def gather_models(cmddir, bf, age_range, logz, vvc_range, av, dmod):
# form the array of models spanning age and v/vc grid space:
# structure is an NxM dimensional matrix of Hess diagrams, indexed
# by i for the ith vector along the rotation rate axis, and by j along
# the jth age axis. Each Hess diagram is normalized to sum to one.
vvc_pts = []
for i, a_vvc in enumerate(vvc_range):
# step through in age, adding an age vector at each point in v/vc space
age_vector = []
for j, an_age in enumerate(age_range):
a_cmd = cmd.CMD(get_cmdf(cmddir, bf, an_age, logz, a_vvc, av, dmod))
model_hess = a_cmd.cmd['Nsim']
model_hess /= np.sum(model_hess)
age_vector.append(model_hess)
vvc_pts.append(np.array(age_vector))
gates = a_cmd.cmd['gate']
ugates = np.unique(gates)
if len(ugates) > 1:
dinds = np.digitize(gates, bins=ugates, right=True)
else:
dinds = None
model = np.array(vvc_pts)
return model, dinds
def genmod_bivvcspread(cmddir, age, mass, vvcmu1, vvcsig1, vvcmu2, vvcsig2, vvclim):
vvcs = np.arange(0.0, vvclim, 0.1)
vvcweights = gen_bigaussweights(vvcs, vvcmu1, vvcsig1, vvcmu2, vvcsig2)
photbase = cmddir.split('/')[1]
bf, av, agebin, logz, dmod = fio.parse_fname(photbase, mode="str")
hess_arr = []
for vvc in vvcs:
a_cmd = cmd.CMD(fio.get_cmdf(cmddir, bf, age, logz, vvc, av, dmod))
mock_hess = a_cmd.cmd['Nsim']
mock_hess /= np.sum(mock_hess)
#if age == ages[0]:
# composite_hess = mock_hess
#else:
# composite_hess += mock_hess
hess_arr.append(mock_hess)
hess_arr = np.array(hess_arr)
#print(len(hess_arr))
#print(len(vvcweights))
composite_hess = np.sum(vvcweights*hess_arr.T, axis=1)
#print(len(composite_hess))
truths = {}
for i, vvc in enumerate(vvcs):
truths[vvc] = mass*vvcweights[i]
return mass*composite_hess, truths
def genmod_agespread(cmddir, mass, agemu, agesig):
ages = np.arange(8.50, 9.50, 0.02)
ageweights = gen_gaussweights(ages, agemu, agesig)
photbase = cmddir.split('/')[1]
bf, av, agebin, logz, dmod = fio.parse_fname(photbase, mode="str")
hess_arr = []
for age in ages:
a_cmd = cmd.CMD(fio.get_cmdf(cmddir, bf, age, logz, 0.0, av, dmod))
mock_hess = a_cmd.cmd['Nsim']
mock_hess /= np.sum(mock_hess)
#if age == ages[0]:
# composite_hess = mock_hess
#else:
# composite_hess += mock_hess
hess_arr.append(mock_hess)
hess_arr = np.array(hess_arr)
#print(len(hess_arr))
#print(len(ageweights))
composite_hess = np.sum(ageweights*hess_arr.T, axis=1)
#print(len(composite_hess))
truth = np.sum(composite_hess)
return mass*composite_hess
def genmod_agebivvcspread(cmddir, mass, agemu, agesig, vvcmu1, vvcsig1, vvcmu2, vvcsig2, vvclim):
ages = np.arange(8.50, 9.50, 0.02)
#mu, sig = 9.00, 0.3
ageweights = gen_gaussweights(ages, agemu, agesig)
vvcs = np.arange(0.0, vvclim, 0.1)
#mu, sig = 0.3, 0.2
vvcweights = gen_bigaussweights(vvcs, vvcmu1, vvcsig1, vvcmu2, vvcsig2)
photbase = cmddir.split('/')[1]
bf, av, agebin, logz, dmod = fio.parse_fname(photbase, mode="str")
composite_hess = cmd.CMD(fio.get_cmdf(cmddir, bf, 9.00, logz, 0.0, av, dmod))
composite_hess = np.zeros(len(composite_hess.cmd['Nsim']))
vvc_pts = []
for i, a_vvc in enumerate(vvcs):
# step through in age, adding an age vector at each point in v/vc space
age_vector = []
for j, an_age in enumerate(ages):
a_cmd = cmd.CMD(fio.get_cmdf(cmddir, bf, an_age, logz, a_vvc, av, dmod))
model_hess = a_cmd.cmd['Nsim']
model_hess /= np.sum(model_hess)
age_vector.append(model_hess)
vvc_pts.append(np.array(age_vector))
model = np.array(vvc_pts)
composite_hess += np.sum((vvcweights[:, np.newaxis])*np.sum(ageweights[:,np.newaxis]*model, axis=1), axis=0)
#print(len(hess_arr))
#print(len(vvcweights))
#composite_hess = np.sum(vvcweights*hess_arr.T, axis=1)
#print(len(composite_hess))
truths = {}
for i, vvc in enumerate(vvcs):
truths[vvc] = mass*vvcweights[i]
return mass*composite_hess, truths
def pgplot(obs, model, cmddir, bf, age, logz, av, dmod, vvclim, weights, filters, age_range, svname=None, log=False, svdir=None):
"""
Creates a MATCH pg style plot of data, model, data-model, and -2lnP map.
"""
composite_cmd = cmd.CMD(fio.get_cmdf(cmddir, bf, age, logz, 0.0, av, dmod))#, ymag='I')
composite_cmd.cmd['Nsim'] = np.zeros(len(composite_cmd.cmd['Nsim']))
vvc_range = np.arange(0.0, vvclim, 0.1)
Nrot = len(vvc_range)
#age_range = np.arange(8.5, 9.5, 0.02)
mu = weights[Nrot]
try:
sigma = weights[Nrot+1]
except IndexError:
sigma = 0.0
ageweights = gen_gaussweights(age_range, mu, sigma)
#for i, avvc in enumerate(vvc_range):
# for j, anage in enumerate(age_range):
# a_cmd = cmd.CMD(fio.get_cmdf(cmddir, bf, anage, logz, avvc, av, dmod))#, ymag='I')
# 1 * (jth age weight, added i times) * ith rotation rate.
# a_cmd.cmd['Nsim'] = (a_cmd.cmd['Nsim'] / np.sum(a_cmd.cmd['Nsim'])) * (best_gweights[j]) * (10**weights[i])
# add each cmd (re-weighted by solutions) to the composite CMD model.
# composite_cmd.cmd['Nsim'] += a_cmd.cmd['Nsim']
vvcweights = weights[:Nrot]
composite_cmd.cmd['Nsim'] += np.sum((10**vvcweights[:, np.newaxis])*np.sum(ageweights[:,np.newaxis]*model, axis=1), axis=0)
composite_cmd.cmd['Nobs'] = obs
# arbitrary vvc used -- just need a file name to break up.
#fn_base = (fio.get_cmdf(cmddir, bf, age, logz, 0.0, av, dmod).split('/')[-1]).split('.out.cmd')[0]
#fehval = float((fn_base.split('logz')[-1]).split('_')[0])
#lageval = float((fn_base.split('_t')[-1]).split('_')[0])
#avval = float((fn_base.split('_av')[-1]).split('_')[0])
#dmodval = float((fn_base.split('_dmod')[-1]).split('_')[0])
print(max(composite_cmd.cmd['Nsim']))
filters, photstrs = match_to_mist(filters)
photstr = photstrs[0]
print(filters)
redmag_name = filters[1]
bluemag_name = filters[0]
color_name = "{:s}-{:s}".format(bluemag_name, redmag_name)
# iso00 = rmm.ISOCMD(round(float(logz), 2), min(vvc_range), ebv= round(float(av), 2)/3.1, photstr=photstr, exttag='TP')
# iso00.set_isodata(round(float(mu), 2), color_name, bluemag_name, dmod=round(float(dmod), 2))
# this try except is fudgy -- onl needed cause v/vc = 0.6 models aren't available on my local machine.
# try:
# iso06 = rmm.ISOCMD(round(float(logz), 2), max(vvc_range), ebv= round(float(av), 2)/3.1, photstr=photstr, exttag='TP')
# except Exception as e:
# iso06 = rmm.ISOCMD(round(float(logz), 2), 0.6, ebv= round(float(av), 2)/3.1, photstr=photstr, exttag='TP')
#iso06.set_isodata(round(float(mu), 2), color_name, bluemag_name, dmod=round(float(dmod), 2))
# (x, y), i.e., (color, red mag) points of each isochrone in a list:
#mist_pts = [
# isoget_colmags(iso00, [color_name, bluemag_name], lage=round(float(mu), 2), dmod=round(float(dmod), 2)),
# isoget_colmags(iso06, [color_name, bluemag_name], lage=round(float(mu), 2), dmod=round(float(dmod), 2))
# ]
# recalculate the d-m and signifigance hesses using the new hesses.
composite_cmd.recalc()
# create a MATCH pg style plot using the .cmd file:
pgcmd_kwargs = {}
#pgcmd_kwargs['mist_pts'] = mist_pts
if svname == None:
if svdir == None:
pgcmd_kwargs['figname'] = os.path.join(cmddir, 'match_pgplot.png')
else:
pgcmd_kwargs['figname'] = os.path.join(cmddir, svdir, 'match_pgplot.png')
else:
if svdir == None:
pgcmd_kwargs['figname'] = os.path.join(cmddir, svname)
else:
pgcmd_kwargs['figname'] = os.path.join(cmddir, svdir, svname)
# four panel plot:
if log:
pgcmd_kwargs['logcounts'] = True
composite_cmd.pgcmd(**pgcmd_kwargs)
else:
composite_cmd.pgcmd(**pgcmd_kwargs)
return
vvclim = round(Nrot / 10.0, 1)
vvc_range = np.arange(0.0, vvclim, 0.1)
age_range = np.arange(7.9, 9.5, 0.02) # lower lim was 8.5
# default, dummy truths; will be reassigned:
truths = {rot: 1e-11 for rot in vvc_range}
mass = 5E4 # mock cluster "mass" or total counts; 1e6 was default
# Generates mock data from the model library to use instead of observed data if told to:
if "mock" in mode:
# case of no rotation distribution:
if mockd_vvcsig1 == 0.0:
# and no age distribution (would be an SSP):
if mockd_agesig == 0.0:
obscmd = cmd.CMD(
fio.get_cmdf(cmddir, bf, mockd_agemu, logz, mockd_vvcmu1,
av, dmod))
obs = obscmd.cmd['Nobs']
# age distribution, no rotation distribution, P(sigtau); in practice will trigger so long as mode is mock-sigtau and agesig !=0.0:
else:
print("GENERATING MOCK DATA WITH AGE SPREAD...")
