On your ubuntu box :

# apt-get install vnc4server fluxbox

Vnc4server is a free [...]]]> If you want to set up a ci build, running ui tests, on a box without any windowing system, you need to pay attention to a few details, here are some of  them :

Set up your ubuntu box and Jenkins/Hudson

On your ubuntu box :

# apt-get install vnc4server fluxbox

Vnc4server is a free vnc server, that will act as your X environment, and fluxbox is a tiny window manager; more than enough for UI testing.

In Jenkins / Hudson, Manage Jenkins -> Plugins -> Install new plugins, choose the xvnc plugin

Starting xvnc [workspace] $ vncserver :10 You will require a password to access your desktops. getpassword error: Invalid argument Password:Starting xvnc [workspace]

# su jenkins
$ vncserver :10You will require a password to access your desktops.

Password:
Verify:
Password too long - only the first 8 characters will be used
xauth:  creating new authority file /var/lib/jenkins/.Xauthority

New 'host:10 (jenkins)' desktop is host:10

Creating default startup script /var/lib/jenkins/.vnc/xstartup
Starting applications specified in /var/lib/jenkins/.vnc/xstartup
Log file is /var/lib/jenkins/.vnc/host:10.log

[workspace] $ vncserver :14

New 'host:14 (jenkins)' desktop is host:14 Starting applications specified in /var/lib/jenkins/.vnc/xstartup Log file is /var/lib/jenkins/.vnc/host:14.log

Have a look at the tested UI while the tests are running

This can be very useful whenever your UI gets stuck (waiting on a dialog to be closed …, in my case I run into the issue of launching an eclipse runtime, and a plugin, Google ADT, was asking to confirm whether or not I wanted to upload usage data– hopefully only the first time) during the tests.

# ps aux |grep "vnc"
jenkins  29591  0.0  0.7  33192 14104 ?        S    21:10   0:00 Xvnc4 :14 -desktop host:14 (jenkins) -auth /var/lib/jenkins/.Xauthority -geometry 1024x768 -depth 16 -rfbwait 30000 -rfbauth /var/lib/jenkins/.vnc/passwd -rfbport 5914 -pn -fp /usr/X11R6/lib/X11/fonts/Type1/,/usr/X11R6/lib/X11/fonts/Speedo/,/usr/X11R6/lib/X11/fonts/misc/,/usr/X11R6/lib/X11/fonts/75dpi/,/usr/X11R6/lib/X11/fonts/100dpi/,/usr/share/fonts/X11/misc/,/usr/share/fonts/X11/Type1/,/usr/share/fonts/X11/75dpi/,/usr/share/fonts/X11/100dpi/ -co /etc/X11/rgb

OK, looks like xvnc is listening on port 5914 in this case.
You can then use a vnc viewer (on ubuntu, vinagre is a good choice) of your choice to connect to your ubuntu box on port 5914.
If the ubuntu box has firewall rules impeding you to connect to 5914, then set up a ssh tunnel :

$ ssh -L 5914:localhost:5914 anthony@ubuntubox
then connect your vnc viewer to localhost:5914 , and the magic begins !

A concrete example :  m2eclipse-android-integration in Jenkins

I actually set up this configuration for the m2eclipse-android-integration project; this project is an eclipse plugin, that extends the capability of m2eclipse to android projects, here is the link to the Jenkins hosting this job

The author, Ricardo Gladwell, designed exclusively integration tests on the UI, since most of the functionality of this eclipse plugin is UI contribution.

Here are the specific problems I met while setting up this job in Jenkins :

• as I previously mentioned, the first time I launched the job against this project, once all the tests passed, the tycho surefire plugin step would not stop… actually the Google ADT plugin in Eclipse, was asking to confirm whether or not I wanted to upload usage data, so I connected to the UI and clicked « ok »;  vnc saved the day…
• the tests were relying on android projects, requiring to have among the latest Android libraries; to be up to date (so that the tests would not fail), I had to run the command :

$ android update sdk --no-ui --obsolete --force

Overall, that was a good exercise to practice CI on a UI project.

Links :

Lucky because it was my first time at EclipseCon, my first major conference in North America (it was located in Santa Clara, in the Californian Silicon Valley) and also because my employer, Compuware, agreed to let me go [...]]]> From the 21st to the 24th of March, 2011, I was a lucky attendee of EclipseCon 2011.

Lucky because it was my first time at EclipseCon, my first major conference in North America (it was located in Santa Clara, in the Californian Silicon Valley) and also because my employer, Compuware, agreed to let me go there, along with 2 of my colleagues.

