Wrong Spectrum Computation

[00sapo]

LibXtract/src/vector.c

Lines 323 to 373 in 41273dd

	default:
	/* MAGNITUDE_SPECTRUM */
	for(n = 0, m = 0; m < M; ++n, ++m)
	{
	if(n==0 && !withDC) /* discard DC and keep Nyquist */
	{
	++n;
	}
	#ifdef USE_OOURA
	/*
	if(n==1 && withDC) // discard Nyquist
	{
	++n;
	}
	*/
	// OOURA discards the always 0 imaginary of DC and Nyquists
	if (n == M && !withDC)
	{
	real = fft[1];
	imag = 0.0;
	}
	else if (n == 0 && withDC) {
	real = fft[0];
	imag = 0.0;
	}
	else {
	real = fft[n * 2];
	imag = fft[n * 2 + 1];
	}
	#else
	if (n == M && !withDC)
	{
	real = fft->imagp[0];
	imag = 0.0;
	}
	else if (n == 0 && withDC) {
	real = fft->realp[0];
	imag = 0.0;
	}
	else {
	real = fft->realp[n];
	imag = fft->imagp[n];
	}
	#endif
	result[m*2] = real;
	result[m*2+1] = imag;
	XTRACT_SET_FREQUENCY;
	XTRACT_GET_MAX;
	}
	break;
	}

I was trying to implement a higher level API in Julia and I've found these two lines that seem wrong to me. Shouldn't it compute the euclidean distance between imag and real? The following two lines seems copied from the previous clause:

LibXtract/src/vector.c

Lines 367 to 368 in 41273dd

	result[m*2] = real;
	result[m*2+1] = imag;

If you confirm, I can do a pull request

[00sapo]

I noted that in master branch there is the following line:

*marker = sqrt(XTRACT_SQ(real) + XTRACT_SQ(imag)) / (double)N;

So, I think develop branch is wrong. I cannot compile master branch, though...