ValueError: Non-square pixels. Please resample data.

[esoPanda]

I got this error and I checked the header:

CDELT1 = -0.00268754582783 CDELT2 = 0.00269177717671 CDELT3 = 0.20000000298

Really, this is not almost_equal? Any idea for a work around?

[keflavich]

The pixel axes only differ by about 0.1%, but that can still be significant over a large enough image. If you really want to assume they're the same, just edit the CDELT1/CDELT2 to be identical. Otherwise, reproject them to a header with truly square pixels

[astrofrog]

Just thinking out loud - maybe we could have a flag for relaxing this constraint to a certain tolerance?

[esoPanda]

Thanks. Why pixels are required to be strictly square?

Indeed, would be good to have an option to relax the matching to really 'almost' equal. For example, if abs(dx) * nx and abs(dy) * nx differs less than one pixel, this can be acceptable?

The small size difference in x and y was introduced when I projected the cube from EQ to Gal (not using python/reproject). In this case I solved by using the original EQ cube, which has square pixels.

[keflavich]

There are a lot of simplifications that we can and do employ when we assume the pixels are square. Since almost all astronomical images on a grid (not HEALPIX stuff) uses square pixels, and it is straightforward to regrid to square pixels, I'm going to tag dealing with non-square pixels as low priority and close this issue.

[kmhess]

Hi, I'm curious is any progress on this has been made? I have the following code:

slice = PathFromCenter(center=SkyCoord(ra=source['pos_x'], dec=source['pos_y'], unit='deg'),
                       length=opt_view, angle=source['kin_pa']*u.deg, width=1*u.arcsec)
mask = fits.open(filename)
mask_pv = extract_pv_slice(mask[0].data, slice, wcs=WCS(mask[0].header, fix=True, translate_units='shd'))

And I get the error:

AssertionError: 
Arrays are not almost equal to 7 decimals
 ACTUAL: 0.0016675802169471255
 DESIRED: 0.0016657532823994607

The error has me stumped. I don't know what I'm supposed to reproject to fix this since the WCS and the data come from the same fits files and the actual/desired really are almost equal. Thanks in advance. Also CDELT1 and CDELT2 in the mask[0].header are identical and equal to 1.6666670000E-3

[kmhess]

I just realized there is an open issue with a similar title. I'll put my question there (apologies, this closed one was the first Issue I found with google).