# Licensed under a 3-clause BSD style license - see LICENSE.rst
import abc
import numpy as np
from astropy import units as u
from astropy.io import fits
from astropy.table import Table
from gammapy.data import GTI
from gammapy.irf import EDispKernel, EDispKernelMap
from gammapy.maps import RegionNDMap
from gammapy.utils.scripts import make_path
from .spectrum import SpectrumDatasetOnOff
__all__ = [
"DatasetReader",
"DatasetWriter",
"OGIPDatasetReader",
"OGIPDatasetWriter",
]
class DatasetReader(abc.ABC):
"""Dataset reader base class."""
@property
@abc.abstractmethod
def tag(self):
pass
@abc.abstractmethod
def read(self):
pass
class DatasetWriter(abc.ABC):
"""Dataset writer base class."""
@property
@abc.abstractmethod
def tag(self):
pass
@abc.abstractmethod
def write(self, dataset):
pass
[docs]class OGIPDatasetWriter(DatasetWriter):
"""Write OGIP files.
If you want to use the written files with Sherpa, you have to use the
``ogip-sherpa`` format. Then all files will be written in units of 'keV' and
'cm2'.
The naming scheme is fixed as following:
* PHA file is named filename.fits
* BKG file is named filename_bkg.fits
* ARF file is named filename_arf.fits
* RMF file is named filename_rmf.fits
Parameters
----------
filename : `pathlib.Path` or str
Filename
format : {"ogip", "ogip-sherpa"}
Which format to use
overwrite : bool
Overwrite existing files?
"""
tag = ["ogip", "ogip-sherpa"]
def __init__(self, filename, format="ogip", overwrite=False):
filename = make_path(filename)
filename.parent.mkdir(exist_ok=True, parents=True)
self.filename = filename
self.format = format
self.overwrite = overwrite
[docs] @staticmethod
def get_filenames(filename):
"""Get filenames.
Parameters
----------
filename : `~pathlib.Path`
Filename
Returns
-------
filenames : dict
Dict of filenames
"""
suffix = "".join(filename.suffixes)
name = filename.name.replace(suffix, "")
name = f"{name}{{}}{suffix}"
return {
"respfile": name.format("_rmf"),
"backfile": name.format("_bkg"),
"ancrfile": name.format("_arf"),
}
[docs] def write(self, dataset):
"""Write dataset to file.
Parameters
----------
dataset : `SpectrumDatasetOnOff`
Dataset to write
"""
filenames = self.get_filenames(self.filename)
self.write_pha(dataset, filename=self.filename)
path = self.filename.parent
self.write_arf(dataset, filename=path / filenames["ancrfile"])
if dataset.counts_off:
self.write_bkg(dataset, filename=path / filenames["backfile"])
if dataset.edisp:
self.write_rmf(dataset, filename=path / filenames["respfile"])
[docs] def write_rmf(self, dataset, filename):
"""Write energy dispersion.
Parameters
----------
dataset : `SpectrumDatasetOnOff`
Dataset to write
filename : str or `Path`
Filename to use
"""
kernel = dataset.edisp.get_edisp_kernel()
kernel.write(filename=filename, overwrite=self.overwrite, format=self.format)
[docs] def write_arf(self, dataset, filename):
"""Write effective area.
Parameters
----------
dataset : `SpectrumDatasetOnOff`
Dataset to write
filename : str or `Path`
Filename to use
"""
aeff = dataset.exposure / dataset.exposure.meta["livetime"]
aeff.write(
filename=filename,
overwrite=self.overwrite,
format=self.format.replace("ogip", "ogip-arf"),
)
[docs] def to_counts_hdulist(self, dataset, is_bkg=False):
"""Convert counts region map to hdulist.
Parameters
----------
dataset : `SpectrumDatasetOnOff`
Dataset to write
is_bkg : bool
Whether to use counts off
"""
counts = dataset.counts_off if is_bkg else dataset.counts
acceptance = dataset.acceptance_off if is_bkg else dataset.acceptance
hdulist = counts.to_hdulist(format=self.format)
table = Table.read(hdulist["SPECTRUM"])
meta = self.get_ogip_meta(dataset, is_bkg=is_bkg)
if dataset.mask_safe is not None:
mask_array = dataset.mask_safe.data[:, 0, 0]
else:
mask_array = np.ones(acceptance.data.size)
table["QUALITY"] = np.logical_not(mask_array)
del table.meta["QUALITY"]
table["BACKSCAL"] = acceptance.data[:, 0, 0]
del table.meta["BACKSCAL"]
# adapt meta data
table.meta.update(meta)
hdulist["SPECTRUM"] = fits.BinTableHDU(table)
return hdulist
[docs] def write_pha(self, dataset, filename):
"""Write counts file.
Parameters
----------
dataset : `SpectrumDatasetOnOff`
Dataset to write
filename : str or `Path`
Filename to use
"""
hdulist = self.to_counts_hdulist(dataset)
if dataset.gti:
hdu = fits.BinTableHDU(dataset.gti.table, name="GTI")
hdulist.append(hdu)
hdulist.writeto(filename, overwrite=self.overwrite)
[docs] def write_bkg(self, dataset, filename):
"""Write off counts file.
