Tom Pierce: Java

Hibernate circular cascading

Tue, 09 Mar 2004 21:43:15 GMT

I'm going to start by saying that I may have missed something somewhere in the Hibernate documentation, FAQs, or forums. But, I don't remember seeing direct references to this.

It seems that is a bad idea to make both ends of an inverse mapping cascade. In this situation, Hibernate will try to cascade multiple times when saving new objects that cause a cascade. The only way I found to prevent it was to remove the cascade attribute on the inverse end of the mapping.

OK, a little less talk and a little more coding please. I had something like this and it was bad:

  <!-- Application class -->
  <class name="net.trin.colorreports.reportcomments.dao.valueobjects.Application">
    <id name="id" column="application_id" type="integer">
      <generator class="identity">
      </generator>
    </id>
    <property name="name" type="string" column="application_name">
    </property>
    
    <set name="reports" inverse="true" lazy="false" cascade="all">
     <key column="application_id"/>
     <one-to-many class="net.trin.colorreports.reportcomments.dao.valueobjects.Report" />
    </set>
  </class>
  
  <!-- Report class -->
  <class name="net.trin.colorreports.reportcomments.dao.valueobjects.Report">
   <id name="id" column="report_id" type="integer">
    <generator class="identity"/>
   </id> 
   
   <property name="name" type="string" column="name"/>
   
   <many-to-one name="application" column="application_id" class="net.trin.colorreports.reportcomments.dao.valueobjects.Application" cascade="save-update" />
   
   <set name="comments" inverse="true" lazy="false" cascade="all">
    <key column="comment_id"/>
    <one-to-many class="net.trin.colorreports.reportcomments.dao.valueobjects.SummaryComment" />
   </set>
  </class>

Notice the cascade attributes on the <set> tags. This created the loop because of the cascade on the <many-to-one> tag I think. Once I removed the cascade attribute from the <set> tags, everything started working beautifully.

Naming a directory TAGS in CVS

Tue, 09 Mar 2004 21:23:36 GMT

I hope that someone smarter than I am will read this and correct me. It seems that if I name a directory "tags" somewhere in my CVS project, it will not be handled correctly by IBM Websphere Studio 5.1.1 (Eclipse 2.1.2). When I try to check that directory (a Java package directory) out of CVS, WSAD says "error fetching file revisions".

I am using CVS for NT, but I AM NOT using a repository prefix. So, I don't have this known error situation. I'm not sure if my issue is a WSAD/Eclipse issue, a CVS for NT issue, or just a CVS issue.

We are working around it by renaming our tags package to taglib.

Fri, 27 Feb 2004 15:32:12 GMT

Oh No! DTO!. Should DTOs have public variables? Or should they have private variables with getters and setters? [Artima Weblogs]

There's an interesting discussion going on in the comments of the blog. Worth reading - I'm sure we've all pondered this question before. Especially when you think of languages like Python where it's downright hard (well, at least contrived) to make member variables private.

Voruta: An Interesting Approach to O/R Mapping

Tue, 10 Feb 2004 17:17:54 GMT

Voruta is an OSS framework for automatically generating data access code. The framework allows you to embed SQL calls in the javadoc associated with DAO methods. The code for the DAO calls is generated dynamically at run time. This is an interesting approach, however it doesn't allow you to externalize the mappings and prevent recompiles when data access queries or methods change.

Don't forget about the Struts HTML XHTML tag

Mon, 26 Jan 2004 19:53:29 GMT

If you are developing a Struts web application and you want to develop XHTML compliant pages, don't forget to add <html:xhtml/> to you pages. If you don't, the other HTML tags aren't guarenteed to generate XHTML compliant output. Further, you must put this tag on every JSP fragment or Tile. The fragments don't inherit the property from their container.

Find all the classes imported in your source files

Tue, 20 Jan 2004 22:12:17 GMT

Do you do your Java development on Unix or on Windows and have Cygwin? Do you explicitly import every class in your source files? If you answered yes to both questions, then you can use this handy command to find all the classes imported by your source files:

find src/. -name "*.java" -type f -exec grep -H -n import {} ; |sed "s/^.*import \(.*\?\);/\1/g" | sort -u

Be sure to change directory (first parameter to the find command).

Don't get fancy with Struts tags!

Mon, 19 Jan 2004 22:37:26 GMT

I just spent too long debugging a little problem I was having with the Struts nested tags. With the nested tags, you can define a "root" bean and then access properties off the root with special nested versions of the other Struts tags. This allows you to access properties of beans with syntax like:

<nested:root name="customer">
Name = <nested:write property="name" />
</nested:root>

Which is really cool, especially with some of the more complex tags because you don't have to refer to the parent bean. In theory, this allows you to also nest references to beans contained within the bean with something like:

<nested:root name="customer">
Name = <nested:write property="name" />

<nested:root name="address">
Address Line 1 = <nested:write property="line1" />
Address Line 2 = <nested:write property="line2" />
City = <nested:write property="city" />
</nested:root>

</nested:root>

However, the more you nest the tags within each other, the more flaky things are likely to get. In my recent debugging session, I made the decision that from now on, nested roots should contain at most 1 nested root within them. If you have more than 1 root defined within a root, all the subsequent tags outside the 1st nested root don't work correctly.

As an example, I don't think you should do this:

<nested:root name="customer">
Name = <nested:write property="name" />

<nested:root name="address">
Address Line 1 = <nested:write property="line1" />
Address Line 2 = <nested:write property="line2" />
City = <nested:write property="city" />
</nested:root>

<nested:present name="altContact">
<nested:root name="altContact">
Name = <nested:write property="name"/>
</nested:root>
</nested:present>

</nested:root>

Because chances are that the tag logic won't determine that altContact is present correctly.

Get the user id from servlet container

Mon, 19 Jan 2004 21:07:57 GMT

Because I don't get to do it often enough, I always forget how to get the user id from the servlet container. This is handy if you are letting the container handle the authentication for your application. All you have to do is call request.getUserPrincipal().getName(). This returns the id as a String. Beautiful.

Common Struts Errors and Solutions

Thu, 15 Jan 2004 16:25:41 GMT

Here's a great page that lists some of the common Struts configuration errors and their solutions.

Show CSS computed styles for HTML page elements

Thu, 15 Jan 2004 15:51:04 GMT

I just found an awesome bookmarklet for Firebird (may work with IE, but it didn't for me). Get it here. Some of its sweetness includes:

style information is shown in a separate window and elements are selected just by mousing over them.
showing you the hierarchy of styles that are being applied to the element
gives you a complete breakdown of the style that's in effect for the element

I've been having some trouble debugging some CSS issues with one of my pages. This is a tremendous help.

Mon, 12 Jan 2004 19:06:19 GMT

Here's a follow-up to "Fighting with Java logging". I've found that alot of the Jakarta libraries and other open source libraries are using Commons Logging to abstract the logging for their library. This makes alot of sense - they don't have to worry about whether you are using JDK 1.4 logging of whether you are using Log4j.

However, if you are using Lumberjack, you are screwed. Let me rephrase that: if you are using Lumberjack, and you don't have time to fiddle with the source for Lumberjack to add the missing method from JDK1.4 or add support for Lumberjack to Commons Logging, then you are screwed. In my case, I'm screwed.

So far, I can't use Commons BeanUtils or StrutsTestCase. Not happy. I guess I should post a question to a Lumberjack list or something.

Idea for providing "uncached" objects after retrieving with Toplink

Fri, 09 Jan 2004 21:47:30 GMT

When you retrieve an object with Toplink, no matter if it's through the ClientSession or a UnitOfWork, Toplink returns an object that it is tracking in its cache. If you change the object in one place, ask for it again in the other, you will get the changed object. In some cases this is great because you can quickly and easily reflect the latest information with incredible performance. However I find myself being confused by this behavior for most of my applications.

My idea is that each of the classes you store with Toplink implements Cloneable. Then, in your DAO layer (you are using a DAO layer, right?) , before you return an object to the caller you call its clone() method. This will ensure that you have a shallow copy of the object that is not associated with Toplink's cache. Then, you can count on 2 retrievals of the same object only reflecting any updates that have actually been committed to the cache/database.

If you have a complex object graph that you are persisting with Toplink and you want the same functionality, you can implement a method to do a deep clone. This would make sure that the entire graph reflects the data at that point in time rather than any changes that are happening in other places.

Emacs, DBVisualizer, and Java

Fri, 09 Jan 2004 18:38:09 GMT

Within the last six months, I've started using DBVisualizer (http://www.minq.se/products/dbvis/) for accessing databases while I'm coding. It's a great front end for any relational database that has a JDBC driver. I highly recommend it.

One of the many cool features is that result sets in the product are displayed in a grid (JTable). You can select values in the grid and copy them to the clipboard to paste in other applications. I frequently paste the data into Emacs for use in Java unit tests.

The problem is that the columns of text are not in an array format. So, I wrote the following Emacs functions to "fix the data up". After you paste the data into Emacs you can immediately run one of the interactive fix-up functions. (One of the thousands of beautiful things about Emacs is that it automatically sets the region around the pasted text.) This will remove the column header and produce a comma delimited list of terms suitable for use in an array.

