intransitione.com

MDRadialProgress Examples

marco.dinacci@intransitione.com (Marco Dinacci) — Fri, 14 Feb 2014 00:00:00 +0000

I dropped by 50% the non-attribution licenses for my MDRadialProgress component until the end of next week, go get one !

MDRadialProgress is an Open Source iOS UIView that allows to represent progress in discrete steps. It is useful in those circumstances where you have a task or anything else that has a fixed number of steps and you want to represent the progress.

Last time I wrote about MDRadialProgress I promised some technical documentation, well here are some examples to make you started with this component:

Setup

First, let's prepare the view and create a frame that will contain the progress view:

self.view.backgroundColor = [UIColor whiteColor];	
CGRect frame = CGRectMake(self.view.center.x - 100, self.view.center.y - 100, 200, 200);
MDRadialProgressView *radialView = [self progressViewWithFrame:frame];
[self.view addSubview:radialView];

That's it, we just created a 200x200 frame and set its coordinates in order for our view to appear centered. We also added it to our main view. That wasn't too hard.

Basic View

Let's start with the most basic view you can possibly make. To the code you've written above, add:

radialView.progressTotal = 7;
radialView.progressCounter = 5;

progressTotal represents the total number of steps in the view. progressCounter keep track of the number of the steps completed. By default, completed progress is represented in green, uncompleted progress in light gray. You can also see that there's a divider between each slice, which has the same colour as the center color (in this case white) and that there's a number in the middle, which represents the completion percentage.

Changing colors

Let's say you want to change the colours, to make the component fits better into your UI.

radialView.theme.incompletedColor = [UIColor colorWithRed:0.15 green:0.3 blue:0.45 alpha:0.4];
radialView.theme.completedColor = [UIColor colorWithRed:0.15 green:0.3 blue:0.45 alpha:1.0];

As you can see, we're actually using the same colour, just with a different alpha value. You don't have to use the same color though. You may be asking yourself, what's the theme property all about ? We'll see that later, for the moment just know that you may want to have multiple radial views on the same screen sharing the same colours and settings. A theme encapsulate those settings.

Showing/hiding percentage and slice dividers

You may not want to show the percentage label in the middle of the view, fair enough, let's just hide it:

radialView.label.hidden = YES;

(This should be part of the theme too, it will probably change in the next release). Also, let's hide the slice dividers so that the progress look like a single uninterrupted arc.

radialView.theme.sliceDividerHidden = YES;

Hiding more stuff

Is the incompleted arc really worth showing ? That's up to you, but in case you want to hide it and show only the progress done, you can, just set its color to clearColor:

radialView.theme.incompletedColor = [UIColor clearColor];

The progress now looks smoother, with no interruptions.

Changing the center color

It might not be very pretty in this example, but you can change the center color yoo if you want:

radialView.theme.centerColor = [UIColor colorWithRed:0.45 green:0.3 blue:0.15 alpha:0.1];

Changing the starting point and direction

As you certainly noticed, the progress start drawing from 12'clock and proceeds in a clockwise direction. You can change both these settings very easily:

radialView.clockwise = NO;
radialView.startingSlice = 2;

The progressCounter in this picture was set to 3.

Changing divider color and thickness

Until now we haven't changed the "thickness" of the area where the progress is drawn. We also never changed the slice divider color.
This example shows you how to implement these customisations. Since I made a few changes since the original example, I'm showing all the code required to reproduce the picture above:

radialView.progressTotal = 7;
radialView.progressCounter = 3;
radialView.theme.thickness = 60;
radialView.theme.sliceDividerColor = [UIColor colorWithRed:0.80 green:0.80 blue:0.80 alpha:0.6];
radialView.theme.incompletedColor = [UIColor colorWithRed:0.15 green:0.3 blue:0.45 alpha:0.4];
radialView.theme.completedColor = [UIColor colorWithRed:0.15 green:0.3 blue:0.45 alpha:1.0];
radialView.label.hidden = YES;

As you can see, our progress view looks quite different this time.

Themes

The last feature I want to show is themes, which can be used when there are multiple views that must share the same settings. We'll first create a MDRadialProgressTheme object:

MDRadialProgressTheme *newTheme = [[MDRadialProgressTheme alloc] init];
newTheme.completedColor = [UIColor colorWithRed:90/255.0 green:212/255.0 blue:39/255.0 alpha:1.0];
newTheme.incompletedColor = [UIColor colorWithRed:164/255.0 green:231/255.0 blue:134/255.0 alpha:1.0];
newTheme.centerColor = [UIColor clearColor];
newTheme.centerColor = [UIColor colorWithRed:224/255.0 green:248/255.0 blue:216/255.0 alpha:1.0];
newTheme.sliceDividerHidden = YES;
newTheme.labelColor = [UIColor blackColor];
newTheme.labelShadowColor = [UIColor whiteColor];

Then, we'll create 4 radial views and assign the theme to each one of them in the constructor:

int x = 50, y = 50;
CGRect frame = CGRectMake(x, y, 80, 80);
MDRadialProgressView *radialView = [[MDRadialProgressView alloc] initWithFrame:frame andTheme:newTheme];
radialView.progressTotal = 4;
radialView.progressCounter = 1;
[self.view addSubview:radialView];

y += 120;

frame = CGRectMake(x, y, 80, 80);
MDRadialProgressView *radialView2 = [[MDRadialProgressView alloc] initWithFrame:frame andTheme:newTheme];
radialView2.progressTotal = 4;
radialView2.progressCounter = 2;
[self.view addSubview:radialView2];

y += 120;

frame = CGRectMake(x, y, 80, 80);
MDRadialProgressView *radialView3 = [[MDRadialProgressView alloc] initWithFrame:frame andTheme:newTheme];
radialView3.progressTotal = 4;
radialView3.progressCounter = 3;
[self.view addSubview:radialView3];

y += 120;

frame = CGRectMake(x, y, 80, 80);
MDRadialProgressView *radialView4 = [[MDRadialProgressView alloc] initWithFrame:frame andTheme:newTheme];
radialView4.progressTotal = 4;
radialView4.progressCounter = 4;
[self.view addSubview:radialView4];

Done ! Don't forget to grab the source on Github and buy a non-attribution license if you require one.

MDRadialProgress 1.0

marco.dinacci@intransitione.com (Marco Dinacci) — Mon, 09 Sep 2013 00:00:00 +0100

Did you know ? you can buy a Non-Attribution License for MDRadialProgress view on my online license store.

Last time I've written about my MDRadialProgress control was when I had just released the very first 0.1 version. After several versions, a few bug fixes and new features, I'm finally releasing a new version that contains all features I wanted to implement for a 1.0 release. With MDRadialProgress now you can:

* Draw progress in separated slices or as a single arc
* Can customize all colors, the distance of the slices and their thickness
* Clockwise and anti-clockwise drawing order
* Can start progress from any slice, not necessarily from the top
* Display a label with the current progress
* Themes support for easier customization of multiple views
* Fully accessible
* Support iOS 6.0+
* Original BSD and Commercial Non-Attribution licenses available (you can use it in commercial software)

The biggest change is probably the license. I realized in the last years that maintaining Open Source software is very consuming and I don't have the time to keep adding features and fixing bugs for free. Although I'm not still entirely sure about this, I decided to switch the license for MDRadialProgressView from MIT to old BSD.

The old BSD license, or 4-clause BSD, contains an attribution clause that has since been removed from the "New" BSD license, that forces the people using the licensed software to provide attribution to the original owner in any advertising material mentioning features or use of the licensed software.
I'm not entirely happy to have included such clause, but I couldn't find a better, existing license to allow me to distribute my software free and at the same time monetize it by forcing attribution on people and companies that don't want to provide it. That said, I'm happy to re-discuss licensing agreement with anybody interested.

MDRadialProgressView is on sale on my online license store and you can buy a single or an unlimited license.

I'll follow soon with a post containing more technical documentation.

Marco

Review of Python For Data Analysis

marco.dinacci@intransitione.com (Marco Dinacci) — Thu, 09 May 2013 00:00:00 +0100

The book is essentially a reference on how to use the pandas Python library. It is well written and can be used both as a tutorial for learning about a particular aspect of the library and as a reference.

The first two chapters introduce the technologies and iPython, an advanced interactive Python shell. The third chapter is about NumPy. NumPy is introduced because it's the basis used by many other scientific libraries, and it helps the readers getting acquainted with array programming. It is also used extensively in pandas.

Pandas, an high-level library for data analysis written by the author itself, is introduced in the fourth chapter. Along with pandas, matplotlib is discussed in order to visualize the data from pandas. There are enough examples where the author use public available data from the internet and show how to manipulate it, clean it with pandas and visualize it with matplotlib. I found these examples much more useful than the dry explanation of the library features.

Overall, I recommend it to anybody with basic Python programming skills wishing to learn more about pandas and matplotlib.

Software does not last

marco.dinacci@intransitione.com (Marco Dinacci) — Wed, 10 Apr 2013 00:00:00 +0100

Yesterday I was scavenging through some old projects of mine and I stumbled upon a video of a program I made in 2009. Here's the video on Youtube:

At that time I had quit my stable job in aerospace to make a sort of Trailblazer clone, my favourite game on my very first computer, a Commodore 16. By trying to bring the game to the 21st century I thought of adding curves and elevation so I implemented the track as a tessellated NURBS surface which was automatically generated from the control points drawn with a basic 2D editor. At this point the software was still a prototype but it was already usable.

Long story short, I never managed to finish the game as it took me too long and I ran out of money. It was a great learning experience though and after this my respect for indie (game) developers has grown tenfold.

This was all great fun but it's not the point of this post. The point is that I cannot run neither the editor not the game anymore.

I have all source code and assets but since I haven't archived the third-party libraries I was using the program doesn't run. I can probably find the old libraries I was using or adapt my code to the last version of these libraries. I could also archive them together with the project files so I could run it in the future (silly me not having done this before). But how far in the future ? I could save the libraries but how can I be sure that in 30 years the environment required to run the program will still be available ? Will Python still be available ? Will I be able to modify assets in Blender ?

This situation reminds me of an ex-collegue who was also an amateur photographer who decided to go back to analog photography as he wanted his photos to last through time and be visible to his grandchildren and their sons. He doesn't believe that JPEG images will be visible in 30 years. Well, nobody knows. I'm sure though that when a new and improved format will emerge (actually JPEG2000 has been available for a while but nobody seems to care, except for some folks working in satellite image processing) it will be possible to convert existing images to the new format. It will be a painful process and probably non professional users will never do it unless the software they use doesn't do it automatically for them.

If JPEG images will be supplanted, how will our legacy view our photos (assuming they have not been converted to a new format) ? Maybe one could include a standalone JPEG viewer along with the photos, but will the software run in 30 years ? I very much doubt.

I find it particularly disappointing that I won't be able to show my (future) children the software I've done in the past. A piece of music can last forever. Of course the medias (tapes, CDs, digital files) will obsolesce but I doubt music notation will ever fall in disuse. A painting can last hundreds of years, maybe thousands with the current conservation and restoration techniques. Constructions can last thousands of years. Words lasts hundreds or thousands of years.

Even if software is intrinsically eternal since it runs on a physical hardware it cannot last.

Music and words last because they are immediately accessible by our senses while software requires a translation medium (hardware) to interpret it so that we can access it.

Can we make software that is directly accessible to our senses without an interpreter which is inevitably condemned to obsolence like modern computers ? Will this require ourselves be the hardware ?

If we were the hardware, we could probably easily pass software to our children. And if we could pass software, will we be able also to transfer knowledge ? Can we store all digital information we produce within ourselves ? Some people apparantly think so.

Biocomputers are getting closer, I wonder if it will be possible one day to have a software that will adapt automatically to the changes in its environment without requiring any human intervention. I guess at that point humans won't be needed anymore...

Marco

MDRadialProgress 0.1

marco.dinacci@intransitione.com (Marco Dinacci) — Tue, 09 Apr 2013 00:00:00 +0100

I've been using (and still do !) several controls from cocoacontrols.com and from CocoaPods. It's great to see so many high quality components freely available, I can't calculate how many hours I've saved by using them.

I'm happy to finally contribute to these projects by releasing a simple custom UIView that can be used to represent progress in discrete steps. The inspiration comes from lumosity, whose iOS app uses a very similar component to represent the steps done in a game session. I'm quite convinced though not 100% sure that they're using an image instead of a customisable component.

The github project comes with an example (see ViewController.m) and documentation. You can install it by copying the MDRadialProgressView.{m|h} into your project or by using Cocoapod by adding this line pod 'MDRadialProgress', '~> 0.1.1' to your project's Podfile.

The simplest way to use it is to leave all properties with their default value and just customise progressTotal and progressCompleted:

// Create an instance of the view with a width and height (and consequently a diameter) of 50px. 
CGRect frame = CGRectMake(30, 30, 50, 50);
MDRadialProgressView *radialView = [[MDRadialProgressView alloc] initWithFrame:frame];

// Total number of steps    
radialView.progressTotal = 5;
// Number of steps completed
radialView.progressCurrent = 2;

// Customise colours and thickness here...see documentation on github

// Add to existing view
[self.view addSubview:radialView];

It's released under the MIT license, source code and documentation are on github.

