SpringOne 2010: How to Build Business Applications using Google Web Toolkit and Spring Roo

I spent the second-last week of October at SpringOne 2GX 2010 in Chicago and I thought some of you might get something useful out of my notes. These aren’t my complete reinterpretations of every slide, but just things I jotted down that I thought were interesting enough to remember or look into further.

How to Build Business Applications using Google Web Toolkit and Spring Roo
presented by Amit Manjhi

Amit works at Google on the GWT (Google Web Toolkit) team. In this session, he showed the ease of creating a GWT app with Spring Roo and then talked through a a whole swathe of GWT best practices.

Roo supports the addition of JSR-303 constraints [PDF] as part of adding a field to an entity.

Creating a GWT webapp from an existing Roo domain model is as simple as typing gwt setup (and hitting enter) in the Roo console.

The default webapp generated by Roo automatically handles validation on the server side. I have to say that the default rendering for errors was pretty poor. The error he showed as an example just appeared at the top of the form and said something like "this field is required", without any reference to which field had caused the error. Room for improvement there, but you certainly get a lot of webapp for doing nothing much at all.

Bookmarkable URLs work out of the box, although I'm not sure if this meant that all URLs are bookmarkable by default or whether it's just easy to make URLs bookmarkable when you need them.

Amit showed a version of the same UI that some Google guys had jazzed-up with some nicer CSS and a few changes to the layout components. It was using some kind of slide transition where, when an item was selected from a list, the list would slide off to the left and the item detail slide in from the right, then vice versa when you went back. Looked very neat.

The Spring Roo addon for GWT generates the site using a whole bunch of best practices as learned by the Google Engineers who’ve been using GWT to develop the AdWords site.

The Model-View-Presenter pattern was presented as a suitable pattern for client-side GWT. This decoupling pattern allows different view implementations to be attached to the same Presenter.

Using DTOs (Data Transfer Objects) (as opposed to sending entities to the webapp) was recommended, though he did note that, coded manually, DTOs typically violate the DRY (don’t repeat yourself) principle. This downside is overcome by creating an empty interface for the DTO, annotated with a @ProxyFor annotation. GWT then automagically creates a proxy object for the entity class named in the annotation and this proxy acts as a DTO. The Google guys call this an Entity Proxy. From what I could tell, the proxy automatically proxies all value fields of the entity. You can provide annotated methods on the DTO interface that allow lazy navigation of entity relationships.

The default Roo project doesn'’t use GWT-RPC, chiefly because of the bandwidth implications when mobile devices are involved. Instead, they use an object on the client called a RequestFactory that talks JSON to a generic RequestFactoryServlet on the server.

RequestFactory receives notifications of the side-effects of its server-side calls and posts these as events on an event bus in the client.

They’ve replaced the Presenter with something they call an Activity that abstracts away a lot of the boilerplate normally required in a Presenter. The common parts of Activities are themselves abstracted out into a generic ActiviyManager.

He highlighted a proposed method for allowing search engine web crawlers to follow Ajax bookmarkable URLs by using #! to prefix the anchor rather than just #. Google have published an article called 'Making AJAX Applications Crawlable' where they discuss the details.

The talk became quite confusing past about half-way when it stopped being obvious (to me at least, having never written a GWT app) which code samples were part of the server and which were run on the client. My kind request to any GWT talk presenters: please introduce each code-sample slide by saying “This code runs on the {server|client}”.

GWT has a nice SafeHtmlUtils class for escaping HTML entered by users to avoid cross-site scripting (XSS) attacks.

GWT 2.1 contains a client-side logging library in the vein of the JDK logging. Spring Roo-generated GWT apps come with a handler that allows client events to be logged on the server. You can also use a GWT <property-provider> element to enable and disable logging at runtime on a per-user scope. (There's some information about property-provider here under the heading 'Elements for Deferred Binding'.)

Want to learn more?

From Amazon...

From Book Depository...