maxGraph/docs/manual.html

1724 lines
74 KiB
HTML

<html>
<head>
<title>mxGraph User Manual - JavaScript Client</title>
<link rel="stylesheet" href="css/manual-styles.css">
<link rel="stylesheet" href="css/manual-colors.css">
<script type="text/javascript" src="js/toc.js"></script>
</head>
<body onload="maketoc(document.getElementById('toc'));">
<h1>mxGraph User Manual &ndash; JavaScript Client</h1>
<br/>
<br/>
<p>mxGraph Version 3.3.0.1 &ndash; 19. May 2015</p>
<p>Copyright (c) JGraph Ltd 2006-2013</p>
<br/>
<p>All rights reserved. No part of this publication may be
reproduced, stored in a retrieval system, or transmitted in any form or
by any means, electronic, mechanical, photocopying, recording or
otherwise, without the prior written permission of the author.</p>
<p>The programs in this book have been included for their
instructional value. They have been tested with care but are not
guaranteed for any particular purpose. The publisher does not offer any
warranties or representations nor does it accept any liabilities with
respect to the programs.</p>
<p>Possession, use, or copying of the software described in this
publication is authorized only pursuant to a valid written license from
JGraph Ltd.</p>
<p>Neither JGraph Ltd. nor its employees are responsible for any
errors that may appear in this publication. The information in this
publication is subject to change without notice.</p>
<p>Java and all Java-based marks are trademarks or registered
trademarks of Sun Microsystems, Inc. in the U.S. and other countries.</p>
<br/>
<h1>Table Of Contents</h1>
<div id="toc"></div>
<br/>
<h1><a name="Introduction"></a>Introduction</h1>
<h2><a name="mxGraph_Products"></a>Product Introduction</h2>
<p>mxGraph is a JavaScript component that provides features aimed at applications that display
interactive <a href="http://en.wikipedia.org/wiki/Diagram">diagrams</a>
and graphs. Note by graphs we mean <a
href="http://en.wikipedia.org/wiki/Graph_(mathematics)">mathematical
graphs</a>, not necessarily <a href="http://en.wikipedia.org/wiki/Charts">charts</a>
(although some charts are graphs). See later section &ldquo;What is a
Graph?&rdquo; for more details.</p>
<p>Being a developer library, mxGraph is not designed specifically
to provide a ready to use application, although many of the examples are
close to being usable applications. mxGraph provides all the commonly
required functionality to draw, interact with and associate a context
with a diagram. mxGraph comes with a number of examples that help explain how a
basic application is put together and showcases
individual features of the library.</p>
<p>Developers integrating the library in their application should
read the section
&ldquo;Pre-requisites&rdquo; below. Given that mxGraph is a component
part of your application, you must understand how JavaScript web
applications are constructed at an architectural level, and how to program
both in JavaScript, as well as any server-side languages used.</p>
<p>mxGraph mainly comprises one JavaScript
file that contains all of the mxGraph functionality. This is loaded into
a HTML web page in a JavaScript section and executes in an HTML
container in the browser. This is an incredibly simple architecture that
only requires a web server capable of serving html pages and a
JavaScript enabled web browser.</p>
<p>The key advantages of this technology are:</p>
<ul>
<li>That no third-party plug-ins are required. This removes
plug-in vendor dependence.</li>
<li>The technologies involved are open and there are many open
implementations, no one vendor can remove a product or technology that
leaves your application unworkable in practise.</li>
<li>Standardized technologies, meaning your application is
deployable to the maximum number of browser users without need for
additional configuration or installation at the client computer. Large
corporate environments often dislike allowing individuals to install
browser plug-ins and do not like to change the standard build rolled
out to all machines.</li>
</ul>
<p><img src="images/mx_man_architecture.png" name="mxgraph_architecture" />
<br/>
<em>The mxGraph components and their relationships</em></p>
<h2><a name="mxgraph_applications"></a>What Applications can
mxGraph be used for?</h2>
<p>Example applications for a graph visualization library include:
process diagrams, workflow and BPM visualization, flowcharts, traffic or
water flow, database and WWW visualization, networks and
telecommunications displays, mapping applications and GIS, UML diagrams,
electronic circuits, VLSI, CAD, financial and social networks, data
mining, biochemistry, ecological cycles, entity and cause-effect
relationships and organisational charts.</p>
<h2><a name="mxgraph_deployment"></a>How is mxGraph deployed?</h2>
<p>In the typical thin-client environment, mxGraph is split into the
client-side JavaScript library and a server-side library in one of the
two supported languages, .NET and Java. The JavaScript library is
contained as part of a larger web application that is delivered to the
browser using a standard web server. All the browser needs is the
ability to run JavaScript to be enabled.</p>
<p>In the third part of this manual, you will see an example of an
html page that embeds the mxGraph library, as well as a simple
application to invoke the library's functionality.</p>
<h2><a name="mxgraph_technologies"></a>mxGraph Technologies</h2>
<p>mxGraph uses JavaScript for the client-side functionality on the
browser. The JavaScript code in turn uses the underlying vector graphics
language on the active browser to render the displayed diagram, SVG for
the more standards compliant browsers and VML in the case of Microsoft
Internet Explorer. mxGraph also includes the feature to render entirely
using html, this limits the range of functionality available, but is
suitable for more simple diagrams.</p>
<p>As a developer you are not exposed to browser specific features.
As mentioned, the vector graphics language varies by browser, so mxGraph
abstracts their features into a common class. Similarly, for event
handling and DOMs. Browsers differ in their implementation of these two
major browser functionalities, mxGraph exposes a constant API over all
browsers and adapts to the inconsistencies behind the scenes.</p>
<h2><a name="mxgraph_licensing"></a>mxGraph Licensing</h2>
<p>The JavaScript client of mxGraph is licensed under a
<a href="http://www.jgraph.com/mxlicense.html"> standard
commercial license</a> For detailed licensing questions you are always advised to
consult a legal professional.
<h2><a name="what_graph"></a>What is a Graph?</h2>
<p>Graph visualization is based on the mathematical theory of
networks, graph theory. If you're seeking JavaScript bar <em>charts</em>,
pie <em>charts</em>, Gantt <em>charts</em>, have a look at the <a
href="http://code.google.com/apis/chart/">Google Charts</a> project
instead, or similar</p>
<p>A graph consists of vertices, also called nodes, and of edges
(the connecting lines between the nodes). Exactly how a graph appears
visually is not defined in graph theory. The term <em>cell</em> will be
used throughout this manual to describe an element of a graph, either
edges, vertices or groups.</p>
<p><img src="images/mx_man_simple_graph.png" /><br/>
<em>A simple Graph</em></p>
<br/>
<p>There are additional definitions in graph theory that provide
useful background when dealing with graphs, they are listed in the
Appendices if of interest to you.</p>
<h3><a name="graph_visualization"></a>Graph Visualization</h3>
<p>Visualization is the process of creating a useful visual
representation of a graph. The scope of visualization functionality is
one of mxGraphs' main strength. mxGraph supports a wide range of
features to enable the display of cells to only be limited by the skill
of the developer and the platform functionality available. Vertices may
be shapes, images, vector drawing, animations, virtually any graphical
operations available in browsers. You can also use HTML mark-up in both
vertices and edges.</p>
<p><img src="images/mx_man_graph_vis.png"><br/>
<em>Graph Visualization of a transport system. (c) Tourizm Maps
2003, http://www.world-maps.co.uk</em></p>
<br/>
<h3><a name="graph_interaction"></a>Graph Interaction</h3>
<p>Interaction is the way in which an application using mxGraph can
alter the graph model through the web application GUI. mxGraph supports
dragging and cloning cells, re-sizing and re-shaping, connecting and
disconnecting, drag and dropping from external sources, editing cell
labels in-place and more. One of the key benefits of mxGraph is the
flexibility of how interaction can be programmed.</p>
<p>Many complex graphical web applications rely on a round-trip to
the server in order to form the display, not only the base display but
also the interaction events. Although this is often given the title of
AJAX functionality, such server reliance is not appropriate for
interaction events. Visual feedback taking longer than about 0.2 seconds
in an application generally seriously impacts the usability. By placing
all of the interaction on the client, mxGraph provides the true feel of
a application, rather than seeming like a dumb remote terminal. It also
allows the possibility of off-line use.</p>
<p><img src="images/mx_man_graph_interaction.png" /><br/>
<em>Selection shading while selecting an area through mouse drag</em></p>
<br/>
<h3><a name="graph_layouts"></a>Graph Layouts</h3>
<p>Graph cells can be drawn anywhere in a simple application,
including on top of one another. Certain applications need to present
their information in a generally ordered, or specifically ordered
structure. This might involve ensuring cells do not overlap and stay at
least a certain distance from one another, or that cells appear in
specific positions relative to other cells, usually the cells they are
connected to by edges. This activity, called the layout application, can
be used in a number of ways to assist users in setting out their graph.