EclipseCon 2011 was great, because

• the main actors of the Eclipse community were there : people I may follow on twitter or ask questions to on several mailing lists; the committers of the projects I use everyday at work; being able to chat with them and ask them questions (even harass them ! no I’m kidding, I did not harass Jeff during 2 days! )
• the talks were diversified and most of them were really interesting, you have below links to some of the talks I attended while there.

Attending this conference made me also realize, maybe once more, that meeting with other developers working on the same technologies, is really inspiring, motivating, and it makes you grow technically.

Ok, you got the picture, I enjoyed the conf !

Here are some of the talks I attended:

• Under the provisioning (p2) and release engineering themes,

1. p2, your savior or your Achilles heel? Everything an Eclipse team needs to know about p2 at EclipseCon 2011
2. Discovering the P2 APIs at EclipseCon 2011
3. Tycho presentation and tutorial at EclipseCon 2011
4. Next Generation Development Infrastructure: Maven, m2eclipse, Nexus & Hudson at EclipseCon 2011
5. Growing an open source project one bugday at a time at EclipseCon 2011

• Under the architecture ,  performance and OSGI themes,

1. Stop the Architecture Erosion of Eclipse And Open Source Projects at EclipseCon 2011
2. Using and Extending Memory Analyzer into Uncharted Waters at EclipseCon 2011
3. OSGi introduction and OSGi 4.3 at EclipseCon2011
4. 10 signs you’re doing OSGi wrong at EclipseCon 2011

• Under the Eclipse tooling theme,

1. Android Development with Eclipse at EclipseCon 2011
2. Getting Eclipse Preferences Under Control in Teams at EclipseCon 2011

• Other really cool talks

1. Keynote on Apache Hadoop at EclipseCon 2011
2. Virgo and RT playing together at EclipseCon 2011
3. Effective Git Tutorial

If you want my opinion, well, I’d say that Tycho, p2, Virgo, git along with gerrit, and Eclipse Memory Analyzer were predominant topics; you could definitely not ignore them.. and not like them neither !

If you did not, you’ll hear from these great technologies in the next few months !

This article explains why I decided to create a new project, and also what I did differently for this new Android app.

First of all, the name [...]]]> On Tuesday 21st of December 2010, I officially launched ReGalAndroid, the follow up of G2Android, aiming at providing G3 and Piwigo support.

This article explains why I decided to create a new project, and also what I did differently for this new Android app.

Why creating a new project ?

First of all, the name of the app (g2android) did not reflect what I wanted it to be: I wanted it not to be specific to a gallery type (G2) but also compatible with G3 and Piwigo.
In the android world, an app is defined by its package name; since I wanted to change from net.dahanne.android.g2android to the new name net.dahanne.android.regalandroid I had to create a new app.
Sure, I could have keep on using the old package name, but it did not make sense, and also I wanted to start refactoring the app modularizing it.

Modularizing an Android app, using maven-android-plugin

In G2Android, one of the things that frustrated me most was the way it was built : it had dependencies on libraries (jars) not part of the android runtime. So I had to include them in a lib directory; adding them to the classpath so that the Android Developer tools knew that G2Android needed them.
Another problem was the modularity of the application : the Gallery2 client API code was mixed with the Android UI code; it could become a nightmare for testing, as for plain Junit test cases, I had to remove the android.jar lib from my classpath (and Android integration testing was not an option as it takes really too long to launch; and does not make sense to test an HTTP client api…)
Starting ReGalAndroid, the first thing I did was modularizing the application :

• one standard jar maven project for the commons classes (Pictures, Albums, etc..) with its regular Junit tests

• one standard jar maven project for each gallery client java api (one for G2 extracted from g2android, and 2 new ones, one for G3 from scratch, and one for Piwigo extracted from Jiwigo, an existing Java Swing client for Piwigo) with their regular Junit tests

• one android maven project with only android related classes, this project heavily depends on the common module and the 3 gallery implementations; this has been possible thanks to the maven android plugin

To ease the process inside Eclipse, you can use Maven Integration for Android Development Tools , it will allow you to launch your Android application from the IDE, just the same as before.

This way, I am now able to test the remote API implementations independently of the main Android app; and also if someone wants to reuse those implementations for any other Java projects, he can just check out those plain java projects.

Logging an android app using slf4j-android

If your Android application depends on plain java projects; those projects certainly do not use Android logging framework.

So if you want to log all your modules in your android app, you can use slf4j : this is meta API where you can choose at compile time which logging API you will use.

For example, in all the logged classes of ReGalAndroid, I’m depending on org.slf4j.Logger and org.slf4j.LoggerFactory , even in the Android module.

