[AstroPy] Possible PyFITS behavior change with scaled data

[Erik Bray]

Hi all,

Speaking of conversion of scaled image data to floats, I have a question 
to put out there regarding the upcoming 3.0.8 release of PyFITS.

PyFITS 3.0.8 will fix a long-standing bug that's been around long enough 
that I'm worried that some people might have workarounds for it in place 
that will break now that the bug is fixed.

This specifically pertains to support for scaled images in image "sections".

For some background, PyFITS has a feature called "sections" that allows 
accessing small portions of an image without reading the entire image 
into memory at once.  For example:

 >>> hdul = pyfits.open('jb1f98q1q_raw.fits')
 >>> hdul.info()
Filename: jb1f98q1q_raw.fits
No.    Name         Type      Cards   Dimensions   Format
0    PRIMARY     PrimaryHDU     205   ()           int16
1    SCI         ImageHDU        95   (4144, 2068)   int16
...
 >>> hdul['SCI',1].header['BZERO']
32768.0
 >>> hdul['SCI',1].section[:100,:100]
array([[ 2278.,  2252.,  2248., ...,  2252.,  2252.,  2253.],
        [ 2277.,  2252.,  2250., ...,  2255.,  2253.,  2258.],
        [ 2277.,  2257.,  2250., ...,  2260.,  2255.,  2257.],
        ..., dtype=float32)

This reads the file in portions so that just the first 100x100 block of 
pixels is read from disk.

This feature used to have very spotty/broken support for scaled image 
data, but now that's been fixed.

Is that fact that this is now fixed likely to break anybody's code? 
Just curious.

Erik