For non-editable graphs, layout application is the process of applying a
layout algorithm to the cells. For interactive graphs, meaning those
that can be edited through the UI, layout application might involve only
allowing users to make changes to certain cells in certain positions, to
re-apply the layout algorithm after each change to the graph, or to
apply the layout when editing is complete.</p>
<p><img src="images/mx_man_graph_layout.png" /><br/>
<em>Layout of a workflow using a horizontal hierarchical layout</em></p>
<p>mxGraph supports a range of tree, force-directed and hierarchical
layouts which will fit most layout needs. See the later section on using
the layouts for more information.</p>
<p>In a client-server architecture there are two options for how
layouts can be run. The Javascript versions provide the ability to run
the layouting entirely on the client, while the same layout
implementation in Java on the server-side enables the option to offload
some processing to the server, if required.</p>
<h3><a name="graph_analysis"></a>Graph Analysis</h3>
<p>Analysis of graphs involves the application of algorithms
determining certain details about the graph structure, for example,
determining all routes or the shortest path between two cells. There are
more complex graph analysis algorithms, these being often applied in
domain specific tasks. Techniques such as clustering, decomposition, and
optimization tend to be targeted at certain fields of science and have
not been implemented in the core mxGraph packages at the current time of
writing.</p>
<p><img src="images/mx_man_graph_analysis.jpg"
name="shortest_path_analyis" /><br/>
<em>Shortest Path Analysis</em></p>
<br/>
<h2><a name="about_manual"></a>About this Manual</h2>
<h3><a name="pre_requisites"></a>Pre-requisites for mxGraph</h3>
<p>To benefit fully from this manual you will need to have a
reasonable understanding of web applications and of the server
technology you wish to deploy using. Deployment examples are available
for each of the server technologies supported, some familiarity with
that server technology is obviously required.</p>
<p>Basic XML knowledge is useful for changing the editor
configuration files that describe the visual and behavioural aspects of
the editor. You will need to understand and implement Javascript coding
and be familiar with object orientated programming principles and modern
software design.</p>
<p>You do not need knowledge of the underlying vector graphics
language that the browser uses, such as SVG, VML or HTML canvas. mxGraph
abstracts the description of the visual component into one API.</p>
<h1><a name="Getting_Started"></a>Getting Started</h1>
<h2><a name="mxgraph_site"></a>The mxGraph Package</h2>
<h3><a name="obtaining_mxgraph"></a>Obtaining mxGraph</h3>
<p>To evaluate mxGraph:</p>
<ul>
<li>Navigate to the <a
href="http://www.jgraph.com/mxdownload.html">evaluation download
request page</a>.</li>
<li>Please note that we require a commercial or organizational email
address to process your request. We're not able to sell mxGraph to
individuals.</li>
<li>When you receive the download details, unzip to your preferred
location.</li>
<li>Within the package under javascript/examples you will find a
number of html files that demonstrate how to invoke the library within
a web page.</li>
<li>To create your own example, copy the basic structure of the
examples and add your extensions/changes to the onload invoked
JavaScript main function.</li>
<li>The evaluation copy is valid for 90 days after download. If
you wish to continue evaluation please use the <a
href="http://jgraph.com/contact.html">contact form</a>.</li>
<li>For commercial evaluations looking for commercial grade
evaluation support, again, please use the <a
href="http://jgraph.com/contact.html">contact form</a>.</li>
</ul>
<p>There is no functional API difference between the evaluation and full
version of mxGraph.</p>
<h3><a name="Installing_the_JGraph_Suite"></a>Installing mxGraph</h3>
<p>Both the evaluation and full versions of mxGraph are delivered as
zip files. Unzip the package to your preferred location, a folder named
mxGraph will be created there, this folder is the root folder of the
mxGraph installation.</p>
<h3><a name="project_structure"></a>Project structure and build
options</h3>
<p>Once unzipped you will be presented with a number of files and
directories in the installation root.</p>
<TABLE WIDTH=642 BORDER=1 BORDERCOLOR="#000000" CELLPADDING=4
CELLSPACING=0>
<COL WIDTH=165>
<COL WIDTH=459>
<THEAD>
<TR VALIGN=TOP>
<TD WIDTH=165>
<p>/doc</p>
</TD>
<TD WIDTH=459>
<p>Documentation root, includes this user manual</p>
</TD>
</TR>
</THEAD>
<TBODY>
<TR VALIGN=TOP>
<TD WIDTH=165>
<p>/dotnet</p>
</TD>
<TD WIDTH=459>
<p>.NET server-side classes</p>
</TD>
</TR>
<TR VALIGN=TOP>
<TD WIDTH=165>
<p>/java</p>
</TD>
<TD WIDTH=459>
<p>Java server-side classes</p>
</TD>
</TR>
<TR VALIGN=TOP>
<TD WIDTH=165>
<p>/javascript</p>
</TD>
<TD WIDTH=459>
<p>JavaScript client functionality.</p>
</TD>
</TR>
<TR VALIGN=TOP>
<TD WIDTH=165>
<p>/javascript/examples</p>
</TD>
<TD WIDTH=459>
<p>HTML examples demonstrating the use of mxGraph</p>
</TD>
</TR>
<TR VALIGN=TOP>
<TD WIDTH=165>
<p>ChangeLog</p>
</TD>
<TD WIDTH=459>
<p>Details of the changes between releases</p>
</TD>
</TR>
<TR VALIGN=TOP>
<TD WIDTH=165>
<p>index.html</p>
</TD>
<TD WIDTH=459>
<p>Basic introduction to the library</p>
</TD>
</TR>
<TR VALIGN=TOP>
<TD WIDTH=165>
<p>license.txt</p>
</TD>
<TD WIDTH=459>
<p>The licensing terms under which you must use the library</p>
</TD>
</TR>
</TBODY>
</TABLE>
<p><em>Table: Project Directory Structure</em></p>
<br/>
<h2><a name="web_applications"></a>JavaScript and Web Applications</h2>
<p>Web applications, specifically the use of JavaScript to attempt
to emulate desktop application-like behaviour in web browsers, is still
a relatively new field of software engineering. There are three main
issues with JavaScript that are perceived to be a barrier to producing
high quality applications, performance, lack of native functionality
available in desktop applications and inconsistent APIs between
browsers.</p>
<p>There has been considerable effort toward developing framework
libraries to solve two of the problems, the functionality and API
issues. The requirements of many of these libraries is driven by both
improving web site design and usability, as well as to assist production
of what we generally refer to as application features (menus, windows,
dialogs, persistence, event handling, etc). They also provide certain
base functionalities missing in JavaScript that desktop application
developers take for granted, such as basic maths and collections
functionality.</p>
<p>Many of these JavaScript frameworks have IDE support for
development nowadays and all of the major browsers now contain
JavaScript debuggers, either natively or as a plug-in. There is no
compilation phase with JavaScript (it is an interpreted language) so
basic typographical errors are often only caught at runtime, unless you
obtain a syntax checking tool in your IDE. So although there is not one
complete package for your JavaScript development needs, there are a
number of vendors providing the individual components you need to
produce JavaScript applications effectively.</p>
<h3><a name="ajax"></a>AJAX</h3>
<p><a href="http://en.wikipedia.org/wiki/AJAX">AJAX</a> is a very
fashionable term applied freely to make JavaScript applications sound
more modern. The idea of AJAX was originally to enable communication
with a server without leaving the client browser hanging in a loop
waiting for the response. Strictly speaking, the basic functioning of
mxGraph involves no AJAX. All of the interaction and model changes are
applied on the client in JavaScript.</p>
<p>Alternative approaches to display interaction diagrams have been
taken that use the AJAX approach, but be weary of the AJAX term to suggest
that the solution is technological more advanced. Going back to the
server for interaction changes, in particular, cause serious delays in
updating the UI, often making the application unusable.</p>
<h3><a name="js_frameworks"></a>Third-Party JavaScript Frameworks</h3>
<h4><a name="GWT"></a>Google Web Toolkit</h4>
<p>Given the apparent complexity from a desktop developer's point of
view, a common approach is to use the <a
href="http://code.google.com/webtoolkit/">Google Web Toolkit (GWT)</a>.
GWT provides many commonly required web application features by
translating Java into JavaScript. This JavaScript can be deployed just
as natively written JavaScript is. GWT also provides the option to
obfuscate the resulting JavaScript. (See later section on obfuscation in
this chapter).</p>
<p>The key advantages of GWT are:</p>
<ul>
<li>Reuses Java developer knowledge,</li>
<li>Allows debugger of the Java source,</li>
<li>Resolves cross-browser issues,</li>
<li>Native code can be interleaved with Java,</li>
<li>Certain compile-time errors can be detected before execution.</li>
</ul>
<p>The key disadvantages of GWT are:</p>
<ul>
<li>Commercial support is not available from the developers,</li>
<li>The higher the level of abstraction, the more difficult it is
to implement features that go against the architecture and to debug
when things go wrong,</li>
<li>It's use may deter from learning and understanding JavaScript,
which is ultimately necessary to implement certain new features.</li>
</ul>
<p>In summary, consider GWT for simpler projects, but be careful of
placing to much emphasis on it as a tool that avoids having to learn
JavaScript properly.</p>
<h4><a name="native_js_frameworks"></a>Native JavaScript Frameworks
and Libraries</h4>
<p>Rather than list and compare every JavaScript framework, please
see the wikipedia entries for <a
href="http://en.wikipedia.org/wiki/List_of_JavaScript_libraries#JavaScript">web
application frameworks</a> and the <a
href="http://en.wikipedia.org/wiki/Comparison_of_JavaScript_frameworks">comparison
of JavaScript</a>. The comparison should not be considered authoritative,
more so it illustrates the types of features provided, such as event
handling, animation, widgets, AJAX request support etc. <a
href="http://javascriptlibraries.com/">This site</a> is also a useful
list of JavaScript libraries, mostly being open/free source licensed.</p>
<p>Be aware that many frameworks add implicit behaviours to make
JavaScript appear more like an OO language and to increase the base
functionality of the language. During the writing of the layout portion
of mxGraph, it was found that this implicit behaviour broke an example
in a very hard to debug manner. Be aware that this may cause
problems and if you select a framework ensure you understand which
implicit behaviours it introduces.</p>
<p>When selecting a framework and/or libraries think about which
frameworks tie you into certain functional behaviour and look for
libraries that provide features such as animation as distinct,
independent blocks, that you can use without being tied into the overall
design.</p>
<h4><a name="integration_js_frameworks"></a>Integration of mxGraph
and JavaScript frameworks</h4>
<p>This area is often misunderstood, put simply, there is no <em>integration</em>
required. Web applications generally comprise one or more <a
href="http://en.wikipedia.org/wiki/Span_and_div"><em>div</em></a>
elements into which the HTML wrapping the JavaScript of the application
is placed. If you create a div as a container for an mxGraph, that area
is a stand-alone display for the mxGraph application. It can communicate
itself with any back-end server, but there is no interdependence between
that div and the rest of the page, other than the area each take up.
This includes event handling, mxGraph can handle the events for its
container, even if the rest of the web page used a completely different
event model. As long as neither mxGraph nor the other libraries and
frameworks on the page introduce implicit behaviours that break one part
of the page, the issue of client integration is not something that needs
analysis.</p>
<p>Integration of the mxGraph back-end functionality, that which
sits at the server-side is the subject of a later chapter.</p>
<h4><a name="extending_mxgraph"></a>Extending mxGraph in JavaScript</h4>
<p>In JavaScript, there are various ways of mapping the Object
Oriented paradigm to language constructs. mxGraph uses a particular
scheme throughout the project, with the following implicit rules:</p>
<ul>
<li>Do not change the built-in prototypes</li>
<li>Donot try to limit the power of the JavaScript language.</li>
</ul>
<p>There are two types of &ldquo;classes&rdquo; in mxGraph; <EM>classes</em>
and <EM>singletons</em> (where only one instance of the class exists).
Singletons are mapped to global objects where the variable name is the
same as the class name. For example, mxConstants is an object with all
the constants defined as object fields. Normal classes are mapped to a
constructor function and a prototype which defines the instance fields
and methods. For example, mxEditor is a function and mxEditor.prototype
is the prototype for the object that the mxEditor function creates. The
<em>mx</em> prefix is a convention that is used for all classes in the
mxGraph package to avoid conflicts with other objects in the global
namespace.</p>
<p>For subclassing, the superclass must provide a constructor that
is either parameterless or handles an invocation with no arguments.
Furthermore, the special constructor field must be redefined after
extending the prototype. For example, the superclass of mxEditor is
mxEventSource. This is represented in JavaScript by first
&ldquo;inheriting&rdquo; all fields and methods from the superclass by
assigning the prototype to an instance of the superclass, eg.</p>
<pre>mxEditor.prototype = new mxEventSource()</pre>
<p>and redefining the constructor field using:</p>
<pre>mxEditor.prototype.constructor = mxEditor</pre>
<p>The latter rule is applied so that the type of an object can be
retrieved via the name of it&rsquo;s constructor using <EM>mxUtils.getFunctionName(obj.constructor)</em>.</p>
<h5><a name="constructor"></a>Constructor</h5>
<p>For subclassing in mxGraph, the same mechanism should be applied.
For example, for subclassing the mxGraph class, first a constructor must
be defined for the new class. The constructor calls the super
constructor with any arguments that it may have using the <em>call</em>
function on the mxGraph function object, passing along explicitly each
argument:</p>
<pre>
function MyGraph(container)
{
mxGraph.call(this, container);
}
</pre>
<p>The prototype of MyGraph inherits from mxGraph as follows. As
usual, the constructor is redefined after extending the superclass:</p>
<pre>MyGraph.prototype = new mxGraph();
MyGraph.prototype.constructor = MyGraph;</pre>
<p>You may want to define the codec associated for the class after
the above code (see I/O section of manual). This code will be executed
at class loading time and makes sure the same codec is used to encode
instances of mxGraph and MyGraph.</p>
<pre>
var codec = mxCodecRegistry.getCodec(mxGraph);
codec.template = new MyGraph();
mxCodecRegistry.register(codec);
</pre>
<h5><a name="functions"></a>Functions</h5>
<p>In the prototype for MyGraph, functions of mxGraph can be
extended as follows.</p>
<pre>
MyGraph.prototype.isSelectable = function(cell)
{
var selectable = mxGraph.prototype.isSelectable.apply(this, arguments);
var geo = this.model.getGeometry(cell);
return selectable &amp;&amp;(geo == null || !geo.relative);
}
</pre>
<p>The supercall in the first line is optional. It is done using the
<em>apply</em> function on the <em>isSelectable</em> function object of
the mxGraph prototype, using the special <em>this</em> andn <em>arguments</em>
variables as parameters. Calls to the superclass function are only
possible if the function is not replaced in the superclass as follows,
which is another way of &ldquo;subclassing&rdquo; in JavaScript.</p>
<pre>
mxGraph.prototype.isSelectable = function(cell)
{
var geo = this.model.getGeometry(cell);
return selectable &amp;&amp; (geo == null || !geo.relative);
}
</pre>
<p>The above scheme is useful if a function definition needs to be
replaced completely.</p>
<p>In order to add new functions and fields to the subclass, the
following code is used. The example below adds a new function to return
the XML representation of the graph model:</p>
<pre>
MyGraph.prototype.getXml = function()
{
var enc = new mxCodec();
return enc.encode(this.getModel());
}
</pre>
<h5><a name="fields"></a>Fields</h5>
<p>Likewise, a new field is declared and defined as follows:</p>
<pre>
MyGraph.prototype.myField = &lsquo;Hello, World!&rsquo;;
</pre>
<p>Note that the value assigned to myField is created only once,
that is, all instances of MyGraph share the same value. If you require
instance-specific values, then the field must be defined in the
constructor instead. For example:</p>
<pre>
function MyGraph(container)
{
mxGraph.call(this, container);
this.myField = [];
}
</pre>
<p>Finally, a new instance of MyGraph is created using the following
code, where container is a DOM node that acts as a container for the
graph view:</p>
<pre>
var graph = new MyGraph(container);
</pre>
<br/>
<h3><a name="general_javascript"></a>General JavaScript Development</h3>
<h4><a name="development_environments"></a>JavaScript Development
Environments</h4>
<p>Please refer to the <a
href="http://extjs.com/learn/Manual:Resources#IDEs">ExtJS manual
entry</a> on the subject.</p>
<h4><a name="debugging_javascript"></a>Debugging JavaScript</h4>
<p>Please refer to the <a
href="http://en.wikipedia.org/wiki/JavaScript#Debugging">wikipedia
entry</a>, or to the <a
href="http://extjs.com/learn/Manual:Resources#Various_debugging_tools">ExtJS
manual entry</a>, for information on the subject.</p>
<h4><a name="javascript_obfuscation"></a>JavaScript Obfuscation</h4>
<p>By default, when you deliver JavaScript to a browser client, you
deliver the entire source to that JavaScript. That JavaScript is then
interpreted and run on the browser. It is not possible to encrypt the
JavaScript to any extent on the client at the point it is run, since the
JavaScript source must be understood by the JavaScript interpretor and
interpreted languages do not have a binary intermediate form.</p>
<p>It would be possible to encrypt the JavaScript in transmission
and have it decrypted and run on the client, but the client would still
be able to access the source after decryption.</p>
<p>We do not obfuscate because the method names form a public API
and I/O would need to understand the obfuscation at both communication
ends.</p>
<h4><a name="namespaces"></a>Namespaces</h4>
<p>The concept of namespaces does not exist in JavaScript, so take
great care when creating new class names. In mxGraph, all of the classes
begin with the prefix &ldquo;mx-&rdquo;, to avoid clashes or overriding
prototypes unintentionally. Prior to starting your application it is
worth creating a prefix particular to your application that you append
to all of the your classes to create a manual &ldquo;namespace&rdquo;.</p>
<br/>
<h2><a name="hello_world"></a>Hello World!</h2>
<p>Hello World in mxGraph consists of a simple client-side example
that displays two connected vertices with the labels &ldquo;Hello&rdquo;
and &ldquo;World!&rdquo;. The example demonstrates the following things:
</p>
<ul>
<li><strong>Creating an HTML page that links the mxGraph
client JavaScript,</strong></li>
<li><strong>Creating a container to place the mxGraph
into,</strong></li>
<li><strong>Adds the required cells to that graph.</strong></li>
</ul>
<p>The source code for the example, helloworld.html, can be found
below and in the examples directory of both the evaluation and full
versions of mxGraph. The HTML source contains two main sections, the
head and the body. These contain the following main elements that you
can consider a template for building a basic mxGraph application:</p>
<ul>
<li><strong>mxBasePath</strong>: This is a JavaScript variable
that defines the directory within which the css, images, resources and js
directories are expected to be found. It is JavaScript code and needs
to be placed with in a <em>script</em> tag. This must come before the
line loading mxClient.js and should not have a trailing slash.</li>
<li><strong>mxClient.js</strong>: This is the path to mxGraph
library. If the HTML file is executed locally, the path might be local
to the computer or a public Internet path. If the html page were
downloaded from a web server, the path would generally be a public
Internet path.</li>
<li><strong>Creation of the container</strong>: At the bottom of
the code, in the body element, the function that is called on loading
the web page is defined (the value of onload). It passes in a div
container as a parameter, that is defined underneath. This div is the
container the mxGraph component will be placed within. In this example
a grid background is applied, as commonly used in diagramming
applications. No other part of the graph visuals are described at
container creation, other than the background and the container width
and height.
<p>Note that the overflow:hidden style should always be used if you
want no scrollbars to appear.</p>
</li>
<li><strong>The entry function</strong>: The main code of the file
is the entry method executed on page load in this case. This is
JavaScript code and must be within a JavaScript <em>script</em>
element. The first lines of any mxGraph application should be to check
the browser is supported and exit appropriately if not. If the browser
is supported, a mxGraph is created within the div container and three
cells are added to the graph between the begin/end update calls.</li>
</ul>
<br/>
<p><img src="images/mx_man_hello_world.png" name="ill_hello_world" />
<br/>
<em>The mxGraph HelloWorld example</em></p>
<br/>
<pre>
&lt;html&gt;
&lt;head&gt;
&lt;title&gt;Hello, World! example for mxGraph&lt;/title&gt;
&lt;!-- Sets the <U>basepath</U> for the library if not in same directory --&gt;
&lt;script type=<em>&quot;text/javascript&quot;</em>&gt;
mxBasePath = '../src';
&lt;/script&gt;
&lt;!-- Loads and <U>initializes</U> the library --&gt;
&lt;script type=<em>&quot;text/javascript&quot;</em> src=<em>&quot;../src/js/mxClient.js&quot;</em>&gt;&lt;/script&gt;
&lt;!-- Example code --&gt;
&lt;script type=<em>&quot;text/javascript&quot;</em>&gt;
// Program starts here. Creates a sample graph in the
// DOM node with the specified ID. This function is invoked
// from the onLoad event handler of the document (see below).
function main(container)
{
// Checks if the browser is supported
if (!mxClient.isBrowserSupported())
{
mxUtils.error('Browser is not supported!', 200, false);
}
else
{
// Creates the graph inside the given container
var graph = new mxGraph(container);
// Enables rubberband selection
new mxRubberband(graph);
// Gets the default parent for inserting new cells. This
// is normally the first child of the root (ie. layer 0).
var parent = graph.getDefaultParent();
// Adds cells to the model in a single step
graph.getModel().beginUpdate();
try
{
var v1 = graph.insertVertex(parent, null,
'Hello,', 20, 20, 80, 30);
var v2 = graph.insertVertex(parent, null,
'World!', 200, 150, 80, 30);
var e1 = graph.insertEdge(parent, null, '', v1, v2);
}
finally
{
// Updates the display
graph.getModel().endUpdate();
}
}
};
&lt;/script&gt;
&lt;/head&gt;
&lt;!-- Page passes the container for the graph to the program --&gt;
&lt;body onload=<em>&quot;main(document.getElementById('graphContainer'))&quot;</em>&gt;
&lt;!-- Creates a container for the graph with a grid wallpaper --&gt;
&lt;div id=<em>&quot;graphContainer&quot;</em>
style=<em>&quot;overflow:hidden;width:321px;height:241px;background:url('editors/images/grid.gif')&quot;</em>&gt;
&lt;/div&gt;
&lt;/body&gt;
&lt;/html&gt;
</pre>
<p>Important concepts to note in this exercise are:</p>
<ul>
<li>mxClient.js is a JavaScript file combining all of the
JavaScript source code of mxGraph. When downloading from a web server,
obtaining all the JavaScript as one file is much faster than as lots of
separate files, due to the overhead of the requests/acknowledgements
required for each file. The speed increase is usually at least x2,
although it varies with the capacity of the server to have parallel
sockets open with one client.</li>
<li>The JavaScript code and its dependencies are all placed within
the <em>head</em> element.</li>
<li>Internet Explorer has, by default, security options enabled
that cause a user prompt when attempting to run JavaScript from the
local file system. This can be disabled in the options menu, but note
that running from the local file system is not a deployment scenario of
mxGraph, this would only happen during development.</li>
<li>Your application can be written and linked into the
application either within the HTML file, or in separate JavaScript
source code that is linked into the html in the way the mxClient.js
file is in the example.</li>
</ul>
<h2><a name="deployment"></a>mxGraph Deployment and Debugging</h2>
<p>There are two versions of the mxclient.js file, one for production use and
a second for development/debugging use. <em>javascript/src/js/mxClient.js</em> is
the production version and <em>javascript/debug/js/mxClient.js</em> is for
development. The first version has all linefeeds stripped to ensure the
file is the minimal size possible. This has the side-effect of breaking most
JavaScript debuggers. During development you are advised to use the debug
version, which has the linefeeds in, enabling debugging in the supported
browsers.</p>
<p>Both mxClient.js files are the entire JavaScript source to mxGraph,
with all of the whitespace and comments removed to reduce file size.
Whilst debugging, it is easier to use the individual source files if you
need to debug into the mxGraph library itself. The source code version of
mxGraph contains the full source in the source.zip file in the
<em>javascript/devel</em> directory. Unzipping this into the mxBasePath and removing
the load of the complete mxClient.js file enables easier debugging of mxGraph.
Note that the mxclient.js file in the source zip is a bootstrap file that
loads all the other JavaScript source code.</p>
<p>The download speed of the client source can be further improved
by compressing the code. All modern browsers support receiving and
uncompressing transmissions compressed at the server end and all good
web servers support detection of those browser that do not support it
and send the uncompressed version as a fallback.</p>
<p>For example, on the Apache web server there is a mod_deflate
module, details of its use can be found from a standard search. The
jgraph.com server uses this module and there have been no reports of
issues in any supported browser.</p>
<p>The use of compression reduces the mxClient.js file size down
from about 600KB to around 130KB. The difference is not noticed by the
user on most modern networks, but there might be situations where the
smaller version would be preferable.</p>
<br/>
<h1><a name="model_cells"></a>mxGraph Model and Cells</h1>
<h2><a name="mxgraph_architecture"></a>Core mxGraph architecture</h2>
<h3><a name="mxgraph_model"></a>The mxGraph Model</h3>
<p>The mxGraph model is the core model that describes the structure
of the graph, the class is called mxGraphModel and is found within the
model package. Additions, changes and removals to and from the graph
structure take place through the graph model API. The model also
provides methods to determine the structure of the graph, as well as
offering methods to set visual states such as visibility, grouping and
style.</p>
<p>However, although the transactions to the model are stored on the
model, mxGraph is designed in such a way that the main public API is
through the mxGraph class. The concept of &ldquo;add this cell to the
graph&rdquo; is a more natural description of the action than &ldquo;add
this cell to the model of the graph&rdquo;. Where it is intuitive,
functions available on the model and cells are duplicated on the graph
and those methods on the graph class are considered the main public API.
Throughout the rest of this manual these key API methods are given a pink background:</p>
<div id="coreapi">
<p>anExampleCoreAPIMethod()</p>
</div>
<p>So, though many of the main API calls are through the mxGraph
class, keep in mind that mxGraphModel is the underlying object that
stores the data structure of your graph.</p>
<p>mxGraph uses a transactional system for making changes to the
model. In the HelloWorld example we saw this code:</p>
<pre>
// Adds cells to the model in a single step
graph.getModel().beginUpdate();
try
{
var v1 = graph.addVertex(parent, null, 'Hello,', 20, 20, 80, 30);
var v2 = graph.addVertex(parent, null, 'World!', 200, 150, 80, 30);
var e1 = graph.addEdge(parent, null, '', v1, v2);
}
finally
{
// Updates the display
graph.getModel().endUpdate();
}
</pre>
<p>to perform the insertion of the 2 vertices and 1 edge. For each
change to the model you make a call to beginUpdate(), make the
appropriate calls to change the model, then call endUpdate() to finalize
the changes and have the change event notifications sent out.</p>
<div id="coreapi">
<p><strong>Key API Methods:</strong></p>
<ul>
<li><strong>mxGraphModel.beginUpdate() </strong>- starts a new
transaction or a sub-transaction.</li>
<li><strong>mxGraphModel.endUpdate()</strong> - completes a
transaction or a sub-transaction.</li>
<li><strong>mxGraph.addVertex()</strong> - Adds a new vertex to
the specified parent cell.</li>
<li><strong>mxGraph.addEdge()</strong> - Adds a new edge to the
specified parent cell.</li>
</ul>
</div>
<p><strong>Note</strong> &ndash; Technically you do not have to
surround your changes with the begin and end update calls. Changes made
outside of this update scope take immediate effect and send out the
notifications immediately. In fact, changes within the update scope
enact on the model straight away, the update scope is there to control
the timing and concatenation of event notifications. Unless the update
wrapping causes code aesthetic issues, it is worth using it by habit to
avoid possible problems with event and undo granularity.</p>
<p>Note the way in which the model changes are wrapped in a try
block and the endUpdate() in a finally block. This ensures the update is
completed, even if there is an error in the model changes. You should
use this pattern wherever you perform model changes for ease of
debugging.</p>
<p>Ignore the reference to the parent cell for now, that will be
explained later in this chapter.</p>
<h3><a name="transaction_model"></a>The Transaction Model</h3>
<p>The sub-transaction in the blue block above refers to the fact
that transactions can be nested. That is, there is a counter in the
model that increments for every <em>beginUpdate</em> call and decrements
for every <em>endUpdate</em> call. After increasing to at least 1, when
this count reaches 0 again, the model transaction is considered complete
and the event notifications of the model change are fired.</p>
<p>This means that every sub-contained section of code can (and
should) be surrounded by the begin/end combination. This provide the
ability in mxGraph to create separate transactions that be used as
&ldquo;library transactions&rdquo;, the ability to create compound
changes and for one set of events to be fired for all the changes and
only one undo created. Automatic layouting is a good example of where
the functionality is required.</p>
<p>In automatic layouting, the user makes changes to the graph,
usually through the user interface, and the application automatically
positions the result according to some rules. The automatic positioning,
the layouting, is a self-contained algorithm between begin/end update
calls that has no knowledge of the specifics of the change. Because all
changes within the begin/end update are made directly to the graph
model, the layout can act upon the state of the model as the change is
in progress.</p>
<p>It is important to distinguish between functionality that acts on
the graph model as part of a compound change and functionality that
reacts to atomic graph change events. In the first case, such as for
automatic layouting, the functionality takes the model as-is and acts
upon it. This method should only be used for parts of compound model
changes. All other parts of the application should only react to model
change events.</p>
<p>Model change events are fired when the last endUpdate call
reduces the counter back down to 0 and indicate that at least one atomic
graph change has occurred. The change event contains complete
information as to what has altered (see later section on <strong>Events</strong>
for more details).</p>
<h4><a name="model_change_methods"></a>The Model Change Methods</h4>
<p>Below is a list of the methods that alter the graph model and
should be placed, directly or indirectly, with the scope of an update:</p>
<ul>
<li>add(parent, child, index)</li>
<li>remove(cell)</li>
<li>setCollapsed(cell, collapsed)</li>
<li>setGeometry(cell, geometry)</li>
<li>setRoot(root)</li>
<li>setStyle(cell, style)</li>
<li>setTerminal(cell, terminal, isSource)</li>
<li>setTerminals(edge,source,target)</li>
<li>setValue(cell, value)</li>
<li>setVisible(cell, visible)</li>
</ul>
<p>Initially, we will just concern ourselves with the add and
remove, as well as the geometry and style editing methods. Note that
these are not core API methods, as usual these methods are on the
mxGraph class, where appropriate, and they perform the update
encapsulation for you.</p>
<p><em>Design Background</em> - Some people are confused by the
presence of visual information being stored by the model. These
attributes comprise cell positioning, visibility and collapsed state.
