Source code for desisim.simexp

from __future__ import absolute_import, division, print_function

import sys
import os.path
import warnings
import datetime, time

import numpy as np

import astropy.table
# See
import astropy.time
from import fits
import fitsio

import desitarget
import desitarget.targetmask
from desitarget.targets import main_cmx_or_sv
from desimodel.focalplane import fiber_area_arcsec2
import desiutil.depend
import desispec.interpolation
import desisim.specsim

#- Reference observing conditions for each of dark, gray, bright
reference_conditions = dict(DARK=dict(), GRAY=dict(), BRIGHT=dict())
reference_conditions['DARK']['SEEING']  = 1.1
reference_conditions['DARK']['EXPTIME'] = 1000
reference_conditions['DARK']['AIRMASS'] = 1.0
reference_conditions['DARK']['MOONFRAC'] = 0.0
reference_conditions['DARK']['MOONALT'] = -60
reference_conditions['DARK']['MOONSEP'] = 180

reference_conditions['GRAY']['SEEING']  = 1.1
reference_conditions['GRAY']['EXPTIME'] = 1000
reference_conditions['GRAY']['AIRMASS'] = 1.0
reference_conditions['GRAY']['MOONFRAC'] = 0.1
reference_conditions['GRAY']['MOONALT']  = 10
reference_conditions['GRAY']['MOONSEP'] = 60

reference_conditions['BRIGHT']['SEEING']  = 1.1
reference_conditions['BRIGHT']['EXPTIME'] = 300
reference_conditions['BRIGHT']['AIRMASS'] = 1.0
reference_conditions['BRIGHT']['MOONFRAC'] = 0.7
reference_conditions['BRIGHT']['MOONALT']  = 60
reference_conditions['BRIGHT']['MOONSEP'] = 50

for objtype in ('LRG', 'QSO', 'ELG'):
    reference_conditions[objtype] = reference_conditions['DARK']
for objtype in ('MWS', 'BGS'):
    reference_conditions[objtype] = reference_conditions['BRIGHT']

[docs]def simarc(arcdata, nspec=5000, nonuniform=False, testslit=False): ''' Simulates an arc lamp exposure Args: arcdata (Table): Table with columns VACUUM_WAVE and ELECTRONS nspec (int, optional) number of spectra to simulate nonuniform (bool, optional): include calibration screen non-uniformity testslit (bool, optional): this argument is undocumented. Returns: (wave, phot, fibermap) wave: 1D[nwave] wavelengths in Angstroms phot: 2D[nspec,nwave] photons observed by CCD (i.e. electrons) fibermap: fibermap Table Note: this bypasses specsim since we don't have an arclamp model in surface brightness units; we only have electrons on the CCD. But it does include the effect of varying fiber sizes. TODO: * add exptime support * update inputs to surface brightness and DESI lamp lines (DESI-2674) * add psfconvolve option ''' wave = arcdata['VACUUM_WAVE'] phot = arcdata['ELECTRONS'] if testslit: fibermap = astropy.table.Table(testslit_fibermap()[0:nspec]) else: fibermap = astropy.table.Table( fibermap.meta['FLAVOR'] = 'arc' fibermap['OBJTYPE'] = 'ARC' x = fibermap['FIBERASSIGN_X'] y = fibermap['FIBERASSIGN_Y'] r = np.sqrt(x**2 + y**2) #----- #- Determine ratio of fiber sizes relative to larges fiber fiber_area = fiber_area_arcsec2(x, y) size_ratio = fiber_area / np.max(fiber_area) #- Correct photons for fiber size phot = np.tile(phot, nspec).reshape(nspec, len(wave)) phot = (phot.T * size_ratio).T #- Apply calibration screen non-uniformity if nonuniform: ratio = _calib_screen_uniformity(radius=r) assert np.all(ratio <= 1) and np.all(ratio > 0.99) phot = (phot.T * ratio).T return wave, phot, fibermap
[docs]def simflat(flatfile, nspec=5000, nonuniform=False, exptime=10, testslit=False, psfconvolve=True, specsim_config_file="desi"): ''' Simulates a flat lamp calibration exposure Args: flatfile (str): filename with flat lamp spectrum data nspec (int, optional): number of spectra to simulate nonuniform (bool, optional): include calibration screen non-uniformity exptime (float, optional): exposure time in seconds psfconvolve (bool, optional): passed to simspec.simulator.Simulator camera_output. if True, convolve with PSF and include per-camera outputs specsim_config_file (str, optional): path to DESI instrument config file. default is desi config in specsim package. Returns: (sim, fibermap) sim: specsim Simulator object fibermap: fibermap Table ''' import astropy.units as u import specsim.simulator from desiutil.log import get_logger log = get_logger()'Reading flat lamp spectrum from {}'.format(flatfile)) sbflux, hdr = fits.getdata(flatfile, header=True) wave = assert len(wave) == len(sbflux) #- Trim to DESI wavelength ranges #- TODO: is there an easier way to get these parameters? try: params = wavemin = params['ccd']['b']['wavemin'] wavemax = params['ccd']['z']['wavemax'] except KeyError: wavemin ='b').wavemin wavemax ='z').wavemax ii = (wavemin <= wave) & (wave <= wavemax) wave = wave[ii] sbflux = sbflux[ii] #- Downsample to 0.2A grid to not blow up memory ww = np.arange(wave[0], wave[-1]+0.1, 0.2) sbflux = desispec.interpolation.resample_flux(ww, wave, sbflux) wave = ww if testslit: fibermap = astropy.table.Table(testslit_fibermap()[0:nspec]) else: fibermap = astropy.table.Table( fibermap.meta['FLAVOR'] = 'flat' fibermap['OBJTYPE'] = 'FLT' x = fibermap['FIBERASSIGN_X'] y = fibermap['FIBERASSIGN_Y'] r = np.sqrt(x**2 + y**2) xy = np.vstack([x, y]).T * #- Convert to unit-ful 2D sbunit = 1e-17 * u.erg / (u.Angstrom * u.s * ** 2 * u.arcsec ** 2) sbflux = np.tile(sbflux, nspec).reshape(nspec, len(wave)) * sbunit if nonuniform: ratio = _calib_screen_uniformity(radius=r) assert np.all(ratio <= 1) and np.all(ratio > 0.99) sbflux = (sbflux.T * ratio).T tmp = np.min(sbflux) / np.max(sbflux)'Adjusting for calibration screen non-uniformity {:.4f}'.format(tmp)) log.debug('Creating specsim configuration') config = _specsim_config_for_wave(wave,specsim_config_file=specsim_config_file) log.debug('Creating specsim simulator for {} spectra'.format(nspec)) # sim = specsim.simulator.Simulator(config, num_fibers=nspec) sim = desisim.specsim.get_simulator(config, num_fibers=nspec, camera_output=psfconvolve) sim.observation.exposure_time = exptime * u.s log.debug('Simulating') sim.simulate(calibration_surface_brightness=sbflux, focal_positions=xy) return sim, fibermap
[docs]def _calib_screen_uniformity(theta=None, radius=None): ''' Returns calibration screen relative non-uniformity as a function of theta (degrees) or focal plane radius (mm) ''' if theta is not None: assert radius is None #- Julien Guy fit to DESI-2761v1 figure 5 #- ratio lamp/sky = 1 - 9.4e-04*theta - 2.1e-03 * theta**2 return 1 - 9.4e-04*theta - 2.1e-03 * theta**2 elif radius is not None: import ps = theta = np.interp(radius, ps['radius'], ps['theta']) return _calib_screen_uniformity(theta=theta) else: raise ValueError('must provide theta or radius')
[docs]def simscience(targets, fiberassign, obsconditions='DARK', expid=None, nspec=None, psfconvolve=True): ''' Simulates a new DESI exposure from surveysim+fiberassign+mock spectra Args: targets (tuple): tuple of (flux[nspec,nwave], wave[nwave], meta[nspec]) fiberassign (Table): fiber assignments table obsconditions (object, optional): observation metadata as str: DARK (default) or GRAY or BRIGHT dict or row of Table with keys:: SEEING (arcsec), EXPTIME (sec), AIRMASS, MOONFRAC (0-1), MOONALT (deg), MOONSEP (deg) Table including EXPID for subselection of which row to use filename with obsconditions Table; expid must also be set expid (int, optional): exposure ID nspec (int, optional): number of spectra to simulate psfconvolve (bool, optional): passed to simspec.simulator.Simulator camera_output. if True, convolve with PSF and include per-camera outputs Returns: (sim, fibermap, meta) sim: specsim.simulate.Simulator object fibermap: Table meta: target metadata truth table See obs.new_exposure() for function to generate new random exposure, independent from surveysim, fiberassignment, and pre-generated mocks. ''' from desiutil.log import get_logger log = get_logger() flux, wave, meta = targets if nspec is not None: fiberassign = fiberassign[0:nspec] flux = flux[0:nspec] meta = meta[0:nspec] assert np.all(fiberassign['TARGETID'] == meta['TARGETID']) fibermap = fibermeta2fibermap(fiberassign, meta) #- Parse