, followed by first contentful paint (FCP). These metrics mark the points in time when the browser renders a given document. This is important to the user because it answers the question: is it happening?

The primary difference between the two metrics is FP marks the first time the browser renders anything for a given document. By contrast, FCP marks the time when the browser renders the first bit of image or text content from the DOM.

Usage example {#example}
------------------------

<pre class="example highlight">
    const observer = new PerformanceObserver(function(list) {
        const perfEntries = list.getEntries();
        for (const perfEntry of perfEntries) {
            // Process entries
            // report back for analytics and monitoring
            // ...
        }
    });

    // register observer for paint timing notifications
    observer.observe({entryTypes: ["paint"]});
</pre>

</div>

Terminology {#sec-terminology}
==============================

<dfn export>Paint</dfn>: the user agent has performed a "paint" (or "render") when it has converted the render tree to pixels on the screen.
Formally, we consider the user agent to have "rendered" a document when it has performed the [=update the rendering=] steps of the event loop.

    NOTE: The rendering pipeline is very complex, and the timestamp should be the latest timestamp the user agent is able to note in this pipeline (best effort). Typically the time at which the frame is submitted to the OS for display is recommended for this API.

A [=generated content pseudo-element=] is a <dfn>paintable pseudo-element</dfn> when all of the following apply:
* The pseudo-element's [=used value|used=] [=visibility=] is <code>visible</code>.
* The pseudo-element's [=used value|used=] [=opacity=] is greater than zero.
* The pseudo-element generates a non-empty [=box=].

A [=CSS image=] |img| is a <dfn>contentful image</dfn> when all of the following apply:
* |img| is [=url valued=].
* |img| is [=available=].

A {{DOMString}} is <dfn>non-empty</dfn> if it contains at least one character excluding [=document white space characters=].

An [=/element=] |target| is <dfn export>contentful</dfn> when one or more of the following apply:
* |target| has a [=text node=] child, representing [=non-empty=] text, and the node's [=used value|used=] [=opacity=] is greater than zero.

    NOTE: this covers the case where a [=typographical pseudo-element=] would override the opacity of the text node.

* |target| is a [=replaced element=] representing an [=available=] [=image=].
* |target| has a [=background-image=] which is a [=contentful image=], and its [=used value|used=] [=background-size=] has non-zero width and height values.
* |target| is a <{canvas}> with its [=context mode=] set to any value other than <code>none</code>.
* |target| is a <{video}> element that [=represents=] its [=poster frame=] or the first video frame and the frame is available.
* |target| is an [=svg element with rendered descendants=].
* |target| is an <{input}> element with a [=non-empty=] <{input/value}> attribute.
* |target| is an [=originating element=] for a [=paintable pseudo-element=] that represents a [=contentful image=] or [=non-empty=] text.

An [=/element=] is <dfn export>timing-eligible</dfn> if it is one of the following:

* an <{img}> element.
* an <{image}> element inside an <{svg}> element.
* a <{video}> element with a [=poster frame=].
* an element with a [=contentful image|contentful=] <a>background-image</a>.
* a text node.

To compute the <dfn>paintable bounding rect</dfn> of [=/element=] |target|, run the following steps:
    1. Let |boundingRect| be the result of running the {{Element/getBoundingClientRect()}} on |target|.
    1. Clip |boundingRect| with the [=document=]'s [=scrolling area=].
    1. Return |boundingRect|.

    NOTE: elements contained by boxes with <code>overflow: scroll</code> or <code>overflow: hidden</code> don't have their [=paintable bounding rect=] clipped, as in both cases the [=/element=] can become visible by scrolling.

An [=/element=] |el| is <dfn>paintable</dfn> when all of the following apply:
* |el| is [=being rendered=].
* |el|'s [=used value|used=] [=visibility=] is <code>visible</code>.
* |el| and all of its ancestors' [=used value|used=] [=opacity=] is greater than zero.

    NOTE: there could be cases where a <code>paintable</code> [=/element=] would not be visible to the user, for example in the case of text that has the same color as its background.
    Those elements would still considered as paintable for the purpose of computing [=first contentful paint=].

* |el|'s [=paintable bounding rect=] intersects with the [=scrolling area=] of the [=document=].

    NOTE: This covers the cases where the element is scaled to zero size, has <code>display: none</code>, or <code>display: contents</code> where the contents resolve to an empty rect.

    NOTE: As a general rule, an [=/element=] is paintable if it is within the viewport, or can potentially be in the viewport as a result of scrolling or zooming.

<dfn export>First paint</dfn> entry contains a {{DOMHighResTimeStamp}} reporting the time when the user agent first rendered after navigation. This excludes the default background paint, but includes non-default background paint and the enclosing box of an iframe. This is the first key moment developers care about in page load – when the user agent has started to render the page.

A [=browsing context=] |ctx| is <dfn>paint-timing eligible</dfn> when one of the following apply:
* |ctx| is a [=top-level browsing context=].
* |ctx| is a [=nested browsing context=], and the user agent has configured |ctx| to report paint timing.

    NOTE: this allows user agents to enable paint-timing only for some of the frames, in addition to the main frame, if they so choose.
    For example, a user agent may decide to disable paint-timing for cross-origin iframes, as in some scenarios their paint-timing might reveal information about the main frame.

The {{PerformancePaintTiming}} interface {#sec-PerformancePaintTiming}
=======================================

<pre class="idl">
    [Exposed=Window]
    interface PerformancePaintTiming : PerformanceEntry {};
</pre>

{{PerformancePaintTiming}} extends the following attributes of {{PerformanceEntry}} interface:

* The {{PerformanceEntry/name}} attribute's getter must return a {{DOMString}} for minimal frame attribution. Possible values of name are:
    * <code>"first-paint"</code>: for [=first paint=]
    * <code>"first-contentful-paint"</code>: for [=first contentful paint=]
* The {{PerformanceEntry/entryType}} attribute's getter must return <code>"paint"</code>.
* The {{PerformanceEntry/startTime}} attribute's getter must return a {{DOMHighResTimeStamp}} of when the paint occured.
* The {{PerformanceEntry/duration}} attribute's getter must return 0.

NOTE: A user agent implementing {{PerformancePaintTiming}} would need to include <code>"paint"</code> in {{PerformanceObserver/supportedEntryTypes}} of a [=realm/global object=] whose [=Window/browsing context=] is [=paint-timing eligible=].
This allows developers to detect support for paint timing for a particular [=browsing context=].

Processing model {#sec-processing-model}
========================================

Associated Image Requests {#sec-associated-image-requests}
----------------------------------------------------------

Each {{Element}} has an <dfn>associated image request</dfn> which is an [=image
request=] or null, initially null.

When the processing model for an {{Element}} <em>element</em> of type
{{HTMLImageElement}}, {{SVGImageElement}}, or {{HTMLVideoElement}} creates a
new image resource (e.g., to be displayed as an image or poster image),
<em>element</em>'s <a>associated image request</a> is set to the <a>image
request</a> of the created image resource.

Note: Every image resource that is obtained from a URL whose
<a spec=url>scheme</a> is equal to "data" has an associated <a>image request</a>
which is not fetched but still needs to be loaded. This request can be the
<a>associated image request</a> of an {{Element}}.

Note: The current language is vague since it does not point to specific
algorithms. This can be made more rigorous when the corresponding processing
models have a more unified processing model.

Every {{Element}} has a list of <dfn>associated background image requests</dfn>
which is initially an empty array. When the processing model for the {{Element}}
<em>element</em>'s style requires a new image resource (to be displayed as
background image), the <a>image request</a> created by the new resource is
appended to <em>element</em>'s <a>associated background image requests</a>.