But in my maven dependencies; I tell my android module to depend on slf4j-android :

<dependency>
<groupid>org.slf4j</groupid>
<artifactid>slf4j-android</artifactid>
<version>1.6.1-RC1</version>
<type>jar</type>
<scope>compile</scope>
</dependency>

That way, each call to LoggerFactory.getLogger(My.class)  in any of my modules will be wrapped to the android.util.log APIs

Cool thing is, if anyone wants to use any of the remote gallery implementations (G2, G3 or Piwigo client) in its Java app, he will be able to use Log4J or logback, or any other slf4j bridge he wants !

Moving from Google code to GitHub for code hosting

I’m quite satisfied with Google Code hosting; in fact I still use it as a wiki and a bug tracker; the truth, I wanted to become familiar with Git, to test its benefits.

As I was not starting from scratch, I wanted to be able to keep the history of my project

There is an easy way to do so in GitHub : create a new project, and then select ‘Importing a Subversion repo’; but unfortunately, it did not keep track of all my tags; so I chose to go with the manual way.

Export google code svn repo to a local git repo

I use this utility called svn2git, I had to check out from GitHub : svn2git

It is written in Ruby, so you have to make sure you have ruby and rubygems installed.

Then, according to your gems directory, you can run it :

/var/lib/gems/1.8/bin/svn2git https://g2android.googlecode.com/svn/  –verbose

All the svn tags are imported as git tags (git tag to check all the tags are there) but the authors are not imported (« (no author) author not found in authors.txt)

To manually change the authors name, I found the solution reading this post.

git filter-branch -f --env-filter "GIT_AUTHOR_NAME='Newname';  GIT_AUTHOR_EMAIL='newemail'; GIT_COMMITER_NAME='Newname';  GIT_COMMITTER_EMAIL='newemail';" HEAD </code>

From now on, you have a local git repo with all your code  from your SVN (including the tags !) in a local git repo; let’s share this one !

First create a new repo on github (using the web interface), then set your local repo as the origin :

git remote add origin git@github.com:anthonydahanne/ReGalAndroid.git

Finally, sync the local and remote , including the tags of the master branch :

git push origin master --tags

And youŕe done migrating from Google code SVN to GitHub !

Migrating your users from a previous android app to a new one

The problem when you’re creating a new android application to replace a previous one, is that the users having the old app installed on their phone will not be notified of the availability of the new one, since the package name of your new application (referenced in your AndroidManifest.xml file) is different :

<manifest xmlns:android="http://schemas.android.com/apk/res/android"
package="net.dahanne.android.g2android" android:versionCode="15" android:versionName="1.6.3">
 
</manifest><manifest xmlns:android="http://schemas.android.com/apk/res/android"
package="net.dahanne.android.regalandroid" android:versionCode="2" android:versionName="1.0.1">
</manifest>

Then the market you’re using (let’s say Android Market) is not aware of your new application replacing the previous one.

For the particular case of the migration from G2Android to ReGalAndroid, I first, during a month, advertised the new app on Twitter, and on the dedicated G2, G3 and Piwigo forums and websites; so that I could have some early adopters.

Then, in a second step, to accelerate the migration, I updated G2Android, with a new activity, launched by default on startup to tell the user that the application is not maintained anymore and that he should tap on the button to download the new application.
Here is the source code to redirect users from the old android app to the new one

At the time of writing this blog post, here are the statistics reported by Android market about both apps :

It means that the migration is quite slow, since many users still have the old application installed on their phones; I may remove the old app from the market in the next few months to make sure they do not use it anymore (please note that ReGalAndroid is also available via other markets and also via direct download)

The final words

The application is open source and if I receive many issues and functional requirements via the ReGalAndroid bug tracker (and I thank the users for that, it clearly influences the path the application is taking) I would love to see new contributors and committers add some code (see the ReGalAndroid github repository); because this is before all what makes the app

Here are the steps I followed to create this project.

Creating this project, I wanted a lightweight Java library, easily embeddable in an Android application  (G2Android not to name it, a [...]]]> I finally started a java library to implement Gallery3 REST remote API; g3-java-client (the name is not really original huh ?!)

Here are the steps I followed to create this project.

The technical choices

Creating this project, I wanted a lightweight Java library, easily embeddable in an Android application  (G2Android not to name it, a Gallery2 client for Android).

So I carefully designed my dependencies, in order to leverage Android already included libraries, such as Apache Http Client, org.json packages (by the way I initially wanted to use the jersey REST client API, but… I wanted to stay as close as possible to the Android platform…)

Finally, I’m developing this library using TDD, and Maven as a CI tool :  I want a really reliable library, for 2 reasons : because it is motivating to create quality code, and also it is really painful to test an android application (you have to start an emulator or AVD and deploy the binary each time you launch a test, and each of these steps can take time) .

The licensing

I chose GPL v3 to make sure my work remains open source and not used in any commercial projects.

I’m using maven-license-plugin to add the license to the source files.

The hosting

I chose Github, mainly because I wanted to get familiar with git : I have heard so many people happy using Git instead of SVN,  and I also wanted to be able to switch easily from trunk to another branch… sounded appealing to me !

To tag the sources, I decided to follow Ulrich’s steps using mvn-release-plugin to tag every new release (in French, but you can follow this link to have an equivalent article in English)

And to host the binaries, I created a maven repo out of  a new Github project following Chas’tutorial; sure I could have asked to host my project on Sonatype’s Nexus for Open Source Software, but as the project is actually really young, I prefer to use my own repo; I guess later on I’ll switch to this better solution.

The commands used

If I had to initialize the workspace from another computer, I would type the following commands :

mkdir g3-java-client
git clone git@github.com:anthonydahanne/g3-java-client.git
cd g3-java-client/
#to make sure everything (compilation, testing) is ok
mvn clean install
#to add the license header to the source files
mvn license:check
mvn license:format
#to commit the changes to github
git commit -a -m 'commit message'
git push origin master
#to release a new version  (be careful there is no file named release.properties at the root of the project)
mvn release:prepare

We just added the license info, committed the snapshot version , committed the stable version and also created a new branch (mvn release:prepare)

Now, we can deploy it to a maven repo

#to deploy the current version to a repository on my local file system :
mvn -DaltDeploymentRepository=snapshot-repo::default::file:../../m2repo/snapshots clean deploy
#to commit this repo to github (I created this Github repo to serve as a maven repo http://github.com/anthonydahanne/m2repo)
cd ../../m2repo/snapshots
git commit -a -m 'new snapshot for g3-java-client'
git push origin master

Pretty clear and easy right ? (except the deploy phase…) all of this thanks to Maven and my pom configuration.

What’s next ?

First, finish implementing all the Gallery3 REST remote API commands; and then, create a new android project (G3Android ?!) to consume this library; using … maven-android-plugin of course !

or you can read it, Asynchronous [...]]]> Tonight I gave a presentation at Android Montreal about the AsyncTasks API, for the Android platform.
Although the main content was « said » during the presentation, you can download the support :

• Asynchronous Tasks with Android (OpenDocument format)
• Asynchronous Tasks with Android (PowerPoint format)
• Asynchronous Tasks with Android (PDF format)

or you can read it, Asynchronous Tasks with Android on slideshare.

If you want to compare the different ways to deal with multi threading on the android platform, you can also have a look at the project I created on Github, AsyncTasksDemo.
Also, you can have a look at the source code of G2Android, in this application I use a lot of AsyncTasks to be able to run worker threads fetching remote content over HTTP (you can particularly have a look at AddPhotosTask)
Many thanks to the Android Montreal volunteers for hosting my presentation! It’s a really nice event with bright developers !

References :

if you’re interested in the AsyncTask API, you MUST read those blog posts from Android developers :

• Painless threading, by Romain Guy
• Multithreading For Performance, by Gilles Debunne

-Xdebug -Xrunjdwp:transport=dt_socket,address=8787,server=y,suspend=y

So for example, if you want to debug an ant launch (launched via « ant -f myBuild.xml » for example), you simply have to set ANT_OPTS; on windows :

set ANT_OPTS=-Xdebug [...]]]> As long as you’re using a java application (ie an application calling your JVM), you can enable the debug mode with these parameters :

-Xdebug -Xrunjdwp:transport=dt_socket,address=8787,server=y,suspend=y

So for example, if you want to debug an ant launch (launched via « ant -f myBuild.xml » for example), you simply have to set ANT_OPTS; on windows :

set ANT_OPTS=-Xdebug -Xrunjdwp:transport=dt_socket,address=8787,server=y,suspend=y

Another example, with JBoss :

set JAVA_OPTS=-Xdebug -Xrunjdwp:transport=dt_socket,address=8787,server=y,suspend=y

and then you can debug your launch with Eclipse, for example, creating a new debug configuration (remote java application on port 8787) !
Option : you can choose whether or not the application should wait for your debugger setting suspend to y or n.

Now, you can still see your albums and pictures, but you can also :

The ergonomy is not yet perfect, as you may have [...]]]> I’m writing this small post just to indicate that the new version of the open source Gallery2 client for the android platform, G2Android, is out !

Now, you can still see your albums and pictures, but you can also :

• upload photos
• create new albums

The ergonomy is not yet perfect, as you may have already noticed, even though almost everything is multi threaded (using AsyncTask), meaning the ui won’t freeze anymore; the next features will probably will allow the user to insert a photo to his gallery from every picture context on the android platform (declaring a « share image » intent)

Stay tuned !