(defun fixup-dbviz-array (start end replace-string)

"Formats the region for use as an array and replaces the regexp match

with the specified replacement string."



  (unwind-protect

      (progn

        (narrow-to-region start end)

        (goto-char (point-min))

        (kill-line 't)



        (while (re-search-forward "\\(.*?\\)\n" nil t)

          (replace-match
replace-string))          

        



        (if (re-search-backward "," nil t)

            (delete-char 1)))

                

    (widen)))



(defun fixup-dbviz-string-array (start end)

"Formats a column of values that are copy and pasted from DB

Visualizer and formats them for use as a String array"

  (interactive "r")

  

  (fixup-dbviz-array start end "\"\1", "))



(defun fixup-dbviz-int-array (start end)

"Formats a column of values that are copy and pasted from DB

Visualizer and formats them for use as a int array"

  (interactive "r")



  (fixup-dbviz-array start end "\\1, "))

Toplink Session Console classpath

Fri, 09 Jan 2004 14:30:13 GMT

Here's a little something that you may or may not find in the documentation about the Toplink Session Console. If you want to use the console to play with your mappings and run queries with them, you have to add your project's build directory to the classpath. If it can't find the objects the mappings point to, you get and ConversionException when you try to load your project into the console.

The way I worked around this was to create a copy of the batch file that loads the Session Console and add my project's build directory to the classpath. Bingo! Everything works.

Toplink Session Profiler

Tue, 06 Jan 2004 21:13:32 GMT

Have you discovered the PerformanceProfiler yet? This is a Toplink class that allows you to add profiling to your log, and it's simply awesome. Here's a sample of the output:

Profile(ReportQuery,
    class=com.fritolay.bsg.pog.db.valueobjects.Customer,
    number of objects=17,
    total time=80,
    local time=80,
    query prepare=20,
    sql execute=50,
    sql generation=10,
    time/object=4,
    objects/second=212,
)
} End Profile

You enable the profiler with code via your Session. You'll probably want to do it on your ServerSession so that you can catch everything, however I can envision a scenario where you might only want to profile one of your ClientSessions. Whatever. The call is setProfiler on session and it takes a SessionProfiler, the parent of PerformanceProfiler.

Have fun!

Toplink Session Logging

Tue, 06 Jan 2004 20:17:39 GMT

Here's a tasty little Toplink morsel to kick of the new year. If you are setting up session logging in Toplink, and you are doing it by hand, you should set the logging options on the ServerSession not the ClientSession. If you set it on the ClientSession, you won't get the SQL statements dumped to your log! How do you like that? Tasty, I know.

Fighting with Java logging

Mon, 15 Dec 2003 23:56:28 GMT

I've been fighting with Java logging alot today. I've found out some interesting, if not disturbing things. I'm also very disappointed.

I'm currently working on a project that must comply with certain customer standards. The customer has standardized on Java 1.3 for their VM. They have also standardized on the Java logging API for logging, but since they are using 1.3, they have decided to implement the logging API using Lumberjack. This gives them the logging API while still using Java 1.3.

I thought this was a bit unusual, so I thought I would use the Jakarta Commons Logging API to abstract the logging interface. This would allow either the Java logging API or log4j to be used as the logging implementation without changing code. A very good thing.

This is where the fun began. It turns out that although the Commons Logging API will front either JDK 1.4 logging API or log4j, it will not front Lumberjack. Why? Well, it's simple. Commons Logging makes a call to getStackTrace on the Exception class if you are using JDK 1.4 style logging. This method is not available prior to JDK 1.4! Since Lumberjack can only be used with JDK 1.2 and 1.3, Commons Logging cannot front Lumberjack! ACK!

So, I punted. I decided I would just use the Java logging API without an abstraction on top of it. Fine. I created a base class for several of my classes that contained methods for handling exceptions. These methods made the appropriate call to log the exception and then threw an application specific exception. This provides an easy and centralized method of logging and wrapping exceptions.

Enter my disappointment. Upon testing I found that the Java logging API was reporting the class logging the messages as my abstract base class! I thought this might be something peculiar with Lumberjack, so I switched to JDK 1.4 and re-ran my tests. Nope. It still does it.

I figured there had to be a way out of this nightmare. I went to the API Javadoc. I found that there is a logp method that takes the class name and the method name of the class logging the message. Perfect! I could just pass in the class name and method name, and I could probably get those dynamically...

WRONG! As it turns out it's easy enough to get the class name - we've all done it. Do something like this.getClass().getName(). Simple. However, in JDK 1.3 there doesn't appear to be a way to dynamically (and generically) derrive the current method name. Surely not! I went to Google. I found this page which shows a way to get the current method name. But wait! What's that call? Is that a call to getStackTrace on Exception? Oh nuts.

SO, here's where I am. In the interest of time (on this project there is not alot) I decided to bite the bullet and take the inaccuracy of having the log reflect the abstract base class when logging. It was a trade off between log accuracy and ease of code maintenance. Maintenance won. I'm going to consider putting the class name as part of the logging message so that the reader will get closer to the problem faster.

I'm beat.

Sun's Java Desktop System

Fri, 05 Dec 2003 05:03:39 GMT

In case you haven't seen it or heard about it, Sun is promoting a Linux distribution they call the Java Desktop System. I'm not sure how Java fits in there other than the distro has the runtime. But, more power to them because they are promoting two of my favorite things.

Struts and Container Managed Persistence

Fri, 05 Dec 2003 04:27:23 GMT

This is a cool idea of how to use Struts with container managed persistence. I haven't had a chance to play with it yet, but it sounds feasible. I'm thinking that you could use this for implementing a poor man's single sign-on. You could programmatically login to the other applications your user would be accessing and finally log in to the container. If authentication failed at any point in the chain, your app would send the user to the login error page.

Running Ant Exec Tasks Asynchronously

Thu, 04 Sep 2003 23:32:43 GMT

Apparently, Ant doesn't allow you to run exec tasks asynchronously. For example, if you start up Tomcat from an Ant build file, it will wait until Tomcat exits before it continues to work through its build file. This is probably not what you want it to do. You probably want it to spawn Tomcat in a new process and keep on trucking (or wait until Tomcat is up and then keep on trucking).

It seems to me this would be easy in Unix with a decent command shell. However, I'm in Windoze. So, no luck there.

But, I found a great thread (ala Google) that gives you a couple of options. I went with the "antRunAsync" option. Worked like a charm! Here's my Tomcat start target:


 <target name="start.tomcat" depends="prepare" if="tomcat-needs-to-start">

  <exec executable="cmd.exe" vmlauncher="false" failonerror="true">

   <env key="PATH" path="C:/WINNT/system32;C:/WINNT;C:/Program Files/SQLLIB/BIN;C:/Program Files/SQLLIB/FUNCTION;C:/Program Files/SQLLIB/SAMPLES/REPL;C:/Program Files/SQLLIB/HELP;C:/PROGRA~1/IBM/IMNNQ"/>

   <arg value="/c ${ant.home}/bin/antRunAsync ${websrv.dir}/bin/startup.bat"/>

  </exec>



  <waitfor>

   <http url="${testurl}"/>

  </waitfor>

 </target>

Dynamically Redeploying With Tomcat 4.1, Ant, and Struts

Thu, 04 Sep 2003 22:06:27 GMT

I've been to battle. My foe is my web application built around Struts 1.1 that I'm deploying on Tomcat 4.1. I want Tomcat to be able to remove the application completely so that I can redeploy it without having to restart my server. Tomcat patently rerfuses to delete the application's directory under "webapps" because the struts.jar in my WEB-INF/lib directory is being used. Used by what? I don't know.

I have emerged victorious. How? Instead of trying to actually remove or undeploy, I decided to just stop the web application, unwar my war on top of the Tomcat deploy directory, and start the web application again. This seems to be a pretty decent work around to the locked struts.jar problem.

I'm automating all this with Tomcat's Ant tasks. Booyah!

Using Ant to Deploy to Tomcat

Thu, 04 Sep 2003 19:32:28 GMT

Did you know that you can have Ant deploy your application to Tomcat 4.1 and automagically deploy/undeploy the application? This prevents you from having to restart Tomcat after a deploy if you've changed the web.xml or other non-reloaded properties. Here's the link to the documentation.

Ant DTD

Thu, 04 Sep 2003 19:30:15 GMT

Want to get your hands on an Ant DTD? Why not have Ant generate it for you? Check out this article.

Wed, 27 Aug 2003 22:00:52 GMT

Here's a little Oracle nugget that a coworker learned today. Oracle version 8.1.6 and lower doesn't like to be involved in wrapped connections. That means that you will have problems using connection pooling with this version of Oracle and prior. Go figure.

Sat, 23 Aug 2003 23:50:37 GMT

Introducing XACML. This Computerworld article provides a very brief introduction to XACML, an access control list language implemented with XML. To attain an understanding of the lingo check out this XACML glossary or if you prefer dive straight into the specification( [CapeBlog]

This is an introductory article. Sounds interesting. This technology provides a standard way to query for and distribute access policies. The Sun site for this technology is here