obs = pr.genmod_agespread(cmddir,
mass=mass,
agemu=mockd_agemu,
agesig=mockd_agesig)
obsweight = np.sum(obs)
# using lower limit of weight search prior for truths of "zero" components.
truths = {
rot: max([0.0, np.log10(obsweight * 1e-4)])
for rot in vvc_range
#ax6 = plt.subplot(gs[1, 0])
#ax7 = plt.subplot(gs[1, -1])
#ax8 = plt.subplot(gs[2, 0])
#ax9 = plt.subplot(gs[2, -1])
#ax10 = plt.subplot(gs[-1, 0])
#ax11 = plt.subplot(gs[-1, 1])
#ax12 = plt.subplot(gs[-1, 2])
#ax13 = plt.subplot(gs[-1, 3])
#ax14 = plt.subplot(gs[-1, 4])
#axa = [ax1,ax2,ax3,ax4,ax5,ax6,ax7,ax8,ax9,ax10,ax11,ax12,ax13,ax14]
bf, age, logz, av, dmod = "0.00", 9.00, '-0.30', '0.0', '0.00'
cmddir = 'output/bf0.00_av0.0_SFR0.01_t9.00_9.02_logZ-0.30_vvc0.4_Tycho_BTycho_V_ex/'
# empty out a hess to use as a composite model later:
composite_cmd = cmd.CMD(fio.get_cmdf(cmddir, bf, age, logz, 0.0, av,
dmod)) #, ymag='I')
composite_cmd.cmd['Nsim'] = np.zeros(len(composite_cmd.cmd['Nsim']))
obsweight = np.sum(composite_cmd.cmd['Nobs'])
# need random weights and mu, sigma
filters = ['Tycho_B', 'Tycho_V']
vvc_range = np.arange(0.0, 1.0, 0.2)
Nrot = len(vvc_range)
age_range = np.arange(8.5, 9.5, 0.2)
mu = np.random.uniform(8.5, 9.3)
sigma = 10**np.random.uniform(-2, 0)
# try:
# sigma = weights[Nrot+1]
# except IndexError:
# sigma = 0.0
def main(cluster_name,
photbase,
jobid,
filters,
dmod,
ebv,
local,
star_num,
truth,
incvvc,
savebest=True,
bestax=None,
justbest=False):
global match_dir
global scrn_dir
global outpath
global plotpath
match_dir = get_workdir(data=cluster_name)
# path to MATCH .scrn files:
scrn_dir = os.path.join(get_scrndir(data=cluster_name), photbase,
'SLURM{:s}'.format(jobid))
# path to MATCH .cmd and .out files:
outpath = os.path.join(get_outdir(data=cluster_name), photbase,
'SLURM{:s}'.format(jobid))
# path to place output plots in:
plotpath = os.path.join(outpath, 'plots')
#print(outpath)
#vbests, vvcrits, qdict, bfdict = marginalize(cluster_name, photbase, sys.argv[3], sys.argv[4], local=local)
#vvcrits, best_dict, qdict, bfdict = marginalize(cluster_name, photbase, sys.argv[3], sys.argv[4], local=local, incvvc=incvvc)
# by default, readssp() will return an AVERAGE (over all runs) of the bestfit parameter value and uncertainty.
# the original, unaveraged values are derived from the sspcombine solutions.
if incvvc == 'all':
vvcrits = [0.0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6]
else:
vvcrits = [incvvc]
best_dict = {}
for vvc in vvcrits:
# read sspcombine output for age results (subsampled solution being used):
starnums, sspbest_age, ssp_ageerr = readssp(data=cluster_name,
solution='subsampled',
fromfit=False,
param='age',
photname_base=photbase,
incvvc=vvc)
# read sspcombines for metallicity results:
starnums, sspbest_logz, ssp_logzerr = readssp(data=cluster_name,
solution='subsampled',
fromfit=False,
param='logZ',
photname_base=photbase,
incvvc=vvc)
# dictionary whose keys are the corresponding vvc values for the best fit values found above (dmod, Av fixed):
best_dict[vvc] = {
'lage': {
'val': float(sspbest_age[0]),
'err': float(ssp_ageerr[0])
},
'feh': {
'val': float(sspbest_logz[0]),
'err': float(ssp_logzerr[0])
},
'dmod': dmod,
'ebv': ebv
}
# if told to plot "truth" values:
if truth:
if 'Fakedata' in cluster_name:
# true age bin (list), logZ +/- spread (list), v/vcrit (float) of artificial data fit to:
true_age = plt_truth(None,
cluster_name,
photbase,
'lage',
justval=True)
true_feh = plt_truth(None,
cluster_name,
photbase,
'logZ',
justval=True)
true_vvc = plt_truth(None,
cluster_name,
photbase,
'vvcrit',
justval=True)
elif 'Hyades' in cluster_name:
# Perryman et al.1998.
true_age = [np.log10(575e6), np.log10(675e6)]
# [Fe/H] -- e.g. deB01 says 0.14 +/- 0.05
true_feh = [0.15, 0.15]
# assume canonically non-rotating:
true_vvc = 0.0
elif 'Praesepe' in cluster_name:
# source???.
true_age = [np.log10(575e6), np.log10(675e6)]
# [Fe/H] source???
true_feh = [0.15, 0.15]
# assume canonically non-rotating:
true_vvc = 0.0
elif 'Pleiades' in cluster_name:
