def __init__(self, data, srcpos, ndir=1, sys=sys):
# Build source parameter struct for PmatPtsrc
self.params = np.zeros([srcpos.shape[-1], ndir, 8], np.float)
self.params[:, :, :2] = srcpos[::-1, None, :].T
self.params[:, :, 5:7] = 1
scan = actscan.ACTScan(data.entry, d=data)
self.psrc = pmat.PmatPtsrc(scan, self.params, sys=sys)
self.pcut = pmat.PmatCut(scan)
# Extract basic offset
self.off0 = data.point_correction
self.off = self.off0 * 1
self.el = np.mean(data.boresight[2, ::100])
self.point_template = data.point_template
def __init__(self, scan, srcpos, ndir=1, perdet=False, sys="cel"):
# Build source parameter struct for PmatPtsrc
self.params = np.zeros([srcpos.shape[-1],ndir,scan.ndet if perdet else 1,8],np.float)
self.params[:,:,:,:2] = srcpos[::-1,None,None,:].T
self.params[:,:,:,5:7] = 1
self.psrc = pmat.PmatPtsrc(scan, self.params, sys=sys)
self.pcut = pmat.PmatCut(scan)
# Extract basic offset. Warning: referring to scan.d is fragile, since
# scan.d is not updated when scan is sliced
self.off0 = scan.d.point_correction
self.off = self.off0*0
self.el = np.mean(scan.boresight[::100,2])
self.point_template = scan.d.point_template
self.cut = scan.cut
def point_source_cut(d, srcs, thresholds=[]):
if len(thresholds) == 0: return []
# Sort-of-circular dependency here. I don't like
# how actdata datasets are incompatible with scans.
# Should I just replace scans with actdata objects?
from . import actscan
# Simulate sources
tod = np.zeros((d.ndet, d.nsamp), np.float32)
srcs = srcs.astype(np.float64)
scan = actscan.ACTScan(d.entry, d=d)
psrc = pmat.PmatPtsrc(scan, srcs)
psrc.forward(tod, srcs)
# Use them to define mask
cuts = [sampcut.from_mask(tod > tr) for tr in thresholds]
return cuts
except (IOError, OSError):
try:
d = actscan.ACTScan(db[filelist[ind]])
if d.ndet == 0 or d.nsamp == 0: raise errors.DataMissing("all samples cut")
except errors.DataMissing as e:
L.debug("Skipped %s (%s)" % (filelist[ind], str(e)))
continue
try:
L.debug("Reading samples")
tod = d.get_samples().astype(dtype)
except errors.DataMissing as e:
L.debug("Skipped %s (%s)" % (filelist[ind], str(e)))
continue
# Measure noise
L.debug("Noise")
ivar = 1/np.array([np.median(onlyfinite(get_desloped_var(blockify(t,20)))) for t in tod])
# Set up pmat for this scan
L.debug("Pmats")
d.noise = bunch.Bunch(ivar = ivar)
Psrc = pmat.PmatPtsrc(d, params.T)
# Extract point-source relevant TOD properties
L.debug("Extract data")
srcdata = Psrc.extract(tod, cut=d.cut, raw=not args.oldformat)
#srcdata.ivars = d.noise.iD[-1]
srcdata.ivars = ivar
L.debug("Writing")
ptsrc_data.write_srcscan(ofile, srcdata)
ivar = 1 / (np.median(np.mean(diff**2, -1), -1) / 2**0.5)
del diff
print "D", id, comm.rank
with bench.mark("actscan"):
scan = actscan.ACTScan(entry, d=d)
print "E", id, comm.rank
with bench.mark("pmat1"):
# Build effective source parameters for this TOD. This will ignore planet motion,
# but this should only be a problem for very near objects like the moon or ISS
src_param = np.zeros((len(srcs), 8))
src_param[:, 0] = srcs.dec
src_param[:, 1] = srcs.ra
src_param[:, 2] = 1
src_param[:, 5:7] = 1
# And use it to build our source model projector. This is only used for the cuts
psrc = pmat.PmatPtsrc(scan, src_param)
print "F", id, comm.rank
with bench.mark("source mask"):
# Find the samples where the sources live
src_mask = np.zeros(tod.shape, np.bool)
# Allow elongating the mask vertically
nypad = 1
dypad = R / 2
if ypad > 0:
nypad = int((2 * ypad) // dypad) + 1
dypad = (2 * ypad) / (nypad - 1)
# Hack: modify detector offsets to apply the y broadening
detoff = scan.offsets.copy()
src_tod = tod * 0
for yi in range(nypad):
yoff = -ypad + yi * dypad