multiple options for obsconditions if isinstance(obsconditions, str): #- DARK GRAY BRIGHT if obsconditions.upper() in reference_conditions:'Using reference {} obsconditions'.format(obsconditions.upper())) obsconditions = reference_conditions[obsconditions.upper()] #- filename elif os.path.exists(obsconditions):'Loading obsconditions from {}'.format(obsconditions.upper())) if obsconditions.endswith('.ecsv'): allobs =, format='ascii.ecsv') else: allobs = #- trim down to just this exposure if (expid is not None) and 'EXPID' in allobs.colnames: obsconditions = allobs[allobs['EXPID'] == expid] else: raise ValueError('unable to select which exposure from obsconditions file') else: raise ValueError('bad obsconditions {}'.format(obsconditions)) elif isinstance(obsconditions, (astropy.table.Table, np.ndarray)): #- trim down to just this exposure if (expid is not None) and ('EXPID' in obsconditions): obsconditions = allobs[allobs['EXPID'] == expid] else: raise ValueError('must provide expid when providing obsconditions as a Table') #- Validate obsconditions keys try: obskeys = set(obsconditions.dtype.names) except AttributeError: obskeys = set(obsconditions.keys()) missing_keys = set(reference_conditions['DARK'].keys()) - obskeys if len(missing_keys) > 0: raise ValueError('obsconditions missing keys {}'.format(missing_keys)) sim = simulate_spectra(wave, flux, fibermap=fibermap, obsconditions=obsconditions, psfconvolve=psfconvolve) return sim, fibermap
[docs]def fibermeta2fibermap(fiberassign, meta): ''' Convert a fiberassign + targeting metadata table into a fibermap Table A future refactor will standardize the column names of fiber assignment, target catalogs, and fibermaps, but in the meantime this is needed. ''' #- Handle DESI_TARGET vs. SV1_DESI_TARGET etc. target_colnames, target_masks, survey = main_cmx_or_sv(fiberassign) targetcol = target_colnames[0] #- DESI_TARGET or SV1_DESI_TARGET desi_mask = target_masks[0] #- desi_mask or sv1_desi_mask #- Copy column names in common fibermap = for c in ['FIBER', 'TARGETID', 'BRICKNAME']: fibermap[c] = fiberassign[c] for c in target_colnames: fibermap[c] = fiberassign[c] for band in ['G', 'R', 'Z', 'W1', 'W2']: key = 'FLUX_'+band fibermap[key] = meta[key] #- TODO: FLUX_IVAR_* #- set OBJTYPE #- TODO: what about MWS science targets that are also standard stars? #- Loop over STD options for backwards/forwards compatibility stdmask = 0 for name in ['STD', 'STD_FSTAR', 'STD_WD' 'STD_FAINT', 'STD_FAINT_BEST', 'STD_BRIGHT', 'STD_BRIGHT_BEST']: if name in desi_mask.names(): stdmask |= desi_mask[name] isSTD = (fiberassign[targetcol] & stdmask) != 0 isSKY = (fiberassign[targetcol] & desi_mask.SKY) != 0 isSCI = (~isSTD & ~isSKY) fibermap['OBJTYPE'][isSKY] = 'SKY' fibermap['OBJTYPE'][isSCI | isSTD] = 'TGT' fibermap['LAMBDAREF'] = 5400.0 fibermap['TARGET_RA'] = fiberassign['TARGET_RA'] fibermap['TARGET_DEC'] = fiberassign['TARGET_DEC'] fibermap['FIBER_RA'] = fiberassign['TARGET_RA'] fibermap['FIBER_DEC'] = fiberassign['TARGET_DEC'] fibermap['FIBERASSIGN_X'] = fiberassign['FIBERASSIGN_X'] fibermap['FIBERASSIGN_Y'] = fiberassign['FIBERASSIGN_Y'] fibermap['DELTA_X'] = 0.0 fibermap['DELTA_Y'] = 0.0 #- TODO: POSITIONER -> LOCATION #- TODO: TARGETCAT (how should we propagate this info into here?) #- TODO: NaNs in fibermap for unassigned positioners targets return fibermap
#------------------------------------------------------------------------- #- specsim related routines
