Is the Google Web Toolkit Right For You?

An anonymous reader writes "The recently released Google Web Toolkit (GWT) is a comprehensive set of APIs and tools that lets you create dynamic Web applications almost entirely in Java code. However, GWT is something of an all-or-nothing approach, targeted at a relatively small niche in Web application development market. This article shows you what GWT can do and will help you decide if it's the best tool to use for your web development."

Google takes more ground...

I, for one, welcome our weather-reporting overlords.

Mingling of server and client code is "unusual"?

"The GWT takes an unusual approach to Web application development. Rather than employing the normal separation of client-side and server-side codebases, GWT provides a Java API that lets you create component-based GUIs and then compile them for display in the user's Web browser."

I think that's how ASP.NET components have worked for years too. So, I wouldn't say that it's unusual unless you're coming from a completely "my text editor is my development environment" world.

Re:Mingling of server and client code is "unusual"

Now I understand why computer science geeks don't "couple". They have a whole different meaning of the word from the rest of the world ;)

Does it have an "I'm feeling lucky" feature?

Does it have a "I'm feeling lucky" feature, where it will deploy your project for the first time without bugs?

Short answer: No.

GWT does have a couple of fairly significant flaws. First among them is its lack of provision for graceful degradation. Best practice in modern Web application development is to create pages that work without JavaScript, and then use it where available to embellish and add extra behavior. In GWT, if JavaScript isn't available, you won't get any UI at all.

In other words, if you want to make sure your site "just works", GWT isn't a good technology to use. If your management team is paying attention, that should pretty much stick a fork in this technology.

Accessibility?

Also, what's the state of the art in screen readers? Do they have sufficient JavaScript support? (*)

If not, using GWT for a corporate web site is just a lawsuit waiting to happen.

(*) This is not an idle question, I'm sincerely interested in real answers and pointers to software.

Re:Accessibility?

If I had mod points today, I'd boost you up. If you go by the ADA provisions*, I think you need to provide enough content in plain old HTML code and text (Javascript doesn't count) so that all the content is available inline and enough links, text boxes and other items (also in plain old HTML code) so that someone can navigate your entire site without any Javascript.

*Section 508 Standards for Electronic and Information Technology;
SubSection 1194

Re:Short answer: No.

Disabled users? Blind users? Screw 'em. If they want to be cripples, that is their business. Why should everyone else suffer?

Ever think that it's not always someone's *choice* whether they can use JavaScript enhancements?

P

Re:Short answer: No.

I disagree with your assertion. I often use lynx (a text-only browser) to access sites, both internal to my company and externally. Sometimes when you are connected via an SSH connection, console is all you have. It is very annoying to me when sites make use of javascript as the only method of navigating a site, especially when it detracts from normal functionality. It is not that difficult to make a very basic site that allows for at least a minimal level of functionality to a text-based user.

Don't even get me started on *flash* sites...

GWT feels like GUI programming

From a web developer point of view, and as the article states several times, the Google Web Toolkit uses an approach similar to Swing, code written for GWT feels like GUI code more than web code, while I don't have anything against this and I believe it does have audience, I think many would prefer a more web-like approach such as Prototype [conio.net] or the Yahoo UI library [yahoo.com].

Why I like YUI

The two things I like about the Yahoo! UI toolkit are that:

- it is a la carte
- it represents pretty advanced, human-readable Javascript, which has allowed me to learn advanced Javascript techniques

The first, a la carte, means that the entire web application does not need to be dedicated to YUI. Recently I incorporated a YUI DHTML window (Panel, in YUI-speak) into an existing ASP.NET application. It required only a few lines of Javascript and some HTML. Ditto for another application (JSF, this time), where I added an AJAX data lookup to dynamically populate sample values in a dropdown list. I love the DOM and Event libraries: they really clean up the cruddy job of adding small DHTML things to a website, regardless of how it's put together.

The second, that YUI is a learning experience, is also very important to me. I've come late to the DHTML game, and most books and tutorials on Javascript deal with the basics, and stuff like OO programming is either glossed over or omitted entirely. However, Javascript is a language that is suited to OO programming in a unique way, IMHO. By studing how the YUI toolkit works and seeing what Javascript is capable of, it has opened the doors for new ways of tackling problems and reducing the number of lines of JS code I write.

HBH

As if..

Using GWT is far closer to developing with Swing or SWT than the usual experience of Web application development,

As if GUIs in Swing or SWT weren't slow and clunky enough. Now Java developers have a slower and clunkier alternative. Yay!