Enjoy and let me know if you find it useful.

Marco

Watchlist Alert in the Mac App Store

marco.dinacci@intransitione.com (Marco Dinacci) — Wed, 06 Feb 2013 00:00:00 +0000

I developed a simple utility application called Watchlist Alert which monitors a user's eBay watchlist and notify him when the items in his watchlist are due to expire.

It started as an app to "scratch an itch" as at one point I was using eBay a lot and I was constantly forgetting about the bids I was watching. With a little bit more effort I polished and published it in the Mac App Store.

I'm planning to distribute it outside the Apple store too, mostly to see how complicated the whole process it.

If you happen to be using it and have any comments or suggestions, please leave them on the support website or just fire me an e-mail from the contact page.

Marco

Cross-platform desktop Java applications, JavaFX is not enough

marco.dinacci@intransitione.com (Marco Dinacci) — Thu, 04 Oct 2012 00:00:00 +0100

I just finished to watch the JavaOne 2012 Technical Keynote. Nice, not particularly impressive though. The "Windows moment" (JVM crash) was quite funny. It happens around the 52nd minute I think. No you can't skip and if you advance too much the video won't play. Flash fails.

What bothers me and the motivation why I'm writing this post is the excessive focus on the UI in JavaFX. They made it look like it is the only concern in a multi-platform app is the UI. WRONG.

The UI is only half of the problem. The other half, which becomes more and more complicated thanks to the advent of app stores and sandboxes is the integration with the host OS.

Where do you store application settings and data on Windows, Linux and OSX ?

The era of storing data in the application bundle itself or in ~/.appname or anywhere you want IS OFFICIALLY OVER.

Apple doesn't allow a sandboxed application to mess with the filesystem. Windows 8 won't either. You want to store settings ? put them in Application Support directory. Want to save documents ? do it in the Documents folder.

There's more, the problem is not only related to writing data. The OSX and Windows 8 sandboxes only allow you full access in the container (AppData on Windows 8) folder. Want to read the pictures folder ? you need to ask the OS where it is and have an entitlement (capability on Windows) declared in the Info.plist (Manifest file on Windows).

How is a Java application supposed to deal with this in a clean, cross-platform manner ?

Well you can't now, "write once run anywhere"...

They didn't show the dreaded GateKeeper warning dialog you get when you launch an application coming from an unidentified developer when they demoed their app on OSX. Why ? Because it wasn't sandboxed.

Moneydance is the first OpenJDK, sandbox-compatible app available in the Mac App Store. It took us months to make it ready. I contributed at least four patches to the OpenJDK OSX port, I had to fork and improve AppBundler because Oracle doesn't accept contributions on this Open Source project (unless it's one line of code...), and write native code where no other Java-only solution existed.
This whole process was much more time consuming that tuning the interface to be more OSX-like.

Fancy graphics and animations are pointless if an application can't be executed at all.

Oracle must improve the Desktop API and start eating its own dog food. Release complex software on the Apple App Store and start working towards a solution for the Windows 8 store, where the situation is even grimmer. If Oracle itself doesn't show that it's possible to deploy a Java app on the app stores who will ? We did our part with Moneydance, but it's not enough.

App stores, whether that's good or bad, are becoming the most accepted and simple way to buy software. What is the point of a cross-platform JavaFX app if you can't deploy it ?

I'm planning to start an Open Source project in the next few weeks to attempts to solve the problem of accessing user directories consistently over multiple platforms. I'll post more info once I'll have a foundation ready.

Marco

OpenJDK weirdness on OSX Leopard

marco.dinacci@intransitione.com (Marco Dinacci) — Wed, 03 Oct 2012 00:00:00 +0100

This post is to document a couple of issues encountered while deploying an OpenJDK 7 based app on OSX Leopard.
The first is a JVM crash caused by a supposedly closed bug in hotspot.
The crash happens when the sun.nio.cs.StreamDecoder.read is compiled at run-time, normally a fraction of a second after the application launches.
The only workaround I've found is to launch your application with the following line:

-XX:CompileCommand=exclude,sun/nio/cs/StreamDecoder,read

which tells hotspot not to compile that method.

The second issue is a NullPointerException in sun.net.www.ParseUtil.encodePath. Here's a partial stacktrace:

Exception in thread "AWT-EventQueue-0" java.lang.NullPointerException
  at sun.net.www.ParseUtil.encodePath(ParseUtil.java:105)
  at sun.misc.URLClassPath$JarLoader.checkResource(URLClassPath.java:734)
  at sun.misc.URLClassPath$JarLoader.getResource(URLClassPath.java:819)
  at sun.misc.URLClassPath.getResource(URLClassPath.java:195)
  at java.net.URLClassLoader$1.run(URLClassLoader.java:358)
  at java.net.URLClassLoader$1.run(URLClassLoader.java:355)
  at java.security.AccessController.doPrivileged(Native Method)
  at java.net.URLClassLoader.findClass(URLClassLoader.java:354)
  at java.lang.ClassLoader.loadClass(ClassLoader.java:423)
  at sun.misc.Launcher$AppClassLoader.loadClass(Launcher.java:308)
  at java.lang.ClassLoader.loadClass(ClassLoader.java:356)
  ...

The line causing the NPE is:

char[] retCC = new char[path.length() * 2 + 16];

which means that somehow the path is null. As bizarre as it may seems this can be fixed using these two JVM options:

-XX:-UseLoopPredicate -XX:-LoopLimitCheck

which you may remember it was used to fix a nasty problem in hotspot in the first public release of Java 7.

Marco

Take your Java application to the Mac App Store.

marco.dinacci@intransitione.com (Marco Dinacci) — Sat, 30 Jun 2012 00:00:00 +0100

The AppBundler fork is still maintained but no longer by me. Forward your queries to the capable folks at the Infinite Kind. Please also note that this article has been last updated on the 29th October 2012 and I'm no longer actively maintaning it. Thanks.

Moneydance is on sale in the App Store!

In this guide I'll go through all the steps required to port your Java Swing application to OpenJDK with the goal of preparing it for the submission into the App Store. Although the App Store guidelines explicitly forbids applications to rely on deprecated or optionally installed technologies (Apple no longer bundles their JDK port so applications can't rely on the user to have it installed), you can still distribute your Java application on the App Store by embedding the OpenJDK 7 OSX port in a native OSX application.

It may sounds obvious but the procedure that follows can be performed only on OSX, not from Windows, Linux or any other OS.

Follow this steps:

Gather the requirements
Make your application sandbox-proof
Migrate existing data and preferences
Build and sign a package for your application bundle

and be sure to take a look at the known problems.

Requirements

OpenJDK

First of all, you need to acquire the OpenJDK OSX port. There are 3(½) ways:

Download the Oracle binary distribution.
Clone the Mercurial repository(recommended). Compile it with Java 6 or 7 if you already have it. Pick OpenJDK7u to get the stability of OpenJDK7 and the backwards updates from OpenJDK 8.
~~Download one of the excellent Henry Gomez's pre-built packages.~~Unfortunately Henry stopped its project but he was kind enough to release all the source code to build the packages.

Don't know which one to choose ?

Getting the source should be the only reasonable option if you're serious about bringing your application to the App Store. However if you want to get started quickly and are not interested in the bleeding edge version, download the Oracle distribution. If you want the latest version of the code but don't want to build it yourself, download one of the Henry Gomez's pre-built images.

Note that the Oracle bundle and the OpenJDK have different licences, the first is released under the Binary Code License agreement , the second under the GPL version 2 with Classpath Exception. Make your choice.

(½) The absolute best setup would be to clone the official repo and then replace the jdk subrepo with our fork which includes more bug fixes for OSX.The situation is a bit different from when I started writing this post. Oracle has integrated all my patches, so unless you require 32bits support the recommended option is to use an OpenJDK binary distribution (and apply the fix for freetype, see below).

AppBundler

Once you have finished downloading the OpenJDK you need to install AppBundler to package your application.

AppBundler is an ant task so just put the downloaded jar in the lib folder in your ant distribution and then follow the instructions to package your application. It still misses a few features like the ability to register your application with a particular file type but it allows to include a Java runtime within the application bundle and building an Info.plist file with most of the required keys.

AppBundler is still missing some basic features like file types association. I recommend instead using our fork.

Note that some keys are compulsory and your application won't be accepted if they're missing so be sure they're present. You must declare at least:

CFBundleIdentifier
CFBundleVersion
NSHumanReadableCopyright
LSApplicationCategoryType

Refer to Apple documentation for the meaning of these keys or just follow this example to be sure they're always included:

<target name="bundle" depends="...">
    <!-- Import the AppBundlerTaks from ant lib directory -->
    <taskdef name="bundleapp" 
           classname="com.oracle.appbundler.AppBundlerTask"/>

    <bundleapp outputdirectory="."
        name="MyApp"
        displayname="MyApp"
        identifier="com.foobarbaz.MyApp"
        shortversion="1.0"
        icon="MyApp.icns"
        mainclassname="MyAppMain"
        copyright="2012 FooBarBaz LLC."
        applicationCategory="public.app-category.entertainment">

        <!-- The directory where your OpenJDK runtime is. -->
        <runtime dir="/path/to/openjdk/Contents/Home"/>

        <!-- The bundleapp task doesn't support classpathref so all 
            the run classpath entries must be stated here too.
        -->
        <classpath file="MyApp.jar"/>
        <classpath file="dependencies.jar"/>
      
        <!-- Workaround since the icon parameter for bundleapp 
             doesn't work. (It's not a bug in AppBundler but 
             in the JavaAppLauncher, see Known Problems).
        -->
        <option value="-Xdock:icon=Contents/Resources/${bundle.icon}"/>

        <!-- OSX specific options, optional -->
        <option value="-Dapple.laf.useScreenMenuBar=true"/>
        <option value="-Dcom.apple.macos.use-file-dialog-packages=true"/>
        <option value="-Dcom.apple.macos.useScreenMenuBar=true"/>
        <option value="-Dapple.awt.application.name=MyApp"/>
        <option value="-Dcom.apple.smallTabs=true"/>

        <option value="-Xmx1024M"/>
    </bundleapp>
</target>

If you decide to use my fork, you can do much more:

<target name="bundle">
  <taskdef name="bundleapp" 
    classpath="appbundler-1.0ea.jar"
    classname="com.oracle.appbundler.AppBundlerTask"/>

    <!-- Note the usage of classpathref to avoid copy-pasting all 
    your classpath entries from another target. -->
  <bundleapp 
        classpathref="runclasspathref"
        name="MyApp"
        displayname="MyApp"
        identifier="com.foobarbaz.MyApp"
        shortversion="1.0"
        version="build 325"
        icon="MyApp.icns"
        mainclassname="MyAppMain"
        copyright="2012 FooBarBaz LLC."
        applicationCategory="public.app-category.entertainment">

      <runtime dir="${runtime}/Contents/Home"/>

      <!-- Specify which architectures you want to support -->
      <arch name="x86_64"/>
      <arch name="i386"/>

      <!-- Register the application as an editor for PNG and JPG files -->
      <bundledocument extensions="png,jpg"
        icon="${bundle.icon}"
        name="Images"
        role="editor">
      </bundledocument>

      <!-- Register the application as a viewer for PDF files -->
      <bundledocument extensions="pdf"
        icon="${bundle.icon}"
        name="PDF files"
        role="viewer">
      </bundledocument>

      <!-- Register the application with your custom format, 
      bundled as a package -->
      <bundledocument extensions="custom"
        icon="${bundle.icon}"
        name="Custom data"
        role="editor"
        isPackage="true">
      </bundledocument>

      <!-- Workaround since the icon parameter for bundleapp doesn't work -->
      <option value="-Xdock:icon=Contents/Resources/${bundle.icon}"/>

      <!-- OSX specific options, optional -->
      <option value="-Dapple.laf.useScreenMenuBar=true"/>
      <option value="-Dcom.apple.macos.use-file-dialog-packages=true"/>
      <option value="-Dcom.apple.macos.useScreenMenuBar=true"/>
      <option value="-Dcom.apple.mrj.application.apple.menu.about.name=${bundle.name}"/>
      <option value="-Dcom.apple.smallTabs=true"/>

      <option value="-Xmx1024M"/>
  </bundleapp>
</target>

See how I used classpathref to declare an external classpath, arch to declare the supported architectures and bundledocument to register the application with particular file types.

Make your application sandbox proof

Starting from the 1st of June all applications submitted to the App Store must be sandbox ready.

For a comprehensive explanation of what the sandbox is refers to the Apple documentation. In short though, the sandbox is an access control mechanism which restricts interaction of your application with the operating system.

The container is the directory where your application can safely read/write the files required for its correct functioning. It is physically located in ~/Library/Containers/yourapp.bundle.id; I doubt this is likely to change anytime soon so if you have a very simple application you may hardcode this path instead of relying on a native library to return it for you. I say may because I can't possibly know how Apple will react if you hardcode the path of the container instead of obtaining it by using the API.

