It’s been a while. If you read this, come over to my new blog.

April 15, 2014 Leave a comment

Oh well. My last post here was about what I consider a first scope for Web Raid Mobile.

By now I’m at version 3.4.3.

For updates on Web Raid Mobile, please look here:


Categories: Uncategorized


February 4, 2011 Leave a comment

It’s been a while since I’ve been making progress with Web Raid Mobile. Seems I wanted too much of a change – configurable monsters instead of a long list of them, animation (which I’m just not able to do well)…

Since I’m playing “Dweller” a lot recently, I guess that even a pretty much unmodified Web Raid would be fun to people out there.

So I’m stripping away a few ideas for the first version and working on a simple first version. Other than the original web raid it will have

– missions
– doors and levers to open them
– a much better dungeon generator
– be “real time” instead of turn based
– have less clever monsters (the AI eats up a lot of processor time).

Categories: Uncategorized


October 21, 2010 Leave a comment

and… while we’re at it, how long are the other times?

I’m sure every developer dealing with Android’s sensors comes across this question: what value should I choose for the update frequencies. And why can’t I just “roughly in msec”?

Strangely, I couldn’t find any good results on the web, so here’s my own test results, done with a HTC Legend:

  • Accelerometer, SENSOR_DELAY_FASTEST: 18-20 msec
  • Accelerometer, SENSOR_DELAY_GAME: 37-39 msec
  • Accelerometer, SENSOR_DELAY_UI: 85-87 msec
  • Accelerometer, SENSOR_DELAY_NORMAL: 215-230 msec
  • Orientation Sensor, SENSOR_DELAY_FASTEST: 16-17 msec
  • Orientation Sensor, SENSOR_DELAY_GAME: 37-39 msec
  • Orientation Sensor, SENSOR_DELAY_UI: 77 msec
  • Orientation Sensor, SENSOR_DELAY_NORMAL: 224-225 msec

If you leave the phone unattended, the orientation sensor will report events only about half as often (but not never).

The test application is a simple sensor listener for the Acceleration sensor that keeps track of time passed and counts the number of invocations of onSensorChanged. Every 1000 (game/normal delay: 100) invocations it writes the time passed. I was measuring time with System.nanoTime, of course.

By the way, the light sensor didn’t want to ever return an event to me, even if the same sensor as a proximity sensor does. Strange.

Categories: Android, Rate, Sensor

Placing monsters randomly with good performance

July 28, 2010 Leave a comment

Now this is not as easy as it seems… Let’s suppose you have a dungeon of nxn squares, about 50% of which is solid walls, and you want to place a number of monsters in this dungeon, randomly… (actually not randomly in my game, but controlled by the letters of a sentence, i.e. the letter A corresponds to a low coordinate, Z to a high coordinate).

The straightforward case to do this is to get random coordinates, check, and repeat this until you find a good space.

This didn’t work out so well with a lowly Android phone / Dalvik VM combination.

What did work well was using a “SnailIterator”:

1. pick a random x/y

2. Move around this x/y, slowly away, like you’d be drawing a snail shell around it. Always check the coordinates and pick the first that fits.

This gave me a much better performance, particularly as the snail iterator also doesn’t construct a single object (thus forcing me not to implement the Iterator contract which would have forced me to e.g. return Dimension objects for next() and kill performance.

Here’s the code (beautifully un-intended by WordPress… I have to learn to change this…):

public class SnailIterator {

private int x;
private int y;

private int xsize;
private int ysize;

private int n;
private int step;
private int count;
private int dx;
private int dy;
private boolean finished;

* Creates a new snail iterator with a given start coordinate and a given size that must not
* be reached.
* @param x        the start x coordinate
* @param y        the start y coordinate
* @param xsize    the x size; all coordinates must be between 0 and xsize-1
* @param ysize    the y size; all coordinates must be between 0 and ysize-1
public SnailIterator(int x, int y, int xsize, int ysize) {
this.x = x;
this.y = y;
this.xsize = xsize;
this.ysize = ysize;

dx = 0;
dy = -1;
n = 1;
step = 0;
count = 0;

finished = x < 0 || x >= xsize || y < 0 || y >= ysize;

* Returns whether getX() and getY() will return valid values
public boolean hasValue() {
return !finished;

* Determines the next values for getX() and getY()
public void next() {
if (step >= n) {
step = 1;
if (count == 2) {
count = 0;
else {
x = x + dx;
y = y + dy;
finished = x < 0 || x >= xsize || y < 0 || y >= ysize;

private void turnRight() {
int xx = dx;
int yy = dy;

if (xx != 0)
dx = 0;
dx = -yy;

if (yy != 0)
dy = 0;
dy = xx;

* Returns the current x value
* @return
public int getX() {
return x;

* Returns the current y value
* @return
public int getY() {
return y;


Categories: Uncategorized

Android Development Tip #1: Don’t use Integer keys

July 25, 2010 Leave a comment

Maybe it’s useful to condense some of my learnings into tips. So here’s #1: Don’t use Integer keys. If you think that e.g. a HashMap is good, try to opt for an array of Anything. If your Integers can grow arbitrarily big, maybe create e.g. an 8-bit Hashkey from your integer, and create an array of ArrayList.

It seems that you should avoid very dynamic objects, like a temporary Integer key created to lookup a HashMap, at all costs. For me the performance of the level generation grew from many minutes to a few seconds after changing that.

Categories: Uncategorized Tags: , , ,

My first dungeon level on the Android emulator :-)

July 22, 2010 Leave a comment

An actual, working, freshly generated level (monster placement is visualized wrongly, some end up in walls...)

Categories: New Feature, Progress

Dungeon Generation works! (well, works a little bit)

July 21, 2010 Leave a comment

After a lot of debugging yesterday, my first test UI that allows to browse dungeon levels that have been generated starts to work.

I like the new way of constructing levels:

  • One single web page for a whole dungeon (this also eats up less web bandwidth compared to what Web Raid is doing)
  • One sentence defines the level tree; some letters mean “this level has one child”, others “this level has two children”, others “this is a leaf level”. This results in much nicer level structures.
  • Monsters and items are placed in these levels in a way that fits their levels: easy beasts in Leve l 1, tough ones in the lowest levels.

It all still takes a bit long (generating a full dungeon takes a minute), but we’ll see about that…

Categories: New Feature, Progress