InnerHTML Bindings for SSR


Applications dealing with rich content, like commerce apps, need to render this content as raw HTML, typically coming from a database or a CMS. Today, the data binding capability always escapes the HTML for security reasons, which makes the rich content data not rendered as expected.

Basic example

The following template render some HTML content as innerHTML

function createMarkup() { return {__html: 'First · Second'}; };

    <div lwc:inner-html={createMarkup}/>

The directive is named inner-html as it mimics what the DOM innerHTML property does. As such, it can be used on any tag where the innerHTML property is available. The HTML fragment inserted is not processed by the LWC compiler and thus LWC components won't be created unless they are exposed as custom elements to the browser.


It is common in customer facing apps to render rich content coming from a CMS or equivalent.

As of today, and according to the LWC documentation, this capability is supported by acting directly with the DOM within renderedCallback(). The templating mechanism does not provide any binding capability that renders some content as raw HTML.

As a side effect, this also breaks with SSR, as renderedCallback() and DOM manipulation won't be allowed when a component is rendered on the server. Moreover, even if it was possible, it would make the client side hydration more complex as the content is not generated as part of the VDOM.

Having to act on the DOM is also not user friendly compared to a binding facility.

Detailed design

It seems that there is a consensus among the most popular libraries to expose the innerHTML property of an element, although the exposed name for this property varies:

React: dangerouslySetInnerHTML attribute.
Angular: [innerHTML].
VueJS: v-html attribute.

The Angular choice of innerHTML feels the most natural, while the React dangerouslyInnerHTML has probably be named this way to not conflict with any potential component property. To avoid any name collision, LWC can prefix it with the technical "lwc:" prefix.

React makes it even more secure by forcing the HTML to be added to a temporary object with a __html property. This avoids undesired assignment from a method that does not sanitize the content:

    <div lwc:inner-html={getUserName()}/>

It is up to the component developer to make sure that the injected HTML is properly sanitized ( Several client side libraries, easily consumable by an LWC application, are available to do the job.

The compiler generates an error if an element has both an innerHTML property and some content defined, like bellow:

    <div lwc:inner-html={myHTML()}>
      Adding content when innerMTML is defined leads to a compilation error


The lwc:inner-html attribute can be set on any tag in an LWC template with the following restrictions:

Implementation details

From an implementation standpoint, the HTML fragment is kept in the host element attribute. When the VDOM comparison finds that this attribute has changed, then it updates the DOM element content by assigning it to the element.innerHTML property.


Note: the security issue is not different from setting the HTML with a DOM operation in renderedCallback(). If DOM manipulation is secured by locker, then this should benefit from it.


Different syntaxes for the innerHTML attribute are possible, but they finally lead to the same results. It is then more a matter of preference and consistency.

Another solution would define a new binding syntax, like for example ${{myhtml()}} or ${htnl:myhtml()}}. But such a syntax could be used everywhere a binding is possible, including attribute values. This could lead to undefined behaviors why not providing any value. The proposed syntax makes sure that the innerHTML binding always applies to the content of an element.

Adoption strategy

Because the attribute uses the reserved lwc: prefix, there is not backward compatibility issue, and there is no potential name conflicts.

How we teach this

inner HTML or raw HTML binding are the industry adopted terms for such a feature.