If your application will require storing user documents in the container or writing temporary files you'll have to acquire the appropriate directories.One route is to write a JNI wrapper and bundle a small native library with your app, another is to use our AppBundler fork which pass the directories to the Java app as environment variables.

If you decide to use our AppBundler, from your Java app you can retrieve the directories using these system properties:

System.getProperty("LibraryDirectory");
System.getProperty("DocumentsDirectory");
System.getProperty("CachesDirectory");
System.getProperty("ApplicationSupportDirectory");
System.getProperty("SandboxEnabled");
System.getProperty("SandboxEnabled");// (the String "true" or "false")

Assuming your app identifier is com.acme.MyApp, the variable values will be:

// If the app is sandboxed:
/Users/user/Library/Containers/com.acme.MyApp/Data/Library
/Users/user/Library/Containers/com.acme.MyApp/Data/Documents
/Users/user/Library/Containers/com.acme.MyApp/Data/Library/Application Support
/Users/user/Library/Containers/com.acme.MyApp/Data/Library/Caches

// if the app is not sandboxed
/Users/user/Library
/Users/user/Documents
/Users/user/Library/Application Support
/Users/user/Library/Caches
false

Otherwise, if you decide to go with the native route, this is how you would create a native library and acquire the directories where you can write and store app-specific data:

UPDATE: javah generate the header files just fine, the problem was due to the fact I was using Apple's JavaVM framework based on JDK 1.6; instead, compile including the headers of the OpenJDK you're using, for example:

-I /Library/Java/JavaVirtualMachines/jdk1.7.0_06.jdk/Contents/Home/include -I /Library/Java/JavaVirtualMachines/jdk1.7.0_06.jdk/Contents/Home/include/darwin

and include the jni header

#include <JavaVM/jni.h>

Thanks to Scott Kovatch for pointing it out.

Here's a simple Makefile to generate the headers:

#!/bin/sh

DIST = jni_headers
CP = foobaz.jar
CLASS = com.foobaz.yourclass

all:
    mkdir -p $(DIST)
    javah -jni -d $(DIST) -cp $(CP) $(CLASS)

clean:
    rm *.o *.class

Now that you have your C header you can create a function that will return for example the Application Support folder for your application, which is located inside the appplication container:

JNIEXPORT jstring JNICALL 
Java_com_foobar_OSXAdapter_getApplicationSupportFolder
(JNIEnv *env, jobject jthis) {
    jstring path;
    
    @autoreleasepool {
        
        JNF_COCOA_ENTER(env);
        
        NSArray *paths = NSSearchPathForDirectoriesInDomains(
                                NSApplicationSupportDirectory, 
                                NSUserDomainMask, YES);
        NSString *basePath = [paths objectAtIndex:0];
        
        // Convert the NSString to a jstring
        const char *cString = [basePath UTF8String];
        path = (*env)->NewStringUTF(env, cString);
        
        JNF_COCOA_EXIT(env);
    }
    
    return path;
}

Migrate existing application data and preferences

If you're not releasing a new application chances is that you are already saving your application data and preferences somewhere, most likely in ~/Library/Application Support/YourApp.

Once your updated application is executed in the sandbox, that folder will be inaccessible for your program, but luckily Apple thought of a migration mechanism, which happens automatically the first time the user start the application. Here's an example of how you would migrate the preferences in ~/Library/Application Support/YourApp to ~/Library/Container/com.yourcompany.YourApp/Data/Library/Application Support/YourApp.

<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
<plist version="1.0">
<dict>
   <key>Move</key>
   <array>
      <string>${Library}/YourApp</string>
  </array>
</dict>
</plist>

Check the Apple documentation for more complex examples.

Build and sign the application bundle

This is where you need to have a Mac developer account. You could technically sign your bundle with a test key but you won't be able to submit your app nor to install it.

The process is made of three steps:

Write an entitlements file to specify the permissions your application requires.
Sign the bundle.
Package it.

The first step require writing a plist file listing the permissions you need. Head over to the Apple website and check the documentation to find out which keys are available.

I'll show a simple example and I'll later demonstrate how to include it in your bundle:

?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
<plist version="1.0">
<dict>
    <!-- Activates the sandbox, required. -->
    <key>com.apple.security.app-sandbox</key>
    <true/>
    <!-- Allow your application read/write access to the user downloads folder -->
    <key>com.apple.security.files.downloads.read-write</key>
    <true/>
    <!-- Allow your application read/write access to the file selected by the user. -->
    <key>com.apple.security.files.user-selected.read-write</key>
    <true/>
    <!-- Allow your application to initiate network requests -->
    <key>com.apple.security.network.client</key>
    <true/>
    <!-- Allow your application to listen for network requests -->
    <key>com.apple.security.network.server</key>
    <true/>
    <!-- Allow your application to use the printer -->
    <key>com.apple.security.print</key>
    <true/>
</dict>
</plist>

It's not very difficult to understand what's going on in the above example, especially with the comments, but com.apple.security.files.user-selected.read-write deserves a more detailed explanation. The user selected file corresponds to the selected file or directory in a file dialog (either open or save). Note that on OSX this means that you're obliged to use the FileDialog instead of the JFileChooser, this is because the former uses the native NSOpenDialog and NSSaveDialog which are compatible with the sandbox.

Now let's see how to sign your bundle (that you have generated using AppBundler):

I assume that you have downloaded and installed the two keys required for signing the bundle from the Apple developer website. Of course you need to be registered as a Mac developer.

First, you have to sign the application bundle, which you do by executing the following command from a terminal:

codesign -v -f -s SIGNING_ID_APP --entitlements YourApp.entitlements YourApp.app

SIGNING_ID_APP is the name of your key which is named something like 3rt Party Mac Developer Application: your name here

Second, you need to sign every jar and dylib (the jvm obiously have plenty) contained in the bundle:

find YourApp.app/Contents/ -type f \( -name "*.jar" -or -name "*.dylib" \) -exec codesign --verbose -f -s SIGNING_ID_APP --entitlements YourApp.entitlements {} \;

It doesn't harm to verify that the signing process went well:

echo "Verify all libraries have been signed..."
find YourApp.app/Contents/ -type f \( -name "*.jar" -or -name "*.dylib" \) -exec codesign --verbose --verify {} \;

Finally you can create a .pkg ready for submission:

productbuild --component YourApp.app /Applications --sign SIGNING_ID_INSTALLER YourApp.pkg

SIGNING_ID_APP is the name of your installer key which is named something like 3rt Party Mac Developer Installer: your name here

Congratulations, your app is ready for submission!

Known problems

Plenty unfortunately, however I fixed most of the ones I've encountered and included the fixes in our forks. The easiest way to get them is to clone our jdk and AppBundler stable forks.

Here's a list of most of the problems I've encountered and their solution, where available:

When I say that an issue has been fixed it means it's in the official Oracle JDK8 or JDK7u repo.

Launch failure (LSOpenURLsWithRole() failed with error...)

UPDATE: this has been fixed.

Mailing list Discussion.

This is not a bug in the AppBundler but it is related to this bug ~~which should be fixed soon~~.

libfreetype absolute reference causes applications inside a sandbox to crash

Mailing list discussion

All Swing/AWT apps won't even open because they rely on the freetype lib installed on the user system but the sandbox will immediately block your application the access to it and your application will crash.

The proper solution would be to fix the OpenJDK makefiles otherwise a quick workaround is to use two nifty tools called otool and install_name_tools:

First, find out where libfontmanager.dylib,located in the lib directory of your VM, expects to find freetype.

Open a shell and type:

    $ otool -L libfontmanager.dylib

Take note of the libfreetype path, likely /usr/X11/lib/libfreetype.6.dylib unless you have built freetype yourself or installed it somewhere else (maybe using MacPorts or a similar tool). Now type:

    $ sudo install_name_tool -change /usr/X11/lib/libfreetype.6.dylib @rpath/libfreetype.dylib libfontmanager.dylib

Type again otool -L libfontmanager.dylib and you should see the change. We have changed the path to where libfontmanager look for libfreetype to a relative one. So now you have to copy the system or your custom-build freetype library inside the JVM in the same location as libfontmanager.dylib.
Relaunch your application, you should at least be able to see it now :)

segmentedTextured buttons not working

Mailing list Discussion.

I didn't find a solution for this, but it seems that increasing the margin around the button make the problem go away. I couldn't afford to have too much empty space though so I just replaced the decorations with an image.

OpenJDK Universal build not available

Patch by Henry Gomez.

AppBundler universal binary

There's no point to have a 32/64 OpenJDK if your JavaAppLauncher only supports a 64bit architecture. Use this AppBundler instead.

Dock Icon defaults to Generic Java Application

UPDATE: I believe this has been fixed too but I haven't found the reference yet.

Dock Icon defaults to Generic Java Application

The (temporary) solution here is to specify both CFBundleIconFile and -Xdock:icon=Contents/Resources/your-icon in your Info.plist file. This could be done easily with AppBundler, I've written an example above.

apple.awt.fileDialogForDirectories property not working.

UPDATE: this patch has been accepted, it's already in OpenJDK 8 and ~~will be soon~~ in OpenJDK 7u

Bug report, discussion and a standalone patch.

JFileChooser broken in the sandbox.

Bug report. Discussion. There is no workaround for this, except to switch to a FileDialog .The bug has been incorrectly closed as duplicate of another issue with Runtime.exec().

Crash when opening an application from a registered file

UPDATE: my patch has been integrated into OpenJDK 8 and backported to OpenJDK 7u.

Bug report and Discussion.

If a user double click on a file to open your application the application will crash before even showing its main window.

awt.brush.metalLook is not observed.

UPDATE: this has been fixed, see webrev.

~~Planned to be fixed in jdk7u6~~.

Sandbox Violation on Runtime Exec

Bug description and ML discussion. Because of the sandbox an application can't create a child process; there are a few workarounds but they all involve writing native code. The real solution is to use posix_spawn.

CA certificates not included in OpenJDK

If your application uses SSL, you may want to look at this discussion (workaround included).

Private API usage

UPDATE: fixed, this will soon be in JDK7u (hopefully 7u8/10). I updated my repo with the (corrected) patch I proposed.

Our first app submission failed because some awt files use some functions defined in some Apple private frameworks. I replaced the functions I've found with public API and added a patch to the bug report.

Bug report and ml discussion.

ActionListener called twice for JMenuItem using ScreenMenuBar

UPDATE: I fixed this in my repo and submitted a patch to Oracle, it will be included in JDK7u8/10.

This causes the ActionListener to be called twice when executing an Action that is bound to a JMenuItem if the property apple.laf.useScreenMenuBar is set to true.

Bug report and patch discussion.

Main menu shortcuts not displayed when using ScreenMenuBar

UPDATE: this has been fixed in JDK8 so soon will be avalaible only in JDK7.

Bug report and discussion.

I fixed this bug in my fork ~~but unfortunately the fix now causes bug #7160951 to manifest again~~. However I will revert my change once the JDK8 fix will be backported to JDK7u

It took me ages to make this post and I try as much as possible to keep it updated, if there are any errors or if you have any questions, leave a comment here or on Twitter or send me an e-mail.

Marco

Forking the OpenJDK for fun and profit.

marco.dinacci@intransitione.com (Marco Dinacci) — Sat, 30 Jun 2012 00:00:00 +0100

Update: 04 November 2012 - The OpenJDK macosx-port has vastly improved in the last few months and all my patches have been accepted so I no longer recommend this fork.

At my current job I'm tasked with bringing a Java Swing application to the App Store.

The way to do it, on OSX, is to bundle a Java distribution within the application as by Apple's terms it is no longer possible to distribute applications that rely on technologies optionally installed. As such, I started looking how to bundle the OpenJDK inside an OSX application.

It turns out AppBundler does exactly that. So, task finished ? no, not quite...In the process of testing and improving our application I discovered several bugs in the OpenJDK and after a few patches we finally decided to maintain our own branch.

Our OpenJDK fork currently provides:

Support for the property apple.awt.fileDialogForDirectories.
Support for opening applications by double-clicking on a file. Current OpenJDK behaviour is a hard crash...
Fix a problem where the native save dialog doesn't save the extension of the file saved.
Fix the JMenuItem not showing accelerators keys.
Add universal binary (i386/x86_64) support.
Shows the actual symbol keys for most of the symbols in the JMenuItem accelerators. Eg. shows Ctrl+[ instead of Ctrl+Open Bracket.

I hear you say, why forking instead of contributing back the changes ?

I did contribute one patch, but my other contributions weren't accepted for several reasons. apple.awt.fileDialogForDirectories is not going to be supported, so there's no interest in adding it. The univesal build change touches the whole OpenJDK not only the jdk repository, it's been contributed originally by somebody else and it is still currently been discussed. Regarding the JMenuItem fix, the real cause of the problem is still unknown, although my change still fixes the problem and correct an existing error. In general, it is a bit difficult to match our need to move quickly and our interests with the slow process of contributing an official patch and the long-term plans at Oracle.