# source???
true_age = [np.log10(100e6), np.log10(125e6)]
# [Fe/H] e.g. Soderblom+ 2009 says 0.03 +/- 0.02
true_feh = [0.00, 0.00]
# assume canonically non-rotating:
true_vvc = 0.0
truths = {'age': true_age, 'feh': true_feh, 'vvc': true_vvc}
else:
truths = None
#plt.clf()
# clearing out pre-existing output files that this script may have produced in prev. runs:
if not justbest:
for f in glob.glob(os.path.join(outpath, 'cmd*.png')):
os.remove(f)
for f in glob.glob(os.path.join(plotpath, 'cmd*.png')):
os.remove(f)
for f in glob.glob(os.path.join(plotpath, '*.txt')):
os.remove(f)
# if req., correct 2MASS filter names to those recognizable by MIST.
for i, afilter in enumerate(filters):
if afilter == 'J':
filters[i] = '2MASS_J'
elif afilter == 'Ks':
filters[i] = '2MASS_Ks'
elif afilter == 'B':
filters[i] = 'Bessell_B'
elif afilter == 'V':
filters[i] = 'Bessell_V'
# Retrieve the magnitude limits used in the fit (from the calcsfh .param file used):
param_fname = glob.glob(
os.path.join(match_dir, 'csfh_param',
'calcsfh_{:s}.param'.format(photbase)))[0]
with open(param_fname, 'r') as f:
lines = f.readlines()
# Color limits:
color_lims = lines[4].split(' ')
minvmi = float(color_lims[3])
maxvmi = float(color_lims[4])
ibin = float(color_lims[0])
vibin = float(color_lims[1])
ilims = lines[6].split(' ')
imin = float(ilims[1])
imax = float(ilims[0])
exclusions = lines[7].split('\n')[0]
ntbins = int(lines[8])
tbins = lines[9:8 + ntbins]
for i, aline in enumerate(tbins):
tbins[i] = map(float, aline.split('\n')[0].split(' ')[-2:])
# Manage exclude gates (right now only works for one region):
# If there are exclude gates...
if int(exclusions[0]) > 0:
# Get the exclude gate points:
expts = map(float, exclusions.split(' ')[1:-1])
ex_x = expts[::2]
ex_y = expts[1::2]
ex_xy = zip(ex_x, ex_y)
else:
expts = None
# Also get the residuals (???):
# For overplotting data points, get the data points CMD x, y values from the photometry file:
photf_v = np.genfromtxt(os.path.join(match_dir, 'photometry',
'{:s}.phot'.format(photbase)),
usecols=(0, ))
photf_i = np.genfromtxt(os.path.join(match_dir, 'photometry',
'{:s}.phot'.format(photbase)),
usecols=(1, ))
photf_vmi = photf_v - photf_i
# Tuple storing the x, y values for plotting the data on a CMD (using color on x, red filter on y):
photf_pts = (photf_vmi, photf_i)
# match uses blue filter on y axis.
photf_pts_alt = (photf_vmi, photf_v)
# store number of data pts if told to:
if star_num:
nstars = len(photf_vmi)
#plt.clf()
bestfit = 99999.9
# Plot best isochrone for each v/vcrit individually:
#=================================================================
allfig = plt.figure()
allax = allfig.add_subplot(111)
# iterate through all v/vcrits used for fitting:
for j, vvcrit in enumerate(vvcrits):
#print(outpath)
# get the .cmd file for making a MATCH pg style plot:
fname = glob.glob(
os.path.join(outpath, '*vvcrit{:.1f}*.cmd'.format(vvcrit)))[0]
# use Phil's code to make a CMD object from the .cmd file:
a_cmd = cmd.CMD(fname)
# store the axis extents of the CMD data
extent = a_cmd.extent
if not justbest:
fig = plt.figure()
ax = fig.add_subplot(111)
# scatter plot of stars from data file (observations):
# w/ num of data points in the plot legend...
if star_num:
ax.scatter(*photf_pts,
lw=0,
s=8,
c='r',
label=r"$N_*$ = " + "{:d}".format(nstars))
ax.legend(loc='best')
# just the scatter plot...
else:
ax.scatter(*photf_pts, lw=0, s=8, c='r')
# X out masked data points.
if expts != None:
for k in range(len(photf_pts[0])):
if (min(ex_x) < photf_vmi[k] <
max(ex_x)) & (min(ex_y) < photf_v[k] < max(ex_y)):
ax.scatter(photf_vmi[k],
photf_i[k],
s=12,
c='k',
marker='x')
# Plot MIST isochrones corresponding to MATCH best-fit parameters.
ax, iso, mist_pts = plotisocmd(ax=ax,
vvcrit=vvcrit,
best_dict=best_dict,
filters=filters,
truths=truths,
extent=extent)
# saving the plot of data points w/ isochrones overlaid:
savename = os.path.join(
plotpath,
'cmd_vvcrit{:.1f}_m2lnP{:.2f}.png'.format(vvcrit, a_cmd.fit))
fig.savefig(savename, dpi=300)
# create a MATCH pg style plot using the .cmd file:
# using photf_pts_alt because MATCH does its CMDs with V-I vs. V & not sure if this is changeable.
a_cmd.pgcmd(photf_pts=photf_pts_alt,
mist_pts=mist_pts,
best_list=[
best_dict[vvcrit]['lage']['val'],
best_dict[vvcrit]['feh']['val']
])
# closing figure before moving on to next plots.
fig.clf()