Parameters
----------
dataset : `SpectrumDatasetOnOff`
Dataset to write
filename : str or `Path`
Filename to use
"""
hdulist = self.to_counts_hdulist(dataset, is_bkg=True)
hdulist.writeto(filename, overwrite=self.overwrite)
[docs]class OGIPDatasetReader(DatasetReader):
"""Read `~gammapy.datasets.SpectrumDatasetOnOff` from OGIP files.
BKG file, ARF, and RMF must be set in the PHA header and be present in
the same folder.
The naming scheme is fixed to the following scheme:
* PHA file is named ``pha_obs{name}.fits``
* BKG file is named ``bkg_obs{name}.fits``
* ARF file is named ``arf_obs{name}.fits``
* RMF file is named ``rmf_obs{name}.fits``
with ``{name}`` the dataset name.
Parameters
----------
filename : str or `~pathlib.Path`
OGIP PHA file to read
"""
tag = "ogip"
def __init__(self, filename):
self.filename = make_path(filename)
[docs] def get_valid_path(self, filename):
"""Get absolute or relative path.
The relative path is with respect to the name of the reference file.
Parameters
----------
filename : str or `Path`
Filename
Returns
-------
filename : `Path`
Valid path
"""
filename = make_path(filename)
if not filename.exists():
return self.filename.parent / filename
else:
return filename
[docs] def get_filenames(self, pha_meta):
"""Get filenames.
Parameters
----------
pha_meta : dict
Meta data from the PHA file
Returns
-------
filenames : dict
Dict with filenames of "arffile", "rmffile" (optional)
and "bkgfile" (optional)
"""
filenames = {"arffile": self.get_valid_path(pha_meta["ANCRFILE"])}
if "BACKFILE" in pha_meta:
filenames["bkgfile"] = self.get_valid_path(pha_meta["BACKFILE"])
if "RESPFILE" in pha_meta:
filenames["rmffile"] = self.get_valid_path(pha_meta["RESPFILE"])
return filenames
[docs] @staticmethod
def read_pha(filename):
"""Read PHA file.
Parameters
----------
filename : str or `Path`
PHA file name
Returns
-------
data : dict
Dict with counts, acceptance and mask_safe
"""
data = {}
with fits.open(filename, memmap=False) as hdulist:
data["counts"] = RegionNDMap.from_hdulist(hdulist, format="ogip")
data["acceptance"] = RegionNDMap.from_hdulist(
hdulist, format="ogip", ogip_column="BACKSCAL"
)
if "GTI" in hdulist:
data["gti"] = GTI(Table.read(hdulist["GTI"]))
data["mask_safe"] = RegionNDMap.from_hdulist(
hdulist, format="ogip", ogip_column="QUALITY"
)
return data
[docs] @staticmethod
def read_bkg(filename):
"""Read PHA background file.
Parameters
----------
filename : str or `Path`
PHA file name
Returns
-------
data : dict
Dict with counts_off and acceptance_off
"""
with fits.open(filename, memmap=False) as hdulist:
counts_off = RegionNDMap.from_hdulist(hdulist, format="ogip")
acceptance_off = RegionNDMap.from_hdulist(
hdulist, ogip_column="BACKSCAL", format="ogip"
)
return {"counts_off": counts_off, "acceptance_off": acceptance_off}
[docs] @staticmethod
def read_rmf(filename, exposure):
"""Read RMF file.
Parameters
----------
filename : str or `Path`
PHA file name
exposure : `RegionNDMap`
Exposure map
Returns
-------
data : `EDispKernelMap`
Dict with edisp
"""
kernel = EDispKernel.read(filename)
edisp = EDispKernelMap.from_edisp_kernel(kernel, geom=exposure.geom)
# TODO: resolve this separate handling of exposure for edisp
edisp.exposure_map.data = exposure.data[:, :, np.newaxis, :]
return edisp
[docs] @staticmethod
def read_arf(filename, livetime):
"""Read ARF file.
Parameters
----------
filename : str or `Path`
PHA file name
livetime : `Quantity`
Livetime
Returns
-------
data : `RegionNDMap`
Exposure map
"""
aeff = RegionNDMap.read(filename, format="ogip-arf")
exposure = aeff * livetime
exposure.meta["livetime"] = livetime
return exposure
[docs] def read(self):
"""Read dataset.
Returns
-------
dataset : SpectrumDatasetOnOff
Spectrum dataset
"""
kwargs = self.read_pha(self.filename)
pha_meta = kwargs["counts"].meta
name = str(pha_meta["OBS_ID"])
livetime = pha_meta["EXPOSURE"] * u.s
filenames = self.get_filenames(pha_meta=pha_meta)
exposure = self.read_arf(filenames["arffile"], livetime=livetime)
if "bkgfile" in filenames:
bkg = self.read_bkg(filenames["bkgfile"])
kwargs.update(bkg)
if "rmffile" in filenames:
kwargs["edisp"] = self.read_rmf(filenames["rmffile"], exposure=exposure)
return SpectrumDatasetOnOff(name=name, exposure=exposure, **kwargs)