In addition to the OpenJDK fork, I also forked AppBundler, reason being, my perceived lack of interest in this project at Oracle (although I did submit patches that have been accepted).

Our AppBundler fork features:

Universal (32/64) JavaAppLauncher (Java launcher stub)

Fixes icon not showing in the dock, linked to bug bug #7159381 in the OpenJDK.

Add classpathref support to the bundleapp task.

Add support for LSArchitecturePriority key.

Allow to specify a custom value for CFBundleVersion instead of the hard-coded "1".

Allow specifying registered file extensions using CFBundleDocumentTypes

If you're a Java developer trying to bring your app to the Mac App Store, go ahead and download our stable jdk and AppBundler forks. For a slightly more experimental code, feel free to clone my jdk and my AppBundler repos on BitBucket.

Marco

OpenJDK on OSX crashes when opening an application with a registered file

marco.dinacci@intransitione.com (Marco Dinacci) — Tue, 12 Jun 2012 00:00:00 +0100

Here's fix for a low priority bug which was unfortunately critical for my application and I suspect for every application that is registered to open a particular set of files.

I'll explain this bug with an example.

Say you download a Java PDF reader and you have the latest Java 7 (or OpenJDK) distribution. You would expect that if you click on a PDF file your PDF reader would open so you could read it, right ?....
No, unfortunately the PDF reader will crash before it could even show its window. You can still open the PDF reader and then open the file, but who would release (and use) an application that crashes so easily ?

The OpenJDK OSX code relies on closures ("blocks" in Objective-C) to defer some events, like the opening of a document when the application is starting, until the application has been fully initialized. Here's an example from QueuingApplicationDelegate.m.

- (void)application:(NSApplication *)theApplication openFiles:(NSArray *)fileNames
{
    [self->queue addObject:^(){
        [realDelegate application:theApplication openFiles:fileNames];
    }];
}

If you already know how blocks are allocated you should have already spotted the error. The block literal is stack allocated, as such the retain that is performed by the container has absolutely no effect. When the function exits, the stack will be deleted and the block with it, leaving you with a dangling pointer inside the array.

The solution is to call copy on the block which will move the memory allocated on the stack to the heap

- (void)application:(NSApplication *)theApplication openFiles:(NSArray *)fileNames
{
    [self->queue addObject:[^(){
        [realDelegate application:theApplication openFiles:fileNames];
    } copy]];
}

Once the block has been used it also has to be released:

NSUInteger i;
NSUInteger count = [self->queue count];
for (i = 0; i < count; i++) {
    void (^event)() = (void (^)())[self->queue objectAtIndex: i];
    event();
    [event release];
}

Download the patch.

View/Hide patch

diff -r 96dd3a52e76c src/macosx/native/sun/osxapp/NSApplicationAWT.m
--- a/src/macosx/native/sun/osxapp/NSApplicationAWT.m   Thu May 31 16:35:25 2012 +0100
+++ b/src/macosx/native/sun/osxapp/NSApplicationAWT.m   Mon Jun 11 16:49:04 2012 +0100
@@ -334,17 +334,17 @@ AWT_ASSERT_APPKIT_THREAD;
 }
 
 @end
 
 
 void OSXAPP_SetApplicationDelegate(id <NSApplicationDelegate> delegate)
 {
 AWT_ASSERT_APPKIT_THREAD;
-    applicationDelegate = delegate;
+    applicationDelegate = [delegate retain];
 
     if (NSApp != nil) {
         [NSApp setDelegate: applicationDelegate];
 
         if (applicationDelegate && qad) {
             [qad processQueuedEventsWithTargetDelegate: applicationDelegate];
             qad = nil;
         }
diff -r 96dd3a52e76c src/macosx/native/sun/osxapp/QueuingApplicationDelegate.m
--- a/src/macosx/native/sun/osxapp/QueuingApplicationDelegate.m Thu May 31 16:35:25 2012 +0100
+++ b/src/macosx/native/sun/osxapp/QueuingApplicationDelegate.m Mon Jun 11 16:49:04 2012 +0100
@@ -104,110 +104,110 @@ static id <NSApplicationDelegate> realDe
     self->queue = nil;
 
     [super dealloc];
 }
 
 
 - (void)_handleOpenURLEvent:(NSAppleEventDescriptor *)openURLEvent withReplyEvent:(NSAppleEventDescriptor *)replyEvent
 {
-    [self->queue addObject:^(){
+    [self->queue addObject:[^(){
         [realDelegate _handleOpenURLEvent:openURLEvent withReplyEvent:replyEvent];
-    }];
+    } copy]];
 }
 
 - (void)application:(NSApplication *)theApplication openFiles:(NSArray *)fileNames
 {
-    [self->queue addObject:^(){
+    [self->queue addObject:[^(){
         [realDelegate application:theApplication openFiles:fileNames];
-    }];
+    } copy]];
 }
 
 - (NSApplicationPrintReply)application:(NSApplication *)application printFiles:(NSArray *)fileNames withSettings:(NSDictionary *)printSettings showPrintPanels:(BOOL)showPrintPanels
 {
     if (!fHandlesDocumentTypes) {
         return NSPrintingCancelled;
     }
 
-    [self->queue addObject:^(){
+    [self->queue addObject:[^(){
         [realDelegate application:application printFiles:fileNames withSettings:printSettings showPrintPanels:showPrintPanels];
-    }];
+    } copy]];
 
     // well, a bit premature, but what else can we do?..
     return NSPrintingSuccess;
 }
 
 - (void)_willFinishLaunching
 {
-    QueuingApplicationDelegate * q = self;
-    [self->queue addObject:^(){
+    [self->queue addObject:[^(){
         [[realDelegate class] _willFinishLaunching];
-    }];
+    } copy]];
 }
 
 - (BOOL)applicationShouldHandleReopen:(NSApplication *)theApplication hasVisibleWindows:(BOOL)flag
 {
-    [self->queue addObject:^(){
+    [self->queue addObject:[^(){
         [realDelegate applicationShouldHandleReopen:theApplication hasVisibleWindows:flag];
-    }];
+    } copy]];
     return YES;
 }
 
 - (NSApplicationTerminateReply)applicationShouldTerminate:(NSApplication *)app
 {
-    [self->queue addObject:^(){
+    [self->queue addObject:[^(){
         [realDelegate applicationShouldTerminate:app];
-    }];
+    } copy]];
     return NSTerminateLater;
 }
 
 - (void)_systemWillPowerOff
 {
-    [self->queue addObject:^(){
+    [self->queue addObject:[^(){
         [[realDelegate class] _systemWillPowerOff];
-    }];
+    } copy]];
 }
 
 - (void)_appDidActivate
 {
-    [self->queue addObject:^(){
+    [self->queue addObject:[^(){
         [[realDelegate class] _appDidActivate];
-    }];
+    } copy]];
 }
 
 - (void)_appDidDeactivate
 {
-    [self->queue addObject:^(){
+    [self->queue addObject:[^(){
         [[realDelegate class] _appDidDeactivate];
-    }];
+    } copy]];
 }
 
 - (void)_appDidHide
 {
-    [self->queue addObject:^(){
+    [self->queue addObject:[^(){
         [[realDelegate class] _appDidHide];
-    }];
+    } copy]];
 }
 
 - (void)_appDidUnhide
 {
-    [self->queue addObject:^(){
+    [self->queue addObject:[^(){
         [[realDelegate class] _appDidUnhide];
-    }];
+    } copy]];
 }
 
 - (void)processQueuedEventsWithTargetDelegate:(id <NSApplicationDelegate>)delegate
 {
+    realDelegate = [delegate retain];
+
     NSUInteger i;
     NSUInteger count = [self->queue count];
 
-    realDelegate = delegate;
-
     for (i = 0; i < count; i++) {
         void (^event)() = (void (^)())[self->queue objectAtIndex: i];
         event();
+        [event release];
     }
 
     [self->queue removeAllObjects];
 }
 
 @end

Marco

Fix for the OpenJDK FileDialog on OSX

marco.dinacci@intransitione.com (Marco Dinacci) — Wed, 30 May 2012 00:00:00 +0100

The OpenJDK FileDialog on OSX does not allow to select directories even if the apple.awt.fileDialogForDirectories property is set to true.

A bug has been reported a bit more than a month ago but it seem there wasn't any intention to fix it as it's been closed as a duplicate of a RFE aimed at improving the FileDialog functionalities. This is all nice and good except for the fact that the RFE is almost two years and a half old so the chances to see 7161437 fixed anytime soon are pretty slim.

For those who can't wait until 7u8 or more to have this basic functionality in, here's a patch that solves the issue.

> See diff and test case

diff -r 6432464ffec2 src/macosx/classes/sun/lwawt/macosx/CFileDialog.java
--- a/src/macosx/classes/sun/lwawt/macosx/CFileDialog.java  Sat May 26 08:01:40 2012 -0700
+++ b/src/macosx/classes/sun/lwawt/macosx/CFileDialog.java  Tue May 29 17:56:51 2012 +0100
@@ -49,20 +49,28 @@ class CFileDialog implements FileDialogP
 
                 navigateApps = true;
 
                 String title = target.getTitle();
                 if (title == null) {
                     title = " ";
                 }
 
+                String chooseDirectoriesValue = System.getProperty("apple.awt.fileDialogForDirectories");
+                boolean chooseDirectories = false;
+                if(chooseDirectoriesValue != null && 
+                        chooseDirectoriesValue.equalsIgnoreCase("true")) {
+                    chooseDirectories = true;
+                }
+
                 String[] userFileNames = nativeRunFileDialog(title,
                         dialogMode,
                         target.isMultipleMode(),
                         navigateApps,
+                        chooseDirectories,
                         target.getFilenameFilter() != null,
                         target.getDirectory(),
                         target.getFile());
 
                 String directory = null;
                 String file = null;
                 File[] files = null;
 
@@ -137,18 +145,18 @@ class CFileDialog implements FileDialogP
             File directoryObj = new File(fileObj.getParent());
             String nameOnly = fileObj.getName();
             ret = ff.accept(directoryObj, nameOnly);
         }
         return ret;
     }
 
     private native String[] nativeRunFileDialog(String title, int mode,
-            boolean multipleMode, boolean shouldNavigateApps, boolean hasFilenameFilter,
-            String directory, String file);
+            boolean multipleMode, boolean shouldNavigateApps, boolean canChooseDirectories, 
+            boolean hasFilenameFilter, String directory, String file);
 
     @Override
     public void setDirectory(String dir) {
     }
 
     @Override
     public void setFile(String file) {
     }
diff -r 6432464ffec2 src/macosx/native/sun/awt/CFileDialog.h
--- a/src/macosx/native/sun/awt/CFileDialog.h   Sat May 26 08:01:40 2012 -0700
+++ b/src/macosx/native/sun/awt/CFileDialog.h   Tue May 29 17:56:51 2012 +0100
@@ -47,16 +47,19 @@
     jint fMode;
 
     // Indicates whether the user can select multiple files
     BOOL fMultipleMode;
 
     // Should we navigate into apps?
     BOOL fNavigateApps;
 
+    // Can the dialog choose directories ?
+    BOOL fChooseDirectories;
+
     // Contains the absolute paths of the selected files as URLs
     NSArray *fURLs;
 }
 
 // Allocator
 - (id) initWithFilter:(jboolean)inHasFilter
            fileDialog:(jobject)inDialog
                 title:(NSString *)inTitle
diff -r 6432464ffec2 src/macosx/native/sun/awt/CFileDialog.m
--- a/src/macosx/native/sun/awt/CFileDialog.m   Sat May 26 08:01:40 2012 -0700
+++ b/src/macosx/native/sun/awt/CFileDialog.m   Tue May 29 17:56:51 2012 +0100
@@ -38,30 +38,32 @@
 - (id)initWithFilter:(jboolean)inHasFilter
           fileDialog:(jobject)inDialog
                title:(NSString *)inTitle
            directory:(NSString *)inPath
                 file:(NSString *)inFile
                 mode:(jint)inMode
         multipleMode:(BOOL)inMultipleMode
       shouldNavigate:(BOOL)inNavigateApps
+canChooseDirectories:(BOOL)inChooseDirectories
              withEnv:(JNIEnv*)env;
 {
     if (self == [super init]) {
         fHasFileFilter = inHasFilter;
         fFileDialog = JNFNewGlobalRef(env, inDialog);
         fDirectory = inPath;
         [fDirectory retain];
         fFile = inFile;
         [fFile retain];
         fTitle = inTitle;
         [fTitle retain];
         fMode = inMode;
         fMultipleMode = inMultipleMode;
         fNavigateApps = inNavigateApps;
+        fChooseDirectories = inChooseDirectories;
         fPanelResult = NSCancelButton;
     }
 
     return self;
 }
 
 -(void) disposer {
     if (fFileDialog != NULL) {
@@ -104,17 +106,17 @@
         if (fNavigateApps) {
             [thePanel setTreatsFilePackagesAsDirectories:YES];
         }
 
         if (fMode == java_awt_FileDialog_LOAD) {
             NSOpenPanel *openPanel = (NSOpenPanel *)thePanel;
             [openPanel setAllowsMultipleSelection:fMultipleMode];
             [openPanel setCanChooseFiles:YES];
-            [openPanel setCanChooseDirectories:NO];
+            [openPanel setCanChooseDirectories:fChooseDirectories];
             [openPanel setCanCreateDirectories:YES];
         }
 
