def overlapping(spec_or_bins_a, spec_or_bins_b):
"""Check if there is any overlap."""
getbins = lambda sob: sob if type(sob) is np.ndarray else wbins(sob)
wbinsa, wbinsb = map(getbins, [spec_or_bins_a, spec_or_bins_b])
ainb0, ainb1 = [mnp.inranges(w, wbinsb, [0, 0]) for w in wbinsa.T]
bina0, bina1 = [mnp.inranges(w, wbinsa, [0, 0]) for w in wbinsb.T]
return np.any(ainb0 | ainb1) or np.any(bina0 | bina1)
def specSnapshot(star,
inst,
trange,
wrange,
n=100,
ax=None,
vCen=None,
maxpts=500,
**kwargs):
if ax is None: ax = plt.gca()
ph, p = io.readphotons(star, inst)
keep = mnp.inranges(p['time'], trange)
p = p[keep]
gtis = np.array([ph['gti'].data['start'], ph['gti'].data['stop']]).T
gt = mnp.range_intersect([trange], gtis)
dt = np.sum(gt[:, 1] - gt[:, 0])
w0, w1, spec, err = sp.smooth_spec(p['wavelength'], p['epera'], n)
spec, err = spec / dt, err / dt
w = (w0 + w1) / 2.0
keep, = np.nonzero(mnp.inranges(w, wrange))
keep = np.insert(keep, [0, len(keep)], [keep[0] - 1, keep[-1] + 1])
w, spec, err = w[keep], spec[keep], err[keep]
if len(w) > maxpts:
keep = un.downsample_even(w, maxpts)
w, spec, err = w[keep], spec[keep], err[keep]
velocify = lambda w: (w - vCen) / vCen * 3e5
if vCen is not None:
w = velocify(w)
ax.set_xlabel('Doppler Velocity [km s$^{-1}$]')
ax.set_xlim(map(velocify, wrange))
else:
ax.set_xlabel('Wavelength [$\AA$]')
ax.set_xlim(wrange)
ax.set_ylabel(fluxlabel)
return pu.errorpoly(w, spec, err, **kwargs)
def stackspecs(specs, range, norm=True, offfac=1.0, xlim=None):
ax = plt.gca()
if xlim is not None: plt.xlim(xlim)
off = 0.0
for spec in specs:
w = utils.wbins(spec)
wmid = (w[:, 0] + w[:, 1]) / 2.0
f = spec['flux']
fac = np.max(f[mnp.inranges(wmid, range)])
if norm:
f /= fac
f += off
l = specplot(w, f)
xlim = ax.get_xlim()
mid = (xlim[0] + xlim[1]) / 2.0
plt.text(mid,
off,
spec.meta['STAR'],
bbox=dict(fc='w', alpha=0.5, lw=0))
if norm:
off += offfac
else:
off += fac * offfac
plt.ylim(0.0, off)
def showFlareStats(star, inst, label, trange, dt=30.0, ax=None):
if ax is None: ax = plt.gca()
trange = np.array(trange)
flareTbl, bands = io.readFlareTbl(star, inst, label)
# make lightcurve
groups = [range(len(bands))]
curve = reduce.auto_curve(star,
inst,
bands,
dt,
appx=False,
groups=groups,
fluxed=True)
t0, t1, flux, err = zip(*curve)[0]
t = (t0 + t1) / 2.0
# narrow lightcurve down to range of interest
keep = (t1 > trange[0]) & (t0 < trange[1])
t, flux, err = t[keep], flux[keep], err[keep]
ymax = flux.max()
# keep only largest flare in time range of interest
keep = (flareTbl['start'] < trange[1]) & (flareTbl['stop'] > trange[0])
flareTbl = flareTbl[keep]
iflare = np.argmax(flareTbl['PEW'])
flare = flareTbl[iflare]
# reference time to start of flare
tref = flare['start']
t, flare['start'], flare['stop'], trange = t - tref, flare[
'start'] - tref, flare['stop'] - tref, trange - tref
# plot lightcurve highlighting flare points
flarepts = mnp.inranges(t, [flare['start'], flare['stop']])
ax.errorbar(t[~flarepts],
