Tassos Bassoukos - Software

XMLTV grabber for Greece, take 2

abassouk@gmail.com (Tassos Bassoukos) — Tue, 26 Feb 2008 12:39:21 +0200

I had previously written an XMLTV grabber for Greece that had several shortcomings (problems with character encodings and scheduling of programmes after midnight), and left it there. Now, a from-scratch rewritten version is available: XMLTV grabber for Greece.

This version uses the improved XMLTV api, hence it should be available in applications that use tv_find_grabbers. Currently it extracts the following items from the Nova site:

Program title
Program description
Start time
Repeat status
Rating
Categories

Optionally, in slow mode it also extracts the following items if they are available:

Synopsis
Image
Credits
Year of production
Duration

I would be interested in hearing from anyone that uses it in non-UTF8 environments (e.g. Windows users). Note that you may need to pass the results through tv_sort in some cases.

Many thanks to vrypan for providing the motive for completing this.

Status of Java Closures

abassouk@gmail.com (Tassos Bassoukos) — Mon, 4 Feb 2008 04:34:28 +0200

Since before Java SE 6 came out, a discussion has been going on in the Java community future directions and extensions of the Java language, with the most energy directed on closures. There are currently two major camps on this issue, the BGGA proposal and the CICE proposal.

CICE

Concise Instance Creation Expressions by Bob Lee, Doug Lea, and Josh Bloch favor adding simple syntactic sugar for anonymous inner class creation, where

List ls = ... ;
Collections.sort(ls, new Comparator() {
  public int compare(String s1, String s2) {
    return s1.length() - s2.length();
  }
});

can be written as

List ls = ... ;
Collections.sort(ls,
  Comparator(String s1, String s2){ return s1.length() - s2.length(); });

Additionally, it allows capture of and access to the local variables of the enclosing block(s). In general it's a small proposal, with few warts, that simply provides for small gains in legibility.

BGGA

BGGA is the proposal by Gilad Bracha, Neal Gafter, James Gosling and Peter von der Ahé, which is much more heavyweight, providing for true closures, function types and user-defined control abstractions. The above example could be written as

List ls = ... ;
Collections.sort(ls,  {String s1, String s2 => int  s1.length() - s2.length()});

However, this proposal leads to some weird interactions with expected behaviour:

...
System.out.println("1");
{=>return;}.invoke();
System.out.println("2");
...

The previous snippet will not print "2", as the return is interpreted as a return from the defining method, not the closure; a new type has been added (java.lang.Unreachable) so that the previous code should not compile. What happens tho if you store the closure and invoke it after the defining method has returned? See the Joshua Bloch presentation at JavaPolis 2007 for a recap of the present problems.

All in all, while BGGA adds new powerful concepts and new problems to the Java language, the community is still undecided on whether the inherent problems with closures are justified by the expressiveness they enable; several voices advocate leaving Java unmodified and switching over to JVM-based languages that have them built in, such as Scala.

Using GWT modules in not-so-bright frameworks

abassouk@gmail.com (Tassos Bassoukos) — Tue, 10 Oct 2006 12:57:55 +0300

I've recently had to use GWT to implement a module for a PostNuke installation. Two problems came up with that - first, the HTML head tag was in a static file somewhere, and I could not modify it and second, the URL that would call the module was not the URL for the files of the module.

To add a GWT module to a static page requires three things:

A module specific element, which usually has a specific id attribute, where its contents will be replaced by the modules controls,
A META tag in the HEAD of the page, specifying the modules that will be loaded,
A SCRIPT tag linking to the gwt.js file.

Requirements (1) and (3) are easily met by modifying the actual HTML page, or in my case, emitting via PHP the HTML elements; however, requirement (2) is a bit harder to implement when one does not have access to the file that contains the HEAD element. Since we're doing AJAX anyway, a Javascript snippet will help us out a bit:

<script type="text/javascript">
   var headTag = document.getElementsByTagName('head')[0];
   var metaTag = document.createElement('meta');
   metaTag.setAttribute('name','gwt:module');
   metaTag.setAttribute('content','com.example.Module');
   headTag.appendChild(metaTag);
   headTag=null;
   metaTag=null;
</script>

Inserting that code before the tag for (3) will ensure that our module appears at least in the DOM. However, we aren't done yet, as the files for each module are resolved by GWT relative to the current location of the browser - which isn't that useful when your URL is something along the lines of "http://example.com/index.php?module=this&q=that" but your GWT modules reside in "/gwt-modules/example".

Googling around did not give any insight on the issue, so I delved a bit into the innards of GWT, and found an undocumented feature that allows one to rebind the base URL for a GWT module: simply use

<meta name="gwt:module" content="http://example.com/gwt-modules/example=com.example.Module" />

instead of

<meta name="gwt:module" content="com.example.Module" />

Obviously, this also allows for modules in different locations.

Smallest code segments?

abassouk@gmail.com (Tassos Bassoukos) — Thu, 20 Jul 2006 03:22:26 +0300

Quick: how long is the shortest program to calculate the first prime number after a given number? The answer might surprise you!