#==================================================================
# For plotting all best fits together (on another matplotlib axis):
#==================================================================
# Only scatter plot data points once (on axis 2/ all isochrones).
if j == 0:
allax.scatter(*photf_pts, lw=0, s=8, c='r')
# Mark excluded region(s)
if expts != None:
for k in range(len(photf_pts[0])):
if (min(ex_x) < photf_vmi[k] < max(ex_x)) & (
min(ex_y) < photf_v[k] < max(ex_y)):
allax.scatter(photf_vmi[k],
photf_i[k],
s=12,
c='k',
marker='x')
# all best isochrones are plotted together (no +/- 1 sigma isochrones are plotted):
iso.isoplot(allax,
xlims=extent[0:2],
ylims=extent[2:],
shade=vvcrit,
legloc='upper right')
# track which cmd has the overall highest lnP & corresponding v/vcrit.
if a_cmd.fit < bestfit:
bestfit = a_cmd.fit
bestvvc = vvcrit
# write best fit solutions +/- uncertainties to a summary file:
# -------------------------------------------------------------
outlines = [
'Hyades\n', '_______\n', 'Tycho B, V\n', '--------\n', '\n', '\n',
'\n', '2MASS J, Ks\n', '-------\n', '\n', '\n', '\n', 'Praesepe\n',
'________\n', 'Tycho B, V\n', '--------\n', '\n', '\n', '\n',
'2MASS J, Ks\n', '-------\n', '\n', '\n', '\n', 'Pleiades\n_',
'_______\n', 'Tycho B, V\n', '--------\n', '\n', '\n', '\n',
'2MASS J, Ks\n', '-------\n', '\n', '\n', '\n'
]
with open('/n/home12/sgossage/match_results.txt', 'rw+') as sumfile:
# get current lines of the summary file:
inlines = sumfile.readlines()
#print(inlines)
#print(len(inlines))
#print(len(outlines))
for i, line in enumerate(inlines):
# maintain lines that already contain best fits
if 'Best' in line:
print(line)
outlines[i] = line
# update best fit line for the current run:
if 'Hyades' in cluster_name:
n = 4
elif 'Praesepe' in cluster_name:
n = 16
elif 'Pleiades' in cluster_name:
n = 28
if '2MASSJK' in photbase:
n += 5
# print these lines when updating...
# ...if using a distribution of roation rates...
if '_rot' in cluster_name:
n += 1
outlines[n] = 'Best (rotation distribution) v/vcrit = {:.1f}, ' \
'age = {:.1f} +/- {:.1f} Myrs,' \
'[Fe/H] = {:.2f} +/- {:.2f}\n'.format(bestvvc,
10**best_dict[bestvvc]['lage']['val']/1e6,
(10**(best_dict[bestvvc]['lage']['val'] + \
best_dict[bestvvc]['lage']['err']) - \
10**best_dict[bestvvc]['lage']['val'])/1e6,
best_dict[bestvvc]['feh']['val'],
best_dict[bestvvc]['feh']['err'])
# ...or else if not using a distribution of rotation rates.
else:
outlines[n] = 'Best v/vcrit = {:.1f}, ' \
'age = {:.1f} +/- {:.1f} Myrs,' \
'[Fe/H] = {:.2f} +/- {:.2f}\n'.format(bestvvc,
(10**best_dict[bestvvc]['lage']['val'])/1e6,
(10**(best_dict[bestvvc]['lage']['val'] + \
best_dict[bestvvc]['lage']['err']) - \
10**best_dict[bestvvc]['lage']['val'])/1e6,
best_dict[bestvvc]['feh']['val'],
best_dict[bestvvc]['feh']['err'])
# return to top of file & then write lines out.
sumfile.seek(0)
sumfile.writelines(outlines)
# After all, save image of the plot containing all isochrones plotted side by side.
savename = os.path.join(plotpath,
'cmdall_bestvvcrit{:.1f}.png'.format(bestvvc))
allfig.savefig(savename, dpi=300)
allfig.clf()
#=========================================================
if bestax == None:
bestfig = plt.figure()
bestax = bestfig.add_subplot(111)
bestax.scatter(*photf_pts, lw=0, s=8, c='r')
if expts != None:
for k in range(len(photf_pts[0])):
if (min(ex_x) < photf_vmi[k] < max(ex_x)) & (min(ex_y) < photf_v[k]
< max(ex_y)):
bestax.scatter(photf_vmi[k],
photf_i[k],
s=12,
c='k',
marker='x')
bestax, bestiso, mist_pts = plotisocmd(ax=bestax,
vvcrit=bestvvc,
best_dict=best_dict,
filters=filters,
truths=truths,
extent=extent)
if savebest:
# save the figure.
bestfig.savefig(os.path.join(
plotpath,
'bestcmd_vvcrit{:.1f}_m2lnP{:.2f}.png'.format(vvcrit, a_cmd.fit)),
dpi=300)
bestfig.clf()
return bestax, allax
def main(cluster_name,
photbase,
jobid,
filters,
dmod,
ebv,
local,
vvc,
outname=None,
savebest=True,
bestax=None,
justbest=False,
txtlbl_kwargs={
"frameon": False,
"fontsize": 14,
"loc": 4
},
texts=['vvc', 'age', 'feh'],
hess_datapts=True,
hess_modeliso=True):
global match_dir
global scrn_dir
global outpath
global plotpath
# path/to/Cluster
match_dir = get_workdir(data=cluster_name)
# path to MATCH .scrn files:
scrn_dir = os.path.join(get_scrndir(data=cluster_name), photbase,
'SLURM{:s}'.format(jobid))
# path to MATCH .cmd and .out files:
outpath = os.path.join(get_outdir(data=cluster_name), photbase,
'SLURM{:s}'.format(jobid))
# path to place output plots in:
plotpath = os.path.join(outpath, 'plots')
#print(outpath)
rot = False
if 'rot' in cluster_name:
rot = True
if outname != None:
# Create corner PDF plot:
if vvc != 'all':
flist = glob.glob(
os.path.join(scrn_dir, '*vvcrit{:.1f}*.scrn'.format(vvc)))
params = ['lage', 'logZ']
else:
flist = glob.glob(os.path.join(scrn_dir, '*.scrn'))
ssp.combine_files(flist, outfile=outname)
# Create a match SSP object from the combined scrns of the given runs:
combineddata = ssp.SSP(outname)
# Corner plots:
# 'lage', 'vvcrit', 'logZ', 'dmod', 'Av'
#pdffig, pdfax = combineddata.pdf_plots(['lage','logZ','vvcrit'], twod=False, quantile=True, cmap=plt.cm.Reds, interpolateq=True)
#plt.close()
#pdffig, pdfax = combineddata.pdf_plots(['lage','logZ','vvcrit'], twod=True, quantile=True, cmap=plt.cm.Reds, interpolateq=True)
#plt.savefig(os.path.join(plotpath, 'pdfcornerplot.png'))
#plt.close()
#vbests, vvcrits, qdict, bfdict = marginalize(cluster_name, photbase, sys.argv[3], sys.argv[4], local=local)
#vvcrits, best_dict, qdict, bfdict = marginalize(cluster_name, photbase, sys.argv[3], sys.argv[4], local=local, incvvc=incvvc)
# by default, readssp() will return an AVERAGE (over all runs) of the bestfit parameter value and uncertainty.
# the original, unaveraged values are derived from the sspcombine solutions.
if vvc == 'all':
vvcrits = [0.0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6]
else:
vvcrits = [vvc]
#print(vvcrits)
best_dict = {}
# for vvc in vvcrits:
# read sspcombine output for age results (subsampled solution being used):
# sspbest_age, ssp_upageerr, ssp_loageerr = readssp(photname_base=photbase, data=cluster_name,
# param='age', vvc=vvc, jobidstr=jobid)
# read sspcombines for metallicity results:
# sspbest_logz, ssp_uplogzerr, ssp_lologzerr = readssp(photname_base=photbase, data=cluster_name,
# param='logZ', vvc=vvc, jobidstr=jobid)
# dictionary whose keys are the corresponding vvc values for the best fit values found above (dmod, Av fixed):
# best_dict[vvc] = {'lage': {'val': sspbest_age, 'uperr': ssp_upageerr, 'loerr': ssp_loageerr},
# 'feh': {'val': sspbest_logz, 'uperr': ssp_uplogzerr, 'loerr': ssp_lologzerr},
# 'dmod': dmod,
# 'ebv': ebv}
#plt.clf()
# clearing out pre-existing output files that this script may have produced in prev. runs:
if not justbest:
for f in glob.glob(os.path.join(outpath, 'cmd*.png')):
os.remove(f)
for f in glob.glob(os.path.join(plotpath, 'cmd*.png')):
os.remove(f)
for f in glob.glob(os.path.join(plotpath, '*.txt')):
os.remove(f)
# if req., correct 2MASS filter names to those recognizable by MIST.
for i, afilter in enumerate(filters):
if afilter == 'J':
filters[i] = '2MASS_J'
elif afilter == 'Ks':
filters[i] = '2MASS_Ks'
elif afilter == 'B':
filters[i] = 'Bessell_B'
elif afilter == 'V':
filters[i] = 'Bessell_V'
# Retrieve the magnitude limits used in the fit (from the calcsfh .param file used):
param_fname = glob.glob(
os.path.join(match_dir, 'csfh_param',
'calcsfh_{:s}.param'.format(photbase)))[0]
with open(param_fname, 'r') as f:
lines = f.readlines()
# Color limits:
color_lims = lines[4].split(' ')
minvmi = float(color_lims[3])
maxvmi = float(color_lims[4])
ibin = float(color_lims[0])
vibin = float(color_lims[1])
ilims = lines[6].split(' ')
imin = float(ilims[1])
imax = float(ilims[0])
exclusions = lines[7].split('\n')[0]
ntbins = int(lines[8])
tbins = lines[9:8 + ntbins]
for i, aline in enumerate(tbins):
tbins[i] = np.array(
list(map(float,
aline.split('\n')[0].split(' ')[-2:])))
# Manage exclude gates (right now only works for one region):
# If there are exclude gates...
if int(exclusions[0]) > 0:
# Get the exclude gate points:
expts = np.array(list(map(float, exclusions.split(' ')[1:-1])))
ex_x = expts[::2]
ex_y = expts[1::2]
ex_xy = zip(ex_x, ex_y)
else:
expts = None
ex_xy = None
# Also get the residuals (???):
# For overplotting data points, get the data points CMD x, y values from the photometry file:
photf_path = os.path.join(match_dir, 'photometry',
'{:s}.phot'.format(photbase))
photf_v_raw = np.genfromtxt(photf_path, usecols=(0, ))
photf_i_raw = np.genfromtxt(photf_path, usecols=(1, ))
if any(photf_v_raw >= 99.999) or any(photf_i_raw >= 99.999):
print("Bad photometric values detected in {:s}. Cutting them out...".
