sw-toolbox
andsw-precache
are deprecated in favor of Workbox.
Please readthis migration guide
for information on upgrading.
Service Worker Precache is a module for generating a service worker that precaches resources. It integrates with your build process. Once configured, it detects all your static resources (HTML, JavaScript, CSS, images, etc.) and generates a hash of each file's contents. Information about each file's URL and versioned hash are stored in the generated service worker file, along with logic to serve those files cache-first, and automatically keep those files up to date when changes are detected in subsequent builds.
Serving your local static resources cache-first means that you can get all the crucial scaffolding for your web app—your App Shell—on the screen without having to wait for any network responses.
The module can be used in JavaScript-based build scripts,
like those written withgulp
,and it also provides a
command-line interface.You can use the module
directly, or if you'd prefer, use one of thewrappers
aroundsw-precache
for specific build environments, like
webpack
.
It can beused alongsidethesw-toolbox
library, which works well when following the App Shell + dynamic content model.
The full documentation is in this README, and the getting started guideprovides a quicker jumping off point.
To learn more about the internals of the generated service worker, you can read this deep-dive byHuang Xuan.
- Install
- Usage
- Runtime Caching
- API
- Methods
- Options Parameter
- cacheId [String]
- clientsClaim [Boolean]
- directoryIndex [String]
- dontCacheBustUrlsMatching [Regex]
- dynamicUrlToDependencies [Object⟨String,Buffer,Array⟨String⟩⟩]
- handleFetch [boolean]
- ignoreUrlParametersMatching [Array⟨Regex⟩]
- importScripts [Array⟨String⟩]
- logger [function]
- maximumFileSizeToCacheInBytes [Number]
- navigateFallback [String]
- navigateFallbackWhitelist [Array⟨RegExp⟩]
- replacePrefix [String]
- runtimeCaching [Array⟨Object⟩]
- skipWaiting [Boolean]
- staticFileGlobs [Array⟨String⟩]
- stripPrefix [String]
- stripPrefixMulti [Object]
- templateFilePath [String]
- verbose [boolean]
- Wrappers and Starter Kits
- Acknowledgements
- Support
- License
Local build integration:
$ npm install --save-dev sw-precache
Global command-line interface:
$ npm install --global sw-precache
-
Make sure your site is served using HTTPS! Service worker functionality is only available on pages that are accessed via HTTPS. (
http://localhost
will also work, to facilitate testing.) The rationale for this restriction is outlined in the "Prefer Secure Origins For Powerful New Features" document. -
Incorporate
sw-precache
into yournode
-based build script.It should work well with eithergulp
orGrunt
,or other build scripts that run onnode
.In fact, we've provided examples of both in thedemo/
directory. Each build script indemo
has a function calledwriteServiceWorkerFile()
that shows how to use the API. Both scripts generate fully-functional JavaScript code that takes care of precaching and fetching all the resources your site needs to function offline. There is also acommand-line interface available, for those using alternate build setups. -
Register the service worker JavaScript.The JavaScript that's generated needs to be registered as the controlling service worker for your pages. This technically only needs to be done from within a top-level "entry" page for your site, since the registration includes a
scope
which will apply to all pages underneath your top-level page.service-worker-registration.js
is a sample script that illustrates the best practices for registering the generated service worker and handling the variouslifecycleevents.
The project'ssamplegulpfile.js
illustrates the full use of sw-precache
in context. (Note that the sample gulpfile.js is the one in thedemo
folder,
not the one in the root of the project.) You can run the sample by cloning this
repo, usingnpm install
to pull in the
dependencies, changing to thedemo/
directory, running`npm bin`/gulp serve-dist
,and
then visitinghttp://localhost:3000.
There's also asampleGruntfile.js
that shows service worker generation in
Grunt. Though, it doesn't run a server on localhost.
Here's a simpler gulp example for a basic use case. It assumes your site's resources are located under
app
and that you'd like to cacheallyour JavaScript, HTML, CSS, and image files.
gulp.task('generate-service-worker',function(callback){
varswPrecache=require('sw-precache');
varrootDir='app';
swPrecache.write(`${rootDir}/service-worker.js`,{
staticFileGlobs:[rootDir+'/**/*.{js,html,css,png,jpg,gif,svg,eot,ttf,woff}'],
stripPrefix:rootDir
},callback);
});
This task will createapp/service-worker.js
,which your client pages need to
registerbefore it can take control of your site's
pages.service-worker-registration.js
is a ready-to-
use script to handle registration.
-
Service worker caching should be considered a progressive enhancement. If you follow the model of conditionally registering a service worker only if it's supported (determined by
if('serviceWorker' in navigator)
), you'll get offline support on browsers with service workers and on browsers that don't support service workers, the offline-specific code will never be called. There's no overhead/breakage for older browsers if you addsw-precache
to your build. -
Allresources that are precached will be fetched by a service worker running in a separate thread as soon as the service worker is installed. You should be judicious in what you list in the
dynamicUrlToDependencies
andstaticFileGlobs
options, since listing files that are non-essential (large images that are not shown on every page, for instance) will result in browsers downloading more data than is strictly necessary. -
Precaching doesn't make sense for all types of resources (see the previous point). Other caching strategies, like those outlined in theOffline Cookbook,can be used in conjunction with
sw-precache
to provide the best experience for your users. If you do implement additional caching logic, put the code in a separate JavaScript file and include it using theimportScripts()
method. -
sw-precache
uses acache-firststrategy, which results in a copy of any cached content being returned without consulting the network. A useful pattern to adopt with this strategy is to display a toast/alert to your users when there's new content available, and give them an opportunity to reload the page to pick up that new content (which the service worker will have added to the cache, and will be available at the next page load). The sampleservice-worker-registration.js
fileillustrates the service worker lifecycle event you can listen for to trigger this message.
For those who would prefer not to usesw-precache
as part of agulp
or
Grunt
build, there's acommand-line interfacewhich supports the
options listedin the API, provided via flags or an
external JavaScript configuration file.
Hypenated flags are converted to camelCaseoptions.
Options starting with--no
prefix negate the boolean value. For example,--no-clients-claim
sets the value ofclientsClaim
tofalse
.
Warning:When usingsw-precache
"by hand", outside of an automated build process, it's your
responsibility to re-run the command each time there's a change to any local resources! Ifsw-precache
is not run again, the previously cached local resources will be reused indefinitely.
Sensible defaults are assumed for options that are not provided. For example, if you are inside
the top-level directory that contains your site's contents, and you'd like to generate a
service-worker.js
file that will automatically precache all of the local files, you can simply run
$ sw-precache
Alternatively, if you'd like to only precache.html
files that live withindist/
,which is a
subdirectory of the current directory, you could run
$ sw-precache --root=dist --static-file-globs='dist/**/*.html'
Note:Be sure to use quotes around parameter values that have special meanings
to your shell (such as the*
characters in the sample command line above,
for example).
Finally, there's support for passing complex configurations using--config <file>
.
Any of the options from the file can be overridden via a command-line flag.
We strongly recommend passing it an external JavaScript file defining config via
module.exports
.
For example, assume there's apath/to/sw-precache-config.js
file that contains:
module.exports={
staticFileGlobs:[
'app/css/**.css',
'app/**.html',
'app/images/**.*',
'app/js/**.js'
],
stripPrefix:'app/',
runtimeCaching:[{
urlPattern:/this\\.is\\.a\\.regex/,
handler:'networkFirst'
}]
};
That file could be passed to the command-line interface, while also setting the
verbose
option, via
$ sw-precache --config=path/to/sw-precache-config.js --verbose
This provides the most flexibility, such as providing a regular expression for
theruntimeCaching.urlPattern
option.
We also support passing in a JSON file for--config
,though this provides
less flexibility:
{
"staticFileGlobs":[
"app/css/**.css",
"app/**.html",
"app/images/**.*",
"app/js/**.js"
],
"stripPrefix":"app/",
"runtimeCaching":[{
"urlPattern":"/express/style/path/(.*)",
"handler":"networkFirst"
}]
}
It's often desireable, even necessary to use precaching and runtime caching together. You may have seen oursw-toolbox
tool, which handles runtime caching, and wondered how to use them together. Fortunately,sw-precache
handles this for you.
Thesw-precache
module has the ability to include thesw-toolbox
code and configuration alongside its own configuration. Using theruntimeCaching
configuration option insw-precache
(see below) is a shortcut that accomplishes what you could do manually by importingsw-toolbox
in your service worker and writing your own routing rules.
Thesw-precache
module exposes two methods:generate
andwrite
.
generate
takes inoptions,generates a service worker
from them and passes the result to a callback function, which must
have the following interface:
callback(error, serviceWorkerString)
In the 1.x releases ofsw-precache
,this was the default and only method
exposed by the module.
Since 2.2.0,generate()
also returns a
Promise
.
write
takes inoptions,generates a service worker from them,
and writes the service worker to a specified file. This method always
invokescallback(error)
.If no error was found, theerror
parameter will
benull
Since 2.2.0,write()
also returns aPromise
.
Both thegenerate()
andwrite()
methods take the same options.
A string used to distinguish the caches created by different web applications that are served off
of the same origin and path. While serving completely different sites from the same URL is not
likely to be an issue in a production environment, it avoids cache-conflicts when testing various
projects all served off ofhttp://localhost
.You may want to set it to, e.g., thename
property from yourpackage.json
.
Default:''
Controls whether or not the generated service worker will call
clients.claim()
inside theactivate
handler.
Callingclients.claim()
allows a newly registered service worker to take
control of a page immediately, instead of having to wait until the next page
navigation.
Default:true
Sets a default filename to return for URL's formatted like directory paths (in
other words, those ending in'/'
).sw-precache
will take that translation
into account and serve the contents a relativedirectoryIndex
file when
there's no other match for a URL ending in'/'
.To turn off this behavior,
setdirectoryIndex
tofalse
ornull
.To override this behavior for one
or more URLs, use thedynamicUrlToDependencies
option to explicitly set up
mappings between a directory URL and a corresponding file.
Default:'index.html'
It's very important that the requestssw-precache
makes to populate your cache
result in the most up-to-date version of a resource at a given URL. Requests
that are fulfilled with out-of-date responses (like those found in your
browser's HTTP cache) can end up being read from the service worker's cache
indefinitely. Jake Archibald'sblog post
provides more context about this problem.
In the interest of avoiding that scenario,sw-precache
will, by default,
append a cache-busting parameter to the end of each URL it requests when
populating or updating its cache. Developers who are explicitly doing "the right
thing" when it comes to setting HTTP caching headers on their responses might
want to opt out of this cache-busting. For example, if all of your static
resources already include versioning information in their URLs (via a tool like
gulp-rev
), and are served with
long-lived HTTP caching headers, then the extra cache-busting URL parameter
is not needed, and can be safely excluded.
dontCacheBustUrlsMatching
gives you a way of opting-in to skipping the cache
busting behavior for a subset of your URLs (or all of them, if a catch-all value
like/./
is used).
If set, then thepathname
of each URL that's prefetched will be matched against this value.
If there's a match, then the URL will be prefetched as-is, without an additional
cache-busting URL parameter appended.
Note: Prior tosw-precache
v5.0.0,dontCacheBustUrlsMatching
matched against
the entire request URL. As of v5.0.0, it only matches against the URL's
pathname.
Default:not set
Maps a dynamic URL string to an array of all the files that URL's contents
depend on. E.g., if the contents of/pages/home
are generated server-side via
the templateslayout.jade
andhome.jade
,then specify'/pages/home': ['layout.jade', 'home.jade']
.The MD5 hash is used to determine whether
/pages/home
has changed will depend on the hashes of bothlayout.jade
and
home.jade
.
An alternative value for the mapping is supported as well. You can specify
a string or a Buffer instance rather than an array of file names. If you use this option, then the
hash of the string/Buffer will be used to determine whether the URL used as a key has changed.
For example,'/pages/dynamic': dynamicStringValue
could be used if the contents of
/pages/dynamic
changes whenever the string stored indynamicStringValue
changes.
Default:{}
Determines whether thefetch
event handler is included in the generated
service worker code. It is useful to set this tofalse
in development builds,
to ensure that features like live reload still work. Otherwise, the content
would always be served from the service worker cache.
Default:true
sw-precache
finds matching cache entries by doing a comparison with the full request URL. It's
common for sites to support URL query parameters that don't affect the site's content and should
be effectively ignored for the purposes of cache matching. One example is the
utm_
-prefixedparameters used for tracking
campaign performance. By default,sw-precache
will ignorekey=value
whenkey
matchesanyof
the regular expressions provided in this option.
To ignore all parameters, use[/./]
.To take all parameters into account when matching, use[]
.
Default:[/^utm_/]
Writes calls toimportScripts()
to the resulting service worker to import the specified scripts.
Default:[]
Specifies a callback function for logging which resources are being precached and
a precache size. Usefunction() {}
if you'd prefer that nothing is logged.
Within agulp
script, it's recommended that you usegulp-util
and pass ingutil.log
.
Default:console.log
Sets the maximum allowed size for a file in the precache list.
Default:2097152
(2 megabytes)
Sets an HTML document to use as a fallback for URLs not found in thesw-precache
cache. This
fallback URL needs to be cached viastaticFileGlobs
ordynamicUrlToDependencies
otherwise it
won't work.
// via staticFileGlobs
staticFileGlobs:['/shell.html']
navigateFallback:'/shell.html'
// via dynamicUrlToDependencies
dynamicUrlToDependencies:{
'/shell':['/shell.hbs']
},
navigateFallback:'/shell'
This comes in handy when used with a web application that performs client-side URL routing using theHistory API.It allows any arbitrary URL that the client generates to map to a fallback cached HTML entry. This fallback entry ideally should serve as an "application shell" that is able to load the appropriate resources client-side, based on the request URL.
Note:This isnotintended to be used to route failed navigations to a
generic "offline fallback" page. ThenavigateFallback
page is used whether the
browser is online or offline. If you want to implement an "offline fallback",
then using an approach similar tothis example
is more appropriate.
Default:''
Works to limit the effect ofnavigateFallback
,so that the fallback only
applies to requests for URLs with paths that match at least one
RegExp
.