[docs]def simulate_spectra(wave, flux, fibermap=None, obsconditions=None, redshift=None, dwave_out=None, seed=None, psfconvolve=True, specsim_config_file = "desi"): ''' Simulates an exposure without reading/writing data files Args: wave (array): 1D wavelengths in Angstroms flux (array): 2D[nspec,nwave] flux in 1e-17 erg/s/cm2/Angstrom or astropy Quantity with flux units fibermap (Table, optional): table from fiberassign or fibermap; uses X/YFOCAL_DESIGN, TARGETID, DESI_TARGET obsconditions(dict-like, optional): observation metadata including SEEING (arcsec), EXPTIME (sec), AIRMASS, MOONFRAC (0-1), MOONALT (deg), MOONSEP (deg) redshift (array-like, optional): list/array with each index being the redshifts for that target seed (int, optional): random seed psfconvolve (bool, optional): passed to simspec.simulator.Simulator camera_output. if True, convolve with PSF and include per-camera outputs specsim_config_file (str, optional): path to DESI instrument config file. default is desi config in specsim package. Returns: A specsim.simulator.Simulator object TODO: galsim support ''' import specsim.simulator import specsim.config import astropy.units as u from astropy.coordinates import SkyCoord from desiutil.log import get_logger log = get_logger('DEBUG') from desiutil.iers import freeze_iers freeze_iers() # Input cosmology to calculate the angular diameter distance of the galaxy's redshift from astropy.cosmology import FlatLambdaCDM LCDM = FlatLambdaCDM(H0=70, Om0=0.3) ang_diam_dist = LCDM.angular_diameter_distance random_state = np.random.RandomState(seed) nspec, nwave = flux.shape #- Convert to unit-ful quantities for specsim if not isinstance(flux, u.Quantity): fluxunits = 1e-17 * u.erg / (u.Angstrom * u.s ***2) flux = flux * fluxunits if not isinstance(wave, u.Quantity): wave = wave * u.Angstrom log.debug('loading specsim desi config {}'.format(specsim_config_file)) config = _specsim_config_for_wave('Angstrom').value, dwave_out=dwave_out, specsim_config_file=specsim_config_file) #- Create simulator log.debug('creating specsim desi simulator') # desi = specsim.simulator.Simulator(config, num_fibers=nspec) desi = desisim.specsim.get_simulator(config, num_fibers=nspec, camera_output=psfconvolve) if obsconditions is None: log.warning('Assuming DARK conditions') obsconditions = reference_conditions['DARK'] elif isinstance(obsconditions, str): obsconditions = reference_conditions[obsconditions.upper()] desi.atmosphere.seeing_fwhm_ref = obsconditions['SEEING'] * u.arcsec desi.observation.exposure_time = obsconditions['EXPTIME'] * u.s desi.atmosphere.airmass = obsconditions['AIRMASS'] desi.atmosphere.moon.moon_phase = np.arccos(2*obsconditions['MOONFRAC']-1)/np.pi desi.atmosphere.moon.moon_zenith = (90 - obsconditions['MOONALT']) * u.deg desi.atmosphere.moon.separation_angle = obsconditions['MOONSEP'] * u.deg try: desi.observation.exposure_start = astropy.time.Time(obsconditions['MJD'], format='mjd')'exposure_start {}'.format(desi.observation.exposure_start.utc.isot)) except KeyError:'MJD not in obsconditions, using DATE-OBS {}'.format(desi.observation.exposure_start.utc.isot)) for obskey in reference_conditions['DARK'].keys(): obsval = obsconditions[obskey] log.debug('obsconditions {} = {}'.format(obskey, obsval)) #- Set fiber locations from meta Table or default fiberpos fiberpos = if fibermap is not None and len(fiberpos) != len(fibermap): ii = np.in1d(fiberpos['FIBER'], fibermap['FIBER']) fiberpos = fiberpos[ii] if fibermap is None: fibermap = astropy.table.Table() fibermap['X'] = fiberpos['X'][0:nspec] fibermap['Y'] = fiberpos['Y'][0:nspec] fibermap['FIBER'] = fiberpos['FIBER'][0:nspec] fibermap['LOCATION'] = fiberpos['LOCATION'][0:nspec] #- Extract fiber locations from meta Table -> xy[nspec,2] assert np.all(fibermap['FIBER'] == fiberpos['FIBER'][0:nspec]) if 'XFOCAL_DESIGN' in fibermap.dtype.names: xy = np.vstack([fibermap['XFOCAL_DESIGN'], fibermap['YFOCAL_DESIGN']]).T * elif 'X' in fibermap.dtype.names: xy = np.vstack([fibermap['X'], fibermap['Y']]).T * else: xy = np.vstack([fiberpos['X'], fiberpos['Y']]).T * if 'TARGETID' in fibermap.dtype.names: unassigned = (fibermap['TARGETID'] == -1) if np.any(unassigned): #- see #- for the units -> array -> units trick xy[unassigned,0] = np.asarray(fiberpos['X'][unassigned], dtype=xy.dtype) * xy[unassigned,1] = np.asarray(fiberpos['Y'][unassigned], dtype=xy.dtype) * #- Determine source types #- TODO: source shapes + galsim instead of fixed types + fiberloss