format(photf_path))
photf_v = photf_v_raw[(photf_v_raw < 99.999) & (photf_i_raw < 99.999)]
photf_i = photf_i_raw[(photf_v_raw < 99.999) & (photf_i_raw < 99.999)]
else:
photf_v = photf_v_raw
photf_i = photf_i_raw
photf_vmi = photf_v - photf_i
# Tuple storing the x, y values for plotting the data on a CMD (using color on x, red filter on y):
photf_pts = (photf_vmi, photf_i)
# match uses blue filter on y axis.
photf_pts_alt = (photf_vmi, photf_v)
bestfit = 99999.9
# Plot best isochrone for each v/vcrit individually:
#=================================================================
allfig = plt.figure()
allax = allfig.add_subplot(111)
probfig, probax = plt.subplots(1, 1)
# iterate through all v/vcrits used for fitting:
for j, vvcrit in enumerate(vvcrits):
print(outpath)
# get the .cmd file for making a MATCH pg style plot:
if 'flat' in cluster_name:
fname = glob.glob(os.path.join(outpath, '*vvcflat*.cmd'))[0]
else:
try:
fname = glob.glob(
os.path.join(outpath, '*vvc{:.1f}*.cmd'.format(vvcrit)))[0]
except IndexError:
fname = glob.glob(
os.path.join(outpath,
'*vvcrit{:.1f}*.cmd'.format(vvcrit)))[0]
# use Phil's code to make a CMD object from the .cmd file:
a_cmd = cmd.CMD(fname, ymag='I')
# store the axis extents of the CMD data
extent = a_cmd.extent
# read sspcombine output for age results (subsampled solution being used):
sspbest_age, ssp_upageerr, ssp_loageerr = readssp(
photname_base=photbase,
data=cluster_name,
param='age',
vvc=vvcrit,
jobidstr=jobid)
# read sspcombines for metallicity results:
sspbest_logz, ssp_uplogzerr, ssp_lologzerr = readssp(
photname_base=photbase,
data=cluster_name,
param='logZ',
vvc=vvcrit,
jobidstr=jobid)
# dictionary whose keys are the corresponding vvc values for the best fit values found above (dmod, Av fixed):
best_dict[vvcrit] = {
'lage': {
'val': sspbest_age,
'uperr': ssp_upageerr,
'loerr': ssp_loageerr
},
'feh': {
'val': sspbest_logz,
'uperr': ssp_uplogzerr,
'loerr': ssp_lologzerr
},
'dmod': dmod,
'ebv': ebv,
'fit': a_cmd.fit
}
# to plot the fit statistic vs. v/vcrit:
probax.scatter(vvcrit, a_cmd.fit)
if True: #not justbest:
fig = plt.figure()
ax = fig.add_subplot(111)
# scatter plot of stars from data file (observations):
ax = pp.plotphot(*photf_pts, ax=ax)
# X out masked data points.
if expts is not None:
for k in range(len(photf_pts[0])):
if (min(ex_x) < photf_vmi[k] <
max(ex_x)) & (min(ex_y) < photf_v[k] < max(ex_y)):
ax.scatter(photf_vmi[k],
photf_i[k],
s=40,
c='k',
marker='x')
# Plot MIST isochrones corresponding to MATCH best-fit parameters.
ax, iso, isou, isol, mist_pts, axlims = plotisocmd(
ax=ax,
vvcrit=vvcrit,
best_dict=best_dict,
filters=filters,
extent=extent,
txtlbl_kwargs=txtlbl_kwargs,
rot=rot,
texts=texts)
ax.set_title('{:s}'.format(cluster_name.split('rot')[0]))
# saving the plot of data points w/ isochrones overlaid:
savename = os.path.join(
plotpath,
'cmd_vvc{:.1f}_m2lnP{:.2f}.png'.format(vvcrit, a_cmd.fit))
fig.savefig(savename, dpi=300)
# create a MATCH pg style plot using the .cmd file:
# using photf_pts_alt because MATCH does its CMDs with V-I vs. V & not sure if this is changeable.
pgcmd_kwargs = {}
if hess_datapts:
pgcmd_kwargs['photf_pts'] = photf_pts_alt
if hess_modeliso:
pgcmd_kwargs['mist_pts'] = mist_pts
pgcmd_kwargs['best_list'] = [
best_dict[vvcrit]['lage']['val'],
best_dict[vvcrit]['feh']['val']
]
pgcmd_kwargs['ymag'] = 'V'
# four panel plot of all hess diagrams:
# a_cmd.pgcmd(**pgcmd_kwargs)
for m in range(4):
# each of the 4 plotted separately:
pgcmd_kwargs['hess_i'] = m
a_cmd.plthess(**pgcmd_kwargs)
# closing figure before moving on to next plots.
fig.clf()
# track which cmd has the overall highest lnP & corresponding v/vcrit.
if a_cmd.fit < bestfit:
bestfit = a_cmd.fit
bestvvc = vvcrit
bestcmd = a_cmd
# save plot of fit statistic vs. v/vc:
probax.set_xlabel(r'$\Omega/\Omega_c$')
probax.set_ylabel(r'$-2\ln$P')
probfig.savefig(os.path.join(plotpath, 'lnpvsrot.png'))
# write best fit solutions +/- uncertainties to a summary file:
# -------------------------------------------------------------
print('WRITING RESULTS...')
