Skinning on its way.

Gbleh. So, before I can release it, the last thing I need to do is to get the board itself (the thing with the pieces on) resizing according to the skin graphic sizes, the sizes specified by the skinner, and dynamically according to how much space there is in the window.

Sounds easy enough, right? Right.

Except, there's a LOT of code that relies on the specific size of the specific skin I used. This is an example:

  public int getBoardSquareX(int mouseX, int mouseY) {
    int result;
    if (viewingDiagonally) {
      // for /...    Offset is board dependent.
      squareWidth = boardWidth / 10;
      squareHeight = boardHeight / 10;
      result = (int)(((mouseY - BOARD_SKIN_VPAD)
       - mouseX*(squareHeight/squareWidth))
       / (-squareHeight) + 8.5);
    }
    else {
      squareWidth = boardWidth / 15;
      // +1 as int rounds down.
      result = (int)(mouseX/squareWidth) + 1;
    }
    return result;
  }

Doesn't look too scary, until you look at BOARD_SKIN_VPAD, which is some padding I added to the top so that the dwarfs heads would draw - in the skinnable version, you can add padding to ANY edge, so that's four times more work. And that "8.5" (it's "-1.5" in getBoardSquareY) is completely unexplained, and arbitrary - I know it was experimentally obtained. And there are 20 functions like this throughout the Board object, with these unexplained equations with their unexplained terms in... what was I thinking? Well, most likely I was thinking "I have one night to write this in ready for the demo, so I can cut a few corners"... and now it's come to bite me in the bum.

Frankly, I feel there must be an easier, more portable way to convert between screen coordinates, board coordinates, and server coordinates (like board coordinates, but sometimes rotated from how they appear on the screen).

Maybe make each square a lightweight component, so I know where the mouse is at any time, if the event handlers don't get confused and trigger in the wrong order - which they probably would and I'd still need to track where put PUT the components. In later versions of Java, you could use imagemaps. That would be interesting, but not an option for me now.

So it looks like I did comment some of it.

Here's one that just made me chuckle, though I suspect in the cold light of day it'll be less amusing..

// Viewing diagonally, rotated 45deg antiCW from
// 1,1 at topleft, so 1,1 at left now.
// Don't need to take octagonalness into account
// as we're measuring relative to board centre.
// halfway down the screen - distance from board
// centre * halfsquares + padding
result = (int)(BOARD_SKIN_HEIGHT/2
 - (squareX - squareY) * Sh/2 + BOARD_SKIN_VPAD);

this is all just a bit too hectic for me,but i can see where all your time is going,so thanks a lot.

It edges closer

Over the last three days I procrastinated for a day (went to watch a movie, did mother's day shopping), programmed for a day, and farted about with a Java Profiler for a day.

In the process, I fixed the following glitches and show-stoppers:
x Diagonal drawing of the guidelines is mucked up.
x Vertical centring is ignored when drawing diagonally.
x Board sizes are ignored
X Components aren't transparent
X Sounds do not play on board
X Out of memory.

[Incidentally, in the following, KiB and MiB are megabytes and kilobytes respectively: it's the newer, non-ambiguous notation]

That last one was why I downloaded and installed a profiler. Turns out that although Sun claims it's the bee's knees when it comes to threading, Java actually can't cope with much of it. See, even if it doesn't use all of it, each thread needs an allocation of 528KiB (400 for Java code, 128 for native code).

I was loading each image and each sound in a separate thread, since that's the sensible way to do it, if you want to update the display once any of the images finishes loading. You can't, otherwise, wait until just one of the images loads: you have to wait until they're all done. Well, you can (and I do) but it involved jumping through hoops, and it's not at all elegant. But, if each image has its own thread, it prettymuch all handles itself, very elegantly.

But, there're 10 sounds and 27 graphics in the default skin. That meant that just to download them (since they all started downloading at the same time) it needed about 37*528KB, or about 19.5MiB. There are also about a further seven threads (it will vary depending on the JVM). Some are created by me, some by the JVM. So just in thread stack allocation, we're using 23MiB.

The default JVM size is only 16Mb, which is way too small. (all these sizes are probably bigger in the later versions, but that just means the problem is scaled up).

So, instead I had to write a version that loaded all the images through a single thread, and not use a thread for the sounds at all.

At just 8 threads, we're only using about 4MiB, which is saner. Running it in the profiler, though, I got it to allocate up to 100MiB at a time, between garbage collections. I don't know if this is going to be a problem in real life though. I can't see how it is allocating this much memory since even the profiler thinks that there's only 4MiB of data being stored.

Anyway, I'm putting memory optimization to one side, and am concentrating on what I hope are the final issues before releasing it:

Significant but easy:
Default piece graphics are tiny [fixed since posting]
Pieces don't resize.

Significant and hard:
Images don't display immediately once loaded.

Minor, may not fix yet:
On resizing the window, the background resizes slowly.
Textarea text is not selectable/copyable.
Default image backgrounds aren't transparent.