In perl, it's about 60 characters:

perl -e '$i=$ARGV[0]+1;$i++while("."x$i)=~/^(..+?)\1+$/;print"$i\n";' 90

Of course, I believe APL will win that contest hands down...

P.S: There's even a perl competition to produce the shortest code for a given problem, with the name Perl Golf...

Character set that understands property files

abassouk@gmail.com (Tassos Bassoukos) — Mon, 12 Dec 2005 11:58:45 +0200

If you are a Java developer in a non-ISO8859-1 environment, you will have encountered the character set problem when editing property files. The spec says that they are in ISO8859-1, with Unicode escape sequences for characters outside that set. While one can set up automatic translation, it is intrusive and not suited to all cases. Being fed up with the problem and the inelegance of all the available solutions, I wrote a custom character set provider which handles the property file encoding as a multi-byte character set.

The source is here, and the prebuilt JAR is here. To use the property character set, drop the JAR file in the extension folder of your JRE, usually $JAVA_HOME/lib/ext and restart (or start) your application. Property files should be assigned the "X-PROP" character set, and all should be ready.

If you are using Eclipse 3.1 or later, simply go to Window → Preferences → General → Content Types → Text → Java Properties files, set the default encoding to X-PROP and select Update to apply the setting.

The code is in the public domain, and I'm not responsible if it munges your files, destroys your computer or causes the next ice age, etc...

Sharks with Lasers on their heads!

abassouk@gmail.com (Tassos Bassoukos) — Tue, 30 Aug 2005 08:55:39 +0300

A recent slashdot discussion that forked off to GUI vs CLI made me think that this split looks artificial, and that a merger of the two would be at least worth investigating.

My idea is to add an optional command-line to each Nautilus window, which will be one text line high and would appear between the status bar and the icon view. It would be a normal terminal running the user's shell, with a number of interaction possible between the shell and the command line:

the current directory of the shell and the nautilus window would be synchronized. The gnome-terminal in Ubuntu breezy at the time of writing has an implementation of current directory detection based on escaped echos at each prompt. Some problems here with the other way of synchronization, as a command to the shell might not be parsed due to the user having typed some text.
file selection on the icon view would result in the addition of the files to the command line. Again, problems will exist if the user types some text. Also, what happens if the use deselects some files?
a special command would take a list of files and show them in the iconic view. Such lists of file scan be generated with find or grep.
a keyboard command would switch the view between the four different states: no command line, 1 line high command line, 50% split between command line and icons, only command line (effectively degenerates into a terminal)

XMLTV grabber for Greece

abassouk@gmail.com (Tassos Bassoukos) — Sun, 14 Nov 2004 04:49:13 +0200

Some time ago I installed MythTV on my computer. One major problem was that it could not find schedules for the greek television stations. As it used XMLTV as the schedule grabber, I went and wrote one for Greece.

So, here it is: tv_grab_el will sceen-scrape NOVA's website and generate a XMLTV - compatible file.

Please be gentle, as a full, slow search for 7 channels will download about 10 megabytes. Also take note that I haven't managed to keep up with the actual XMLTV progress for several months, so while it works for me, it may not work for you.

Notes: Don't forget that you will need an XMLTV installation to run this. Try the -cache option, it should increase the speed of subsequent downloads, with the drawback that you will miss updates on the schedules. The documentation is embedded in the file. I've have had problem reports when running on Windows.

From a different age...

abassouk@gmail.com (Tassos Bassoukos) — Fri, 29 Oct 2004 06:18:45 +0300

I was rummaging around my old and not-so-well-kept archives, when I found two examples of my early programming skills (or to be precise, my lack of them...)

The first example is a rudimentary assembler written in AmigaBasic here. Major features include non-structured programming (the whole thing is implemented as a single if-elseif block, mostly due to the fact that AmigaBASIC was crap and you couldn't do indirect jumps), not allowing includes, having extremely simple assembly-time arithmetics and structures, and not being able to link. This was written in a summer somewhere around 1990 or 1991 - at that time I had a disassembler on my hands, but no assembler, and wanted to go a step up from AmigaBASIC. Noteworthy is that I remember this to be a huge piece of code, easily stretching the limits of my capabilities when I wrote it but I just paged thru it and it is just 623 lines of code.

The second example is a FORTH interpreter featuring non-optimized just-in-time code generation. Actually, it would be quite easy to improve speed by an order of magintude now that I think about it, as FORTH uses a stack and most of the instructions just manipulate the stack. The zip file contains the core in assembly (about 1200 instructions), the add-on numerical computation words and interfaces to several Amiga libraries. This was written during the summer of 1994 or 1995 and no, don't ask me about the implementation of the internals, I'd need to re-read and understand the whole thing from scratch.

Forth was such a pleasant language to write programs in, but very hard to read. I don't mean this in the perl kind of sense; FORTH generally encourages bottom-up design, and you will find that you are using a quite different, custom-built language when you start implementing the higher levels of you application.

I'm considering these files to be in the public domain, however I would appreciate it if you would left me a note here.