-matthew

Feeling Inadequate

I had to stop after the third line... "Level: Advanced"

we just weren't right for each other

I mean, I tried to make it work, but I couldn't ever get her to give me a callback...

haxe

There's been several JS code generators recently. Apart from GWT there's also haXe [haxe.org] which is more ambitious. haXe can also be used on the Server Side and includes some facilities for Dynamic programming (whereas Java is strongly typed 100% of the time).

YAHOO UI Toolkit

We've already got an MVC LAMP architecture here at work and I have limited Javascript development until the framework was further along. But now the designers are getting edgy because I won't let them incorporate Javascript.

My issue was that I wanted a unified javascript library so we didn't have redundant code snippets all over the place and some way to standardize the development. So after a little searching, I found the Yahoo UI toolkit that is still in beta. It's a unified javascript library released under a BSD license that has been tested in multiple browsers.

No I have a unified library, tested in multiple browsers and standardized that all I have to do is drop into a directory and hook into the framework.

I know this is probably unrelated but since we were talking about toolkits, I thought this would be handy to mention this as well since it saved us a month of work.

Why not learn the tools instead

Hmm, all the examples in the article look pretty much the same than they were written in JavaScript, but in a more complex way. Why not actually learn the trade. Its easier to fix problems if you're working with the actual code that runs in the browser, not the "meta code".

Re:Why not learn the tools instead

Almost modded up, but hopefully someone else will take care of it for me.

Expanding on parent's point, a lot of problems I see in my short time in web development is that too many people are getting into it not by learning basics (like how to build a well-formed XHTML/HTML document with DTD and such, or how to make an image swap sources onmouseover or whatever) but by diving straight into frameworks. I understand the want (and need, in some case) to make programming of all flavors more non-programmer friendly, but without that base foundation we'll end up with a bunch of forums full of "how do i make it do this" questions that are elementary in nature and, even worse, a bunch of web apps that are riddled with problems in security, UI, or other. There's no harm in asking questions, but when everyone is asking the same question that is answered in chapter 2 of any good HTML book, that's a lot of wasted time.

I'm not saying everyone needs to learn how to build Slashcode from the ground up using only Notepad, Mountain Dew, and a bag of Doritos, but learning the basics first then going to a framework to speed up your work on complex projects would seem like a better option. It will almost always be cheaper and faster to write simple things in the base language, but so many are so fixed on frameworks they wouldn't know how to do that.

Google Map's GWT API 1.1.5

Hey, this is a bit of a shameless plug for my open source project, but it's on topic to GWT. Speaking to the subject of the post, regarding the utility of GWT, I point to my API and API like it so say there is probably a larger market for GWT than this fellow is speculating, and frankly from my experience with the interface I don't see how he could reasonablely charictarize it as all-or-none (the project google generates is example of how to integrate google GWT into your current HTML using named elements and replacing them....). Features: Fully implemented all the major classes including the new Geocoder and events. Available for download on SourceForge (http://sourceforge.net/projects/gwt/). For anyone interested in helping help please e-mail me (developers, testers, documenters welcome).

Article misses the point a bit?

I would never, ever, ever even _think_ of trying to write an Ajax application. I hate javascript way too much for that, and the whole asynchronous web-page dynamicness scares the shit out of me. However, I can stand writing GUI code in Java, and I already have some knowledge of how it works. GWT enables me to write super-cool Ajax websites without worrying a bit about javascript. I think this is the true power of GWT: it enables _application_ developers (rather than web developers) to write nice web applications without learning a whole new skillset. This is a win for developers because we can do cool things without learning a whole new trade, and it's a win for consumers because web apps can be written by skilled application developers instead of photoshop gurus.

Also, just because GWT provides some server-side machinery doesn't mean you have to use it. GWT will happily make asynchronous calls to any old URL, so you can write your backend in PHP or Perl or C or whatever you like. Personally, I've been playing with GWT using PHP/SQLite as my backend (following Juan Hurtado's great tutorial [googlepages.com]). It works great, and since I'm already comfortable with Java, PHP and SQL, it doesn't require a steep learning curve to create very nice web apps.

That said, I will certainly be taking a deeper look at TFA later and perhaps following its example to learn how to use the Java server-side machinery. Could be interesting.

