def main(inf, outf, config): with open(inf, "r", newline="") as f: cadns, times, fluxs = utils.csv_column_read(f, FIELDS, casts=CASTS, start=config.start, end=config.end) # Do the thing. times, fluxs = highpass(times, fluxs, config) utils.csv_column_write(outf, [cadns, times, fluxs], FIELDS)
def main(inf, outf, config): with open(inf, "r", newline="") as inf: cadns, times, fluxs = utils.csv_column_read(inf, FIELDS, start=config.start, end=config.end, casts=CASTS) # Decorrelate flux. fluxs, _ = decorrelate_data(config.tp, times, fluxs) utils.csv_column_write(outf, [cadns, times, fluxs], FIELDS)
def plot_fft(fig, config):
CASTS = [float, float]
FIELDS = ["frequency", config.pgtype]
# Iterate over all inputs.
nfiles = len(config.ifiles)
for i, ifile in enumerate(config.ifiles):
# Get the periodogram information.
with open(ifile, "r", newline="") as f:
fx, fy = utils.csv_column_read(f, FIELDS, casts=CASTS)
# Create a latex axis.
ax = utils.latexify(fig.add_subplot("%d%d%d" % (nfiles, 1, i + 1)))
# Plot the periodogram.
ax.plot(fx, fy, color="k", linestyle="-", marker="None")
ax.set_axisbelow(True)
ax.set_xlabel("Frequency ($\mu$Hz)")
# Make sure the axis names are correct.
if config.pgtype == "amplitude":
ax.set_ylabel("Amplitude (ppm)")
elif config.pgtype == "psd":
ax.set_ylabel("PSD (ppm$^2$ $\mu$Hz$^{-1}$)")
if config.maxx > config.minx >= 0:
ax.set_xlim([config.minx, config.maxx])
if config.maxy > config.miny >= 0:
ax.set_ylim([config.miny, config.maxy])
plt.legend()
fig.tight_layout()
def main(inf, outf, config): with open(inf, "r", newline="") as inf: cadns, times, fluxs = utils.csv_column_read(inf, FIELDS, start=config.start, end=config.end, casts=CASTS) # To PPM. fluxs = (fluxs / fluxs.mean() - 1) * 1e6 utils.csv_column_write(outf, [cadns, times, fluxs], FIELDS)
def main(inf, outf, config): with open(inf, "r", newline="") as f: times, fluxs = utils.csv_column_read(f, FIELDS, casts=CASTS, start=config.start, end=config.end) fx, fy = utils.lombscargle_amplitude(times, fluxs, mult=config.mult, upper=config.upper) # Output the FFT. utils.csv_column_write(outf, [fx, fy], ["frequency", "amplitude"])
def main(inf, outf, config): with open(inf, "r", newline="") as inf: cadns, times, fluxs = utils.csv_column_read(inf, FIELDS, start=config.start, end=config.end, casts=CASTS) # Remove outliers. filt = reject_outliers(times, fluxs, config) cadns = cadns[filt] times = times[filt] fluxs = fluxs[filt] utils.csv_column_write(outf, [cadns, times, fluxs], FIELDS)
def main(inf, outf, config): FIELDS = ["frequency", "amplitude"] CASTS = [float, float] with open(inf, "r", newline="") as f: fx, fy = utils.csv_column_read(f, FIELDS, casts=CASTS) # Do the thing. fx, fy = utils.raw_to_psd(fx, fy, config.variance) # Output the PSD. FIELDS[1] = "psd" utils.csv_column_write(outf, [fx, fy], FIELDS)
def plot_echelle(config, fig, ifile):
CASTS = [float, float]
FIELDS = ["frequency", config.pgtype]
if config.deltanu is None:
raise NotImplementedError("∆𝜈 autocorrelation not implemented")
# Get the cadence information.
with open(ifile, "r", newline="") as f:
fx, fy = utils.csv_column_read(f, FIELDS, casts=CASTS)
# Filter start and end.
filt = numpy.ones_like(fx, dtype=bool)
if config.start is not None:
filt &= fx >= config.start * config.deltanu
if config.end is not None:
filt &= fx <= (config.end + 1) * config.deltanu
fx = fx[filt]
fy = fy[filt]
# Convert from power to amplitude.
if config.pgtype == "psd":
pass
#fy *= numpy.diff(fx).mean()
#fy **= 0.5
#pass
# Convert to wrapped echelle.
fx += config.off
fx %= config.deltanu
# Sort the arrays.
sfilt = numpy.argsort(fx)
fx = fx[sfilt]
fy = fy[sfilt]
