1 /* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 2 -*- */
2 /* vim: set ts=8 sts=2 et sw=2 tw=80: */
3 /* This Source Code Form is subject to the terms of the Mozilla Public
4 * License, v. 2.0. If a copy of the MPL was not distributed with this
5 * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
7 /* Provides checked integers, detecting integer overflow and divide-by-0. */
9 #ifndef mozilla_CheckedInt_h
10 #define mozilla_CheckedInt_h
13 #include "mozilla/Assertions.h"
14 #include "mozilla/IntegerTypeTraits.h"
18 template<typename T
> class CheckedInt
;
23 * Step 1: manually record supported types
25 * What's nontrivial here is that there are different families of integer
26 * types: basic integer types and stdint types. It is merrily undefined which
27 * types from one family may be just typedefs for a type from another family.
29 * For example, on GCC 4.6, aside from the basic integer types, the only other
30 * type that isn't just a typedef for some of them, is int8_t.
33 struct UnsupportedType
{};
35 template<typename IntegerType
>
36 struct IsSupportedPass2
38 static const bool value
= false;
41 template<typename IntegerType
>
44 static const bool value
= IsSupportedPass2
<IntegerType
>::value
;
48 struct IsSupported
<int8_t>
49 { static const bool value
= true; };
52 struct IsSupported
<uint8_t>
53 { static const bool value
= true; };
56 struct IsSupported
<int16_t>
57 { static const bool value
= true; };
60 struct IsSupported
<uint16_t>
61 { static const bool value
= true; };
64 struct IsSupported
<int32_t>
65 { static const bool value
= true; };
68 struct IsSupported
<uint32_t>
69 { static const bool value
= true; };
72 struct IsSupported
<int64_t>
73 { static const bool value
= true; };
76 struct IsSupported
<uint64_t>
77 { static const bool value
= true; };
81 struct IsSupportedPass2
<char>
82 { static const bool value
= true; };
85 struct IsSupportedPass2
<signed char>
86 { static const bool value
= true; };
89 struct IsSupportedPass2
<unsigned char>
90 { static const bool value
= true; };
93 struct IsSupportedPass2
<short>
94 { static const bool value
= true; };
97 struct IsSupportedPass2
<unsigned short>
98 { static const bool value
= true; };
101 struct IsSupportedPass2
<int>
102 { static const bool value
= true; };
105 struct IsSupportedPass2
<unsigned int>
106 { static const bool value
= true; };
109 struct IsSupportedPass2
<long>
110 { static const bool value
= true; };
113 struct IsSupportedPass2
<unsigned long>
114 { static const bool value
= true; };
117 struct IsSupportedPass2
<long long>
118 { static const bool value
= true; };
121 struct IsSupportedPass2
<unsigned long long>
122 { static const bool value
= true; };
125 * Step 2: Implement the actual validity checks.
127 * Ideas taken from IntegerLib, code different.
130 template<typename IntegerType
, size_t Size
= sizeof(IntegerType
)>
131 struct TwiceBiggerType
133 typedef typename
detail::StdintTypeForSizeAndSignedness
<
134 sizeof(IntegerType
) * 2,
135 IsSigned
<IntegerType
>::value
139 template<typename IntegerType
>
140 struct TwiceBiggerType
<IntegerType
, 8>
142 typedef UnsupportedType Type
;
149 // In C++, right bit shifts on negative values is undefined by the standard.
150 // Notice that signed-to-unsigned conversions are always well-defined in the
151 // standard, as the value congruent modulo 2**n as expected. By contrast,
152 // unsigned-to-signed is only well-defined if the value is representable.
153 return bool(typename MakeUnsigned
<T
>::Type(aX
) >>
154 PositionOfSignBit
<T
>::value
);
157 // Bitwise ops may return a larger type, so it's good to use this inline
158 // helper guaranteeing that the result is really of type T.
