Next chapter
Previous chapter

2.1   Review of Basic Driver/Mudlib Interaction

1. When an object is first loaded into memory, the driver will call create() in native muds and reset() in compat muds.   A creator uses create() or reset() to give initial values to the object.
2. At an interval setup by the game administrator, the driver calls the function reset().   This allows the object to regenerate monsters and such.   Notice that in a compat mud, the same function is used to set up initial values as is used to reset the room.
3. Any time a living object comes in contact with an object of any sort, the driver calls init() in the newly encountered object.   This allows newly encountered objects to give living objects commands to execute through the add_action() efun, as well as perform other actions which should happen whenever a living thing encounters a given object.
4. The driver defines a set of functions known as efuns which are available to all objects in the game.   Examples of commonly used efuns are: this_player(), this_object(), write(), say(), etc.

2.2 The Driver Cycle

Once the game has booted up and is properly functioning (the boot up process will be discussed in a future, advanced LPC textbook), the driver enters a loop which does not terminate until the shutdown() efun is legally called or a bug causes the driver program to crash.   First off, the driver handles any new incoming connections and passes control of the connection to a login object.   After that, the driver puts together a table of commands which have been entered by users since the last cycle of the driver.   After the command table is assembled, all messages scheduled to be sent to the connection from the last driver cycle are sent out to the user.   At this point, the driver goes through the table of commands to be executed and executes each set of commands each object has stored there.   The driver ends its cycle by calling the function heart_beat() in every object with a heart_beat() set and finally performing all pending call outs.   This chapter will not deal with the handling of connections, but instead will focus on how the driver handles user commands and heartbeats and call outs.

2.3 User Commands

The driver starts at the top of the list of living objects with pending commands, and successively performs each command it typed by calling the function associated with the command and passing any arguments the command giver gave as arguments to the function.   As the driver starts with the commands issued by a new living object, the command giver variable is changed to be equal to the new living object, so that during the sequence of functions initiated by that command, the efun this_player() returns the object which issued the command.

Let's look at the command buffer for an example player.   Since the execution of his last command, Bozo has typed "north" and "tell descartes when is the next reboot".   The command "north" is associated with the function "Do_Move()" in the room Bozo is in (the command "north" is automatically setup by the set_exits() efun in that room).   The command "tell" is not specifically listed as a command for the player, however, in the player object there is a function called "cmd_hook()" which is associated with the command "", which matches any possible user input.

Once the driver gets down to Bozo, the command giver variable is set to the object which is Bozo.   Then, seeing Bozo typed "north" and the function "north" is associated with, the driver calls Bozo's_Room- >Do_Move(0).   An argument of 0 is passed to the function since Bozo only typed the command "north" with no arguments.   The room naturally calls some functions it needs, all the while such that the efun this_player() returns the object which is Bozo.   Eventually, the room object will call move_player() in Bozo, which in turn calls the move_object() efun.   This efun is responsible for changing an object's environment.

When the environment of an object changes, the commands available to it from objects in its previous environment as well as from its previous environment are removed from the object.   Once that is done, the driver calls the efun init() in the new environment as well as in each object in the new environment.   During each of these calls to init(), the object Bozo is still the command giver.   Thus all add_action() efuns from this move will apply to Bozo.   Once all those calls are done, control passes back from the move_object() efun to the move_player() lfun in Bozo. move_player() returns control back to Do_Move() in the old room, which returns 1 to signify to the driver that the command action was successful.   If the Do_Move() function had returned 0 for some reason, the driver would have written "What?" (or whatever your driver's default bad command message is) to Bozo.

Once the first command returns 1, the driver proceeds on to Bozo's second command, following much the same structure.   Note that with "tell descartes when is the next reboot", the driver passes "descartes when is the next reboot" to the function associated with tell.   That function in turn has to decide what to do with that argument.   After that command returns either 1 or 0, the driver then proceeds on to the next living object with commands pending, and so on until all living objects with pending commands have had their commands performed.

2.4 The Efuns set_heart_beat() and call_out()

The most common use for heartbeats in the mudlib is to heal players and monsters and perform combat.   Once the driver has finished dealing with the command list, it goes through the heartbeat list calling heart_beat() in each object in the list.   So for a player, for example, the driver will call heart_beat() in the player which will:

1. age the player
2. heal the player according to a heal rate
3. check to see if there are any hunted, hunting, or attacking objects around
4. perform an attack if step 3 returns true.
5. any other things which need to happen automatically roughly every second

Note that the more objects which have heartbeats, the more processing which has to happen every cycle the mud is up.   Objects with heartbeats are thus known as the major hog of CPU time on muds.  

The call_out() efun is used to perform timed function calls which do not need to happen as often as heartbeats, or which just happen once.   Call outs let you specify the function in an object you want called.   The general formula for call outs is:

call_out(func, time, args);

The third argument specifying arguments is optional.   The first argument is a string representing the name of the function to be called.   The second argument is how many seconds should pass before the function gets called.

Practically speaking, when an object calls call_out(), it is added to a list of objects with pending call outs with the amount of time of the call out and the name of the function to be called.   Each cycle of the driver, the time is counted down until it becomes time for the function to be called. When the time comes, the driver removes the object from the list of objects with pending call outs and performs the call to the call out function, passing any special args originally specified by the call out function.

If you want a to remove a pending call before it occurs, you need to use the remove_call_out() efun, passing the name of the function being called out.   The driver will remove the next pending call out to that function.   This means you may have some ambiguity if more than one call out is pending for the same function.

In order to make a call out cyclical, you must reissue the call_out() efun in the function you called out, since the driver automatically removes the function from the call out table when a call out is performed.   Example:

void foo() { call_out("hello", 10); }

void hello() { call_out("hello", 10); }

will set up hello() to be called every 10 seconds after foo() is first called. There are several things to be careful about here.   First, you must watch to make sure you do not structure your call outs to be recursive in any unintended fashion.   Second, compare what a set_heart_beat() does when compared directly to what call_out() does.

set_heart_beat():

1. Adds this_object() to a table listing objects with heartbeats.
2. The function heart_beat() in this_object() gets called every single driver cycle.

call_out():

1. Adds this_object(), the name of a function in this_object(), a time delay, and a set of arguments to a table listing functions with pending call outs.  
2. The function named is called only once, and that call comes after the specified delay.

As you can see, there is a much greater memory overhead associated with call outs for part (a), yet that there is a much greater CPU overhead associated with heartbeats as shown in part (b), assuming that the delay for the call out is greater than a single driver cycle.

Clearly, you do not want to be issuing 1 second call outs, for then you get the worst of both worlds.   Similarly, you do not want to be having heart beats in objects that can perform the same functions with call outs of a greater duration than 1 second.   I personally have heard much talk about at what point you should use a call out over a heartbeat.   What I have mostly heard is that for single calls or for cycles of a duration greater than 10 seconds, it is best to use a call out.   For repetitive calls of durations less than 10 seconds, you are better off using heartbeats.   I do not know if this is true, but I do not think following this can do any harm.

2.5 Summary

Copyright (c) George Reese 1993