Generalities On Entities
Introduction
In this tutorial I will try to sum up a lot of the maint things a MOD programmer for the Source engine should know about making entities. This is stuff I have learned through experience and that is not usually explained (or explained well) in existing Valve Docs. There's a few bits and pieces of information that will save you time when trying to create entities, but they are not really all related to the same thing.
Some of the things I say here may be dead wrong, in wich case you'll be nice to notify me in the discussion page so I can edit this here. I do not consider myself an expert or an authority, I'm merely trying to share some tips to help you understand a bit faster what source entities are all about.
You will not make anything with this tutorial, no weapons, or flashy pieces of eye candy. I will also assume you are fluent in C++ (meaning that if you just read a book on C++, you should go out and practice your skills first before reading this).
I also will not cover creating Weapon entities, new Player entities, or NPCs, as those are already covered better and in more details in other tutorials/posts.
Ok, here we go!
Naming Conventions
Before anything else, I must emphasize a point wich is naming conventions. In the Source SDK, all class names begin with a capital C. Furthermore, on the client, the names begin with C_ to differentiate them from their server counterparts. So you would have: Server: CMyEntity Client: C_MyEntity You should follow that naming convention, and also look into hungarian notation if you haven't done so already, learn it, and use it, it sames you a LOT of time, because just by looking at the name of a variable, you know it's type and wether it belongs to a class or not. It saves you the 5-30 seconds that MSVC 2003 takes to find the definition of the variable when you use the right click menu and select "Go to Definition". And believe me, those few seconds are precious time for a programmer. You should also do well to use tabs. If you are lazy (like I sometimes am), there is a shortcut in MSVC 2003 to tabulize a certain selection of text. It should be in Edit->Advanced Options->Tabulize or some similar name. Learn the shortcut, and use it a lot, it makes things a lot easier to read, and also allows you to spot missing brackets, if you have any.
What Base Entity is right for me?
Ah, this is a question wich I have asked myself many times when creating new entities. This is usually my starting point when trying to make something new: You look around in the different existing entities and try to pick the best one to use as a Base Class. There's many possibilities offered to you, the simplest one being CBaseEntity of course, because every entity is derived from it anyways, and it is the mother of all entities anyways.
CBaseEntity: CBaseEntity, as said above, is the base for every other class out there. I rarely use it as a base though, I have done it sometimes, but usually you want to use something a little more evolved. You should base off CBaseEntity when you need someting really basic, that doesn't use a Model, and that is usually going to be Static. An example for that would be say... A Spawning entity, wich you only need to know the position and team number of.
CBaseAnimating: Ah, this one already offers many possibilities. First of all, this entity has a model. This might sound a little obvious but not all entities have Models. I think this is the right time to point out something important for later: ENTITIES WITHOUT MODELS DO NOT GET SENT OVER THE NETWORK BY DEFAULT. You might even have the corresponding Client entity for it all set up, and have network tables all done, fine and dandy, but it will NOT transmit ANYTHING if it doesn't have a model. The simple workaround is to add the Entity flag EFL_FORCE_CHECK_TRANSMIT using the AddEFlags() function in that entitiy's Spawn(). That will make sure it does transmit it's information to all clients normally.
Back to CBaseAnimating: CBaseAnimating offers many interesting functions and possibilities.
SetModel(): SetModel is simply a function to set a certain entity's model! You need to use the Precache() function to PreCache (read load into memory) the model beforehand. Precaching is usually done in the Precache() function, wich gets called up when the entity is first loaded. You also need to use the PRECACHE_REGISTER macro to tell the engine to Precache your entity (we'll come back to that later).
SetSequence()/LookupSequence():At some point I'm sure, you're going to want your model to animate! Well these functions do just that. SetSequence() takes the number of the animation. It starts (I believe) at 0, -1 being the invalid Sequence number, and the order is determined by how you compiled the model. The list in Model Viewer is an easy way to find out the number for the animation sequence you want to use. You can also use LookupSequence() wich takes the name of the sequence as a parameter and returns the corresponding number, or -1 if it did not find any animation with the correct name.
CBaseTrigger etc...: We now enter the domain of map entities (entities primarly used by mappers.) There's a couple different types of those, but I have not had the chance yet to explore them all. There is one important thing to note about map entities though, and that is the following function:
KeyValue():This is a function defined in CBaseEntity, wich you must override if you want to use it. This function is called up by the engine during map load, when map entities are created. When a mapper makes an entity in Hammer, he fills in a certain number of values (name, angles, parent, etc...). Those are called KeyValues, and go in pairs of name/value. You don't really need to "register" these pairs anywhere. All you need to do is add your own if statement in your KeyValue function to check if the entity has your custom KeyValue set, and do what you want with the value. Then you'll need to set up your FGD file for your mappers, so they can fill the value. I won't dwell on this here, Valve has a few examples of setting up FGD files in their own Entities tutorials, and it's pretty easy to set it up by just looking at how existing entities are done.
Right. Now that you have the class set up, you might want it to DO something:
Think Functions
Think functions could be defined as Functions that dictate the behaviour of an entity. They quite literally THINK for the function, and make it change states, and so on.
SetThink() and SetNextThink():This assumes you know what a function pointer is. If you don't, go refresh your memory and come back when you have. In setting up your entity, you will want for it to change, maybe change animations, or do something to players, or other entities... There's a host of things an entity can do, and i leave that up to you. You can have many think functions, but here's another one of those hard-learned tips: THINK FUNCTIONS DO NOT EXIST ON THE CLIENT SIDE. At least not in the same manner that they do on the server. Ok, let's get an example so you'll understand it more clearly: I want to have an entity that does two things: When there is no player around, it will play it's idle animation, and do nothing more. When a player comes into range of it, it will start to emit puffs of smoke. So I will make two think functions: IdleThink and SmokeThink. These should return void and have no argument list like so:
void IdleThink(void); void SmokeThink(void);
When the entity spawns, I will do this:
SetThink( &CMyEntity::IdleThink ); SetNextThink( gpGlobals->curtime + 0.5 );
What this does is tell the Engine that this entity will run the IdleThink functions in .5 seconds. Now in IdleThink, we will look for a player near us (Use a SpereQuery for example), if we find one, we will do:
SetThink( &CMyEntity::SmokeThink ); SetNextThink(gpGlobals->curtime);
If you tell SetNextThink that the next think time is NOW, the function will get ran the next server frame. If we do not find a player near us, it is important that we still call SetNextThink:
SetNextThink(gpGlobals->curtime);
Otherwise, things will get messed up after that. I am not sure of the exact effect but i do believe that your entity will STOP THINKING if you do not set a new NextThink time.
Right, this is all server side. Do not forget to use the DEFINE_THINKFUNC macro for all your think functions (we'll come back to that later as well).
Now for the Client Side: As I said before, the client does not have the same System for entities to think. You cannot define multiple think functions here, as SetThink will not work properly on this side. Instead, you must use the following:
ClientThink()/SetNextClientThink():These are the equivalent of the server system, except you only have one function at your disposal here (wich sucks a bit for organisation, but oh well). This will get called based on the SetNextClientThink() time you set in Spawn, and in subsequent calls to ClientThink().
Simulate():Here's the Commentary in c_baseentity.cpp: Once-per-frame stuff should go in Simulate(). That really sums it up. I personally use Simulate() mainly for Particle effects, or graphical effects of any sorts. Beware of what you run in Simulate(), as said before, this gets run EVERY FRAME, meaning you should avoid creating 50 new particules every Frame ;)
That's about it I believe. Think functions are an essential part of any entity and you should learn to use them well.
Macros
I want to dedicate an entire chapter to these Macros, because they are what causes problems with entities at least 60% of the time. Badly set up macros will give you errors messages, crash your game, give you erratic behaviour, and so forth. I'm going to try to sum up most of them and a few more tips as best i can:
LINK_ENTITY_TO_CLASS():This is your main macro. You do not have an entity until you call this macro. This macro will NAME your entity for the game. Anytime you call GetClassname() on an entity, this is the name that will be returned on the server side. There is a subtility however, wich is that , on the client side, the entity name will not be the one you give to LINK_ENTITY_TO_CLASS unless you make a DATADESC. Instead, it returns (client-side) something like "class C_MyEntity".
PRECACHE_REGISTER():This is used to tell the engine that it needs to precache your entity when it loads. You pass it the name of your entity (the same you passed to LINK_ENTITY_TO_CLASS() ).
DECLARE_CLASS():You need to put this in your Class definition, in a public section. This gives you access to the BaseClass Macro, wich is very handy. I believe it also helps to set up some networking stuff.
DECLARE_DATADESC()/BEGIN_DATADESC()/END_DATADESC()/DEFINE_XXX(): This set of macros is used for two main things: First, you need it if you're going to have thinkfunctions. Second, I believe it takes care of saving any variable that is defined in the set for Single Player Games, and restoring their values when you load a saved game. You'll want to go look around at examples to see how all of these macros are used, but they are pretty straightforward.
Networking Entities: I was going to explain a little bit about networking entities and networked variables and such, but there is already a lot of nice tutorials by Valve that explain it all in detail. See the Networking Section in this Wiki's Programming Section.
In Conclusion
I have only brushed the surface here. I have been working with this SDK for over 8 months now, and I learn something new about it everytime I open it. It is complex, it is vast, and you will not learn all about it by just reading tutorials. The best way to do it has always been for me to JUST DO IT. You take an idea, something you want to add to your MOD. You look around source with a few keywords to see what's already been done. You search on a few forums and see if it's already been discussed. You look at tutorial sites and see if anything relates to it. Once you have a good idea on how you're gonna do it, well you just dive in the deep end and start coding it!
I hope you've learned something from this tutorial. Once again, if I mislead you in anyway, it is not intentional. This is all of what I have learned through months of trial and error and I hope it speeds up your progress in learning more about this SDK.
Good Luck. Imperio59