GWT vs. Echo2

This seems like a useful article. At work we were just discussing the possibilities of using the GWT. Currently we use NextApp's Echo1 and are also pondering upgrading to Echo2 [nextapp.com]. The demo app for Echo2 is actually mighty impressive; I'd say moreso than GWT's. Though I wonder how much of it is just more time spent on fancy graphics as opposed to a better quality framework.

Why GWT Isn't A Good Framework

As a disclaimer, I'm a huge fan of Ruby on Rails, and not at all a fan of Java.

The problem with the GWT and other framworks like it as it ignores the reality that browsers today suck. IE's rendering engine is suckier than Monica Lewinsky holding a Dyson at the event horizon of a black hole. Firefox doesn't quite yet pass Acid2, but is as close to a reference platform as one can get. Safari shows promise, but it has a weak JavaScript environment that doesn't support things like ContentEditable. The whole problem is that the GWT assumes a much more stable platform than actually exists.

The real challenge for web application developers is that there are no frameworks (that I know of) that provide for things like fully semantic code, graceful degradation of capabilities, and full separation of content, behavior, and presentation. (For why that separation is important A List Apart has a great article on the subject [alistapart.com].) Not even Ruby on Rails gets this right by default.

GWT tries too hard to abstract the actual code that user agents see from the code the programmers create - and that level of abstraction just doesn't work yet. Just like trying to translate a passage in French to English and Japanese with a machine translator, the GWT tries to take Java code and translate it into a mish-mash of XHTML, CSS, and JavaScript - and the results are as mangled as one would expect.

Until someone comes along with a framework that creates clean, semantic code with full separation of behavior, presentation, and content, web application developers have to be mindful of their code and do a lot by hand. Frameworks can save time, but they also cause a trade-off in terms of code quality and compatibilty. The GWT goes too far in that balance IMHO, and isn't something I'd use to develop public applications. Like ASP.NET, it's too reliant on abstracting XHTML/CSS/JavaScript from what the programmers deal with, and that always leads to bloated masses of code that frustrate users and hog bandwidth.

Re:Why GWT Isn't A Good Framework

IE's rendering engine is suckier than Monica Lewinsky holding a Dyson at the event horizon of a black hole

Mod -1: Trying too hard.

Seriously though, you say GWT tries to take Java code and translate it into a mish-mash of XHTML, CSS, and JavaScript - and the results are as mangled as one would expect. and then go on to say Until someone comes along with a framework that creates clean, semantic code with full separation of behavior, presentation, and content.... Isn't that kind of contradictory? If its spitting out xhtml, css and javascript, that seems like content, presentation, and behavior are all clearly defined.

Re:Why GWT Isn't A Good Framework

OK, ignoring the fact that GWT uses Java as its initiation language (it could use Ruby or C++ or PHP) - I still have to disagree on a number of points.

The problem with the GWT and other framworks like it as it ignores the reality that browsers today suck.

Right, which is why something like GWT is nice because you don't really care that the browser sucks, you write your code and it works. Graceful degradation isn't really an issue in this case ... GWT supports a number of browsers identically (i.e. your code will run identically to the end-user). If you're looking for graceful degradation to plain HTML w/o JS, then that's a bit of a pipe dream, since you're not talking the same application. That's analagous to writing a GUI app (on Windows or OS X or Linux/GTK/Qt) and having it gracefully degrade to a console application. Last I checked, anything that's more than a simple app doesn't do anything like that, and for good reasons. Endless backwards compatibility is a case of diminishing returns and while it may be "nice", it isn't practical or economical.

GWT tries to take Java code and translate it into a mish-mash ... and the results are as mangled as one would expect.

Um, no. The results are a web-application that functions as you would expect. It is unknown if Google Calendar or Google Spreadsheet are using GWT, but according to Google, the pain they experienced writing Google Maps & GMail played a part in developing GWT - so those types of applications are certainly on the drawing board for GWT. I wouldn't call those apps mangled or anything along those lines. They're quite possibly best-of-breed.

I also disagree that Google's approach with GWT:

...always leads to bloated masses of code that frustrate users and hog bandwidth.

On the contrary, I think in reality, hand-coding, by the masses of programmers that cannot be experts at JavaScript and Browser Nuances, has already created bloated masses of code that not only frustrate users & hog bandwidth, but are also difficult to maintain, practically impossible to debug effectively, and a huge drain on an organization's resources. Its time to change that and I think GWT takes steps in the right direction. I haven't heard of masses of users being frustrated by Google Maps, Mail, or Calendar ... at least not because its bloated or hogs bandwidth.