table source_types = get_source_types(fibermap) # source types are sky elg lrg qso bgs star , they # are only used in specsim.fiberloss for the desi.instrument.fiberloss_method="table" method desi.instrument.fiberloss_method = 'fastsim' log.debug('running simulation with {} fiber loss method'.format(desi.instrument.fiberloss_method)) unique_source_types = set(source_types) comment_line="source types:" for u in set(source_types) : comment_line+=" {} {}".format(np.sum(source_types==u),u) log.debug(comment_line) source_fraction=None source_half_light_radius=None source_minor_major_axis_ratio=None source_position_angle=None if desi.instrument.fiberloss_method == 'fastsim' or desi.instrument.fiberloss_method == 'galsim' : # the following parameters are used only with fastsim and galsim methods elgs=(source_types=="elg") lrgs=(source_types=="lrg") bgss=(source_types=="bgs") if np.sum(lrgs)>0 or np.sum(elgs)>0: log.warning("the half light radii are fixed here for LRGs and ELGs (and not magnitude or redshift dependent)") if np.sum(bgss)>0 and redshift is None: log.warning("the half light radii are fixed here for BGS (as redshifts weren't supplied)") # BGS parameters based on SDSS main sample, in g-band # see analysis from J. Moustakas in # # B/T (bulge-to-total ratio): 0.48 (0.36 - 0.59). # Bulge Sersic n: 2.27 (1.12 - 3.60). # log10 (Bulge Half-light radius): 0.11 (-0.077 - 0.307) arcsec # log10 (Disk Half-light radius): 0.67 (0.54 - 0.82) arcsec # This gives # bulge_fraction = 0.48 # disk_fraction = 0.52 # bulge_half_light_radius = 1.3 arcsec # disk_half_light_radius = 4.7 arcsec # note we use De Vaucouleurs' law , which correspond to a Sersic index n=4 # source_fraction[:,0] is DISK profile (exponential) fraction # source_fraction[:,1] is BULGE profile (devaucouleurs) fraction # 1 - np.sum(source_fraction,axis=1) is POINT source profile fraction # see specsim.GalsimFiberlossCalculator.create_source routine source_fraction=np.zeros((nspec,2)) source_fraction[elgs,0]=1. # ELG are disk only source_fraction[lrgs,1]=1. # LRG are bulge only source_fraction[bgss,0]=0.52 # disk comp in BGS source_fraction[bgss,1]=0.48 # bulge comp in BGS # source_half_light_radius[:,0] is the half light radius in arcsec for the DISK profile # source_half_light_radius[:,1] is the half light radius in arcsec for the BULGE profile # see specsim.GalsimFiberlossCalculator.create_source routine source_half_light_radius=np.zeros((nspec,2)) source_half_light_radius[elgs,0]=0.45 # ELG are disk only, arcsec source_half_light_radius[lrgs,1]=1. # LRG are bulge only, arcsec # 4.7 is angular size of z=0.1 disk, and 1.3 is angular size of z=0.1 bulge bgs_disk_z01 = 4.7 # in arcsec bgs_bulge_z01 = 1.3 # in arcsec # Convert to angular size of the objects in this sample with given redshifts if redshift is None: angscales = np.ones(np.sum(bgss)) else: bgs_redshifts = redshift[bgss] # Avoid infinities if np.any(bgs_redshifts <= 0.): bgs_redshifts[bgs_redshifts <= 0.] = 0.0001 angscales = ( ang_diam_dist(0.1) / ang_diam_dist(bgs_redshifts) ).value source_half_light_radius[bgss,0]= bgs_disk_z01 * angscales # disk comp in BGS, arcsec source_half_light_radius[bgss,1]= bgs_bulge_z01 * angscales # bulge comp in BGS, arcsec if desi.instrument.fiberloss_method == 'galsim' : # the following parameters are used only with galsim method # source_minor_major_axis_ratio[:,0] is the axis ratio for the DISK profile # source_minor_major_axis_ratio[:,1] is the axis ratio for the BULGE profile # see specsim.GalsimFiberlossCalculator.create_source routine source_minor_major_axis_ratio=np.zeros((nspec,2)) source_minor_major_axis_ratio[elgs,0]=0.7 source_minor_major_axis_ratio[lrgs,1]=0.7 source_minor_major_axis_ratio[bgss,1]=0.7 # the source position angle is in degrees # see specsim.GalsimFiberlossCalculator.create_source routine source_position_angle = np.zeros((nspec,2)) random_angles = 360.*random_state.uniform(size=nspec) source_position_angle[elgs,0]=random_angles[elgs] source_position_angle[lrgs,1]=random_angles[lrgs] source_position_angle[bgss,1]=random_angles[bgss] #- Work around randomness in specsim quickfiberloss calculations #- while not impacting global random state. #- See randstate = np.random.get_state() np.random.seed(seed) desi.simulate(source_fluxes=flux, focal_positions=xy, source_types=source_types, source_fraction=source_fraction, source_half_light_radius=source_half_light_radius, source_minor_major_axis_ratio=source_minor_major_axis_ratio, source_position_angle=source_position_angle) np.random.set_state(randstate) return desi
[docs]def _specsim_config_for_wave(wave, dwave_out=None, specsim_config_file = "desi"): ''' Generate specsim config object for a given wavelength grid Args: wave: array of linearly spaced wavelengths in Angstroms Options: specsim_config_file: (str) path to DESI instrument config file. default is desi config in specsim package. Returns: specsim Configuration object with wavelength parameters set to match this input wavelength grid ''' import specsim.config dwave = round(np.mean(np.diff(wave)), 3) assert np.allclose(dwave, np.diff(wave), rtol=1e-6, atol=1e-3) config = specsim.config.load_config(specsim_config_file) config.wavelength_grid.min = wave[0] config.wavelength_grid.max = wave[-1] + dwave/2.0 config.wavelength_grid.step = dwave if dwave_out is None: dwave_out = 1.0 config.instrument.cameras.b.constants.output_pixel_size = "{:.3f} Angstrom".format(dwave_out) config.instrument.cameras.r.constants.output_pixel_size = "{:.3f} Angstrom".format(dwave_out) config.instrument.cameras.z.constants.output_pixel_size = "{:.3f} Angstrom".format(dwave_out) config.update() return config
[docs]def get_source_types(fibermap): ''' Return a list of specsim source types based upon fibermap['DESI_TARGET'] Args: fibermap: fibermap Table including DESI_TARGET column Returns array of source_types 'sky', 'elg', 'lrg', 'qso', 'star' Unassigned fibers fibermap['TARGETID'] == -1 will be treated as 'sky' If fibermap.meta['FLAVOR'] = 'arc' or 'flat', returned source types will match that flavor, though specsim doesn't use those as source_types TODO: specsim/desimodel doesn't have a fiber input loss model for BGS yet, so BGS targets get source_type = 'lrg' (!) ''' from desiutil.log import get_logger log = get_logger() if ('DESI_TARGET' not in fibermap.dtype.names) and \ ('SV1_DESI_TARGET' not in fibermap.dtype.names): log.warning("(SV1_)DESI_TARGET not in fibermap table; using source_type='star' for everything") return np.array(['star',] * len(fibermap)) target_colnames, target_masks, survey = main_cmx_or_sv(fibermap) targetcol = target_colnames[0] #- DESI_TARGET or SV1_DESI_TARGET tm = target_masks[0] #- desi_mask or sv1_desi_mask source_type = np.zeros(len(fibermap), dtype='U4') assert np.all(source_type == '') if 'TARGETID' in fibermap.dtype.names: unassigned = fibermap['TARGETID'] == -1 source_type[unassigned] = 'sky' source_type[(fibermap['OBJTYPE'] == 'FLT')] = 'FLAT' source_type[(fibermap['OBJTYPE'] == 'ARC')] = 'ARC' source_type[(fibermap[targetcol] & tm.SKY) != 0] = 'sky' source_type[(fibermap[targetcol] & tm.ELG) != 0] = 'elg' source_type[(fibermap[targetcol] & tm.LRG) != 0] = 'lrg' source_type[(fibermap[targetcol] & tm.QSO) != 0] = 'qso' source_type[(fibermap[targetcol] & tm.BGS_ANY) != 0] = 'bgs' starmask = 0 for name in ['STD', 'STD_FSTAR', 'STD_WD', 'MWS_ANY', 'STD_FAINT', 'STD_FAINT_BEST', 'STD_BRIGHT', 'STD_BRIGHT_BEST']: if name in desitarget.targetmask.desi_mask.names(): starmask |= desitarget.targetmask.desi_mask[name] source_type[(fibermap[targetcol] & starmask) != 0] = 'star' #- Simulate unassigned fibers as sky ## TODO: when fiberassign and desitarget are updated, use ## desitarget.targetmask.desi_mask.NO_TARGET to identify these source_type[fibermap['TARGETID'] < 0] = 'sky' assert not np.any(source_type == '') for name in sorted(np.unique(source_type)): n = np.count_nonzero(source_type == name) log.debug('{} {} targets'.format(name, n)) return source_type
