Converting bitset to signed integer

[jars121]

Hi,

I use a generic 32-bit bitset to perform some bit manipulation, before converting the bitset into a useable integer. The process works as intended for unsigned integers, but I'm unable to get the desired result if the intended output integer is signed.

I use a 32-bit bitset as the data will never be larger than 32-bits; but it can be anywhere from 1 bit up to 32 bits. Code as follows:

//Let's use the hexadecimal value 0xE008 here as an example
std::bitset<32> bitset(0xE008);

//Converting this to a uint32_t works as expected:
uint32_t unsigned = bitset.to_ulong();

qDebug() << "unsigned expected: 57352";
qDebug() << "unsigned actual: " << unsigned;

//Attempts to instead convert it to a signed integer don't work as expected:
int32_t signed = static_cast<int32_t>(bitset.to_ulong());

qDebug() << "signed expected: -8184";
qDebug() << "signed actual: " << signed;

The resultant debugs:

unsigned expected: 57352
unsigned actual: 57352
signed expected: -8184
signed actual: 57352

I've read a couple of threads regarding the use of std::bitset with signed integers but haven't made any progress. Is there something simple I'm missing here?

[Bonnie]

I think 0xE008 only equals -8184 when being a signed 16bit integer?

[jars121]

Ah of course you're right.

I just changed 'signed' from int32_t to int16_t and it gives -8184 as expected.

I guess I need to check whether the unsigned value is greater than 32767, and if so, then set the left-most/most significant bits accordingly to convert the value to a negative.

For example:

if the value is 7FFF (32,767), the value can be converted to a uin32_t as shown above. However, if the value is 8000 (32,768), set the left-most 16 bits to 1 and the resultant int32_t will be -32768 as expected.

Thanks for pointing me in the right direction!

[Bonnie]

@jars121
I don't fully understand you but I think you can just convert it to a signed 16bit integer and then assign the 16 bit one to a 32bit integer.

[jars121]

Yes you're right, I suppose I was just considering instances where the value is greater than 16 bits, in which case some additional manual processing would be required.