Unobtrusive JavaScript

I'd prefer to see a toolkit that provides unobtrusive JavaScript [adactio.com]. Basically, you'd tag (with a specific class) the HTML elements that you'd want to act as UI widgets. Then include the JavaScript file, and the JavaScript file does all the work. This provides graceful degradation -- if they don't have JavaScript, they just get the original HTML. And it lets you use all your standard HTML tools, so your designers can still handle all the look and feel, without having to deal with anything besides HTML and CSS.

XML11 looks more promising

Uses java byte code instead of source so it's more stable. http://video.google.com/videoplay?docid=-317158218 7051229467 [google.com]

application development, web interface

Interesting, I'm not really a web developer except for a bit of custom javascript DHTML stuff I've done.
(Yes, most of my web experience comes from the days when it was just called DHTML..)

This is the first time I've peeked into a platform like this... it's really interesting how close it is to writing a GUI application, like in SWT or something. (Again, I've only done a little bit of SWT, since I don't prefer Java most of the time, but it definitely has its place.)

But I find it amazing how it seems totally incidental that it happens to use the web for rendering its interface. What it really makes me think of is the possibility of writing multiple back-ends for it. Imagine using it to write applications that can compile to both a GUI environment, using native Win32/GTK/Cocoa elements, and then also be able to easily re-compile to a web application. Not sure there'd be any point to that, come to think of it, but the idea sounds cool for some reason. ;-)

Or imagine writing an application which is totally local, but happens to use a browser for its interface. It could embed a mini HTTP server in it.. well, I know this has been done before, but this is the first time I've seen an API that actually makes the idea somewhat attractive to me. It would solve the multiplatform GUI problem by just using whatever browser the user happens to prefer.

Anyways, now if only I could use GWT from Python instead of Java.. ;-)
(Same goes for SWT by the way.. the PySWT project isn't very useful yet. Sigh..)

Google Web ...

Did you say "Google Web Rootkit"?

GWT Analysis

A competing framework Thinwire[www.thinwire.com] has another analysis of GWT.
http://www.thinwire.com/thinwire_vs_gwt.html [thinwire.com]

GWT Interview

Here's an interview with someone in the middle of integrating Google Web Toolkit (GWT) with a Spring MVC application [oreillynet.com]. In the interview Michael Podrazik provides some insight and tips for people interested in starting out with GWT.

Huh?

You haven't seen bad OO code? (maybe YOU'RE the bastard who writes it) You don't know how to write good non-OO code? It exists.

There is an OO equivalent to spaghetti code, and it's fearsome.

Re:When I hear OO ... When I hear Java

What do you think when you hear "Code Generator?"

Thats my biggest beef with the way this kit works. The JSNI interface seems like a pure hack to start with doing things like embedding javascript code in a java file using code like the following:

private native void applyEffect(Element element, String effectName) /*-{

// Trigger named Scriptaculous effect
$wnd.Effect[effectName](element);
}-*/;

I find these kinds of toolkits get you up and going quickly, especially if you are new. However, the first time you run into something the toolkit can't handle, the black box nature means your SOL.

Re:When I hear OO ... When I hear Java

What do you think when you hear "Code Generator?"

When I hear "code generator", I think compiler.

Re:When I hear OO ... When I hear Java

When I hear someone say what you said I think... "another hack who's never programmed anything, never will, but read some shit on slashdot about it and now believes he's an expert"

There's plenty of unreadable, unusable OO code out there, as well as plenty of well performing Java.

Re:When I hear OO ... When I hear Java

We are taking advantage of it. We really look forward to the ability to develop rich Web Apps using a standard IDE and Java "OO" paradigm. ==== Beaches & Casinos: http://buddytrace.com/ [buddytrace.com]

Re:When I hear OO ... When I hear Java

Umm... Java IS an Object Orientated language... (unless you mean something else by OO, in which case I would love to be enlightened..)

Ahem

*Kaffe* [kaffe.org]

There's not a lot of tolerace for criticism of corporations, is there?
I think that most coders feel the same way.

Re:When I hear OO ... When I hear Java

Java's performance isn't "poor" — as is frequently pointed out on this site, this is largely an outdated myth. As for not truly open source, it's a definate issue, yes, although there is open implementations [wikipedia.org].

Re:In all fairness

Bah. Until certain whiners in the OSS community gets off their asses and work on the existing OSS implementations instead of demanding a non-OSS implementation should be opened up, it's simply not worth it debating with them like that.

It's not like GCC was written by AT&T.