#------------------------------------------------------------------------- #- I/O related routines
[docs]def read_fiberassign(tilefile_or_id, indir=None): ''' Returns fiberassignment table for tileid Args: tilefile_or_id (int or str): tileid (int) or full path to tile file (str) Returns: fiberassignment Table from HDU 1 ''' #- tileid is full path to file instead of int ID or just filename if isinstance(tilefile_or_id, str) and os.path.exists(tilefile_or_id): return if indir is None: indir = os.path.join(os.environ['DESI_TARGETS'], 'fiberassign') if isinstance(tilefile_or_id, (int, np.int32, np.int64)): tilefile = os.path.join(indir, 'tile_{:05d}.fits'.format(tilefile_or_id)) else: tilefile = os.path.join(indir, tilefile_or_id) return, 'FIBER_ASSIGNMENTS')
#------------------------------------------------------------------------- #- Move this to def testslit_fibermap(): # from WBS 1.6 PDR Fiber Slit document # science slit has 20 bundles of 25 fibers # test slit has 1 fiber per bundle except in the middle where it is fully populated nspectro=10 testslit_nspec_per_spectro=20 testslit_nspec = nspectro*testslit_nspec_per_spectro # fibermap = np.zeros(testslit_nspec, fibermap = fibermap['FIBER'] = np.zeros((testslit_nspec)).astype(int) fibermap['SPECTROID'] = np.zeros((testslit_nspec)).astype(int) for spectro in range(nspectro) : fiber_index=testslit_nspec_per_spectro*spectro first_fiber_id=500*spectro for b in range(20) : # Fibers at Top of top block or Bottom of bottom block if b <= 10: fibermap['FIBER'][fiber_index] = 25*b + first_fiber_id else: fibermap['FIBER'][fiber_index] = 25*b + 24 + first_fiber_id fibermap['SPECTROID'][fiber_index] = spectro fiber_index+=1 return fibermap #------------------------------------------------------------------------- #- MOVE THESE TO (?) #-------------------------------------------------------------------------
[docs]def get_mock_spectra(fiberassign, mockdir=None, nside=64, obscon=None): ''' Args: fiberassign: table loaded from fiberassign tile file Options: mockdir (str): base directory under which files are found nside (int): healpix nside for file directory grouping obscon (str): (observing conditions) None/dark/bright extra dir level Returns (flux, wave, meta) tuple ''' nspec = len(fiberassign) flux = None meta = None wave = None objmeta = None target_colnames, target_masks, survey = main_cmx_or_sv(fiberassign) targetcol = target_colnames[0] #- DESI_TARGET or SV1_DESI_TARGET desi_mask = target_masks[0] #- desi_mask or sv1_desi_mask issky = (fiberassign[targetcol] & desi_mask.SKY) != 0 skyids = fiberassign['TARGETID'][issky] #- check several ways in which an unassigned fiber might appear unassigned = np.isnan(fiberassign['TARGET_RA']) unassigned |= np.isnan(fiberassign['TARGET_DEC']) unassigned |= (fiberassign['TARGETID'] < 0) ## TODO: check desi_mask.NO_TARGET once that bit exists for truthfile, targetids in zip(*targets2truthfiles( fiberassign[~unassigned], basedir=mockdir, nside=nside, obscon=obscon)): #- Sky fibers aren't in the truth files ok = ~np.in1d(targetids, skyids) tmpflux, tmpwave, tmpmeta, tmpobjmeta = read_mock_spectra(truthfile, targetids[ok]) if flux is None: nwave = tmpflux.shape[1] flux = np.zeros((nspec, nwave)) meta = np.zeros(nspec, dtype=tmpmeta.dtype) meta['TARGETID'] = -1 wave = tmpwave.astype('f8') objmeta = dict() for key in tmpobjmeta.keys(): objmeta[key] = list() ii = np.in1d(fiberassign['TARGETID'], tmpmeta['TARGETID']) flux[ii] = tmpflux meta[ii] = tmpmeta assert np.all(wave == tmpwave) for key in tmpobjmeta.keys(): if key not in objmeta: objmeta[key] = list() objmeta[key].append(tmpobjmeta[key]) #- Stack the per-objtype meta tables for key in objmeta.keys(): objmeta[key] = astropy.table.Table(np.hstack(objmeta[key])) #- Set meta['TARGETID'] for sky fibers #- TODO: other things to set? meta['TARGETID'][issky] = skyids meta['TARGETID'][unassigned] = fiberassign['TARGETID'][unassigned] assert np.all(fiberassign['TARGETID'] == meta['TARGETID']) return flux, wave, astropy.table.Table(meta), objmeta
