Unibbl

Unity networking

Unity provides a set of APIs to help with networking. At least they seem to provide a solution that would work for short multiplayer matches

Marketing pitch (or high level info, if you'd call it that)

Network Reference Guide

If all you needed is a few GET/POST requests a Q&A here will put you on the right track (maybe also look here)

Photon

I guess everybody who played with Unity for a bit have heard of Photon. Service looks like it's got a free-something with cloud storage so you don't need to run your own server.

Analytics

It looks like we can integrate Flurry and TapStream (free version of ). Flurry doesn't look very easy to integrate but there are free tools to help with this.

Miscellani

There is free stuff lying around which may not scale up to a big multiplayer game or intensive analytics, but can help with testing depending on what kind of multiplayer features you want.

000webhost.com - free web hosting. As far as I could judge, without ads.

freesqldatabase

I was looking for a way to:

Find out which object in the 3D world is "under a mouse click"

Find the coordinates of a mouse hit in the 3D world.

Physics.Raycast can help you here. It is well documented too. But there's a new system shipping with Unity 4.6 - and you probably want to use this instead.

I did some testing and fixes as players reported that the player character was "sliding".

The coloured balls show input magnitude:

red above 1.0

white under 1.0

size of ball is 0.1 x input magnitude.

With the Input class Unity performs input interpolation, which means that a key press doesn't set input to 0 or 1.

I found the reason for the "sliding" effect: after releasing the keyboard, input interpolation causes the character to move a little further.

I then played a little "game" to test competing input schemes: walking from right to left, trying to come as close as possible (but not drop into) the edge of a shallow depression in the terrain.

In the topmost image input is normalised (original method) which increases the undesirable effect. As you can see I failed to stop on the edge every single time.

In the middle image input is used 'as is'. The stopping distance (in white) is something like 50 cm. A lot but I can get pretty close to the edge without falling.

Duplicating a level to create another, similar level sharing the same setup is intuitive but doesn't work very well if we later add functionality that we want shared across levels.

A solution is to create a so called "setup prefab" to factorise common functionality not fully expressed in code.

Caveats:

If our setup prefab has object references, these will break.

A couple of pics showing how we're keeping buildings out of the main level file.

Reasons we don't want buildings in the main level file:

How Blender manages hierarchies. Notably, it doesn't allow name collisions and has a bit of a quirky approach to hierarchic groupings.

Don't want to collapse modifier stacks until late in the project life cycle.

Triangle count. To keep import speedy it's better to break everything down into several files.

So, building dummies in the main level file are replaced by footprints (top). Sometimes these marks are also used to level the ground under buildings using skinwrap.

Bottom pic shows how boolean ops are used. Although booleans are applied in blender since Unity imports everything cutout shapes end up in the editor. Not clean but not a big deal, just uncheck the box next to MeshRenderer in properties. Likewise with footprints.

A useful idiom when you play the same animation over and over.

I was wondering why every other call to my 'attack' animation failed. I used a regular CrossFade() and didn't have much reason to suspect a conflict with another script...

So, we can play a whole animation by calling CrossFade(ANIM) at every game frame, right? This actually has no incidence (except the redundant calls are often wasteful) ... until the animation has finished playing.

So, we need something to make an animation repeat before it has ended. Unity has Rewind().

Gist:

Tip: re-position CharacterController on KO

Here's a tip that's rather patchy but may help if you don't want to destroy your NPCs on KO.

The problem that this video illustrates is how a KO animation tends to move actual geometry away from the character controller. Later when other agents try to walk past the KO'd agent, they encounter an invisible obstacle.

A simple, drastic solution is to delete the game object. An equally simple, less drastic solution is to disable the character controller.

But what if you still want some collision to happen and don't want to fuss with rag-dolls or the bone hierarchy? Something that's not too complex, but will look 80% good?

Well, you can move the centre of CharacterController. Here I also reduce the radius so the KO'd NPC can be easily walked over.

It may be a good idea to animate the modification, notably if the scene is busy with other NPCs nearby.

What happens when your javascript (aka UnityScript) class name clashes with the name of a UnityEngine class?

Well, not much until you try to extend from your own custom class. Somehow the UnityEngine namespace gets imported implicitly, so you'll end up getting this error:

Cannot extend final type UnityEngine.Xxx

Where Xxx of course is the name of your custom class.

Game AI: CD.01