12.21.9. - Referring to data within your subroutines
In the file p_fitsize.h is the definition of the data structure:
struct f_data {
unsigned d_flags; /* Used internally */
float d_w2; /* The square of the weight */
double d_y; /* The dependent variable */
double d_ys; /* Used internally */
float d_xx[1]; /* Independent variable(s) */
};
extern struct f_data *f_data, *dp;
#define INC(d) (d = (struct f_data * ) (((char *) d) + dpsize))
The pointer
f_data is the base of an array of these structures.
The pointer
dp contains the address of the array element associated
with the current point when fitting.
The macro definition
INC(d) illustrates how to increment a pointer to an element in the
f_data array by one element.
Such machinations are necessary since the size
of an array element depends on the number of independent variables
configured.
A fragment of code with which you might loop through all the data points follows:
register int i;
register struct f_data *d;
for (d = f_data, i = 0; i < npts; i++, INC(d)) {
if (d->d_flags&D_DONT_FIT)
continue;
d-d_xx[0] =
d-d_xx[1] =
d-d_xx[2] =
d-d_y =
d-d_w2 =
}
This example assumes three independent variables.
Of course, the example
still requires the right-hand side of the assignment expressions.
The externally defined integer
npts contains the current number
of data points.
The data flags have the
D_DONT_FIT bit set if
the points are not in the current fit range.