This method is [...]]]> In this article, I invite the reader to discover the different methods to access http resources from the Android platform.
These methods can be adapted to access web services (using REST) or simply to download files !

First Method : getting an input stream given a simple url from Android using HttpURLConnection

This method is the most basic one : it allows you, using the basic HttpUrlConnection, ( contained in java.net) to get an InputStream from an Url :

private InputStream downloadUrl(String url) {
		HttpURLConnection con = null;
		URL url;
		InputStream is=null;
		try {
			url = new URL(url);
			con = (HttpURLConnection) url.openConnection();
			con.setReadTimeout(10000 /* milliseconds */);
			con.setConnectTimeout(15000 /* milliseconds */);
			con.setRequestMethod("GET");
			con.setDoInput(true);
			con.addRequestProperty("Referer", "http://blog.dahanne.net");
			// Start the query
			con.connect();
			is = con.getInputStream();
		}catch (IOException e) {
                        //handle the exception !
			e.printStackTrace();
		}
		return is;
 
	}

You can also use the Post method, sending data in the HTTP POST payload :

private InputStream downloadUrl(String url) {
                InputStream myInputStream =null;
		StringBuilder sb = new StringBuilder();
                //adding some data to send along with the request to the server
		sb.append("name=Anthony");
		URL url;
		try {
			url = new URL(url);
			HttpURLConnection conn = (HttpURLConnection) url.openConnection();
			conn.setDoOutput(true);
			conn.setRequestMethod("POST");
			OutputStreamWriter wr = new OutputStreamWriter(conn
					.getOutputStream());
                        // this is were we're adding post data to the request
                        wr.write(sb.toString());
			wr.flush();
			myInputStream = conn.getInputStream();
			wr.close();
		} catch (Exception e) {
                        //handle the exception !
			Log.d(TAG,e.getMessage());
		}
                return myInputStream;
}

But there are better ways to achieve that, using Apache HttpClient, included in android.jar (no need to add another jar, it’s included in android core)

Second Method : getting an input stream given a simple url from Android using HttpClient

Why is it a better to do it ? because the simpler, the better ! See by yourself :

public static InputStream getInputStreamFromUrl(String url) {
		InputStream content = null;
		try {
			HttpGet httpGet = new HttpGet(url);
			HttpClient httpclient = new DefaultHttpClient();
			// Execute HTTP Get Request
			HttpResponse response = httpclient.execute(httpGet);
			content = response.getEntity().getContent();
                } catch (Exception e) {
			//handle the exception !
		}
		return content;
}

But you maybe wondering if it’s still easy with HTTP Post method ? You won’t be deceived !

public static InputStream getInputStreamFromUrl(String url) {
		InputStream content = null;
		try {
          		HttpClient httpclient = new DefaultHttpClient();
			HttpPost httpPost = new HttpPost(url);
			List nameValuePairs = new ArrayList(1);
                        //this is where you add your data to the post method
                        nameValuePairs.add(new BasicNameValuePair(
			"name", "anthony"));
			httpPost.setEntity(new UrlEncodedFormEntity(nameValuePairs));
			// Execute HTTP Post Request
			HttpResponse response = httpclient.execute(httpPost);
			content = response.getEntity().getContent();
		        return content;
	        }
}

But what if you want to read a cookie from the response ? And how can you send a cookie back to the server for the next request ?

Reading / Sending a cookie along with the requests

Using Apache HttpClient, it’s easy to retrieve cookies ! Everything is in the headers after all !

[...]
Cookie sessionCookie =null;
HttpResponse response = httpclient.execute(httpPost);
Header[] allHeaders = response.getAllHeaders();
CookieOrigin origin = new CookieOrigin(host, port,path, false);
for (Header header : allHeaders) {
			List parse = cookieSpecBase.parse(header, origin);
			for (Cookie cookie : parse) {
				// THE cookie
				if (cookie.getName().equals(COOKIE_I_WAS_LOOKING_FOR)
						&& cookie.getValue() != null && cookie.getValue() != "") {
					sessionCookie = cookie;
				}
			}
}

To send a cookie along with your request, keep it simple :

HttpPost httpPost = new HttpPost(url);
CookieSpecBase cookieSpecBase = new BrowserCompatSpec();
List cookies = new ArrayList();
cookies.add(sessionCookie);
List cookieHeader = cookieSpecBase.formatCookies(cookies);
// Setting the cookie
httpPost.setHeader(cookieHeader.get(0));

What about the resulting InputStream ? You definitely want to transform it into a String or an Drawable (to set it to an ImageView for example !) don’t you ?