flux[~flarepts],
err[~flarepts],
fmt='ko',
capsize=0)
ax.errorbar(t[flarepts],
flux[flarepts],
err[flarepts],
fmt='rd',
capsize=0)
# reverse-engineer mean flux
luminosity = flare['energy'] / flare['PEW']
dist = sc.quickval(rc.proppath, star, 'dist')
mnflux = luminosity / 4 / np.pi / (dist * 3.08567758e18)**2
# plot mean flux
ax.axhline(mnflux, color='gray', ls='--')
tmid = np.mean(t[~flarepts])
mnfluxstr = '$\overline{F}$ = %.1e' % mnflux
ax.text(tmid, mnflux + ymax / 20.0, mnfluxstr, ha='center')
# # plot duration
# dt = flare['stop'] - flare['start']
# y = 1.05 * np.max(flux + err)
# ax.annotate('', xy=(flare['start'], y), xytext=(flare['stop'], y), arrowprops=dict(arrowstyle='<->', color='r'))
# # ax.arrow(flare['start'], y, dt, 0.0, length_includes_head=True)
# # ax.arrow(flare['start'], y, dt, 0.0, length_includes_head=True, head_starts_at_zero=True)
# ax.text(tmidFlare, y*1.02, '{:.0f} s'.format(dt), ha='center', color='r')
# fill area under flare
# tmidFlare = np.sum(t[flarepts]*flux[flarepts])/np.sum(flux[flarepts])
lo = [mnflux] * np.sum(flarepts)
ax.fill_between(t[flarepts], lo, flux[flarepts], color='r', alpha=0.3)
y = 0.2 * np.max(flux)
intlbl = 'equiv. width = {:.0f} s\nenergy = {:.1g} erg'.format(
flare['PEW'], flare['energy'])
ax.text(0.05,
0.95,
intlbl,
ha='left',
va='top',
color='r',
transform=ax.transAxes)
# axes
ax.set_xlabel('Time [s]')
ax.set_ylabel(linefluxlabel)
def spectrumMovieFrames(star,
inst,
band,
trange,
dt,
smoothfac,
axspec,
axcurve,
folder,
dpi=80,
velocityplot=False,
reftrange=None,
dryRun=False,
ylim=None):
ph, photons = io.readphotons(star, inst)
band, trange, reftrange = map(np.asarray, [band, trange, reftrange])
fig = axcurve.get_figure()
figwidth = fig.get_figwidth()
axwidth = axcurve.get_position().width * figwidth
axPix = axwidth * dpi
def goodN(Nphotons):
n = 100 # SN of 10
if Nphotons / n > axPix: # don't need to sample more finely than the pixel scale
n = Nphotons / axPix
elif Nphotons / n < 20: # want at least some resolution
n = max(Nphotons / 20, 9) # but no less than SN of 3ish
return n
# re-reference times to start of time range
tref = trange[0]
photons['time'] -= tref
if reftrange is not None: reftrange -= tref
trange -= tref
if velocityplot:
velocify = lambda w: (w - velocityplot) / velocityplot * 3e5
vband = velocify(band)
p = photons
tkeep = [
max(p['time'][0], -trange[1] * 0.5),
min(p['time'][-1], trange[1] * 1.5)
]
dw = band[1] - band[0]
wkeep = [band[0] - 5 * dw, band[1] + 5 * dw]
# get rid of superfluous counts
keep = mnp.inranges(p['time'], tkeep) & mnp.inranges(
p['wavelength'], wkeep)
p = p[keep]
## make lightcurve and set up initial plot
nlc = goodN(
np.sum(
mnp.inranges(p['wavelength'], band)
& mnp.inranges(p['time'], trange)))
t0, t1, lc, lcerr = sp.smooth_curve(p['time'],
p['wavelength'],
p['epera'],
nlc,
bands=[band],
trange=tkeep)
tlc = (t0 + t1) / 2.0
axcurve.set_xlim(trange)
axcurve.set_xlabel('Time [s]')