         [thePanel setDelegate:self];
         fPanelResult = [thePanel runModalForDirectory:fDirectory file:fFile];
         [thePanel setDelegate:nil];
 
         if ([self userClickedOK]) {
@@ -177,17 +179,17 @@
  * Class:     sun_lwawt_macosx_CFileDialog
  * Method:    nativeRunFileDialog
  * Signature: (Ljava/lang/String;ILjava/io/FilenameFilter;
  *             Ljava/lang/String;Ljava/lang/String;)[Ljava/lang/String;
  */
 JNIEXPORT jobjectArray JNICALL
 Java_sun_lwawt_macosx_CFileDialog_nativeRunFileDialog
 (JNIEnv *env, jobject peer, jstring title, jint mode, jboolean multipleMode,
- jboolean navigateApps, jboolean hasFilter, jstring directory, jstring file)
+ jboolean navigateApps, jboolean chooseDirectories, jboolean hasFilter, jstring directory, jstring file)
 {
     jobjectArray returnValue = NULL;
 
 JNF_COCOA_ENTER(env);
     NSString *dialogTitle = JNFJavaToNSString(env, title);
     if ([dialogTitle length] == 0) {
         dialogTitle = @" ";
     }
@@ -195,16 +197,17 @@ JNF_COCOA_ENTER(env);
     CFileDialog *dialogDelegate = [[CFileDialog alloc] initWithFilter:hasFilter
                                                            fileDialog:peer
                                                                 title:dialogTitle
                                                             directory:JNFJavaToNSString(env, directory)
                                                                  file:JNFJavaToNSString(env, file)
                                                                  mode:mode
                                                          multipleMode:multipleMode
                                                        shouldNavigate:navigateApps
+                                                 canChooseDirectories:chooseDirectories
                                                               withEnv:env];
 
     [JNFRunLoop performOnMainThread:@selector(safeSaveOrLoad)
                                  on:dialogDelegate
                          withObject:nil
                       waitUntilDone:YES];
 
     if ([dialogDelegate userClickedOK]) {

import java.awt.FileDialog;
import java.io.File;
import javax.swing.JFrame;

public class FileDialogTest extends JFrame {

    public String selectFolder() {
        FileDialog fd = new FileDialog(this, "Select Folder Test", FileDialog.LOAD); 
        fd.setDirectory(System.getProperty("user.home")); 
        fd.setLocation(50,50);
        fd.setVisible(true); 

        return fd.getFile();
    }

    public static void main(String[] args) {
        System.setProperty("apple.awt.fileDialogForDirectories", "true");

        FileDialogTest fdt = new FileDialogTest();
        String selectedFolder = fdt.selectFolder();

        System.out.println("The selected file was: " + selectedFolder);

        System.setProperty("apple.awt.fileDialogForDirectories", "false");
        System.exit(0);
    }

}

Marco

Group Bill, my first app in the iTunes Store.

marco.dinacci@intransitione.com (Marco Dinacci) — Thu, 01 Mar 2012 00:00:00 +0000

Group Bill is available to download from the iTunes Store!

After having worked for four months at a reasonably big native iOS application and not seeing it yet on the store because of a client I'm not too happy with, I finally had my first one being accepted today.

I read nightmare stories about app submissions but I have to say it's been quite easy in my opinion and the review process last less than a week (actually the review itself was done in one day), so I'm very happy with it.

Group Bill is a simple app to split the restaurant bill, there are already millions available that perform the same task but mine does it with a twist as it uses a trivial but effective formula devised by a british mathematician called Matt Parker.

He explains the details much better than I ever could in this video on Youtube:

I'll update it to French and Spanish and then I'll be ready to start a bigger project. If you want to contribute a translation in your language please let me know (use the share link on the bottom or the support page) I will include you in the credits.

Marco

A look at the IT job market

marco.dinacci@intransitione.com (Marco Dinacci) — Thu, 10 Nov 2011 00:00:00 +0000

Update! the ruby source code is on github.

Introduction

Few weeks ago I bought an amazing book whose first chapter is an introduction to the Ruby language. I've never really been interested in Ruby before as I'm perfectly happy with Python but I think there's always something to learn when studying a new language. With this in mind, and considering that I'm looking for a job, I thought to learn Ruby by writing a tool to analyze the technical skills most required by the market today. I selected the website that in my opinion offers the more interesting jobs, careers.stackoverflow.com, did a bit of scraping and made some visualisations to help me understand the data collected (which, BTW, is limited only to the last 5 weeks).

The data

First, let's see what are the technologies required the most:

Although the data is limited (again, only 5 weeks), we can make some interesting comparisons:

There's more demand of Java developers than C# developers.
More jobs require Android knowledge rather than iOS, maybe this is also why Java > C#.
MySQL beats any other database, including NoSQL ones.
Javascript is the most frequent tag, immediately followed by Java.
Ruby is less requested than I thought, but maybe requests are distributed among the tags ruby and ruby-on-rails, I don't know.

This knowledge is useful but too superficial. Most of the companies are looking for candidates with a knowledge of a set of technologies strongly related between them so I made a graph to see which groups of technologies are the most requested. This visualization takes the 20 most occurring tags and analyze which other tags appear with. I shortened some of the words to make the visualization more readable, eg. ror stands for Ruby on Rails while js represents Javascript. Each circle represents a tag, the number in each circle represent the number of times that tag appears with the tag label on the left. (If you don't like the colors reload the page).

Now I think it's possible to appreciate more which set of technologies are currently requested and which one a developer should focus in order to increase his/her changes of landing a job of the ones proposed on careers.stackoverflow.com.

Trends ?

For frontend web developers that know HTML and CSS it is also necessary to know Javascript and a big advantage is to know JQuery which demonstrates to be the leading Javascript framework. If you're a frontend developer and want to learn a server-side scripting language, PHP (instead of Ruby or Python) seems to be the most requested.

Python is very often associated with Java and C++ but also with Django and Javascript confirming the language versatility. It seems though that companies prefer to use it mostly outside the web, probably as a "second language requirement" for Java or C++ developers. If you know Python already and want to apply it on the web, study Django.

Nothing surprising in the Microsoft world, .net, C#, Asp.net and SQL server are the technologies required in companies that relies on Microsoft technologies.

It's interesting to see C++ and Java so often together. Companies are probably looking for candidates with a knowledge of either one or the other but I think that while a C++ developer can pick up Java pretty fast I've some doubts about the opposite being so easy.

Knowledge of both Android an iOS seems to be well appreciated by companies that probably prefer to have only one developer working on the same application for both platforms. Windows mobile and blackberry skills are not particularly requested.

The Spring framework is often demanded when working with Java EE. Interesting to note that Java skills are often requested along with a scripting language like Python, Perl or Groovy. They aren't probably necessary but a candidate with a knowledge of at least one scripting language may have the head over a candidate that does not.

Finally, since I don't live in a big city with a lot of job opportunities, I'm particularly interested in telecommuting jobs:

Guess I should relocate... :-|

The Code

The code to scrape the data and to visualize it is on github.

I think making this kind of tools is a great exercise to learn a new programming language and I really enjoyed doing it in Ruby and Javascript.

Movie Suggestions Beta

marco.dinacci@intransitione.com (Marco Dinacci) — Thu, 14 Jul 2011 00:00:00 +0100

UPDATE: here is the source code of the RottenTomatoes API wrapper. It's not a full wrapper but it covers most of the API.

I just released my first Android application on the Android market. Technically it's the second but it's the first one on the market.

Movie Suggestions is a simple application that uses a couple of web services to find a list of movies that you may like. Once you select a movie you can then read some comments from reviewers and see the trailer on Youtube, where available.

Graphics and functionalities are pretty basic at the moment but hopefully the only one thing it does it does it well. In any case, I made it to learn the Android framework and it's been a great exercise.

I used the Rottentomatoes API to search and retrieve movie informations and another web service to look for the trailer on Youtube. This last one actually saved me a lot of work.

The application is free and will always be but I'm not yet releasing the source code as it uses some libraries I've written that I don't want to release. I will definitely release the wrapper around the Rottentomatoes API as soon as I'll find the time.

Advanced ListView: populating a list with images and text

marco.dinacci@intransitione.com (Marco Dinacci) — Tue, 10 May 2011 00:00:00 +0100

The Android documentation says that the ListView widget is easy to use. It is true if you just want to display a simple list of strings but as soon as you want to customize your list things become more complicated.

The following is a tutorial on how to write a ListView that displays a static list of images and strings, similar to the settings list on your phone.

For the impatients, here's the Eclipse project code and a picture of the final result:

Let's start with the Activity code. First of all, we extends from ListActivity instead of Activity so we can easily supply our custom adapter:

public class AdvancedListViewActivity extends ListActivity {

    @Override
    public void onCreate(Bundle savedInstanceState) {
        super.onCreate(savedInstanceState);
        setContentView(R.layout.main);
        
        Context ctx = getApplicationContext();
    Resources res = ctx.getResources();

    String[] options = res.getStringArray(R.array.country_names);
    TypedArray icons = res.obtainTypedArray(R.array.country_icons);
        
    setListAdapter(new ImageAndTextAdapter(ctx, R.layout.main_list_item, options, icons));
    }
}

In the onCreate we also create an array of strings, which contains the country names, and a TypedArray, which will contain our Drawable flags.

The arrays are created from an XML file, here's the content of the countries.xml file.

<?xml version="1.0" encoding="utf-8"?>
<resources>
    <string-array name="country_names">
        <item>Bhutan</item>
        <item>Colombia</item>
        <item>Italy</item>
        <item>Jamaica</item>
        <item>Kazakhstan</item>
        <item>Kenya</item>
    </string-array>
    <array name="country_icons">
        <item>@drawable/bhutan</item>
        <item>@drawable/colombia</item>
        <item>@drawable/italy</item>
        <item>@drawable/jamaica</item>
        <item>@drawable/kazakhstan</item>
        <item>@drawable/kenya</item>
    </array>
</resources>

Now we're ready to create the adapter. The official documentation for Adapter says:
An Adapter object acts as a bridge between an AdapterView and the underlying data for that view. The Adapter provides access to the data items. The Adapter is also responsible for making a View for each item in the data set.

There are already various subclasses of Adapter, we're going to extend on ArrayAdapter which is a concrete BaseAdapter that is backed by an array of arbitrary objects.

public class ImageAndTextAdapter extends ArrayAdapter<String> {

    private LayoutInflater mInflater;
    
    private String[] mStrings;
    private TypedArray mIcons;
    
    private int mViewResourceId;
    
    public ImageAndTextAdapter(Context ctx, int viewResourceId,
            String[] strings, TypedArray icons) {
        super(ctx, viewResourceId, strings);
        
        mInflater = (LayoutInflater)ctx.getSystemService(
                Context.LAYOUT_INFLATER_SERVICE);
        
        mStrings = strings;
        mIcons = icons;
        
        mViewResourceId = viewResourceId;
    }

    @Override
    public int getCount() {
        return mStrings.length;
    }

    @Override
    public String getItem(int position) {
        return mStrings[position];
    }

    @Override
    public long getItemId(int position) {
        return position;
    }

    @Override
    public View getView(int position, View convertView, ViewGroup parent) {
        convertView = mInflater.inflate(mViewResourceId, null);
        
        ImageView iv = (ImageView)convertView.findViewById(R.id.option_icon);
        iv.setImageDrawable(mIcons.getDrawable(position));

        TextView tv = (TextView)convertView.findViewById(R.id.option_text);
        tv.setText(mStrings[position]);
        
        return convertView;
    }
}

The constructor accepts a Context, the id of the layout that will be used for every row (more on this soon), an array of strings (the country names) and a TypedArray (our flags).

The getView method is where we build a row for the list. We first use a LayoutInflater to create a View from XML, then we retrieve the country flag as a Drawable and the country name as a String and we use them to populate the ImageView and TextView that we've declared in the layout.

The layout for the list rows is the following:

<?xml version="1.0" encoding="utf-8"?>
<LinearLayout xmlns:android="http://schemas.android.com/apk/res/android">
    <ImageView
    android:id="@+id/option_icon"
    android:layout_width="48dp"
    android:layout_height="fill_parent"/>
    <TextView
       android:id="@+id/option_text"
        android:layout_width="fill_parent"
        android:layout_height="fill_parent"
        android:padding="10dp"
        android:textSize="16dp" >
    </TextView>
</LinearLayout>

As a reference, this is the content of the main layout:

<?xml version="1.0" encoding="utf-8"?>
<LinearLayout xmlns:android="http://schemas.android.com/apk/res/android"
    android:orientation="vertical"
    android:layout_width="fill_parent"
    android:layout_height="fill_parent"
    >
<ListView android:id="@android:id/list" 
    android:layout_height="wrap_content" 
    android:layout_width="fill_parent"
    />
</LinearLayout>

Note that the ListView ID must be exactly

@android:id/list

or you'll get a RuntimeException.

