astropy.io.fits; ValueError: The header keyword 'description' with its value is too long

I am using astropy with anaconda on a Mac with Python 3.6.

astroquery.version: 0.3.9 astropy.version: 3.2.1

I have used astroquery to download a table from Vizier but when I try to write out the table I get the error:

File "/Users/rgm/anaconda2/envs/py36/lib/python3.6/site-packages/astropy/io/fits/card.py", line 990, in _format_image
    'too long'.format(self.keyword))
ValueError: The header keyword 'description' with its value is too long

This looks like a problem with the description of one of the columns. I have investigated the description lengths and trimmed the longest one. This did not help.

I then tried to replac the descriptions with an empty string and this did not help either.

result.columns[icol].description = ‘’"

If I knew which column was causing the problem, I could remove it.

See example code below

# python 3 future proofing;see Python 2 to 3 cheatsheet
from __future__ import print_function, division
from builtins import input
try:
   input = raw_input
except NameError:
   pass

import numpy as np

from astropy.table import Table
import astropy.units as u
import astropy.coordinates as coord

import astroquery
print('astroquery.__version__:', astroquery.__version__)

import astropy
print('astropy.__version__:', astropy.__version__)


from astroquery.vizier import Vizier
from astropy.config import get_config
from astropy.config import get_config_dir
from astropy.utils.data import conf

print(get_config('conf'))
print('get_config_dir(): ', get_config_dir())

config_vizier = get_config('astroquery.vizier')
for key,value in config_vizier.items():
    print(key, value)

Vizier.TIMEOUT = 300
Vizier.ROW_LIMIT = -1

catalog = 'II/317/cfhtls_w'
ra = 35.665
dec = -4.795
radec = coord.SkyCoord(ra=ra, dec=dec,
                       unit=(u.deg, u.deg),
                       frame='icrs')
print(radec)
vizier = Vizier(columns=['**'])
vizier.ROW_LIMIT = -1
width = 0.1
height = 0.1
result = vizier.query_region(radec,
                             width=width*u.deg, height=height*u.deg,
                             catalog=catalog)

print('Number of rows returned', len(result))

# result is a TableList; since one catalogue use [0]
result = result[0]
print('Number of rows returned', len(result))
print()

result.info()

trimAll = True
print()
for icol, column in enumerate(result.itercols()):
    print(icol,':', column.description,
          ';', len(column.description))
    if len(column.description) >= 40:
        print()
        print(icol)
        print(result.columns[icol].unit)
        print(result.columns[icol].description)
        print()
        result.columns[icol].description = \
            result.columns[icol].description[0:38]
        print(icol,':', column.description,
              ';', len(column.description))

        result.columns[icol].description = ''

    if trimAll:
        result.columns[icol].description = ''

result.info()

result.write('tmp.fits')

[keflavich]

Yeah, this seems to be simply a writer issue. Not all tables can be written by all writers; for example, most votables can’t be written as ascii.ipac files because of forbidden characters in the column names. It would be nice if we had some “automatic fixing” options (e.g., allowing description truncation in this case), but that’s low on the wishlist. I’m in favor of improving the error message and/or documentation around the error message, but leaving the code untouched.

[bsipocz]

This has bitten me in the past, too, but I don’t think the way forward is to patch astroquery. It returns a perfectly valid Table, and after that it’s the users decision of what they do with it, what format they write it out, etc. Thus we shouldn’t make assumptions it would be fits rather than some other format that would happily round trip with the description in meta. @keflavich - what is your take on this?

So I think it’s indeed the first writer here that should return a more useful error message, pointing out the problem less cryptically.