Add workaround for nans in times during injection

src/qusi/internal/light_curve_dataset.py

@@ -197,6 +197,7 @@ def inject_light_curve(
 ) -> LightCurveObservation:
     (
         fluxes_with_injected_signal,
+        injected_light_curve_times,
         _,
         _,
     ) = inject_signal_into_light_curve_with_intermediates(
@@ -208,9 +209,10 @@ def inject_light_curve(
         baseline_flux_estimation_method=BaselineFluxEstimationMethod.MEDIAN,
     )
     injected_light_curve = LightCurve.new(
-        times=injectee_observation.light_curve.times,
+        times=injected_light_curve_times,
         fluxes=fluxes_with_injected_signal,
     )
+    # TODO: Quickly hacked in times with nans removed. Should be handled elsewhere.
     injected_observation = LightCurveObservation.new(
         light_curve=injected_light_curve, label=injectable_observation.label
     )
@@ -383,11 +385,19 @@ def inject_signal_into_light_curve_with_intermediates(
                                             for scaling the signal magnifications.
     :return: The fluxes with the injected signal, the offset signal times, and the signal flux.
     """
-    minimum_light_curve_time = np.min(light_curve_times)
+    light_curve_nan_flux_indexes = np.isnan(light_curve_times) | np.isnan(light_curve_fluxes)
+    light_curve_times = light_curve_times[~light_curve_nan_flux_indexes]
+    light_curve_fluxes = light_curve_fluxes[~light_curve_nan_flux_indexes]
+    signal_nan_flux_indexes = np.isnan(signal_times) | np.isnan(signal_magnifications)
+    signal_times = signal_times[~signal_nan_flux_indexes]
+    signal_magnifications = signal_magnifications[~signal_nan_flux_indexes]
+    # TODO: Remove quick hack of removing nans and add a more proper handling.
+
+    minimum_light_curve_time = np.nanmin(light_curve_times)
     relative_light_curve_times = light_curve_times - minimum_light_curve_time
-    relative_signal_times = signal_times - np.min(signal_times)
-    signal_time_length = np.max(relative_signal_times)
-    light_curve_time_length = np.max(relative_light_curve_times)
+    relative_signal_times = signal_times - np.nanmin(signal_times)
+    signal_time_length = np.nanmax(relative_signal_times)
+    light_curve_time_length = np.nanmax(relative_light_curve_times)
     time_length_difference = light_curve_time_length - signal_time_length
     signal_start_offset = (
                                   np.random.random() * time_length_difference
@@ -452,7 +462,7 @@ def inject_signal_into_light_curve_with_intermediates(
         )
     interpolated_signal_fluxes = signal_flux_interpolator(light_curve_times)
     fluxes_with_injected_signal = light_curve_fluxes + interpolated_signal_fluxes
-    return fluxes_with_injected_signal, offset_signal_times, signal_fluxes
+    return fluxes_with_injected_signal, light_curve_times, offset_signal_times, signal_fluxes
 
 
 def move_path_to_nvme(path: Path) -> Path: