skia2/include/core/SkScalar.h
reed@android.com 187d559590 add dox for scalarcomparetype
use internal mutable version of drawPath when we've cons'd up a path on behalf
of a rect



git-svn-id: http://skia.googlecode.com/svn/trunk@258 2bbb7eff-a529-9590-31e7-b0007b416f81
2009-07-08 14:03:56 +00:00

254 lines
10 KiB
C

/*
* Copyright (C) 2006 The Android Open Source Project
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef SkScalar_DEFINED
#define SkScalar_DEFINED
#include "SkFixed.h"
/** \file SkScalar.h
Types and macros for the data type SkScalar. This is the fractional numeric type
that, depending on the compile-time flag SK_SCALAR_IS_FLOAT, may be implemented
either as an IEEE float, or as a 16.16 SkFixed. The macros in this file are written
to allow the calling code to manipulate SkScalar values without knowing which representation
is in effect.
*/
#ifdef SK_SCALAR_IS_FLOAT
#include "SkFloatingPoint.h"
/** SkScalar is our type for fractional values and coordinates. Depending on
compile configurations, it is either represented as an IEEE float, or
as a 16.16 fixed point integer.
*/
typedef float SkScalar;
extern const uint32_t gIEEENotANumber;
extern const uint32_t gIEEEInfinity;
/** SK_Scalar1 is defined to be 1.0 represented as an SkScalar
*/
#define SK_Scalar1 (1.0f)
/** SK_Scalar1 is defined to be 1/2 represented as an SkScalar
*/
#define SK_ScalarHalf (0.5f)
/** SK_ScalarInfinity is defined to be infinity as an SkScalar
*/
#define SK_ScalarInfinity (*(const float*)&gIEEEInfinity)
/** SK_ScalarMax is defined to be the largest value representable as an SkScalar
*/
#define SK_ScalarMax (3.4028235e+38f)
/** SK_ScalarMin is defined to be the smallest value representable as an SkScalar
*/
#define SK_ScalarMin (1.1754944e-38f)
/** SK_ScalarNaN is defined to be 'Not a Number' as an SkScalar
*/
#define SK_ScalarNaN (*(const float*)(const void*)&gIEEENotANumber)
/** SkScalarIsNaN(n) returns true if argument is not a number
*/
static inline bool SkScalarIsNaN(float x) { return x != x; }
/** SkIntToScalar(n) returns its integer argument as an SkScalar
*/
#define SkIntToScalar(n) ((float)(n))
/** SkFixedToScalar(n) returns its SkFixed argument as an SkScalar
*/
#define SkFixedToScalar(x) SkFixedToFloat(x)
/** SkScalarToFixed(n) returns its SkScalar argument as an SkFixed
*/
#define SkScalarToFixed(x) SkFloatToFixed(x)
#define SkScalarToFloat(n) (n)
#define SkFloatToScalar(n) (n)
#define SkScalarToDouble(n) (double)(n)
#define SkDoubleToScalar(n) (float)(n)
/** SkScalarFraction(x) returns the signed fractional part of the argument
*/
#define SkScalarFraction(x) sk_float_mod(x, 1.0f)
/** Rounds the SkScalar to the nearest integer value
*/
#define SkScalarRound(x) sk_float_round2int(x)
/** Returns the smallest integer that is >= the specified SkScalar
*/
#define SkScalarCeil(x) sk_float_ceil2int(x)
/** Returns the largest integer that is <= the specified SkScalar
*/
#define SkScalarFloor(x) sk_float_floor2int(x)
