CRD Argument Type
Many effects arguments are a CRD type of argument. This stands for ''Continuous Random Distribution'',
and is used to assign a variable argument. In other words, the argument's value is not fixed, but
is determined at the time the property of an instance of an object first needs it. This allows
effects to have a controlled randomness to them: a variable speed and direction for where various
bits fly, etc.
A CRD always has four sub-arguments, e.g.:
The meaning of the second and third arguments, which are always numbers,
varies depending on the type of distribution property specified.
The four types for this first argument are:
So these first three arguments compute a random number.
The final, fourth argument is a boolean, and can affect the sign of this computed number.
Specifically, if this argument is 1 (true), then there is a 50/50 chance that the computed number
will be negated. If 0 (false), no negation "coin-flip" takes place.
Some examples:
NOTE: this distribution is often misused in the Battlefield code, e.g.
Variance specifies the range in which about 4/5ths of the numbers generated will fall.
In our example, the 2 is the average time to live desired, and a variance of 3.5 means that
this value can vary fairly widely; about 4/5ths of the numbers generated will
be in the range +3.5 to -3.5 from the mean of 2, i.e. -1.5 to 5.5, with 1/5th of the numbers
appearing above and below this range.
If you want to test out this function, here is pseudocode:
Go to the All Properties list, All Types list, Class file list, CON file list, or Main scripting page.
CRD_NONE/3/0/0
The first argument says what sort of distribution will happen with the numbers that follow, and the
next two arguments are how to vary these numbers. The last argument is called symmetry, and can affect
any of the distributions; it will be explained later.
ObjectTemplate.TimeToLive CRD_NONE/240/0/0
For the exppack, the time to live before it blows up by itself is always 240 seconds. This sort of
CRD can be directly replaced by the number itself, e.g. on the ships the landing craft
TimeToLive property is set to
30, even though this property's argument is normally a CRD.
ObjectTemplate.TimeToLive CRD_UNIFORM/2/4/0
For depth charges, the time to live before it blows up by itself is between 2 and 4 seconds.
ObjectTemplate.rotationalSpeedInDof CRD_EXPONENTIAL/-20/0/1
The exponential distribution basically gives a distribution curve that drops off sharply and then
levels out.
It is computed by -mean*ln( random(0,1) ). In other words, a random number
between 0 and 1 (not including 0 or 1) is generated, the natural logarithm of this number is computed, and
the result is multiplied by the negative of the mean (the average),
the second argument (and first number) passed in.
For example, if the random number generated was 0.1, the natural log is -2.3, so since
the mean was -20, the final
number returned would be -(-20)*-2.3 = -46. Since the symmetric flag, the last argument, is set to 1 (true),
there is a 50/50 chance that this result will be negated, i.e. be returned as 46 instead of -46.
ObjectTemplate.initRotation CRD_EXPONENTIAL/0/180/1
is a common use. However, this setting actually always returns a value of 0, since that is the
mean specified. The "180" argument is ignored.
ObjectTemplate.timeToLive CRD_NORMAL/2/3.5/0
A normal distribution is a classic bell curve shape. The second argument is the mean, the average
value, and the third argument is the variance, how fast the distribution curve drops off into its
bell shape. The fourth argument again acts
to force symmetry if set, though the bell curve has its own symmetry when the mean is 0.
do {
u1 = getUniform(); // random number between 0 and 1
v1 = 2*u1 - 1; // random number between -1 and 1
u2 = getUniform(); // random number between 0 and 1
v2 = 2*u2 - 1; // random number between -1 and 1
w = (v1)*(v1) + (v2)*(v2);
} while (w > 1);
return mean + variance*v1*sqrt((-2*log(w))/w);
where the mean and variance are the first two numbers passed in.
The loop finds a point inside a circle of radius one. The resulting values are then used
to generate a normal distribution value, which is multiplied by the variance and has the
mean added to it.