persistence.js

persistence.js is a asynchronous Javascript object-relational mapper library. It can be used both in the web browser and on the server using node.js. It currently supports 4 types of data stores:

HTML5 WebSQL database, a somewhat controversial part of HTML5 that is supported in Webkit browsers, specifically on mobile devices, including iPhone, Android and Palm's WebOS.
Google Gears, a browser plug-in that adds a number of feature to the browser, including a in-browser database.
MySQL, using the node-mysql, node.js module on the server.
In-memory, as a fallback. Keeps the database in memory and is cleaned upon a page refresh (or server restart).

There is also an experimental support for Qt 4.7 Declarative UI framework (QML) which is an extension to JavaScript.

For browser use, persistence.js has no dependencies on any other frameworks, other than the Google Gears initialization script, in case you want to enable Gears support.

Plug-ins

There are a few persistence.js plug-ins available that add functionality:

persistence.search.js, adds simple full-text search capabilities, see docs/search.md for more information.
persistence.migrations.js, supports data migrations (changes to the database schema), see docs/migrations.md for more information.
persistence.sync.js, supports database synchronization with a remote server, see docs/sync.md for more information.

A Brief Intro to Async Programming

In browsers, Javascript and the web page's rendering engine share a single thread. The result of this is that only one thing can happen at a time. If a database query would be performed synchronously, like in many other programming environments like Java and PHP the browser would freeze from the moment the query was issued until the results came back. Therefore, many APIs in Javascript are defined as asynchronous APIs, which mean that they do not block when an "expensive" computation is performed, but instead provide the call with a function that will be invoked once the result is known. In the meantime, the browser can perform other duties.

For instance, a synchronous database call call would look as follows:

var results = db.query("SELECT * FROM Table");
for(...) { ... }

The execution of the first statement could take half a second, during which the browser doesn't do anything else. By contrast, the asynchronous version looks as follows:

db.query("SELECT * FROM Table", function(results) {
  for(...) { ... }
});

Note that there will be a delay between the db.query call and the result being available and that while the database is processing the query, the execution of the Javascript continues. To make this clear, consider the following program:

db.query("SELECT * FROM Table", function(results) {
  console.log("hello");
});
console.log("world");

Although one could assume this would print "hello", followed by "world", the result will likely be that "world" is printed before "hello", because "hello" is only printed when the results from the query are available. This is a tricky thing about asynchronous programming that a Javascript developer will have to get used to.

Using persistence.js in the browser

Browser support

Modern webkit browsers (Google Chrome and Safari)
Firefox (through Google Gears)
Android browser (tested on 1.6 and 2.1)
iPhone browser (iPhone OS 3+)
Palm WebOS (tested on 1.4.0)

(The following is being worked on:) Internet Explorer is likely not supported (untested) because it lacks __defineGetter__ and __defineSetter__ support, which persistence.js uses heavily. This may change in IE 8.

Setting up

To use persistence.js you need to clone the git repository:

git clone git://github.com/zefhemel/persistencejs.git

To use it you need to copy lib/persistence.js to your web directory, as well as any data stores you want to use. Note that the mysql and websql stores both depend on the sql store. A typical setup requires you to copy at least lib/persistence.js, lib/persistence.store.sql.js and lib/persistence.store.websql.js to your web directory. You can then load them as follows:

<script src="persistence.js" type="application/javascript"></script>
<script src="persistence.store.sql.js" type="application/javascript"></script>
<script src="persistence.store.websql.js" type="application/javascript"></script>

Setup your database

You need to explicitly configure the data store you want to use, configuration of the data store is store-specific. The WebSQL store (which includes Google Gears support) is configured as follows:

persistence.store.websql.config(persistence, 'yourdbname', 'A database description', 5 * 1024 * 1024);

The first argument is always supposed to be persistence. The second in your database name (it will create it if it does not already exist, the third is a description for you database, the last argument is the maximum size of your database in bytes (5MB in this example).

If you're using the in-memory store, you can configure it as follows:

persistence.store.memory.config(persistence);

Schema definition

A data model is declared using persistence.define. The following two definitions define a Task and Category entity with a few simple properties. The property types are based on SQLite types, specifically supported types are (but any SQLite type is supported):

TEXT: for textual data
INT: for numeric values
BOOL: for boolean values (true or false)
DATE: for date/time value (with precision of 1 second)
JSON: a special type that can be used to store arbitrary JSON data. Note that this data can not be used to filter or sort in any sensible way. If internal changes are made to a JSON property, persistence.js may not register them. Therefore, a manual call to anObj.markDirty('jsonPropertyName') is required before calling persistence.flush.

