[AstroPy] FITS files and column descriptions

Perry Greenfield stsci.perry at gmail.com
Mon Nov 11 08:51:05 EST 2013

What TOPCAT is doing is in no way part of the FITS standard. As far as I can tell, it isn't even listed as a registered convention. It's exactly this sort of thing that drives us nuts regarding FITS. Somehow people expect us to read and support anything anyone happens to stuff into a FITS file and handle its semantics correctly.

Sorry, it's not going to happen within the FITS module itself. If you want to use it, you need to layer something over FITS to handle it.


On Nov 7, 2013, at 4:59 PM, Paul Kuin wrote:

> Some years ago, I was leading the NASA Astronomical Data Center, now gone. I worked with CDS to have a way to present a readable table along with the ability to automatically generate a FITS file.  That lead to the "ReadMe" files that accompany the catalogs from CDS, which is storing the metadata. 
> I think it is important to make sure no metadata are lost in the conversion process. Sure, the FITS standard says so-and-so, but in practice, one has to make amends so the standard can evolve. 
> I have only used the astropy saving to tables a little bit because I got annoyed that there was loss of (meta) data.  
> The reason I write is, that there needs to be some serious thought about this issue. as flagged by Yannick. My (biased) feeling  is that you cannot go around having a second metadata file when going to ascii. Doing otherwise will lead to a plethora of ways to save the data file. Better to keep the data file clean, and put the jumk in a meta data file.
> Cheers, 
>    Paul
> On Thu, Nov 7, 2013 at 9:20 PM, Yannick Roehlly <yannick.roehlly at lam.fr> wrote:
> Hi all,
> I have some remarks / questions concerning the column descriptions in
> FITS table files (and VO).
> 1. Long column description
> The FITS format requires that a column description must be 68
> characters or less. Some tools like topcat / stills allows longer
> descriptions because they use a kind of VO-table metada section in the
> FITS header. If you open such file with astropy.io.fits and save it
> again, you loose the long description. In fact, it's not an astropy
> problem as these description as not in the "real FITS" column
> descriptions in the header (TCOMxx), it is just because astropy does
> not save the "VO-table" part of the topcat file.
> Maybe it's worth raising a warning when opening such files, like "Hey,
> you may loose metadata!" because one may be surprise not to have all
> its column description.

More information about the AstroPy mailing list