Converting the InputStream into a Drawable in Android

The Drawable class already handles that for you :

Drawable d = Drawable.createFromStream(myInputStream, "nameOfMyResource");

Converting the InputStream into a String in Android

This is some classic java stuff (don’t tell about how easier it is in Ruby.. I know … but hey ! Java SE7 at the rescue with NIO !!! maybe one day in 2010 ! )

BufferedReader rd = new BufferedReader(new InputStreamReader(myInputStreamToReadIntoAString), 4096);
String line;
StringBuilder sb =  new StringBuilder();
while ((line = rd.readLine()) != null) {
		sb.append(line);
}
rd.close();
String contentOfMyInputStream = sb.toString()
 
That's it folks ! If you have any other methods to achieve these goals, feel free to share them sending a comment !

During his talk, the speaker described the audience the future of a Java platform, because it is actually going to be there for a long time now. He mentioned [...]]]> Vous pouvez lire cet article en français sur le blog de valtech.

The Changing Nature of Enterprise Java Application Development, 9:00 – 9:45 by Adrian Colyer

During his talk, the speaker described the audience the future of a Java platform, because it is actually going to be there for a long time now.
He mentioned the dynamic languages compatible with the JVM : Groovy, JRuby, Scala, and Clojure.
He mainly focused on Groovy (which belongs to SpringSource by the way), even if he was not sure if whether JRuby or Groovy would win the « war of dynamic languages ».
For the web platform, he mentioned Grails, built on top of Groovy, Rails running on top of Ruby, and Lift built on top of Scala.
He finished his talk introducing OSGi and Spring DM; the future of enterpise is application may be in the cloud : easily to deploy and scale applications are to come.

The android runtime environment, 10:30 – 11:20, by Joerg Pleumann

The speaker first made an introduction to the Android system, as a platform competing windows mobile and palm.
More technically speaking, it is based on a Linux kernel (along with necessary drivers : display, camera, sound, etc…).
On top of it come the Android Runtime and libraries; the talk is about the Android Runtime and its Dalvik VM.
Dalvik VM : CPU between 250 and 500 MHz, RAM >=64 (equivalent to a 1999 PC), but low power consumption is critical.
The Dalvik VM is really efficient, using explicit registers to keep temporary values: the immediate benefit is the number of lines of the bytecode resulting from the DalvikVM compilation : nearly 30% less than a standard VM compilation.
A dex file (Dalvik Executable) contains several files, and is not compressed.
JNI is ready, a sdk for using JNI is about to come out in the next days; also an Android Debug Bridge is available from the beginning.
The core libraries
dalvik vm specific libraries : dalvik.*
java compatibility libraries : java.* and javax.*
and some apache libraries(mainly for http) : org.apache.*
The core libraries are partly taken from Apache Harmony (a JDK implementation by Apache), partly written from scratch; and some of them have been optimized (using JNI)
The core libraries are mostly compatible to a subset of J2SE 5.0.
JDBC 2.0 is supported, but the SQLite driver is limited., but one can use the alternative in android.database.sqlite which are better integrated.
To conclude with, with Android you can reuse your java knowledge and tools.

Android Application Model, 11:30 – 12:20, by Dominik Gruntz

4 main components :activities, services, content providers and broadcast receivers.
A single class, uses intents to communicate with the other activities; when you launch an activity, you can (or not) expect a result; if this is the case, the method onIntentResult will be called.
You can also define intent filters to describe the intents an activity can handle.
An intent specification is on its way : http://www.openintents.org
An application is packaged using an APK (Android package).
Every activity has a lifecycle containing 4 steps : new, running, paused, stopped, killed and the corresponding methods (onCreate, onStop etc..)
For each activity a new process (with a single thread) is created.
The Looper.loop() dispathes the actions to a queue.
If an activity does not conume events in 5sec, the system issue a warning message.
A service is running background and can communicate with an activity (in process if it’s the same application (APK)) or via AIDL.
Finally, a content provider is the only way to share date between Android applications; it can using SQLite, a file, or a remote storage.

Scalable Agile Web Development: REST meets JCR meets OSGI, by Michael Marth

JCR : JSR 270 for accessing content repository : located on a file or database; can be versioned; node ahave read/write group properties, etc…
Apache Sling is a web framework, on top of REST/Scripting (Ruby, Groovy, etc…)/OSGi(in this case Apache Felix Container, to switch on/off part of the frameework live)/JCR.
So basically Apache Sling allows the user to create/edit/read/update content in a restful manner (http methods), the data being created is then stored on the JCR repository, organized in a very similar WebUrlFriendly way (and then you can manipulate the data through rest or webdav and so on, thanks to JCR)
The demo was really smooth and interesting !

