FITS table documentation uses non-standard FITS unit

Description

While reading the documentation about fits table data, I noticed that there is a non-standard unit used: https://github.com/astropy/astropy/blame/main/docs/io/fits/usage/table.rst#L346

   >>> col2 = fits.Column(name='counts', format='J', unit='DN', array=counts)

Expected behavior

I’d expect that the documentation sticks to the standard.

Actual behavior

DN is not a standard FITS unit, and even not directly supported by astropy.units. For example, the following does not work:

counts = [0., 1.]
col2 = fits.Column(name='counts', format='J', unit='DN', array=counts)
tbl = QTable([col2])

It may sound nitpicking, but we should not encourage people to use phantasy units. The name “counts” would suggest as unit count.

[pllim]

@olebole , are you interested in submitting a PR? 😸

[jdavies-st]

DN are not fantasy units, though it does seem that they exist in some parallel, mystical place, not of this earth. Most HST, JWST and lots of Spitzer science data use DN to indicate counts. That’s why that unit was added to astropy.units (currently unreleased), so that the data from these telescopes could be read in easily as a Quantity.

Agree that the example in the docs should use a FITS standard unit. Just because there’s lots of existing (and soon-to-exist) data using DN, doesn’t mean we should encourage its use.

What does it mean? Some say Data Number. Some say Dumb Number. (h.t. @eslavich)

The history is that astropy.io.fits came via pyfits from STScI where DN was and is common usage for counts amongst the scientists and engineers, and all that pre-dated any standard for a count unit from the FITS working group. FITS had no standard for units of counts up through 2005, and only adopted count, ct or adu in 2008.

So yeah, let’s change the documentation example.