Fix for Tabular-fits reader/writer fix #1113
Conversation
kelle
requested review from
eteq,
rosteen,
nmearl and
keflavich
as code owners
|
At some point as part of the |
|
I think the trouble is in |
|
Added this to the reader and all is now well tab.meta = fits.getheader(file_obj, 0) |
| tab = Table.read(file_obj, format='fits', hdu=hdu) | ||
|
|
||
| # Get the header from HDU0 | ||
| tab.meta = fits.getheader(file_obj, 0) |
There was a problem hiding this comment.
I think this is breaking compression support since Table.read() in those cases closes the file_obj on return, so you can no longer read from it. Probably need to get away without using the table io at all here as well, i.e. use read_fileobj_or_hdulist as wcs1d_fits_loader does and read both tab and header from hdulist (not sure if Table directly supports this).
| # Read it in and check against the original | ||
| with fits.open(tmpfile) as hdulist: | ||
| assert hdulist[0].header['NAXIS'] == 0 |
There was a problem hiding this comment.
Is this no longer working? Would be worrying, because an "empty" PrimaryHDU should always have NAXIS = 0!
This PR modifies the
tabular-fitswriter to put the header information in BOTH HDU0 and HDU1 based on the conversation in #905 .Other folks are welcome to modify this PR.
Broke out the testing of the
update_headermethod into its own test.fits compressedcurrently failing but I think that might be a file/memory handling thing which I may have accidentally broken in the test file.Remaining Todo:
fits_compressedtestupdate_headerbehavior and tests.