Category: Articles

The enemy HP values have been removed from above the enemies on the game screen, and a damage value that floats over the enemy when hit has been added.  I created a new class for overlay text, which is used by both the 2D and 3D game screen.  This prevents the code from being duplicated for things like the status text and enemy damage.  I added a new property to the enemy that is the dying state and counter.  This was needed because the damage value does not display if the enemy is not alive.  This kept the last damage number from displaying.  Now I’ve changed it so that the damage value displays if the enemy is alive or dying, so the final damage number now correctly displays.  The dying counter will also be needed to display a death/blowup animation.  On the 2D game display, I added logic so that the enemy just fades away using an alpha color value based on the percentage of the dying animation passed.

I created a simple graphic for the projectile.  I mapped this graphic to a simple 2D plane in Blender.  It’s important to remember that all points of the Blender model must have a Z coordinate equal or less than 0 (non-positive) to appear.  This may be backwards for my game, since I am rendering the 3D world with the Y axis facing downward, which makes translating from the 2D to 3D coordinates simpler.  Remember, the model coordinates are relative to the small orange dot, which is not necessarily the location where the axes cross in Blender.  It is a good idea to put the orange dot at coordinate 0,0,0 in Blender to keep from getting confused.  Therefore, all vertices must be moved in edit mode (not object mode) for the changes to appear in the game.  Similarly, scaling an object in Blender’s object mode will have not effect on the model in the game.  The model must be scaled in edit mode with all vertices selected for the size change to appear in the game.

Also, I converted the List of loaded Model objects to a Dictionary object.  I thought that a Dictionary was the same as a hash table, but Chris Gardner (@freestylecoder) explained to me that Dictionaries don’t allow multiple values to be assigned to a key like a hash table.  He said that Dictionaries should be used like symbolic links, which is what I am doing with my constants which represent each model in the game.  The previous method I was using was kludgy and bad style, which was adding the Model to the List and setting the constant ID of that object to the order number which it was added to the list.  That process was a hassle to maintain and a recipe for disaster later.

Not much coding was done over the past few days, since I have been working on rebuilding a PC.  I also upgraded my main PC to Windows 8, which required a complete wipe of my primary drive, which means I had to install Visual C Sharp 2010 Express, XNA Game Studio, Blender, Gimp, TortoiseHg/Mecurial, and other programs again.  One issue with installing GameStudio in Windows 8 is that the Games for Marketplace Client must be installed for the Game Studio installer to succeed.

Spent some time cleaning up the project directory.  I got rid of all references to the old game name, so now the name in the repository is now BlastingBits.  Just changing the name required all files in my project directory to be recommited and uploaded to the repository.  Unfortunately, due to the directory structure, my repository had four levels of BlastingBits folders (BlastingBits/BlastingBits/BlastingBits/BlastingBits).  Using the TortoiseHg rename option, I changed it to one level of BlastingBits and the rest in an “src” directory.  Again, the repository saw the moved files as files being deleted and added, so it required a full upload of my entire project directory again.  Not sure if there is any way around that, since I used the TortoiseHg rename option.

After the cleanup, I reimported all of the source files and assets.  The project was still trying to create an executable with the old name, so it was easier to just recreate the project and import the sources.

One issue that I was able to resolve was the warning about the textures associated with the models.  This is the “Asset was built two times with different settings” warning when running the game.  I figured out that this warning was due to the image being compiled for the model and then again as a regular texture.  I was able to get this warning to go away by setting the “Build Action” option to “None” for the textures used by the models.  This apparently prevents these images from being used as regular sprites in the game.

I’ve been working on updating the second armor set.  I added some additional changes to the helmet, and I started working on the body piece.  One lesson I learned the hard way is to model both the front and side at the same time.  I only modeled the front at first, so now it is really difficult to model the sides since there are so many vertices now.

Using some of the graphics that I have already created, I made a promotional graphic that I used on my main homepage.  This looks much better than the in-game screenshot that I had before.  I had to render the helmet again in Blender to add the circle plane below with the helmet’s shadow.  One important tip that I learned is that shadow in Blender are only displayed when using a “Spot” light source.  This same graphic and background were also used to update the theme for this development blog.

I was able to get my models to display on the 3D game screen, and I was able to make it move around using the thumbstick for X/Y axis and the LT/RT triggers for Z axis.  Using an orthographic projection, I was able to make the model appear as a static image.  For now, my screen is mapped to 24×16 game world coordinates, since my player model is only two Blender units tall.  This will be really helpful when converting the game world coordinates to the 3D space, because I should be able to make a 1-to-1 correlation between the 2D bounding boxes and the rendered 3D objects.  Unfortunately, my model still displayed on its side, with the positive Z axis pointing out of the screen.  I spent quite a while playing with the FBX export settings, but I was unable to get it to display correctly.  Then I finally just rotated the model manually in Blender and exported, and still got the same result with the model on its side.  After multiple rebuilds and exports, I finally figured out that the object’s vertices must be rotated in edit mode for the rotation to take effect.  By doing this, I was able to get the player model to display correctly.  It is also important to note that all of the object’s coordinates are in relation to the orange anchor point in Blender, so it is best to move that anchor point to the origin first.  I think it may be better to go ahead and programmatically rotate the FBX object, just in case I get models from another source.  Other models will still have the same problem, and I don’t want to manually rotate the vertices of every model I use.  I also noticed that XNA does matrix operations in order when multiplying.  When rotating an object, it is important to apply the rotation matrix first before translating it.  I haven’t found an equivalent to pushing/popping matrices as in OpenGL.  It is also important to have the model in Blender centered on the rotation point.  This will prevent having to programmatically center it first.

After I got the blocks of the game world displaying correctly, I copied in all of the controller handling code from my existing GameScreen class.  All of the display code from the game objects such as enemies, collectibles, and blocks is the same, except that I have to divide all of the position location coordinates by the tile size (48 pixels) to covert the object’s location from screen coordinates to world coordinates.  The screen coordinates are 1280×720 and the world coordinates are 24×16.  I think the world coordinates are much easier to deal with, since one Blender unit is exactly one unit in the world space.  The only problem is that my collision boxes are Rectangle objects which are handled in screen coordinates.  I can’t convert these Rectangles to world coordinates because the Rectangle only accepts integers and doesn’t handle floats.  I could write my own collision box class and methods that accept floats, but that would be more work to do with no real benefit.  There may be a slight performance improvement, but I don’t think it would be noticeable.  On the other hand, it may be less efficient to do float based collision detection, since floating point operations are generally more expensive.

Using the models that I generated for the body, hands, boots, and helmet, I went ahead and updated the model on the equipment screen.  Currently, this is just a static image that I rendered by importing all of the armor pieces in Blender.  These pieces complete most of the player’s body, but I will probably need another player model to connect the pieces of equipment.  I’m hoping I can use the bone/armature information from the player to set the armor pieces in real time.  That way a player can have any combination of armor pieces equipped and have it display in real time on their game character.

One other change I made was to pull in all of the “model” (as in MVC) into its own class.  Since the World class already contained the data for the game objects, I used that class.  This just required me to add the Player class into the world class.  The reason for doing this was to have all of the non-graphical game information in one location.  This allows me to pass the “model” to the two different views that I have created, which are the 2D sprite view and the 3D view using the imported FBX models.  Ideally, I should be able to switch between the 2D view and 3D view instantaneously while playing the game, since they are both using the same data model.  I added an interface called GameWorldHandler that just has methods for handling the World object.  This allows me do determine if I should pass the World object to the current screen on each update.  The two game screens and the equipment screen implement this interface, since those require the game World (and the Player object) to display.  However, screens like the title screen and credits screen are not GameWorldHandlers, so those do not implement that interface.  I could have used an abstract class, but those screens already inherit from my ResistorKit Screen class.

Finally, I have started working on modeling the second armor set.  Currently, I just have the helmet done.  I may want to round out the edges some more, but I don’t want to add too may polygon faces because it could possibly impact performance.  Plus, the details of the model won’t be noticeable since the player is so small.  However, the details may be noticeable on the equipment screen where I plan to display the model in full size.