outlines = [
'Hyades\n', '_______\n', 'Tycho B, V\n', '--------\n', '\n', '\n',
'\n', '2MASS J, Ks\n', '-------\n', '\n', '\n', '\n', 'Praesepe\n',
'________\n', 'Tycho B, V\n', '--------\n', '\n', '\n', '\n',
'2MASS J, Ks\n', '-------\n', '\n', '\n', '\n', 'Pleiades\n_',
'_______\n', 'Tycho B, V\n', '--------\n', '\n', '\n', '\n',
'2MASS J, Ks\n', '-------\n', '\n', '\n', '\n'
]
# writing results to a .txt file for reference (don't add fake data results though):
if 'Fakedata' not in cluster_name:
with open('/n/home12/sgossage/match_results.txt', 'r+') as sumfile:
# get current lines of the summary file:
inlines = sumfile.readlines()
# print(inlines)
#print(len(inlines))
#print(len(outlines))
for i, line in enumerate(inlines):
# maintain lines that already contain best fits
if ('Best' in line) | ('v/vcrit = 0.0' in line):
print(line)
outlines[i] = line
# update best fit line for the current run:
if 'Hyades' in cluster_name:
n = 4
elif 'Praesepe' in cluster_name:
n = 16
elif 'Pleiades' in cluster_name:
n = 28
if '2MASSJK' in photbase:
n += 5
# print these lines when updating...
# ...if using a distribution of roation rates...
if ('_rot' in cluster_name) | ('_flat' in cluster_name):
n += 2
outlines[n] = 'Best (rotation distribution) v/vcrit = {:.1f}, ' \
'age = {:.1f} + {:.1f} - {:.1f} Myrs,' \
'[Fe/H] = {:.2f} + {:.2f} - {:.2f}, -2lnP ={:.2f} \n'.format(bestvvc,
(10**best_dict[bestvvc]['lage']['val'])/1e6,
(10**(best_dict[bestvvc]['lage']['val'] + \
best_dict[bestvvc]['lage']['uperr']) - \
10**best_dict[bestvvc]['lage']['val'])/1e6,
(10**best_dict[bestvvc]['lage']['val'] - \
10**(best_dict[bestvvc]['lage']['val'] - \
best_dict[bestvvc]['lage']['loerr']))/1e6,
best_dict[bestvvc]['feh']['val'],
best_dict[bestvvc]['feh']['uperr'],
best_dict[bestvvc]['feh']['loerr'],
best_dict[bestvvc]['fit'])
# ...or else if not using a distribution of rotation rates.
else:
outlines[n] = 'Best v/vcrit = {:.1f}, ' \
'age = {:.1f} + {:.1f} - {:.1f} Myrs,' \
'[Fe/H] = {:.2f} + {:.2f} - {:.2f}, -2lnP = {:.2f}\n'.format(bestvvc,
(10**best_dict[bestvvc]['lage']['val'])/1e6,
(10**(best_dict[bestvvc]['lage']['val'] + \
best_dict[bestvvc]['lage']['uperr']) - \
10**best_dict[bestvvc]['lage']['val'])/1e6,
(10**best_dict[bestvvc]['lage']['val'] - \
10**(best_dict[bestvvc]['lage']['val'] - \
best_dict[bestvvc]['lage']['loerr']))/1e6,
best_dict[bestvvc]['feh']['val'],
best_dict[bestvvc]['feh']['uperr'],
best_dict[bestvvc]['feh']['loerr'],
best_dict[bestvvc]['fit'])
n += 1
outlines[n] = 'v/vcrit = {:.1f}, ' \
'age = {:.1f} + {:.1f} - {:.1f} Myrs,' \
'[Fe/H] = {:.2f} + {:.2f} - {:.2f}, -2lnP = {:.2f}\n'.format(0.0,
(10**best_dict[0.0]['lage']['val'])/1e6,
(10**(best_dict[0.0]['lage']['val'] + \
best_dict[0.0]['lage']['uperr']) - \
10**best_dict[0.0]['lage']['val'])/1e6,
(10**best_dict[0.0]['lage']['val'] - \
10**(best_dict[0.0]['lage']['val'] - \
best_dict[0.0]['lage']['loerr']))/1e6,
best_dict[0.0]['feh']['val'],
best_dict[0.0]['feh']['uperr'],
best_dict[0.0]['feh']['loerr'],
best_dict[0.0]['fit'])
# return to top of file & then write lines out.
sumfile.seek(0)
sumfile.writelines(outlines)
#=========================================================
if bestax == None:
bestfig = plt.figure()
bestax = bestfig.add_subplot(111)
bestax.set_title('{:s}'.format(cluster_name.split('rot')[0]))
#bestax.scatter(*photf_pts, lw=0, s=8, c='r')
bestax = pp.plotphot(*photf_pts, ax=bestax)
# if expts != None:
## for k in range(len(photf_pts[0])):
# if (min(ex_x) < photf_vmi[k] < max(ex_x)) & (min(ex_y) < photf_v[k] < max(ex_y)):
# bestax.scatter(photf_vmi[k], photf_i[k], s=40, c='k', marker='x')
bestax, bestiso, bestisou, bestisol, mist_pts, axlims = plotisocmd(
ax=bestax,
vvcrit=bestvvc,
best_dict=best_dict,
filters=filters,
extent=extent,
txtlbl_kwargs=txtlbl_kwargs,
rot=rot,
texts=texts)
if savebest:
# save the figure.
bestfig.savefig(os.path.join(
plotpath,
'bestcmd_vvcrit{:.1f}_m2lnP{:.2f}.png'.format(vvcrit, a_cmd.fit)),
dpi=300)
bestfig.clf()
return bestax, allax, (
bestiso, bestisou,
bestisol), bestcmd, photf_pts, mist_pts, ex_xy, axlims