Example use:

var Task = persistence.define('Task', {
  name: "TEXT",
  description: "TEXT",
  done: "BOOL"
});

var Category = persistence.define('Category', {
  name: "TEXT",
  metaData: "JSON"
});

var Tag = persistence.define('Task', {
  name: "TEXT"
});

The returned values are constructor functions and can be used to create new instances of these entities later:

Relationships between entities are defined using the constructor function's hasMany call:

// This defines a one-to-many relationship:
Category.hasMany('tasks', Task, 'category');
// These two definitions define a many-to-many relationship
Task.hasMany('tags', Tag, 'tasks');
Tag.hasMany('tasks', Task, 'tags');

The first statement defines a tasks relationship on category objects containing a QueryCollection (see the section on query collections later) of Tasks, it also defines an inverse relationship on Task objects with the name category. The last two statements define a many-to-many relationships between Task and Tag. Task gets a tags property (a QueryCollection) containing all its tags and vice versa, Tag gets a tasks property containing all of its tasks.

The defined entity definitions are synchronized (activated) with the database using a persistence.schemaSync call, which takes a callback function (with a newly created transaction as an argument), that is called when the schema synchronization has completed, the callback is optional.

persistence.schemaSync();
// or
persistence.schemaSync(function(tx) { 
  // tx is the transaction object of the transaction that was
  // automatically started
});

There is also a migrations plugin you can check out, documentation can be found in persistence.migrations.docs.md file.

Creating and manipulating objects

New objects can be instantiated with the constructor functions. Optionally, an object with initial property values can be passed as well, or the properties may be set later:

var task = new Task();
var category = new Category({name: "My category"});
category.metaData = {rating: 5};
var tag = new Tag();
tag.name = "work";

Many-to-one relationships are accessed using their specified name, e.g.: task.category = category;

One-to-many and many-to-many relationships are access and manipulated through the QueryCollection API that will be discussed later:

task.tags.add(tag);
tasks.tags.remove(tag)l
tasks.tags.list(tx, function(allTags) { console.log(allTags); });

Persisting/removing objects

Similar to hibernate, persistence.js uses a tracking mechanism to determine which objects' changes have to be persisted to the datase. All objects retrieved from the database are automatically tracked for changes. New entities can be tracked to be persisted using the persistence.add function:

var c = new Category({name: "Main category"});
persistence.add(c);
for ( var i = 0; i < 5; i++) {
  var t = new Task();
  t.name = 'Task ' + i;
  t.done = i % 2 == 0;
  t.category = c;
  persistence.add(t);
}

Objects can also be removed from the database:

persistence.remove(c);

All changes made to tracked objects can be flushed to the database by using persistence.flush, which takes a transaction object and callback function as arguments. A new transaction can be started using persistence.transaction:

persistence.transaction(function(tx) {
  persistence.flush(tx, function() {
    alert('Done flushing!');
  });
});

For convenience, it is also possible to not specify a transaction or callback, in that case a new transaction will be started automatically. For instance:

persistence.flush();
// or, with callback
persistence.flush(function() {
  alert('Done flushing');
});

Note that when no callback is defined, the flushing still happens asynchronously.

Important: Changes and new objects will not be persisted until you explicitly call persistence.flush(). The exception to this rule is using the list(...) method on a database QueryCollection, which also flushes first, although this behavior may change in the future.

Dumping and restoring data

persistence.dump can be used to create an object containing a full dump of a database. Naturally, it is adviced to only do this with smaller databases. Example:

persistence.dump(tx, [Task, Category], function(dump) {
  console.log(dump);
});

The tx is left out, a new transaction will be started for the operation. If the second argument is left out, dump defaults to dumping all defined entities.

The dump format is:

{"entity-name": [list of instances],
 ...}

persistence.load is used to restore the dump produced by persistence.dump. Usage:

persistence.load(tx, dumpObj, function() {
  alert('Dump restored!');
});

The tx argument can be left out to automatically start a new transaction. Note that persistence.load does not empty the database first, it simply attempts to add all objects to the database. If objects with, e.g. the same ID already exist, this will fail.

Similarly, persistence.loadFromJson and persistence.dumpToJson respectively load and dump all the database's data as JSON strings.

Entity constructor functions

The constructor function returned by a persistence.define call cannot only be used to instantiate new objects, it also has some useful methods of its own:

EntityName.all([session]) returns a query collection containing all persisted instances of that object. The session argument is optional and only required when persistence.js is used in multi-session mode.
EntityName.load([session], [tx], id, callback) loads an particular object from the database by id or returns null if it has not been found.
EntityName.findBy([session], [tx], property, value, callback) searches for a particular object based on a property value (this is assumed to be unique), the callback function is called with the found object or null if it has not been found.