axcurve.set_ylabel('Integrated Flux \n[erg cm$^{-2}$ s$^{-1}$]')
inrange = mnp.inranges(tlc, trange)
pu.errorpoly(tlc[inrange],
lc[inrange],
lcerr[inrange],
'k-',
ax=axcurve,
alpha=0.3,
ealpha=0.15)
# make spectrum frames
T = dt * smoothfac
nframes = int(round((trange[1] - trange[0] - T) / dt))
t1s = np.linspace(trange[0] + T, trange[1], nframes)
t0s = t1s - T
wList, specList, errList = [], [], []
for t0, t1 in zip(t0s, t1s):
inInterval = mnp.inranges(p['time'], [t0, t1])
pt = p[inInterval]
n = goodN(np.sum(mnp.inranges(pt['wavelength'], band)))
w0, w1, spec, err = sp.smooth_spec(pt['wavelength'], pt['epera'], n,
wkeep)
wList.append((w0 + w1) / 2.0)
specList.append(spec / T)
errList.append(err / T)
## set up spectrum plot
if velocityplot:
axspec.set_xlabel('Doppler Velocity [km s$^{-1}$]')
axspec.set_xlim(vband)
else:
axspec.set_xlabel('Wavelength [$\AA$]')
axspec.set_xlim(band)
axspec.set_ylabel(fluxlabel)
if ylim is None:
ymin = min([np.min(s - e) for s, e in zip(specList, errList)])
ymax = max([np.max(s + e) for s, e in zip(specList, errList)])
axspec.set_ylim(ymin, ymax)
else:
axspec.set_ylim(ylim)
# compute and plot reference spectrum, if desired
if reftrange is not None:
gtis = np.array([ph['gti'].data['start'], ph['gti'].data['stop']
]).T - tref
gt = mnp.range_intersect([reftrange], gtis)
Tref = np.sum(gt[:, 1] - gt[:, 0])
keep = mnp.inranges(photons['time'], reftrange) & mnp.inranges(
photons['wavelength'], wkeep)
pt = photons[keep]
nref = goodN(np.sum(mnp.inranges(pt['wavelength'], band)))
w0, w1, spec, _ = sp.smooth_spec(pt['wavelength'], pt['epera'], nref,
wkeep)
w = (w0 + w1) / 2.0
if velocityplot:
w = velocify(w)
spec = spec / Tref
axspec.plot(w, spec, 'k-', alpha=0.3)
# make folder to save frames in if it doesn't exist
if not os.path.exists(folder):
os.mkdir(folder)
## loop through frames
if dryRun:
nframes = dryRun
for i in range(nframes):
# plot time range on lightcurve
span = axcurve.axvspan(t0s[i], t1s[i], color='k', alpha=0.2)
inrange = mnp.inranges(tlc, [t0s[i], t1s[i]])
linelc, polylc = pu.errorpoly(tlc[inrange],
lc[inrange],
lcerr[inrange],
'k-',
ax=axcurve,
ealpha=0.3)
# plot spectrum
w, spec, err = wList[i], specList[i], errList[i]
inrange = mnp.inranges(w, band)
inrange = np.nonzero(inrange)[0]
inrange = np.insert(inrange, [0, len(inrange)],
[inrange[0] - 1, inrange[-1] + 1])
ww, ss, ee = w[inrange], spec[inrange], err[inrange]
ss[[0, -1]] = np.interp(band, ww, spec[inrange])
ee[[0, -1]] = np.interp(band, ww, err[inrange])
ww[[0, -1]] = band
if velocityplot:
ww = velocify(ww)
linespec, polyspec = pu.errorpoly(ww,
ss,
ee,
'k-',
ax=axspec,
ealpha=0.2)
# save frame
path = os.path.join(folder, '{:04d}.png'.format(i))
fig.savefig(path, dpi=dpi)
# remove plots
[obj.remove() for obj in [span, linelc, polylc, linespec, polyspec]]
def lightcurveCompendium(stars='hosts',
figure=None,
flarecut=2.0,
flarelabel='SiIV',
dt=30.0,
colorful=False):
"""
Create a compendium of lightcurves for the specified stars highlighting flares.
"""
fig = plt.gcf() if figure is None else figure
if colorful:
colors = itertools.cycle(['b', 'g', 'r'])
else:
colors = itertools.cycle(['k'])
inst = 'hst_cos_g130m'
if stars == 'hosts':
stars = filter(
lambda s: len(db.findfiles('u', inst, 'corrtag_a', s)) >= 4,
rc.observed)
## SET UP AXES
## -----------
# setup axes with just bottom axis showing
fig.set_facecolor('w')
fig.add_axes((0.23, 0.14, 0.72, 0.84), frameon=False)
ax = fig.axes[0]
xax = ax.get_xaxis()
xax.tick_bottom()
yax = ax.get_yaxis()
yax.set_visible(False)
ax.set_clip_on(False)
fntsz = mpl.rcParams['font.size']
spacedata = (fntsz / 72.0 / fig.get_figwidth()) / 0.8 * 14000
# common keywords to use in errorbar plot
ekwds = dict(fmt='.', capsize=0.0)
alphaNF = 0.1 if colorful else 0.4
## MAKE LIGHTCURVES
## ----------------
bands = rc.flare_bands[inst]
offset, offsets = 0.0, []
for star, color in zip(stars, colors):
# get flare info
flares, bands = io.readFlareTbl(star, inst, flarelabel)
curve = reduce.auto_curve(star,
inst,
dt=30.0,
bands=bands,
groups=[range(len(bands))])
t0, t1, cps, err = zip(*curve)[0]
t = (t0 + t1) / 2.0
# get rid of gaps
s, j, _ = mnp.shorten_jumps(t, maxjump=10 * dt, newjump=30 * dt)
# identify flare pts
flares = flares[flares['PEWratio'] > flarecut]
flare_ranges = np.array([flares['start'], flares['stop']]).T
flarepts = mnp.inranges(t, flare_ranges)
# normalize the data to median
med = np.median(cps[~flarepts])
y = cps / med
y -= 1.0
e = err / med
# normalize data to max - min
ymax = np.max(y)
y /= ymax
e /= ymax
# cull negative outliers
good = y > -ymax
s, y, e, flarepts = [a[good] for a in [s, y, e, flarepts]]
# offset data in y
offset = offset - np.min(y - e)
offsets.append(offset)
yo = y.copy() + offset
# plot data
ax.errorbar(s[~flarepts],
yo[~flarepts],
e[~flarepts],
alpha=alphaNF,
color=color,
**ekwds)
ax.axhline(offset, linestyle='--', color='k', alpha=0.5)
flarecolor = 'r' if color == 'k' else color
ax.errorbar(s[flarepts],
yo[flarepts],
e[flarepts],
color=flarecolor,
**ekwds)
# make arrow
xy = (0.0, offset)
xytext = (0.0, offset + 1.1)
arrowprops = dict(arrowstyle='<-', color=color)
ax.annotate('', xy=xy, xytext=xytext, arrowprops=arrowprops)
# label arrow
ymaxstr = '{:4.1f}'.format(ymax + 1.0)
ax.text(-0.5 * spacedata,
offset,
'1.0',
va='bottom',
ha='right',
color=color,
fontsize=fntsz * 0.8)
ax.text(-0.5 * spacedata,
offset + 1.1,
ymaxstr,
va='top',
ha='right',
color=color,
fontsize=fntsz * 0.8)
# label star
starlbl = starprops['name tex'][star]
ax.text(-3.0 * spacedata,
offset + 0.5,
starlbl,
va='center',
ha='right',
color=color)
# increase offset
offset += np.max(y + e)
offset += 0.3
ax.autoscale(axis='both', tight=True)
# add x axis line and label
ax.axhline(0, color='k')
ax.set_xlabel('Time, Observation Gaps Shortened (s)')
def killnegatives(spectbl,
sep_insts=False,
quickndirty=False,
minSN=None,
res_limit=1.0):
"""
Removes negative bins by summing with adjacent bins until there are no negative bins left. I.e. the resolution in
negative areas is degraded until the flux is no longer negative.