That's it, you can download the Eclipse project code or post a comment if you have any question.

How to detect motion on an Android device

marco.dinacci@intransitione.com (Marco Dinacci) — Wed, 04 May 2011 00:00:00 +0100

For learning purposes I've written PicInMotion, a small Android application that automatically takes a picture when it detects motion in the camera preview.

This article explains how the simple motion detection mechanism works.

Update: source code is available at github, be sure to clone the standalone repository.

Strategy

Dead simple. I store a frame and compare it to the next, if the differences in the two frames are above a certain threshold then motion is detected. Common image detection techniques are far more complicated than what I'm using here and image processing operations (like erode, dilate and some other form of noise reduction) are normally applied to frames in order to reduce digital noise.

Since I wanted to focus on learning Android more than some new image processing techniques, I kept the image operations to the absolute minimum.

Acquiring data

First, we need to acquire the raw image data from the camera. To do it, we implement the Camera.PreviewCallback and override the onPreviewFrame method. This method is called every time a new frame is ready.

Here's the code:

@Override
public void onPreviewFrame(byte[] data, Camera camera) {
    if (mMotionDetection.detect(data)) {
        // The delay is necessary to avoid taking a picture while in the 
        // middle of taking another. This problem can causes some phones
        // to reboot.
        // I set PICTURE_DELAY to 3000 ms, your degree may vary.
        // mReferenceTime is simply 0 at the beginning.
        long now = System.currentTimeMillis();
    if (now > mReferenceTime + PICTURE_DELAY) {              
            mReferenceTime = now + PICTURE_DELAY;
        camera.takePicture(null, null, this);
    } else {
        Log.i(TAG, "Not taking picture because not enough time has "
        + "passed since the creation of the Surface");
    }
    }
}

data is a byte array containing the raw data coming from the camera hardware. Its format may change depending on the device and if the hardware supports multiple formats you can change it using:

Camera.Parameters parameters = mCamera.getParameters();
parameters.setPreviewFormat(...);
mCamera.setParameters(parameters);

In this post I'll assume the image format is PixelFormat.YCbCr_420_SP (now deprecated and to be replaced withImageFormat.NV21.

Detecting movement

The advantage of using ImageFormat.NV21 is that being a YUV based, not RGB based, format, the first height * width bytes contains the luminance data. You can think of the luminance data as a grayscale version of your image.

We can perform the motion detection on the luminance data only, it will be much faster than comparing two RGB images.

public boolean detect(byte[] data) {        
    // Create the "background" picture, the one that will be used
    // to check the next frame against.
    if(mBackground == null) {
        mBackground = AndroidImageFactory.createImage(data,   
            mSize.value, mPixelFormat.value);
    Log.i(TAG, "Creating background image");
    return false;
    }
        
    boolean motionDetected = false;
        
    mAndroidImage=AndroidImageFactory.createImage(data,
        mSize.value, mPixelFormat.value);
        
    motionDetected = mAndroidImage.isDifferent(mBackground, 
        mPixelThreshold.value, mThreshold.value);
   
    // Replace the current image with the background.
    // This is an oversimplification, you would normally blend the two
    // images to create a new one.
    mBackground = mAndroidImage;
    
    return motionDetected;
}

AndroidImageFactory is a simple factory class I've created that create a specific class depending on the image format. An extract:

public static AndroidImage createImage(byte[] data, Size<Integer,
    Integer> size, int format) {
    
    AndroidImage im = null;
        
    String imgType = "";
        
    switch (format) {
        case IMAGE_FORMAT_NV21:
        im = new AndroidImage_NV21(data,size);
        imgType = "NV_21";
        break;
   
    ...

AndroidImage and AndroidImage_NV21 are also classes I've created and that are part of my commons library. I'm not releasing yet the source code for these classes but you don't need it anyway.

Comparing the images by luminance values

This is the bulk of the article, here the isDifferent method in my AndroidImage_NV21 class.

@Override
public boolean isDifferent(AndroidImage other, int pixel_threshold, 
    int threshold) {
        
        if(!assertImage(other))
            return false;
        
    byte[] otherData = other.get();
    int totDifferentPixels = 0;
        
    int size= mHeight * mWidth;
    for (int i = 0, ij=0; i < mHeight; i++) {
        for (int j = 0; j < mWidth; j++,ij++) {
            int pix = (0xff & ((int) mData[ij])) - 16;
        int otherPix = (0xff & ((int) otherData[ij])) - 16;
                
        if (pix < 0) pix = 0;
        if (pix > 255) pix = 255;
        if (otherPix < 0) otherPix = 0;
        if (otherPix > 255) otherPix = 255;

        if(Math.abs(pix - otherPix) >= pixel_threshold)
            totDifferentPixels++;
        }
    }
        
    if(totDifferentPixels == 0) totDifferentPixels = 1;
    Log.d(TAG, "Number of different pixels: " + totDifferentPixels + "> " 
            + (100 / ( size / totDifferentPixels) ) + "%");
        
    return totDifferentPixels > threshold;
}

I use two different thresholds. The first acts on the pixel values and could be defined as:

reference_pixel - threshold < new_pixel < reference_pixel + threshold

So the new_pixel must be bigger than the reference_pixel minus the threshold but smaller than new_pixel plus threshold.

If the pixel is different I increment totDifferentPixels. At the end I check how many different pixels I have, if the value is greater than the second threshold, then the image is different.

As a reference, I use 25 as value for pixel value threshold (~10% difference) and 3% (9216 for a 640x480 image) as total pixel threshold.

Questions ? write a comment.

Adding control to the OpenGL spinning cube

marco.dinacci@intransitione.com (Marco Dinacci) — Mon, 02 May 2011 00:00:00 +0100

This demo extends on the Create a spinning cube OpenGL to show how to control the cube using a D-pad. We're going to increment the speed rotation along the X and Y axes using the directional keys of the D-pad.

The biggest change is that we now have our custom class that extends the SurfaceView. We do this so we can override the onKeyUp method and be notified when the user uses the D-pad.

The onCreate of our Activity looks like this:

public class SpinningCubeActivity2 extends Activity {
    @Override
    public void onCreate(Bundle savedInstanceState) {
        super.onCreate(savedInstanceState);
        
        // go fullscreen
        requestWindowFeature(Window.FEATURE_NO_TITLE);
        getWindow().setFlags(WindowManager.LayoutParams.FLAG_FULLSCREEN,
                        WindowManager.LayoutParams.FLAG_FULLSCREEN);
        
        // create our custom view
        GLSurfaceView view = new OpenGLSurfaceView(this);
        view.setRenderer((Renderer)view);
        setContentView(view);
    }
}

Our new GLSurfaceView also implements the Renderer interface. The onSurfaceCreated and onSurfaceChanged method are exactly the same as in the previous recipe, most of the changes occur in the onDrawFrame as we introduce four new parameters:

mXrot and mYrot to control the rotation of the cube along the X and Y axis.
mXspeed and mYSpeed to store the speed of the rotation along the X and Y axis.

Each time the user click on a D-pad button we alter the speed of the cube by modifying these parameters. Here's the full code of our new class:

class OpenGLSurfaceView extends GLSurfaceView implements Renderer {

    private Cube mCube;
    private float mXrot;
    private float mYrot;
    private float mXspeed;
    private float mYspeed;
        
    public OpenGLSurfaceView(Context context) {
        super(context);
        
        // give focus to the GLSurfaceView    
        requestFocus();
        setFocusableInTouchMode(true);
                
        mCube = new Cube();
    }

    @Override
    public void onDrawFrame(GL10 gl) {
        gl.glClear(GL10.GL_COLOR_BUFFER_BIT | GL10.GL_DEPTH_BUFFER_BIT);        
        gl.glLoadIdentity();
               
        gl.glTranslatef(0.0f, 0.0f, -10.0f);
                
        gl.glRotatef(mXrot, 1.0f, 0.0f, 0.0f);  
        gl.glRotatef(mYrot, 0.0f, 1.0f, 0.0f);
        
        mCube.draw(gl);
                
        gl.glLoadIdentity();                                    
                
        mXrot += mXspeed;
        mYrot += mYspeed;
    }

    @Override
    public boolean onKeyUp(int keyCode, KeyEvent event) {
        if(keyCode == KeyEvent.KEYCODE_DPAD_LEFT)
            mYspeed -= 0.1f;
        else if(keyCode == KeyEvent.KEYCODE_DPAD_RIGHT)
            mYspeed += 0.1f;
        else if(keyCode == KeyEvent.KEYCODE_DPAD_UP)
            mXspeed -= 0.1f;
        else if(keyCode == KeyEvent.KEYCODE_DPAD_DOWN)
            mXspeed += 0.1f;
             
        return true;
     }

     // unchanged
     @Override
     public void onSurfaceCreated(GL10 gl, EGLConfig config) {
         gl.glClearColor(0.0f, 0.0f, 0.0f, 0.5f); 
                
         gl.glClearDepthf(1.0f);
         gl.glEnable(GL10.GL_DEPTH_TEST);
         gl.glDepthFunc(GL10.GL_LEQUAL);

         gl.glHint(GL10.GL_PERSPECTIVE_CORRECTION_HINT,
                  GL10.GL_NICEST);        
     }

     
     // unchanged
     @Override
     public void onSurfaceChanged(GL10 gl, int width, int height) {
         gl.glViewport(0, 0, width, height);
         gl.glMatrixMode(GL10.GL_PROJECTION);
         gl.glLoadIdentity();
         GLU.gluPerspective(gl, 45.0f, (float)width / (float)height, 0.1f, 100.0f);
         gl.glViewport(0, 0, width, height);

         gl.glMatrixMode(GL10.GL_MODELVIEW);
         gl.glLoadIdentity();
     }
}

The Cube is inherited from the previous recipe. Don't forget to call the requestFocus() and setFocusableInTouchMode(true) in the constructor of the view or else the key events will not be received. The complete code is available.

"touch" a file on Android

marco.dinacci@intransitione.com (Marco Dinacci) — Sat, 30 Apr 2011 00:00:00 +0100

This snippet mimics the behaviour of the unix touch command on the Android platform:

public static void touch(File file) throws IOException {
    if(!file.exists()) {
        File parent = file.getParentFile();
        if(parent != null) 
            if(!parent.exists())
	        if(!parent.mkdirs())
	            throw new IOException("Cannot create parent
                    directories for file: " + file);
		
		file.createNewFile();
    }
		
    boolean success = file.setLastModified(System.currentTimeMillis());
    if (!success)
        throw new IOException("Unable to set the last modification time for " + file);
}

Taking a picture with android.media.Camera

marco.dinacci@intransitione.com (Marco Dinacci) — Sat, 30 Apr 2011 00:00:00 +0100

Sometimes you may want more control over the stages involved when taking a picture or you may want to access and modify the raw image data acquired by the camera. In these cases, using a simple Intent to take a picture is not enough.

We're going to create a new Activity and customize the view to make it full screen inside the onCreate method.

public class TakePictureActivity extends Activity {
    private Preview mCameraView;

    @Override
    public void onCreate(Bundle savedInstanceState) {
        super.onCreate(savedInstanceState);
        
        // Force screen in landscape mode as showing a video in 
        // potrait mode is not easily doable on all devices
        setRequestedOrientation(ActivityInfo.SCREEN_ORIENTATION_LANDSCAPE);
                
        // Hide window title and go fullscreen
        requestWindowFeature(Window.FEATURE_NO_TITLE);
        getWindow().addFlags(WindowManager.LayoutParams.FLAG_FULLSCREEN);

        mCameraView= new Preview(this);
        setContentView(mCameraView);
    }
}

The Preview class is the bulk of the recipe. It handles the Surface where the pixels are drawn and the Camera object.

We define a ClickListener in the constructor so the user can take a picture by just tapping once on the screen. Once we get the notification of the click we take a picture passing as parameters four (all optional) callbacks.

class Preview extends SurfaceView implements SurfaceHolder.Callback, PictureCallback  {

    private SurfaceHolder mHolder;
    private Camera mCamera;
    private RawCallback mRawCallback;

    public Preview(Context context) {
        super(context);

        mHolder = getHolder();
        mHolder.addCallback(this);
        mHolder.setType(SurfaceHolder.SURFACE_TYPE_PUSH_BUFFERS);
        mRawCallback = new RawCallback();
            
        setOnClickListener(new OnClickListener() {
                        
            @Override
            public void onClick(View v) {
                mCamera.takePicture(mRawCallback, mRawCallback, null, 
                                Preview.this);
            }
        });
    }

The Preview class implement the SurfaceHolder.Callbackinterface in order to be notified when underlying surface is created, changed and destroyed. We'll use these callbacks to properly handle the Camera object.