And of course the methods to define relationships to other entities:

EntityName.hasMany(property, Entity, inverseProperty) defines a 1:N or N:M relationship (depending on the inverse property)
EntityName.hasOne(property, Entity) defines a 1:1 or N:1 relationship

Query collections

A core concept of persistence.js is the QueryCollection. A QueryCollection represents a (sometimes) virtual collection that can be filtered, ordered or paginated. QueryCollections are somewhate inspired by Google AppEngine's Query class. A QueryCollection has the following methods:

filter(property, operator, value)
Returns a new QueryCollection that adds a filter, filtering a certain property based on an operator and value. Supported operators are '=', '!=', '<', '<=', '>', '>=', 'in' and 'not in'. Example: .filter('done', '=', true)
order(property, ascending)
Returns a new QueryCollection that will order its results by the property specified in either an ascending (ascending === true) or descending (ascending === false) order.
limit(n)
Returns a new QueryCollection that limits the size of the result set to n items. Useful for pagination.
skip(n)
Returns a new QueryCollection that skips the first n results. Useful for pagination.
prefetch(rel)
Returns a new QueryCollection that prefetches entities linked through relationship rel, note that this only works for one-to-one and many-to-one relationships.
add(obj)
Adds object obj to the collection.
remove(obj)
Removes object obj from the collection.
list([tx], callback)
Asynchronously fetches the results matching the formulated query. Once retrieved, the callback function is invoked with an array of entity objects as argument.
each([tx], eachCallback)
Asynchronously fetches the results matching the formulated query. Once retrieved, the eachCallback function is invoked on each element of the result objects.
forEach([tx], eachCallback)
Alias for each
one([tx], callback) Asynchronously fetches the first element of the collection, or null if none.
destroyAll([tx], callback) Asynchronously removes all the items in the collection. Important: this does not only remove the items from the collection, but removes the items themselves!
count([tx], callback) Asynchronously counts the number of items in the collection. The arguments passed to the callback function is the number of items.

Query collections are returned by:

EntityName.all(), e.g. Task.all()
one-to-many and many-to-many relationships, e.g. task.tags

Example:

var allTasks = Task.all().filter("done", '=', true).prefetch("category").order("name", false).limit(10);
    
allTasks.list(null, function (results) {
    results.forEach(function (r) {
        console.log(r.name)
        window.task = r;
    });
});

Using persistence.js on the server

Installing persistence.js on node is easy using npm:

npm install persistencejs

Sadly the node.js server environment requires slight changes to persistence.js to make it work with multiple database connections:

A Session object needs to be passed as an extra argument to certain method calls, typically as a first argument.
Methods previously called on the persistence object itself are now called on the Session object.

An example node.js application is included in test/node-blog.js.

Setup

You need to require two modules, the persistence.js library itself and the MySQL backend module.

var persistence = require('persistencejs/persistence').persistence;
var persistenceStore = require('persistencejs/persistence.store.mysql');

Then, you configure the database settings to use:

persistenceStore.config(persistence, 'localhost', 'dbname', 'username', 'password');

Subsequently, for every connection you handle (assuming you're building a sever), you call the persistenceStore.getSession() method:

var session = persistenceBackend.getSession();

This session is what you pass around, typically together with a transaction object. Note that currently you can only have one transaction open per session and transactions cannot be nested.

session.transaction(function(tx) {
  ...
});

Defining your data model

Defining your data model is done in exactly the same way as regular persistence.js:

var Task = persistence.define('Task', {
  name: "TEXT",
  description: "TEXT",
  done: "BOOL"
});

A schemaSync is typically performed as follows:

session.schemaSync(tx, function() {
  ...
});

Creating and manipulating objects

Creating and manipulating objects is done much the same way as with regular persistence.js, except that in the entity's constructor you need to reference the Session again:

var t = new Task(session);
...
session.add(t);

session.flush(tx, function() {
  ...
});

Query collections

Query collections work the same way as in regular persistence.js with the exception of the Entity.all() method that now also requires a Session to be passed to it:

Task.all(session).filter('done', '=', true).list(tx, function(tasks) {
  ...
});

Closing the session

After usage, you need to close your session:

session.close();

Bugs and Contributions

If you find a bug, please report it. or fork the project, fix the problem and send me a pull request. For a list of planned features and open issues, have a look at the issue tracker.

For support and discussion, please join the persistence.js Google Group.

Thanks goes to the people listed in AUTHORS for their contributions.

If you use GWT (the Google Web Toolkit), be sure to have a look at Dennis Z. Jiang's GWT persistence.js wrapper.

License

This work is licensed under the MIT license.

Support this work

You can support this project by flattering it:

mhoffmann/persistencejs

persistence.js

Plug-ins

A Brief Intro to Async Programming

Using persistence.js in the browser

Browser support

Setting up

Setup your database

Schema definition

Creating and manipulating objects

Persisting/removing objects

Dumping and restoring data

Entity constructor functions

Query collections

Using persistence.js on the server

Setup

Defining your data model

Creating and manipulating objects

Query collections

Closing the session

Bugs and Contributions

License

Support this work