This option is useful if you want to fallback to the cached App Shell for certain specific subsections of your site, but not have that behavior apply to all of your site's URLs.
For example, if you would like to havenavigateFallback
only apply to
navigation requests to URLs whose path begins with/guide/
(e.g.https://example.com/guide/1234
), the following configuration could be
used:
navigateFallback:'/shell',
navigateFallbackWhitelist:[/^\/guide\//]
If set to[]
(the default), the whitelist will be effectively bypassed, and
navigateFallback
will apply to all navigation requests, regardless of URL.
Default:[]
Replaces a specified string at the beginning of path URL's at runtime. Use this
option when you are serving static files from a different directory at runtime
than you are at build time. For example, if your local files are under
dist/app/
but your static asset root is at/public/
,you'd strip 'dist/app/'
and replace it with '/public/'.
Default:''
Configures runtime caching for dynamic content. If you use this option, thesw-toolbox
library configured with the caching strategies you specify will automatically be included in
your generated service worker file.
EachObject
in theArray
needs aurlPattern
,which is either a
RegExp
or a string, following the conventions of thesw-toolbox
library's
routing configuration.Also required is
ahandler
,which should be either a string corresponding to one of the
built-in handlersunder thetoolbox.
namespace, or a function corresponding to your custom
request handler.
Optionally,method
can be added to specify one of thesupported HTTP methods(default:'get'
). There is also
support foroptions
,which corresponds to the same options supported by a
sw-toolbox
handler.
For example, the following defines runtime caching behavior for two different URL patterns. It uses a
different handler for each, and specifies a dedicated cache with maximum size for requests
that match/articles/
:
runtimeCaching:[{
urlPattern:/^https:\/\/example\.com\/api/,
handler:'networkFirst'
},{
urlPattern:/\/articles\//,
handler:'fastest',
options:{
cache:{
maxEntries:10,
name:'articles-cache'
}
}
}]
Thesw-precache
+sw-toolbox
explainerhas
more information about how and why you'd use both libraries together.
Default:[]
Controls whether or not the generated service worker will call
skipWaiting()
inside theinstall
handler.
By default, when there's an update to a previously installed
service worker, then the new service worker delays activation and stays in a
waiting
state until all pages controlled by the old service worker are
unloaded. CallingskipWaiting()
allows a newly registered service worker to
bypass thewaiting
state.
WhenskipWaiting
istrue
,the new service worker'sactivate
handler will
be called immediately, and any out of date cache entries from the previous
service worker will be deleted. Please keep this in mind if you rely on older
cached resources to be available throughout the page's lifetime, because, for
example, youdefer the loading of some resources
until they're needed at runtime.
Default:true
An array of one or more string patterns that will be passed in to
glob
.
All files matching these globs will be automatically precached by the generated service worker.
You'll almost always want to specify something for this.
Default:[]
Removes a specified string from the beginning of path URL's at runtime. Use this
option when there's a discrepancy between a relative path at build time and
the same path at run time. For example, if all your local files are under
dist/app/
and your web root is also atdist/app/
,you'd strip that prefix
from the start of each local file's path in order to get the correct relative
URL.
Default:''
Maps multiple strings to be stripped and replaced from the beginning of URL paths at runtime.
Use this option when you have multiple discrepancies between relative paths at build time and
the same path at run time.
IfstripPrefix
andreplacePrefix
are not equal to''
,they are automatically added to this option.
stripPrefixMulti:{
'www-root/public-precached/':'public/',
'www-root/public/':'public/'
}
Default:{}
The path to the (lo-dash) template used to
generateservice-worker.js
.If you need to add additional functionality to the
generated service worker code, it's recommended that you use the
importScripts
option to include extra JavaScript rather than
using a different template. But if you do need to change the basic generated
service worker code, please make a copy of theoriginal template,
modify it locally, and use this option to point to your template file.
Default:service-worker.tmpl
(in the directory that this module lives in)
Determines whether there's log output for each individual static/dynamic resource that's precached. Even if this is set to false, there will be a final log entry indicating the total size of all precached resources.
Default:false
While it's possible to use thesw-precache
module's API directly within any
JavaScript environment, several wrappers have been developed by members of the
community tailored to specific build environments. They include:
There are also several starter kits or scaffolding projects that incorporate
sw-precache
into their build process, giving you a full service worker out of
the box. The include:
While there are not always ready-to-use wrappers for specific environments, this list contains some recipes to integratesw-precache
in your workflow:
Thanks toSindre Sorhusand Addy Osmanifor their advice and code reviews. Jake Archibaldwas kind enough to review the service worker logic.
Copyright © 2017 Google, Inc.
Licensed under theApache License, Version 2.0(the "License" ); you may not use this file except in compliance with the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.