float vs. double in polynomial functions

[David.Robbins@…]

It appears that the functions formerly 'psDPolynomial?D' and 'psPolynomial?D'
have been merged into one data type. However, the evaluation functions only
use psF64 arguments and return psF64. Shouldn't the psF32 part of psPolynomial
either be dropped or the eval functions changed correspondingly? Also, the
allocators for psPolynomial only take 'int' elements. Shouldn't this also be
changed to unsigned long or at least unsigned int?

[Paul Price]

The evaluation functions use double precision values as arguments and return so
that the float and double version of the polynomials can be amalgamated into a
single function. We keep the float version around in the interest of speed; the
double in the interest of precision.

I think the allocators should be unsigned, like you say.

typedef struct {

psPolynomialType type; /< Polynomial type
psElemType ctype; /< Polynomial precision
unsigned int n; /< Number of terms
union {

psF32 *F32; /< Coefficients
psF64 *F64; /< Coefficients

} coeff;
union {

psF32 *F32; /< Error in coefficients
psF64 *F64; /< Error in coefficients

} coeffErr;
char *mask; /< Coefficient mask

} psPolynomial1D;

psPolynomial1D *psPolynomial1DAlloc(psPolynomialType type,

psElemType ctype,
unsigned int nX);

(etc)

Also noticed the spline just below has the same issue, so changed to:

typedef struct {

unsigned int n; /< Number of spline pieces
psPolynomial1D spline; /< Array of n pointers to splines
psVector *knots; /< The boundaries between each spline

piece. Size is n+1.
} psSpline1D;

psSpline1D *psSpline1DAlloc(unsigned int n, unsigned int order, float min, float
max);
psSpline1D *psSpline1DAllocGeneric(const psVector *bounds, unsigned int order);

[Paul Price]

Removed the support for both F32 and F64. Now polynomials only contain F64.