# Smooth using a Savgol filter.
#itp = scipy.interpolate.interp1d(fx, fy, kind='linear')
#fy = scipy.signal.savgol_filter(itp(fx), config.width, 6)
win = scipy.signal.boxcar(config.width)
fy = scipy.signal.convolve(fy, win, mode="same")
# Plot.
ax = utils.latexify(fig.add_subplot(111))
ax.plot(fx, fy, color="r")
ax.set_ylabel("PSD (ppm$^2$$\mu$Hz$^{-1}$)")
ax.set_xlabel("Frequency mod %s ($\mu$Hz)" % (config.deltanu,))
plt.legend()
fig.tight_layout()
def main(files, outf, config):
TIME = []
FLUX = []
# We split the filename into <fname>[:<start>:<end>].
for fname in files:
start = 0
end = -1
# Allow slices in fnames.
if ":" in fname:
fname, start, end = fname.split(":")
# Float.
start = float(start)
end = float(end)
# Open file.
with open(fname, "r", newline="") as f:
times, fluxs = utils.csv_column_read(f, FIELDS, casts=CASTS)
# Times.
offset = times - times.min()
# Allow negative slices.
if end < 0:
end = offset.max() + (end + 1)
# Filter.
filt = (start <= offset) & (offset <= end)
times = times[filt]
fluxs = fluxs[filt]
# Append.
TIME = numpy.append(TIME, times)
FLUX = numpy.append(FLUX, fluxs)
# Fake the cadence numbers.
CADN = numpy.arange(TIME.shape[0]) + 1
# We need to sort by the time. This is a bit of a pain, because the operation
# of "sorting rows" isn't a thing in numpy. We need to switch out to Python
# lists.
tosort = numpy.array([TIME, FLUX]).T
tosort = sorted(list(tosort), key=lambda arr: arr[0])
TIME, FLUX = numpy.array(tosort).T
utils.csv_column_write(outf, [CADN, TIME, FLUX], ["cadence"] + FIELDS)
def main(inf, outf, config): with open(inf, "r", newline="") as f: cadns, times, fluxs = utils.csv_column_read(f, FIELDS, casts=CASTS, start=config.start, end=config.end) diffs = numpy.diff(times).copy() assert(numpy.all(diffs >= 0)) for idx, gap in zip(*numpy.where(diffs > config.width), diffs[diffs > config.width]): # We only remove the gap to the scale of the median. gap -= gap % numpy.median(diffs) # Remove the gap. This isn't a concurrent modification because the diffs aren't views. times[idx+1:] -= gap # Output the FFT. utils.csv_column_write(outf, [cadns, times, fluxs], FIELDS)
def main(files, outf, config): # Cadences. cadns = [] times = [] fluxs = [] for fname in files: with open(fname, "r", newline="") as f: # Get CSV. _cadn, _time, _flux = utils.csv_column_read(f, FIELDS, casts=CASTS, start=config.start, end=config.end) # Convert to numpy arrays. cadns.append(_cadn) times.append(_time) fluxs.append(_flux) # Get unique set. inter = cadns[0] for cadn in cadns[1:]: inter = np.intersect1d(inter, cadn) # Filter out each array. for i, cadn in enumerate(cadns): filt = np.in1d(cadn, inter) cadns[i] = cadns[i][filt] times[i] = times[i][filt] fluxs[i] = fluxs[i][filt] # Convert filtered version to arrays (now everything is the same shape). cadns = np.median(cadns, axis=0) times = np.median(times, axis=0) fluxs = np.array(fluxs, dtype=float) # Combine the flux values. fluxs = config.reduce(fluxs) utils.csv_column_write(outf, [cadns, times, fluxs], FIELDS)
def plot_echelle(config, fig, ifile):
CASTS = [float, float]
FIELDS = ["frequency", config.pgtype]
if config.deltanu is None:
raise NotImplementedError("∆𝜈 autocorrelation not implemented")
# Get the cadence information.
with open(ifile, "r", newline="") as f:
fx, fy = utils.csv_column_read(f, FIELDS, casts=CASTS)
# Filter start and end.
filt = numpy.ones_like(fx, dtype=bool)
if config.start is not None:
filt &= fx >= config.start * config.deltanu
if config.end is not None:
filt &= fx <= (config.end + 1) * config.deltanu
fx = fx[filt]
fy = fy[filt]
# Convert from power to amplitude.
if config.pgtype == "psd":
pass
#fy *= numpy.diff(fx).mean()
#fy **= 0.5
# First we bin the transform.
width = config.smoothwidth
bins = int(fx.ptp() // width)
newfy = numpy.zeros(bins)
for i in range(bins):
rnge = (i*width <= fx) & (fx < (i+1)*width)
newfy[i] = fy[rnge].sum()
fy = newfy
fx = numpy.linspace(fx.min(), fx.max(), bins)
# Then we compute the grid sizes.
delta = numpy.diff(fx).mean()
rows = int(fx.ptp() // config.deltanu)
cols = int(config.deltanu // delta)