Parameters
----------
spectbl
Returns
-------
newtbl
A new bare-bones spectbl that has bin edges, flux, and error
WARNING: the obs date, instrument, etc. columns will all get set to default values, as will all of the
metadata except for 'star'
"""
# if not np.any(spectbl['flux'] < 0):
# return spectbl
if sep_insts:
return inst_by_inst(spectbl,
killnegatives,
sep_insts=False,
quickndirty=quickndirty,
minSN=minSN)
if hasgaps(spectbl):
return gap_by_gap(spectbl,
killnegatives,
sep_insts=False,
quickndirty=quickndirty,
minSN=minSN)
w0, w1, f_dsty, e_dsty = [
spectbl[s].copy() for s in ['w0', 'w1', 'flux', 'error']
]
line_bands = np.vstack(rc.line_bands.values())
line_bands = line_bands[np.argsort(line_bands[:, 0]), :]
untouchable = mnp.inranges(w0, line_bands) | mnp.inranges(w1, line_bands)
if minSN is None:
untouchable = untouchable & (f_dsty >= 0)
else:
untouchable = untouchable & (f_dsty / e_dsty >= minSN)
dw = w1 - w0
f, e = f_dsty * dw, e_dsty * dw
v = e**2
# I had this kind of vectorized once, but ultimately I think it just made for terrible readability with little gain in speed
# print "bad bins remaining:"
while True:
if minSN is None:
n = np.sum(f < 0)
else:
n = np.sum(f / np.sqrt(v) < minSN)
# print n
if n == 0:
break
# find the worst offending point
imin = np.argmin(f / np.sqrt(v)) if minSN else np.argmin(f)
# integrate bins progressively outward until it no longer offends
i0, i1 = imin - 1, imin + 1
w0bin, w1bin, fbin, vbin = w0[imin], w1[imin], f[imin], v[imin]
side = 0
while True:
if minSN is None:
if fbin >= 0:
break
else:
if fbin / sqrt(vbin) >= minSN:
break
# check if we should stop integrating outward on either side
stop_at_0 = i0 < 0 or untouchable[i0]
stop_at_1 = i1 > len(f) - 1 or untouchable[i1]
# if can't integrate further outward, then set fbin to 0 if it is still negative and break
if stop_at_0 and stop_at_1:
if fbin < 0:
fbin, vbin = 0, 0
break
# else incorporate the next bin
if side == 0 and not stop_at_0:
fbin += f[i0]
vbin += v[i0]
i0 -= 1
elif not stop_at_1:
fbin += f[i1]
vbin += v[i1]
i1 += 1
# switch sides if possible
if stop_at_0:
side = 1
elif stop_at_1:
side = 0
else:
side = not side
# replace the appropriate section of the vectors
bad_block = slice(i0 + 1, i1)
arrays = w0, w1, f, v, untouchable
wrng = w1[i1 - 1] - w0[i0 + 1]
if wrng > res_limit:
w0in = np.arange(w0[i0 + 1], w1[i1 - 1], res_limit)
w1in = np.append(w0in[1:], w1[i1 - 1])
dw = w1in - w0in
fin = fbin * dw / wrng
vin = vbin * dw / wrng
inserts = w0in, w1in, fin, vin, [False] * len(fin)
else:
inserts = w0[i0 + 1], w1[i1 - 1], fbin, vbin, False
new_arrays = []
for a, value in zip(arrays, inserts):
a = np.delete(a, bad_block)
a = np.insert(a, i0 + 1, value)
new_arrays.append(a)
w0, w1, f, v, untouchable = new_arrays
# return a spectbl
dw = w1 - w0
f_dsty, e_dsty = f / dw, np.sqrt(v) / dw
bins = np.array([w0, w1]).T
newspec = rebin(spectbl, bins)
newspec['flux'] = f_dsty * newspec['flux'].unit
newspec['error'] = e_dsty * newspec['error'].unit
return newspec