1. Introduction
This section is non-normative. The LargestContentfulPaint API enables developers to gain visibility into the loading and rendering process of the web pages, in order for them to be able to optimize it.
Developers today don’t have a reliable metric that correlated with their user’s visual rendering experience. Existing metrics such as First Paint and First Contentful Paint focus on initial rendering, but don’t take into account the importance of the painted content, and therefore may indicate times in which the user still does not consider the page useful.
Largest Contentful Paint (LCP) aims to be a new page-load metric that:
-
better correlates with user experience than the existing page-load metrics
-
is easy to understand and reason about
-
reduces the chance of gaming
The largest paint during the loading process of the page is likely to signify a meaningful event from the user’s perspective, and is therefore something we want to expose by default to developers, enabling performance teams, analytics providers and lab-based measurement tools to collect those metrics without requiring extra annotation work by the folks creating the content itself.
The API relies heavily on [ELEMENT-TIMING], which can be thought of as the low-level primitive that this high-level feature is built on top of. For cases where the content creators are willing to annotate their content and indicate the important points in the page’s loading cycle, Element Timing is the API that will provide them more control over the elements that get reported.
1.1. Elements exposed
The Largest Contentful Paint API will only expose element types that are already exposed by the Element Timing API. In this case, there is no need to annotate them with the elementtiming
attribute.
1.2. Largest content
The algorithm used for this API keeps track of the content seen so far. Whenever a new largest content is found, a new entry is created for it. Whenever content is removed, that content is no longer considered by the algorithm. In particular, if the content removed was the largest, then a new entry is created for the new largest. The algorithm terminates whenever scroll or input events occur, since those are likely to introduce new content into the website.
1.3. Usage example
The following example shows an image and a large body of text. The developer then registers an observer that gets candidate entries for the largest paint while the page is loading.
< img src = "large_image.jpg" > < p id = 'large-paragraph' > This is large body of text.</ p > ...< script > const observer= new PerformanceObserver(( list) => { let perfEntries= list. getEntries(); let lastEntry= perfEntries[ perfEntries. length- 1 ]; // Process the latest candidate for largest contentful paint }); observer. observe({ entryTypes: [ 'largest-contentful-paint' ]}); </ script >
1.4. Limitations
The LargestContentfulPaint API is based on heuristics. As such, it is error prone. It has the following problems:
-
The algorithm halts when it detects certain types of user inputs. However, this means that the algorithm will not capture the main content if the user input occurs before the main content is displayed. In fact, the algorithm may produce meaningless results or no results at all if user input occurs very early.
-
To account for splash screens, content cannot be considered as the largest once it is removed. This presents problems for websites with large image carousels where images rotate automatically. If the image is removed when the next one is painted and the carousel is the largest content, the algorithm will continuously update based on carousel updates.
2. Largest Contentful Paint
Largest Contentful Paint involves the following new interface:
2.1. LargestContentfulPaint
interface
[Exposed =Window ]interface :
LargestContentfulPaint PerformanceEntry {readonly attribute DOMHighResTimeStamp ;
renderTime readonly attribute DOMHighResTimeStamp ;
loadTime readonly attribute unsigned long ;
size readonly attribute DOMString ;
id readonly attribute DOMString ;
url readonly attribute Element ?; [
element Default ]object (); };
toJSON
Each LargestContentfulPaint
object has these associated concepts:
-
A renderTime, initially set to 0.
-
A size, initially set to 0.
-
A loadTime, initially set to 0.
-
An id, initially set to the empty string.
-
A url, initially set to the empty string.
-
An element containing the associated
Element
, initially set to
.null
The entryType
attribute’s getter must return the DOMString
.
The name
attribute’s getter must return the value it was initialized to.
The startTime
attribute’s getter must return the value of the context object’s renderTime if it is not 0, and the value of the context object’s loadTime otherwise.
The duration
attribute’s getter must return 0.
The renderTime
attribute must return the value of the context object’s renderTime.
The loadTime
attribute must return the value of the context object’s loadTime.
The size
attribute must return the value of the context object’s size.
The id
attribute must return the value of the context object’s id.
The url
attribute must return the value of the context object’s url.
The element
attribute’s getter must return the value returned by running the get an element algorithm with element and null as inputs.
Note: The above algorithm defines that an element that is no longer descendant of the Document
will no longer be returned by element
's attribute getter, including elements that are inside a shadow DOM.
This specification also extends Document
by adding to it a largest contentful paint size concept, initially set to 0, and a largest content, initially set to null.
It also adds an associated content map, which is initially an empty map. The content map will be filled with entries with the following format:
-
The key will be a pair with an
Element
as the first item and aRequest
as the second item. This allows identifying the content. The second item will be null for text content. -
The value will be a map which contains information that is required to fill up the
LargestContentfulPaint
entry. This allows exposing a newLargestContentfulPaint
entry when content is removed from the page.
