SpringLobby release

Long time no post, but now there is a reason to make one. A new SpringLobby has been released and again I've been working on a new feature for it. This time a somewhat bigger feature then just displaying a map's heightmap.

The thing I couldn't stand about SpringLobby was it's map selector. Or rather, it's lack thereof. The map selector was a simple combobox. That's very nice if you have only 10 maps and can remember their names by heart, but when you have 650 maps and can never remember maps by their name it's about the worst selection mechanism I can think of.

So... a while ago I decided I'd write a map selector to fix this "major flaw", and also to actually gain some hands on experience with the wxWidgets windowing toolkit, and with multithreaded programming.

Here is the end result:


The selector allows you to pick a basic set of maps (all, currently played, recently played) and shows these in a draggable grid to the right. The sort order in both directions can be chosen arbitrarily (I'm not yet entirely sure of the use of this), and of course it's also possible to quickly filter the shown maps on any text in their name or description. Combined with the ease of recognizing maps by minimap I think this makes for a powerful map selector.

Comments, suggestions and bug reports are welcome!

WolfenQt

Today I spotted this post about Qt widget rendering in OpenGL on the Trolltech blog. You can read it, or you can just watch the video:



It links to another post, which already describes features that would be sufficient for an experiment. What Samuel Rødal shows in his posts would mean basically all Qt widgets would be able to render to an arbitrary OpenGL surface. This suggests it would be interesting to experiment with Qt integration in Spring. Basically any GUI toolkit would be a huge improvement over the current GUI of Spring, and the fact that we'd get it for free with even a company (Nokia) backing it, would be a huge bonus.

Besides a replacement for Spring's native GUI, LUA bindings could be supplied to allow LUA widgets and gadgets to create a consistent look and feel throughout the engine, no matter whether a piece of functionality is hardcoded in the engine or supplied by a LUA script.

Could this mean the end of the "Bermuda GUI"?

String concatenation, now in near infinite time!

One particular thing has always surprised me in languages like Java and C#. Consider the following piece of code, taken from this file in ChanServ (a bot in the Spring lobby server for moderation/administration purposes.)

// precondition: there exists a logfile with filename ,
// which may contain a very large amount of lines, which are
// to be inserted in a MySQL database.

ArrayList<Long> stamps = new ArrayList<Long>();
ArrayList<String> lines = new ArrayList<String>();

String update = "";
BufferedReader in = new BufferedReader(new FileReader(file));
String line;

int lineCount = 0;
while ((line = in.readLine()) != null) {
    if (line.trim().equals("")) continue; // empty line

    long stamp;
    try {
        stamp = Long.parseLong(line.substring(0, line.indexOf(' ')));
    } catch (NumberFormatException e) {
        Log.error("Line from log file " + file.getName() + " does not contain time stamp. Line is: \"" + line + "\"");
        return ;
    }

    if (lineCount == 0) {
        update += "INSERT INTO `" + name + "` (stamp, line) values (?, ?)";
    } else {
        update += ","+ Misc.EOL + "(?, ?)";              
    }
     
    stamps.add(stamp);
    lines.add(line.substring(line.indexOf(' ')+1, line.length()));
     
    lineCount++;
}
in.close();

update += ";";

// it continues by creating a prepared SQL statement and setting
// the parameters using the stamps and lines ArrayLists...

Now, do you already see the issue? Don't worry if you don't - I've seen the same issue in various projects (including commercial projects) already, so you're not the only one who doesn't see it.

Let me give you a hint related to this specific code: One day the box hosting the Spring lobby server went down. So when it came back up I began to restart the lobby server software. First the lobby server itself. All was fine. Then ChanServ, the bot which contained the above code back then. I ran it's startup script, and waited..., and kept waiting...! The script didn't finish. At all.

Any clue yet?

No? Let me explain.

What - after some debugging - happened to be the case, was that this code in ChanServ was being invoked to import some rather large log files into the database. Now take a look at the code again: for every line in the log file, it appends a small string to update. This wouldn't be bad, if appending to a string in Java and .NET wasn't O(n), with n the length of the total string. In other words, the algorithm above as a whole was O(n²) instead of O(n) with respect to the number of lines in the file! (For example, this means that if importing 1000 lines takes 1 second, then importing 1 million lines takes, no, not half an hour, but a whopping 278 hour!)

It was trivial to rewrite the above code using StringBuilder. Afterwards, ChanServ started up and imported the log files into the database in a few minutes only.

StringBuilder update = new StringBuilder();

while ((line = in.readLine()) != null) {

    // ...

    if (lineCount == 0) {
        update.append( "INSERT INTO `" + name + "` (stamp, line) values (?, ?)" );
    } else {
        update.append( ","+ Misc.EOL + "(?, ?)" );
    }

  // ...
}

All in all, I want to urge you as C# or Java developer (or any other language with immutable strings) to keep in mind the algorithmic complexity of string concatenation. In simpler terms, I want you to use StringBuilder / StringBuffer whenever you are concatenating more then a few hundred characters into a string, to prevent nasty suprises like the one above.