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), unless saved to localStorage.
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.
There are a few persistence.js
plug-ins available that add functionality:
persistence.search.js
, adds simple full-text search capabilities, seedocs/search.md
for more information.persistence.migrations.js
, supports data migrations (changes to the database schema), seedocs/migrations.md
for more information.persistence.sync.js
, supports database synchronization with a remote server, seedocs/sync.md
for more information.jquery.persistence.js
, adds jQuery integration, including jQuery-mobile ajax request interception and re-routing to persistencejs, seedocs/jquery.md
for more information anddemo/jquerymobile
for a simple demo.
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.
- Modern webkit browsers (Google Chrome and Safari)
- Firefox (through Google Gears)
- Opera
- Android browser (tested on 1.6 and 2.x)
- iPhone browser (iPhone OS 3+)
- Palm WebOS (tested on 1.4.0)
- Other browsers supporting
localStorage
(e.g. Firefox)
(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 9.
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>
If you want to use the in-memory store (in combination with
localStorage
) you also need the persistence.store.memory.js
included.
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).
The in-memory store is offered as a fallback for browsers that do not support any of the other supported stores (e.g. WebSQL or Gears). In principal, it only keeps data in memory, which means that navigating away from the page (including a reload or tab close) will result in the loss of all data.
A way around this is using the persistence.saveToLocalStorage
and
persistence.loadFromLocalStorage
functions that can save the entire
database to the localStorage, which
is persisted indefinitely (similar to WebSQL).
If you're going to use the in-memory store, you can configure it as follows:
persistence.store.memory.config(persistence);
Then, if desired, current data can be loaded from the localStorage using:
persistence.loadFromLocalStorage(function() {
alert("All data loaded!");
});
And saved using:
persistence.saveToLocalStorage(function() {
alert("All data saved!");
});
Drawbacks of the in-memory store:
- Performance: All actions that are typically performed by a database (sorting, filtering), are now all performed in-memory using Javascript.
- Limited database size: Loading and saving requires serialization of
all data from and to JSON, which gets more expensive as your dataset
grows. Most browsers have a maximum size of 5MB for
localStorage
. - Synchronous behavior: Although the API is asynchronous, all persistence actions will be performed synchronously on the main Javascript thread, which may make the browser less responsive.
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 dataINT
: for numeric valuesBOOL
: for boolean values (true
orfalse
)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 aJSON
property,persistence.js
may not register them. Therefore, a manual call toanObj.markDirty('jsonPropertyName')
is required before callingpersistence.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('Tag', {
name: "TEXT"
});
The returned values are constructor functions and can be used to create new instances of these entities later.
It is possible to create indexes on one or more columns using
EntityName.index
, for instance:
Task.index('done');
Task.index(['done', 'name']);
These indexes can also be used to impose unique constraints :
Task.index(['done', 'name'],{unique:true});
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 Task
s, 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.
You can also define mix-ins and apply them to entities of the model.
A mix-in definition is similar to an entity definition, except using
defineMixin
rather than just define
. For example:
var Annotatable = persistence.defineMixin('Annotatable', {
lastAnnotated: "DATE"
});
You can define relationships between mix-in and entities. For example:
// A normal entity
var Note = persistence.define('Note', {
text: "TEXT"
});
// relationship between a mix-in and a normal entity
Annotatable.hasMany('notes', Note, 'annotated');
Once you have defined a mix-in, you can apply it to any entity of your model,
with the Entity.is(mixin)
method. For example:
Project.is(Annotatable);
Task.is(Annotatable);
Now, your Project
and Task
entities have an additional lastAnnotated
property.
They also have a one to many relationship called notes
to the Note
entity.
And you can also traverse the reverse relationship from a Note
to its annotated
object.
Note that annotated
is a polymorphic relationship as it may yield either a Project
or a Task
(or any other entity which is `Annotatable').
Note: Prefetch is not allowed (yet) on a relationship that targets a mixin. In the example above
you cannot prefetch the annotated
relationship when querying the Note
entity.
Notes: this feature is very experimental at this stage. It needs more testing. Support for "is a" relationships (classical inheritance) is also in the works.
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);
tasks.tags.list(tx, function(allTags) { console.log(allTags); });
Similar to hibernate, persistence.js
uses a tracking mechanism to determine which objects' changes have to
be persisted to the database. 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.
The library supports two kinds of 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.
