[Feature] Modify signal_interpolate for input with NaNs and extrapolation

NEWS.rst

@@ -2,6 +2,32 @@ News
 =====
 
 
+
+
+
+0.2.1
+-------------------
+New Features
++++++++++++++
+
+* Allow for input with NaNs and extrapolation in `signal_interpolate()`
+* Add argument `method` in `find_outliers()`
+
+
+
+
+
+0.2.0
+-------------------
+New Features
++++++++++++++
+
+* Add new time-domain measures in `hrv_time()`: `Prc20NN`, `Prc80NN`, `MinNN`, and `MaxNN`
+
+
+
+
+
 0.1.6
 -------------------
 

neurokit2/signal/signal_interpolate.py

@@ -3,11 +3,11 @@
 import scipy.interpolate
 
 
-def signal_interpolate(x_values, y_values, x_new=None, method="quadratic"):
+def signal_interpolate(
+    x_values=None, y_values=None, x_new=None, method="quadratic", fill_value=None
+):
     """**Interpolate a signal**
-
     Interpolate a signal using different methods.
-
     Parameters
     ----------
     x_values : Union[list, np.array, pd.Series]
@@ -29,27 +29,28 @@ def signal_interpolate(x_values, y_values, x_new=None, method="quadratic"):
         order of the spline interpolator to use.
         See `here <https://docs.scipy.org/doc/scipy/reference/generated/scipy.interpolate.
         PchipInterpolator.html>`_ for details on the ``"monotone_cubic"`` method.
-
+    fill_value : float or tuple or str
+        If a ndarray (or float), this value will be used to fill in for
+        requested points outside of the data range.
+        If a two-element tuple, then the first element is used as a fill value
+        for x_new < x[0] and the second element is used for x_new > x[-1].
+        If “extrapolate”, then points outside the data range will be extrapolated.
+        If not provided, then the default is ([y_values[0]], [y_values[-1]]).
     Returns
     -------
     array
         Vector of interpolated samples.
-
     Examples
     --------
     .. ipython:: python
-
       import numpy as np
       import neurokit2 as nk
       import matplotlib.pyplot as plt
-
       # Generate Simulated Signal
       signal = nk.signal_simulate(duration=2, sampling_rate=10)
-
       # We want to interpolate to 2000 samples
       x_values = np.linspace(0, 2000, num=len(signal), endpoint=False)
       x_new = np.linspace(0, 2000, num=2000, endpoint=False)
-
       # Visualize all interpolation methods
       @savefig p_signal_interpolate1.png scale=100%
       nk.signal_plot([
@@ -65,9 +66,22 @@ def signal_interpolate(x_values, y_values, x_new=None, method="quadratic"):
       plt.scatter(x_values, signal, label="original datapoints", zorder=3)
       @suppress
       plt.close()
-
     """
     # Sanity checks
+    if x_values is None and y_values is None:
+        raise ValueError(
+            "NeuroKit error: signal_interpolate(): x_values or y_values must be provided."
+        )
+    elif x_values is None or y_values is None:
+        # for interpolating NaNs
+        if y_values is None:
+            y_values = x_values
+        x_values = np.arange(0, len(y_values))
+        if x_new is None:
+            x_new = x_values
+        y_finite = np.where(np.invert(np.isnan(y_values)))[0]
+        x_values = x_values[y_finite]
+        y_values = y_values[y_finite]
     if len(x_values) != len(y_values):
         raise ValueError("x_values and y_values must be of the same length.")
 
@@ -77,23 +91,22 @@ def signal_interpolate(x_values, y_values, x_new=None, method="quadratic"):
     else:
         if len(x_values) == len(x_new):
             return y_values
-
     if method == "monotone_cubic":
         interpolation_function = scipy.interpolate.PchipInterpolator(
             x_values, y_values, extrapolate=True
         )
     else:
+        if fill_value is None:
+            fill_value = ([y_values[0]], [y_values[-1]])
         interpolation_function = scipy.interpolate.interp1d(
             x_values,
             y_values,
             kind=method,
             bounds_error=False,
-            fill_value=([y_values[0]], [y_values[-1]]),
+            fill_value=fill_value,
         )
-
     if isinstance(x_new, int):
         x_new = np.linspace(x_values[0], x_values[-1], x_new)
-
     interpolated = interpolation_function(x_new)
 
     if method == "monotone_cubic":
@@ -102,5 +115,4 @@ def signal_interpolate(x_values, y_values, x_new=None, method="quadratic"):
         # scipy.interpolate.interp1d with fill_value=([y_values[0]], [y_values[-1]].
         interpolated[: int(x_values[0])] = interpolated[int(x_values[0])]
         interpolated[int(x_values[-1]) :] = interpolated[int(x_values[-1])]
-
     return interpolated