The model stores the default state of these attributes, providing a
common place to set them on a per-cell basis, whereas, views can
override the values on a per-view basis. The model is simply the first
common place in the architecture where these attributes can be set on a
global basis. Remember, this is a graph <em>visualization</em> library,
the visualization part is the core functionality.</p>
<h5><a name="inserting_cells"></a>Inserting Cells</h5>
<p>The three graph cells created in the <CODE>HelloWorld</CODE>
application are two vertices and one edge connecting the vertices. If
you are not familiar with basic graph theory and its terminology, please
see the <a href="http://en.wikipedia.org/wiki/Graph_theory">wikipedia
entry</a>.</p>
<p>You can add vertices and edges using the add() method on the
model. However, for the purposes of general usage of this library, learn
that mxGraph.insertVertex() and mxGraph.insertEdge() are the core public
API for adding cells. The function of the model requires that the cell
to be added is already created, whereas the mxGraph.insertVertex()
creates the cell for you.</p>
<div id="coreapi">
<p><strong>Core API functions:</strong></p>
<ul>
<li><strong>mxGraph.insertVertex(</strong><strong>parent,
id, value, x, y, width, height, style</strong><strong>)</strong> &ndash; creates
and inserts a new vertex into the model, within a begin/end update
call.</li>
<li><strong>mxGraph.insertEdge(</strong><strong>parent,
id, value, source, target, style</strong><strong>)</strong><strong>
&ndash; </strong>creates and inserts a new edge into the model, within a
begin/end update call.</li>
</ul>
</div>
<p><code>mxGraph.insertVertex()</code> will create an mxCell object
and return it from the method used. The parameters of the function are:</p>
<ul>
<li><strong>parent</strong> &ndash; the cell which is the immediate
parent of the new cell in the group structure. We will address the
group structure shortly, but for now use <code>graph.getDefaultParent();</code>
as your default parent, as used in the HelloWorld example.</li>
<li><strong>id</strong> &ndash; this is a global unique identifier
that describes the cell, it is always a string. This is primarily for
referencing the cells in the persistent output externally. If you do not
wish to maintain ids yourself, pass null into this parameter and ensure
that mxGraphModel.isCreateIds() returns true. This way the model will
manage the ids and ensure they are unique.</li>
<li><strong>value</strong> &ndash; this is the user object of the
cell. User object are simply that, just objects, but form the objects
that allow you to associate the business logic of an application with
the visual representation of mxGraph. They will be described in more
detail later in this manual, however, to start with if you use a string
as the user object, this will be displayed as the label on the vertex or
edge.</li>
<li><strong>x, y, width, height</strong> &ndash; as the names
suggest, these are the x and y position of the top left corner of the
vertex and its width and height.</li>
<li><strong>style</strong> &ndash; the style description to be
applied to this vertex. Styles will be described in more detail shortly,
but at a simple level this parameter is a string that follows a
particular format. In the string appears zero or more style names and
some number of key/value pairs that override the global style or set a
new style. Until we create custom styles, we will just use those
currently available.</li>
</ul>
<p>With the edge addition method, the identically named parameters
perform the same function as in the vertex addition method. The source
and target parameters define the vertices to which the edge is
connected. Note that the source and target vertices should already have
been inserted into the model.</p>
<h3><a name="mxcell"></a>mxCell</h3>
<p>mxCell is the cell object for both vertices and edges. mxCell
duplicates many of the functions available in the model. The key
difference in usage is that using the model methods creates the
appropriate event notifications and undo, using the cell makes the
change but there is no record of the change. This can be useful for
temporary visual effects such as animations or changes on a mouse over,
for example. As a general rule though, use the model editing API unless
you encounter a specific problem with this mechanism.</p>
<p>When creating a new cell, three things are required in the
constructor, a value (user object), a geometry and a style. We will now
explore these 3 concepts before returning to the cell.</p>
<h4><a name="styles"></a>Styles</h4>
<p>The concept of styles and stylesheets in conceptually similar to
CSS stylesheets, though note that CSS are actually used in mxGraph, but
only to affect global styles in the DOM of the HTML page. Open up the
util.mxConstants.js file in your editor and search for the first match
on &ldquo;STYLE_&rdquo;. If you scroll down you will see a large number
of strings defined for all the various styles available with this
prefix. Some of styles apply to vertices, some to edges and some to
both. As you can see, these define visual attributes on the element they
act upon.</p>
<p>The mxStylesheet holds one object, styles, which is a hashtable
mapping style names to an array of styles:</p>
<p><img src="images/mx_man_styles.png" name="graphics5"/><br/>
<em>Style arrays within the styles collection</em></p>
<br/>
<p>In the above image the blue box represents the styles hashtable
in mxStyleSheet. The string 'defaultVertex' is the key to an array of
string/value pairs, which are the actual styles. Note that mxGraph
creates two default styles, one for vertices and one for edges. If you
look back to the helloworld example, no style was passed into the
optional style parameter of insertVertex or insertEdge. In this case the
default style would be used for those cells.</p>
<h5><a name="setting_cell_style"></a>Setting the Style of a Cell</h5>
<p>If you wanted to specify a style other than the default for a
cell, you must pass that new style either to the cell when it is created
(mxGraph's insertVertex and insertEdge both have an optional parameter
for this) or pass that style to the cell using model.setStyle().</p>
<p>The style that you pass has the form stylename. ,note that the
stylenames and key/value pairs may be in any order. Below are examples
to demonstrate this concept, adapting the insertVertex call we saw in
helloworld:</p>
<ol>
<li>
<p>A new style called 'ROUNDED' has been created, to apply this to
a vertex:</p>
<pre>var v1 = graph.insertVertex(parent, null, 'Hello', 20, 20, 80, 30, 'ROUNDED');</pre>
</li>
<li>
<p>To create a new vertex with the ROUNDED style, overriding the
stroke and fill colors:</p>
<pre>var v1 = graph.insertVertex(parent, null, 'Hello', 20, 20, 80, 30, 'ROUNDED;strokeColor=red;fillColor=green');</pre>
</li>
<li>
<p>To create a new vertex with no global style, but with local
stroke and fill colors:</p>
<pre>var v1 = graph.insertVertex(parent, null, 'Hello', 20, 20, 80, 30, ';strokeColor=red;fillColor=green');</pre>
</li>
<li>
<p>To create a vertex that uses the defaultVertex style, but a
local value of the fill color:</p>
<pre>var v1 = graph.insertVertex(parent, null, 'Hello', 20, 20, 80, 30, 'defaultVertex;fillColor=blue');</pre>
</li>
</ol>
<br/>
<p>Note that default style must be explicitly named in this case,
missing the style out sets no global style on the cell when the
semi-colon starts the string. If the string starts with no semi-colon,
the default style is used.</p>
<p>Again, the mxGraph class provides utility functions that form the
core API for accessing and changing the styles of cells:</p>
<div id="coreapi">
<p><strong>Core API functions:</strong></p>
<ul>
<li>
<strong>mxGraph.setCellStyle(style, cells)</strong> &ndash; Sets
the style for the array of cells, encapsulated in a begin/end update.
</li>
<li>
<strong>mxGraph.getCellStyle(cell)</strong> &ndash; Returns the
style for the specified cell, merging the styles from any local style
and the default style for that cell type.
</li>
</ul>
</div>
<h5><a name="new_global_style"></a>Creating a New Global Style</h5>
<p>To create the ROUNDED global style described above, you can
follow this template to create a style and register it with
mxStyleSheet:</p>
<pre>
var style = new Object();
style[mxConstants.STYLE_SHAPE] = mxConstants.SHAPE_RECTANGLE;
style[mxConstants.STYLE_OPACITY] = 50;
style[mxConstants.STYLE_FONTCOLOR]= '#774400';
graph.getStylesheet().putCellStyle('ROUNDED',style);
</pre>
<h4><a name="geometry"></a>Geometry</h4>
<p>In the helloworld example we saw the position and size of the
vertices passed into the insertVertex function. The coordinate system in
JavaScript is x is positive to the right and y is positive downwards,
and in terms of the graph, the positioning is absolute to the container
within which the mxGraph is placed.</p>
<p>The reason for a separate mxGeometry class, as opposed to simply
having the mxRectangle class store this information, is that the edges
also have geometry information.</p>
<p>The width and height values are ignored for edges and the x and y
values relate to the positioning of the edge label. In addition, edges
have the concept of control points. These are intermediate points along
the edge that the edge is drawn as passing through. The use of control
points is sometimes referred to as <strong>edge routing</strong>.</p>
<p><img src="images/mx_man_edge_routing.png" name="ill_edge_routing"/><br/>
<em>An edge routed by 2 control points</em></p>
<p>There are two more important additional concepts in geometry,
relative positioning and offsets</p>
<h5><a name="relative_positioning"></a>Relative Positioning</h5>
<p>By default, the x and y position of a vertex is the offset of the
top left point of the bounding rectangle of the parent to the top left
point of the bounding rectangle of the cell itself. The concept of
parents and groups is discussed later in this chapter, but without going
into too much detail, if a cell does not have cell parent, the graph
container is its parent for positioning purposes.</p>
<p><img src="images/mx_man_non_relative_pos.png"
name="ill_non_realtive_pos"/><br/>
<em>Non-relative vertex positioning</em></p>
<br/>
<p>For an edge, in non-relative mode, which is the default mode, the
edge label position is the absolute offset from the graph origin.</p>
<p><img src="images/mx_man_non_realtive_edge_pos.png"
name="ill_non_rel_edge_pos"/><br/>
<em>Non-relative edge label positioning</em></p>
<br/>
<p>For vertices in relative mode, (x,y) is the proportion along the
parent cell's (width, height) where the cell's origin lies. (0,0) is the
same origin as the parent, (1,1) places the origin at the bottom right
corner of the parent. The same relative positioning extends below 0 and
above 1 for both dimensions. This positioning is useful for keeping
child cells fixed relative to the overall parent cell size.</p>
<p><img src="images/mx_man_rel_vert_pos.png" name="rel_vert_pos"/><br/>
<em>Relative vertex positions</em></p>
<br/>
<p>Lastly, edge labels in relative mode are palced based on the
positioning from the center of the edge. The x-coordinate is the
relative distance from the source end of the edge, at -1, to the target
end of the edge, at 1. The y co-ordinate is the pixel offset orthogonal
from the edge. The diagram below shows the values of x,y for various
edge labels in relative mode. Note that for a straight edge, the
calculations are simple. For edges with multiple control points, the
edge has to be traced along its segments (a segment being the line
between end points and/or control points) to find the correct distance
along the edge. The y value is the orthogonal offset from that segment.</p>
<p>Switching relative positioning on for edge labels is a common
preference for applications. Navigate to the mxGraph.insertEdge()
function in mxGraph, you will see this calls createEdge(). In
createEdge() the geometry is set relative for every edge created using
this prototype. This is partly the reason for the amount of helper
functions in mxGraph, they enable easy changing of the default
behaviour. You should try to use the mxGraph class API as much as
possible to provide this benefit in your applications.</p>
<h5><a name="offsets"></a>Offsets</h5>
<p>The offset field in mxGeometry is an absolute x,y offset applied
to the cell <strong>label</strong>. In the case of edge labels, the
offset is always applied after the edge label has been calculated
according to the relative flag in the above section.</p>
<div id="coreapi">
<p><strong>Core API functions:</strong></p>
<ul><li><strong>mxGraph.resizeCell(cell, bounds)</strong> &ndash; Resizes
the specified cell to the specified bounds, within a begin/end update
call.</li>
<li><strong>mxGraph.resizeCells(cells, bounds)</strong> &ndash;
Resizes each of the cells in the cells array to the corresponding entry
in the bounds array, within a begin/end update call.</li>
</ul>
</div>
<h4><a name="user_objects"></a>User Objects</h4>
<p>The User object is what gives mxGraph diagrams a context, it
stores the business logic associated with a visual cell. In the
HelloWorld example the user object has just been a string, in this case
it simply represents the label that will be displayed for that cell. In
more complex applications, these user objects will be objects instead.
Some attribute of that object will generally be the label that the
visual cell will display, the rest of the object describes logic
relating to the application domain.</p>
<p>Using the example of a simple workflow or process application,
say we have the graph below (<a
href="http://www.jgraph.com/demo/mxgraph/editors/workfloweditor.html">this
example is available online</a>, select the Swimlanes example from the tasks
window):</p>
<p><img src="images/mx_man_simple_workflow.png"
name="ill_simple_workflow"/><br/>
<em>A simple workflow</em></p>
<br/>
<p>Typically, this workflow will exist on some application server
and/or database. The browser user connects to that server, or some
front-end server linked to the application server and the user's web
application requests the &ldquo;order&rdquo; workflow. The server
obtains the data of that workflow and transmits it to the client.</p>
<p>mxGraph supports the process of populating the model on the
server-side and transmitting to the client, and back again. See the
later chapter on &ldquo;I/O and Server Communication&rdquo;.</p>
<p>The data transmitted will be both the visual model (the diagram)
as well as the business logic (mostly contained in the user objects).
The client will initially show the diagram above. If the user has
permission to edit this workflow they will normally be able to do two
things, 1) edit the diagram, add and remove vertices, as well as
changing the connections, and 2) edit the user objects of the cells
(vertices and/or edges).</p>
<p>In the online demo, if you right click and select properties of
the &ldquo;Check Inventory&rdquo; diamond you will see this dialog:</p>
<p><img src="images/mx_man_vertex_props.png" name="ill_vertex_props"/><br/>
<em>The properties of a vertex</em></p>
<br/>
<p>These properties show the geometry, label, ID etc, but a dialog
could just as easily show the user object of the cell. There might be a
reference to some process on the workflow engine as to how the inventory
is actually checked. This might be an application specific mechanism for
both the server and client to assign some identification to remote
method calls. Another value might be the type of object that process
returned, maybe a boolean or an integer to indicate stock level in this
case. Given that return type, it is possible to enforce constraints with
the diagram and provide visual alerts of if, say, the outgoing edges
decision check does not correspond to the return type of the vertex.</p>
<p>Next, as an example, the user objects of the outgoing edges might
contain a label and a boolean state. Again, the mxGraph-based editor
might provide the means to alter the boolean value. On the server, when
executing the process, it might follow the edges that correspond to the
boolean value returned by the decision node.</p>
<p>Keep in mind that the above example is very domain specific, it
is there to explain how the user object maps to the business logic of
the application. It visualizes how mxGraph creates what we term a <strong>contextual
graph</strong>. The context is formed by the connections between vertices and the
business logic stored within the user objects. A typical application
receives the visual and business logic from a sever, may allow editing
of both, then transmits both back to the server for persistence and/or
execution.</p>
<h4><a name="cell_types"></a>Cell Types</h4>
<p>As described previously, mxGraph is the primary API for using
this library and the same concept applies to cells. One basic state of
the cell not exposed on the graph is whether a cell is a vertex or an
edge, this call be performed on the cell or on the model.</p>
<p>There are two boolean flags on mxCell, vertex and edge, and the
helper methods set one of these to true when the cell is created.
isVertex(), isEdge() on mxIGraphModel are what the model uses to
determine a cell's type, there are not separate objects for either type.
Technically, it is possible to switch the type of a cell at runtime, but
take care to invalidate the cell state (see later section) after
changing the type. Also, be aware that the geometry object variable
means different things to vertices and edges. Generally, it is not
recommended to change a cell type at runtime.</p>
<h3><a name="group_structure"></a>Group Structure</h3>
<p>Grouping, within mxGraph, is the concept of logically associating
cells with one another. This is commonly referred to as the concept of
sub-graphs in many graph toolkits. Grouping involves one or more
vertices or edges becoming children of a parent vertex or edge (usually
a vertex) in the graph model data structure. Grouping allows mxGraph to
provide a number of useful features:</p>
<ul>
<li>Sub-graphs, the concept of a logically separate graph that is
displayed in the higher level graph as a cell per sub-graph.</li>
<li>Expanding and collapsing. Collapsing is the ability to replace
a collection of grouped cells visually with just their parent cell.
Expanding is the reverse of this. This behaviour can be seen by
clicking the small &ldquo;-&rdquo; in the top left corner of the group
cells of the swimlanes example in the <a
href="http://www.jgraph.com/demo/mxgraph/editors/workfloweditor.html">online
workfloweditor</a> example. This is described in the C<em>omplexity
Management</em> section below.</li>
<li>Layering. Layering is the concept of assigning cells to a
particular z-order layer within the graph display.</li>
<li>Drill down, step up. These concepts allow sub-graphs to be
visualized and edited as if they are a complete graph. In the <em>User
Objects</em> section we saw the &ldquo;check inventory&rdquo; vertex as a
single cell. Take, for example, the case where a developer is
describing each of the vertices in the process as the software
processes that perform the task. The application might have an option
to drill down into the check inventory vertex. This would result in a
new graph appearing that describes in detail how exactly the system
checks the inventory. The graph might have the title of the parent
&ldquo;check inventory&rdquo; vertex to indicate it is a child, as well
as the option to step-up back to the next level up.</li>
</ul>
<p>In grouping, cells are assigned a parent cell. In the simplest
case, all cells have the default parent as their parent. The default
parent is an invisible cell with the same bounds as the graph. This is
the cell returned by graph.getDefaultParent() in the helloworld example.
The x,y position of a vertex is its position relative to its parent, so
in the case of default grouping (all cells sharing the default parent)
the cell positioning is also the absolute co-ordinates on the graph
component. In the case all cells being added to the default root, the
group structure logically looks like, in the case of the helloworld
example, the diagram below.</p>
<p>Note the addition of the Layer 0 cell, this is the default
indirection in the group structure that allows layer changes with the
requirement of additional cells. We include it below for correctness,
but in later group diagrams it will be omitted.</p>
<p><a name="ill_mx_man_hello_struct"><img src="images/mx_man_hello_struct.png"
WIDTH=441 HEIGHT=241 BORDER=0></a><br/>
<em>The group structure of the helloworld example</em></p>
<br/>
<p>Also, note that the position of the edge label (x,y in geometry)
is relative to the parent cell.</p>
<p>If we go back to the simple workflow example in the User Objects
section, we can see what grouping might look like visually. In the
example the group cells represent people and the child vertices
represent tasks assigned to those people. In this example the logical
group structure looks like this:</p>
<p><img src="images/mx_man_log_group_struct.png"
name="ill_mx_man_log_group_struct"/><br/>
<em>The logical group structure of the workflow example</em></p>
<br/>
<p>The workflow action vertices are the yellow children and the
swimlane group vertices are marked blue.</p>
<p>Inserting cells into the group structure is achieved using the
parent parameter of the insertVertex and insertEdge functions on the
mxGraph class. These functions set the parent cell on the child
accordingly and, importantly, informs the parent cell of its new child.</p>
<p>Altering the group structure is performed via the
mxGraph.groupCells() and mxGraph.ungroupCells() functions.</p>
<div id="coreapi">
<p><strong>Core API functions:</strong></p>
<ul><li><strong>mxGraph.groupCells(group, border, cells)</strong> &ndash;
Adds the specified cells to the specified group, within a begin/end
update</li>
<li><strong>mxGraph.ungroupCells(cells)</strong> &ndash; Removes the
specified cells from their parent and adds them to their parent's
parent. Any group empty after the operation are deleted. The operation
occurs within a begin/end update.</li>
</ul>
</div>
<h3><a name="complexity_management"></a>Complexity Management</h3>
<p>There are two primary reasons to control the number of cells
displayed at any one time. The first is performance, drawing more and
more cells will reach performance usability limits at some point on any
platform. The second reason is ease of use, a human can only comprehend
a certain amount of information. All of the concepts associated with
grouping, listed above, can be used to reduce the complexity of
information on the screen for the user.</p>
<h4><a name="folding"></a>Folding</h4>
<p>Folding is the collective term we use for expanding and
collapsing groups. We say a cell is folded by making it's child vertices
invisible. There are a number of functions relating to this feature:</p>
<div id="coreapi">
<p><strong>Core API function:</strong></p>
<ul><li><strong>mxGraph.foldCells(collapse, recurse, cells)</strong>
&ndash; States the collapsed state of the specificed cells, within a
begin/end update.</li></ul>
</div>
<p><strong>Folding related functions:</strong></p>
<p><strong>mxGraph.isCellFoldable(cell, collapse)</strong> &ndash;
By default true for cells with children.</p>
<p><strong>mxGraph.isCellCollapsed(cell)</strong> &ndash; Returns
the folded state of the cell</p>
<p>When a group cell is collapsed, three things occur by default:</p>
<ul>
<li>The children of that cell become invisible.</li>
<li>The group bounds of the group cell is used. Within mxGeometry
there is a alternativeBounds field and in groups cells, by default
store a separate bounds for their collapsed and expanded states. The
switch between these instances is invoked by mxGraph.swapBounds() and
this is handled for you within a foldCells() call. This allows
collapsed groups to be resized whilst when expanded again the size
looks correct using the pre-collapsed size.</li>
<li>Edge promotion occurs, by default. Edge promotion means
displaying edges that connect to children within the collapsed group
that also connect to cells outside of the collapsed group, by making
them appear to connect to the collapsed parent.</li>
</ul>
<p><img src="images/mx_man_expand_swim.png"
name="ill_mx_man_expand_swim"/><br/>
<em>Expanded swimlane</em></p>
<p><img src="images/mx_man_collapse_swim.png"
name="ill_mx_man_collapse_swim"/><br/>
<em>Collapsed Swimlane</em></p>
<p>The above two images demonstrate these three concepts. In its
expanded state the upper group cell displays a small box in the top left
hand corner with a &ldquo;-&rdquo; character inside. This indicates that
clicking on this box collapses the group cell. Doing this we get the
bottom image where the group cell takes on its collapsed size. Child
vertices and edge that do not leave the group cell are made invisible.
Finally, edges that exit the group cell are promoted to appear to be
connected to the collapsed group cell. Clicking on the &ldquo;+&rdquo;
character that now appears within the box expands the group cell and
brings it back to its original state of the top image.</p>
<p>Using the mxGraph.foldCells() function, you can achieve the same
result programmatically as clicking on the expand/collapse symbols. One
common usage of this is when the application zooms out a specific
amount, clusters of cells are grouped and the grouped cell collapsed
(very often without the &ldquo;-&rdquo; box since the application is
controlling the folding). This way fewer, larger cells are visible to
the user, each one representing their children cells logically. You
might then provide a mechanism to zoom into a group, which expands it in
the process. You might also provide drill-down/step-up, explained next.</p>
<h4><a name="sub_graph_drilling"></a>Sub-Graphs, Drill-Down /
Step-Up</h4>
<p>Sometimes, as an alternative to expand/collapse, or possibly in
combination with it, your graph will be composed of a number of graphs,
nested into a hierarchy. Below we see a simple example:</p>
<p><img src="images/mx_man_drill_down.png"
name="ill_mx_man_drill_down"/><br/>
<em>An example top level workflow</em></p>
<br/>
<p>This simple workflow consists of three high level steps.
Obviously, the individual steps contain a number of sub-steps and we
will look at a sub-graph of the <em>Solve Bug</em> cell.</p>
<p>Under the <em>Solve Bug</em> vertex we have created a number of
children to represent the process of solving a bug in more detail, in
this case the process of solving a bug on the <a
href="http://en.wikipedia.org/wiki/Starship_Enterprise">Starship
Enterprise</a>.</p>
<p>In this example, which uses the GraphEditor example, the menu
option shown selected in the above image invokes
mxGraph.enterGroup(cell), which is one of the pair of core API functions
for sub-graphs.</p>
<div id="coreapi">
<p><strong>Core API functions:</strong></p>
<ul><li><strong>mxGraph.enterGroup(cell)</strong> &ndash; Makes the
specified cell the new root of the display area.</li>
<li><strong>mxGraph.exitGroup()</strong> - Makes the parent of the
current root cell, if any, the new root cell.</li>
<li><strong>mxGraph.home()</strong> - Exits all groups, making the
default parent the root cell.</li>
</ul>
</div>
<p>The root cell of the graph has been, up to now, the default
parent vertex to all first-level cells. Using these functions you can
make any group cell in the group structure the root cell, so that the
children of that parent appear in the display as the complete graph.</p>
<p><img src="images/mx_man_drilling.png" name="ill_mx_man_drilling"/><br/>
<em>Result of drilling down into the Solve Bug vertex</em></p>
<P>The same graph expanded using folding instead looks like:</p>
<p><img src="images/mx_man_top_level.png" name="ill_mx_mantop_level"
WIDTH=695 HEIGHT=227 BORDER=0></p>
<p>Exiting the group using the <em>shape-&gt;exit group</em> option,
which invokes mxGraph.exitGroup, brings you back to the original 3
vertex top level graph.</p>
<h4><a name="layer_filter"></a>Layering and Filtering</h4>
<p>In mxGraph, like many graphical applications, there is the
concept of z-order. That is, the order of objects as you look into the
screen direction. Objects can be behind or in front of other objects and
if they overlap and are opaque then the back-most object will be
partially or complete obscured. Look back to the <a href="#ill_mx_man_hello_struct">
graph structure of HelloWorld illustration</a> above. Children cells are stored under
parents in a deterministic order (by default the order in which you add
them).</p>
<p>If we move the cells in the HelloWorld example we see the
following result:</p>
<p><img src="images/mx_man_overlap.png" name="ill_mx_man_overlap"/><br/>
<em>Overlapped vertices</em></p>
<p>It can be seen that the <em>World</em> vertex is in front of the
<em>Hello</em> vertex. This is because the <em>World</em> vertex has a
higher child index than the <em>Hello</em> vertex, at positions 1 and 0
respectively in the ordered collection that holds the children of the
root cell.</p>
<p>To change order we use mxGraph.orderCells.</p>
<div id="coreapi">
<p><strong>Core API function:</strong></p>
<ul><li><strong>mxGraph.orderCells(back, cells)</strong> &ndash; Moves
the array of cells to the front or back of their siblings, depending on
the flag, within a begin/end update.</li></ul>
</div>
<p>A sibling cell in mxGraph is any cell that shares the same
parent. So by invoking this on the <em>Hello</em> vertex it would then
overlap the <em>World</em> Vertex.</p>
<p>Ordering and grouping can be extended to form logically layered
groups. The cells are drawn via a depth-first search. Take the
HelloWorld example again and imagine that both the <em>Hello</em> and <em>World
</em>vertices have some hierarchy of children underneath them. The <em>Hello</em>
vertex and all of its children will be drawn before the <em>World</em>
vertex or any of its children. If <em>Hello</em> and <em>World</em> were
invisible group cells you then have two hierarchies of cells, one being
drawn entirely before the other. You can also switch the order of the
hierarchies by simply switching the order of the invisible group cells.</p>
<p>The concept of layering is demonstrated in the layers.html
example. Here buttons are used to set the visibility of group layer
cells. This example ties very closely into the concept of filtering.</p>
<p>In filtering cells with some particular attribute are displayed.
One option to provide filtering functionality is to check some state
before rendering the cells. Another method, if the filtering conditions
are simple and known in advance, is to assign filterable cells by
groups. Making the groups visible and invisible performs this filtering
operation.</p>
<br/>
<hr size="1">
&copy; 2006-2014 by JGraph Ltd.
</body>
</html>