worker-pouch
var db = new PouchDB('mydb', {adapter: 'worker'});
Adapter plugin to use PouchDB over web workers. Transparently proxies all PouchDB API requests to a web worker, so that the most expensive database operations are run in a separate thread. Supports Firefox and Chrome.
Basically, worker-pouch allows you use the PouchDB API like you normally would, but your UI will suffer fewer hiccups, because any blocking operations (such as IndexedDB or checksumming) are run inside of the worker. You don't even need to set up the worker yourself, because the script is loaded in a Blob URL.
The worker-pouch adapter passes the full PouchDB test suite. It requires PouchDB 5.0.0+.
IE, Edge, Safari, and iOS are not supported due to browser bugs. Luckily, Firefox and Chrome are the browsers that benefit the most from web workers. There is also an API to detect browser support.
Usage
$ npm install worker-pouch
When you npm install worker-pouch
, the client JS file is available at node_modules/worker-pouch/dist/pouchdb.worker-pouch.js
. Or you can just download it from Github above (in which case, it will be available as window.workerPouch
).
Then include it in your HTML, after PouchDB:
<script src="pouchdb.js"></script>
<script src="pouchdb.worker-pouch.js"></script>
Then you can create a worker-powered PouchDB using:
var db = new PouchDB('mydb', {adapter: 'worker'});
In Browserify
The same rules apply, but you have to notify PouchDB of the new adapter:
var PouchDB = require('pouchdb');
PouchDB.adapter('worker', require('worker-pouch'));
Performance benefits
These numbers were recorded using this site. The test involved inserting 10000 PouchDB documents, and was run on a 2013 MacBook Air. Browser data was deleted between each test.
| Time (ms) | Blocked the DOM? |
---|
Chrome 48 | | |
put() - normal | 50070 | No |
put() - worker | 56993 | No |
bulkDocs() - normal | 2740 | Yes |
bulkDocs() - worker | 3454 | No |
Firefox 43 | | |
put() - normal | 39595 | No |
put() - worker | 41425 | No |
bulkDocs() - normal | 1027 | Yes |
bulkDocs() - worker | 1130 | No |
Basic takeaway: put()
s avoid DOM-blocking (due to using many smaller transactions), but are much slower than bulkDocs()
. With worker-pouch, though, you can get nearly all the speed benefit of bulkDocs()
without blocking the DOM.
(Note that by "blocked the DOM," I mean froze the animated GIF for a significant amount of time - at least a half-second. A single dropped frame was not penalized. Try the test yourself, and you'll see the difference is pretty stark.)
Detecting browser support
This plugin doesn't support all browsers. So it provides a special API to dianogose whether or not the current browser supports worker-pouch. Here's how you can use it:
var workerPouch = require('worker-pouch');
workerPouch.isSupportedBrowser().then(function (supported) {
var db;
if (supported) {
db = new PouchDB('mydb', {adapter: 'worker'});
} else {
db = new PouchDB('mydb');
}
}).catch(console.log.bind(console));
The isSupportedBrowser()
API returns a Promise for a boolean, which will be true
if the browser is supported and false
otherwise.
If you are using this method to return the PouchDB object itself from a Promise, be sure to wrap it in an object, to avoid "circular promise" errors:
var workerPouch = require('worker-pouch');
workerPouch.isSupportedBrowser().then(function (supported) {
if (supported) {
return {db: new PouchDB('mydb', {adapter: 'worker'})};
} else {
return {db: new PouchDB('mydb')};
}
}).then(function (dbWrapper) {
var db = dbWrapper.db;
}).catch(console.log.bind(console));
Debugging
worker-pouch uses debug for logging. So in the browser, you can enable debugging by using PouchDB's logger:
PouchDB.debug.enable('pouchdb:worker:*');
Q & A
Wait, doesn't PouchDB already work in a web worker?
Yes, you can use pure PouchDB inside of a web worker. But the point of this plugin is to let you use PouchDB from outside a web worker, and then have it transparently proxy to another PouchDB that is isolated in a web worker.
What browsers are supported?
Only those browsers that 1) allow blob URLs for web worker scripts, and 2) allow IndexedDB inside of a web worker. Today, that means Chrome and Firefox.
Can I use it with other plugins?
Not right now, although map/reduce is supported.
Don't I pay a heavy cost of structured cloning due to worker messages?
Yes, but apparently this cost is less than that of IndexedDB, because the DOM is significanty less blocked when using worker-pouch. Another thing to keep in mind is that PouchDB's internal document representation in IndexedDB is more complex than the PouchDB documents you insert. So you clone a small PouchDB object to send it to the worker, and then inside the worker it's exploded into a more complex IndexedDB object. IndexedDB itself has to clone as well, but the more complex cloning is done inside the worker.
Does replication occur inside the worker?
It's a bit subtle. The answer is yes, if you do this:
var local = new PouchDB('local', {adapter: 'worker'});
local.replicate.to('http://example.com/db');
However, the answer is no if you do:
var local = new PouchDB('local', {adapter: 'worker'});
var remote = new PouchDB('http://example.com/db');
local.replicate.to(remote);
The reason is that when you create a remote PouchDB using new PouchDB('http://example.com/db')
, then that runs inside the UI thread. However, when you .replicate.to('http://example.com/db')
, then that string is passed ver-batim to the worker thread, where worker-pouch
becomes responsible for creating the remote PouchDB. Hence replication will occur inside of the worker thread.
In general, if you are very concerned about performance implications of what runs inside of the woker vs what runs outside of the worker, you are encouraged to not use worker-pouch
and to instead just run PouchDB inside a worker and handle message-passing yourself (might I recommend promise-worker?). This is the only way to really ensure that all PouchDB operations are isolated to the worker.
Building
npm install
npm run build
Your plugin is now located at dist/pouchdb.worker-pouch.js
and dist/pouchdb.worker-pouch.min.js
and is ready for distribution.
Testing
In the browser
Run npm run dev
and then point your favorite browser to http://127.0.0.1:8000/test/index.html.
The query param ?grep=mysearch
will search for tests matching mysearch
.
Automated browser tests
You can run e.g.
CLIENT=selenium:firefox npm test
CLIENT=selenium:phantomjs npm test
This will run the tests automatically and the process will exit with a 0 or a 1 when it's done. Firefox uses IndexedDB, and PhantomJS uses WebSQL.