Bump light-curve-feature to 0.15.0 (#701)

* Bump light-curve-feature to 0.15.0

* Add phase_features to Periodogram, Chi2Pvar and LaflerKinmanStringLength

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

---------

Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>

Author: hombit

CHANGELOG.md

@@ -9,11 +9,17 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Added
 
---
+- `Periodogram` now accepts a `phase_features` argument: features extracted from the light curve
+  phase-folded at the best period
+  ([#701](https://github.com/light-curve/light-curve-python/pull/701)).
+- New feature `Chi2Pvar`: probability of variability from the chi-squared test
+  ([#701](https://github.com/light-curve/light-curve-python/pull/701)).
+- New feature `LaflerKinmanStringLength`: smoothness of the light curve
+  ([#701](https://github.com/light-curve/light-curve-python/pull/701)).
 
 ### Changed
 
---
+- **Breaking** Underlying `nuts-rs` crate has updated, so specific non-linear optimization results may change https://github.com/light-curve/light-curve-python/pull/701
 
 ### Deprecated
 

light-curve/Cargo.lock

@@ -75,7 +75,7 @@ arrow-schema = "58"
 comfy-table = "7"
 
 [dependencies.light-curve-feature]
-version = "0.14.0"
+version = "0.15.0"
 default-features = false
 
 [dependencies.light-curve-dmdt]

light-curve/Cargo.toml

@@ -1611,6 +1611,8 @@ transform : None
     }
 }
 
+evaluator!(Chi2Pvar, lcf::Chi2Pvar, StockTransformer::Identity);
+
 evaluator!(Cusum, lcf::Cusum, StockTransformer::Identity);
 
 evaluator!(Eta, lcf::Eta, StockTransformer::Identity);
@@ -1673,6 +1675,12 @@ quantile : positive float
 
 evaluator!(Kurtosis, lcf::Kurtosis, StockTransformer::Arcsinh);
 
+evaluator!(
+    LaflerKinmanStringLength,
+    lcf::LaflerKinmanStringLength,
+    StockTransformer::Identity
+);
+
 evaluator!(LinearFit, lcf::LinearFit, StockTransformer::Identity);
 
 evaluator!(LinearTrend, lcf::LinearTrend, StockTransformer::Identity);
@@ -1932,6 +1940,7 @@ impl Periodogram {
         freqs: Option<Bound<PyAny>>,
         fast: Option<bool>,
         features: Option<Bound<PyAny>>,
+        phase_features: Option<Bound<PyAny>>,
         normalization: PeriodogramNormalization,
     ) -> PyResult<(LcfPeriodogram<f32>, LcfPeriodogram<f64>)> {
         let mut eval_f32 = match peaks {
@@ -2065,8 +2074,16 @@ impl Periodogram {
         if let Some(features) = features {
             for x in features.try_iter()? {
                 let py_feature = x?.cast::<PyFeatureEvaluator>()?.borrow();
-                eval_f32.add_feature(py_feature.feature_evaluator_f32.clone());
-                eval_f64.add_feature(py_feature.feature_evaluator_f64.clone());
+                eval_f32.add_spectrum_feature(py_feature.feature_evaluator_f32.clone());
+                eval_f64.add_spectrum_feature(py_feature.feature_evaluator_f64.clone());
+            }
+        }
+
+        if let Some(phase_features) = phase_features {
+            for x in phase_features.try_iter()? {
+                let py_feature = x?.cast::<PyFeatureEvaluator>()?.borrow();
+                eval_f32.add_phase_feature(py_feature.feature_evaluator_f32.clone());
+                eval_f64.add_phase_feature(py_feature.feature_evaluator_f64.clone());
             }
         }
 
@@ -2127,6 +2144,7 @@ impl Periodogram {
         freqs = None,
         fast = true,
         features = None,
+        phase_features = None,
         normalization = "psd",
         transform = None,
     ))]
@@ -2138,6 +2156,7 @@ impl Periodogram {
         freqs: Option<Bound<PyAny>>,
         fast: Option<bool>,
         features: Option<Bound<PyAny>>,
+        phase_features: Option<Bound<PyAny>>,
         normalization: &str,
         transform: Option<Bound<PyAny>>,
     ) -> PyResult<(Self, PyFeatureEvaluator)> {
@@ -2155,6 +2174,7 @@ impl Periodogram {
             freqs,
             fast,
             features,
+            phase_features,
             normalization,
         )?;
         Ok((
@@ -2238,9 +2258,14 @@ fast : bool or None, optional
     Use "Fast" (approximate and FFT-based) or direct periodogram algorithm,
     default is {default_fast}
 features : iterable or None, optional
-    Features to extract from periodogram considering it as a time-series,
-    default is None which means no additional features
-    Features to extract from periodogram considering it as a time-series
+    Features extracted from the periodogram power spectrum, treating it as a
+    time-series (frequency as time, power as magnitude).
+    Default is None which means no additional spectrum features.
+phase_features : iterable or None, optional
+    Features to extract from the light curve phase-folded at the best period.
+    Phase runs from 0 to 1 with phase 0 at the magnitude minimum.
+    Feature names are prefixed with `period_folded_`.
+    Default is None which means no phase features.
 normalization : str, optional
     Normalization of the periodogram power. Affects `power()`,
     `freq_power()`, and feature extraction via `__call__()`.

light-curve/src/features.rs

@@ -58,6 +58,7 @@ fn light_curve(py: Python, m: Bound<PyModule>) -> PyResult<()> {
     m.add_class::<f::BazinFit>()?;
     m.add_class::<f::BeyondNStd>()?;
     m.add_class::<f::Bins>()?;
+    m.add_class::<f::Chi2Pvar>()?;
     m.add_class::<f::Cusum>()?;
     m.add_class::<f::Duration>()?;
     m.add_class::<f::Eta>()?;
@@ -66,6 +67,7 @@ fn light_curve(py: Python, m: Bound<PyModule>) -> PyResult<()> {
     m.add_class::<f::JsonDeserializedFeature>()?;
     m.add_class::<f::InterPercentileRange>()?;
     m.add_class::<f::Kurtosis>()?;
+    m.add_class::<f::LaflerKinmanStringLength>()?;
     m.add_class::<f::LinearFit>()?;
     m.add_class::<f::LinearTrend>()?;
     m.add_class::<f::LinexpFit>()?;

light-curve/src/lib.rs

@@ -3,7 +3,7 @@
 from numpy.testing import assert_allclose
 from scipy.signal import lombscargle
 
-from light_curve.light_curve_ext import Periodogram
+from light_curve.light_curve_ext import Duration, Periodogram
 
 
 def test_vs_lombscargle():
@@ -121,6 +121,21 @@ def test_normalization_default_is_psd():
     assert_allclose(power_default, power_psd)
 
 
+def test_phase_features_duration():
+    """Duration of phase-folded dense light curve must be close to 1."""
+    period = 2.0
+    # Dense, regular sampling over many periods so the phase folds uniformly
+    t = np.linspace(0, 100 * period, 10_000)
+    m = np.sin(2 * np.pi * t / period)
+
+    fe = Periodogram(peaks=1, fast=True, phase_features=[Duration()])
+    result = fe(t, m, sorted=True)
+
+    # result layout: period_0, snr_0, period_folded_duration
+    period_folded_duration = result[-1]
+    assert_allclose(period_folded_duration, 1.0, atol=1e-3)
+
+
 def test_normalization_invalid():
     """Invalid normalization string should raise ValueError."""
     with pytest.raises(ValueError, match="normalization must be one of"):