Batch Workflows with TestGroup

TestGroup holds labelled collections of objects (GOOD / BAD / ANY) so you can clean, analyse, and visualise many stars with one call instead of a manual loop.

from varistar import TimeSeries, LightCurve, TestGroup
from _synthetic import make_sinusoidal_lightcurve

lightcurves = []
for i, seed in enumerate([1, 2, 3, 4, 5, 6]):
    df = make_sinusoidal_lightcurve(period=2.0 + 0.3 * i, seed=seed)
    ts = TimeSeries(magnitude="mag I", time_scale="HJD")
    ts.load_data_from_df(df, data_id=f"star_{i:03d}")
    ts.clean_by_error(max_error=0.05)
    lightcurves.append(LightCurve(ts))

group = TestGroup(good_ts=lightcurves[:4], bad_ts=lightcurves[4:], name="synthetic_sample")
group.summary()

[clean_by_error] Removed 0 points (err > 0.05).
[clean_by_error] Removed 0 points (err > 0.05).
[clean_by_error] Removed 0 points (err > 0.05).
[clean_by_error] Removed 0 points (err > 0.05).
[clean_by_error] Removed 0 points (err > 0.05).
[clean_by_error] Removed 0 points (err > 0.05).
TestGroup 'synthetic_sample'  |  GOOD: 4  BAD: 2  ANY/LC: 0  (total: 6)

Bulk period-finding

apply() calls an unbound method on every registered object. find_best_period also fills in is_harmonic, which we’ll use below:

group.apply(LightCurve.find_best_period)

Mosaic view

group.plot_mosaic(LightCurve.plot_phased, n_cols=3, fit_model="fourier")

Filtering and export

short_period = group.filter_by("periods", lambda p: p[0] < 3.0, target="any")
short_period.summary()

TestGroup 'synthetic_sample[filtered]'  |  GOOD: 0  BAD: 0  ANY/LC: 4  (total: 4)

group.export_periods()

shape: (6, 4)

timeseries_id	best_period	n_candidates	status
str	f64	i64	str
"unknown"	1.998568	20	"GOOD"
"unknown"	2.299374	20	"GOOD"
"unknown"	2.598417	20	"GOOD"
"unknown"	2.899632	20	"GOOD"
"unknown"	3.19618	20	"BAD"
"unknown"	3.504806	20	"BAD"

Next: ML Feature Pipeline.