gecko-dev/dom/bindings/PrimitiveConversions.h

359 lines
10 KiB
C++

/* -*- Mode: C++; tab-width: 2; indent-tabs-mode: nil; c-basic-offset: 2 -*-*/
/* vim: set ts=2 sw=2 et tw=79: */
/* This Source Code Form is subject to the terms of the Mozilla Public
* License, v. 2.0. If a copy of the MPL was not distributed with this file,
* You can obtain one at http://mozilla.org/MPL/2.0/. */
/**
* Conversions from jsval to primitive values
*/
#ifndef mozilla_dom_PrimitiveConversions_h
#define mozilla_dom_PrimitiveConversions_h
#include <limits>
#include <math.h>
#include <stdint.h>
#include "jsapi.h"
#include "mozilla/Assertions.h"
#include "mozilla/ErrorResult.h"
#include "mozilla/FloatingPoint.h"
namespace mozilla {
namespace dom {
template<typename T>
struct TypeName {
};
template<>
struct TypeName<int8_t> {
static const char* value() {
return "byte";
}
};
template<>
struct TypeName<uint8_t> {
static const char* value() {
return "octet";
}
};
template<>
struct TypeName<int16_t> {
static const char* value() {
return "short";
}
};
template<>
struct TypeName<uint16_t> {
static const char* value() {
return "unsigned short";
}
};
template<>
struct TypeName<int32_t> {
static const char* value() {
return "long";
}
};
template<>
struct TypeName<uint32_t> {
static const char* value() {
return "unsigned long";
}
};
template<>
struct TypeName<int64_t> {
static const char* value() {
return "long long";
}
};
template<>
struct TypeName<uint64_t> {
static const char* value() {
return "unsigned long long";
}
};
enum ConversionBehavior {
eDefault,
eEnforceRange,
eClamp
};
template<typename T, ConversionBehavior B>
struct PrimitiveConversionTraits {
};
template<typename T>
struct DisallowedConversion {
typedef int jstype;
typedef int intermediateType;
private:
static inline bool converter(JSContext* cx, JS::Handle<JS::Value> v,
jstype* retval) {
MOZ_CRASH("This should never be instantiated!");
}
};
struct PrimitiveConversionTraits_smallInt {
// The output of JS::ToInt32 is determined as follows:
// 1) The value is converted to a double
// 2) Anything that's not a finite double returns 0
// 3) The double is rounded towards zero to the nearest integer
// 4) The resulting integer is reduced mod 2^32. The output of this
// operation is an integer in the range [0, 2^32).
// 5) If the resulting number is >= 2^31, 2^32 is subtracted from it.
//
// The result of all this is a number in the range [-2^31, 2^31)
//
// WebIDL conversions for the 8-bit, 16-bit, and 32-bit integer types
// are defined in the same way, except that step 4 uses reduction mod
// 2^8 and 2^16 for the 8-bit and 16-bit types respectively, and step 5
// is only done for the signed types.
//
// C/C++ define integer conversion semantics to unsigned types as taking
// your input integer mod (1 + largest value representable in the
// unsigned type). Since 2^32 is zero mod 2^8, 2^16, and 2^32,
// converting to the unsigned int of the relevant width will correctly
// perform step 4; in particular, the 2^32 possibly subtracted in step 5
// will become 0.
//
// Once we have step 4 done, we're just going to assume 2s-complement
// representation and cast directly to the type we really want.
//
// So we can cast directly for all unsigned types and for int32_t; for
// the smaller-width signed types we need to cast through the
// corresponding unsigned type.
typedef int32_t jstype;
typedef int32_t intermediateType;
static inline bool converter(JSContext* cx, JS::Handle<JS::Value> v,
jstype* retval) {
return JS::ToInt32(cx, v, retval);
}
};
template<>
struct PrimitiveConversionTraits<int8_t, eDefault> : PrimitiveConversionTraits_smallInt {
typedef uint8_t intermediateType;
};
template<>
struct PrimitiveConversionTraits<uint8_t, eDefault> : PrimitiveConversionTraits_smallInt {
};
template<>
struct PrimitiveConversionTraits<int16_t, eDefault> : PrimitiveConversionTraits_smallInt {
typedef uint16_t intermediateType;
};
template<>
struct PrimitiveConversionTraits<uint16_t, eDefault> : PrimitiveConversionTraits_smallInt {
};
template<>
struct PrimitiveConversionTraits<int32_t, eDefault> : PrimitiveConversionTraits_smallInt {
};
template<>
struct PrimitiveConversionTraits<uint32_t, eDefault> : PrimitiveConversionTraits_smallInt {
};
template<>
struct PrimitiveConversionTraits<int64_t, eDefault> {
typedef int64_t jstype;
typedef int64_t intermediateType;
static inline bool converter(JSContext* cx, JS::Handle<JS::Value> v,
jstype* retval) {
return JS::ToInt64(cx, v, retval);
}
};
template<>
struct PrimitiveConversionTraits<uint64_t, eDefault> {
typedef uint64_t jstype;
typedef uint64_t intermediateType;
static inline bool converter(JSContext* cx, JS::Handle<JS::Value> v,
jstype* retval) {
return JS::ToUint64(cx, v, retval);
}
};
template<typename T>
