[MINC-development] micopy_all_var_values dies if non-standard variables are present in MINC files
Rick Hoge
minc-development@bic.mni.mcgill.ca
Tue, 3 Feb 2004 22:32:35 -0500
Thanks a lot for the suggestion - it turns out I did have an old (but
not that old) version of the minc libraries. These files work fine
with the current version. Also appreciate the info on the dimension
vars - this has always been a bit mysterious to me but is making more
sense...
Rick
On Feb 3, 2004, at 10:28 PM, Peter NEELIN wrote:
> Hi Rick,
>
> On Fri, 23 Jan 2004, Rick Hoge wrote:
>
>> For a while we've been motivated to add 'custom' variables to MINC
>> files to embed information about the experiment or analysis in a
>> convenient way. It seems though that non-standard variables prevent
>> some of the MINC tools from running.
>
> They shouldn't. That is one of the ideas behind minc. It should be
> extensible, and as long as you don't use any standard variable names,
> you
> should be okay. (One caution - I don't know if variables that are
> indexed
> by image dimensions are always handled properly by programs that change
> the dimensions, like mincresample.)
>
> Anyway, I tried resampling the file fileWithNewVariable.mnc and it
> worked
> fine on the SGIs in the BIC. Do you have an older version of
> mincresample?
> Can you trace the problem down? The fact that it dies without an error
> makes it hard to diagnose - the MINC package entry point just tells you
> that a NetCDF function failed when called by the minc function
> micopy_all_var_values, but NetCDF obviously did not report the error
> as it
> should have. Are you certain that you have not simply run out of disk
> space (or quota)? NetCDF does have a habit of swallowing that sort of
> error.
>
> You could try compiling a debugging version of mincresample and trace
> it
> to find out where the problem occurs.
>
> Are you using NetCDF 3.5? I think that it does a better job of
> reporting
> errors than 3.2.
>
> The only comment that I would have about your file is that you do not
> need
> to have the dimension variables (xspace, etc) indexed by anything if
> the
> spacing is regular. If you are not going to fill in the information,
> then
> you should not have the indexing. Right now, the information is
> contradictory and so your file is strictly speaking not valid.
>
> For the imagemin/max variables, you do not need to have them indexed by
> zspace if you are going to scale all of the images the same way.
>
> Peter
> ----
> Peter Neelin (neelin@bic.mni.mcgill.ca)
>
>
> _______________________________________________
> MINC-development mailing list
> MINC-development@bic.mni.mcgill.ca
> http://www.bic.mni.mcgill.ca/mailman/listinfo/minc-development