/** Returns the absolute value of the specified SkScalar
*/
#define SkScalarAbs(x) sk_float_abs(x)
/** Returns the value pinned between 0 and max inclusive
*/
inline SkScalar SkScalarClampMax(SkScalar x, SkScalar max) {
return x < 0 ? 0 : x > max ? max : x;
}
/** Returns the value pinned between min and max inclusive
*/
inline SkScalar SkScalarPin(SkScalar x, SkScalar min, SkScalar max) {
return x < min ? min : x > max ? max : x;
}
/** Returns the specified SkScalar squared (x*x)
*/
inline SkScalar SkScalarSquare(SkScalar x) { return x * x; }
/** Returns the product of two SkScalars
*/
#define SkScalarMul(a, b) ((float)(a) * (b))
/** Returns the product of two SkScalars plus a third SkScalar
*/
#define SkScalarMulAdd(a, b, c) ((float)(a) * (b) + (c))
/** Returns the product of a SkScalar and an int rounded to the nearest integer value
*/
#define SkScalarMulRound(a, b) SkScalarRound((float)(a) * (b))
/** Returns the product of a SkScalar and an int promoted to the next larger int
*/
#define SkScalarMulCeil(a, b) SkScalarCeil((float)(a) * (b))
/** Returns the product of a SkScalar and an int truncated to the next smaller int
*/
#define SkScalarMulFloor(a, b) SkScalarFloor((float)(a) * (b))
/** Returns the quotient of two SkScalars (a/b)
*/
#define SkScalarDiv(a, b) ((float)(a) / (b))
/** Returns the mod of two SkScalars (a mod b)
*/
#define SkScalarMod(x,y) sk_float_mod(x,y)
/** Returns the product of the first two arguments, divided by the third argument
*/
#define SkScalarMulDiv(a, b, c) ((float)(a) * (b) / (c))
/** Returns the multiplicative inverse of the SkScalar (1/x)
*/
#define SkScalarInvert(x) (SK_Scalar1 / (x))
#define SkScalarFastInvert(x) (SK_Scalar1 / (x))
/** Returns the square root of the SkScalar
*/
#define SkScalarSqrt(x) sk_float_sqrt(x)
/** Returns the average of two SkScalars (a+b)/2
*/
#define SkScalarAve(a, b) (((a) + (b)) * 0.5f)
/** Returns the geometric mean of two SkScalars
*/
#define SkScalarMean(a, b) sk_float_sqrt((float)(a) * (b))
/** Returns one half of the specified SkScalar
*/
#define SkScalarHalf(a) ((a) * 0.5f)
#define SK_ScalarSqrt2 1.41421356f
#define SK_ScalarPI 3.14159265f
#define SK_ScalarTanPIOver8 0.414213562f
#define SK_ScalarRoot2Over2 0.707106781f
#define SkDegreesToRadians(degrees) ((degrees) * (SK_ScalarPI / 180))
float SkScalarSinCos(SkScalar radians, SkScalar* cosValue);
#define SkScalarSin(radians) (float)sk_float_sin(radians)
#define SkScalarCos(radians) (float)sk_float_cos(radians)
#define SkScalarTan(radians) (float)sk_float_tan(radians)
#define SkScalarASin(val) (float)sk_float_asin(val)
#define SkScalarACos(val) (float)sk_float_acos(val)
#define SkScalarATan2(y, x) (float)sk_float_atan2(y,x)
#define SkScalarExp(x) (float)sk_float_exp(x)
#define SkScalarLog(x) (float)sk_float_log(x)
inline SkScalar SkMaxScalar(SkScalar a, SkScalar b) { return a > b ? a : b; }
inline SkScalar SkMinScalar(SkScalar a, SkScalar b) { return a < b ? a : b; }
