def generate_gaussian_line(err=1.0, h=10.0): xx = R.random.randn(1000)*err ii = R.random.randint(2, 997) print ii xc = R.arange(1000) eterm = -((xc-ii)**2)/(2.0*(2.5**2)) eterm = R.clip(eterm, -10.0, 0.0) xx += R.exp(eterm)*h return xx
def gauss(x, sigma): # Do all the derivatives at the same time eterm = -(x**2)/(2.0*sigma**2) eterm = R.clip(eterm, -10.0, 0.0) # Avoid nastiness invs2pi = 1./R.sqrt(2.0*R.pi) norm = invs2pi * (1./sigma) d0 = R.exp(eterm)*norm d1 = -x * d0/sigma**2 return (d0, d1)
def generate_derivatives(xx, sigma=1.0): # First set up the convolution dimensions npix = xx.shape[0] npix_full = npix*2 xx1 = R.zeros((npix_full, ), dtype='f8') xx1[0:npix] = xx[:] # Copy into padded array # Set up the coordinate arrays # We can happily misuse fftfreq here, for simplicity xc = R.fft.fftfreq(npix_full, 1./npix_full) # Ok, compute first and second derivative kernels # Note, this is not optimized for memory etc... d0p, d1p = gauss(xc+0.5, sigma) d0m, d1m = gauss(xc-0.5, sigma) kernel1 = d0p - d0m kernel2 = d1p - d1m # FFT everything in sight xx1_ft = R.fft.rfft(xx1) kernel1_ft = R.fft.rfft(kernel1) kernel2_ft = R.fft.rfft(kernel2) # r1 = R.fft.irfft(xx1_ft* kernel1_ft) r2 = R.fft.irfft(xx1_ft* kernel2_ft) return r1[0:npix], r2[0:npix]
def determine_oned_line_points(r1, r2, lo=0.05, eps=0.01): r2min = r2.min() w1 = R.nonzero(r2 < (lo*r2min)) xx = -r1[w1]/r2[w1] w2 = R.nonzero(R.absolute(xx) < (0.5+eps)) # Put in an epsilon correction return (w1[0])[w2]
import numpy as np
import mangle
import ROOTpy as R
npoints1 = 1000
ra1 = np.linspace(0.0, 360.0, num=npoints1, endpoint=False)
dec1 = np.linspace(-0.99999, 1.0, num=npoints1, endpoint=False)
dec1 = 90.0 - np.arccos(dec1) * 180.0 / np.pi
ra, dec = np.meshgrid(ra1, dec1)
ra = ra.flatten()
dec = dec.flatten()
foot = mangle.Mangle("boss_survey.ply")
ids = foot.get_polyids(ra, dec)
ra = ra[ids >= 0]
dec = dec[ids >= 0]
R.forprint("grid.txt", "%20.10e %20.10e\n", "w", ra, dec)
