C Programming/tgmath.h

tgmath.h is a Standard C header that defines many type-generic macros that can be used for a variety of mathematical operations. This header also includes math.h and complex.h. For all of the functions in the math.h and complex.h headers that do not have an f (float) or l (long double) suffix, and whose corresponding type is double (with the exception of modf()), there is a corresponding macro.

Type-generic macro

A type generic macro is something which allows calling a function whose type is determined by the type of argument in the macro. This means, for example, x is declared as an int data type but $tan$ has been called this way:

 tan((float)x)

then this expression will have a type float.
Also, if any one of the parameters or arguments of a type-generic macro is complex, it will call a complex function, otherwise a real function will be called. The type of function that is called, ultimately depends upon the final converted type of parameter..

Dependency Graph

The flowchart below shows the Dependency graph for tgmath.h.

Dependency Graph

Functions from math.h

The functions listed below handle real arguments and return real output only. If complex arguments are passed to these functions, NaN or other special numbers are returned.

Name	Description
acos	inverse cosine
asin	inverse sine
atan	one-parameter inverse tangent
atan2	two-parameter inverse tangent
ceil	ceiling, the smallest integer not less than parameter
cos	cosine
cosh	hyperbolic cosine
exp	exponential function
fabs	absolute value of a floating-point number
floor	floor, the largest integer not greater than parameter
fmod	floating-point remainder
frexp	decomposes a number into significand and a power of 2
ldexp	scale floating-point number by exponent (see article)
log	natural logarithm
log10	base-10 logarithm
modf(x,p)	returns fractional part of x and stores integral part where pointer p points to
pow(x,y)	raise x to the power of y, xy
sin	sine
sinh	hyperbolic sine
sqrt	square root, returns non-negative square-root of the number
tan	tangent
tanh	hyperbolic tangent
isgreater	If x > y, it returns 1, otherwise 0
isnan(x)	returns 1, if x is NaN (Not A Number)
islessequal	returns 1, if x is less than or equal to y, otherwise returns 0
isinf(x)	represents whether x is infinite
signbit	represents whether x is negative (returns 8 if x is less than 0, otherwise returns 0)

Functions from complex.h

The functions defined below handle complex parameters, but do not return complex output.

Name	Description
cacos	inverse cosine
casin	inverse sine
catan	one-parameter inverse tangent
catan2	two-parameter inverse tangent
cceil	ceiling, the smallest integer not less than parameter
ccos	cosine
ccosh	hyperbolic cosine
cexp	exponential function
cabs	absolute value of a complex floating-point number
cfloor	floor, the largest integer not greater than parameter
cfmod	floating-point remainder
cfrexp	decomposes a number into significand and a power of 2
cldexp	scale floating-point number by exponent (see article)
clog	natural logarithm
clog10	base-10 logarithm
cmodf(x,p)	returns fractional part of x and stores integral part where pointer p points to
cpow(x,y)	raise x to the power of y, xy
csin	sine
csinh	hyperbolic sine
csqrt	square root, returns non-negative square-root of the number
ctan	tangent
ctanh	hyperbolic tangent

Usage example

The code below illustrates usage of ,atan function defined in tgmath.h, which computes the inverse of tangent of a number defined in the domain of tangent.

#include <stdio.h>
#include <tgmath.h>
int main()
{
        float ang, ans;
        scanf("%f", &ang);
        ans = atan(ang);
        printf("%0.3f\n", ans);
        return 0;
}

Notable differences

The similar functions defined here have notable difference when it comes to return value of "tricky" numbers. For example, using sqrt to compute square root of -25 returns -NaN(Not a Number), whereas, csqrt returns 0.000000. Such differences may be noticed in other functions also.

Rationale

This header file is mainly included while calculating mathematical functions.As it includes both math.h and complex.h, the problem arising due to inconsistent input is solved. Including any of the header files, individually, involves inconsistent outputs for some inputs.

External Links

• Documentation http://clang.llvm.org/doxygen/tgmath_8h-source.html
• http://www.unix.com/man-page/OpenSolaris/3head/tgmath.h/
• http://publib.boulder.ibm.com/infocenter/zos/v1r11/index.jsp?topic=/com.ibm.zos.r11.bpxbd00/tgmathh.htm
• nlib.h http://read.pudn.com/downloads170/sourcecode/math/784268/nlib.h__.htm