I'd like to voice my thoughts on this, but first a fairly big caveat: I am far from a GWT expert.
I'd dabbled with it for work-related purposes at my last job, but the focus of what I worked on was definitely not GWT. (One of our products was a SaaS app whose UI is built almost exclusively on GWT; our other product also used GWT for admin-related UIs, which were embedded in a thick-client app.) I'm familiar with a lot of the concepts, but I've never had a need to build something ground-up with GWT; so perhaps a lot of my experience with GWT+Roo has been tainted by the fact that I'm still fairly new on the GWT learning curve. That being said, I generally pride myself on learning quickly :) I was looking for a way to jumpstart a project I've been working on that would allow me to use GWT on the client to talk to a fairly straightforward data back-end. I didn't want to spend bundles of time writing tedious plumbing code, as I really wanted to get up and running ASAP; Roo seemed perfect since it handles a lot of the boilerplate codegen, etc. Indeed, it is ideal for this; I had a very basic CRUD app up and running with my handful of entities in a matter of minutes. (Incidentally - one of the details that I think is perhaps often overlooked is the fact that you also get a nicely codegenned build/test system; this is typically a lot of annoying plumbing as well that Just Worked with Roo.) It was only after the first ten minutes that I was exposed to the full brunt of the learning curve... I too have found that any attempt to stray from the cookie-cutter CRUD app requires a fair bit of self-investment in reading the generated code, etc. One of the things I'm currently fighting is that the MVP patterns that are generated by Roo do not quite match the GWT 2.1 docs and what they outline for their Hello World app. In particular, the bits around how activity mappers are implemented in Roo-generated code are quite different from the docs. (Actually, a fair bit of thanks goes to you, Amir! Your post on GWT 2.1 MVP with Gin addresses a lot of my recent confusion around the subject: check out https://groups.google.com/group/google-web-toolkit/browse_thread/thread/2ed199a42c500156 for those who haven't seen the post.) What I actually found to be more sensible was to create a parallel ground-up GWT module/entrypoint, while trying to reuse at least some of the generated code, where possible. Largely, I've found that I cannot; the code is fairly tightly-bound to the notion of a master/details-style CRUD interface; while this would be a huge time saver for a number of different types of apps, it's unfortunately of little value to me, at least as I've been delving into it the past week or two. However, none of that inherently bothers me; there *is* still a bit of usable generated code (specifically the entity proxies and related code); there are certainly a few annoying bugs I ran into (the issue with adding new finders is something I ran into right away), but I was able to at last get past these for now. My biggest concern about GWT +Roo is the utter lack of documentation of the GWT-specific pieces as they relate to a Roo-built app. I've basically had to try and take the knowledge gained from the GWT docs (e.g., re: Activities/Places-style MVP) and map it to the generated code, with an attempt to understand what was generated and why. I think at least a basic discussion on typical customizations (e.g., adding a finder) or limitations of the generated code would be a must for docs to cover. I'd also like to see a bit of analysis into the basic architecture of the generated UI code; this would have helped me a lot (though perhaps this was just me also trying to learn the GWT 2.1 patterns in conjunction with what Roo generated). Additionally, I have also observed that a lot of the forum posts about GWT+Roo tend to remain unanswered or have been thus far identified as known issues (suggesting a lack of project maturity). That being said - while I'm troubled by the paucity of docs - I am well aware that this is all fairly bleeding-edge stuff, and I'm posting my experiences simply to share them, and hardly to bash GWT or Roo. I am very intrigued by the possibilities of what the toolkit could offer. If nothing else, at least as it stands today, Roo+GWT generates everything just shy of a usable UI for your code, leaving you to implement those pieces yourself. This coupled with all the data- binding goodness that now exists in GWT 2.1 I think can still make for a powerful toolkit, and can make Roo+GWT quite successful. So, in short - am I having better luck? I have to say, I'm afraid I'm not :( However, I'm still excited by what's there - because what is there *does* work, and that means it's already saved me time by not having to re-invent boilerplate plumbing. Sankalp On Nov 3, 4:26 pm, Amir Kashani <amirkash...@gmail.com> wrote: > I've been playing with Roo & GWT for a couple of days now, which > certainly doesn't make me an expert, but I'd like to share my > experience and see how it compares with others. > > The application that I'm trying to create has about 8 entities that > require a CRUD like interface. The rest of the application requires a > more specific (non-CRUD) UI. My hope was that Roo would help generate > the CRUD portions, just allowing for some basic customizations, and > let me handle the rest. So far, I seem to be fighting it more than > it's helped. Some examples: > > - You can't modify the generated ui.xml files -- Roo will simply > clobber them as soon as it makes the next change for that entity. For > example, I don't want to list the ID field in the list view, but I > don't see a way to remove it. There's a JIRA open about this, that was > deferred to post 1.1.0 (ROO-937). > > - New dynamic finders don't seem to be added to the respective Request > interface. Further, and this may be a Roo thing, they return > TypedQuery rather than a List or the entity, which I don't imagine the > RequestFactory framework can handle. There's another JIRA than covers > this, I believe (ROO-1595) > > - Adding custom finders doesn't work, because again, Roo will clobber > the Request interface. I tried removing the @RooGwtMirroredFrom > annotation from the Request, but Roo just readied it for me. Probably > under the previous JIRA. > > Lastly, the documentation is atrocious. The Roo reference guide only > mentions GWT in its appendix of commands, and other guides only go as > far as a simple "gwt setup". There's no mention of the expectations, > limitations, guide to customization, or anything else. I've seen > several, yet unanswered posts on the Roo forum about customization, so > I imagine I'm not alone here. > > I've been a very loyal GWT user since 1.3, and I'm very pleased with > the ongoing progress, including the gamut of features introduced in > 2.1, so I hope nobody takes this as GWT bashing. However, unless I'm > really missing something, the implementation of Roo+GWT falls far > short of the hype it's received since I/O. In the current form, it > makes for a very impressive demo: type a few commands and a full > functioning database editor pops out. But as soon as you try to do > anything outside of what's generated, it's very hard, if not > impossible. > > So, it's only been a few days since final release, but what do other > people think? Is anyone having better luck? > > Thanks. > > - Amir -- You received this message because you are subscribed to the Google Groups "Google Web Toolkit" group. To post to this group, send email to google-web-tool...@googlegroups.com. To unsubscribe from this group, send email to google-web-toolkit+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/google-web-toolkit?hl=en.