The charm of Mockito: Test Spies in action, 14:30 – 15:20, by Szczepan Faber

The speaker begins asking who uses JMock : 2 or 3 people among an audience of 60 people, half of the audience using EasyMock.
So let’s start with EasyMock : coding in TDD a dictionnary, (and its history) depending on a translator : the translator is mocked.
With easymock the speaker keeps it old fadhioned : createMock/expect/replay/assert/verify , (his test extends Assert) and then refactor a bit in the @Before and @After test methods, nothing special here.
Then Mockito : mock/when/assert/verify : it’s a lighter syntax indeeed!
Szczepan insisted on the fact that usually developers using easymock create replayAll and verifyAll methods called in the @After method; which make Easymok fail when many behaviours are attached.
The speaker uses BDD style template (given/when/then) in eclipse, when creating a test.
Mockito eases the use of a given/when/then syntax when writing a test.
Has good point of failure ?
The failure trace in mockito is linked to the source code and more explicit about the error, it does not whow internal classes exceptions
Mockito or mockito-like for python, flex, javascript, C++; there are mockito extensions
He concluded with : « Do TDD and write great tests with whatever tools that work for you »
May the force be with you !

Eclipse Galileo and JBoss Tools, 16:30 – 16:50, by Max Andersen

Galileo (3.5) is a huge release.
Must-sees : Eclipse Memory Analyser (to see which plugin consumes all the memory for example), using jmap
XML and XSL editor is really better.
P2 (eclipse update manager) works for real; you shouldn’t use dropins anymore (single zip files to unpackage in /plugins and /features), for offline installation plugins providers should suggest copies of the update site, then you may use P2 to explore these files, as if you were online).

First a nice little video from Java One, introducing the history of Java ! Very funny how Java conquered the world [...]]]> The following sessions were given during Jazoon 09, on Tuesday the 23rd of June 2009, vous pouvez lire la traduction en français sur le blog de Valtech

James Gosling keynote : 9:50 to 10:45

First a nice little video from Java One, introducing the history of Java ! Very funny how Java conquered the world !
Then James Gosling begins his speech.
Today there are more JREs installed than people on earth ! It can be be smart cards, desktops, phones… 6M Java developers today in the world … 15M JRE downloads / week.
You can now view Java as « Learn One, Work everywhere » because it’s really spread everywhere now !
James Gosling then describes the brazilian healthcare, where every piece of paper is now on the network : prescriptions, x rays, everything is now on the network (thanks to Java !), ebay, metropolitan networks cards, web apps, games… Java is everywhere !
But the main thing is not java as the language, but the JVM that now hosts Ruby, Scala, JavaScript, etc…
James then showed us a bit of java code : a servlet for Java EE6 running on GlashFish v3, changed, saved and instantly shown just refreshing the browser ! No more deployment bottleneck !
He also introduced Netbeans, as the best IDE ! version 6.7 now integrates with project kenai, a sourceforge like (you can check in/out merge your files, etc… from Netbeans to kenai), also some better maven support.
Java is also used in the real time world, even in a car that James turns into electric : a lincoln (lincvolt.org) which is hybrid, connected to the internet, running java code for measurement and control.
Now down to the performance part : Java is highly optimized, genrally beats C/C++, GC is a lot faster than malloc/free !
Multithreading is the future of software designing : because the processors speed is not evolving anymore, we’ve got to think multicore !
Java is also running on cellphones, JSR 248, Mobile Service Architecture aggregates all Java mobile specs.
JDK7 will be modular (see Mark Reinhold’s blog), dynamic language support will be available too (JSR 292?); waiting for Java 7, there are still Java SE 6 updates : better browser and desktop integration are on their way !
Of course, we got a few slides about JavaFX, which can run on desktops, mobile phones, set top boxes, etc…
To finish his speech, James reminded us with all the java communities : netbeans.org, java.net, glassfish.org, etc…

Java EE 6 11:00 – 11:50 by Roberto Chinnici from Sun

Java EE6 was too complicated, so they decided to make it more easy to use, more powerful, and more flexible.
That’s why they created the profiles, letting the users choose the features they need.
For example the web profile includes everything for the web development, including the EJB 3.1 lite, and bean validation.

Servlet 3.0

New annotations @WebServlet, @WebFilter, @WebListener, the possibility to register new web components just adding them to WEB-INF/lib directory, a new async mode using just a single thread for many requests

JAX-RS 1.1 : RestFul webservices

This API is there to abstract low level HTTP details : using @GET and @POST (etc..) you can define your retrieval and creation REST operations, JAX-RS translates the XML to the java entities (and back) automatically for you (you can also use JSON)

Bean Validation 1.0 (JSR-303)

This API is here to validate fields (text inputs for example)
Seems to be really verbose just to define constraints… We’ll see…

