Primary HDU checksum data must be np.array?

I have some code which attempts to update the FITS checksums:

 output = "out.fits"
 with fits.open(self.filepath, do_not_scale_image_data=True) as hdus:
         for hdu in hdus:
               fits.append(output, hdu.data, hdu.header, checksum=True)

For a file which has no data in an HDU (primary), this code now raises an exception requiring that the data array be an np.array rather than None.

There's a relatively easy "work" around something like this:

     data = hdu.data if hdu.data is not None else np.array([])

but I wasn't positive it's correct.

Since it used to be possible to round trip these HDUs, I thought I should raise it as an issue and possible bug.

I (think) this problem is exposed with:

astropy 4.0.0.dev12788+g4166b6a94 py36_0 http://ssb.stsci.edu/conda-dev
numpy 1.16.3 py36hacdab7b_0
numpy-base 1.16.3 py36h6575580_0

It's my fault, sorry. This is because of 6800eec : _makehdu relies on _BaseHDU(data, header) being able to use the HDU detection from the header, and there is no test for this...

I removed _BaseHDU.__new__ because when reading from a FITS file this HDU detection was done twice. So I'm hesitating between just reverting this commit, or adding a classmethod which would explicitly call the class detection. I will look into this later today.

In [2]:  output = "out.fits" 
   ...:  with fits.open('astropy/io/fits/tests/data/test0.fits', do_not_scale_image_data=True) as hdus: 
   ...:          for hdu in hdus: 
   ...:                fits.append(output, hdu.data, hdu.header, checksum=True) 
   ...:                                                                                                                                                                                        
---------------------------------------------------------------------------
KeyError                                  Traceback (most recent call last)
<ipython-input-2-579e0ac3115e> in <module>
      2 with fits.open('astropy/io/fits/tests/data/test0.fits', do_not_scale_image_data=True) as hdus:
      3         for hdu in hdus:
----> 4               fits.append(output, hdu.data, hdu.header, checksum=True)

~/dev/astropy/astropy/io/fits/convenience.py in append(filename, data, header, checksum, verify, **kwargs)
    619         # output to the empty object.
    620         #
--> 621         writeto(filename, data, header, checksum=checksum, **kwargs)
    622     else:
    623         hdu = _makehdu(data, header)

~/dev/astropy/astropy/utils/decorators.py in wrapper(*args, **kwargs)
    519                             kwargs[new_name[i]] = value
    520 
--> 521             return function(*args, **kwargs)
    522 
    523         return wrapper

~/dev/astropy/astropy/io/fits/convenience.py in writeto(filename, data, header, output_verify, overwrite, checksum)
    414     """
    415 
--> 416     hdu = _makehdu(data, header)
    417     if hdu.is_image and not isinstance(hdu, PrimaryHDU):
    418         hdu = PrimaryHDU(data, header=header)

~/dev/astropy/astropy/io/fits/convenience.py in _makehdu(data, header)
   1045             hdu = ImageHDU(data, header=header)
   1046         else:
-> 1047             raise KeyError('Data must be a numpy array.')
   1048     return hdu
   1049 

KeyError: 'Data must be a numpy array.'

EDIT: xref #8502

Add test reproducing astropy#8660

Navigation Menu

Repository navigation

Primary HDU checksum data must be np.array? #8660

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Primary HDU checksum data must be np.array? #8660

Description

Activity

pllim commented on May 6, 2019

saimn commented on May 6, 2019

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions