convolution debugging

[sophiathl]

1. If Gaussian convolution, then the fwhm argument ought to be set
2. If psf kernel convolution, then for the fits images without corresponding kernel, a Gaussian convolution is done. That means that both --fwhm and --kernels ought to be set in the psf kernel convolution case.

Do we want the 2nd option as such? Perhaps a message to the user of the like: "psf kernels not corresponding to the number of the input images - the --fwhm argument should be set to get the gaussian convolution running instead, for those images without corresponding psf kernels"

Also, how about not going through Gaussian convolution, if --kernels option set, unless the user sets also the --fwhm?

[jeffctaylor]

All of this sounds good to me.

Regarding the --kernels option, I was wondering if it should also take an argument which would be a directory containing the kernels that the user wants to use. I'm thinking of a user who wants to convolve to different instruments/wavelengths - our current setup would mean that she has to rename an entire set of kernels and place them in the /kernels subdirectory each time. If she could specify the directory where the kernels are located, then she could easily run the script as many times as necessary while changing a single argument each time.

Let me know how this sounds to you - I'm thinking only from a computational perspective without any real knowledge of the kind of practical usage this script will receive.

[sophiathl]

I agree - this is extremely thoughtful, once more!

[jeffctaylor]

Oh, nice to know that I can close an issue from another repository, but this has not been completely resolved. I still need to deal with the actual issue, not just my comment.