Split array of m bytes into chunks of n bytes - c++

Question

I'm working on a program that manipulates brain data. It recieves a value represents the current magnitude of 8 commonly-recognized types of EEG (brain-waves). This data value is output as a series of eight 3-byte unsigned integers in little-endian format.
Here is a piece of my code:
if (extendedCodeLevel == 0 && code == ASIC_EEG_POWER_CODE)
{
fprintf(arq4, "EXCODE level: %d CODE: 0x%02X vLength: %d\n", extendedCodeLevel, code, valueLength );
fprintf(arq4, "Data value(s):" );
for( i=0; i<valueLength; i++ ) fprintf(arq4, " %d", value[0] & 0xFF );
}
The value value[0] is my output. It is the series of bytes that represents the brain waves. The current output file contains is the following data:
EXCODE level: 0x00 CODE: 0x83 vLength: 24
Data value(s): 16 2 17 5 3 2 22 1 2 1 0 0 0 4 0 0 3 0 0 5 1 0 4 8
What I need is to divide the sequence of bytes above into 3-byte chunks, to identify the EEG. The wave delta is represented by the first 3-byte sequence, theta is represented by the next bytes, and so on. How can I do it?

Answer 1

Assuming that you know that your input will always be exactly eight three-bit integers, all you need is a simple loop that reads three bytes from the input and writes them out as a four-byte value. The easiest way to do this is to treat the input as an array of bytes and then pull bytes off of this array in groups of three.
// Convert an array of eight 3-byte integers into an array
// of eight 4-byte integers.
void convert_3to4(const void* input, void* output)
{
uint32_t tmp;
uint32_t* pOut = output;
uint8_t* pIn = input;
int i;
for (i=0; i<24; i+=3)
{
tmp = pIn[i];
tmp += (pIn[i+1] << 8);
tmp += (pIn[i+2] << 16);
pOut[((i+2) / 3)] = tmp;
}
}

Answer 2

Like this? The last bytes are not be printed if are not aligned by 3. Do you need them?
for( i=0; i<valueLength; i+=3 ) fprintf(arq4, "%d %d %d - ", value[i] & 0xFF,
value[i+1] & 0xFF,
value[i+2] & 0xFF );

Answer 3

Converting eight 3-byte little endian character-steams into eight 4-byte integers is fairly trivial:
for( int i = 0; i < 24; ++i )
{
output[ i & 0x07 ] |= input[ i ] << ( i & 0x18 );
}
I think that (untested) code will do it. Assuming input is a 24-entry char array, and output is an eight-entry int array.

Answer 4

You might try s.th. like this:
union _32BitValue
{
uint8_t bytes[4];
uint32_t uval;
}
size_t extractUint32From3ByteSegemented(const std::vector<uint8_t>& rawData, size_t index, uint32_t& result)
{
// May be do some checks, if the vector size fits extracting the data from it,
// throwing exception or return 0 etc. ...
_32BitValue tmp;
tmp.bytes[0] = 0;
tmp.bytes[1] = rawData[index + 2];
tmp.bytes[2] = rawData[index + 1];
tmp.bytes[3] = rawData[index];
result = ntohl(tmp.uval);
return index + 3;
}
The code used to parse the values from the raw data array:
size_t index = 0;
std::vector<uint8_t> rawData = readRawData(); // Provide such method to read the raw data into the vector
std::vector<uint32_t> myDataValues;
while(index < rawData.size())
{
uint32_t extractedValue;
index = extractUint32From3ByteSegemented(rawData,index,extractedValue);
// Depending on what error detection you choose do check for index returned
// != 0, or catch exception ...
myDataValues.push_back(extractedValue);
}
// Continue with calculations on the extracted values ...
Using the left shift operator and addition as shown in other answers will do the trick as well. But IMHO this sample shows clearly what's going on. It fills the unions byte array with a value in big-endian (network) order and uses ntohl() to retrieve the result in the host machine's used format (big- or little-endian) portably.

