One of the things I'm working on for my day job right now is the ability to display an Excel-like UI for editing a form. So I did a search on Google tonight and found DHTML Kitchen. On this site, they had exactly what I was looking for - a howto for creating a tabbed panel system. So I've used this example to create a prototype of what I can do. This is pretty slick b/c now I can give the users a UI that looks like the Excel they're used to, and I can use the same ValidatorForm for the entire page. It even supports remembering which tab you last selected, and also allows navigation to a tab. The DHTML Kitchen also appears to have all kinds of other goodies to checkout. I'll definitely be adding it to my list of cool bookmarks.
If someone could verify that this prototype works in IE 5.5 - that'd be awesome! This is the browser we have to support at work and all my browsers are 6.0+. In return, I offer you the source in a single zip file :-) After playing with this a bit after posting - it seems like it's got a couple of issues in IE 6. The first is that a double line shows up at the bottom of the top tabs after refreshing. Clicking on any tab at the top snaps the tab bar back into place. The second is performance - it's taking 3-8 seconds to load the page - yikes! I'm still going to use it though, and hopefully optimize and fix these issues later.
My first attempt at using ConvertUtils is turning out to be a painful one - most likely due to my own ignorance. Let's see if you can help me out. I have a Hibernate Bag, which is really a java.util.List on my User object. When I run the User object through XDoclet, I create a UserForm (extends ValidatorForm). The form has an ArrayList on for any instances of List or Set on the object (set through a custom struts_form.xdt template). So I created a ListConverter to convert a List object to an ArrayList. Sounds pretty simple right?! Here's my ListConverter.java:
public class ListConverter implements Converter {
//~ Instance fields ========================================================
protected Log log = LogFactory.getLog(ListConverter.class);
//~ Methods ================================================================
/**
* Convert a List to an ArrayList
*
* @param type the class type to output
* @param value the object to convert
*/
public Object convert(Class type, Object value) {
if (log.isDebugEnabled()) {
log.debug("entering 'convert' method");
}
// for a null value, return null
if (value == null) {
return null;
} else if (value instanceof Set && (type == Set.class)) {
return new ArrayList((Set) value);
} else if (value instanceof List && (type == List.class)) {
return new ArrayList((List) value);
} else {
throw new ConversionException("Could not convert " + value
+ " to ArrayList!");
}
}
}
When I run BeanUtils.copyProperties(userForm, user), I get:
Could not convert cirrus.hibernate.collections.Bag@e2892b to ArrayList!
On another class, where I am trying to convert a List of Longs, I get:
Could not convert [-1, 1, 30129] to ArrayList!
I'm registering my custom converter in a static block of my BaseManager class. My *Manager classes do all the conversions, so this seems logical:
static {
ConvertUtils.register(new StringConverter(), String.class);
ConvertUtils.register(new LongConverter(), Long.class);
ConvertUtils.register(new ListConverter(), ArrayList.class);
if (log.isDebugEnabled()) {
log.debug("Converters registered...");
}
}
Since I'm in a major time crunch, I'll try simply making my getter/setters on my UserForm to be List. I'd like to use ConvertUtils though, so hopefully someone has a solution.
This brings me to a RANT and I think it's my first official one. My last three projects have always started small, and the goal has always been a prototype of functionality. A prototype that turns into a production system. All fricken three of them. All were supposed to take about 3 months to develop initially. The last two projects took at least 6 months. This one has a one month deadline, but the scope is a lot smaller than the previous two. But still, it's always the same scenario - the clients want a prototype, but turn it into a production system. Since it's a prototype, I tend to write "workarounds" for design patterns (see above) that I can't figure out. Is this good? It probably doesn't hurt since no one will ever look at my code - right?! When's the last time you looked at a co-workers code? (The more == the better). And the truth is, as long as it works - it's probably good enough. However, I as a developer, get heartburn when I think about maintaining the system that I created under the impression that it was a prototype. Maybe one of these days I'll figure out all the best practices to creating a robust web application, and then I'll know everything - so I won't have to write workarounds for my lack of knowledge. If you see some pigs flying, you can think to yourself - "Wow, Raible must've figured it all out" ;-)
I saw Russ's "Links To Me" link on his site, and tried it for this site. There I found a great idea from Patrick Chanezon regarding WebTest:
...once you go declarative, you can begin to build tools to generate the declaration.
One could build a Mozilla XUL based tool that would generate this XML based on an interactive testing session in the browser.
Now THAT is a great idea! Any takers?
I'm implementing an interesting feature this morning. When a user views the error-page of the application, an e-mail is sent to an administrator with a StackTrace. By error-page, I mean the "errorPage" attribute in a JSP's page declaration.
<%@ page language="java" errorPage="/errorPage.jsp"
contentType="text/html; charset=utf-8" %>
I'm using the mailer taglib to do this, but can't help thinking that Log4j already provides similar functionality. Doesn't it have an SMTPAppender or something like that. I briefly scanned their site, but didn't see anything. If you know how to configure this functionality - hook me up!
After being severly disappointed in my purchase of Connectix's Virtual PC 6, I've returned it. Since I bought it online and downloaded it, it was as simple as sending an e-mail and destroying the software. I had no problems getting rid of the software as it's too damn slow to be tolerable. I don't blame this on Virtual PC, I blame it on the slowness of Apple's CPU. Megahertz don't matter - baaahhh! When I get my 2GHz PowerBook, I might re-purchase this product, but no need for now. I'll settle for using Microsoft's RDC to get my Windows emulation.
So if you're a Mac fan and you're purchasing a new Mac with the hopes of running Virtual PC - you'd better get the dual processor G4, you're gonna need it!
Rather than pulling an all nighter like I have been the last couple of weekends, I decided to go to bed at a reasonable hour and get up early instead. So here I am, woke up at 4:30 and I get to work all damn day today. Should be a blast... While I'm up, here's an interesting post from the xdoclet-devel mailing list:
XJavaDoc is pretty mature by now. -But so is Eclipse. One of the Eclipse modules, JDT, is an engine similar to XJavaDoc. It can be used outside the Eclipse IDE.
The nice thing about JDT is that it has diff/merge support too (I haven't verified this, but Erich Gamma said told me it has that). That would make it possible to edit generated sources and not lose them.
What do you think about evaluating JDT for XDoclet 2?
Also, take a look at the Hibernator source:
http://tinyurl.com/49zk
Hibernator uses Eclipse's java parser, is fairly small, and is therefore an excellent source of inspiration if we want to investigate this path further for XDoclet 2.
Emphasis added by me to indicate my favorite part.
