Treat numbers in IRF table columns as `ndmin=1`

It is currently impossible using astropy to read or write 1d arrays with a single entry. They will be written and read as single numbers instead of 1d arrays of length 1.

This means, that IRFs cannot be written with single bins in any of the dimensions, which is most important for the FOV offset angle for point-like IRFs.

Issue at astropy: https://github.com/astropy/astropy/issues/10765

Gammapy should treat single numbers in columns like THETA_LOW, THETA_HI as 1d arrays with a single entry, e.g. by calling something like u.Quantity(table['THETA_LO'], ndmin=1).

CUrrently this fails due to len being called on a scalar:

In [5]: EffectiveAreaTable2D.read('./pyirf_eventdisplay.fits.gz', hdu='EFFECTIVE_AREA')
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-5-1f951cbdf154> in <module>
----> 1 EffectiveAreaTable2D.read('./pyirf_eventdisplay.fits.gz', hdu='EFFECTIVE_AREA')

~/.local/anaconda/envs/pyirf-dev/lib/python3.7/site-packages/gammapy/irf/effective_area.py in read(cls, filename, hdu)
    413         """Read from file."""
    414         with fits.open(make_path(filename), memmap=False) as hdulist:
--> 415             return cls.from_hdulist(hdulist, hdu=hdu)
    416 
    417     def to_effective_area_table(self, offset, energy=None):

~/.local/anaconda/envs/pyirf-dev/lib/python3.7/site-packages/gammapy/irf/effective_area.py in from_hdulist(cls, hdulist, hdu)
    407     def from_hdulist(cls, hdulist, hdu="EFFECTIVE AREA"):
    408         """Create from `~astropy.io.fits.HDUList`."""
--> 409         return cls.from_table(Table.read(hdulist[hdu]))
    410 
    411     @classmethod

~/.local/anaconda/envs/pyirf-dev/lib/python3.7/site-packages/gammapy/irf/effective_area.py in from_table(cls, table)
    401             offset_hi=table["THETA_HI"].quantity[0],
    402             data=table["EFFAREA"].quantity[0].transpose(),
--> 403             meta=table.meta,
    404         )
    405 

~/.local/anaconda/envs/pyirf-dev/lib/python3.7/site-packages/gammapy/irf/effective_area.py in __init__(self, energy_lo, energy_hi, offset_lo, offset_hi, data, meta, interp_kwargs)
    369             offset_axis = MapAxis.from_nodes(offset_lo, interp="lin", name="offset")
    370         else:
--> 371             offset_edges = edges_from_lo_hi(offset_lo, offset_hi)
    372             offset_axis = MapAxis.from_edges(offset_edges, interp="lin", name="offset")
    373 

~/.local/anaconda/envs/pyirf-dev/lib/python3.7/site-packages/gammapy/maps/utils.py in edges_from_lo_hi(edges_lo, edges_hi)
    153     edges = edges_lo.copy()
    154     try:
--> 155         edges = edges.insert(len(edges), edges_hi[-1])
    156     except AttributeError:
    157         edges = np.insert(edges, len(edges), edges_hi[-1])

~/.local/anaconda/envs/pyirf-dev/lib/python3.7/site-packages/astropy/units/quantity.py in __len__(self)
   1064         if self.isscalar:
   1065             raise TypeError("'{cls}' object with a scalar value has no "
-> 1066                             "len()".format(cls=self.__class__.__name__))
   1067         else:
   1068             return len(self.value)

TypeError: 'Quantity' object with a scalar value has no len()

[maxnoe]

ACtually the fix is about to be merged and should be released with astropy 4.0.2

[maxnoe]

@adonath since the fix is now published in astropy 4.0.2, I will close this as not a gammapy issue.

If you want to support older gammapy versions, you could implement something, but I wouldn’t.