def _get_maskpoly_from_row(self, table_row):
"""
Get a mask polygon from a table row
Parameters
----------
table_row : `np.ndarray`
Row of a mask table with RAs and Decs
Returns
-------
maskpoly : `healsparse.Polygon`
"""
mask_poly = healsparse.Polygon(ra=np.array([
table_row['ra_1'], table_row['ra_2'], table_row['ra_3'],
table_row['ra_4']
]),
dec=np.array([
table_row['dec_1'],
table_row['dec_2'],
table_row['dec_3'],
table_row['dec_4']
]),
value=1)
return mask_poly
def load_polygons(*, data, values, bands=None):
"""
load a set of polygons (rectangles) from the input
data. EDGEBLEED is skipped for 'u' and 'Y' bands
Parameters
----------
data: array with fields
Must have
ra_1, dec_1
ra_2, dec_2
ra_3, dec_3
ra_4, dec_4
badpix
"""
values = _extract_values(values, data.size)
has_bands = 'band' in data.dtype.names
has_badpix = 'badpix' in data.dtype.names
if bands is not None and not has_bands:
raise ValueError('bands= sent but no bands present in input data')
EDGEBLEED = 128
polygons = []
for i in range(data.size):
idata = data[i]
if has_bands and bands is not None:
band = idata['band'].strip()
if band not in bands:
continue
if (has_bands and has_badpix and band in ['u', 'Y']
and idata['badpix'] == EDGEBLEED):
print('skipping %s EDGEBLEED' % band)
continue
ra, dec = _extract_vert(idata)
polygon = hs.Polygon(
ra=ra,
dec=dec,
value=values[i],
)
polygons.append(polygon)
return polygons
def _render_catalog(self, maskmap, maskcat, indices):
"""
Render part of a maskcat into a mask map.
Parameters
----------
maskmap : `healsparse.HealSparseMap`
The map must be of `healsparse.WIDE_MASK` type
maskcat : `np.ndarray`
Catalog of region information
indices : `np.ndarray`
Array of indices from maskcat to render
"""
box = np.isnan(maskcat['radius'][indices])
circ = ~box
ra = maskcat['ra'][indices[circ]]
dec = maskcat['dec'][indices[circ]]
radius = maskcat['radius'][indices[circ]]
shapes = []
for i in range(ra.size):
shapes.append(
healsparse.Circle(ra=ra[i],
dec=dec[i],
radius=radius[i],
value=[self.config.default_bit]))
ra = maskcat['ra'][indices[box]]
dec = maskcat['dec'][indices[box]]
width = maskcat['width'][indices[box]] / np.cos(
np.deg2rad(maskcat['dec'][indices[box]]))
height = maskcat['height'][indices[box]]
for i in range(ra.size):
shapes.append(
healsparse.Polygon(ra=[
ra[i] - width[i] / 2., ra[i] - width[i] / 2.,
ra[i] + width[i] / 2., ra[i] + width[i] / 2.
],
dec=[
dec[i] - height[i] / 2.,
dec[i] + height[i] / 2.,
dec[i] + height[i] / 2.,
dec[i] - height[i] / 2.