Note: A user agent probably wants to implement the Document
's associated concepts using a priority queue or binary search tree to avoid the O(n) cost of finding the largest size within content map when largest content is removed.
3. Processing model
3.1. Potentially add LargestContentfulPaint entry
Note: A user agent implementing the Largest Contentful Paint API would need to include
in supportedEntryTypes
for Window
contexts.
This allows developers to detect support for the API.
In order to potentially add a LargestContentfulPaint
entry, the user agent must run the following steps:
- Input
-
intersectionRect, a
DOMRectReadOnly
imageRequest, a
Request
renderTime, a DOMHighResTimestamp
loadTime, a DOMHighResTimestamp
element, an Element
document, a Document
- Output
-
None
-
Let contentIdentifier be the pair (element, imageRequest).
-
If document’s content map contains contentIdentifier, return.
-
Let window be document’s relevant global object.
-
If either of window’s has dispatched scroll event or has dispatched input event is true, return.
-
Let url be the empty string.
-
If imageRequest is not null, set url to be imageRequest’s request URL.
-
Let id be element’s element id.
-
Let width be intersectionRect’s
width
. -
Let height be intersectionRect’s
height
. -
Let size be
width
.* height -
If imageRequest is not null, run the following steps:
-
Let naturalWidth and naturalHeight be the outputs of running the same steps for an
img
'snaturalWidth
andnaturalHeight
attribute getters, but using imageRequest as the image. -
Let naturalSize be
naturalWidth
.* naturalHeight -
Let displayWidth and displayHeight be the outputs of running the same steps for an
img
'swidth
andheight
attribute getters, but using imageRequest as the image. -
Let displaySize be
displayWidth
.* displayHeight -
Let penaltyFactor be
min
.( displaySize, naturalSize) / displaySize -
Multiply size by penaltyFactor.
-
-
Let contentInfo be a map with contentInfo["size"] = size, contentInfo["url"] = url, contentInfo["id"] = id, contentInfo["renderTime"] = renderTime, and contentInfo["loadTime"] = loadTime.
-
Add a new entry with contentIdentifier as the key and contentInfo as the value to document’s content map.
-
If size is smaller or equal to document’s largest contentful paint size, return.
-
Create a LargestContentfulPaint entry with element and contentInfo as inputs.
-
3.2. Remove element content
In order to remove element content, the user agent must run the following steps:
- Input
-
element, an Element
document, a Document
- Output
-
None
-
For each entry of document’s content map:
-
If entry’s key is a pair whose first item is equal to element, remove entry from document’s content map.
-
-
If document’s largest content is a pair whose first item is not equal to element, then return.
-
Let largestSize be 0, let largestContentIdentifier be null, and let largestContentInfo be null.
-
For each key → value of document’s content map:
-
If value["size"] is greater than largestSize, set largestSize to value["size"], largestContentIdentifier to key, and largestContentInfo to value.
-
-
If largestContentIdentifier is not null, create a LargestContentfulPaint entry with largestContentIdentifier and largestContentInfo as inputs.
-
3.3. Create a LargestContentfulPaint entry
In order to create a LargestContentfulPaint
entry, the user agent must run the following steps:
- Input
-
contentIdentifier, a pair
contentInfo, a map
- Output
-
None
-
Set document’s largest content to contentIdentifier.
-
Set document’s largest contentful paint size to contentInfo["size"].
-
Let entry be a new
LargestContentfulPaint
entry, with it’ssize
set to contentInfo["size"],url
set to contentInfo["url"],id
set to contentInfo["id"],renderTime
set to contentInfo["renderTime"],loadTime
set to contentInfo["loadTime"], and itselement
set to contentIdentifier’s first item. -
Queue the PerformanceEntry entry.
-
3.4. Modifications to the DOM specification
This section will be removed once the [DOM] specification has been modified.
Right after step 1, we add the following step:
-
If target’s relevant global object is a
Window
object, event’stype
isscroll
and itsisTrusted
is false, set target’s relevant global object's has dispatched scroll event to true.
-
Call the algorithm to remove element content passing in node and node’s node document.
background image changes or changes in image src
should trigger removal from the Document
's content map. <https://github.com/WICG/largest-contentful-paint/issues/41>
3.5. Modifications to the HTML specification
This section will be removed once the [HTML] specification has been modified.Each Window
has has dispatched scroll event, a boolean which is initially set to false.
4. Security & privacy considerations
This API relies on Element Timing for its underlying primitives. LCP may expose some element not exposed by Element Timing in case that they are smaller than Element Timing’s limits, but are still the largest elements to be painted up until that point in the page’s loading. That does not seem to expose any sensitive information beyond what Element Timing already enables.