Answer 5

What I need is, instead of displaying the whole sequence of 24 bytes, I need to get the 3-byte sequences separately.
You can easily copy the 1d byte array to the desired 2d shape.
Example:
#include <inttypes.h>
#include <stdio.h>
#include <string.h>
int main() {
/* make up data */
uint8_t bytes[] =
{ 16, 2, 17,
5, 3, 2,
22, 1, 2,
1, 0, 0,
0, 4, 0,
0, 3, 0,
0, 5, 1,
0, 4, 8 };
int32_t n_bytes = sizeof(bytes);
int32_t chunksize = 3;
int32_t n_chunks = n_bytes/chunksize + (n_bytes%chunksize ? 1 : 0);
/* chunkify */
uint8_t chunks[n_chunks][chunksize];
memset(chunks, 0, sizeof(uint8_t[n_chunks][chunksize]));
memcpy(chunks, bytes, n_bytes);
/* print result */
size_t i, j;
for (i = 0; i < n_chunks; i++)
{
for (j = 0; j < chunksize; j++)
printf("%02hhd ", chunks[i][j]);
printf("\n");
}
return 0;
}
The output is:
16 02 17
05 03 02
22 01 02
01 00 00
00 04 00
00 03 00
00 05 01
00 04 08

Answer 6

I used some of the examples here to come up with a solution, so I thought I'd share it. It could be a basis for an interface so that objects can transmit copies of themselves over a network with the hton and ntoh functions, which is actually what I am trying to do.
#include <iostream>
#include <string>
#include <exception>
#include <arpa/inet.h>
using namespace std;
void DispLength(string name, size_t var){
cout << "The size of " << name << " is : " << var << endl;
}
typedef int8_t byte;
class Bytes {
public:
Bytes(void* data_ptr, size_t size)
: size_(size)
{ this->bytes_ = (byte*)data_ptr; }
~Bytes(){ bytes_ = NULL; } // Caller is responsible for data deletion.
const byte& operator[] (int idx){
if((size_t)idx <= size_ && idx >= 0)
return bytes_[idx];
else
throw exception();
}
int32_t ret32(int idx) //-- Return a 32 bit value starting at index idx
{
int32_t* ret_ptr = (int32_t*)&((*this)[idx]);
int32_t ret = *ret_ptr;
return ret;
}
int64_t ret64(int idx) //-- Return a 64 bit value starting at index idx
{
int64_t* ret_ptr = (int64_t*)&((*this)[idx]);
int64_t ret = *ret_ptr;
return ret;
}
template <typename T>
T retVal(int idx) //-- Return a value of type T starting at index idx
{
T* T_ptr = (T*)&((*this)[idx]);
T T_ret = *T_ptr;
return T_ret;
}
protected:
Bytes() : bytes_(NULL), size_(0) {}
private:
byte* bytes_; //-- pointer used to scan for bytes
size_t size_;
};
int main(int argc, char** argv){
long double LDouble = 1.0;
Bytes bytes(&LDouble, sizeof(LDouble));
DispLength(string("LDouble"), sizeof(LDouble));
DispLength(string("bytes"), sizeof(bytes));
cout << "As a long double LDouble is " << LDouble << endl;
for( int i = 0; i < 16; i++){
cout << "Byte " << i << " : " << bytes[i] << endl;
}
cout << "Through the eyes of bytes : " <<
(long double) bytes.retVal<long double>(0) << endl;
return 0;
}

Answer 7

you can use bit manipulation operators
I would use, not following actual code, just show example
(for I =0 until 7){
temp val = Value && 111 //AND operation with 111
Value = Value >> 3; //to shift right
}

Answer 8

Some self documenting, maintainable code might look something like this (untested).
typedef union
{
struct {
uint8_t padding;
uint8_t value[3];
} raw;
int32_t data;
} Measurement;
void convert(uint8_t* rawValues, int32_t* convertedValues, int convertedSize)
{
Measurement sample;
int i;
memset(&sample, '\0', sizeof(sample));
for(i=0; i<convertedSize; ++i)
{
memcpy(&sample.raw.value[0], &rawValues[i*sizeof(sample.raw.value)], sizeof(sample.raw.value));
convertedValues[i]=sample.data;
}
}