Reputation: 4114
I want to convert some numbers from __int64
to char
. For this purpose I want to use the function __i64toa_s
, as suggested by MS. The function expects a char buffer for the return value. How can I determine the minimal needed size of this buffer? I don't want to get errors because of a too small buffer, but I don't want to waste space, too.
Thank you!
Upvotes: 0
Views: 596
Reputation: 1359
I would suggest a more generic meta-function, which I use myself whenever I'm in need.
It calculates the maximum storage size for any numerical integral type in radix 10 by facilitating std::numeric_limits<>
:
/** @short Metafunction that calculates the maximum storage size for a string
* representation of a numerical type in base10, eventually including the sign.
*
* @tparam T integral type, forwarded to std::numeric_limits<>
* @tparam null_inclusive whether to include the terminating null-character
* into the storage size
*/
template<typename T, bool null_inclusive = true>
struct digits10_storage_size:
std::integral_constant<
std::size_t,
// note: digits10 returns the number of deciaml digits that the type can
// represent without loss of precision, i.e. for char we get 2 digits (0..99)
//
// so we add here 1 to get 3 digits because [100..255] are 3 digits.
// we add another time 1 because of an eventual sign (+/-) that a conversion
// function could produce, plus finally the trailing \0-character eventually
std::numeric_limits<T>::digits10 + 2 + null_inclusive
>
{};
Advantages I see:
Usage:
char szNum[digits10_storage_size<__int64>::value];
_i64toa_s(i, szNum, 10);
// ... or use "unsafe" version because of correctly precomputed buffer:
_i64toa(i, szNum, 10);
Upvotes: 1
Reputation: 612794
Assuming decimal output, you need a buffer of at least 21 characters. You need to allow for 19 digits for a 64 bit integer represented as decimal, a minus sign, and a null terminator.
However, since you are using C++, it would seem to me to make much more sense to use a C++ solution and avoid null-terminated C strings. I see no reason to use __i64toa_s
. Use a string stream instead.
Upvotes: 1
Reputation: 55887
You can simply count, how many digits is in __int64
value. By loop, or by log10(value) + 1
, if log10
is overloaded for __int64
.
Upvotes: 2