This document defines the will-change CSS property, which allows an author to inform the UA ahead of time of what kinds of changes they are likely to make to an element. This allows the UA to optimize how they handle the element ahead of time, performing potentially-expensive work preparing for an animation before the animation actually begins.
CSS is a language for describing the rendering of structured documents
(such as HTML and XML)
on screen, on paper, in speech, etc.
Status of this document
This section describes the status of this document at the time of its publication.
Other documents may supersede this document.
A list of current W3C publications and the latest revision of this technical report
can be found in the W3C technical reports index at http://www.w3.org/TR/.
This document was produced by the CSS Working Group
(part of the Style Activity)
as a Candidate Recommendation. This document is intended to become a W3C Recommendation.
This document will remain a Candidate Recommendation at least until in order
to ensure the opportunity for wide review.
Publication as a Candidate Recommendation does not imply endorsement by the W3C Membership. This is a draft document and may be updated, replaced or obsoleted by other documents at any time. It is inappropriate to cite this document as other than work in progress.
The (archived)
public mailing list [email protected]
(see instructions)
is preferred for discussion of this specification.
When sending e-mail,
please put the text “css-will-change” in the subject,
preferably like this:
“[css-will-change] …summary of comment…”
A test suite is under development but as yet there is no Implementation Report.
Modern CSS renderers perform a number of complex optimizations in order to render webpages quickly and efficiently.
Unfortunately, employing these optimizations often has a non-trivial start-up cost,
which can have a negative impact on the responsiveness of a page.
For example, when using CSS 3D Transforms to move an element around the screen,
the element and its contents might be promoted to a “layer”,
where they can render independently from the rest of the page and be composited in later.
This isolates the rendering of the content so that the rest of the page doesn’t have to be rerendered
if the element’s transform is the only thing that changes between frames,
and often provides significant speed benefits.
However, setting up the element in a fresh layer is a relatively expensive operation,
which can delay the start of a transform animation by a noticeable fraction of a second.
The will-change property defined in this specification allows an author to declare ahead-of-time what properties are likely to change in the future,
so the UA can set up the appropriate optimizations some time before they’re needed.
This way, when the actual change happens,
the page updates in a snappy manner.
The will-change property,
like all performance hints,
can be somewhat difficult to learn how to use “properly”,
particularly since it has very little, if any, effect an author can directly detect.
However, there are several simple “Dos and Don’ts”
which hopefully will help develop a good intuition about how to use will-change well.
Don’t Spam will-change Across Too Many Properties or Elements
A common initial response to seeing will-change is to assume that code like this is a good idea:
After all, this tells the browser to go ahead and optimize everything,
which has to be good right?
Wrong. The browser already tries as hard as it can to optimize everything.
Telling it to do so explicitly doesn’t help anything,
and in fact has the capacity to do a lot of harm;
some of the stronger optimizations that are likely to be tied to will-change
end up using a lot of a machine’s resources,
and when overused like this can cause the page to slow down or even crash.
In addition, will-change does have some side-effects,
and it’s very unlikely that pages actually want all those side-effects on every element.
Using will-change directly in a stylesheet
implies that the targeted elements are always a few moments away from changing.
This is usually not what you actually mean;
instead, will-change should usually be flipped on and off via scripting
before and after the change occurs (see
Don’t Waste Resources On Elements That Have Stopped Changing).
However, there are some common circumstances in which it is appropriate to use will-change directly in a stylesheet.
For example,
specifying will-change for a small number of persistent UI elements in a page
which should react snappily to the user
is appropriate:
body > .sidebar {
will-change: transform;
/* Will use 'transform' to slide it out
when the user requests. */
}
Because this is limited to a small number of elements,
the fact that the optimization is rarely actually used
doesn’t hurt very much.
Sometimes an element really does change a property nearly constantly.
Perhaps it responds to the user’s mouse movements,
or just regularly takes some action that causes an animation.
In this case, just declaring the will-change value in the stylesheet is fine,
as it accurately describes that the element will regularly/constantly change,
and so should be kept optimized.
Another common bad pattern is to apply will-change to an element
immediately before starting the animation or property change that it’s meant to help with.
Unfortunately, most of those optimizations need time to be applied,
and so they don’t have enough time to set-up when this is done,
and the will-change has little to no effect.
Instead, find some way to predict at least slightly ahead of time that something will change,
and set will-changethen.
For example,
if an element is going to change when a user clicks on it,
setting will-change on hover will usually give at least 200 milliseconds
for the optimizations to be set up,
as human reaction time is relatively slow.
This can be done either via scripting,
or rather simply with a CSS rule:
However, a rule like that is useless if the effect is going to happen on hover.
In cases like these, it is often still possible to find some way to predict the action before it occurs.
For example, hovering an ancestor may give enough lead time:
Don’t Waste Resources On Elements That Have Stopped Changing
Because the optimizations browsers use for changing some properties are expensive,
browsers remove them and revert to normal behavior as soon as they can in normal circumstances.
However, will-change will generally override this behavior,
maintaining the optimizations for much longer than the browser would otherwise do.
As such, whenever you add will-change to an element,
especially via scripting,
don’t forget to remove it after the element is done changing,
so the browser can recover whatever resources the optimizations are claiming.
2.
Hinting at Future Behavior: the will-change property
The will-change property provides a rendering hint to the user agent,
stating what kinds of changes the author expects to perform on the element.
This allows the user agent to perform ahead-of-time any optimizations necessary for rendering those changes smoothly,
avoiding “jank” when the author does begin changing or animating that feature.
Different browsers can use the information from will-change in different ways,
and even a single browser might use it in different ways at different time.
For example, a browser that promotes elements to their own “GPU layer”
when they have will-change: transform specified
might avoid doing that when there are too many elements declaring that,
to avoid exhausting GPU memory.
Values have the following meanings:
auto
Expresses no particular intent;
the user agent should apply whatever heuristics and optimizations it normally does.
scroll-position
Indicates that the author expects to animate or change the scroll position of the element in the near future.
For example, browsers often only render the content in the "scroll window" on a scrollable element,
and some of the content past that window,
balancing memory and time savings from the skipped rendering against making scrolling look nice.
A browser might take this value as a signal to expand the range of content around the scroll window that is rendered,
so that longer/faster scrolls can be done smoothly.
contents
Indicates that the author expects to animate or change something about the element’s contents in the near future.
For example, browsers often “cache” rendering of elements over time,
because most things don’t change very often,
or only change their position.
However, if an element does change its contents continually,
producing and maintaining this cache is a waste of time.
A browser might take this value as a signal to cache less aggressively on the element,
or avoid caching at all and just continually re-render the element from scratch.
This value is mostly intended to help browsers optimize JS-based animations of content,
which change aspects of an element’s contents many times per second.
This kind of optimization, when possible,
is already done automatically by browsers when declarative animations are used.
Note: This value more-or-less applies to the entire subtree of the element its declared on,
as it indicates the browser should count on *any* of the descendants changing in some way.
Using this on an element “high up” in your document might be very bad for your page’s performance;
try to only use this on elements near the “bottom” of your document tree,
containing as little of the document as possible.
Indicates that the author expects to animate or change the property with the given name on the element in the near future.
If the property given is a shorthand,
it indicates the expectation for all the longhands the shorthand expands to.
Note: Note that most properties will have no effect when specified,
as the user agent doesn’t perform any special optimizations for changes in most properties.
It is still safe to specify them, though;
it’ll simply have no effect.
For example, browsers often handle elements with transform set to a non-initial value very differently from normal elements,
perhaps rendering them to their own “GPU layer”
or using other mechanisms to make it easier to quickly make the sort of transformations that transform can produce.
A browser might take a value of transform as a signal
that it should go ahead and promote the element to its own layer immediately,
before the element starts to be transformed,
to avoid any delay involved in rerendering the old and new layers.
If any non-initial value of a property would create a stacking context on the element,
specifying that property in will-change must create a stacking context on the element.
If any non-initial value of a property would cause the element to generate a containing block for fixed-position elements,
specifying that property in will-change must cause the element to generate a containing block for fixed-position elements.
If a non-initial value of a property would cause rendering differences on the element
(such as using a different anti-aliasing strategy for text),
the user agent should use that alternate rendering when the property is specified in will-change,
to avoid sudden rendering differences when the property is eventually changed.
For example, setting opacity to any value other than 1 creates a stacking context on the element.
Thus, setting will-change: opacity also creates a stacking context,
even if opacity is currently still equal to 1.
The will-change property has no direct effect on the element it is specified on,
beyond the creation of stacking contexts and containing blocks as specified above.
It is solely a rendering hint to the user agent,
allowing it set up potentially-expensive optimizations for certain types of changes
before the changes actually start occurring.
3. Acknowledgements
Thanks to Benoit Girard for originally suggesting the will-animate property,
and doing a lot of the initial design work.
4. Changes since the April 29 2014 Working Draft
Added an explanatory section giving guidance on how to use will-change well.
Specified the behavior of shorthands
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.
Advisements are normative sections styled to evoke special attention and are
set apart from other normative text with <strong class="advisement">, like
this:
UAs MUST provide an accessible alternative.
Conformance classes
Conformance to this specification
is defined for three conformance classes:
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 CSS2.1 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
[email protected]
mailing list.
CR exit criteria
For this specification to be advanced to Proposed Recommendation,
there must be at least two independent, interoperable implementations
of each feature. Each feature may be implemented by a different set of
products, there is no requirement that all features be implemented by
a single product. For the purposes of this criterion, we define the
following terms:
independent
each implementation must be developed by a
different party and cannot share, reuse, or derive from code
used by another qualifying implementation. Sections of code that
have no bearing on the implementation of this specification are
exempt from this requirement.
interoperable
passing the respective test case(s) in the
official CSS test suite, or, if the implementation is not a Web
browser, an equivalent test. Every relevant test in the test
suite should have an equivalent test created if such a user
agent (UA) is to be used to claim interoperability. In addition
if such a UA is to be used to claim interoperability, then there
must one or more additional UAs which can also pass those
equivalent tests in the same way for the purpose of
interoperability. The equivalent tests must be made publicly
available for the purposes of peer review.
implementation
a user agent which:
implements the specification.
is available to the general public. The implementation may
be a shipping product or other publicly available version
(i.e., beta version, preview release, or "nightly build").
Non-shipping product releases must have implemented the
feature(s) for a period of at least one month in order to
demonstrate stability.
is not experimental (i.e., a version specifically designed
to pass the test suite and is not intended for normal usage
going forward).
The specification will remain Candidate Recommendation for at least
six months.