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. Thesession
argument is optional and only required whenpersistence.js
is used in multi-session mode.EntityName.load([session], [tx], id, callback)
loads an particular object from the database by id or returnsnull
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 ornull
if it has not been found.EntityName.index([col1, col2, ..., colN], options)
creates an index on a column of a combination of columns, for faster searching. If options.unique is true, the index will impose a unique constraint on the values of the columns.
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
Entity instances also have a few predefined properties and methods you should be aware of:
obj.id
, contains the identifier of your entity, this is a automatically generated (approximation of a) UUID. You should never write to this property.obj.fetch(prop, callback)
, if an object has ahasOne
relationship to another which has not yet been fetched from the database (e.g. whenprefetch
wasn't used), you can fetch in manually usingfetch
. When the property object is retrieved the callback function is invoked with the result, the result is also cached in the entity object itself.obj.selectJSON([tx], propertySpec, callback)
, sometime you need to extract a subset of data from an entity. You for instance need to post a JSON representation of your entity, but do not want to include all properties.selectJSON
allows you to do that. ThepropertySpec
arguments expects an array with property names. Some examples:['id', 'name']
, will return an object with the id and name property of this entity['*']
, will return an object with all the properties of this entity, not recursive['project.name']
, will return an object with a project property which has a name property containing the project name (hasOne relationship)['project.[id, name]']
, will return an object with a project property which has an id and name property containing the project name (hasOne relationship)['tags.name']
, will return an object with an arraytags
property containing objects each with a single property: name
A core concept of persistence.js
is the QueryCollection
. A
QueryCollection
represents a (sometimes) virtual collection that can
be filtered, ordered or paginated. QueryCollection
s are somewhate
inspired by Google AppEngine's Query
class.
A QueryCollection
has the following methods:
filter(property, operator, value)
Returns a newQueryCollection
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)
or(filter)
Returns a newQueryCollection
that contains items either matching the filters specified before callingor
, or the filter represented in the argument. Thefilter
argument is of aFilter
type, there are three types of filters:persistence.PropertyFilter
, which filters on properties (internally called whenfilter(...)
is used.
Example:new persistence.PropertyFilter('done', '=', true)
persistence.AndFilter
, which is passed two filter objects as arguments, both of which should be true. Example:new persistence.AndFilter(new persistence.PropertyFilter('done', '=', true), new persistence.PropertyFilter('archived', '=', true))
persistence.OrFilter
, which is passed two filter objects as arguments, one of which should be true. Example:new persistence.OrFilter(new persistence.PropertyFilter('done', '=', true), new persistence.PropertyFilter('archived', '=', true))
and(filter)
same asor(filter)
except that both conditions should hold for items to be in the collection.order(property, ascending)
Returns a newQueryCollection
that will order its results by the property specified in either an ascending (ascending === true) or descending (ascending === false) order.limit(n)
Returns a newQueryCollection
that limits the size of the result set ton
items. Useful for pagination.skip(n)
Returns a newQueryCollection
that skips the firstn
results. Useful for pagination.prefetch(rel)
Returns a newQueryCollection
that prefetches entities linked through relationshiprel
, note that this only works for one-to-one and many-to-one relationships.add(obj)
Adds objectobj
to the collection.remove(obj)
Removes objectobj
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, theeachCallback
function is invoked on each element of the result objects.forEach([tx], eachCallback)
Alias foreach
one([tx], callback)
Asynchronously fetches the first element of the collection, ornull
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 thecallback
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;
});
});
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 theSession
object.
An example node.js
application is included in test/node-blog.js
.
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', 3306, 'dbname', 'username', 'password');
Subsequently, for every connection you handle (assuming you're
building a sever), you call the persistenceStore.getSession()
method:
var session = persistenceStore.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) {
...
});
persistence.js
works in autocommit mode by default.
You can override this behavior and enable explicit commit and rollback
by passing true as first argument to persistence.transaction
.
You can then use the following two methods to control the transaction:
transaction.commit(session, callback)
commits the changes.transaction.rollback(session, callback)
rollbacks the changes.
Typical code will look like:
session.transaction(true, function(tx) {
// create/update/delete objects
modifyThings(session, tx, function(err, result) {
if (err) {
// something went wrong
tx.rollback(session, function() {
console.log('changes have been rolled back: ' + ex.message);
});
}
else {
// success
tx.commit(session, function() {
console.log('changes have been committed: ' result);
});
});
});
Explicit commit and rollback is only supported on MySQL (server side) for now.
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 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 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) {
...
});
After usage, you need to close your session:
session.close();
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
This work is licensed under the MIT license.
You can support this project by flattering it: