CSS Flexbox Grids for Your Apps

ChocolateChip-UI provides a convenient grid system based on CSS flexbox. This means your grid columns use flex to determine their widths. This gives you responsive layouts. Because the grids use flexbox, there are no floats involved, and no need to clear floats, etc. Also all columns are automatically equal height in the same grid.

To make a grid, just put the class grid on a div:

// Define a grid:
<div class='grid'></div>

This gives you a grid row. Inside this you can create grid columns with the class col.

// Define a grid with 3 columns:
<div class='grid'>
   <div class='col'>1</div>
   <div class='col'>2</div>
   <div class='col'>3</div>
</div>

By default all columns have a flex value of 1, so they will all have the same width. You can give columns a specific flex value using specific classes that ChocolateChip-UI uses. These are as follows:

<div class='flex-1'></div>
<div class='flex-2'></div>
<div class='flex-3'></div>
<div class='flex-4'></div>
<div class='flex-5'></div>
<div class='flex-6'></div>
<div class='flex-7'></div>
<div class='flex-8'></div>
<div class='flex-9'></div>
<div class='flex-10'></div>

 

Live Example using “flex-1”

By giving columns various flex values you can create complex layouts for you app

 

Columns Based on Total Flex Value of 10

 

Equal Height Columns

As we mentioned, columns in the same grid row are always equal height.

 

Column Gutters

You can give your grid, both rows and columns, some preset gutters. To do so you can put one of two classes on the grid row: gutter-5 or gutter-10.

 

If you want some other gutter value, you can create your own class. Below is a possible class you could use:

.gutter-15 {
   margin: 15px;
}

Then you can use your class on your grids like with our own gutter classes.

Centered Grids

You can center the columns to the grid row, creating space around them, using the center class:

 

Fixed Width Columns

By their nature, flex columns are dynamic. They flex to fill whatever space is available on screen. This is great for mobile devices, such as when the user changes the orientation of a mobile device from portrait to landscape. You can rest assure that you layout will work equally well in all situations. However, sometimes you may need to give a particular column a fixed width. You can do this by giving it a special class. On this class you will define the width you want, and you will also set its flex value to 0:

.fixedWidth {
  -webkit-flex: none;
  flex: none;
  width: 250px;
}

Putting the above class on a column will force it to always have a width of 250 pixels. Any other columns in its grid row will use whatever space is left after that column is rendered. Below is an example of fixed columns:

 

Learning About Layout in ChocolateChip-UI

To learn more about layouts in ChocolateChip-UI, visit the following links:

• Layout
• Grids
• Cards
• Buttons

From ChocolateChip-UI to TruckJS

It all started with a simple DOM library for mobile use. I named it ChocolateChip, because I liked chocolate a lot and because if was really small.

On this blog, one post after another, I created prototypes of how to replicate iOS with markup, CSS and some JavaScript. After many posts I had the makings of a framework. These were all random, so I needed to bundle it all together. ChocolateChip was the glue, so I named the framework ChocolateChip-UI. That was six years ago.

As time passed, ChocolateChip-UI grew to accommodate new layouts, widgets and support for Android and Windows Phone, and could switch out to use jQuery. And then we made jQuery the default. So many changes…

Recently I was thinking about where to take ChocolateChip-UI. It seemed like it had everything it needed. I went and got some specialty chocolates and some good coffee and sat down to brood. I thought about the difference between Backbone and Angular. Backbone at least had models, where as Angular does not. Because of that, Angular has to run constant polls to compare the state of the raw JavaScript objects. The more you use Angular’s data binding, the more polls you have going and everything starts slowing down, especially on mobile.

In a system with models, it’s the model’s responsibility to inform the system when its data is modified. The system does not need to be running polls to observe the model. It should provide encapsulation and abstraction, protecting data from accidental modification by the system/developer. And so, I began tinkering until I had a model. I created some uses for it, fine tuned it, added some more features, etc. It was an object factory that returned a new model instance without having to use the “new” keyword.

I was pretty happy that ChocolateChip-UI would now have a model, but that got me to thinking – maybe ChocolateChip-UI could use a controller module too. So I looked at how the competition was dealing with controller behavior. And, in the end, I decided to take a completely different approach. I was creating a framework for mobile apps. These have unique challenges because they are mobile. So decided to base this work around the concept of mediators. These sit in the middle of an app’s interactions, like this image:

I used ChocolateChip-UI pubsub methods as the basis for my mediator module. But mediators have a number of unique features. You can control how many times a mediator can run or when it can run. Then I went back to my model code and updated it so that a model would automatically notify the mediator layer that it had changed. Like the model, this was also an object factory reusing the same pattern.

So now I had a model-controller for ChocolateChip-UI. I had always thought of ChocolateChip-UI as just a view. But it was a very passive and disconnected view because it had nothing built into it to enable working with other MVC frameworks. This meant that to use these required various workaround. It was not optimal, but it worked. I started thinking that maybe I could create a view module in the same way that I had create the model and controller modules. My idea was to enable a more component like pattern. I popped out the template parser from ChocolateChip-UI and put it into a new object factory. The object that this created knew about an element that it was associated with. This would be the container for any template rendering. It would automatically parse the element’s child nodes to extract a template. Or you could define a template in the view’s initializer. And you could define events in the view. And finally, you could bind a view to a model.

In the background the code would generate a mediator that would listen for any changes to the bound model. When a change would occur, it would update the rendered view.

So, I now had a model view controller framework. They had no dependency on ChocolateChip-UI. At the moment they were running on top of jQuery. While I was creating these modules, I kept comparing functions to mechanical ones. So, when I was done I realized I had something totally different from ChocolateChip-UI. It was time for a name change. I thought about vehicle names, but in the end I settled on TruckJS because trucks are powerful, they can carry heavy loads. You can count on trucks.

I then though I’d give a shot at rewriting ChocolateChip using the new patterns I used for TruckJS. But I was going to do one important thing. The new DOM library would be identical in API to jQuery so developers wouldn’t have to learn anything new to use it. In the end I had a DOM library less than half the size of jQuery with 90% of the functionality. But instead of the old Ajax syntax, I used the new Fetch API for remove server interaction. And instead of jQuery’s deferred object implementation, I went with the ECMAScript 6 Promises API. Fetch uses ES6 promises, so this was a perfect fit.

Then I though about routing, so I built a router. Then I added in data formatting and form validation. After that I began porting the widgets over. I wound up rewriting every single widget, changing their behavior to return objects and enabling the form-based widget to work with the new form validation.

TruckJS is a standalone framework without any dependency on other libraries or frameworks. You do not need jQuery, etc. It provides everything you need to build mobile web apps. But, if you really want to use jQuery with Truck, you can. You just have to load jQuery first. When Truck detects jQuery, it switches to using it instead of its own DOM library. I’m not sure why anyone would want to use jQuery. Its bulky and slow on mobile devices. But there’s always those special guys who think something can’t be perfect unless jQuery is loaded.

I’ve put up a site, TruckJS.io, with lots of documentation. And it’s available on Github. If you like ChocolateChip-UI, I think you’re really going to like TruckJS. It gives you the same tradition of great layouts and widgets for creating cross platform mobile web apps.

 

ChocolateChip-UI 3.0.3 Supports jQuery

As of version 3.0.3, ChocolateChip-UI now supports jQuery 2.0.3. We tried earlier versions of 2.x, but there where performance issues that prevented us from offering it as an option. jQuery 2.0.3 has proven to be a good option for mobile Web apps as far as size and speed. As such, going forward we will be supporting jQuery with the latest version of ChocolateChip-UI.

This means you can use other frameworks or plugins with jQuery dependencies and still take advantage of the great features in ChocolateChip-UI for creating cross-platform Web apps for iOS 7, Android Jelly Bean and Windows Phone 8.

For more information about jQuery support in ChocolateChip-UI, please visit ChocolateChip-UI.com and read the documentation for jQuery Support.

Back and Next Buttons Revisited

So, many moons ago, I wrote a post about how to make iOS style “Back” and “Next” buttons using CSS. Only problem was that technique used a span tag to create the pointer. For some reason it slipped my mind that I could use the CSS generated content property with the :before pseudo-element. Rather than being bothered with having to stick spans into buttons to turn them into pointers, I now introduce you to a more sane method of creating “Back” and “Next” buttons with just CSS.

Now you might be wondering how we would generate the node we need to style as a pointer. Using the CSS content property we can generate automatic content inside the button, removing the necessity of having a tag sitting there for no other reason. Unfortunately the content property only generates alpha-numeric content. We can’t use it to create a span or any other tag. But we don’t have to use a normal character as the content. A space will suffice to create the desired pointer. We make it display as a block element and then style it as if it were a normal pointer. Problem solved.

To accomplish our goal we need to switch out any reference to < span.pointer for :before

So, here’s the styles for a normal button, followed by the styles for the pointer.

.button.bordered.back:before {
	content: " ";
	display: block;
	z-index: 0;
	background-image: 
		-webkit-gradient(linear, left top, right bottom, 
			from(#92a1bf), 
			color-stop(0.3, #798aad),
			color-stop(0.51, #6276a0), 
			color-stop(0.51, #556a97), 
			color-stop(0.75, #566c98),
			to(#546993)); 
	border-left: solid 1px #484e59;
	border-bottom: solid 1px #9aa5bb;
	-webkit-border-top-left-radius: 5px;
	-webkit-border-bottom-right-radius: 5px;
	-webkit-border-bottom-left-radius: 4px;
	height: 23.5px;
	width: 23.5px;
	display: inline-block;
	-webkit-transform: rotate(45deg);
	-webkit-mask-image: 
		-webkit-gradient(linear, left bottom, right top, 
			from(#000000), 
			color-stop(0.5,#000000), 
			color-stop(0.5, transparent), 
			to(transparent));
	position: absolute;
	left: -9px;
	top: 2.5px;
	-webkit-background-clip: content;
}
.button.bordered.back:hover:before, .button.bordered.back.touched:before {
	background-image: 
		-webkit-gradient(linear, left top, right bottom, 
			from(#7d88a5), 
			color-stop(0.3, #58698c),
			color-stop(0.51, #3a4e78), 
			color-stop(0.51, #253c6a), 
			color-stop(0.75, #273f6d),
			to(#273f6d)); 
}

You can check out the finished version online or download the source code.

ChocolateChip-UI

Learn more about ChocolateChip-UI

ChocolateChip.js is now ChUI (pronounced “chewy”)

In the course of time I’ve created a lot of POCs (proof of concepts) on this blog. Some where fairly good, some where flaky one-offs. In the process I also created a small JavaScript library — ChocolateChip.js — that I used for actualizing most of these demos. Seeing how many people come to this humble blog in search of solutions to their development projects, I though about taking all the good ideas in these demos and putting them together as a one stop solution for mobile Web app development. I’m therefore today announcing ChocolateChip-UI. At present it’s a beta, but I intend to keep developing it to add more useful features over time. I

ChocolateChip-UI consists of ChocolateChip.js plus two other files: ChUI.js and ChUI.css. ChUI.js is a collection of JavaScript methods built on top of ChocolateChip.js. ChUI.js provides controls and widgets enlivened with behaviors needed for Web app development. ChUI.css is the magical CSS that makes simple markup look and act in a amazing way.

ChocolateChip-UI also introduces a new concept: WAML (Web App Markup Language). This is a specialized set of tags and attributes that get around the limits of HTML. HTML tags are really about creating documents, like books and other text publications. In contrast, WAML is a collection of tags and attributes that make sense for Web app development. ChUI.js and ChUI.css are built around the implementation of WAML as the key to how ChocolateChip-UI works.

WAML takes the paradigms of mobile application development and transfers those over to the development of mobile Web apps. There’s no need to smother and bury HTML tags with tons of classes trying to make markup meant for publishing documents work for creating applications. At the same time, WAML is really just a superset of HTML5 markup. You can therefore use any HTML5 tags and attributes along with WAML and ChocolateChip-UI to implement your solution.

ChocolateChip-UI is about making Web app development more straightforward. It provides common controls which have built in functionalities. Instead of having to figure out how to build these yourself, you can spend that time providing the data you want through ChocolateChip-UI’s controls easy to use controls.

The interfaces and controls which ChocolateChip-UI provides are all created using only markup and CSS. No images are required, no gifs or pngs. This means that everything you build with ChocolateChip-UI will be resolution independent. It will look good on a handheld mobile device and on a big HDMI screen. ChocolateChip-UI also provides a set of 52 SVG icons for use with buttons. Because these are vector based, they too are resolution independent. ChocolateChip-UI is therefore the first resolution independent mobile Web app framework.

The Holy Grail of Mobile Layout

The Holy Grail of Mobile Layout

The perfect mobile layout should allow the presence of a navbar, a content area and maybe a footer toolbar. This layout should resize with orientation changes such that the bottom toolbar is always at the bottom and the content area resizes its width and height to fit the new dimensions. This is tough because, first off, there is not fixed positioning, and secondly, there is no single finger scrolling in Webkit for CSS defined scrolling regions.

it’s taken me a while to come up with a solution to these layout requirements. Any of you trying to make a Web app feel like a native mobile app know the frustration of trying to deal with orientation change and complex layout. Not being able to make a toxic mix of viewport meta tags, CSS media queries and onorientationchange events work to give me a fluidly resizing layout when the mobile device’s orientation changed drove me to despair. I compensated early on using JavaScript to resize everything, except I had to pay a performance penalty, especially if an app was complex. After a lot of sweat and blood and endless trial and error, I finally came up with a CSS only solution to making a layout that resizes smoothly with orientation changes.

So, the first big mistake, one that everyone and their brother recommends, is not to use the meta tag viewport settings of device width and device height. Instead, just use the initial scale, maximum scale and user scalability:

<meta name="viewport" content="initial-scale=1.0, maximum-scale=1.0, user-scalable=no">

The reason is that no matter what you set the width and height to with CSS, Webkit insists on thinking the document should have the height and width from when the document first loaded. This happens even if you use a CSS value of “width: 100%”. Switching from portrait to landscape, you’ll see that your layout fails to fill the width of the landscape screen:

iPhone Portrait Layout

iPhone Landscape Layout

As you can see with the landscape orientation the layout fails to fill the available width of the device. This app has its root elements’ widths set to 100% and a meta tag of:

<meta name="viewport" content="width=device-width; height=device-height; initial-scale=1.0, maximum-scale=1.0, user-scalable=no">

Removing the device width and height from the meta tag will give us the fluid layout we seek:

<meta name="viewport" content="initial-scale=1.0, maximum-scale=1.0, user-scalable=no">

With the above change, when switching this document loaded in portrait orientation to landscape mode orientation we get a resized document the way we would like:

iPhone Landscape Layout Correctly Resized

OK, so we have a navbar on the top, but what if we want a toolbar on the bottom? No problem. We can accomplish that with a few adjustments. To fix the toolbar to the bottom we’ll use absolute positioning. One thing to remember about this layout technique, if you are going to have navbars or toolbars, you’ll need to adjust the height of the content so that it fits the space left over by the bars. I do this using a class or class on the content section. A navbar or toolbar has a default height of 45 pixels, so you’d want to position your content that distance from a navbar or toolbar. To make this work, we’re going to have to use absolute positioning on the content section and on any bottom toolbar. One thing rather frustrating about using absolute positioning on block elements is that it causes them to collapse to what their content is, even with they have a width of 100%. We can get around this by setting the appropriate side to 0 pixels. For a bottom toolbar this could be something like this:

.toolbar.bottom {
    height: 45px;
    position: absolute;
    bottom: 0;
    left: 0;
    right: 0;
}

With the above measurement, we have a toolbar that is 45px high fixed to the bottom and stretching from side to side. This toolbar will maintain these dimensions and placement even with an orientation change. We’ll need to do something similar for the content area. Like I said, we’ll need to use absolute positioning and account for any space taken up by a nav or toolbar. I use classes to do this, such as “withNavbar”, or “withNavbar withBottomToolbar”, or “withBottomToolbar”. The CSS for these classes would be something like this:

/* Definition for content area that fills the screen: */
#content {
    position: absolute;
    top: 0;
    right: 0;
    bottom: 0;
    left: 0;
}
#content.withNavbar {
    top: 45px;
    right: 0;
    bottom: 0;
    left: 0;
}
#content.withBottomToolbar {
    top: 0;
    right: 0;
    bottom: 45px;
    left: 0;
}
#content.withNavbar.withBottomToolbar {
    top: 45px;
    right: 0;
    bottom: 45px;
    left: 0;
}

We also need to make sure that the document’s content doesn’t bleed down below the bottom toolbar. We do this by wrapping the entire document in a div. I give it an id of #main:

<body>
	<div id="main">
		<div class="navbar">
			<h1>The Title</h1>
		</div>
		<div id="content" class="withNavbar withBottomToolbar">
			<ul class="table-view">
				<li><span class="title">Item One</span></li>
				<li><span class="title">Item Two</span></li>
				<li><span class="title">Item Three</span></li>
				<li><span class="title">Item Four</span></li>
				<li><span class="title">Item Five</span></li>
				<li><span class="title">Item Six</span></li>
				<li><span class="title">Item Seven</span></li>
				<li><span class="title">Item Eight</span></li>
			</ul>
		</div>
		<div class="toolbar placement-bottom">
			<div class="button">Edit</div>
			<div class="button">Save</div>
		</div>
	</div>
</body>

The styles for the div#main would be:

#main {
	width: 100%;
	display: -webkit-box;
	margin: 0 0 0 0;
	padding: 0;
	height: 100%;
	-webkit-box-orient: vertical;
	-webkit-box-align: stretch;
	-webkit-box-sizing: border-box;
	-webkit-transition: all 0.25s  ease-in-out;
	overflow: hidden;
	position: absolute;
	top: 0;
	left: 0;
	bottom: 0;
	right: 0;
}

Although the above arrangement solves the layout problems for mobile interfaces with automatic adjustment with orientation changes, it presents a new problem. That’s the problem of making the data in the content area reachable by the user. You’ll need to add a way for the user to scroll just that area. Normally the whole document would scroll with the browser, with the result that the top part would scroll out of view. Since we’ve fixed the toolbar to the bottom and have the height of the content area control by absolute position affecting its top, right, bottom and left, we need to provide a way to enable scrolling. Actually we could do this very simply using the CSS overflow property: #content { overflow-y: auto; }. However, on mobile Webkit this requires a two finger gesture. Most users would not expect this. They’ll use a single finger and think that the content is not scrollable.

There’s a simple way to provide single finger scrolling with a JavaScript module called “iScroll” created by Matteo Spinelli. This is a great implementation which provides deceleration animation of the scrolling, just like iOS, Android and other mobile OSes. One thing to be aware of when using iScroll, you don’t use it directly on the content container as this would cause that contain to slide up and down over the navbar and toolbar. Instead you’ll need to add an extra div tag inside it and use iScroll on that. iScroll will then enable scrolling of the div inside the content div. There are a number of options, my favorite is for desktop simulation. Here how you would do it:

<div id="content" class="withNavbar withBottomToolbar">
    <div id="scrollwrapper">
        <!-- Content In Here -->
    </div>
</div>
.........

<script type="text/javascript">
    var scroller = new iScroll(’scrollwrapper’, { desktopCompatibility : 
        true });
</script>

This gives you a basic mobile Web interface with a top navbar, a bottom toolbar and a content area with scroll that resizes quickly with orientation change.

Since publishing this article, I’ve created a framework for creating mobile Web apps. It uses many of the techniques explored in my blog posts. You can learn more about the framework by visiting ChocolateChip-UI.com. There are demos you can try in your desktop browser or on your mobile device.

Practical Examples of the Flexible Box Model

If there’s one thing in CSS3 that really gives me warm fuzzies, it’s the flexible box model. What it does is provides a way to create many layout affects that otherwise would require workarounds with float or positioning or both. And there are many types of layouts that without out it would require JavaScript to create the same effect. Sadly, at present the flexible box model is only implemented in Firefox, Safari and Chrome. As it turns out, Microsoft’s IE team approached the W3C standards teams with a proposal for a grid-based layout module. It appears to be based on experience implementing the Silverlight grid layout system, which is a good thing. Silverlight grids are a powerful layout tool and is the most important feature for creating interfaces with Silverlight. Because this grid proposal is very solid, the W3C has dropped all further work on the flexible box model and instead are moving ahead in finalizing the draft for the grid layout module. If you read that speck, you’ll realize how much we need a CSS grid system. For now, however, we do have the flexible box model working in Firefox, Chrome and Safari, as well as on mobile Webkit, which includes iOS, Android, WebOS, and Blackberry. This means that if you’re doing mobile development you can use the flexible box model today to help you implement your layout needs. Once you start using the flexible box model, you’ll hate having to go back to situations where it’s not available.

The flexible box model got a lot of attention from a post by Alex Russell back in August 2009. The good thing is that in his post he showed some of the practical things that the flexible box model addresses and provided a stylesheet with a small subset of the flexible box model properties that people could use right away. His post has been replicated and linked to all over the Web. Unfortunately a lot of people who are not familiar with the spec for the flexible box model are under the impression that what’s in his stylesheet is it. So, with that in mind I’m loosing another stylesheet on the public that allows a greater utilization of the flexible box model. I’ll also give you a layout where a lot of these are in use so that you can see how they work.

Before I get going let me say this. The flexible box model is not perfect. It isn’t supported by IE9 or Opera. And now it looks like the W3C is bypassing it for a more robust grid system. Personally, I’d like to see something like the Silverlight grid system combined with aspects of the flexible box model. That would be the best of both approaches. If you’re targeting the mobile Web, in other words you intend on target Webkit browsers, you can expect robust support for the flexible box model. I doubt it’s going to go away just because the W3C doesn’t include it in any recommendation. I expect it to remain in Webkit for the foreseeable future because it’s already been in use for a number of years.

The flexible box model is about the way a collection of child nodes gets laid out in relation to each other and their parent node. To achieve its layout goals, the flexible box model offers properties for the parent node as well as for the child nodes that determine the layout characteristics of the child nodes.

To work with the flexible box model you need to define its display property to box:

#myDiv {
    display: -webkit-box;
    display: -moz-box;
}

Elements with display set to box can order their child nodes in two orientations: vertical or horizontal. If no orientation is provided, the box defaults to vertical:

#myDiv {
    display: -webkit-box;
   -webkit-box-orient: horizontal;
    display: -moz-box;
    -moz-box-orient: horizontal;
}

Besides defining the orientation, you can also designate what direction the child elements stack, either normal or reverse. A box direction of reverse means the elements display in the opposite order in which they physically appear in the document, so that the first would be last and the last first. By default the box direction is normal stacking order.

Then there is the packing order. This allows you to define how the child elements are packed inside the parent. By default they are packed in from the beginning of the parent. For vertical orientation this is the top and for horizontal orientation this is the left. Other packing orders are end, center and justify. If the packing order is end, for a vertical orientation the elements will stack up from the bottom. For a box with horizontal alignment, the boxes will stack from the right. A packing order of center means that for vertical orientation the child elements will be centered vertical with any left over space displayed equally at the top and bottom. For horizontal orientation a packing order of center means the elements are centered horizontally with left over space equally divided on the left and right. A packing order of justified means that any available space is spread equally between the child elements, for vertical orientation this is vertical spacing, and for horizontal orientation this is horizontal spacing.

A box with alignment set to "start"

A box with alignment set to "end"

A box with alignment set to "center"

And elements can have their box alignment defined. The default is stretch. This stretches the dimensions of the child elements to fill the parent box. If the alignment is start, for horizontal orientation the elements are aligned to the top of the parent. If the orientation is vertical, they are aligned to the left of the parent. An alignment of end will align elements with horizontal alignment to the bottom of the parent. With vertical orientation they will be aligned to the right of the parent. An alignment of center will align horizontal elements along the horizontal center of the parent and for vertical orientation it will center them along the vertical center of the parent. There is also an alignment of baseline, but this is only for horizontal orientation. It aligns the elements along their horizontal baseline.

Then there is a set of properties that you can define on the child elements. These are flex and ordinal group. The flex property tells the browser how to deal with the dimensions of the child element. If you have three child elements and they don’t fill their parent, giving one of the a value of flex:1 will cause that element to take up all the left over space. If you gave one element flex: 1 and another one flex: 2, the available space would be divided up such that the element with flex: 1 would get one third of the available space and the element with flex: 2 would get two thirds of the available space. Of course the element with no flex value would default to whatever its width is. The ordinal group property allows you to designate groups of elements so that they appear in a different order than their document order. I’m not so sure about the practical use for this property. I’ve thought about it for months and have not been able to come up with a use case where I would need it. But it’s there.

I’ve created a flexible box model stylesheet with classes for all the various box properties. This means you can add the classes to an element to build out the definition you need for an element. I’ve also put together a interactive test case where you can dynamically toggle a number of the box properties and see their effect in real time. You can try it out online if you’re using Firefox, Chrome or Safari, or you can download the working example to dissect and learn.

Multiple Background Images & Animation

Multiple background images allow one to build up complex layered visual effects using one element. In the past such effects would require each background image rendered inside its own element, these being either nested or stacked on top of each other with positioning. When I was working on creating the progress bar example, I ran to a problem where I couldn’t seem to get a specific layer to animate. I could figure out what was going on. Recently, after looking at the code and fiddling with it, I found out what the problem was. Animation of background images works a little differently from other CSS properties. Because multiple backgrounds consists of multiple definitions, it works more like other properties with multiple property values: margins, borders, padding. For example, you can animate all borders on an element, otherwise you can animate a single border. Unfortunately, with multiple backgrounds there is no easy way to single out an individual background image for animation as you can with a border, margin or padding property.

Now the thing to bare in mind when dealing with multiple background images is the stacking order. The first image defined is the topmost image, and the last image defined is the bottommost. Same thing with animations for multiple background images. So, in the case of the animated progress bar, I wanted to animate the bottommost gradient image which consisted of white slanted bars. The rest of the progress bar was fine stable. Well, except that I used a single animation which would instead affect the topmost layer, making it impossible for me to get the stacking layer they way I wanted. Here is the gradient:

background-color: rgb(56,138,213);
background-image:
	-webkit-gradient(linear, 18 0, 0 10,
		color-stop(0.23, rgba(255,255,255,0)),
		color-stop(0.3, rgba(255,255,255,0.8)),
		color-stop(0.3, rgba(255,255,255,1)),
		color-stop(0.7, rgba(255,255,255,1)),
		color-stop(0.7, rgba(255,255,255,0.8)),
		color-stop(0.77, rgba(255,255,255,0))),
	-webkit-gradient(linear, 0 0, 0 100%,
		color-stop(0, rgba(255,255,255,.8)),
		color-stop(0.45, rgba(255,255,255,.05)),
		color-stop(0.55, rgba(0,0,0,.05)),
		color-stop(0.85, rgba(0,0,0,.2)),
		color-stop(0.98, rgba(0,0,0,.5))),
	-webkit-gradient(linear, 0 0, 0 100%,
		color-stop(0.20, transparent),
		color-stop(0.20, rgba(255,255,255,.5)),
		color-stop(0.32, rgba(255,255,255,.5)),
		color-stop(0.32, transparent));

To animate this, I was using a key-frame animation that moved the x axis of the background gradient across the element from 0 to 100% in a loop:

@-webkit-keyframes progressBarAnim {
	0% { background-position-x:  0%; }
	100% { background-position-x: 100%; }
}

If you ran my previous example of the progress bar, it animates as you would expect, except one problem. The topmost gradient layer is the part that should be on the bottom. Yet every time I positioned it to where I wanted it, it wouldn’t animate. In my mind I was thinking that the key frame animation defined above would be animating all the background images. Actually, it doesn’t. It only animates the topmost image background, which is why I had to have the part I wanted to animate on top. Unfortunately I didn’t figure this out until recently. After a lot of fiddling and digging into what was going on with animation of multiple background images, I finally sorted it all out. When defining animations you have to start from the top and work you way down. You don’t have to include an animation for every background image, only down to the layer you need to. But you do have to include all the layers above the animated one up to the topmost one. So here’s my new stacking order for the CSS the way I actually wanted it:

background-color: rgb(56,138,213);
background-image:
	-webkit-gradient(linear, 0 0, 0 100%,
		color-stop(0, rgba(255,255,255,.8)),
		color-stop(0.45, rgba(255,255,255,.05)),
		color-stop(0.55, rgba(0,0,0,.05)),
		color-stop(0.85, rgba(0,0,0,.2)),
		color-stop(0.98, rgba(0,0,0,.5))),
	-webkit-gradient(linear, 0 0, 0 100%,
		color-stop(0.20, transparent),
		color-stop(0.20, rgba(255,255,255,.5)),
		color-stop(0.32, rgba(255,255,255,.5)),
		color-stop(0.32, transparent)),
	-webkit-gradient(linear, 18 0, 0 10,
		color-stop(0.23, rgba(255,255,255,0)),
		color-stop(0.3, rgba(255,255,255,0.8)),
		color-stop(0.3, rgba(255,255,255,1)),
		color-stop(0.7, rgba(255,255,255,1)),
		color-stop(0.7, rgba(255,255,255,0.8)),
		color-stop(0.77, rgba(255,255,255,0)));

And here’s the key frame animation, notice the extra animations for the upper layers:

@-webkit-keyframes progressBarAnim {
	0% { background-position-x:  0%, 0%, 0%; }
	100% { background-position-x: 0%, 0%, 100%; }
}

So remember, in the above key frame animation the first value if for the topmost layer and the last value is for the bottom most layer.

To show the difference in the stacking order, I changed the white stripe that goes horizontally across the progress bar to black. In the wrong stacking order version you can see that its actually behind the white stripes. The corrected version with multiple background animations has it correctly on top:

progress bar with incorrect stacking order

progress bar with correct stacking order

Of course, the final product would not be a black streak, it would be white. This was just to show how the stacking order was previously wrong.

Subpixel Rendering

You know the problem. No matter what you do you can’t get two elements to line up properly. The connecting points are always off by one pixel.

Everyone doing Web development at some point or other comes across a layout problem where no matter what you do, you can’t get two elements to align perfectly. One or the other is always off by one pixel. I was pulling my hair out try to get the pointers on the back and next buttons to align perfectly. They just didn’t look perfect. Worse still, when I used the browser’s “Zoom In” command from the view menu, I could clearly see that the lines did not connect properly.

After hours of fiddling with element sizes and positioning, I was on the verge of giving up. It was then that I remembered similar layout problems that I dealt with when doing Silverlight development. Silverlight is Microsoft’s vector-based, Flash killer/non-killer plugin for creating RIAs. For whatever reason, versiond before 4.0 had terrible problems with exact positioning of elements, causing frequent one pixel disconnects when rendered to screen. The only way to resolve this was to use subpixel rendering. This was accomplished by positioning an element by using partial pixel values, such as 1.5 or 1.25. This would force Silverlight to output the element with subpixel rendering, eliminating the visual disconnect.

OK, so what the heck is subpixel rendering? You experience it everyday with the browser’s font smoothing. You know it as anti-aliasing. The browser looks at the bézier curves of the font and when it sees that a line passes though a pixel, it looks at how much of the pixel is intersected. Depending on the percentage, the browser outputs a percentage of the font’s color. Less means the pixel gets less of the font’s color. For the human eye this creates the illusion of smoother curves.

You can use this same technique to trigger subpixel rendering on an element by giving it percentage-based position, or percentage-based dimensions. Here are some examples:

.button {
    position: absolute;
    left: 0px;
    top: 2.5px;
    height: 23.5px;
    width: 23.5px;
}

Here’s a image of my next button with the browser zoomed in. As you can see the pointer doesn’t line up perfectly with the rest of the button. This caused a slightly noticeable disconnect at normal size as well.

Now here’s the same button with the pointer using position set to top: 2.5px;:

Subpixel rendering solved the connect problem I had at all zoom levels, including at normal size. Depending on your problem, subpixel positioning may be enough, or subpixel dimensions may be enough, or you may need to do both. Using subpixel values can help resolve problems when your layouts are not coming out pixel perfect.

You can try this out online or download the source code.

Today’s new technology terms:

subpixel:

   A pixel rendered with a shade of an adjacent element’s color to make it appear as if the element occupies part of that pixel’s space.

subpixelate:

   To force the browser to render an element with subpixel values.

subpixelation:

   The act of forcing an element to render with subpixel values or the condition of being rendered with subpixel values.

Making an iPhone Switch Control without Images

Works on desktop Safari, Chrome and Firefox, iPhone, iPod Touch and iPad.

On the iPhone and iPad, Apple uses a control called switch. It’s actually a different take on the checkbox. Like radio buttons, checkboxes do not lend themselves to touch interfaces, especially guys with fat fingers, cough! Instead of making us suffer with those dinky checkboxes, Apple uses a more visual cue to what the user is actually doing, switching something on or off. As a matter of fact, that’s exactly how the control is labeled: “on” or “off”. They’re really easy to use, just swipe your finger to throw the switch, done. In case you’re not sure what I’m talking about, here they are:

OK, so all the mobile Web frameworks have a switch control. And I hate them all. They either do an instant switch between the on and off state, using an image sprite, or they do this really lame thing where they animate the horizontal background position of the image on a checkbox with its default styles removed. None of those implementations feels the same as when you swipe the switch control in a native iOS app.

So what am I going to do? I tell you, I’m going to throw the friggin’ image out and build the whole control from scratch using just HTML, CSS3 and some JavaScript to make it work. Bada-bing! To start with, here’s the basic markup for a checkbox:

<div class="checkbox unchecked" id="sleepSwitch">
	<div class="on">ON</div>
	<div class="thumb"><span></span></div>
	<div class="off">OFF</div>
	<input type="checkbox" value="ZZZZZZZZ!" offvalue="But, I need more sleep!">
</div>

As we did when we created iPhone style radios buttons, we’re using real checkboxes in our iPhone switch controls. And like in the radio button example, we’ll set the checkbox input’s display value to “none”. We’ll use CSS3 properties to style the markup to look like a real iOS switch control and we’ll attach event listeners to set the input checkbox’s check state to true or false, depending on whether we want it to be selected or not.

To create this switch control we’ll need to style the frame named “checkbox” with rounded corners. Notice that the markup above contains three parts: the on state, the thumb and the off state. The rounded frame will only be wide enough to show one state plus the thumb. Using CSS3 transitions and transforms, a click or touch will cause the three elements to side back and forth within the rounded frame. For positioning the switch’s elements and sliding them back and forth we’re going to use CSS3 3d transforms on the x axis. Here is the CSS to make this happen:

/* Checkbox */
.checkbox {
	display: -webkit-box;
	-webkit-box-orient: horizontal;
	-webkit-box-pack:justify;
	-webkit-box-sizing: border-box;
	-webkit-tap-highlight-color: transparent;
	width: 94px;
	overflow: hidden;
	-webkit-border-radius: 6px;
	text-align: center;
	line-height: 28px;
	cursor: pointer;
	-webkit-user-select: none;
	position: absolute;
	right: 10px;
	top: 7px;
}
.checkbox > input[type="checkbox"] {
	display: none;
}
.checkbox .thumb {
	-webkit-border-radius: 7px;
	position: relative;
	z-index: 3;
	border: solid 1px #919191;
	-webkit-transition: all 0.125s  ease-in-out;
	-webkit-transform: translate3d(0px,0%,0%);
}
.checkbox .thumb span {
	display: block;
	-webkit-box-sizing: border-box;
	height: 25px;
	width: 38px;
	border-top: solid 1px #efefef;
	-webkit-border-radius: 6px;
	background-image: -webkit-gradient(linear, left top, left bottom, from(#cecece), to(#fbfbfb));
	border-top: solid 1px #efefef;
	position:relative;
}
.checkbox .on {
	color: #fff;
	background-image: 
		-webkit-gradient(linear, left top, left bottom, 
			from(#295ab2), 
			to(#76adfc));
	width: 54px;
	padding-right: 4px;
	border: solid 1px #093889;
	-webkit-border-top-left-radius: 6px;
	-webkit-border-bottom-left-radius: 6px;
	margin-right: -6px;
	height: 25px;
	-webkit-transition: all 0.125s  ease-in-out;
	position: relative;
	-webkit-transform: translate3d(0px,0%,0%);
}
.checkbox .off {
	color: #666;
	background-image: -webkit-gradient(linear, left top, left bottom, from(#b5b5b5), color-stop(0.50, #fff));
	width: 54px;
	padding-left: 4px;
	border: solid 1px #a1a1a1;
	-webkit-border-top-right-radius: 6px;
	-webkit-border-bottom-right-radius: 6px;
	margin-left: -6px;
	height: 25px;
	-webkit-transition: all 0.125s  ease-in-out;
	position: relative;
	-webkit-transform: translate3d(-54px,0%,0%);
}
.checkbox.unchecked .thumb {
	-webkit-transform: translate3d(-54px,0%,0%);
}
.checkbox.checked .thumb {
	-webkit-transform: translate3d(0px,0%,0%);
}
.checkbox.unchecked .on {
	-webkit-transform: translate3d(-60px,0%,0%);
}
.checkbox.checked .on {
	-webkit-transform: translate3d(0px,0%,0%);
}
.checkbox.unchecked .off {
	-webkit-transform: translate3d(-54px,0%,0%);
}
.checkbox.checked .off {
	-webkit-transform: translate3d(6px,0%,0%);
}
/* For Very Important changes, use the orange checkbox */
.checkboxBase.important .on {
	border: solid 1px #d87100;
	background-image: -webkit-gradient(linear, left top, left bottom, from(#e75f00), color-stop(.5, #ff9c12));
}
/* End Checkbox */

To make the switch more realistic, I’m transforming all three pieces of the switch at the same time. This gives the switch a more realistic feeling. Notice the comment in at the end of the above CSS about the important class. You can use this to indicate a switch that makes a very important change. This class changes the default switch’s blue color to bright orange. This is the color Apple uses to show that a switch’s action is very important.

Having the CSS defined for the look and animation brings us close to the finished control, but we need to write some JavaScript to make the switch interactive. The JavaScript needs to do two things: toggle the classes “checked” and “unchecked” on the switch, and toggle the checked value of the checkbox between true and false. I’m using the ChocolateChip JavaScript framework to do this. You can switch my code to whatever library you want. If you know basic JavaScript, it shouldn’t be hard. Here’s the JavaScript to make it happen:

Element.prototype.toggleSwitch = function() {
	if (this.hasClass("switch")) {
		if (this.last().checked === true) {
			this.last().checked = false;
			this.toggleClass("checked", "unchecked");
		} else {
			this.last().checked = true;
			this.toggleClass("checked", "unchecked");
		}
	} else {
		return false;
	}
};

The last() used in the code above is a ChocolateChip method to return the last child of the control, which happens to be the checkbox input. That way we can set its checked state to true or false.

Now that we have the code to setup up the switch control, we can make it work as follows:

$(".switch").forEach(function(checkbox) {
	checkbox.bind("click", function() {
		this.toggleSwitch();
	});
	
	checkbox.bind("touchstart", function(e) {
		e.preventDefault();
		this.toggleSwitch();
	});
}); 

That’s it to make the switch switchable. But to make it do something you’d need a bit more as well. In my example, I’m getting some values from the switch and outputting it to a response field like this:

$(".switch").forEach(function(checkbox) {
	checkbox.bind("click", function() {
		if (this.last().checked === true) {
			$("#switchResponse").fill(
				this.last().getAttribute("value"));
		} else {
			$("#switchResponse").fill(
				this.last().getAttribute("offvalue"));
		} 
	});
	
	checkbox.bind("touchstart", function(e) {
		if (this.last().checked === true) {
			$("#switchResponse").fill(
				this.last().getAttribute("value"));
		} else {
			$("#switchResponse").fill(
				this.last().getAttribute("offvalue"));
		}
	});
}); 

You can try this out online or download the source code.