161 BinaryComplement(T aX
)
168 bool IsTSigned
= IsSigned
<T
>::value
,
169 bool IsUSigned
= IsSigned
<U
>::value
>
170 struct DoesRangeContainRange
174 template<typename T
, typename U
, bool Signedness
>
175 struct DoesRangeContainRange
<T
, U
, Signedness
, Signedness
>
177 static const bool value
= sizeof(T
) >= sizeof(U
);
180 template<typename T
, typename U
>
181 struct DoesRangeContainRange
<T
, U
, true, false>
183 static const bool value
= sizeof(T
) > sizeof(U
);
186 template<typename T
, typename U
>
187 struct DoesRangeContainRange
<T
, U
, false, true>
189 static const bool value
= false;
194 bool IsTSigned
= IsSigned
<T
>::value
,
195 bool IsUSigned
= IsSigned
<U
>::value
,
196 bool DoesTRangeContainURange
= DoesRangeContainRange
<T
, U
>::value
>
197 struct IsInRangeImpl
{};
199 template<typename T
, typename U
, bool IsTSigned
, bool IsUSigned
>
200 struct IsInRangeImpl
<T
, U
, IsTSigned
, IsUSigned
, true>
208 template<typename T
, typename U
>
209 struct IsInRangeImpl
<T
, U
, true, true, false>
211 static bool run(U aX
)
213 return aX
<= MaxValue
<T
>::value
&& aX
>= MinValue
<T
>::value
;
217 template<typename T
, typename U
>
218 struct IsInRangeImpl
<T
, U
, false, false, false>
220 static bool run(U aX
)
222 return aX
<= MaxValue
<T
>::value
;
226 template<typename T
, typename U
>
227 struct IsInRangeImpl
<T
, U
, true, false, false>
229 static bool run(U aX
)
231 return sizeof(T
) > sizeof(U
) || aX
<= U(MaxValue
<T
>::value
);
235 template<typename T
, typename U
>
236 struct IsInRangeImpl
<T
, U
, false, true, false>
238 static bool run(U aX
)
240 return sizeof(T
) >= sizeof(U
)
242 : aX
>= 0 && aX
<= U(MaxValue
<T
>::value
);
246 template<typename T
, typename U
>
250 return IsInRangeImpl
<T
, U
>::run(aX
);
255 IsAddValid(T aX
, T aY
)
257 // Addition is valid if the sign of aX+aY is equal to either that of aX or
258 // that of aY. Since the value of aX+aY is undefined if we have a signed
259 // type, we compute it using the unsigned type of the same size. Beware!
260 // These bitwise operations can return a larger integer type, if T was a
261 // small type like int8_t, so we explicitly cast to T.
263 typename MakeUnsigned
<T
>::Type ux
= aX
;
264 typename MakeUnsigned
<T
>::Type uy
= aY
;
265 typename MakeUnsigned
<T
>::Type result
= ux
+ uy
;
266 return IsSigned
<T
>::value
267 ? HasSignBit(BinaryComplement(T((result
^ aX
) & (result
^ aY
))))
268 : BinaryComplement(aX
) >= aY
;
273 IsSubValid(T aX
, T aY
)
275 // Subtraction is valid if either aX and aY have same sign, or aX-aY and aX
276 // have same sign. Since the value of aX-aY is undefined if we have a signed
277 // type, we compute it using the unsigned type of the same size.
278 typename MakeUnsigned
<T
>::Type ux
= aX
;
279 typename MakeUnsigned
<T
>::Type uy
= aY
;
280 typename MakeUnsigned
<T
>::Type result
= ux
- uy
;
282 return IsSigned
<T
>::value
283 ? HasSignBit(BinaryComplement(T((result
^ aX
) & (aX
^ aY
))))
288 bool IsTSigned
= IsSigned
<T
>::value
,
289 bool TwiceBiggerTypeIsSupported
=
290 IsSupported
<typename TwiceBiggerType
<T
>::Type
>::value
>
291 struct IsMulValidImpl
{};
293 template<typename T
, bool IsTSigned
>
294 struct IsMulValidImpl
<T
, IsTSigned
, true>
296 static bool run(T aX
, T aY
)
298 typedef typename TwiceBiggerType
<T
>::Type TwiceBiggerType
;
299 TwiceBiggerType product
= TwiceBiggerType(aX
) * TwiceBiggerType(aY
);
300 return IsInRange
<T
>(product
);
305 struct IsMulValidImpl
<T
, true, false>
307 static bool run(T aX
, T aY
)
309 const T max
= MaxValue
<T
>::value
;
310 const T min
= MinValue
<T
>::value
;
312 if (aX
== 0 || aY
== 0) {
321 // If we reach this point, we know that aX < 0.
329 struct IsMulValidImpl
<T
, false, false>
331 static bool run(T aX
, T aY
)
333 return aY
== 0 || aX
<= MaxValue
<T
>::value
/ aY
;
339 IsMulValid(T aX
, T aY
)
341 return IsMulValidImpl
<T
>::run(aX
, aY
);
346 IsDivValid(T aX
, T aY
)
348 // Keep in mind that in the signed case, min/-1 is invalid because
351 !(IsSigned
<T
>::value
&& aX
== MinValue
<T
>::value
&& aY
== T(-1));
354 template<typename T
, bool IsTSigned
= IsSigned
<T
>::value
>
355 struct IsModValidImpl
;
359 IsModValid(T aX
, T aY
)
361 return IsModValidImpl
<T
>::run(aX
, aY
);
365 * Mod is pretty simple.
366 * For now, let's just use the ANSI C definition:
367 * If aX or aY are negative, the results are implementation defined.
368 * Consider these invalid.
369 * Undefined for aY=0.
370 * The result will never exceed either aX or aY.
372 * Checking that aX>=0 is a warning when T is unsigned.
376 struct IsModValidImpl
<T
, false>
378 static inline bool run(T aX
, T aY
)
385 struct IsModValidImpl
<T
, true>
387 static inline bool run(T aX
, T aY
)
396 template<typename T
, bool IsSigned
= IsSigned
<T
>::value
>
400 struct NegateImpl
<T
, false>
402 static CheckedInt
<T
> negate(const CheckedInt
<T
>& aVal
)
404 // Handle negation separately for signed/unsigned, for simpler code and to
405 // avoid an MSVC warning negating an unsigned value.
406 return CheckedInt
<T
>(0, aVal
.isValid() && aVal
.mValue
== 0);
411 struct NegateImpl
<T
, true>
413 static CheckedInt
<T
> negate(const CheckedInt
<T
>& aVal
)
415 // Watch out for the min-value, which (with twos-complement) can't be
416 // negated as -min-value is then (max-value + 1).
417 if (!aVal
.isValid() || aVal
.mValue
== MinValue
<T
>::value
) {
418 return CheckedInt
<T
>(aVal
.mValue
, false);
420 return CheckedInt
<T
>(-aVal
.mValue
, true);
424 } // namespace detail
428 * Step 3: Now define the CheckedInt class.
433 * @brief Integer wrapper class checking for integer overflow and other errors
434 * @param T the integer type to wrap. Can be any type among the following:
435 * - any basic integer type such as |int|
436 * - any stdint type such as |int8_t|
438 * This class implements guarded integer arithmetic. Do a computation, check
439 * that isValid() returns true, you then have a guarantee that no problem, such
440 * as integer overflow, happened during this computation, and you can call
441 * value() to get the plain integer value.
443 * The arithmetic operators in this class are guaranteed not to raise a signal
444 * (e.g. in case of a division by zero).
446 * For example, suppose that you want to implement a function that computes
447 * (aX+aY)/aZ, that doesn't crash if aZ==0, and that reports on error (divide by
448 * zero or integer overflow). You could code it as follows:
450 bool computeXPlusYOverZ(int aX, int aY, int aZ, int* aResult)
452 CheckedInt<int> checkedResult = (CheckedInt<int>(aX) + aY) / aZ;
453 if (checkedResult.isValid()) {
454 *aResult = checkedResult.value();
462 * Implicit conversion from plain integers to checked integers is allowed. The
463 * plain integer is checked to be in range before being casted to the
464 * destination type. This means that the following lines all compile, and the
465 * resulting CheckedInts are correctly detected as valid or invalid:
467 // 1 is of type int, is found to be in range for uint8_t, x is valid
468 CheckedInt<uint8_t> x(1);
469 // -1 is of type int, is found not to be in range for uint8_t, x is invalid
470 CheckedInt<uint8_t> x(-1);
471 // -1 is of type int, is found to be in range for int8_t, x is valid
472 CheckedInt<int8_t> x(-1);
473 // 1000 is of type int16_t, is found not to be in range for int8_t,
475 CheckedInt<int8_t> x(int16_t(1000));
476 // 3123456789 is of type uint32_t, is found not to be in range for int32_t,
478 CheckedInt<int32_t> x(uint32_t(3123456789));
480 * Implicit conversion from
481 * checked integers to plain integers is not allowed. As shown in the
482 * above example, to get the value of a checked integer as a normal integer,
485 * Arithmetic operations between checked and plain integers is allowed; the
486 * result type is the type of the checked integer.
