FetchEvent: respondWith() method
Note: This feature is only available in Service Workers.
The respondWith()
method of
FetchEvent
prevents the browser's default fetch handling, and
allows you to provide a promise for a Response
yourself.
In most cases you can provide any response that the receiver understands. For example,
if an <img>
initiates the request, the response body needs to be
image data. For security reasons, there are a few global rules:
- You can only return
Response
objects oftype
"opaque"
if thefetchEvent.request
object'smode
is"no-cors"
. This prevents the leaking of private data. - You can only return
Response
objects oftype
"opaqueredirect"
if thefetchEvent.request
object'smode
is"manual"
. - You cannot return
Response
objects oftype
"cors"
if thefetchEvent.request
object'smode
is"same-origin"
.
Specifying the final URL of a resource
From Firefox 59 onwards, when a service worker provides a Response
to
FetchEvent.respondWith()
, the Response.url
value will be
propagated to the intercepted network request as the final resolved URL. If the
Response.url
value is the empty string, then the
FetchEvent.request.url
is used as the final URL.
In the past the FetchEvent.request.url
was used as the
final URL in all cases. The provided Response.url
was effectively
ignored.
This means, for example, if a service worker intercepts a stylesheet or worker script,
then the provided Response.url
will be used to resolve any relative
@import
or
importScripts()
subresource loads
(Firefox bug 1222008).
For most types of network request this change has no impact because you can't observe the final URL. There are a few, though, where it does matter:
- If a
fetch()
is intercepted, then you can observe the final URL on the result'sResponse.url
. - If a worker script is
intercepted, then the final URL is used to set
self.location
and used as the base URL for relative URLs in the worker script. - If a stylesheet is intercepted, then the final URL is used as the base URL for
resolving relative
@import
loads.
Note that navigation requests for Windows
and
iframes
do NOT use the final URL. The way the HTML
specification handles redirects for navigations ends up using the request URL for the
resulting Window.location
. This means sites can still provide an
"alternate" view of a web page when offline without changing the user-visible URL.
Syntax
respondWith(response)
Parameters
Return value
None (undefined
).
Exceptions
NetworkError
DOMException
-
Returned if a network error is triggered on certain combinations of
FetchEvent.request.mode
andResponse.type
values, as hinted at in the "global rules" listed above. InvalidStateError
DOMException
-
Returned if the event has not been dispatched or
respondWith()
has already been invoked.
Examples
This fetch event tries to return a response from the cache API, falling back to the network otherwise.
addEventListener("fetch", (event) => {
// Prevent the default, and handle the request ourselves.
event.respondWith(
(async () => {
// Try to get the response from a cache.
const cachedResponse = await caches.match(event.request);
// Return it if we found one.
if (cachedResponse) return cachedResponse;
// If we didn't find a match in the cache, use the network.
return fetch(event.request);
})(),
);
});
Note: caches.match()
is a
convenience method. Equivalent functionality is to call
cache.match()
on each cache (in the order returned by
caches.keys()
) until a
Response
is returned.
Specifications
Specification |
---|
Service Workers # fetch-event-respondwith |
Browser compatibility
BCD tables only load in the browser