FITS Binary Table to Astropy Table round trip re-orders COMMENT cards

I’m really not sure if this could be considered a true bug or not, but I imagine it will cause some users some frustration. When a FITS Binary file is round-tripped from an Astropy Table back to a FITS file (using Astropy.table.Table.read and fits.table_to_hdu) the COMMENT cards all get shuffled to the bottom of the header. A lot of HST data uses these COMMENT cards like section titles, not sure if other data tends to do this. I’m using astropy 3.1.dev21825 in this example.

Original, near the top of the header:

Output after round tripping, at the bottom of the header:

[saimn]

Table.meta is an OrderedDict by default I think. The issue has more to do with the way COMMENT/HISTORY/BLANK are managed by the fits.Header code (by default normal cards are inserted before these cards I think, hence the use of end=True).

[pllim]

Keeping the order (in a way that makes sense) gets complicated when header changes, e.g., add/remove keyword/comment/history/column. Sara and I talked about this and at some point concluded that it is easier to just show people how to re-order their header via some manual scripting. 😬