[docs]def read_mock_spectra(truthfile, targetids, mockdir=None): ''' Reads mock spectra from a truth file Args: truthfile (str): full path to a mocks truth-\*.fits file targetids (array-like): targetids to load from that file mockdir: ??? Returns (flux, wave, truth) tuples: flux[nspec, nwave]: flux in 1e-17 erg/s/cm2/Angstrom wave[nwave]: wavelengths in Angstroms truth[nspec]: metadata truth table ''' if len(targetids) != len(np.unique(targetids)): from desiutil.log import get_logger log = get_logger() log.error("Requested TARGETIDs for {} are not unique".format( os.path.basename(truthfile))) #- doesn't return a real ndarray, causing problems #- with the reordering downstream so use fitsio instead # with, memmap=False) as fx: # truth = fx['TRUTH'].data # wave = fx['WAVE'].data # flux = fx['FLUX'].data objtruth = dict() with fitsio.FITS(truthfile) as fx: truth = fx['TRUTH'].read() wave = fx['WAVE'].read() flux = fx['FLUX'].read() if 'OBJTYPE' in truth.dtype.names: # output of desisim.obs.new_exposure objtype = [oo.decode('ascii').strip().upper() for oo in truth['OBJTYPE']] else: # output of objtype = [oo.decode('ascii').strip().upper() for oo in truth['TEMPLATETYPE']] for obj in set(objtype): extname = 'TRUTH_{}'.format(obj) if extname in fx: objtruth[obj] = fx[extname].read() missing = np.in1d(targetids, truth['TARGETID'], invert=True) if np.any(missing): missingids = targetids[missing] raise ValueError('Targets missing from {}: {}'.format(truthfile, missingids)) #- Trim to just the spectra for these targetids ii = np.in1d(truth['TARGETID'], targetids) flux = flux[ii] truth = truth[ii] if bool(objtruth): for obj in objtruth.keys(): ii = np.in1d(objtruth[obj]['TARGETID'], targetids) objtruth[obj] = objtruth[obj][ii] assert set(targetids) == set(truth['TARGETID']) #- sort truth to match order of input targetids # it doesn't matter if objtruth is sorted if len(targetids) == len(truth['TARGETID']): i = np.argsort(targetids) j = np.argsort(truth['TARGETID']) k = np.argsort(i) reordered_truth = truth[j[k]] reordered_flux = flux[j[k]] else: #- Slower, but works even with repeated TARGETIDs ii = np.argsort(truth['TARGETID']) sorted_truthids = truth['TARGETID'][ii] reordered_flux = np.empty(shape=(len(targetids), flux.shape[1]), dtype=flux.dtype) reordered_truth = np.empty(shape=(len(targetids),), dtype=truth.dtype) for j, tx in enumerate(targetids): k = np.searchsorted(sorted_truthids, tx) reordered_flux[j] = flux[ii[k]] reordered_truth[j] = truth[ii[k]] assert np.all(reordered_truth['TARGETID'] == targetids) wave = reordered_flux = return reordered_flux, wave, reordered_truth, objtruth
[docs]def targets2truthfiles(targets, basedir, nside=64, obscon=None): ''' Return list of mock truth files that contain these targets Args: targets: table with TARGETID column, e.g. from fiber assignment basedir: base directory under which files are found Options: nside (int): healpix nside for file directory grouping obscon (str): (observing conditions) None/dark/bright extra dir level Returns (truthfiles, targetids): truthfiles: list of truth filenames targetids: list of lists of targetids in each truthfile i.e. targetids[i] is the list of targetids from targets['TARGETID'] that are in truthfiles[i] ''' import healpy #import as mockio from import find_target_files assert nside >= 2 #- TODO: what should be done with assignments without targets? targets = targets[targets['TARGETID'] != -1] theta = np.radians(90-targets['TARGET_DEC']) phi = np.radians(targets['TARGET_RA']) pixels = healpy.ang2pix(nside, theta, phi, nest=True) truthfiles = list() targetids = list() for ipix in sorted(np.unique(pixels)): filename = find_target_files(basedir, flavor='truth', obscon=obscon, hp=ipix, nside=nside, mock=True) truthfiles.append(filename) ii = (pixels == ipix) targetids.append(np.asarray(targets['TARGETID'][ii])) return truthfiles, targetids