struct PrimitiveConversionTraits_Limits {
static inline T min() {
return std::numeric_limits<T>::min();
}
static inline T max() {
return std::numeric_limits<T>::max();
}
};
template<>
struct PrimitiveConversionTraits_Limits<int64_t> {
static inline int64_t min() {
return -(1LL << 53);
}
static inline int64_t max() {
return (1LL << 53);
}
};
template<>
struct PrimitiveConversionTraits_Limits<uint64_t> {
static inline uint64_t min() {
return 0;
}
static inline uint64_t max() {
return (1LL << 53);
}
};
template<typename T, bool (*Enforce)(JSContext* cx, const double& d, T* retval)>
struct PrimitiveConversionTraits_ToCheckedIntHelper {
typedef T jstype;
typedef T intermediateType;
static inline bool converter(JSContext* cx, JS::Handle<JS::Value> v,
jstype* retval) {
double intermediate;
if (!JS::ToNumber(cx, v, &intermediate)) {
return false;
}
return Enforce(cx, intermediate, retval);
}
};
template<typename T>
inline bool
PrimitiveConversionTraits_EnforceRange(JSContext* cx, const double& d, T* retval)
{
static_assert(std::numeric_limits<T>::is_integer,
"This can only be applied to integers!");
if (!mozilla::IsFinite(d)) {
return ThrowErrorMessage(cx, MSG_ENFORCE_RANGE_NON_FINITE, TypeName<T>::value());
}
bool neg = (d < 0);
double rounded = floor(neg ? -d : d);
rounded = neg ? -rounded : rounded;
if (rounded < PrimitiveConversionTraits_Limits<T>::min() ||
rounded > PrimitiveConversionTraits_Limits<T>::max()) {
return ThrowErrorMessage(cx, MSG_ENFORCE_RANGE_OUT_OF_RANGE, TypeName<T>::value());
}
*retval = static_cast<T>(rounded);
return true;
}
template<typename T>
struct PrimitiveConversionTraits<T, eEnforceRange> :
public PrimitiveConversionTraits_ToCheckedIntHelper<T, PrimitiveConversionTraits_EnforceRange<T> > {
};
template<typename T>
inline bool
PrimitiveConversionTraits_Clamp(JSContext* cx, const double& d, T* retval)
{
static_assert(std::numeric_limits<T>::is_integer,
"This can only be applied to integers!");
if (mozilla::IsNaN(d)) {
*retval = 0;
return true;
}
if (d >= PrimitiveConversionTraits_Limits<T>::max()) {
*retval = PrimitiveConversionTraits_Limits<T>::max();
return true;
}
if (d <= PrimitiveConversionTraits_Limits<T>::min()) {
*retval = PrimitiveConversionTraits_Limits<T>::min();
return true;
}
MOZ_ASSERT(mozilla::IsFinite(d));
// Banker's rounding (round ties towards even).
// We move away from 0 by 0.5f and then truncate. That gets us the right
// answer for any starting value except plus or minus N.5. With a starting
// value of that form, we now have plus or minus N+1. If N is odd, this is
// the correct result. If N is even, plus or minus N is the correct result.
double toTruncate = (d < 0) ? d - 0.5 : d + 0.5;
T truncated = static_cast<T>(toTruncate);
if (truncated == toTruncate) {
/*
* It was a tie (since moving away from 0 by 0.5 gave us the exact integer
* we want). Since we rounded away from 0, we either already have an even
* number or we have an odd number but the number we want is one closer to
* 0. So just unconditionally masking out the ones bit should do the trick
* to get us the value we want.
*/
truncated &= ~1;
}
*retval = truncated;
return true;
}
template<typename T>
struct PrimitiveConversionTraits<T, eClamp> :
public PrimitiveConversionTraits_ToCheckedIntHelper<T, PrimitiveConversionTraits_Clamp<T> > {
};
template<ConversionBehavior B>
struct PrimitiveConversionTraits<bool, B> : public DisallowedConversion<bool> {};
template<>
struct PrimitiveConversionTraits<bool, eDefault> {
typedef bool jstype;
typedef bool intermediateType;
static inline bool converter(JSContext* /* unused */, JS::Handle<JS::Value> v,
jstype* retval) {
*retval = JS::ToBoolean(v);
return true;
}
};
template<ConversionBehavior B>
struct PrimitiveConversionTraits<float, B> : public DisallowedConversion<float> {};
template<ConversionBehavior B>
struct PrimitiveConversionTraits<double, B> : public DisallowedConversion<double> {};
struct PrimitiveConversionTraits_float {
typedef double jstype;
typedef double intermediateType;
static inline bool converter(JSContext* cx, JS::Handle<JS::Value> v,
jstype* retval) {
return JS::ToNumber(cx, v, retval);
}
};
template<>
struct PrimitiveConversionTraits<float, eDefault> : PrimitiveConversionTraits_float {
};
template<>
struct PrimitiveConversionTraits<double, eDefault> : PrimitiveConversionTraits_float {
};
template<typename T, ConversionBehavior B>
bool ValueToPrimitive(JSContext* cx, JS::Handle<JS::Value> v, T* retval)
{
typename PrimitiveConversionTraits<T, B>::jstype t;
if (!PrimitiveConversionTraits<T, B>::converter(cx, v, &t))
return false;
*retval = static_cast<T>(
static_cast<typename PrimitiveConversionTraits<T, B>::intermediateType>(t));
return true;
}
} // namespace dom
} // namespace mozilla
#endif /* mozilla_dom_PrimitiveConversions_h */