Solved:
x Diagonal drawing of the guidelines is mucked up.
x Vertical centring is ignored when drawing diagonally.
x Board sizes are ignored - why?
X components aren't transparent
X Out of memory, most likely due to too many threads.
X Sounds do not play on board
X Default piece graphics are tiddly.
X Board doesn't resize with the window.
X Board squares don't resize with the window.
X Board layout doesn't resize unti rotated.
X Board layout doesn't use padding until rotated.
X Mouse to BoardPos highlight/selection doesn't use padding.
X Piece redraw rectangle is misaligned.
X Pieces don't resize.
X Pieces are misaligned in the square at extreme sizes
X Pieces aren't drawn within the insets area.
x Diagonal guidelines break when horiz padding=0
X Insets should embed itself in any existing padding.
X Insets aren't respected or used.
X Piece alignment breaks when horiz padding=0
X Diagonal piece layout applies topPadding twice.
X Pieces don't resize properly when using the minimum size.

Significant and hard
? Images don't display immediately once loaded.

Minor
On resizing the window, it takes time for the background to reappear.
Textarea text is not selectable/copyable.
Gridelines don't stop at the board edges.
Padding should probably scale as a percentage of prefsize, rather than being absolute.

Solved:
X Images don't display immediately once loaded.
X Guidelines shouldn't display once board images have loaded.
X Main chat should have a background - gives "offscreen null"

No critical bugs left.

I'm now within spitting distance of getting the next version public. I'm at the stage of uploading it to the server and seeing how well it works.

And I find that one rather large problem stands in my way. OK, it's another critical bug.

See, the java thingy is looking in the current html path for the skin files. And the current path for the java file is www.thudgame.com/node - which doesn't exist, it's a virtual directory peopled from the database by the content management system.

So, I'll have to force it to look in another folder. Probably /files or /skins or somesuch.

Argh, why is nothing ever simple? Still at least java's URL object should let me do that fairly simply.

Hrm, "fairly" was clearly overoptimistic. To change the path of a URL, youcan't just do myURL.setPath("new/path"); because that would be too easy.

Instead you have to do:

String myPath = "new/path";
String tmpString = myURL.getFile();
int oldFileIndex = tmpString.lastindexOf("/");
if (oldFileIndex != -1) {
  myPath += tmpString.substring(oldFileIndex);
}
myURL.set(
  myURL.getProtocol(),
  myURL.getHost(),
  myURL.getPort(),
  myPath,
  myURL.getRef()
);

Now, sure, you could do that in one line, but it would be pug ugly.

Mind you the above is pretty damn ugly too. Geesh, sometimes I despise this language.

[Edit: and of course, the above wouldn't work anyway, unless you can *guarantee* that the last thing on the url is a filename that you want to keep. If it's a directory or something, then it should be removed. So, the above is really an over-simplification.]

[Edit2: ...oh, no, that function is protected, you can't use it after all. In fact, you can't change URLs at all, you have to create a new one, like so:

String myPath = "new/path";
String tmpString = myURL.getFile();
int oldFileIndex = tmpString.lastindexOf("/");
if (oldFileIndex != -1) {
  myPath += tmpString.substring(oldFileIndex);
}
try {
  myURL = new URL(
    myURL.getProtocol(),
    myURL.getHost(),
    myURL.getPort(),
    myPath,
  );
}
catch (java.net.MalformedURLException e) {
  System.out.println("Bad URL!");
}

Isn't Java great?]

Hi! Who are you? :)You're reading my dev blog.

I like people who do that

So as a li'l present, here's a link to the new client, uploaded and maybe-working, but not tested because I want to sleep.

Please let me know if it works

it worked,but i didn't have anyone to play against,so can't tell if the actual board ran well.not many people look at this forum,maybe you should set up a time soon when we can try it out,so all of us know when to be on and we can all test it out for you

Hrm. Yeah, that's better than my original plan, which was to just test it a *little* bit, then put it live, then when people said "hai it no wurkz" I'd shout "hahalol sux2bu!" and run away.

I guess I should do some cunning log analysis and find which part of the day I'm prettymuch guaranteed people online, and ask people to test it then!

Ah, but our server logging doesn't have timestamps. *rewrites logging to be more modular and cool* OK, now it does.

[Edit: Incidentally, this is why I hate Java but love PHP. PHP makes stuff like that dreamy-simple. I want to rewrite all my logging stuff from scratch, changing 300 lines of code? No problem, ten minutes' work, works first time, no knock-on effects.]

[Aside: should that be "ten minutes' work", "ten minute's work" or "ten minutes work"? I feel that it's a plural possessive (like: "the work of ten minutes"), so should be the former: but Google disagrees with me]

As far as your grammatical question goes, I'd definately agree with the "ten minutes' work" - google knows nothing

MS

ten minutes' work???
hmmm,ten minutes work
ten minutes' work
hmmmmmmmmm
it would seem to be possessive,wouldn't it.
although on the other hand can minutes possess something?

Well, my sister reckons we're all right too, and she's my own personal Ultimate English Language Authority on account of being a poet and knowing what rules to break when.

She was on the front page of a national newspaper the other day, you know!

(The newspaper is called "Mental Health Wales". Yes, that's the honest truth, and sums up my family quite well.)

Think I'll announce that new version in the announcements forum.

congrats to her,and am glad we have such an expert to lead us on our grammatical quests

i am learning java at uni to! i have been learnin for 8 weeks now! and what can i do?

sod all.

i cant remember any of it

my ears bleed everytime i hear the "j" word

even when it is said with regards to coffee???