# Fill in the grid from the binned data.
grid = numpy.zeros([rows, cols])
for y, _ in enumerate(grid):
grid[y,...] = fy[y*cols:(y+1)*cols]
ax = utils.latexify(fig.add_subplot(111))
#ax.imshow(-grid, cmap="gray", interpolation="nearest", origin="bottom")
#ax.imshow(grid, cmap="viridis", interpolation="nearest", origin="bottom")
ax.pcolormesh(grid, cmap="viridis")
#ax.pcolormesh(-grid, cmap="gray")
#ax.xaxis.set_ticks(numpy.arange)
ax.xaxis.set_ticks
ax.set_xlim([0, config.deltanu])
ax.xaxis.set_ticks(numpy.arange(*ax.get_xlim(), step=20))
ax.xaxis.set_ticks(numpy.arange(*ax.get_xlim(), step=5), minor=True)
ax.set_ylim([config.start, config.end-config.start])
ax.yaxis.set_ticks(numpy.arange(*ax.get_ylim(), step=1)+1)
#ax.yaxis.set_ticks(numpy.arange(*ax.get_ylim(), step=0.25), minor=True)
#ax.set_ylabel("Frequency ($\mu$Hz)")
ax.set_ylabel("Mode Order")
ax.set_xlabel("Frequency mod %s ($\mu$Hz)" % (config.deltanu,))
plt.legend()
fig.tight_layout()
def main(inf, outf, config): with open(inf, "r", newline="") as f: times, fluxs = utils.csv_column_read(f, FIELDS, casts=CASTS, start=config.start, end=config.end) # Output the variance. print(fluxs.var(), file=outf)
def plot_lc(config, fig, ifile):
# Get the cadence information.
with open(ifile, "r", newline="") as f:
times, fluxs = utils.csv_column_read(f, FIELDS, casts=CASTS, start=config.start, end=config.end)
# Times and fluxes.
xs = times
ys = fluxs
# Convert flux to ppm.
#ys = ys / ys.mean() - 1
#ys *= 1e6
# Figure out time-related offsets.
offset = np.min(times)
xs -= offset
if config.timestamp is not None:
config.timestamp -= offset
if config.fft:
fx, fy = utils.lombscargle_amplitude(xs, ys, upper=config.high_freq)
if config.fftout:
with open(config.fftout, "w", newline="") as f:
utils.csv_column_write(f, [fx, fy], ["frequency", "amplitude"])
fx, fy = utils.raw_to_psd(fx, fy, fluxs.var())
if config.lc:
if config.period is not None:
# TODO: We should allow for showing more than one phase.
xs = (xs % config.period) / config.period
xs = (xs + config.phase) % 1.0
# Bin the folded phase plot by taking the average of ranges.
if config.bins is not None:
size = 1.0 / config.bins
nys = np.zeros(config.bins)
for i in range(config.bins):
rnge = (i*size <= xs) & (xs < (i+1)*size)
nys[i] = np.median(ys[rnge])
ys = nys
xs = np.arange(config.bins) * size
# Replication.
xs = np.tile(xs, config.width) + np.repeat(np.arange(config.width), xs.shape[0])
ys = np.tile(ys, config.width)
if config.fft and config.lc:
ax1 = utils.latexify(fig.add_subplot(211))
else:
ax1 = utils.latexify(fig.add_subplot(111))
if config.lc:
if not (config.period or config.bins):
ax1.plot(xs, ys, color="0.5", linestyle="-", marker="None")
#ax1.plot(xs, ys, color="k", linestyle="None", marker="+", label=r"Kepler/K2 Halo Photometry")
ax1.set_xlabel("Time ($d$)")
else:
# TODO: We should overlay a binned version.
if config.timestamp is not None:
predicted = (config.timestamp % config.period) / config.period
predicted = (predicted + config.phase) % 1.0
ax1.xaxis.set_ticks(predicted + np.arange(config.width), minor=True)
ax1.xaxis.grid(True, which="minor", color="r", linestyle="--", linewidth=2)
ax1.plot(xs, ys, color="0.5", linestyle="None", marker="o", label=r"Kepler/K2 Halo Photometry")
ax1.set_xlabel("Phase")
ax1.set_ylabel(r"Intensity (ppm)")
if config.lc and config.title:
ax1.set_title(r"Light Curve [%s] # %s" % (description(config), config.comment or ""))
if config.fft:
if config.lc:
ax2 = utils.latexify(fig.add_subplot(212))
else:
ax2 = ax1
ax2.plot(fx, fy, color="k", linestyle="-", marker="None")
#ax2.xaxis.set_ticks(np.arange(*ax2.get_xlim(), step=1))
#ax2.xaxis.set_ticks(np.arange(*ax2.get_xlim(), step=0.25), minor=True)
#ax2.xaxis.grid(True, which="major", color="k", linestyle="--")
#ax2.xaxis.grid(True, which="minor", color="k", linestyle=":")
ax2.set_axisbelow(True)
#ax2.set_xlabel("Frequency ($d^{-1}$)")
ax2.set_xlabel("Frequency ($\mu$Hz)")
#ax2.set_ylabel("Amplitude (ppm)")
ax2.set_ylabel("PDF (ppm$^2$ $\mu$Hz$^{-1}$)")
if config.maxx > config.minx:
ax2.set_xlim([config.minx, config.maxx])
if config.maxy > config.miny:
ax2.set_ylim([config.miny, config.maxy])
plt.legend()
fig.tight_layout()