EJB 3.1

No intefaces anymore ! One source file per bean.
New timer annotation @Schedule allowing some cron-like scheduling in the container : a great example provided is the case of a cache refreshing ! No more Mbeans called externally with crons, everything can be integrated in the application code !
EJB 3.1 lite can run on Java SE, great for testing and also in a war file !
Oups, we’re running out of time, the speaker skipped JPA 2.0

« Design Patterns » in Dynamic Languages 12:00 – 12:50 by Neal Ford from ThoughtWorks (nealford.com)

After comparing mixing dynamic languages with whisky blending, Neal Ford reminded us about the history of Design Patterns : basically for « making C++ code suck less » !
But the point is : the design pattern designed in 1994 are not always useful today : for example the iterator is quite complicated, whereas today in dynamic languages, like Ruby a collection.each{[x| puts x} is so much more… efficient !
Command design pattern is already built in closures !
The same for strategy, template, interpreter (the example showed how easy it is to implement a DSL in Groovy, defining new behaviors in the Integer class), decorator, recorder, adapter
Then Neal described some other desgn pattern for dynamic languages : the null pattern (?. in groovy) and the aridifier pattern.
To conclude with, Neal suggest that we should try the tools provided with the languages we’re using instead of blindly trying to copy the patterns designed for older, other languages !

Gradle : A build system for Java 14:00 – 14:50, by Hans Dockter

First the speaker reminds us of Ant and Maven.
With Ant, it’s always been repetitive, with a lot of copy and paste when a new project starts, due to Ant limited domain model (resources, target, properties,)
There is no place for listener in an Ant build…
A gradle script can import old ant build scripts.
With Gradle, you can call ant tasks, and even implement listeners on your ant task… An ant task called in a gradle scrip can even be aware of the next task that will be called !
Maven, on its side is a build by convention framework, which has introduced dependency management.
But Maven uses XML, and 25 lines of XML are required to launch a simple build !
Maven uses transitive dependency whereas Gradle is able to skip tasks (which can be handy when a Maven dependency is broken…)
Gradle aims to be a build language rather than a build framework; because when a framework does not implement a functionality, you spend your time fighting against the framework to make him do something it was not intended to do.
You can define your maven build with a groovy script using gradle.
With gradle, it becomes easy to define usually manual and entreprise specific tasks like release management or deployment.
Grandle can use Apache Ivy or Maven dependencies, it can even deploy a build to a maven repository.
There is also a Gradle Wrapper to start using Gradle without installing it : the Continuous integration machine will download Gradle for you and execute the gradle script (which is a shell script downloading gradle) !
What’s next? Gradle 1.0 will be out in Autumn.
Q&A :
What open source projects are already using Gradle ?
Grails, but others are to come.

Next Generation Enterprise Builds: Maven, Mercury, and Tycho (for OSGi builds) by Jason van Zyl from Sonatype

Maven 3.0

Better integration in IDEs, faster, rewritten artifact resolution system, etc…
Maven 3.0 is more focused on integration tests, provides a new version of the plugin manager, a queryable lifecycle to make the tools able to predict what’s going to be done and also able to ask to execute a particular step.
Maven 3.0 uses Mercury : a new repository and transport layer, that can be used with others systems (Ruby gems, Eclipse P2 dependency management)

M2eclipse

M2eclipse has also been improved, now m2eclipse won’t call every maven tasks on a a save action, it will only call the appropriate ones.
The m2eclipse team would like the users to be able to resolve every maven operations from Eclipse,
M2eclipse is now compatible with Flex (out by october 2009).

Nexus

Only 5% of the maven users use a repository manager.
Nexus is also for binary sharing inside the enterprise, it a repository manager, it’s like a source control management, but for binaries !
Nexus main advantage is also the caching system it provides.
Nexus can be queryed through m2eclipse, Nexus API can be used through REST, which makes it very easy to use.

Hudson

For its next version, hudson will integrate :d rools workflow integration (with a workflow editor), a rest layer to query hudson, unifying hudson m2eclipse, nexus better maven integration

Testing Zen 16:00 – 16:20 by Thomas Mueller, Day

In this session Thomas gave us the reasons why we should test our code.
Then he continued giving us an example of fuzz testing (data generation to test)
Code coverage (with Emma) to find untested and dead code.
Database testing is often hard : a pain to install… so you can use a Java SQL Database like HSQL (a good one), H2 (the best, and also the one the speaker is working on ! ) or Derby (the worse).

Creating Performance Test Data with benerator, by Volker Bergmann

Usually performance testers prefer testing with production data, but sometimes it is not possible (for privacy reasons or if the system is new).
In those cases, you have to use a data generator, like Benerator that can create and anonymize data, to DB, file, LDAP, JCR, etc…

To finish the day, we had a really nice speech from Neal Ford, about technology predictability, that was real fun !