],
value=[self.config.default_bit]))
healsparse.realize_geom(shapes, maskmap)
def _make_nexp_map(self):
"""
Make the number-of-exposures map
Returns
-------
nexp_map : `healsparse.HealSparseMap`
"""
# Figure out coverage
ipnest = hp.ang2pix(self.config.nside_coverage,
self.centers['ra_center'], self.centers['dec_center'],
nest=True, lonlat=True)
pixels = np.unique(ipnest)
# Initialize the map and memory
nexp_map = healsparse.HealSparseMap.make_empty(self.config.nside_coverage,
self.config.nside,
np.int32,
sentinel=0,
cov_pixels=pixels)
bad = ((self.table['naxis1'] < self.config.border*2) |
(self.table['naxis2'] < self.config.border*2))
# Loop over boxes
for i, wcs in enumerate(self.wcs_list):
if bad[i]:
continue
x_coords = np.array([self.config.border,
self.config.border,
self.table['naxis1'][i] - self.config.border,
self.table['naxis1'][i] - self.config.border])
y_coords = np.array([self.config.border,
self.table['naxis2'][i] - self.config.border,
self.table['naxis2'][i] - self.config.border,
self.config.border])
ra, dec = wcs.image2sky(x_coords, y_coords)
poly = healsparse.Polygon(ra=ra, dec=dec,
value=1)
try:
nexp_map[poly.get_pixels(nside=nexp_map.nside_sparse)] += 1
except ValueError:
print('Bad WCS mapping for %d/%d' %
(self.table[self.config.exp_field][i],
self.table[self.config.ccd_field][i]))
return nexp_map
def _extract(self):
ra = []
dec = []
npair = len(self.data) // 2
ra = np.zeros(npair)
dec = np.zeros(npair)
for i in range(npair):
ira = i * 2
idec = i * 2 + 1
ra[i] = self.data[ira]
dec[i] = self.data[idec]
self._geom = hs.Polygon(
ra=ra,
dec=dec,
value=self.value,
)
def worker(image, return_dict, procnum):
#Read image header and get coordinates of the image
print(i, 'working with image', image)
hdr = pf.open(image)
w = astrowcs.WCS(hdr[1].header)
corners_image = w.calc_footprint(center=False)
corners_pixels = w.calc_footprint(center=True)
nx = hdr[1].header['naxis1']
ny = hdr[1].header['naxis2']
coords1 = corners_image[0]
coords2 = corners_image[1]
coords3 = corners_image[2]
coords4 = corners_image[3]
#Read image data
image_data = pf.getdata(image)
#Define healsparse polygon within the image
ra = [coords1[0], coords4[0], coords3[0], coords2[0]]
dec = [coords1[1], coords4[1], coords3[1], coords2[1]]
raC = (np.max(ra) + np.min(ra)) / 2.
decC = (np.max(dec) + np.min(dec)) / 2.
nside = 2**17
poly = hs.Polygon(ra=ra, dec=dec, value=1)
smap = poly.get_map(nside=nside, dtype=np.int16)
a = smap.validPixels
b = hp.nest2ring(nside, a)
#a are pixels in NEST, b in RING
#Get center coordinates of each pixel
raF, decF = hp.pix2ang(nside, a, lonlat=True, nest=True)
#Get nx, ny in image from healsparse pixels
myhdr = hdr[1].header
wcs = wc.WCS(myhdr)
xn, yn = wcs.sky2image(raF, decF)
#Get associated weight values
values = []
for x, y in zip(xn, yn):
values.append(image_data[int(y - 1), int(x - 1)])
values = np.array(values)
#Define healsparse map
hsp_map_2 = hs.HealSparseMap.makeEmpty(512, nside, dtype=np.int16)
hsp_map_2.updateValues(a, values)
#Degrade to nside=4096
low_res_hsp = hsp_map_2.degrade(4096)
j = low_res_hsp.validPixels
test_values_2 = low_res_hsp.getValuePixel(j)
#Uncomment if you want in RING format
#k=hp.nest2ring(4096,j)
#hp_aux = np.zeros(hp.nside2npix(4096))+hp.UNSEEN
#hp_aux[j] = test_values_2
hdr.close()
return_dict[procnum] = np.array([j, test_values_2]).transpose()
def build_region_input_map(self, indices, tilename=None, hpix=None):
"""
Build input map for a given tile or hpix.
Must specify tilename or hpix.
Parameters
----------
indices : `np.ndarray`
Indices of WCS table
tilename : `str`, optional
Name of tile
hpix : `int`, optional
Healpix number
Returns
-------
input_map : `healsparse.HealSparseMap`
Wide mask map
"""
if tilename is None and hpix is None:
raise RuntimeError("Must specify one of tilename or hpix.")
if tilename is not None and hpix is not None:
raise RuntimeError("Must specify one of tilename or hpix.")
