天行健 君子當自強而不息

Storing Player Information

Players in the game are only allowed to move around and swing their weapons (hitting
other players). The server will want to track every player’s current state (walking,
standing still, swinging their weapons, or being hurt), the coordinates in the world,
the direction they are facing, and the speed they are walking (if they are walking).

This player data is stored inside a structure called sPlayer. Because all connected
players in the game need their own set of unique data, an array of sPlayer structures
are allocated to store the information. Both the number of player structures
to allocate and the number of players to allow to join the game session are stored
in the macro MAX_PLAYERS, which is currently set to 8.

The sPlayer structure is as follows (with supporting state definition macros):

The variables in sPlayer are self-explanatory, except for latency. Remember that
latency is the delay resulting from network transmission. By storing the time it takes
for a message to go from the server to the client (and vice versa), time-based calculations
become more synchronized between the server and client.

Speaking of time-based calculations, that's the purpose of the last_update_time variable. Whenever
the server updates all players, it needs to know the time that has elapsed between
updates. Every time a player state is changed (from the client), the last_update_time variable is set
to the current time (minus the latency time).

Time is also used to control actions. If a player swings a weapon, the server refuses
to accept further state changes from the client until the swing weapon state is
cleared. How does the state clear? After a set amount of time, that’s how! After one
second passes, the update player cycle clears the player’s state back to idle, allowing
the client to begin sending new state-change messages.

On the subject of sending messages, take a look at how the server deals with the
incoming network messages.

Handling Messages

You’ve already seen DirectPlay messages in action, but now you focus on the game
action messages (state changes). Because DirectPlay has only three functions of
interest when handling incoming network messages (create_player, destroy_player, and
receive), the server will need to convert the incoming networking message to messages
more suited to game-play.

The server receives messages from clients via the DirectPlay network’s receive function.
Those messages are stored in the pReceiveData buffer contained within the
DPNMSG_RECEIVE structure passed to the receive function. That buffer is cast into a
more usable game message, which is stuffed into the game message queue.

The server game code doesn’t deal directly with network messages. Those are handled
by a small subset of functions that take the incoming messages and convert
them into game messages (which are entered into the message queue). The server
game code works with those game messages.

Because there can be many different types of game messages, a generic message
container structure is needed. Each message starts with a header that stores
the type of message, the total size of the message data (in bytes) including the
header, and a DirectPlay player identification number that is usually set to the
player sending the message.

I’ve taken the liberty of separating the header into another structure, making
it possible to reuse the header in every game message:

#define MESSAGES_PER_FRAME 64 // number of messages to process per frame
#define MAX_MESSAGES 1024 // number of message to allocate for message queue

typedef struct sMsgHeader
{
  long type; // type of message (MSG_*)
  long size; // size of data to send
  DPNID player_id; // player performing action
} *sMsgHeaderPtr;

Because there can be many different game messages, you first need a generic message
container capable of holding all the different game messages. This generic
message container is a structure as follows:

typedef struct sMsg // the message queue message structure
{
  sMsgHeader header;
  char data[512];
} *sMsgPtr;

Pretty basic, isn’t it? The sMsg structure needs to contain only the message header
and an array of chars used to store the specific message data. To use a specific message,
you can cast the sMsg structure into another structure to access the data.

For example, here is a structure that represents a state-change message:

typedef struct sStateChangeMsg
{
  sMsgHeader header;

  long state; // State message (STATE_*)
  float x_pos, y_pos, z_pos;
  float direction;
  float speed;
  long latency;
} *sStateChangeMsgPtr;
 

To cast the sMsg structure that contains a state-change message into a usable
sStateChangeMsg structure, you can use this code bit:

sMsg Msg; // Assuming message contains data
sStateChangeMsg *scm = (sStateChangeMsg*) Msg;

// Access state-change message data
scm->state = STATE_IDLE;
scm->direction = 1.57f;

In addition to the state-change message, the following message structures are used
in the network game:

typedef struct sCreatePlayerMsg
{
  sMsgHeader header;
  float x_pos, y_pos, z_pos; // create player coordinates
  float direction;
} *sCreatePlayerMsgPtr;

typedef struct sRequestPlayerInfoMsg
{
  sMsgHeader header;
  DPNID request_player_id; // which player to request
} *sRequestPlayerInfoMsgPtr;

typedef struct sDestroyPlayerMsg
{
  sMsgHeader header;
} *sDestroyPlayerMsgPtr;

Each message also has a related macro that both the server and client use. Those
message macros are the values store in the sMsgHeader::type variable. Those message
type macros are as follows:

#define MSG_CREATE_PLAYER    1
#define MSG_SEND_PLAYER_INFO 2
#define MSG_DESTROY_PLAYER   3
#define MSG_STATE_CHANGE     4

You see each message in action in the sections “Processing Game Messages” and
“Working with Game Clients,” later in this chapter, but for now, check out how the
server maintains these game-related messages.