import astropy
import numpy as np
from astropy.nddata import StdDevUncertainty
from astropy.wcs import WCS
from ndcube import NDCube
from punchbowl.data import NormalizedMetadata
from punchbowl.data.punch_io import encode_outliers
from punchbowl.prefect import punch_task
from punchbowl.util import average_datetime
[docs]
def _merge_ndcubes(cubes: list[NDCube | None], reference_cube_index: int | None = None) -> NDCube:
"""Create a merged data product from a set of input data, weighting by uncertainty."""
if cubes is None:
return None
if reference_cube_index is None:
reference_cube_index = len(cubes)//2 - 1
data_stack = np.stack([cube.data for cube in cubes], axis=0)
uncertainty_stack = np.array([cube.uncertainty.array for cube in cubes])
# Ignores negative / zero / infinite uncertainty, or places where the input data is exactly zero
uncertainty_was_invalid = ~np.isfinite(uncertainty_stack) + (uncertainty_stack == 0)
uncertainty_stack_was_bad = (uncertainty_stack <= 0) + ~np.isfinite(uncertainty_stack)
uncertainty_stack[uncertainty_stack_was_bad] = 1e64
uncertainty_stack[((data_stack == 0) + np.isnan(data_stack)) & uncertainty_was_invalid] = np.nan
weight_stack = 1/np.square(uncertainty_stack)
new_data = np.nansum(data_stack * weight_stack, axis=0) / np.nansum(weight_stack, axis=0)
uncertainty_stack[uncertainty_stack_was_bad] = np.nan
final_uncertainty = np.sqrt(np.nanmean(uncertainty_stack**2, axis=0))
# Restores uncertainty of zero or nan to infinite
final_uncertainty[final_uncertainty == 0] = np.inf
final_uncertainty[~np.isfinite(final_uncertainty)] = np.inf
new_data[np.isnan(new_data)] = 0
return NDCube(data=new_data, uncertainty=StdDevUncertainty(final_uncertainty), \
wcs=cubes[reference_cube_index].wcs)
[docs]
@punch_task
def merge_many_polarized_task(data: list[NDCube | None], trefoil_wcs: WCS, level: str = "2",
product_code: str = "PTM", maintain_nans: bool = False) -> NDCube:
"""Merge many task and carefully combine uncertainties."""
data_layers, uncertainty_layers = [], []
for polarization in [-60, 0, 60]:
polar_data = [d for d in data if d is not None and hasattr(d, "meta") and d.meta["POLAR"].value == polarization]
if len(polar_data) > 0:
data_merged = _merge_ndcubes(polar_data)
data_merged.meta = NormalizedMetadata.load_template(product_code, level=level)
if maintain_nans:
data_stack = np.stack([d.data for d in polar_data], axis=-1)
was_nan = np.all(np.isnan(data_stack), axis=-1)
data_merged.data[was_nan] = np.nan
else:
data_merged = NDCube(data = np.zeros((4096, 4096)),
uncertainty = StdDevUncertainty(np.full((4096, 4096), np.inf)),
wcs = trefoil_wcs,
meta = NormalizedMetadata.load_template(product_code, level=level))
data_layers.append(data_merged.data)
uncertainty_layers.append(data_merged.uncertainty.array)
trefoil_3d_wcs = astropy.wcs.utils.add_stokes_axis_to_wcs(trefoil_wcs, 2)
output_cube = NDCube(data=np.stack(data_layers, axis=0),
uncertainty=StdDevUncertainty(np.stack(uncertainty_layers, axis=0)),
wcs = trefoil_3d_wcs,
meta = NormalizedMetadata.load_template(product_code, level=level))
output_cube.meta["OUTLIER"] = encode_outliers([d for d in data if d is not None])
output_cube.meta["DATE-OBS"] = average_datetime([d.meta.datetime for d in data if d is not None],
).strftime("%Y-%m-%dT%H:%M:%S.%f")[:-3]
output_cube.meta["DATE-AVG"] = average_datetime([d.meta.datetime for d in data if d is not None],
).strftime("%Y-%m-%dT%H:%M:%S.%f")[:-3]
output_cube.meta["DATE-BEG"] = min([d.meta.datetime for d in data if d is not None],
).strftime("%Y-%m-%dT%H:%M:%S.%f")[:-3]
output_cube.meta["DATE-END"] = max([d.meta.datetime for d in data if d is not None],
).strftime("%Y-%m-%dT%H:%M:%S.%f")[:-3]
return output_cube
[docs]
@punch_task
def merge_many_clear_task(
data: list[NDCube | None], trefoil_wcs: WCS, level: str = "2", product_code: str = "CTM",
maintain_nans: bool = False) -> NDCube:
"""Merge many task and carefully combine uncertainties."""
if len(data) > 0:
data = [d for d in data if d is not None]
data_merged = _merge_ndcubes(data)
data_merged.meta = NormalizedMetadata.load_template(product_code, level=level)
if maintain_nans:
data_stack = np.stack([d.data for d in data], axis=-1)
was_nan = np.all(np.isnan(data_stack), axis=-1)
data_merged.data[was_nan] = np.nan
else:
data_merged = NDCube(data = np.zeros((4096, 4096)),
uncertainty = StdDevUncertainty(np.full((4096, 4096), np.inf)),
wcs = trefoil_wcs,
meta = NormalizedMetadata.load_template(product_code, level=level))
data_merged.meta["OUTLIER"] = encode_outliers([d for d in data if d is not None])
data_merged.meta["DATE-OBS"] = average_datetime([d.meta.datetime for d in data if d is not None],
).strftime("%Y-%m-%dT%H:%M:%S.%f")[:-3]
data_merged.meta["DATE-AVG"] = average_datetime([d.meta.datetime for d in data if d is not None],
).strftime("%Y-%m-%dT%H:%M:%S.%f")[:-3]
data_merged.meta["DATE-BEG"] = min([d.meta.datetime for d in data if d is not None],
).strftime("%Y-%m-%dT%H:%M:%S.%f")[:-3]
data_merged.meta["DATE-END"] = max([d.meta.datetime for d in data if d is not None],
).strftime("%Y-%m-%dT%H:%M:%S.%f")[:-3]
return data_merged