if self.config.use_two_amps:
maxbits = len(indices) * 2
else:
maxbits = len(indices)
region_input_map = healsparse.HealSparseMap.make_empty(
nside_coverage=self.nside_coverage_region,
nside_sparse=self.config.nside,
dtype=healsparse.WIDE_MASK,
wide_mask_maxbits=maxbits)
# Pre-match mask tables if available
expnums = np.unique(dg.table[self.config.exp_field][indices])
if dg.streak_table is not None:
_, streak_b = esutil.numpy_util.match(
expnums, dg.streak_table[self.config.exp_field])
if dg.bleed_table is not None:
_, bleed_b = esutil.numpy_util.match(
expnums, dg.bleed_table[self.config.exp_field])
if dg.satstar_table is not None:
_, satstar_b = esutil.numpy_util.match(
expnums, dg.satstar_table[self.config.exp_field])
metadata = {}
for i, ind in enumerate(indices):
if self.config.use_two_amps:
# Everything needs to be done with 2 amps
bit_a = i * 2
bit_b = i * 2 + 1
metadata['B%04dCCD' %
(bit_a)] = dg.table[self.config.ccd_field][ind]
metadata['B%04dCCD' %
(bit_b)] = dg.table[self.config.ccd_field][ind]
metadata['B%04dEXP' %
(bit_a)] = dg.table[self.config.exp_field][ind]
metadata['B%04dEXP' %
(bit_b)] = dg.table[self.config.exp_field][ind]
metadata['B%04dAMP' % (bit_a)] = 'A'
metadata['B%04dAMP' % (bit_b)] = 'B'
metadata['B%04dWT' % (bit_a)] = self._compute_weight(
self.config.skyvar_field + 'a', ind)
metadata['B%04dWT' % (bit_b)] = self._compute_weight(
self.config.skyvar_field + 'b', ind)
else:
bit = i
metadata['B%04dCCD' %
(bit)] = dg.table[self.config.ccd_field][ind]
metadata['B%04dEXP' %
(bit)] = dg.table[self.config.exp_field][ind]
metadata['B%04dWT' % (bit)] = self._compute_weight(
self.config.skyvar_field, ind)
wcs = dg.wcs_list[ind]
if wcs is None:
continue
if self.config.use_wcs:
# Use the WCS and render the ccd boxes
if self.config.use_two_amps:
x_coords_a = np.array([
self.config.amp_boundary, self.config.amp_boundary,
dg.table['naxis1'][ind] - self.config.border,
dg.table['naxis1'][ind] - self.config.border
])
x_coords_b = np.array([
self.config.border, self.config.border,
self.config.amp_boundary, self.config.amp_boundary
])
else:
x_coords = np.array([
self.config.border, self.config.border,
dg.table['naxis1'][ind] - self.config.border,
dg.table['naxis1'][ind] - self.config.border
])
y_coords = np.array([
self.config.border,
dg.table['naxis2'][ind] - self.config.border,
dg.table['naxis2'][ind] - self.config.border,
self.config.border
])
if self.config.use_two_amps:
ra_a, dec_a = wcs.image2sky(x_coords_a, y_coords)
ra_b, dec_b = wcs.image2sky(x_coords_b, y_coords)
poly_a = healsparse.Polygon(ra=ra_a,
dec=dec_a,
value=[bit_a])
poly_b = healsparse.Polygon(ra=ra_b,
dec=dec_b,
value=[bit_b])
else:
ra, dec = wcs.image2sky(x_coords, y_coords)
poly = healsparse.Polygon(ra=ra, dec=dec, value=[bit])
else:
# Don't use the WCS and use the bounding box specified in the table.
# This is only possible with 1-amp mode.
ra = np.array([
dg.table[field][ind]
for field in self.config.ra_corner_fields
])
dec = np.array([
dg.table[field][ind]
for field in self.config.dec_corner_fields
])
poly = healsparse.Polygon(ra=ra, dec=dec, value=[bit])
# Check if we have additional masking
if (dg.streak_table is not None or dg.bleed_table is not None
or dg.satstar_table is not None):
if self.config.use_two_amps:
poly_map_a = poly_a.get_map_like(region_input_map)
poly_map_b = poly_b.get_map_like(region_input_map)
else:
poly_map = poly.get_map_like(region_input_map)
mask_reg_list = []
if dg.streak_table is not None:
sinds, = np.where(
(dg.streak_table[self.config.exp_field][streak_b] ==
dg.table[self.config.exp_field][ind])
& (dg.streak_table[self.config.ccd_field][streak_b] ==
dg.table[self.config.ccd_field][ind]))
for sind in streak_b[sinds]:
mask_reg_list.append(
self._get_maskpoly_from_row(dg.streak_table[sind]))
if dg.bleed_table is not None:
binds, = np.where(
(dg.bleed_table[self.config.exp_field][bleed_b] ==
dg.table[self.config.exp_field][ind])
& (dg.bleed_table[self.config.ccd_field][bleed_b] ==
dg.table[self.config.ccd_field][ind]))
for bind in bleed_b[binds]:
mask_reg_list.append(
self._get_maskpoly_from_row(dg.bleed_table[bind]))
if dg.satstar_table is not None:
sinds, = np.where(
(dg.satstar_table[self.config.exp_field][satstar_b] ==
dg.table[self.config.exp_field][ind])
& (dg.satstar_table[self.config.ccd_field][satstar_b]
== dg.table[self.config.ccd_field][ind]))
for sind in satstar_b[sinds]:
mask_reg_list.append(
self._get_maskcircle_from_row(
dg.satstar_table[sind]))
mask_map = healsparse.HealSparseMap.make_empty(
nside_coverage=self.nside_coverage_region,
nside_sparse=self.config.nside,
dtype=np.uint8)
healsparse.realize_geom(mask_reg_list, mask_map)
if self.config.use_two_amps:
poly_map_a.apply_mask(mask_map)
poly_map_b.apply_mask(mask_map)
pixels_a = poly_map_a.valid_pixels
pixels_b = poly_map_b.valid_pixels
else:
poly_map.apply_mask(mask_map)
pixels = poly_map.valid_pixels
else:
# With no masking we can do a slightly faster version direct
# with the pixels
if self.config.use_two_amps:
pixels_a = poly_a.get_pixels(nside=self.config.nside)
pixels_b = poly_b.get_pixels(nside=self.config.nside)
else:
pixels = poly.get_pixels(nside=self.config.nside)
# Check for bad amps -- only in two amp mode
if self.config.use_two_amps:
if int(dg.table[self.config.ccd_field][ind]) in list(
self.config.bad_amps):
ba = self.config.bad_amps[int(
dg.table[self.config.ccd_field][ind])]
for b in ba:
if b.lower() == 'a':
pixels_a = np.array([], dtype=np.int64)
elif b.lower() == 'b':
pixels_b = np.array([], dtype=np.int64)
if int(dg.table[self.config.ccd_field]
[ind]) in self.config.bad_ccds:
if self.config.use_two_amps:
pixels_a = np.array([], dtype=np.int64)
pixels_b = np.array([], dtype=np.int64)
else:
pixels = np.array([], dtype=np.int64)
# tilename implies DES, and use two amps
if tilename is not None and self.config.use_two_amps:
tind, = np.where(dg.tile_info['tilename'] == tilename)
for pixels, bit in zip([pixels_a, pixels_b], [bit_a, bit_b]):
pixra, pixdec = hp.pix2ang(self.config.nside,
pixels,
lonlat=True,
nest=True)
if dg.tile_info['crossra0'][tind] == 'Y':
# Special for cross-ra0, where uramin will be very large
uramin = dg.tile_info['uramin'][tind] - 360.0
pixra_rot = pixra.copy()
hi, = np.where(pixra > 180.0)
pixra_rot[hi] -= 360.0
ok = ((pixra_rot > uramin) &
(pixra_rot < dg.tile_info['uramax'][tind]) &
(pixdec > dg.tile_info['udecmin'][tind]) &
(pixdec <= dg.tile_info['udecmax'][tind]))
else:
ok = ((pixra > dg.tile_info['uramin'][tind]) &
(pixra <= dg.tile_info['uramax'][tind]) &
(pixdec > dg.tile_info['udecmin'][tind]) &
(pixdec <= dg.tile_info['udecmax'][tind]))
region_input_map.set_bits_pix(pixels[ok], [bit])
else:
# healpix mode
bit_shift = 2 * int(
np.round(np.log2(
self.config.nside / self.config.nside_run)))
npixels = 2**bit_shift
pixel_min = hpix * npixels
pixel_max = (hpix + 1) * npixels - 1
if self.config.use_two_amps:
for pixels, bit in zip([pixels_a, pixels_b],
[bit_a, bit_b]):
ok = ((pixels >= pixel_min) & (pixels <= pixel_max))
region_input_map.set_bits_pix(pixels[ok], [bit])
else:
ok = ((pixels >= pixel_min) & (pixels <= pixel_max))
region_input_map.set_bits_pix(pixels[ok], [bit])
region_input_map.metadata = metadata
return region_input_map
def run(self, sky_map, tract, band, coadd_dict, input_map_dict, visit_summary_dict):
"""Run the healsparse property task.
Parameters
----------
sky_map : Sky map object
tract : `int`
Tract number.
band : `str`
Band name for logging.
coadd_dict : `dict` [`int`: `lsst.daf.butler.DeferredDatasetHandle`]
Dictionary of coadd exposure references. Keys are patch numbers.
input_map_dict : `dict` [`int`: `lsst.daf.butler.DeferredDatasetHandle`]
Dictionary of input map references. Keys are patch numbers.
visit_summary_dict : `dict` [`int`: `lsst.afw.table.ExposureCatalog`]
Dictionary of visit summary tables. Keys are visit numbers.
Raises
------
RepeatableQuantumError
If visit_summary_dict is missing any visits or detectors found in an
input map. This leads to an inconsistency between what is in the coadd
(via the input map) and the visit summary tables which contain data
to compute the maps.
"""
tract_info = sky_map[tract]
tract_maps_initialized = False
for patch in input_map_dict.keys():
self.log.info("Making maps for band %s, tract %d, patch %d.",
band, tract, patch)
patch_info = tract_info[patch]
input_map = input_map_dict[patch].get()
coadd_photo_calib = coadd_dict[patch].get(component="photoCalib")
coadd_inputs = coadd_dict[patch].get(component="coaddInputs")
coadd_zeropoint = 2.5*np.log10(coadd_photo_calib.getInstFluxAtZeroMagnitude())
# Crop input_map to the inner polygon of the patch
poly_vertices = patch_info.getInnerSkyPolygon(tract_info.getWcs()).getVertices()
patch_radec = self._vertices_to_radec(poly_vertices)
patch_poly = hsp.Polygon(ra=patch_radec[:, 0], dec=patch_radec[:, 1],
value=np.arange(input_map.wide_mask_maxbits))
patch_poly_map = patch_poly.get_map_like(input_map)
input_map = hsp.and_intersection([input_map, patch_poly_map])
if not tract_maps_initialized:
# We use the first input map nside information to initialize
# the tract maps
nside_coverage = self._compute_nside_coverage_tract(tract_info)
nside = input_map.nside_sparse
do_compute_approx_psf = False
# Initialize the tract maps
for property_map in self.property_maps:
property_map.initialize_tract_maps(nside_coverage, nside)
if property_map.requires_psf:
do_compute_approx_psf = True
tract_maps_initialized = True
valid_pixels, vpix_ra, vpix_dec = input_map.valid_pixels_pos(return_pixels=True)
# Check if there are no valid pixels for the inner (unique) patch region
if valid_pixels.size == 0:
continue
# Initialize the value accumulators
for property_map in self.property_maps:
property_map.initialize_values(valid_pixels.size)
property_map.zeropoint = coadd_zeropoint
# Initialize the weight and counter accumulators
total_weights = np.zeros(valid_pixels.size)
total_inputs = np.zeros(valid_pixels.size, dtype=np.int32)
for bit, ccd_row in enumerate(coadd_inputs.ccds):
# Which pixels in the map are used by this visit/detector
inmap, = np.where(input_map.check_bits_pix(valid_pixels, [bit]))
# Check if there are any valid pixels in the map from this deteector.
if inmap.size == 0:
continue
# visit, detector_id, weight = input_dict[bit]
visit = ccd_row["visit"]
detector_id = ccd_row["ccd"]
weight = ccd_row["weight"]
x, y = ccd_row.getWcs().skyToPixelArray(vpix_ra[inmap], vpix_dec[inmap], degrees=True)
scalings = self._compute_calib_scale(ccd_row, x, y)
if do_compute_approx_psf:
psf_array = compute_approx_psf_size_and_shape(ccd_row, vpix_ra[inmap], vpix_dec[inmap])
else:
psf_array = None
total_weights[inmap] += weight
total_inputs[inmap] += 1
# Retrieve the correct visitSummary row
if visit not in visit_summary_dict:
msg = f"Visit {visit} not found in visit_summaries."
raise pipeBase.RepeatableQuantumError(msg)
row = visit_summary_dict[visit].find(detector_id)
if row is None:
msg = f"Visit {visit} / detector_id {detector_id} not found in visit_summaries."
raise pipeBase.RepeatableQuantumError(msg)
# Accumulate the values
for property_map in self.property_maps:
property_map.accumulate_values(inmap,
vpix_ra[inmap],
vpix_dec[inmap],
weight,
scalings,
row,
psf_array=psf_array)
# Finalize the mean values and set the tract maps
for property_map in self.property_maps:
property_map.finalize_mean_values(total_weights, total_inputs)
property_map.set_map_values(valid_pixels)
def build_ccd_input_map(self, bbox, wcs, ccds):
"""Build a map from ccd valid polygons or bounding boxes.
Parameters
----------
bbox : `lsst.geom.Box2I`
Bounding box for region to build input map.
wcs : `lsst.afw.geom.SkyWcs`
WCS object for region to build input map.
ccds : `lsst.afw.table.ExposureCatalog`
Exposure catalog with ccd data from coadd inputs.
"""
with warnings.catch_warnings():
# Healsparse will emit a warning if nside coverage is greater than
# 128. In the case of generating patch input maps, and not global
# maps, high nside coverage works fine, so we can suppress this
# warning.
warnings.simplefilter("ignore")
self.ccd_input_map = hsp.HealSparseMap.make_empty(nside_coverage=self.config.nside_coverage,
nside_sparse=self.config.nside,
dtype=hsp.WIDE_MASK,
wide_mask_maxbits=len(ccds))
self._wcs = wcs
self._bbox = bbox
self._ccds = ccds
pixel_scale = wcs.getPixelScale().asArcseconds()
hpix_area_arcsec2 = hp.nside2pixarea(self.config.nside, degrees=True)*(3600.**2.)
self._min_bad = self.config.bad_mask_min_coverage*hpix_area_arcsec2/(pixel_scale**2.)
metadata = {}
self._bits_per_visit_ccd = {}
self._bits_per_visit = defaultdict(list)
for bit, ccd_row in enumerate(ccds):
metadata[f"B{bit:04d}CCD"] = ccd_row["ccd"]
metadata[f"B{bit:04d}VIS"] = ccd_row["visit"]
metadata[f"B{bit:04d}WT"] = ccd_row["weight"]
self._bits_per_visit_ccd[(ccd_row["visit"], ccd_row["ccd"])] = bit
self._bits_per_visit[ccd_row["visit"]].append(bit)
ccd_poly = ccd_row.getValidPolygon()
if ccd_poly is None:
ccd_poly = afwGeom.Polygon(lsst.geom.Box2D(ccd_row.getBBox()))
# Detectors need to be rendered with their own wcs.
ccd_poly_radec = self._pixels_to_radec(ccd_row.getWcs(), ccd_poly.convexHull().getVertices())
# Create a ccd healsparse polygon
poly = hsp.Polygon(ra=ccd_poly_radec[: -1, 0],
dec=ccd_poly_radec[: -1, 1],
value=[bit])
self.ccd_input_map.set_bits_pix(poly.get_pixels(nside=self.ccd_input_map.nside_sparse),
[bit])
# Cut down to the overall bounding box with associated wcs.
bbox_afw_poly = afwGeom.Polygon(lsst.geom.Box2D(bbox))
bbox_poly_radec = self._pixels_to_radec(self._wcs,
bbox_afw_poly.convexHull().getVertices())
bbox_poly = hsp.Polygon(ra=bbox_poly_radec[: -1, 0], dec=bbox_poly_radec[: -1, 1],
value=np.arange(self.ccd_input_map.wide_mask_maxbits))
bbox_poly_map = bbox_poly.get_map_like(self.ccd_input_map)
self.ccd_input_map = hsp.and_intersection([self.ccd_input_map, bbox_poly_map])
self.ccd_input_map.metadata = metadata
# Create a temporary map to hold the count of bad pixels in each healpix pixel
self._ccd_input_pixels = self.ccd_input_map.valid_pixels
dtype = [(f"v{visit}", np.int64) for visit in self._bits_per_visit.keys()]
with warnings.catch_warnings():
# Healsparse will emit a warning if nside coverage is greater than
# 128. In the case of generating patch input maps, and not global
# maps, high nside coverage works fine, so we can suppress this
# warning.
warnings.simplefilter("ignore")
self._ccd_input_bad_count_map = hsp.HealSparseMap.make_empty(
nside_coverage=self.config.nside_coverage,
nside_sparse=self.config.nside,
dtype=dtype,
primary=dtype[0][0])
# Don't set input bad map if there are no ccds which overlap the bbox.
if len(self._ccd_input_pixels) > 0:
self._ccd_input_bad_count_map[self._ccd_input_pixels] = np.zeros(1, dtype=dtype)
def setUp(self):
tract = 0
band = 'r'
patch = 0
visits = [100, 101]
# Good to test crossing 0.
ra_center = 0.0
dec_center = -45.0
pixel_scale = 0.2
coadd_zp = 27.0
# Generate a mock skymap with one patch
config = DiscreteSkyMap.ConfigClass()
config.raList = [ra_center]
config.decList = [dec_center]
config.radiusList = [150 * pixel_scale / 3600.]
