Inspectfs relies on libudev for querying device information but I wrote some C++ wrapper classes because I want the project to contain as much C++ and as less C code as possible.
Soon you’ll be able to find the code and more on github, and if you want to join me, post a message here!

I sort of finished the backend of my model translation tool (It reads an Enterprise Architect Class Diagram [/buzzwordalert] ), and now it’s time for the client. The problem with that is that I always get bored when the interesting stuff is done. In this case, The Interesting Stuff is figuring out how the Class Diagram parts of UML fit together and how to translate that to usable interfaces so that it can be transformed into code.

Luckyly for me, another interesting part is coming up: I have to create an Eclipse plug-in that uses the backend and produce kick-ass quality Java code of it. And reverse engineering, ofcourse. Not that I’m willing to be a rival of any other tool like, say, for instance the Eclispe Modeling Framework. I just want to know how stuff like that is done and what problems are encountered while creating code for that purpose, and I secretly hope that I will be able to contribute to the EMF after my graduation internship because after reading only the beginning of the Eclipse Modeling Framework I’m very excited about it.

I also ordered a book on creating Eclipse plug-ins because I couldn’t find much of a tutorial so I’ll be reading a lot, the next couple of weeks. But until that time I am a little off my game here with translating models and all. Lost in translation.

As it turns out so far, Visual Studio provides a Class Designer tool, but that’s not as extensive as I’d hoped. On the other hand there are the DSL Tools which enables me to create a Domain Specific Language that provides the necessary functionality. In fact, one could create a DSL that generates Java code – or whatever language, but that’s an obvious property of any DSL.
The Eclipse Modeling Framework has way more functionality, but can – as far as I know right now – only generate Java code.
The third tool, the Enterprise Architect, is an UML modeling tool that also provides very basic code generation. Unfortunately I must say that I’m not satisfied at all with the generated code by the Enterprise Architect, and even less satisfied with the plug-ins they provide for Eclipse and Visual Studio. In the generated Java code, enums and interfaces are apparently ignored; in C# code, the namespaces are not created and class methods are generated but getters and setters are neither recognized nor generated (this also goes for Java).

So I have decided that I’m going to write at least an Eclipse plug-in that reads the model from Enterprise Architect and creates decent hand-written-quality code for the domain classes. And given time, I’ll also create a Visual Studio plug-in that does the same. After that, the next step would be to extend these plug-ins with functionality to generate persistence and GUI code. Another thing is that it should be possible to alter the code after it’s been generated and send the changes back to the model. I’ll incorporate that as well in the first version of the plug-ins.

That leaves me with a lot of work todo, so I won’t get bored the next couple of months

Although these things all have a high nerd factor, I’ve a plan that can top that. In the summer holidays, I’ve decided to go to Hacking At Random, which will be in The Netherlands this year. I think that my friends and classmates (who also will be there) and I will have a *lot* of fun. As a matter of fact, I promise to show you some pictures afterwards.

While I’m thinking about it, my learning-by-cheating idea didn’t go as well as I thought: I didn’t finish that little project (but alas, that’s one of many). What did come out of it is Sentry. The Sentry core is actually nothing more than a library loader, but is has some nice features. Each plug-in (or library) can provide a number of hookpoints, where other plug-in commands can hook onto. Nice thing about this is that you can create a completely event-driven application, and anyone is free to create their own plug-ins. At this moment, I’m using it to create an IRC bot that is kind of able to control what’s happening in a channel and also to give my c++ knowledge a boost If you’re interested: you can take a quick look at my websvn subdomain, the repository is sentry_cpp, and it’s also possible to svn co it from the svn subdomain.

That’s all for now, happy 2009 y’all!

Some time after I finished the part of high school we call ‘brugklas’, I came across a teacher who, to my surprise, encouraged using cheating notes. As I soon would realise, the man ment for us to create cheating notes before the exam, and not use them during one. That way, he said, you learn to summarize the things you need to study for that exam, and you’ll remember it, too.

The man was right. So, now that’s all behind me, I can use this story to do some cheating, while pretending to learn from it as well The cheating is about a trivia channel on irc. I’m writing a bot in Java that can parse irssi logfiles, and is also able to learn from the questions the trivia bot asks. The things I’m hoping to learn from this: How to simulate a human being that answers the questions (that means, among other things: do not answer every question correctly, but also that is makes some sense), and what is the nicest, most efficient way to learn questions; how do I determine the correct way to answer, etc.

So: $learning_by_cheating++

A classmate told me to use valgrind to test Skaar for memory leaks. Valgrind is a suite of tools for debugging and profiling programs, so that should work, and there I went:

valgrind --log-file=./valgrind.log --leak-check=full ./skaar

After shutting down, it turned out that Skaar had left something behind, and not that modest (43Kb) either. It’s actually not that simple to prevent memory leaks, I learned. The obvious cases seem to scream: free(3) or delete me, but the not so obvious cases keep their mouths perfectly shut..