488 * Checked integers of different types cannot be used in the same arithmetic
491 * There are convenience typedefs for all stdint types, of the following form
492 * (these are just 2 examples):
494 typedef CheckedInt<int32_t> CheckedInt32;
495 typedef CheckedInt<uint16_t> CheckedUint16;
506 CheckedInt(U aValue
, bool aIsValid
) : mValue(aValue
), mIsValid(aIsValid
)
508 static_assert(detail::IsSupported
<T
>::value
&&
509 detail::IsSupported
<U
>::value
,
510 "This type is not supported by CheckedInt");
513 friend struct detail::NegateImpl
<T
>;
517 * Constructs a checked integer with given @a value. The checked integer is
518 * initialized as valid or invalid depending on whether the @a value
521 * This constructor is not explicit. Instead, the type of its argument is a
522 * separate template parameter, ensuring that no conversion is performed
523 * before this constructor is actually called. As explained in the above
524 * documentation for class CheckedInt, this constructor checks that its
530 mIsValid(detail::IsInRange
<T
>(aValue
))
532 static_assert(detail::IsSupported
<T
>::value
&&
533 detail::IsSupported
<U
>::value
,
534 "This type is not supported by CheckedInt");
538 friend class CheckedInt
;
541 CheckedInt
<U
> toChecked() const
543 CheckedInt
<U
> ret(mValue
);
544 ret
.mIsValid
= ret
.mIsValid
&& mIsValid
;
548 /** Constructs a valid checked integer with initial value 0 */
549 CheckedInt() : mValue(0), mIsValid(true)
551 static_assert(detail::IsSupported
<T
>::value
,
552 "This type is not supported by CheckedInt");
555 /** @returns the actual value */
558 MOZ_ASSERT(mIsValid
, "Invalid checked integer (division by zero or integer overflow)");
563 * @returns true if the checked integer is valid, i.e. is not the result
564 * of an invalid operation or of an operation involving an invalid checked
573 friend CheckedInt
<U
> operator +(const CheckedInt
<U
>& aLhs
,
574 const CheckedInt
<U
>& aRhs
);
576 CheckedInt
& operator +=(U aRhs
);
579 friend CheckedInt
<U
> operator -(const CheckedInt
<U
>& aLhs
,
580 const CheckedInt
<U
>& aRhs
);
582 CheckedInt
& operator -=(U aRhs
);
585 friend CheckedInt
<U
> operator *(const CheckedInt
<U
>& aLhs
,
586 const CheckedInt
<U
>& aRhs
);
588 CheckedInt
& operator *=(U aRhs
);
591 friend CheckedInt
<U
> operator /(const CheckedInt
<U
>& aLhs
,
592 const CheckedInt
<U
>& aRhs
);
594 CheckedInt
& operator /=(U aRhs
);
597 friend CheckedInt
<U
> operator %(const CheckedInt
<U
>& aLhs
,
598 const CheckedInt
<U
>& aRhs
);
600 CheckedInt
& operator %=(U aRhs
);
602 CheckedInt
operator -() const
604 return detail::NegateImpl
<T
>::negate(*this);
608 * @returns true if the left and right hand sides are valid
609 * and have the same value.
611 * Note that these semantics are the reason why we don't offer
612 * a operator!=. Indeed, we'd want to have a!=b be equivalent to !(a==b)
613 * but that would mean that whenever a or b is invalid, a!=b
614 * is always true, which would be very confusing.
616 * For similar reasons, operators <, >, <=, >= would be very tricky to
617 * specify, so we just avoid offering them.
619 * Notice that these == semantics are made more reasonable by these facts:
620 * 1. a==b implies equality at the raw data level
621 * (the converse is false, as a==b is never true among invalids)
622 * 2. This is similar to the behavior of IEEE floats, where a==b
623 * means that a and b have the same value *and* neither is NaN.
625 bool operator ==(const CheckedInt
& aOther
) const
627 return mIsValid
&& aOther
.mIsValid
&& mValue
== aOther
.mValue
;
631 CheckedInt
& operator++()
638 CheckedInt
operator++(int)
640 CheckedInt tmp
= *this;
646 CheckedInt
& operator--()
653 CheckedInt
operator--(int)
655 CheckedInt tmp
= *this;
662 * The !=, <, <=, >, >= operators are disabled:
663 * see the comment on operator==.
665 template<typename U
> bool operator !=(U aOther
) const MOZ_DELETE
;
666 template<typename U
> bool operator < (U aOther
) const MOZ_DELETE
;
667 template<typename U
> bool operator <=(U aOther
) const MOZ_DELETE
;
668 template<typename U
> bool operator > (U aOther
) const MOZ_DELETE
;
669 template<typename U
> bool operator >=(U aOther
) const MOZ_DELETE
;
672 #define MOZ_CHECKEDINT_BASIC_BINARY_OPERATOR(NAME, OP) \
673 template<typename T> \
674 inline CheckedInt<T> \
675 operator OP(const CheckedInt<T>& aLhs, const CheckedInt<T>& aRhs) \
677 if (!detail::Is##NAME##Valid(aLhs.mValue, aRhs.mValue)) { \
678 return CheckedInt<T>(0, false); \
680 return CheckedInt<T>(aLhs.mValue OP aRhs.mValue, \
681 aLhs.mIsValid && aRhs.mIsValid); \
684 MOZ_CHECKEDINT_BASIC_BINARY_OPERATOR(Add
, +)
685 MOZ_CHECKEDINT_BASIC_BINARY_OPERATOR(Sub
, -)
686 MOZ_CHECKEDINT_BASIC_BINARY_OPERATOR(Mul
, *)
687 MOZ_CHECKEDINT_BASIC_BINARY_OPERATOR(Div
, /)
688 MOZ_CHECKEDINT_BASIC_BINARY_OPERATOR(Mod
, %)
690 #undef MOZ_CHECKEDINT_BASIC_BINARY_OPERATOR
692 // Implement castToCheckedInt<T>(x), making sure that
693 // - it allows x to be either a CheckedInt<T> or any integer type
694 // that can be casted to T
695 // - if x is already a CheckedInt<T>, we just return a reference to it,
696 // instead of copying it (optimization)
700 template<typename T
, typename U
>
701 struct CastToCheckedIntImpl
703 typedef CheckedInt
<T
> ReturnType
;
704 static CheckedInt
<T
> run(U aU
) { return aU
; }
708 struct CastToCheckedIntImpl
<T
, CheckedInt
<T
> >
710 typedef const CheckedInt
<T
>& ReturnType
;
711 static const CheckedInt
<T
>& run(const CheckedInt
<T
>& aU
) { return aU
; }
714 } // namespace detail
716 template<typename T
, typename U
>
717 inline typename
detail::CastToCheckedIntImpl
<T
, U
>::ReturnType
718 castToCheckedInt(U aU
)
720 static_assert(detail::IsSupported
<T
>::value
&&
721 detail::IsSupported
<U
>::value
,
722 "This type is not supported by CheckedInt");
723 return detail::CastToCheckedIntImpl
<T
, U
>::run(aU
);
726 #define MOZ_CHECKEDINT_CONVENIENCE_BINARY_OPERATORS(OP, COMPOUND_OP) \
727 template<typename T> \
728 template<typename U> \
729 CheckedInt<T>& CheckedInt<T>::operator COMPOUND_OP(U aRhs) \
731 *this = *this OP castToCheckedInt<T>(aRhs); \
734 template<typename T, typename U> \
735 inline CheckedInt<T> operator OP(const CheckedInt<T>& aLhs, U aRhs) \
737 return aLhs OP castToCheckedInt<T>(aRhs); \
739 template<typename T, typename U> \
740 inline CheckedInt<T> operator OP(U aLhs, const CheckedInt<T>& aRhs) \
742 return castToCheckedInt<T>(aLhs) OP aRhs; \
745 MOZ_CHECKEDINT_CONVENIENCE_BINARY_OPERATORS(+, +=)
746 MOZ_CHECKEDINT_CONVENIENCE_BINARY_OPERATORS(*, *=)
747 MOZ_CHECKEDINT_CONVENIENCE_BINARY_OPERATORS(-, -=)
748 MOZ_CHECKEDINT_CONVENIENCE_BINARY_OPERATORS(/, /=)
749 MOZ_CHECKEDINT_CONVENIENCE_BINARY_OPERATORS(%, %=)
751 #undef MOZ_CHECKEDINT_CONVENIENCE_BINARY_OPERATORS
753 template<typename T
, typename U
>
755 operator ==(const CheckedInt
<T
>& aLhs
, U aRhs
)
757 return aLhs
== castToCheckedInt
<T
>(aRhs
);
760 template<typename T
, typename U
>
762 operator ==(U aLhs
, const CheckedInt
<T
>& aRhs
)
764 return castToCheckedInt
<T
>(aLhs
) == aRhs
;
767 // Convenience typedefs.
768 typedef CheckedInt
<int8_t> CheckedInt8
;
769 typedef CheckedInt
<uint8_t> CheckedUint8
;
770 typedef CheckedInt
<int16_t> CheckedInt16
;
771 typedef CheckedInt
<uint16_t> CheckedUint16
;
772 typedef CheckedInt
<int32_t> CheckedInt32
;
773 typedef CheckedInt
<uint32_t> CheckedUint32
;
774 typedef CheckedInt
<int64_t> CheckedInt64
;
775 typedef CheckedInt
<uint64_t> CheckedUint64
;
777 } // namespace mozilla
779 #endif /* mozilla_CheckedInt_h */