Menu

Doing That Drag Thang

February 27, 2002

Antoine Quint

Introduction

In last month's article, we took a wee trip in the exciting lands of SMIL-powered SVG animation. In that article, we used XML elements to achieve our goals. Today I will show you around a place that might sound a little scary, but that's just as much fun when you take the time to imagine how many possibilities it offers: it is time to take a look at scripting SVG, for all the nifty interactions that declarative SVG Animation could not handle.

As an XML application, SVG benefits from the Document Object Model. The DOM is an object-oriented API for reading from and writing to an XML document. Even if you've never heard of the DOM, you might have had some unfortunate experience with its wayward sibling, DHTML. DHTML really was the combination of HTML, CSS, JavaScript, and a DOM. What made DHTML such a headache is that the two main browser vendors had different DOMs, neither being compliant with the DOM as specified by the W3C. Recent versions of major browsers now support the W3C DOM Level 2, just like the Adobe SVG Viewer, which also offers support for the SVG DOM. If you need more formal introductions to the DOM, I would strongly suggest a bit of preparatory reading before delving into this article. The O'Reilly Network has some great articles by Scott Andrew LePera on scripting DOM Level 2 in an XHTML context (see parts one and two), and the W3C is the official source for the various DOM specifications.

What's in it for us?

Scripting SVG opens up many new possibilities. While client-side scripting is a well-established practice in different environments (especially DHTML and Flash ActionScript), I believe the SVG scripting environment offers a more comprehensive and standards-based approach. Adobe's SVG Viewer version 3.0 offers stable and powerful tools for us to work with in a way that has never been possible before.

First, the basic API is the W3C DOM Level 2 and offers a generic approach to XML scripting. We have the possibility to script every element and attribute in a consistent way rather than deal with specific features that have been made available by the custom API of a (legacy) browser or that of Flash 5 ActionScript. For instance, while Flash 5 ActionScript allows for duplication of existing movie clips and symbols, it does not allow for on-the-fly creation of graphic objects or completely generic graphic properties access (say, updating a stop-color of a gradient fill). Also, Adobe's Viewer comes with the Mozilla JavaScript engine -- making it immune from the pecularities of the host web browser's JavaScript support -- and on top of that offers the most advanced features of JavaScript.

The DOM Level 2 surely is a great basis for scripting SVG, but being a generic API, it does not handle any of the graphics-oriented needs of SVG scripting. You might well want to have a DOM method for computing an object's bounding box or applying a matrix transform to a given coordinate. Well, SVG has its own extension of the core DOM that does, among other things, just that -- the SVG DOM. In this article we take a short stroll through the SVG DOM.

I will show you how to build a simple and common graphics-oriented interaction, a drag. A word of warning before we get started though: this is an introductory example and does not handle zooming, panning, viewBox, or any other subtleties. I promise we will get back to the dragging in a more generic and powerful way later when we get more familiar with the SVG DOM. Still, this example will show you how to handle events, get mouse pointer coordinates, update elements' CSS properties, and work with two of the basic classes of the SVG DOM, SVGPoint and SVGMatrix. Download the source code for the example, and then let's get started.

The concepts of dragging

Take a look at the SVG example before we proceed (requires SVG plug-in)

Being able to drag an object around the screen is quite a powerful thing. And it is quite simple to achieve. In fact, interactivities are often simple enough to code as long as you have a firm understanding of what needs to happen when you click here and there. The basic idea is that once you have clicked on a shape, it should follow your mouse around the screen until you release your mouse click. A first approach would be to simply apply the mouse coordinates to the shape, but that would result in always dragging by the upper left corner of the shape (the SVG coordinate system being left to right and y-down). So what you want to do when starting the drag is to compute the distance from the top-left corner of your shape to the position of your mouse cursor. Then, when you want to update the shape's coordinates on moving your mouse around, you'll just have to take the mouse coordinates and subtract that distance from it.

Catching events in SVG

In order to have certain pieces of code executed when we click or move the mouse around, we need to have a mechanism so that our code could be automatically informed of mouse activities. Luckily, SVG provides event listeners. You might have heard of these before and probably even used these in DHTML with attributes like onmouseover. An event listener is always be focused on what's going on in your SVG, and when something noteworthy happens it tells you. In our case, we need to listen to three different events (all-mouse related): mousedown (when pushing a mouse button), mousemove (when moving the mouse), and mouseup (when releasing a mouse click). These are the events, and the event listeners are attributes with an "on" prefix. If we want to execute an initialization script when we encounter a mousedown event on our shape, we could just write onmousedown="some_function()". In our demo, it is reflected in this bit of SVG:

<g id="target" onmousedown="initDrag()">

Then all we have to do is implement all we need done for initialization purposes in the initDrag() function (we'll see how it looks later on). Similarly you can see how we handle mousemove and mouseup events:

<g id="background" onmousemove="drag()" onmouseup="endDrag()" style="pointer-events: none;">

Adding script to SVG

Now that we have established a bridge from SVG to JavaScript code, it is time to take a look at the code itself. How do we actually get to write JavaScript code with SVG? SVG has a <script> element that allows for either inline coding within a CDATA section (to delimit non-XML portions of the document) or a link to a separate JavaScript library. In our case we xlink:href to a library called drag.es, which keeps the SVG code cleaner:

<script a3:scriptImplementation="Adobe" type="text/ecmascript" xlink:href="drag.es" />

What about that a3:scriptImplementation="Adobe" bit? The scriptImplementation attribute is an SVG extension provided by Adobe (and cleanly introduced as part of their namespace) to allow us to tell the Adobe SVG Viewer to use its own scripting engine rather than the hosting browser's (and believe me, you really want to do that). Now open up the drag.es file and pay close attention.

Wading through the code

I will not go through every single line of this (short) script. However, I believe the file is clearly commented and that comments and the articles I have recommended at the beginning should fill in neatly. Let's concentrate on code specific to the SVG DOM. Before we actually get into the event handler functions, there is one bit of code that is executed when loading the file that's worth taking a look:

var offset = root.createSVGPoint();

The root variable is a global pointer to the root <svg> element of our document. As the root element, this element has special powers and has a method called createSVGPoint() that we make use of here. This method, quite simply, creates an SVGPoint and returns it, making our global offset variable an SVGPoint itself. But what's an SVGPoint? It is one of the few "datatype" objects featured in the SVG DOM and is a representation for a point. As to the role of the offset variable, it will be used later in the script to keep track of the offset of the dragging session.

We said before that the initDrag() function was called when clicking on our draggable shape. The use of this function is to compute the dragging offset and apply a few style changes to our composition. We start off with an interesting line of code:

var matrix = target.getCTM();

The getCTM() method is a neat function that returns the "current transformation matrix", as an SVGMatrix datatype object, of the node we call it on. As you probably know, most SVG elements feature the transform attribute in which you can specify a matrix or pre-set types of transformations (like a translation). In our example, the position of the "target" group is defined with such an attribute:

<g id="target" transform="translate(80,70)" style="pointer-events: all">

The SVGMatrix returned by getCTM() helps handling data stored in the transform attribute. In this case, it would have been easy to just parse the string to find out the translation, but there are cases when you have inline matrix multiplications that would require a lot more work. So we'll use our SVGMatrix here. Remember what an SVG matrix looks like: (a, b, c, d, e, f) with e and f being the fields relative to x and y translations. Thus, if we want to read 80 and 70 from the attribute, we can simply use matrix.e and matrix.f now that we have stored the matrix in the matrix variable.

Now that we have the original position of the draggable object before any dragging is done, we need to find out the position of the mouse so that we can compute the dragging offset. For this, we have created another function called getMouse(). Here we call it:

var mouse = getMouse(evt);

You will probably notice that the getMouse() function takes an argument evt that we have not used or declared before. This is because the DOM offers a mechanism for inspecting the event that got sent to our function. evt is a name commonly given for the events object that is implicitly and automatically passed to all event handling functions. The events object really is quite helpful and holds information like a reference to the node that received the event, mouse coordinates, and other neat things. Looking at the code for getMouse() we see that kind of thing:

var position = root.createSVGPoint();

position.x = evt.clientX;

position.y = evt.clientY;

return position;

clientX and clientY are two fields that the SVG DOM provides for us to be able to track mouse positions. The position these fields give us are computed relative to the top-left of the SVG rendering area (the Adobe SVG Viewer within your browser) and do not take into account zooming or panning. getMouse() returns an SVGPoint storing the mouse coordinate for the event provided as a parameter. Now that we have both the mouse coordinates, we can go back to our event handling initDrag() function and compute the offset:

offset.x = matrix.e - mouse.x;

offset.y = matrix.f - mouse.y;

There we are. To finish things off we will make a crucial adjustment to the CSS properties of the background and the target layers so that the dragging goes smoothly. Now that our draggable element has received the initiating event (mousedown), it is important that we make sure that it will not receive any more events that could conflict with the ones our background layer expects. To prevent an SVG element (and its children) from receiving events, one has to set its CSS pointer-events property to "none". But why do we have to do that?

We have set our SVG so that the background layer handles the mousemove event. If the draggable shape still receives events, it will prevent graphics underneath (our background layer) from receiving events. Then why did we not let the draggable shape handle mousemove itself? Well, if the same layer receives both mousedown and mousemove events, our SVG Viewer might not have enough time to go through all the code in the mousedown event handler function (here, initDrag()) before processing the mousemove event handler function (here, drag()). It happens really often that you click and start moving your mouse around in the same millisecond. In that case, our precious offset will not have had enough time to be computed and our much-coveted effect will be ruined, sacrebleu! If JavaScript offered anything similar to Java's synchronized, life would have been easier. Hence we have to do this:

target.style.setProperty('pointer-events', 'none');

background.style.setProperty('pointer-events', 'all');

If you've made it thus far, then you're a courageous SVGer; this last bit really was tricky. Now that the offset is computed, and we're sure event handling was being taken care of as expected, we can do the easy, i.e., the drag itself. Our appropriately-named drag() does this quite well by getting the mouse coordinates every time we move our mouse, computing the new position of our draggable shape taking into account our pre-computed offset, and finally writing to the SVG transform attribute in order to have the graphics updated. Here's how it goes:

// gets the pointer position

var mouse = getMouse(evt);

var x = mouse.x + offset.x;

var y = mouse.y + offset.y;

// updating the matrix

target.setAttribute('transform', 'translate(' + x + ',' + y + ')');

That was easy. The last thing left for us to do is to handle the mouseup event with our endDrag() function. All we need is to reset the pointer-events values to what they were originally. So this really is only the inverse of what we have done in initDrag(), updating the CSS values of the target and background layers.

Wrapping it all up

I hope this simple dragging interaction has uncovered before your starry-eyed faces the power and simplicity of the SVG DOM. It's only the beginning, though. SVG DOM scripting will be a recurring theme in this column since it is an unending gold source. In the next few columns, we will make this code generic, handle zoom and pan (for that "scalable" part), increase performances, and bridge SVG and JavaScript in a more elegant way. Until then, take it easy and à bientôt!