    @Override
    public void surfaceChanged(SurfaceHolder holder, int format, int width,
            int height) {

        Camera.Parameters parameters = mCamera.getParameters();
        parameters.setPreviewSize(width, height);
        mCamera.setParameters(parameters);
                
        mCamera.startPreview();
    }

    @Override
    public void surfaceCreated(SurfaceHolder holder) {
        mCamera = Camera.open();
            
        configure(mCamera);

        try {
            mCamera.setPreviewDisplay(holder);
        } catch (IOException exception) {
            closeCamera();
        }
    }

    @Override
    public void surfaceDestroyed(SurfaceHolder holder) {
        closeCamera();
    }

As soon as the camera is created we call configure in order to set the parameters the camera will use to take a picture. Things like flash mode, effects, picture format, picture size, scene mode and so on. Since not all devices support all kind of features always ask which features are supported before setting them.

    private void configure(Camera camera) {
        Camera.Parameters params = camera.getParameters();

        // Configure image format. RGB_565 is the most common format.
        List<Integer> formats = params.getSupportedPictureFormats();
        if (formats.contains(PixelFormat.RGB_565))
            params.setPictureFormat(PixelFormat.RGB_565);
        else
            params.setPictureFormat(PixelFormat.JPEG);

        // Choose the biggest picture size supported by the hardware
        List<Size> sizes = params.getSupportedPictureSizes();
        Camera.Size size = sizes.get(sizes.size()-1);
        params.setPictureSize(size.width, size.height);

        List<String> flashModes = params.getSupportedFlashModes();
        if (flashModes.size() > 0)
            params.setFlashMode(Camera.Parameters.FLASH_MODE_AUTO);

        // Action mode take pictures of fast moving objects
        List<String> sceneModes = params.getSupportedSceneModes();
        if (sceneModes.contains(Camera.Parameters.SCENE_MODE_ACTION))
            params.setSceneMode(Camera.Parameters.SCENE_MODE_ACTION);
        else
            params.setSceneMode(Camera.Parameters.SCENE_MODE_AUTO);

        // if you choose FOCUS_MODE_AUTO remember to call autoFocus() on
        // the Camera object before taking a picture 
        params.setFocusMode(Camera.Parameters.FOCUS_MODE_FIXED);
        
        camera.setParameters(params);
    }

Attention! Once the surface is destroyed we must close the camera and free its resources:

    private void closeCamera() {
        if (mCamera != null) {
            mCamera.stopPreview();
            mCamera.release();
            mCamera = null;
        }
    }

The Jpeg callback is the last one called, this is where we restart the preview and save the file on disk.

    @Override
    public void onPictureTaken(byte[] jpeg, Camera camera) {
        // now that all the callbacks have been called it is safe to resume the preview
        mCamera.startPreview();
                
        saveFile(jpeg);
    }
}

Finally we implement the ShutterCallback and again PictureCallback to receive the uncompressed raw image data.

class RawCallback implements ShutterCallback, PictureCallback {

    @Override
    public void onShutter() {
        // notify the user, normally with a sound, that the picture has 
        // been taken
    }

    @Override
    public void onPictureTaken(byte[] data, Camera camera) {
        // manipulate uncompressed image data
    }       
}

Done.

Playing an audio file with Android

marco.dinacci@intransitione.com (Marco Dinacci) — Sat, 30 Apr 2011 00:00:00 +0100

Update: The complete source code for this demo is available for download.

Playing an audio file is as easy as setting up a MediaPlayer and a MediaController.

First create a new activity that implements the MediaPlayerControl interface.

 
public class PlayAudioActivity extends Activity implements MediaPlayerControl {
        private MediaController mMediaController;
        private MediaPlayer mMediaPlayer;
        private Handler mHandler = new Handler();

In the onCreate method we create and configure a MediaPlayer and a MediaController. The first is the object that perform the typical operations on an audio file like playing, pausing and seeking. The second is a view containing the buttons that launch the just mentioned operations through our MediaPlayerControl class. Let's see the onCreate code:

    @Override
    public void onCreate(Bundle savedInstanceState) {
        super.onCreate(savedInstanceState);
        setContentView(R.layout.main);

        mMediaPlayer = new MediaPlayer();
        mMediaController = new MediaController(this);
        mMediaController.setMediaPlayer(PlayAudioActivity.this);
        mMediaController.setAnchorView(findViewById(R.id.audioView));
        
        String audioFile = "" ; 
        try {
            mMediaPlayer.setDataSource(audioFile);
            mMediaPlayer.prepare();
        } catch (IOException e) {
            Log.e("PlayAudioDemo", "Could not open file " + audioFile + " for playback.", e);
        }
        
        mMediaPlayer.setOnPreparedListener(new OnPreparedListener() {
            @Override
            public void onPrepared(MediaPlayer mp) {
                    mHandler.post(new Runnable() {
                            public void run() {
                                    mMediaController.show(10000);
                                    mMediaPlayer.start();
                            }
                    });
            }
        });
    }

In addition to configuring our MediaController and MediaPlayer we create an anonymous OnPreparedListener in order to start the player only when the media source is ready for playback.

Remember to cleanup the media player when the Activity is destroyed.

    @Override
    protected void onDestroy() {
        super.onDestroy();
        mMediaPlayer.stop();
        mMediaPlayer.release();
    }

At last we implement the MediaPlayerControl interface. The code is very straightforward:

    @Override
    public boolean canPause() {
        return true;
    }

    @Override
    public boolean canSeekBackward() {
        return false;
    }

    @Override
    public boolean canSeekForward() {
        return false;
    }

    @Override    
    public int getBufferPercentage() {
        int percentage = (mMediaPlayer.getCurrentPosition() * 100) / mMediaPlayer.getDuration();
        
        return percentage;
    }

    @Override
    public int getCurrentPosition() {
        return mMediaPlayer.getCurrentPosition();
    }

    @Override
    public int getDuration() {
        return mMediaPlayer.getDuration();
    }

    @Override
    public boolean isPlaying() {
        return mMediaPlayer.isPlaying();
    }

    @Override
    public void pause() {
        if(mMediaPlayer.isPlaying())
            mMediaPlayer.pause();
    }

    @Override
    public void seekTo(int pos) {
        mMediaPlayer.seekTo(pos);
    }

    @Override
    public void start() {
        mMediaPlayer.start();
    }

    @Override
    public boolean onTouchEvent(MotionEvent event) {
        mMediaController.show();
        
        return false;
    }
}

As a final touch we override the onTouchEvent in order to show the MediaController buttons when the user click on the screen.

Since we create our MediaController programmatically, the layout is very simple:

<?xml version="1.0" encoding="utf-8"?>
<LinearLayout xmlns:android="http://schemas.android.com/apk/res/android"
    android:orientation="vertical"
    android:layout_width="fill_parent"
    android:layout_height="fill_parent"
    android:id="@+id/audioView">
</LinearLayout>

Play a video on Youtube using an Intent

marco.dinacci@intransitione.com (Marco Dinacci) — Sat, 30 Apr 2011 00:00:00 +0100

Warning! For this demo to work the user needs the Youtube application installed on his device.

public void onCreate(Bundle savedInstanceState) {
    super.onCreate(savedInstanceState);
    setContentView(R.layout.main);
    
    String video_path = "http://www.youtube.com/watch?v=opZ69P-0Jbc";
    Uri uri = Uri.parse(video_path);
        
    // With this line the Youtube application, if installed, will launch immediately.
    // Without it you will be prompted with a list of the application to choose.
    uri = Uri.parse("vnd.youtube:"  + uri.getQueryParameter("v"));
        
    Intent intent = new Intent(Intent.ACTION_VIEW, uri);
    startActivity(intent);
}

The example uses a standard Youtube URL. The

uri.getQueryParameter("v")

is used to extract the video ID from the URI itself, in this example the ID is opZ69P-0Jbc

Grabbing a video using MediaRecorder

marco.dinacci@intransitione.com (Marco Dinacci) — Sat, 30 Apr 2011 00:00:00 +0100

Update! The full source code for this demo is available.

The MediaRecorder is normally used to perform audio and/or video recording. The class has a straightforward API but as it's based on a simple state machine the methods must be called in the proper order in order to avoid IllegalStateException from popping up.

Create a new Activity and override the onCreate method with the following:

    @Override
    public void onCreate(Bundle savedInstanceState) {
        super.onCreate(savedInstanceState);
        setContentView(R.layout.media_recorder_recipe);
        
        // we shall take the video in landscape orientation
        setRequestedOrientation(ActivityInfo.SCREEN_ORIENTATION_LANDSCAPE);
        
        mSurfaceView = (SurfaceView) findViewById(R.id.surfaceView);
        mHolder = mSurfaceView.getHolder();
        mHolder.addCallback(this);
        mHolder.setType(SurfaceHolder.SURFACE_TYPE_PUSH_BUFFERS);
        
        mToggleButton = (ToggleButton) findViewById(R.id.toggleRecordingButton);
        mToggleButton.setOnClickListener(new OnClickListener() {
                @Override
                // toggle video recording
                public void onClick(View v) {
                        if (((ToggleButton)v).isChecked())
                                mMediaRecorder.start();
                        else {
                                mMediaRecorder.stop();
                                mMediaRecorder.reset();
                                try {
                                        initRecorder(mHolder.getSurface());
                                } catch (IOException e) {
                                        e.printStackTrace();
                                }
                        }
                }
        });
    }

The preview frames from the camera will be displayed on a SurfaceView. Recording is controlled by a toggle button. After the recording is over, we stop the MediaRecorder. Since the stop method resets all the state machine variables in order to be able to grab another video we reset the state machine and call our initRecorder once more.

initRecorder is where we configure the MediaRecorder and the camera:

    /* Init the MediaRecorder, the order the methods are called is vital to
     * its correct functioning. 
     */
    private void initRecorder(Surface surface) throws IOException {
        // It is very important to unlock the camera before doing setCamera
        // or it will results in a black preview
        if(mCamera == null) {
            mCamera = Camera.open();
            mCamera.unlock();
        }

        if(mMediaRecorder == null)
            mMediaRecorder = new MediaRecorder();
        
        mMediaRecorder.setPreviewDisplay(surface);
        mMediaRecorder.setCamera(mCamera);
             
        mMediaRecorder.setVideoSource(MediaRecorder.VideoSource.CAMERA);
        mMediaRecorder.setOutputFormat(MediaRecorder.OutputFormat.DEFAULT);
        File file = createFile();
             
        mMediaRecorder.setOutputFile(file.getAbsolutePath());
              
        // No limit. Don't forget to check the space on disk.
        mMediaRecorder.setMaxDuration(-1);
        mMediaRecorder.setVideoFrameRate(15);
        
        mMediaRecorder.setVideoEncoder(MediaRecorder.VideoEncoder.DEFAULT);
              
        try {
            mMediaRecorder.prepare();
        } catch (IllegalStateException e) {
            // This is thrown if the previous calls are not called with the 
            // proper order
            e.printStackTrace();
        }
                
        mInitSuccesful = true;
    }

It is important to create and unlock a Camera object before of the creation of a MediaRecorder. setPreviewDisplay and setCamera must be called immediately after the creation of the MediaRecorder. The choice of the format and the output file is obligatory. Other options are optional but they must be called in the order outlined in the code above.

The MediaRecorder is best initialized when the surface has been created. We register our Activity as a SurfaceHolder.Callback listener in order to be notified of this and override the surfaceCreated method to call our initialization code:

        @Override
        public void surfaceCreated(SurfaceHolder holder) {
            try {
                if(!mInitSuccesful)
                    initRecorder(mHolder.getSurface());
                } catch (IOException e) {
                    // TODO Auto-generated catch block
                    e.printStackTrace();
                }
        }

When you're done with the surface, don't forget to release the resources, the camera is a shared object and may be used by other applications as well:

        private void shutdown() {
            // Release MediaRecorder and especially the Camera as it's a shared
            // object that can be used by other applications
            mMediaRecorder.reset();
            mMediaRecorder.release();
            mCamera.release();
              
            // once the objects have been released they can't be reused
            mMediaRecorder = null;
            mCamera = null;
        }

Override the surfaceDestroyed method so the previous code can be called automatically when the user is done with the Activity:

        @Override
        public void surfaceDestroyed(SurfaceHolder holder) {
            shutdown();
        }

Create a spinning cube with OpenGL ES and Android

marco.dinacci@intransitione.com (Marco Dinacci) — Sat, 30 Apr 2011 00:00:00 +0100

Android supports 3D graphics via the OpenGL ES API, a flavor of OpenGL specifically designed for embedded devices.