NOTE: An {{Element}} can have several [=image requests=], e.g. if its
<a>background-image</a> property has multiple values. For instance, in the
following example, a single <a>background-image</a> property produces four
[=image requests=], each of which will be recorded and reported by the
algorithms below.

```html
<!DOCTYPE html>
<style>
div {
  background-image: url(https://images.example/background1.png),
                    url(https://images.example/background2.png);
}
</style>
<div></div>
<div></div>
```

Recording paint timing {#sec-recording-paint-timing}
--------------------------------------------------------

A <dfn>pending image record</dfn> is a [=struct=] with the following
[=struct/items=]:

    * <dfn for="pending image record">element</dfn>, an {{Element}}
    * <dfn for="pending image record">request</dfn>, an [=image request=]
    * <dfn for="pending image record">loadTime</dfn>, a {{DOMHighResTimeStamp}}

Each {{Element}} has a <dfn>set of owned text nodes</dfn>, which is an [=ordered set=] of {{Text}} nodes, initially empty.

Each {{Document}} has a <dfn>set of previously reported paints</dfn>, which is an [=ordered set=] of [=strings=], initially empty.

Each {{Document}} has an <dfn>images pending rendering</dfn>, which is a [=/list=] of [=pending image records=], initally empty.

Each {{Document}} has a <dfn>set of elements with rendered text</dfn>, which is an [=ordered set=] of {{Element}}s, initially empty.

<h4 id="sec-modifications-CSS">Modifications to the CSS specification</h4>

Whenever an <a>image request</a> in an {{Element}} <em>element</em>'s <a>associated background image requests</a> becomes <a>completely available</a>, run the algorithm to <a>process an image that finished loading</a> with <em>element</em> and <a>image request</a> as inputs.


<h4 id="sec-modifications-dom">Modifications to the HTML specification</h4>

When an {{Element}} <em>element</em>'s <a>associated image request</a> has become <a>completely available</a>, run the algorithm to <a>process an image that finished loading</a> passing in <em>element</em> and its <a>associated image request</a> as inputs.

<div algorithm="text aggregation">
When the user agent paints a {{Text}} node |text| for the first time, it should execute the following steps:

    * If |text| will not be painted due to the font face being in its <a>font block period</a>, then return.
    * Let |element| be the {{Element}} which determines the <a>containing block</a> of |text|.
    * <a for="set">Append</a> |text| to |element|'s <a>set of owned text nodes</a>.
</div>


<h4 dfn>Process image that finished loading</h4>

<div algorithm="image element loaded">
To <dfn>process an image that finished loading</dfn> given an {{Element}} |element| and an [=image request=] |imageRequest|:
    1. Let |root| be |element|'s [=tree/root=].
    1. If |root| is not a {{Document}}, return.
    1. Let |now| be the [=current high resolution time=] given |element|'s <a>relevant global object</a>.
    1. Let |record| be a [=pending image record=] with [=pending image record/element=] |element|, [=pending image record/request=] |imageRequest| and [=pending image record/loadTime=] |now|.
    1. If |imageRequest| is a data URL [[RFC2397]], or the <a>timing allow check</a> succeeds for |imageRequest|'s resource, add |record| to |root|'s [=images pending rendering=].
    1. Otherwise, [=report element timing=] given |root|, |now|, «|record|» and «».
</div>

Reporting paint timing {#sec-reporting-paint-timing}
--------------------------------------------------------

<h4 dfn export>First Contentful Paint</h4>
<div algorithm="Should report first contentful paint">
    To know whether [=Document=] |document| <dfn>should report first contentful paint</dfn>, perform the following steps:
    1. If |document|'s [=set of previously reported paints=] contains <code>"first-contentful-paint"</code>, then return false.
    1. If |document| contains at least one [=/element=] that is both [=paintable=] and [=contentful=], then return true.
    1. Otherwise, return false.
</div>

