Client-side tools for loading a dump from a CouchDB/PouchDB database.
For dumping, check out pouchdb-dump-cli to dump from the command line, or pouchdb-replication-stream to dump from within your Node.js application.
This method is typically much faster than standard replication, because it uses fewer HTTP requests. So it's a great way to quickly load an initial state for your database.
To use this plugin, include it after pouchdb.js
in your HTML page:
<script src="pouchdb.js"></script>
<script src="pouchdb.load.js"></script>
Or install from Bower:
bower install pouchdb-load
Or to use it in Node.js, just npm install it:
npm install pouchdb-load
And then attach it to the PouchDB
object:
var PouchDB = require('pouchdb');
PouchDB.plugin(require('pouchdb-load'));
This plugin exposes a single method on your database, load()
:
This method returns a Promise or calls your callback, if you prefer the callback style.
You can give it a URL pointing to a single dump file:
var db = new PouchDB('my-awesome-db');
db.load('http://example.com/my-dump-file.txt').then(function () {
// done loading!
}).catch(function (err) {
// HTTP error or something like that
});
This will read the entire file into memory, though. Assuming you used the --split
option when you dumped your database, you can also load multiple files by using Promise.all
. For instance, let's say you had 5 files, named
'my-dump-file_00000000.txt'
through 'my-dump-file_00000004.txt'
. You would do:
var dumpFiles = [
'my-dump-file_00000000.txt',
'my-dump-file_00000001.txt',
'my-dump-file_00000002.txt',
'my-dump-file_00000003.txt',
'my-dump-file_00000004.txt',
];
PouchDB.utils.Promise.all(dumpFiles.map(function (dumpFile) {
return db.load('http://example.com/' + dumpFile);
})).then(function () {
// done loading!
}).catch(function (err) {
// HTTP error or something like that
});
This will load them all simultaneously. You can also load them all in a series:
var series = PouchDB.utils.Promise.resolve();
dumpFiles.forEach(function (dumpFile) {
series = series.then(function () {
return db.load('http://example.com/' + dumpFile);
});
});
series.then(function () {
// done loading!
}).catch(function (err) {
// HTTP error or something like that
});
Instead of a URL, you can also load directly from a string. This is useful if you used pouchdb-replication-stream
to dump directly to a string, or if you are loading your dumpfile through some other mechanism than ajax (websockets, WebRTC, etc.):
var db = new PouchDB('my-awesome-db');
var myDumpedString = getDumpedStringSomehow();
db.load(myDumpedString).then(function () {
// done loading!
}).catch(function (err) {
// any possible errors
});
Normally the load()
operation doesn't write any checkpoints, meaning that if you switch from load()
to normal replication, then it will start reading all the changes from the remote CouchDB from the beginning of time. This is slow, so to avoid it, use the proxy
option:
db.load('http://example.com/my-dump-file.txt', {
proxy: 'http://mysite.com/mydb'
}).then(function () {
// done loading! handoff to regular replication
return db.replicate.from('http://mysite.com/mydb');
}).catch(function (err) {
// HTTP error or something like that
});
This will tell the plugin that the dumpfile 'http://example.com/my-dump-file.txt'
is just a proxy for 'http://mysite.com/mydb'
. So when you pick up replication again, it won't start from 0 but rather will start from the last checkpoint reported by the dump file.
If your replication also involves a filter
function, you should pass that in as filter
as well (so that the correct checkpoint can be written):
function filterFun(doc) {
/* your cool filter function here */
}
db.load('http://example.com/my-dump-file.txt', {
proxy: 'http://mysite.com/mydb',
filter: filterFun
}).then(function () {
// done loading! handoff to regular replication
return db.replicate.from('http://mysite.com/mydb', {filter: filterFun});
}).catch(function (err) {
// HTTP error or something like that
});
The same goes for view
and query_params
.
You can also include ajax options in the options
:
db.load('myfile.txt', {
ajax: {
timeout: 30000
}
});
The ajax options themselves are described in the PouchDB documentation.
NPM Browser uses pouchdb-load to load a bunch of static files from Amazon S3, which is how it's able to replicate all of NPM so quickly. Here is the relevant code, which does the dump, checkpointing, and handoff to regular replication (as described above).
The load()
operation is idempotent, meaning that you can run it over and over again, and it won't create duplicate documents in the target database.
However, it's inefficient to run the load()
every time the user starts your app. So if you'd like, you can use "local documents" to remember whether or not this database has already been loaded:
db.get('_local/initial_load_complete').catch(function (err) {
if (err.status !== 404) { // 404 means not found
throw err;
}
db.load(/* ... */).then(function () {
return db.put({_id: '_local/initial_load_complete'});
});
}).then(function () {
// at this point, we are sure that
// initial replication is complete
}).catch(function (err) {
// handle unexpected errors
});
This code first checks for a local document called '_local/initial_load_complete'
. If the document is not found, then it calls dump()
, then puts the local doc to mark that it's complete. Else it finishes.
(Local documents are non-replicated PouchDB/CouchDB documents that are useful for storing local state or configuration files. To create a local document, you simply prefix '_local/'
to the document _id
.)
npm install
npm run build
This will run the tests in Node using LevelDB:
npm test
You can also check for 100% code coverage using:
npm run coverage
npm run test-local
To test in your own browser, or
npm run test-browser
To test in PhantomJS.