This article is not an OpenGL tutorial, it assumes the reader has already basic OpenGL knowledge. The final result will look like this:

First we write a new Activity and in the onCreate method we create the two fundamental objects we need to use the OpenGL API: a GLSurfaceView and a Renderer.

public class OpenGLDemoActivity extends Activity {
        
    @Override
    public void onCreate(Bundle savedInstanceState) {
        super.onCreate(savedInstanceState);
        
        // Go fullscreen
        requestWindowFeature(Window.FEATURE_NO_TITLE);
        getWindow().setFlags(WindowManager.LayoutParams.FLAG_FULLSCREEN,
                        WindowManager.LayoutParams.FLAG_FULLSCREEN);
        
        GLSurfaceView view = new GLSurfaceView(this);
        view.setRenderer(new OpenGLRenderer());
        setContentView(view);
    }
}

And this is our Renderer that uses a simple Cube object we'll describe later do display a spinning cube:

class OpenGLRenderer implements Renderer {

        private Cube mCube = new Cube();
        private float mCubeRotation;

        @Override
        public void onSurfaceCreated(GL10 gl, EGLConfig config) {
            gl.glClearColor(0.0f, 0.0f, 0.0f, 0.5f); 
                
            gl.glClearDepthf(1.0f);
            gl.glEnable(GL10.GL_DEPTH_TEST);
            gl.glDepthFunc(GL10.GL_LEQUAL);

            gl.glHint(GL10.GL_PERSPECTIVE_CORRECTION_HINT,
                      GL10.GL_NICEST);
                
        }

        @Override
        public void onDrawFrame(GL10 gl) {
            gl.glClear(GL10.GL_COLOR_BUFFER_BIT | GL10.GL_DEPTH_BUFFER_BIT);        
            gl.glLoadIdentity();
            
            gl.glTranslatef(0.0f, 0.0f, -10.0f);
            gl.glRotatef(mCubeRotation, 1.0f, 1.0f, 1.0f);
                
            mCube.draw(gl);
               
            gl.glLoadIdentity();                                    
                
            mCubeRotation -= 0.15f; 
        }

        @Override
        public void onSurfaceChanged(GL10 gl, int width, int height) {
            gl.glViewport(0, 0, width, height);
            gl.glMatrixMode(GL10.GL_PROJECTION);
            gl.glLoadIdentity();
            GLU.gluPerspective(gl, 45.0f, (float)width / (float)height, 0.1f, 100.0f);
            gl.glViewport(0, 0, width, height);

            gl.glMatrixMode(GL10.GL_MODELVIEW);
            gl.glLoadIdentity();
        }
}

Our onSurfaceChanged and onDrawFrame methods are basically the equivalent of the GLUT glutReshapeFunc and glutDisplayFunc. The first is called when the surface is resized, for instance when the phone switch between landscape and portrait mode, the second is called every frame and that's where we put the code to draw our cube.

class Cube {
                
    private FloatBuffer mVertexBuffer;
    private FloatBuffer mColorBuffer;
    private ByteBuffer  mIndexBuffer;
        
    private float vertices[] = {
                                -1.0f, -1.0f, -1.0f,
                                1.0f, -1.0f, -1.0f,
                                1.0f,  1.0f, -1.0f,
                                -1.0f, 1.0f, -1.0f,
                                -1.0f, -1.0f,  1.0f,
                                1.0f, -1.0f,  1.0f,
                                1.0f,  1.0f,  1.0f,
                                -1.0f,  1.0f,  1.0f
                                };
    private float colors[] = {
                               0.0f,  1.0f,  0.0f,  1.0f,
                               0.0f,  1.0f,  0.0f,  1.0f,
                               1.0f,  0.5f,  0.0f,  1.0f,
                               1.0f,  0.5f,  0.0f,  1.0f,
                               1.0f,  0.0f,  0.0f,  1.0f,
                               1.0f,  0.0f,  0.0f,  1.0f,
                               0.0f,  0.0f,  1.0f,  1.0f,
                               1.0f,  0.0f,  1.0f,  1.0f
                            };
   
    private byte indices[] = {
                              0, 4, 5, 0, 5, 1,
                              1, 5, 6, 1, 6, 2,
                              2, 6, 7, 2, 7, 3,
                              3, 7, 4, 3, 4, 0,
                              4, 7, 6, 4, 6, 5,
                              3, 0, 1, 3, 1, 2
                              };
                
    public Cube() {
            ByteBuffer byteBuf = ByteBuffer.allocateDirect(vertices.length * 4);
            byteBuf.order(ByteOrder.nativeOrder());
            mVertexBuffer = byteBuf.asFloatBuffer();
            mVertexBuffer.put(vertices);
            mVertexBuffer.position(0);
                
            byteBuf = ByteBuffer.allocateDirect(colors.length * 4);
            byteBuf.order(ByteOrder.nativeOrder());
            mColorBuffer = byteBuf.asFloatBuffer();
            mColorBuffer.put(colors);
            mColorBuffer.position(0);
                
            mIndexBuffer = ByteBuffer.allocateDirect(indices.length);
            mIndexBuffer.put(indices);
            mIndexBuffer.position(0);
    }

    public void draw(GL10 gl) {             
            gl.glFrontFace(GL10.GL_CW);
            
            gl.glVertexPointer(3, GL10.GL_FLOAT, 0, mVertexBuffer);
            gl.glColorPointer(4, GL10.GL_FLOAT, 0, mColorBuffer);
            
            gl.glEnableClientState(GL10.GL_VERTEX_ARRAY);
            gl.glEnableClientState(GL10.GL_COLOR_ARRAY);
             
            gl.glDrawElements(GL10.GL_TRIANGLES, 36, GL10.GL_UNSIGNED_BYTE, 
                            mIndexBuffer);
                
            gl.glDisableClientState(GL10.GL_VERTEX_ARRAY);
            gl.glDisableClientState(GL10.GL_COLOR_ARRAY);
    }
}

The Cube uses two FloatBuffer objects to store vertex and color informations and a ByteBuffer to store the face indices. In order for the buffers to work it is important to set their order according to the endianness of the platform using the order method. Once the buffers have been filled with the values from the arrays, the internal cursor must be restored to the beginning of the data using buffer.position(0).

PyQT playground: image gallery with zoom

marco.dinacci@intransitione.com (Marco Dinacci) — Wed, 28 Jan 2009 00:00:00 +0000

In this article I'm going to create a small PyQT based application that displays images in a grid and that zoom the image hovered by the mouse.

This is how the final application looks like:

The first class we need is an ImageGallery, which is a mere QDialog using a QGridLayout.

class ImageGallery(QDialog):

    def __init__(self, parent=None):
        super(QDialog, self).__init__(parent)
        self.setWindowTitle("Image Gallery")
        self.setLayout(QGridLayout(self))

    def populate(self, pics, size, imagesPerRow=4,
                 flags=Qt.KeepAspectRatioByExpanding):
        row = col = 0
        for pic in pics:
            label = ImageLabel("")
            pixmap = QPixmap(pic)
            pixmap = pixmap.scaled(size, flags)
            label.setPixmap(pixmap)
            self.layout().addWidget(label, row, col)
            col +=1
            if col % imagesPerRow == 0:
                row += 1
                col = 0

The code is very simple, the class has only one public method, populate, which iterates over the images passed in the argument pics, create a QPixmap for each of them, and add them to the layout.

The interesting bit is the line where I create the label:

label = ImageLabel("")

This line creates a new instance of the ImageLabel, which is listed here below:

class ImageLabel(QLabel):
    """ This widget displays an ImagePopup when the mouse enters its region """
    def enterEvent(self, event):
        self.p = ImagePopup(self)
        self.p.show()
        event.accept()

The ImageLabel is a normal QLabel except that I redefined the enterEvent in order to show an ImagePopup when the mouse enters the label.

Finally, the ImagePopup class, responsible for the creation of the zoomed popup image.

class ImagePopup(QLabel):
    """
    The ImagePopup class is a QLabel that displays a popup, zoomed image
    on top of another label.
    """
    def __init__(self, parent):
        super(QLabel, self).__init__(parent)

        # set pixmap and size, which is the double of the original pixmap
        thumb = parent.pixmap()
        imageSize = thumb.size()
        imageSize.setWidth(imageSize.width()*2)
        imageSize.setHeight(imageSize.height()*2)
        self.setPixmap(thumb.scaled(imageSize,Qt.KeepAspectRatioByExpanding))

        # center the zoomed image on the thumb
        position = self.cursor().pos()
        position.setX(position.x() - thumb.size().width())
        position.setY(position.y() - thumb.size().height())
        self.move(position)

        # FramelessWindowHint may not work on some window managers on Linux
        # so I force also the flag X11BypassWindowManagerHint
        self.setWindowFlags(Qt.Popup | Qt.WindowStaysOnTopHint
                            | Qt.FramelessWindowHint
                            | Qt.X11BypassWindowManagerHint)

    def leaveEvent(self, event):
        """ When the mouse leave this widget, destroy it. """
        self.destroy()

That's it, it's very simple as you can see. You can download the python code or a zip file containing the code and the images.

Thanks for reading.

Embedding Mplayer in a PyGTK application

marco.dinacci@intransitione.com (Marco Dinacci) — Sat, 16 Feb 2008 00:00:00 +0000

You can find the complete (but fairly hackish) source code for the article here: AudioExtractor.zip

Recently I wanted to extract the sound track from a MP4 file. You actually need a single command-line to do that, either using Mplayer or GStreamer.

It's amazing what you can do with GStreamer pipelines, but unfortunately I ran into several problems trying to play H.264 video files with AAC audio (the files are produced by a Sanyo camera and make Totem crash after about few seconds...) so I gave up and I took a look at MPlayer. After all I really didn't need all the GStreamer functionalities.

A single line actually does the job:

mplayer -ao pcm:file=$OUTPUT_FILE $INPUT_FILE

Couldn't be easier.

Nevertheless, I wanted to command the extraction from a GUI as it makes it more user friendly for people that don't deal with shells everyday. Additionally, I didn't want the MPlayer window to popup, I actually wanted to embed Mplayer inside my application.

Again, it turns out that a single line does the job:

mplayer -ao pcm:file=$OUTPUT_FILE -wid $WINDOW_XID

The -wid option tells Mplayer to connect to the window represented by the ID $WINDOW_XID. This ID is actually the identifier given by the X Server to the window. We'll see later how to retrieve it.

With this setup, the video is shown in the application, the problem is that there is no indication about the progress. What I wanted was to periodically ask to MPlayer the status of the playback, and displaying it in a progress bar. Checking the MPlayer's man page I discovered that it can be controlled by putting it into a slave mode and by sending him the commands through a named pipe. Bingo !

To start the mplayer in slave mode, I modified the command line to look like this:

mplayer -ao pcm:file=$OUTPUT_FILE -wid $WINDOW_XID -slave -idle \n
-input file=$FIFO

where FIFO is the absolute path of the named pipe where the commands are sent (for a complete list of commands, type mplayer -input cmdlist).

On the Python side, you have to embed the video in a GtkDrawingArea. The documentation for this component just says: The GtkDrawingArea widget is used for creating custom user interface elements. It's essentially a blank widget; you can draw on widget->window

It is actually everything I need, as the only thing required to do is to get the X ID of the window and pass it to Mplayer. You do it like this:

xid = canvas.window.xid # where canvas is an instance of GtkDrawingArea

We have almost everything required to start Mplayer, the only missing thing is the FIFO. In Python, you create a FIFO using

os.mkfifo("path_to_fifo")

The full code to start Mplayer in another process would then be:

command = "mplayer -ao pcm:file=%s -wid %i -slave -idle \n
-input file=%s" % (outputFile,xid,fifo)
command = command.split().append("input_video_file.mp4")
subprocess.Popen(command, stdout=logfile, stderr=logfile)

I added the input file after the call to split because if there are whitespaces in the filename or in a parent directory, then the filename would break down in multiple tokens with the call to split.

Now that MPlayer has started, we have to tell him what to do. To do that, we write to the pipe:

mplayerClient = open(FIFO,"w")
mplayerClient.write("play\n") # commands must ends with a newline

(Actually, I still haven't get MPlayer to not start automatically the playback, even if I'm using the -idle option. Please post a comment if you can shed a light on the subject...)

At this point the application is showing the video, but as we said before, we also want to update a progress bar to show the progress, and to do that we've to repeatedly ask the status to Mplayer.

The easiest way to do that, since I was using Gtk, is to use the gobject.timeout_add call, like this:

# call the myUpdateFunction each UPDATE_INTERVAL milliseconds
gobject.timeout_add(UPDATE_INTERVAL, myUpdateFunction)

and myUpdateFunction is defined as:

def myUpdateFunction():
    keepGoing = True
    mplayerClient.write("get_percent_pos\n")
    mplayerClient.flush() # really important, or commands won't be always sent !
    line = logfile.readline()
    if line and line.startswith("ANS_PERCENT_POSITION"):
    # +1 because progress starts from 0, it would end to 99 otherwise
        progress = int(line[line.index('=')+1:-1]) + 1
        progressBar.set_text("%s %%" % progress)
        progressBar.update(progress/100.0)
    else:
        keepGoing = False

return keepGoing

That's it, progressBar is a normal GtkProgressBar object and logfile is where I previously redirected Mplayer stdout and stderr.

The above code won't actually works out of the box as there are some corner cases you have to deal with, for example it can happen that the update function is called before MPlayer starts to play the video (which should actually happen with the -idle option, but I couldn't get it to work). In that case line would be an empty string and the playback will immediately stop as myUpdateFunction will return False.

Note that the program also pops up a dialog asking you if you want to burn the file with Serpentine or save it to a different location.

Thanks for reading