<h4 dfn export>Mark paint timing</h4>

<div algorithm="Mark paint timing">
    When asked to [=mark paint timing=] given a [=Document=] |document| as input, perform the following steps:
    1. If the [=document=]'s [=Document/browsing context=] is not [=paint-timing eligible=], return.
    1. Let |paintTimestamp| be the [=current high resolution time=] given |document|'s [=relevant global object=].
    1. Let |paintedImages| be a new [=ordered set=]
    1. Let |paintedTextNodes| be a new [=ordered set=]
    1. For each |record| in |doc|'s [=images pending rendering=] list:
        1. If |record|'s [=pending image record/request=] is [=available=] and ready to be painted, then run the following steps:
            1. Append |record| to |paintedImages|.
            1. Remove |record| from |doc|'s <a>images pending rendering</a> list.
    1. For each {{Element}} |element| in |doc|'s <a>descendants</a>:
        1. If |element| is contained in |doc|'s <a>set of elements with rendered text</a>, continue.
        1. If |element|'s <a>set of owned text nodes</a> is empty, continue.
        1. [=set/Append=] |element| to |doc|'s <a>set of elements with rendered text</a>.
        1. [=set/Append=] |element| to |paintedTextNodes|.
    1. Let |reportedPaints| be the |document|'s [=set of previously reported paints=].
    1. If |reportedPaints| does not contain <code>"first-paint"</code>, and the user agent is configured to mark [=first paint=], then invoke the [[#report-paint-timing]] algorithm with |document|, <code>"first-paint"</code>, and |paintTimestamp|.

        NOTE: [=First paint=] excludes the default background paint, but includes non-default background paint.

            ISSUE: This should be turned into a normative note.

    1. If |document| [=should report first contentful paint=], then:
        1. Invoke the [[#report-paint-timing]] algorithm with |document|, <code>"first-contentful-paint"</code>, and |paintTimestamp| as arguments.

        NOTE: A parent frame should not be aware of the paint events from its child iframes, and vice versa. This means that a frame that contains just iframes will have [=first paint=] (due to the enclosing boxes of the iframes) but no [=first contentful paint=].

        NOTE: A [=document=] is not guaranteed to mark <code>"first-paint"</code> or <code>"first-contentful-paint"</code>. A completely blank [=document=] may never mark [=first paint=], and a [=document=] containing only elements that are not [=contentful=] may never mark [=first contentful paint=].

        NOTE: The marking of [=first paint=] is optional. User-agents implementing paint timing should at the very least mark [=first contentful paint=].

    1. [=Report largest contentful paint=] given |document|, |paintTimestamp|,
        |paintedImages| and |paintedTextNodes|.
    1. [=Report element timing=] given |document|, |paintTimestamp|,
        |paintedImages| and |paintedTextNodes|.
</div>

<h4 dfn>Report paint timing</h4>

<div algorithm="Report paint timing">
    When asked to [=report paint timing=] given |document|, |paintType|, and |paintTimestamp| as arguments, perform the following steps:
    1. Create a <a spec=webidl>new</a> {{PerformancePaintTiming}} object |newEntry| with |document|'s [=relevant realm=] and set its attributes as follows:
        1. Set |newEntry|'s {{PerformanceEntry/name}} attribute to |paintType|.
        1. Set |newEntry|'s {{PerformanceEntry/entryType}} attribute to <code>"paint"</code>.
        1. Set |newEntry|'s {{PerformanceEntry/startTime}} attribute to |paintTimestamp|.
        1. Set |newEntry|'s {{PerformanceEntry/duration}} attribute to 0.
    1. <a href="https://w3c.github.io/performance-timeline/#dfn-queue-a-performanceentry">Add the PerformanceEntry</a> |newEntry| object.
    1. [=list/Append=] |paintType| to |document|'s [=set of previously reported paints=].
</div>


Common algorithms {#sec-common-algorithms}
------------------------------------------

<h4 dfn export>Exposed for paint timing</h4>

<div algorithm="Exposed for paint timing">

    To determine whether an [=/Element=] |element| is [=exposed for paint timing=], given a [=Document=] or null |document|, perform the following steps:

    1. If |element| is not [=connected=], return false.
    1. If |document| is null, let |document| be |element|'s [=relevant settings object=]'s [=relevant global object=]'s [=associated document=].
    1. If |document| is not [=fully active=], return false.
    1. If |element|'s [=tree/root=] is not equal to |document|, return false.
    1. Return true.
</div>



<!-- ============================================================ -->
<h2 id=acknowledgements>Acknowledgements</h2>
<!-- ============================================================ -->

Special thanks to <a href="https://github.com/w3c/paint-timing/graphs/contributors">all the contributors</a> for their technical input and suggestions that led to improvements to this
specification.