Obvious case:
char* line = (char*)malloc(513);
memset(line, 0, 513);
strcpy(line, "PRIVMSG #mychannel :foo\r\n");
...
free(line);

Not so obvious case:
// in RFC1459.cpp
JoinMessage* joinmessage = new JoinMessage(this, line);
return (AbstractMessage*)joinmessage;

// somewhat later in Skaar.cpp
AbstractMessage* message = protocol->translateIncoming(rawmsg);
...
delete message;


I’m wondering if this is the right way to end the AbstractMessage*‘s life, by the way. Who’s responsibility is it to get rid of the message? I could go with this solution, that says: “Here is your object, have fun”. I could also create a function in the protocol that destroys the message, something like RFC1459::destroyMessage(AbstractMessage* message), that says “Here is your object, but you have to let me take care of it’s end. Actually, I have no clue which one is the best choice, if there is a best choice here..

But, back to business: all I have to do now is just test every possibility to see if there are any leaks. Or I could just watch it when I write code, which also sounds nice.
Another tool that I’ve come to value is gdb, The GNU Project Debugger. This tool allows you to see what happens ‘inside’ the program while it executes. And yes, I stole that line from gnu.org
To start using gdb is not very hard, just type gdb yourprogram, and that’s that. To actually start executing you program, type run once gdb has started.

The HUbuntu project started last week (actually, it hasn’t gotten off it’s feet for a few weeks), and at the moment we are thinking about how to implement easy access to our own repository. In this repository students will be able to find all the software they need for the courses given at the HU, and we want to make that process really easy for them. We already came up with the principal of meta-packages, and today I did some research in that area. The idea is to create a meta-package for each course and determine the right dependencies for that course. I think that a lot of work has to be done, but the general idea is finished. Yay, isn’t that great?

I also made some progress with my or-mapping. For mattijs’ peace I will post a link to the Object class, just as he asked me to: link. It is password-protected, but if you ask nicely, I will provide you with an account. If you create an Object containing an array of Objects in it, the intermediate tables that are needed are now automagically created, but I still have to build that for arrays of primitive PHP types.

After starting up the HULK for about a fortnight, we started to bring up ideas for projects (so if anyone has an idea…). Actually, while I’m writing this I realize that I haven’t told you anything yet about the HULK (hmm.. weird).
Well, HULK stands for Hogeschool Utrecht Linux Kennisgroep, and consists of a few classmates and yours truly, who try to give Linux a shot in our university, the Hogeschool of Utrecht. The website is in Dutch, so if you want to follow up on it, you have to be able to read it.

We thought about some really cool projects like building our own Ubuntu distribution, HUbuntu , setting up Linux on mobile devices, teaching Unix, and also some smaller projects like building a logbook to keep track of what we’re doing and to provide student services (SSH access, web servers, database servers, etc.). Apart from that, I’ve started learning c(++), and for practice I’m building an IRC client, called skaar.

Next week, together with Mattijs, I’m starting on the logbook which will be built in PHP, so that we can start logging quickly.

Keeping you posted!

Before I started to write on this code, I divided it in three parts:

1. Creating Objects
2. Updating Objects
3. Deleting Objects

I finished a test version of the first part today. Basically, this is how it works: There is a class Object that should be extended by every class that has to be in a database. If you create a class extending this one, it will also be created in the database (if it doesn’t exist yet) as a table. For example, see this class:

 

/**
 * foo
 * bar
 * @table_name=__TEST__
 *
 */
class Test extends Object {    

     /**
     * @nullable = false
     */
    private $firstname; 

    private $lastname; 

    /**
    * @type = string(12)
    * @nullable = false
    */
    private $dateofbirth; 

    private $city; 

    function __construct() {
       parent::__construct();
   } 

    public function getFirstName(){
   	return $this->firstname;
   } 

    public function setFirstname($firstname){
   	$this->firstname = $firstname;
   } 

    public function getLastname(){
   	return $this->lastname;
   } 

    public function setLastname($lastname){
   	$this->lastname = $lastname;
   } 

    public function getDateofbirth(){
   	return $this->dateofbirth;
   } 

    public function setDateofbirth($dob){
   	$this->dateofbirth = $dob;
   } 

    public function getCity(){
   	return $this->city;
   } 

    public function setCity($city){
   	$this->city = $city;
   }
}

Will result in this table:

Pretty cool, huh? I still have to do some work, but the base is there. As you can see, you can also use some kind of annotations to set the information for the table and the columns in the database, but there are also default settings if they are omitted.
Enough for today, I’m going to celebrate my weekend and “Oud & nieuw”, as we call New Year’s Eve in Holland Soon more about this!

Strangely enough, at that time there was no framework or abstraction layer available for PHP, so I started building one myself. The idea was for it to work like EJB, and the result was satisfactory. Then doom hit me, because I lost all that code So now I’m going to start all over again, with a slightly different angle: I am going to write a framework that communicates with the Zend Framework, but with a little extra: the Zend Framework does not provide creation and automatic querying for your objects in the database, and that’s what will be the extra in my framework.