#else
typedef SkFixed SkScalar;
#define SK_Scalar1 SK_Fixed1
#define SK_ScalarHalf SK_FixedHalf
#define SK_ScalarInfinity SK_FixedMax
#define SK_ScalarMax SK_FixedMax
#define SK_ScalarMin SK_FixedMin
#define SK_ScalarNaN SK_FixedNaN
#define SkScalarIsNaN(x) ((x) == SK_FixedNaN)
#define SkIntToScalar(n) SkIntToFixed(n)
#define SkFixedToScalar(x) (x)
#define SkScalarToFixed(x) (x)
#ifdef SK_CAN_USE_FLOAT
#define SkScalarToFloat(n) SkFixedToFloat(n)
#define SkFloatToScalar(n) SkFloatToFixed(n)
#define SkScalarToDouble(n) SkFixedToDouble(n)
#define SkDoubleToScalar(n) SkDoubleToFixed(n)
#endif
#define SkScalarFraction(x) SkFixedFraction(x)
#define SkScalarRound(x) SkFixedRound(x)
#define SkScalarCeil(x) SkFixedCeil(x)
#define SkScalarFloor(x) SkFixedFloor(x)
#define SkScalarAbs(x) SkFixedAbs(x)
#define SkScalarClampMax(x, max) SkClampMax(x, max)
#define SkScalarPin(x, min, max) SkPin32(x, min, max)
#define SkScalarSquare(x) SkFixedSquare(x)
#define SkScalarMul(a, b) SkFixedMul(a, b)
#define SkScalarMulAdd(a, b, c) SkFixedMulAdd(a, b, c)
#define SkScalarMulRound(a, b) SkFixedMulCommon(a, b, SK_FixedHalf)
#define SkScalarMulCeil(a, b) SkFixedMulCommon(a, b, SK_Fixed1 - 1)
#define SkScalarMulFloor(a, b) SkFixedMulCommon(a, b, 0)
#define SkScalarDiv(a, b) SkFixedDiv(a, b)
#define SkScalarMod(a, b) SkFixedMod(a, b)
#define SkScalarMulDiv(a, b, c) SkMulDiv(a, b, c)
#define SkScalarInvert(x) SkFixedInvert(x)
#define SkScalarFastInvert(x) SkFixedFastInvert(x)
#define SkScalarSqrt(x) SkFixedSqrt(x)
#define SkScalarAve(a, b) SkFixedAve(a, b)
#define SkScalarMean(a, b) SkFixedMean(a, b)
#define SkScalarHalf(a) ((a) >> 1)
#define SK_ScalarSqrt2 SK_FixedSqrt2
#define SK_ScalarPI SK_FixedPI
#define SK_ScalarTanPIOver8 SK_FixedTanPIOver8
#define SK_ScalarRoot2Over2 SK_FixedRoot2Over2
#define SkDegreesToRadians(degrees) SkFractMul(degrees, SK_FractPIOver180)
#define SkScalarSinCos(radians, cosPtr) SkFixedSinCos(radians, cosPtr)
#define SkScalarSin(radians) SkFixedSin(radians)
#define SkScalarCos(radians) SkFixedCos(radians)
#define SkScalarTan(val) SkFixedTan(val)
#define SkScalarASin(val) SkFixedASin(val)
#define SkScalarACos(val) SkFixedACos(val)
#define SkScalarATan2(y, x) SkFixedATan2(y,x)
#define SkScalarExp(x) SkFixedExp(x)
#define SkScalarLog(x) SkFixedLog(x)
#define SkMaxScalar(a, b) SkMax32(a, b)
#define SkMinScalar(a, b) SkMin32(a, b)
#endif
#define SK_ScalarNearlyZero (SK_Scalar1 / (1 << 12))
/* <= is slower than < for floats, so we use < for our tolerance test
*/
static inline bool SkScalarNearlyZero(SkScalar x,
SkScalar tolerance = SK_ScalarNearlyZero) {
SkASSERT(tolerance > 0);
return SkScalarAbs(x) < tolerance;
}
/** Linearly interpolate between A and B, based on t.
If t is 0, return A
If t is 1, return B
else interpolate.
t must be [0..SK_Scalar1]
*/
static inline SkScalar SkScalarInterp(SkScalar A, SkScalar B, SkScalar t) {
SkASSERT(t >= 0 && t <= SK_Scalar1);
return A + SkScalarMul(B - A, t);
}
#endif