config.patchInnerDimensions = (350, 350)
config.patchBorder = 50
config.tractOverlap = 0.0
config.pixelScale = pixel_scale
sky_map = DiscreteSkyMap(config)
visit_summaries = [
makeMockVisitSummary(visit,
ra_center=ra_center,
dec_center=dec_center) for visit in visits
]
visit_summary_refs = [
MockVisitSummaryReference(visit_summary, visit)
for visit_summary, visit in zip(visit_summaries, visits)
]
self.visit_summary_dict = {
visit: ref.get()
for ref, visit in zip(visit_summary_refs, visits)
}
# Generate an input map. Note that this does not need to be consistent
# with the visit_summary projections, we're just tracking values.
input_map = hsp.HealSparseMap.make_empty(
nside_coverage=256,
nside_sparse=32768,
dtype=hsp.WIDE_MASK,
wide_mask_maxbits=len(visits) * 2)
patch_poly = afwGeom.Polygon(
geom.Box2D(sky_map[tract][patch].getOuterBBox()))
sph_pts = sky_map[tract].getWcs().pixelToSky(
patch_poly.convexHull().getVertices())
patch_poly_radec = np.array([(sph.getRa().asDegrees(),
sph.getDec().asDegrees())
for sph in sph_pts])
poly = hsp.Polygon(ra=patch_poly_radec[:-1, 0],
dec=patch_poly_radec[:-1, 1],
value=[0])
poly_pixels = poly.get_pixels(nside=input_map.nside_sparse)
# The input map has full coverage for bits 0 and 1
input_map.set_bits_pix(poly_pixels, [0])
input_map.set_bits_pix(poly_pixels, [1])
input_map_ref = MockInputMapReference(input_map,
patch=patch,
tract=tract)
self.input_map_dict = {patch: input_map_ref}
coadd = afwImage.ExposureF(sky_map[tract][patch].getOuterBBox(),
sky_map[tract].getWcs())
instFluxMag0 = 10.**(coadd_zp / 2.5)
pc = afwImage.makePhotoCalibFromCalibZeroPoint(instFluxMag0)
coadd.setPhotoCalib(pc)
# Mock the coadd input ccd table
schema = afwTable.ExposureTable.makeMinimalSchema()
schema.addField("ccd", type="I")
schema.addField("visit", type="I")
schema.addField("weight", type="F")
ccds = afwTable.ExposureCatalog(schema)
ccds.resize(2)
ccds['id'] = np.arange(2)
ccds['visit'][0] = visits[0]
ccds['visit'][1] = visits[1]
ccds['ccd'][0] = 0
ccds['ccd'][1] = 1
ccds['weight'] = 10.0
for ccd_row in ccds:
summary = self.visit_summary_dict[ccd_row['visit']].find(
ccd_row['ccd'])
ccd_row.setWcs(summary.getWcs())
ccd_row.setPsf(summary.getPsf())
ccd_row.setBBox(summary.getBBox())
ccd_row.setPhotoCalib(summary.getPhotoCalib())
inputs = afwImage.CoaddInputs()
inputs.ccds = ccds
coadd.getInfo().setCoaddInputs(inputs)
coadd_ref = MockCoaddReference(coadd, patch=patch, tract=tract)
self.coadd_dict = {patch: coadd_ref}
self.tract = tract
self.band = band
self.sky_map = sky_map
self.input_map = input_map
def build_patch_input_map(self, butler, tract_wcs, patch_info, ccds,
nside_coverage_patch):
"""
Build the patch input map.
Parameters
----------
butler : `lsst.daf.persistence.Butler`
gen2 butler
tract_wcs : `lsst.afw.geom.SkyWcs`
WCS object for the tract
patch_info : `lsst.skymap.PatchInfo`
Patch info object
ccds : `lsst.afw.table.ExposureCatalog`
Catalog of ccd information
nside_coverage_patch : `int`
Healpix nside for the coverage map
Returns
-------
patch_input_map : `healsparse.HealSparseMap`
Healsparse map encoding input ccd information.
"""
patch_input_map = healsparse.HealSparseMap.make_empty(
nside_coverage=nside_coverage_patch,
nside_sparse=self.config.nside,
dtype=healsparse.WIDE_MASK,
wide_mask_maxbits=len(ccds))
if self.config.use_calexp_mask:
# pixel_scale = tract_wcs.getPixelScale().asArcseconds()
hpix_area_arcsec2 = hp.nside2pixarea(patch_input_map.nside_sparse,
degrees=True) * (3600.**2.)
bad_mask = afwImage.Mask.getPlaneBitMask(
self.config.bad_mask_planes)
metadata = {}
for bit, ccd in enumerate(ccds):
metadata['B%04dCCD' % (bit)] = ccd['ccd']
metadata['B%04dVIS' % (bit)] = ccd['visit']
metadata['B%04dWT' % (bit)] = ccd['weight']
wcs = ccd.getWcs()
ccd_poly = ccd.getValidPolygon()
if ccd_poly is None:
ccd_poly = afwGeom.Polygon(lsst.geom.Box2D(ccd.getBBox()))
ccd_poly_radec = pixels_to_radec(
wcs,
ccd_poly.convexHull().getVertices())
# Use polygons for all of these
poly = healsparse.Polygon(ra=ccd_poly_radec[:-1, 0],
dec=ccd_poly_radec[:-1, 1],
value=[bit])
poly_map = poly.get_map_like(patch_input_map)
dataId = {
self.config.detector_id_name: int(ccd['ccd']),
self.config.visit_id_name: int(ccd['visit'])
}
if self.config.use_calexp_mask:
calexp = butler.get('calexp', dataId=dataId)
calexp_metadata = calexp.getMetadata()
mask = calexp.getMask()
pixel_scale = tract_wcs.getPixelScale().asArcseconds()
min_bad = self.config.bad_mask_coverage * hpix_area_arcsec2 / (
pixel_scale**2.)
bad_pixels = np.where(mask.getArray() & bad_mask)
# Convert bad pixels to position
bad_radec = xy_to_radec(wcs, bad_pixels[1].astype(np.float64),
bad_pixels[0].astype(np.float64))
# Convert position to healpixel
bad_hpix = hp.ang2pix(patch_input_map.nside_sparse,
bad_radec[:, 0],
bad_radec[:, 1],
lonlat=True,
nest=True)
min_bad_hpix = bad_hpix.min()
bad_hpix_count = np.zeros(bad_hpix.max() - min_bad_hpix + 1,
dtype=np.int32)
np.add.at(bad_hpix_count, bad_hpix - min_bad_hpix, 1)
hpix_to_mask = min_bad_hpix + np.where(
bad_hpix_count > min_bad)[0]
poly_map.clear_bits_pix(hpix_to_mask, [bit])
if 'SKYLEVEL' not in calexp_metadata:
# We must recompute skylevel, skysigma
skylevel, skysigma = self._compute_skylevel(
butler, dataId, calexp)
else:
skylevel = calexp_metadata['SKYLEVEL']
skysigma = calexp_metadata['SKYSIGMA']
else:
calexp_metadata = butler.get('calexp_md', dataId=dataId)
if 'SKYLEVEL' not in calexp_metadata:
# We want to log this
skylevel = 0.0
skysigma = 0.0
else:
skylevel = calexp_metadata['SKYLEVEL']
skysigma = calexp_metadata['SKYSIGMA']
metadata['B%04dSLV' % (bit)] = skylevel
metadata['B%04dSSG' % (bit)] = skysigma
metadata['B%04dBGM' % (bit)] = calexp_metadata['BGMEAN']
metadata['B%04dBGV' % (bit)] = calexp_metadata['BGVAR']
# Now we have the full masked ccd map, set the appropriate bit
patch_input_map.set_bits_pix(poly_map.valid_pixels, [bit])
# Now cut down to the inner tract polygon
poly_vertices = patch_info.getInnerSkyPolygon(tract_wcs).getVertices()
patch_radec = vertices_to_radec(poly_vertices)
patch_poly = healsparse.Polygon(ra=patch_radec[:, 0],
dec=patch_radec[:, 1],
value=np.arange(
patch_input_map.wide_mask_maxbits))
# Realize the patch polygon
patch_poly_map = patch_poly.get_map_like(patch_input_map)
patch_input_map = healsparse.and_intersection(
[patch_input_map, patch_poly_map])
# And set the metadata
patch_input_map.metadata = metadata
return patch_input_map
